Lecture Notes in Physics

Editorial Board
R. Beig, Wien, Austria
J. Ehlers, Potsdam, Germany
U. Frisch, Nice, France
K. Hepp, Zürich, Switzerland
W. Hillebrandt, Garching, Germany
D. Imboden, Zürich, Switzerland
R. L. Jaffe, Cambridge, MA, USA
R. Kippenhahn, Göttingen, Germany
R. Lipowsky, Golm, Germany
H. v. Löhneysen, Karlsruhe, Germany
I. Ojima, Kyoto, Japan
H. A. Weidenmüller, Heidelberg, Germany
J. Wess, München, Germany
J. Zittartz, Köln, Germany

3
Berlin
Heidelberg
New York
Barcelona
Hong Kong
London
Milan
Paris
Singapore
Tokyo
Editorial Policy
The series Lecture Notes in Physics (LNP), founded in 1969, reports new developments in
physics research and teaching -- quickly, informally but with a high quality. Manuscripts
to be considered for publication are topical volumes consisting of a limited number of
contributions, carefully edited and closely related to each other. Each contribution should
contain at least partly original and previously unpublished material, be written in a clear,
pedagogical style and aimed at a broader readership, especially graduate students and
nonspecialist researchers wishing to familiarize themselves with the topic concerned. For
this reason, traditional proceedings cannot be considered for this series though volumes
to appear in this series are often based on material presented at conferences, workshops
and schools (in exceptional cases the original papers and/or those not included in the
printed book may be added on an accompanying CD ROM, together with the abstracts
of posters and other material suitable for publication, e.g. large tables, colour pictures,
program codes, etc.).

Acceptance
A project can only be accepted tentatively for publication, by both the editorial board and the
publisher, following thorough examination of the material submitted. The book proposal
sent to the publisher should consist at least of a preliminary table of contents outlining the
structure of the book together with abstracts of all contributions to be included.
Final acceptance is issued by the series editor in charge, in consultation with the publisher,
only after receiving the complete manuscript. Final acceptance, possibly requiring minor
corrections, usually follows the tentative acceptance unless the final manuscript differs
significantly from expectations (project outline). In particular, the series editors are entitled
to reject individual contributions if they do not meet the high quality standards of this
series. The final manuscript must be camera-ready, and should include both an informative
introduction and a sufficiently detailed subject index.

Contractual Aspects
Publication in LNP is free of charge. There is no formal contract, no royalties are paid,
and no bulk orders are required, although special discounts are offered in this case. The
volume editors receive jointly 30 free copies for their personal use and are entitled, as are the
contributing authors, to purchase Springer books at a reduced rate. The publisher secures
the copyright for each volume. As a rule, no reprints of individual contributions can be
supplied.

Manuscript Submission
The manuscript in its final and approved version must be submitted in camera-ready form.
The corresponding electronic source files are also required for the production process, in
particular the online version. Technical assistance in compiling the final manuscript can be
provided by the publisher’s production editor(s), especially with regard to the publisher’s
own Latex macro package which has been specially designed for this series.

Online Version/ LNP Homepage

LNP homepage (list of available titles, aims and scope, editorial contacts etc.):
[Link]
LNP online (abstracts, full-texts, subscriptions etc.):
[Link]
C. F. Barenghi R. J. Donnelly W. F. Vinen (Eds.)

Quantized
Vortex Dynamics
and
Superfluid Turbulence

13
Editors
C.F. Barenghi
University of Newcastle
Mathematics Department
Newcastle NE1 7RU, United Kingdom

R.J. Donnelly
University of Oregon
Physics Department
Eugene, OR 97403, USA

W.F. Vinen
University of Birmingham
Physics Department
Birmingham B15 2TT, United Kingdom

Cover picture: Tangle of quantized vortex filaments computed in a periodic box

by D. Kivotides, D. Samuels and C.F. Barenghi.
Library of Congress Cataloging-in-Publication Data applied for.

Die Deutsche Bibliothek - CIP-Einheitsaufnahme

Quantized vortex dynamics and superfluid turbulence / C. F. Barenghi ...

(ed.). - Berlin ; Heidelberg ; New York ; Barcelona ; Hong Kong ; London ;
Milan ; Paris ; Singapore ; Tokyo : Springer, 2001
(Lecture notes in physics ; 571)
(Physics and astronomy online library)
ISBN 3-540-42226-9
ISSN 0075-8450
ISBN 3-540-42226-9 Springer-Verlag Berlin Heidelberg New York

This work is subject to copyright. All rights are reserved, whether the whole or part of the
material is concerned, specifically the rights of translation, reprinting, reuse of illustra-
tions, recitation, broadcasting, reproduction on microfilm or in any other way, and
storage in data banks. Duplication of this publication or parts thereof is permitted only
under the provisions of the German Copyright Law of September 9, 1965, in its current
version, and permission for use must always be obtained from Springer-Verlag. Violations
are liable for prosecution under the German Copyright Law. Springer-Verlag Berlin Hei-
delberg New York
a member of BertelsmannSpringer Science+Business Media GmbH [Link]
c Springer-Verlag Berlin Heidelberg 2001
Printed in Germany The use of general descriptive names, registered names, trademarks,
etc. in this publication does not imply, even in the absence of a specific statement, that such
names are exempt from the relevant protective laws and regulations and therefore free for
general use.
Typesetting: Data conversion by Steingraeber GmbH, Heidelberg
Cover design: design & production, Heidelberg
Printed on acid-free paper
SPIN: 10792065 55/3141/du - 5 4 3 2 1 0
Preface

This book springs from the programme Quantized Vortex Dynamics and Super-
fluid Turbulence held at the Isaac Newton Institute for Mathematical Sciences
(University of Cambridge) in August 2000. What motivated the programme was
the recognition that two recent developments have moved the study of quan-
tized vorticity, traditionally carried out within the low-temperature physics and
condensed-matter physics communities, into a new era.
The first development is the increasing contact with classical fluid dynamics
and its ideas and methods. For example, some current experiments with he-
lium II now deal with very classical issues, such as the measurement of velocity
spectra and turbulence decay rates. The evidence from these experiments and
many others is that superfluid turbulence and classical turbulence share many
features. The challenge is now to explain these similarities and explore the time
scales and length scales over which they hold true. The observed classical aspects
have also attracted attention to the role played by the flow of the normal fluid,
which was somewhat neglected in the past because of the lack of direct flow
visualization. Increased computing power is also making it possible to study the
coupled motion of superfluid vortices and normal fluids. Another contact with
classical physics arises through the interest in the study of superfluid vortex re-
connections. Reconnections have been studied for some time in the contexts of
classical fluid dynamics and magneto-hydrodynamics (MHD), and it is useful to
learn from the experience acquired in other fields.
The second development arises from atomic physics and is the discovery of
Bose–Einstein condensation in confined clouds of alkali atoms. The study of
superfluidity and quantized vorticity is now possible in a wide range of other
systems besides helium II. The rapid progress in this area has given momentum
to the use of the Gross–Pitaevskii Equation or Nonlinear Schroedinger Equation
(NLSE). Researchers have become more aware of the approximations and limi-
tations involved in the NLSE model, but also of its range of validity and great
power of prediction. The use of the NLSE has become more established, and
the NLSE is proving to be a powerful tool for modeling problems such as vortex
nucleation, reconnections and even turbulence.
A further development arises from the results of preliminary theory and ex-
periments in turbulent Helium 3 which suggest that there are significant differ-
ences with turbulence in Helium 4 and these are likely to be explored in the
future.
VI

It is apparent from this background that the contributions to this book come
from investigators with a wide range of backgrounds and expertise: condensed-
matter physics and low-temperature physics, classical fluid dynamics and applied
mathematics, MHD, atomic physics, and engineering (for the applications of
helium II as a cryogenic coolant).
The book is divided into topical chapters. Each chapter begins with one or
two introductory review articles, which are suitable for students and new inves-
tigators interested in entering the field. The introductory articles are followed
by shorter, more specialized papers.
Chapter 1 introduces us to the problem of quantized vorticity and super-
fluid turbulence, and it summarizes the key aspects and problems which are
currently studied. Chapter 2 is devoted to turbulence experiments. Chapter 3
considers the fundamental problem of friction and vortex dynamics. The theory
of superfluid turbulence and the interpretation of the experimental results is the
subject of Chap. 4. Chapter 5 is devoted to the application of the NLSE model
to superfluidity and vortices. Chapter 6 moves away from helium and considers
Bose–Einstein Condensation and vortices in the context of alkali atoms. Chap-
ter 7 is concerned with some aspects of classical turbulence and MHD which
are relevant in the study of superfluid turbulence. Finally, Chap. 8 deals with
Helium 3 and other systems.
We are grateful for the support and encouragement of Professor Keith Mof-
fatt, Director of the Newton Institute, and we would like to thank Tracey Andrew
who helped in the preparation of the manuscripts for publication.

Newcastle, Eugene and Birmingham, Carlo Barenghi

June 2001 Russ Donnelly
Joe Vinen
Contents

Part I Introduction

Introduction to Superfluid Vortices and Turbulence

C.F. Barenghi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1 The Two-Fluid Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Quantized Vortex Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3 Modelling the Vortex Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1 Microscopic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Mesoscopic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3 Macroscopic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4 Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.1 Turbulent Counterflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2 Turbulent Coflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5 Motion of Superfluid Vortices for a Given Normal Fluid . . . . . . . . . . . . . 10
6 Motion of the Normal Fluid at Given Superfluid Vortices . . . . . . . . . . . . 11
7 Fully Coupled Motion of Superfluid Vortices and Normal Fluid . . . . . . . 12
8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Part II Turbulence Experiments

An Introduction to Experiments on Superfluid Turbulence

R.J. Donnelly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2 Update on Pipe Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Update on Towed Grid Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.1 The Nature of Grid Turbulence in Helium II . . . . . . . . . . . . . . . . . . 20
3.2 Four Regimes of Decaying Grid Turbulence in Helium II . . . . . . . . 23
4 Agenda for the Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1 The University of Oregon 6 cm Wind Tunnel . . . . . . . . . . . . . . . . . . 24
4.2 Wind Tunnels for Model Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Tow Tanks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Challenges for the Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.1 The Challenge of Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5.2 Challenges for Understanding Counterflow Turbulence . . . . . . . . . . 29
VIII Contents

5.3
Challenges for Understanding Periodic
Boundary Layer Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.4 Instrumentation to Detect Vortices Below 1 K . . . . . . . . . . . . . . . . . 33
5.5 The Normal Fluid and the Vortex Tangle . . . . . . . . . . . . . . . . . . . . . 33
5.6 Flow over Blunt Objects, Testing Models such as Submarines . . . 34
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

The Experimental Evidence for Vortex Nucleation in 4He

É. Varoquaux, O. Avenel, Y. Mukharsky, P. Hakonen . . . . . . . . . . . . . . . . . . . 36
1 Single Vortex Nucleation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2 Multiple Slips and Collapses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Applications of Superfluid Helium
in Large-Scale Superconducting Systems
S.W. Van Sciver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2 Superconducting Systems That Use He II Cooling . . . . . . . . . . . . . . . . . . 53
2.1 Accelerator Magnet System for LHC . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.2 High Field Solenoid for the NHMFL 45-T Hybrid . . . . . . . . . . . . . . 54
2.3 RF Cavity Systems for the TESLA Electron Collider . . . . . . . . . . . 55
3 Application Relevant He II Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.1 Second Sound Pulse Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2 Transient and Steady Transport in the Mutual Friction Regime . . 58
3.3 The He II Energy Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4 Fluid Dynamics of Forced Flow He II . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.5 He II/Vapor Two Phase Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.6 Fountain Effect (Fluid Management) . . . . . . . . . . . . . . . . . . . . . . . . . 62
4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

The Temperature Dependent Drag Crisis

on a Sphere in Flowing Helium II
Y.S. Choi, M.R. Smith, S.W. Van Sciver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2 Experimental Apparatus and Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Experiments on Quantized Turbulence at mK Temperatures

S.I. Davis, P.C. Hendry, P.V.E. McClintock, H. Nichol . . . . . . . . . . . . . . . . . 73
1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2 Creation and Detection of Vortices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3 The Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4 Preliminary Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
 Contents IX

6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Grid-Generated He II Turbulence
in a Finite Channel – Experiment
J.J. Niemela, L. Skrbek, S.R. Stalp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Intermittent Switching Between Turbulent and Potential Flow
Around a Sphere in He II at mK Temperatures
M. Niemetz, H. Kerscher, W. Schoepe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
1 Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
2.1 Stable Turbulent Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
2.2 Intermittent Switching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.3 Turbulent Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
2.4 Laminar Phases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

Part III Vortex Dynamics

Vortex Filament Methods for Superfluids

D.C. Samuels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2 Vortex Filament Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
2.1 The Biot–Savart Law and the Local Induction Approximation . . . 99
2.2 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
2.3 Meshing of the Filaments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3 Reconnections of Filaments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4 Analysis of the Superfluid Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5 Alternative Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6 Conclusions: What Needs to Be Done . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Introduction to HVBK Dynamics
D.D. Holm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
1 HVBK Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
2 Incompressible Renormalized HVBK Flows . . . . . . . . . . . . . . . . . . . . . . . . 120
3 Rotating Frame Renormalized HVBK Equations . . . . . . . . . . . . . . . . . . . 123
Appendix: Lie-Poisson Hamiltonian Formulation . . . . . . . . . . . . . . . . . . . . . . . 125
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
X Contents

Magnus Force, Aharonov–Bohm Effect,

and Berry Phase in Superfluids
E. Sonin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
2 Gross–Pitaevskii Theory and Two-Fluid Hydrodynamics . . . . . . . . . . . . 132
3 Interaction of Phonons with a Vortex in Hydrodynamics . . . . . . . . . . . . 133
4 Momentum Balance in the Two-Fluid Hydrodynamics . . . . . . . . . . . . . . 135
5 Magnus Force and the Berry Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

Using the HVBK Model

to Investigate the Couette Flow of Helium II
K.L. Henderson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
2 Linear Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3 Nonlinear Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.1 Infinite Cylinder Assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
3.2 Unit Aspect Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

Part IV Turbulence Theory

An Introduction to the Theory of Superfluid Turbulence

W.F. Vinen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
2 Counterflow Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
3 Grid Turbulence in Superfluid Helium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
3.1 Measurements of the Decay of Vortex Lines,
and the Quasi-classical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
3.2 Superfluid Turbulence on Length Scales
Larger than the Vortex Line Spacing . . . . . . . . . . . . . . . . . . . . . . . . . 153
3.3 The Turbulent Energy Spectra in Superfluid Grid Turbulence . . . 154
3.4 Superfluid Turbulence at Very Low Temperatures . . . . . . . . . . . . . . 155
3.5 Dissipation at Higher Temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

Numerical Methods for Coupled Normal-Fluid

and Superfluid Flows in Helium II
O.C. Idowu, D. Kivotides, C.F. Barenghi, D.C. Samuels . . . . . . . . . . . . . . . . 162
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
2 The Self-Consistent Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
3 Numerical Methods for 2-D Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
3.1 The Normal-Fluid Flow in 2-D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
3.2 Delta Function Forcing on a Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
 Contents XI

3.3Extrapolation of the Normal-Fluid Flow in the Neighbourhood

of the Superfluid Vortex Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
3.4 Numerical Stability and Time Stepping . . . . . . . . . . . . . . . . . . . . . . . 168
4 Results in 2-D Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5 Numerical Methods for 3-D Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.1 The Free Normal-Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.2 The Superfluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.3 The Interaction Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.4 Preliminary Results in 3-D Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

From Vortex Reconnections to Quantum Turbulence

T. Lipniacki . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
2 Vortex Motion Following Reconnection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
2.1 The Case vns = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
2.2 The Case vns = const = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
3 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

Vortices and Stability in Superfluid Boundary Layers

S.P. Godfrey, D.C. Samuels, C.F. Barenghi . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
1.1 The Two-Fluid Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
2 Boundary Layer Vortices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
2.1 Properties of the Vortex Line Solutions . . . . . . . . . . . . . . . . . . . . . . . 186
2.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
3 Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
3.1 Linear Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
3.2 Stability Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

Grid Generated He II Turbulence in a Finite Channel –

Theoretical Interpretation
L. Skrbek, J.J. Niemela . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
2 The Spectral Decay Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

Vortex Tangle Dynamics Without Mutual Friction

in Superfluid 4 He
M. Tsubota, T. Araki, S.K. Nemirovskii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
2 Vortex Wave Cascade Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
XII Contents

3 Cascade Process in the Vortex Tangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

3.1 Decay of the Vortex Tangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
3.2 Comparison with the Vinen’s Equation . . . . . . . . . . . . . . . . . . . . . . . 203
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

Applications of the Gaussian Model of the Vortex Tangle

in the Superfluid Turbulent He II
S.K. Nemirovskii, M.V. Nedoboiko . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
2 Constructing the Trial Distribution Function . . . . . . . . . . . . . . . . . . . . . . . 206
3 Hydrodynamic Impulse of the Vortex Tangle . . . . . . . . . . . . . . . . . . . . . . . 208
4 Energy of the Vortex Tangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

Stochastic Dynamics of a Vortex Loop.

Thermal Equilibrium
S.K. Nemirovskii, L.P. Kondaurova, M. Tsubota . . . . . . . . . . . . . . . . . . . . . . . 212
1 Introduction and Scientific Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
2 Langevin Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
3 Fokker–Planck Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
4 Possible Violation of Thermal Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . 217
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

Stochastic Dynamics of Vortex Loop.

Large-Scale Stirring Force
S.K. Nemirovskii, [Link]. Baltsevich . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
2 Analytical Investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
3 Conservation Laws and Pair Correlators . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
4 Some Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

Nonequilibrium Vortex Dynamics in Superfluid Phase

Transitions and Superfluid Turbulence
H.-C. Chu, G.A. Williams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
2 Quenched Superfluid Transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
3 Superfluid Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
4 Three Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
 Contents XIII

Part V The NLSE and Superfluidity

The Nonlinear Schrödinger Equation as a Model of Superfluidity

P.H. Roberts, N.G. Berloff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
2 The Fluid Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
3 Shortcomings of the GP Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
4 Vortices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
5 Superfluid Turbulence; Vortex Line Reconnection . . . . . . . . . . . . . . . . . . . 243
6 Intrinsic Vortex Nucleation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
7 Capture of Impurities by Vortex Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
8 Nonlocal Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256

Vortex Nucleation and Limit Speed for a Flow Passing Nonlinearly

Around a Disk in the Nonlinear Schrödinger Equation
S. Rica . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
1 Introduction and Formulation of the Problem . . . . . . . . . . . . . . . . . . . . . 258
2 Critical Velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
3 Flow Around a Disk via a Janzen–Rayleigh Expansion . . . . . . . . . . . . . . 262
4 Unstable Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
5 The Euler–Tricomi Equation near the Transonic Region . . . . . . . . . . . . 265
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267

Vortices in Nonlocal Condensate Models of Superfluid Helium

N.G. Berloff, P.H. Roberts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
2 Applicability of the Generalized Gross–Pitaevskii Model . . . . . . . . . . . . . 269
3 Nonlocal Nonlinear Schrödinger Equation . . . . . . . . . . . . . . . . . . . . . . . . . 271
4 Vortex Nucleation and Roton Emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

Ginzburg–Landau Description of Vortex Nucleation

in a Rotating Superfluid
I. Aranson, V. Steinberg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
2 Spin-Up and Nucleation of Vortices in Superfluid Helium . . . . . . . . . . . . 277
3 Stability of Multicharged Vortices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
4 Nucleation of Vortices by Rapid Thermal Quench . . . . . . . . . . . . . . . . . . 281
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
XIV Contents

Weak Turbulence Theory for the Gross–Pitaevskii Equation

S. Nazarenko, Y. Lvov, R. West . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
1 Motivation and Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
2 Weak Turbulence Theory for NLSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
3 Linear Dynamics of the GPE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
3.1 Without a Condensate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
3.2 With a Condensate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
4 Applicability of WKB Descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
5 Weakly Nonlinear GPE Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

Dissipative Vortex Dynamics and Magnus Force

L.M. Pismen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
1 Basic Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
2 Magnus Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
3 Three-Dimensional Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
4 Failure of Mechanistic Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296

Transition to Dissipation
in Two- and Three-Dimensional Superflows
C. Huepe, C. Nore, M.-E. Brachet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
2 Definition of the System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
3 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
4 Bifurcation Diagram and Scaling in 2D . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
5 Subcriticality and Vortex-Stretching in 3D . . . . . . . . . . . . . . . . . . . . . . . . 301
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303

Part VI Bose–Einstein Condensation

Motion of Objects Through Dilute Bose–Einstein Condensates

C.S. Adams, B. Jackson, M. Leadbeater, J.F. McCann, T. Winiecki . . . . . . 307
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
2 Fluid Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
3 Time-Independent Solutions in the Object Frame . . . . . . . . . . . . . . . . . . 309
4 The Critical Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
5 Vortex Shedding and Drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
6 The Critical Velocity in Inhomogeneous Condensates . . . . . . . . . . . . . . . 314
7 Comparison to Ions in Helium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
 Contents XV

Stability of a Vortex in a Rotating

Trapped Bose–Einstein Condensate
A.L. Fetter, A.A. Svidzinsky . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
1 Time-Dependent Gross–Pitaevskii Equation . . . . . . . . . . . . . . . . . . . . . . . 320
1.1 Equivalent Hydrodynamics of Compressible Isentropic Fluid . . . . . 320
1.2 Thomas–Fermi Limit for Large Condensates . . . . . . . . . . . . . . . . . . . 321
2 Energy of a Vortex in a Large Rotating Trap . . . . . . . . . . . . . . . . . . . . . . 321
3 Small-Amplitude Excitation of a Vortex in a Rotating Trap . . . . . . . . . . 322
3.1 Stability of a Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
3.2 Splitting of Normal-Mode Frequencies Caused by a Vortex . . . . . . 323
4 Vortex Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
4.1 Dynamics of Straight Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
4.2 Inclusion of Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
Kinetics of Strongly Non-equilibrium
Bose–Einstein Condensation
B. Svistunov . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
2 Kinetic Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
3 Coherent Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
4 External Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

Quantum Nucleation of Phase Slips

in Bose–Einstein Condensates
H.P. Büchler, V.B. Geshkenbein, G. Blatter . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
2 Effective Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
3 Finite Size Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340

Part VII Vortex Reconnections and Classical Aspects

Vortex Reconnection in Normal and Superfluids

J. Koplik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
2 Some Vortex Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
3 The Importance of Reconnection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
4 Reconnection in Normal Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
5 Reconnection in Superfluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
XVI Contents

Helicity in MHD and Hydro Reconnection

A. Brandenburg, R.M. Kerr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
2 Dissipation of Energy and Helicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
3 Interlocked Flux Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
4 Orthogonal Vortex Tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365

Tropicity and Complexity Measures for Vortex Tangles

R.L. Ricca . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
1 Vortex Structures and Tangles
in Classical and Quantized Vortex Flows . . . . . . . . . . . . . . . . . . . . . . . . . . 366
2 Measures of Tropicity for Vortex Tangles:
Tubeness, Sheetness and Bulkiness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
3 Measures of Geometric Complexity:
Directional Alignment and Writhing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
4 Algebraic Measure of Structural Complexity:
Average Crossing Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
5 Measures of Topological Entanglement:
Kinetic Helicity and Directional Linking . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
6 Relationships Between Complexity Measures and Energy Levels . . . . . . 371
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
The Geometry of Magnetic and Vortex Reconnection
G. Hornig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
2 Magnetic Reconnection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
3 Vortex Reconnection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379

Current-Sheet Formation
near a Hyperbolic Magnetic Neutral Line
B.K. Shivamoggi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
2 Current-Sheet Formation at a Hyperbolic Magnetic Neutral Line
in a Stagnation-Point Plasma Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
3 Effect of a Uniform Shear–Strain in the Plasma Flow . . . . . . . . . . . . . . . 385
4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387

Nonlocality in Turbulence
A. Tsinober . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
1 Introduction and Simple Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
2 Different Aspects of Nonlocality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
2.1 Direct Coupling Between Large and Small Scales . . . . . . . . . . . . . . 392
3 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
 Contents XVII

Part VIII Helium 3 and Other Systems

Quantized Vorticity in Superfluid 3 He-A:

Structure and Dynamics
R. Blaauwgeers, V.B. Eltsov, M. Krusius, J. Ruohio, R. Schanen . . . . . . . . 399
1 Superfluid 3 He . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
2 Order-Parameter Texture and Superflow in 3 He-A . . . . . . . . . . . . . . . . . . 401
3 Double-Quantum Vortex Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
4 Vortex Sheet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
5 Dynamic Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
6 Summary and Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419

Vortices in Metastable 4 He Films

R. Blossey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
1 Wetting Properties of 4 He on Weak-Binding Alkali Metals . . . . . . . . . . . 421
1.1 Wetting Transitions of Liquid Helium . . . . . . . . . . . . . . . . . . . . . . . . 421
1.2 Interface Model Description of Wetting Transitions . . . . . . . . . . . . . 422
2 Lifetime of an Undercooled Film . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
3 Application to 4 He/Cs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427

Quantum Hall Effect Breakdown Steps and Possible Analogies

with Classical and Superfluid Hydrodynamics
L. Eaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
2 Model and Comparison with Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . 430
3 Analogies with Classical and Quantum Fluids . . . . . . . . . . . . . . . . . . . . . . 434
4 The Breakdown Steps and Their Relation
to Other Types of QHE Breakdown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
Atomic Bose Condensate with a Spin Structure:
The Use of the Bloch State
H. Kuratsuji . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
2 Order Parameter and Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
3 Hydrodynamical Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440
4 Vortex State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
4.1 The Profile of a Single Vortex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
4.2 Vortex Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
XVIII Contents

Quantum Dynamics of Vortex–Antivortex Pairs in a Circular Box

V. Penna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
2 Canonical Quantization of Planar Vortices . . . . . . . . . . . . . . . . . . . . . . . . . 446
2.1 The Spectrum of Unbounded Vortex Pairs . . . . . . . . . . . . . . . . . . . . 447
3 Pair Quantum Dynamics in a Circular Box . . . . . . . . . . . . . . . . . . . . . . . . 448
3.1 Spectral Structure of Low Energy States . . . . . . . . . . . . . . . . . . . . . . 449
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451

Index of Topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453

List of Contributors

C.S. Adams R.J. Donnelly

Physics Department Physics Department
University of Durham University of Oregon
Durham DH1 3LE, UK Eugene, OR 97403, USA
[Link]@[Link] russ@[Link]

C.F. Barenghi
University of Newcastle L. Eaves
Mathematics Department School of Physics and Astronomy
Newcastle upon Tyne, NE1 7RU University of Nottingham
UK Nottingham NG7 2RD, UK
[Link]@[Link] [Link]@[Link]

N. Berloff
Mathematics Department A. Fetter
University of California Geballe laboratory for Advanced
Los Angeles, CA 90095-1555, USA Materials
nberloff@[Link] Stanford University
Stanford, CA 94305-4045, USA
R. Blossey fetter@[Link]
Department of Physics
University of Essen
45117 Essen, Germany K.L. Henderson
blossey@[Link] Faculty of Computer Studies and
Mathematics
A. Brandenburg University of the West of England
NORDITA Bristol BS16 1QY, UK
Blegdamsvej 17 [Link]@[Link]
2100 Copenaghen, Denmark
brandenb@[Link]
D.D. Holm
H.P. Buechler Theoretical Division
Theoretical Physics Mail Stop B284
ETH Los Alamos National Laboratory
8093 Zuerich, Switzerland Los Alamos NM 87545, USA
buechler@[Link] holm@[Link]
XX List of Contributors

G. Hornig M. Leadbeater
Dept. Theoretical Physics IV Physics Department
Ruhr-Universitaet Bochum University of Durham
44780 Bochum, Germany Durham DH1 3LE, UK
gh@[Link] [Link]@[Link]

C. Huepe T. Lipniacki
James Frank Institute Institute of Fundamental Technologi-
University of Chicago cal Research
5640 S. Ellis Avenue Świȩtokrzyska St. 21
Chicago, IL 60637, USA 00-049 Warsaw, Poland
cristian@[Link] tlipnia@[Link]

O. Idowu P.V.E. McClintock

Center for Turbulence Research University of Lancaster
Stanford University Physics Department
Stanford, CA 94305-3030, USA Lancaster LA1 4YB, UK
idowu@[Link] [Link]@[Link]

D. Kivotides
S. Nazarenko
Mathematics Department
Mathematics Institute
University of Newcastle
University of Warwick
Newcastle NE1 7RU, UK
Coventry CV4 7AL, UK
[Link]@[Link]
snazar@[Link]

J. Koplik
Levich Institute, T-1M S.K. Nemirovskii
City College of New York Institute of Thermosphysics
New York, NY 10031, USA 630090 Novosibirsk, Russia
koplik@[Link] nem@[Link]
koplik@[Link]
J.J. Niemela
M. Krusius Physics Department
Low Temperature Laboratory University of Oregon
Helsinki University of Technology Eugene, OR 97403, USA
02015 HUT joe@[Link]
Finland
krusius@[Link] M. Niemetz
Institut für Experimentelle
H. Kuratsuji und Angewandte Physik
Department of Physics Universität Regensburg
Ritsumeikan University 93040 Regensburg, Germany
Kusatsu City 525-8577, Japan [Link]@
kra@[Link] [Link]
 List of Contributors XXI

V. Penna B.K. Shivamoggi

Dipartimento di Fisica University of Central Florida
and INFM Orlando, FL 32816, USA
Politecnico di Torino ijjmms@[Link]
[Link] Duca degli Abruzzi 24
10129 Torino, Italy E. Sonin
penna@[Link] Racah Institute of Physics
Hebrew University of Jerusalem
L.M. Pismen Givat Ram,
Department of Chemical Engineering Jerusalem 91904, Israel
and Minerva Center for Nonlinear sonin@[Link]
Physics of Complex Systems
Technion - Israel Institute L. Skrbek
of Technology KFNT MFF UK
32000 Haifa, Israel Charles University
V Holesovickach 2
S. Rica 180 00 Prague 8, Czech Republic
CMM CNRS UCHILE skrbek@[Link]
Av Blanco Encalada 2120
Santiago, Chile B. Svistunov
rica@[Link] Russian Research Center
rica@[Link] Kurchatov Institute
123182 Moscow, Russia
L.R. Ricca svist@[Link]
Mathematics Department
University College London A. Tsinober
Gower Street Faculty of Engineering
London WC1E 6BT, UK Tel Aviv University
ricca@[Link] Tel Aviv, Israel
tsinober@[Link]
P.H. Roberts
Mathematics Department M. Tsubota
University of California Department of Physics
Los Angeles, CA 90095-1555, USA Osaka City University
roberts@[Link] Osaka 558-8585, Japan
tsubota@[Link]
W. Schoepe
Institut für Experimentelle und S.W. Van Sciver
Angewandte Physik National High Magnetic Field
Universität Regensburg Laboratory
93040 Regensburg, Germany Florida State University
[Link]@ Tallahassee FL 32310, USA
[Link] vnsciver@[Link]
XXII List of Contributors

É. Varoquaux University of Birmingham

CNRS-Université Paris-Sud Birmingham B15 2TT, UK
Laboratoire de Physique des Solides [Link]@[Link]
Bâtiment 510
F-91405 Orsay Cedex, France G. Williams
varoquaux@[Link] Department of Physics and Astronomy
University of California
W.F. Vinen Los Angeles, CA 90095, USA
Physics Department gaw@[Link]
Introduction to Superfluid Vortices
and Turbulence

Carlo F. Barenghi

Mathematics Department, University of Newcastle, Newcastle NE1 7RU, UK

1 The Two-Fluid Model

My aim in this article is to introduce the basic properties of quantized vortex lines
in Helium II and summarize the main experimental observations of superfluid
turbulence. Then I shall discuss a selection of the theoretical methods used to
study quantized vorticity and turbulence and the results obtained using these
methods.
The liquid state of 4 He exists in two phases: a high temperature phase called
Helium I, and a low temperature phase, called Helium II. The two phases are
separated by a transition called the lambda transition, which occurs at the crit-
ical temperature T = Tλ = 2.172 K at saturated vapour pressure and marks
the onset of Bose Einstein condensation (BEC) and quantum order. The phe-
nomenon of BEC is described in the article of Stringari. Helium I is a classical
fluid which obeys the ordinary Navier - Stokes equations. Hereafter the focus of
attention is only Helium II.
A simple, phenomenological model which explains the motion of Helium II
is the two - fluid theory of Tisza and Landau [1]. In this model Helium II is
described as the intimate mixture of two fluid components which penetrate each
others, the normal fluid and the superfluid. Each fluid component has its own
density and velocity field, ρn and vn for the normal fluid and ρs and vs for
the superfluid. The total density of Helium II is ρ = ρn + ρs . The superfluid
component is irrotational, and, since it carries nor entropy nor viscosity, is similar
to a classical, inviscid Euler fluid. The normal fluid component is a gas of thermal
excitations called phonons and rotons depending on the wavenumber. The normal
fluid carries the entire entropy and viscosity of Helium II and is similar to a
classical, viscous Navier - Stokes fluid.
The relative proportion of normal fluid and superfluid is determined by the
absolute temperature T . At T = 0 Helium II is entirely superfluid: ρs /ρ = 1 and
ρn /ρ = 0. If the temperature is increased the superfluid fraction decreases and
the normal fluid fraction increases, until, at T = Tλ , Helium II becomes entirely
normal: ρs /ρ = 0 and ρn /ρ = 1. The temperature dependence of the two fluid’s
fractions is nonlinear: for example ρn /ρ drops from 100 percent at T = Tλ to
55 percent at 2.0 K and to 7.5 percent at 1.4 K, and is effectively negligible at
temperatures below 1 K.
The two - fluid model explains many observed phenomena. Among them
it is worth mentioning second sound and thermal counterflow because they are
important in the study of turbulence. Ordinary sound in Helium II is called first

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 3–14, 2001.

c Springer-Verlag Berlin Heidelberg 2001
4 C.F. Barenghi

sound. A first sound wave is an oscillation of density ρ and pressure P in which

temperature T and entropy S remain almost constant and vn and vs move in
phase with each others. A second sound wave, on the contrary, is an oscillation
of T and S in which ρ and P remain almost constant and vn and vs move in
antiphase. Second sound is used to detect quantized vorticity.
Thermal counterflow is Helium II’s special way to transfer heat. Consider a
channel which is closed at one end and open to the helium bath at the other
end. At the closed end a resistor dissipates a known heat flux W . In an ordinary
fluid, such as water or Helium I, heat is transferred away from the resistor by
conduction, provided that one is careful to prevent convective motion, so the
heat flux W is proportional to the temperature gradient ∇T and there is a well
defined thermal conductivity at small W . In Helium II the heat is carried by
the normal fluid away from the resistor, W = ρST vn . Because of the presence
of the closed end, however, the mass flux is zero, j = ρs vs + ρn vn = 0, so some
superfluid must flow toward the resistor to conserve mass, vs = −(ρn /ρs )vn . In
this way a counterflow vns = vn − vs is generated which is proportional to the
applied heat flux, vns = W/(ρs ST ). If W exceeds a critical value, then superfluid
turbulence is generated. Turbulence limits the heat transfer properties of Helium
II, so it is relevant in the engineering applications, as described in the article of
Van Sciver.

2 Quantized Vortex Lines

The quantization of the circulation of the superfluid makes the hydrodynamics of
helium II particularly interesting [2]. Superfluid vortex lines appear when helium
II rotates or moves faster than a critical velocity. This vortex nucleation process
has been the subject of many investigations and is described in the articles of
Adams and Rica. Superfluid vortex lines can be spatially organized (laminar
vortex flows) or disorganized (turbulent vortex tangles). The key property of a
superfluid vortex line, discovered by Onsager and developed by Feynman, is that
the circulation is quantised, that is to say

h
vs · dl = , (1)
C m

where C is a circular path around the axis of the vortex. The ratio Γ = h/m of
Plank’s constant and helium’s mass is called the quantum of circulation and has
value Γ = 9.97 × 10−4 cm2 /sec.
The simplest way to create superfluid vortex lines is to rotate a cylinder
filled with Helium II at constant angular velocity Ω. Provided that Ω is large
enough, superfluid vortex line appear and form on ordered array of areal density
n = 2Ω/Γ , all vortex lines being aligned along the axis of rotation. In this way
the superfluid mimics the vorticity 2Ω of the solid body rotating normal fluid,
each vortex line contributing one quantum to the total circulation.
 Superfluid Vortices and Turbulence 5

3 Modelling the Vortex Lines

3.1 Microscopic Model
Using cylindrical coordinates r, φ, z it follows from equation (1) that the super-
fluid velocity around the axis of a vortex line is vs,φ = Γ/2πr, hence it diverges
as r → 0. To understand what happens as r → 0 it is instructive to use the
nonlinear Schroedinger equation (NLSE), which is discussed in the article of
Roberts. The NLSE models a gas of bosons which interact with each others via
a delta function repulsive potential of strength V0 :

∂ψ 2 2
i =− ∇ ψ − mEψ + V0 ψ|ψ|2 , (8)
∂t 2m
Here E is the energy per unit mass and  = h/2π. The wavefunction ψ can be
written as ψ = AeiΦ in terms of an amplitude A and a phase Φ. In this way
one can define the condensate’s density ρBEC = mA2 and velocity vBEC =
(/m)∇Φ; the last relation confirms that the superfluid is irrotational as en-
visaged by Landau. This transformation establishes the hydrodynamics of the
model: equation (8) is equivalent to a continuity equation and an Euler equation
(modified by the so called quantum pressure term).
The NLSE has a vortex solution: if Φ is the azimuthal angle φ then we have
vBEC,φ = Γ/(2πr) which is the Onsager - Feynman vortex. Substitution into
the NLSE yields a differential equation for ρBEC . One finds that ρBEC tends
to the bulk value m2 E/V0 for r → ∞, and that ρBEC → 0 for r → 0. The
characteristic distance over which ρBEC changes from its bulk value to zero is
a0 ≈ 10−8 cm. This distance is called the vortex core parameter. We conclude
that the superfluid vortex line is hollow at the core. Geometrically, a vortex
line transforms the volume occupied by the superfluid into a multiply connected
region.
Hereafter we identify ρBEC with ρs at absolute zero and vBEC with vs . It
must be noted that this identification is convenient from the point of view of
having a simple hydrodynamics model but is not entirely correct. The reason is
that Helium II is a dense fluid, not the weakly interacting Bose gas described
by the NLSE, so the condensate is not the same as the superfluid component.
One should compare the case of Helium II (in which the NLSE is a rather
approximate model) with the case of BEC in clouds of trapped alkali atoms (in
which the NLSE is a better model because the bosons’ interaction is weaker).
Another drawback of the NLSE is that it fails to describe the observed dispersion
relation of Helium II at high momenta. If one studies small oscillations of the
uniform solution of the NLSE and interprets them as thermal excitations, one
finds a dispersion relation E = E(p) in which the energy E is proportional
to the momentum p at small p (phonons) and then becomes quadratic in p at
high p (free particles). The spectrum of excitations observed in Helium II is
different, because the phonon part is followed by the rotons’ minimum. Despite
these shortcomings, of which one must be aware, the NLSE is much used as a
convenient hydrodynamical model of Helium II at T = 0.
6 C.F. Barenghi

3.2 Mesoscopic Model

An important feature of the NLSE model is that it makes visible what happens
on the microscopic scale (≈ 10−8 cm) of the vortex core parameter. This makes
the NLSE a useful tool to investigate phenomena such as vortex nucleation
and vortex reconnections. There are however other problems in which one is
concerned with the behaviour of a large number of vortex lines. In these problems
the resolution of what happens on the microscopic scale is not necessary and can
even be a waste of computer resources. It is more appropriate to use the classical
vortex dynamics model.
The vortex dynamics model was pioneered by Schwarz [3] and consists in
representing a vortex line as a curve s = s(ξ, t) in three-dimenensional space,
where ξ is arclength and t is time. We call s = ds/dξ and note that the vectors
s, s and s × s are perpendicular to each others and point along the tangent,
normal and binormal respectively. To determine the equation of motion of s
we must identify the forces acting upon the line: the Magnus force fM and, at
nonzero temperature, the drag force fD .
The Magnus force arises when a body with circulation about it moves in
a flow: the circulation creates an increased total velocity of fluid on one side,
which results in excess pressure from the other side. Since the key ingredient is
the circulation rather than the details of the body, we apply the concept to a
vortex line and write
fM = ρs Γ s × (vL − vs,tot ), (2)
where vL = ds/dt is the velocity of the line in the laboratory frame and vs,tot is
the total velocity of the surrounding superfluid, also in the laboratory frame. The
velocity vs,tot consists of two parts: any superfluid velocity applied externally
and the self - induced velocity of the vortex line.

vs,tot = vs + vi , (3)

The self induced velocity vi describes the motion which a vortex line induces
onto itself because of its own curvature and is determined by the Biot - Savart
(BS) law 
Γ (z − s) × dz
vi (s) = , (4)
4π |z − s|3
The Biot - Savart law is sometimes replaced by the Local Induction Approxima-
tion (LIA), which is
vi ≈ βs × s , (5)
where β = Γ/(4π)log(1/(|s |a0 )). Since |s | = 1/R where R is the local radius
of curvature, we have
Γ R
vi ≈ ln( )b̂, (6)
4π a0
where b̂ is the binormal.
The drag force fD arises from the mutual friction between the superfluid
vortex lines and the normal fluid [4]. Normal fluid flowing with velocity vn past
 Superfluid Vortices and Turbulence 7

a vortex core exerts a frictional force fD per unit length on the superfluid in the
neighborhood of the core given by

fD = −αρs Γ s × [s × (vn − vs,tot )] − α ρs Γ s × (vn − vs,tot ), (7)

The dimensionless parameters α and α are temperature dependent and are of-
ten written in terms of mutual friction coefficients B and B  defined by α =
ρn B/(2ρ) and α = ρn B  /(2ρ). The values of B and B  are known from experi-
ments. Samuels and Donnelly [5] showed that, at least in the high temperature
range in which most experiments are performed, the friction arises from the
scattering of rotons from the velocity field of a vortex line. The calculation of
the mutual friction parameters over the entire temperature range is still an open
question and has subtle aspects, as explained in the article by Sonin. An impor-
tant effect of the mutual friction is that it modifies the propagation of second
sound. By measuring the second sound attenuation one can determine the su-
perfluid vortex line density L0 , defined as the length of vortex line per unit
volume.
Now that fM and fD are identified we can make use of the fact that the sum
of all forces is zero as the line’s inertia is negligible:

fD + fM = 0, (8)

Hence, solving for ds/dt, we obtain Schwarz’s equation

ds
= vs + vi + αs × (vn − vs − vi ) + α (vn − vs − vi ), (9)
dt
An algorithm to numerically simulate the time evolution of any arbitrary
configuration of vortex lines can be developed on the basis of Schwarz’s equa-
tion and is described in the article of Samuels. Here it suffices to say that an
initial vortex configuration is discretized into N points. The time evolution of
each point is calculated using (9), given externally applied fields vs and vn and
given the temperature T , which determines the friction coefficients α and α .
The transverse part of the mutual friction, proportional to α , is smaller and is
sometimes neglected. The number of points N and the time step must be allowed
to vary during the evolution to take into account the appearance of regions of
high or low curvature. Note that, if one uses the BS law, the computational time
is proportional to N 2 , while, if one uses the LIA, this time is only proportional to
N . Numerical simulations of vortex tangles based on the BS are therefore compu-
tationally expensive. However the use of the LIA can give misleading results [6].
Finally the numerical simulation must be able to perform vortex reconnections
when two vortex lines come sufficiently close to each others. This process is an
arbitrary assumption in the context of the dynamics of vortex filaments, but is
justified by a microscopic calculation [7] performed using the NLSE, as explained
in the article by Koplik.
8 C.F. Barenghi

3.3 Macroscopic Model

Besides the NLSE model (in which the vortex core is visible) and the vortex
dynamics model (in which the core is not visible but the vortex line is) there
is a third macroscopic model in which the individual vortex lines are not visi-
ble and Helium II is considered as a continuous vortex flow. The third model,
called the HVBK model[8] [9], is useful to describe laminar flows in which the
vortex line are spatially organized. Examples are solid body rotation, flows in
an rotating annulus or cavity, and Taylor - Couette flow. The HVBK model is
a generalization of Landau’s equations to include the presence of vortices. The
fluid particles of the model are assumed to be large enough to be threaded by
many vortex lines which are aligned in the same direction. In this way the indi-
vidual vortex lines are not visible, the superfluid is treated as a continuum and
we can define a macroscopic, nonzero superfluid vorticity ωs , despite the fact
that, microscopically, the superfluid velocity field obeys ∇ × vs = 0. Clearly the
HVBK equations are valid only if the vortex lines are organized spatially and not
randomly oriented, and if the length scales of the flow under consideration are
much bigger than the average separation between the vortex lines. An example
is the simple case of Helium II inside a rotating cylinder, for which ωs = 2Ωẑ.
The incompressible HVBK equations are
∂vn 1 ρs ρs
+ (vn · ∇)vn = − ∇P − S∇T + νn ∇2 vn + F, (10)
∂t ρ ρn ρ
∂vs 1 ρn
+ (vs · ∇)vs = − ∇P + S∇T + T − F, (11)
∂t ρ ρ
where we have defined
ωs = ∇ × vs , (12)

B B
F= ω
 s × [ωs × (vn − vs − νs ∇ × ω
 s )] + ωs × (vn − vs − νs ∇ × ω
 s ), (13)
2 2
ω
 s = ωs /|ωs |, (14)
T = −νs ωs × (∇ × ω
s ), (15)
νs = (Γ/4π) log(b0 /a0 ), (16)
The quantities F, T and νs are respectively the friction force, the tension
force and the vortex tension parameter, and b0 = (2ωs /Γ )−1/2 is the intervor-
tex spacing. Note that νs has the same dimension of a kinematic viscosity, but
physically it is very different: it is related to the ability of a superfluid fluid
particle to oscillate because of the vortex waves which can be excited along the
vortex lines threading the fluid particle itself. Note that without F and T the
HVBK equations are formally the same as the original two - fluid equations of
Landau. The HVBK equations have interesting limits. If T → Tλ then ρs → 0 so
the normal fluid equation (10) becomes the classical Navier - Stokes equation. If
T → 0 then ρn → 0 so the superfluid equation (11) describes a pure superflow;
 Superfluid Vortices and Turbulence 9

by setting Plank’s constant equal to zero we have then νs = 0 and the pure
superflow equation becomes the classical Euler equation.
The HVBK model has been used with success to study the transition from
Couette flow to Taylor vortex flow [10]: Barenghi’s predictions [11] of the critical
Reynolds number of the transition and its temperature dependence were con-
firmed by the experiments [12]. Taylor - Couette flow has also been studied in
the nonlinear Taylor vortex flow regime [13] [14] and the results are in agreement
with the observations, providing a further test of the theory. These results are
described in the article by Henderson.

4 Turbulence
Superfluid turbulence manifests itself as a tangle of vortex lines and can be
generated in many ways. Turbulent thermal counterflow was the first turbulent
flow which was studied in detail in a series of pioneering papers by Vinen [15].
Since this flow has no classical analogy it deserves a separate discussion. Other
ways to generate turbulence are more classical in character, and we refer to them
as turbulent coflows.

4.1 Turbulent Counterflows

As said in the first section, laminar counterflow breaks down if the heat flux W
exceeds a critical value Wc . Corresponding to Wc there is a critical counterflow
velocity vns = vc1 = Wc /(ρs ST ). At this critical velocity a vortex tangle appears.
The simplest way to characterize the tangle is to measure or compute its vortex
line density L0 . A great number of measurements of L0 were performed in pipes
2
and channels and showed that L0 = γvns for vns > vc1 where γ is some temper-
ature dependent parameter. Further measurements in circular pipes indicated
that there exists a second critical velocity vc2 at which the vortex line density
L0 becomes suddenly larger. The region of weak turbulence vc1 < vns < vc2 and
the region of strong turbulence v > vc2 are called the T-1 and the T-2 turbu-
lent states respectively [16]. On the theoretical side, the numerical simulations of
Schwarz [3] based on the vortex dyanmics approach confirmed the existence of
a self - sustaining vortex tangle driven by a constant, spatially uniform vns , and
gave values of L0 consistent with the observations in the T-2 state. The nature of
the first turbulent states and the physical meaning of the transition at vns = vc2
were a puzzle until Melotte and Barenghi [17] showed that the transition from
the T-1 to the T-2 state is related to the onset of normal fluid turbulence (see
section 6).

4.2 Turbulent Coflows

Turbulence can also be induced in more traditional ways by driving a mass flow,
spinning discs or propellers, towing a grid or a sphere, using shocks, ultrasound,
jets and rotating cylinders (Taylor - Couette flow). Considering all these results
10 C.F. Barenghi

together, the general trend is that the slow, laminar flow of Helium II, with or
without vortices, tends to be rather different from the flow of a classical fluid,
but when Helium II moves fast and is driven turbulent it seems to behave like a
classical turbulent flow.
To characterize the turbulence we use the Reynolds number Re = U L/ν
where L is the length scale, U the velocity scale and ν the kinematic viscosity.
Examples of the observed classical features of Helium II turbulence are the fol-
lowing. Mass flow rates and pressure drops at Re ≈ 106 can be well described
by using classical relations for high Reynolds number classical flows [18]. Ex-
periments on large scale turbulent vortex rings at Re ≈ 4 × 104 detect normal
fluid vorticity and superfluid vorticity moving together as a single structure [19]
Experiments on turbulent Taylor - Couette flow at Re ≈ 4×103 show the typical
structures of classical turbulent Taylor - Couette flow [20]. Experiments on the
decay of superfluid vorticity created by towing a grid show that the decay in
time obeys the same laws as of the decay of classical turbulence [21]. Moreover
the decay appears to be independent of temperatures in the explored range,
from the lambda region down to 1.4 K, where the normal fluid fraction is only
7.5 percent. Experiments on turbulence created by rotating blades [22] show the
classical Kolmogorov −5/3 power spectrum in the temperature range explored,
from the lambda region down to T = 1.4 K again. Finally experiments on the
drag on a moving sphere (Re ≈ 105 ) show the same drag crisis observed in a
classical fluid [23].
The temperature independence of these observations is interesting. The nor-
mal fluid must be responsible for these classical aspects, but the dynamical im-
portance of the normal fluid should be related to the fraction ρn /ρ, so it should
be negligible at temperatures as low as 1.4 K. Since a large number of quantized
vortex lines must be present in these turbulent flow, it is speculated that they
are able to lock together the two fluid components of Helium II into a single
fluid which behaves somewhat like a classical turbulent fluid. This is a topic of
much current interest‘[24] [25] and is discussed in the article of Vinen.

5 Motion of Superfluid Vortices for a Given Normal Fluid

In the next two sections we discuss some selected examples about the interac-
tion of superfluid vortices and the normal fluid. The numerical calculations done
by Schwarz using the vortex dynamics model (section 3.2) were concernd with
thermal counterflow. The calculations were successful; above all, they gave great
insight into the nature of turbulence, confirming the existence of a vortex tangle,
and brought into attention the issue of vortex reconnections. However Schwarz’s
approach had an important limitation: it assumed that the quantity vn − vs
is constant in time and space. The work of Schwarz was followed up by other
calculations which made different assumptions about the normal fluid: uniform
flow [26], Poiseuille flow [26] [27], a Gaussian vortex [28], ABC flows [29]. All these
calculations were kinematic in character, because the driving field (vn for mass
flow and vn − vs for counterflow) were imposed at the beginning of the calcula-
 Superfluid Vortices and Turbulence 11

tion and never changed, neglecting the back reaction of the superfluid vortices.
Not surprising, the vortex tangles calculatedusing different driving fields looked
different from each others. The success of the original calculation of Schwarz in
reproducing a vortex line density L0 consistent with the experiments in the T-2
state was probably due to the fact that the uniform vn − vs profile used by
Schwarz modelled well the average, flattened turbulence profile in the channel
flows under consideration.
Despite this limitation, kinematic calculations are clearly useful and can shed
light on important physical mechanisms. A particularly interesting mechanism
which is relevant to turbulence is the Ostermeier- Glaberson instability. This
is an instability of Kelvin vortex waves which takes place if the component of
the normal fluid velocity in the direction parallel to the vortex lines exceeds a
critical value. The instability was first observed by Cheng, Cromar and Don-
nelly [30], but is was Ostermeier and Glaberson [31] who explained it and it
was Samuels [28] who realized its importance in turbulence. Another physical
mechanism, which is apparent in the kinematic numerical calculations [28] [29]
is vorticity matching: once the Ostermeier - Glaberson instability has generated
superfluid vortex lines by extracting energy from the normal fluid, then the vor-
tex lines become attracted to the regions of concentrated normal fluid vorticity.
Therefore, although the local superfluid velocity pattern in the bundles is very
complicated, the averaged vorticity ωs is similar to the vorticity of the driving
normal fluid. A further interesting application of the Ostermeier - Glaberson
instability is that it creates a damping length scale  for superfluid turbulence:
superfluid structures at length scale smaller than  will lose energy to the normal
fluid and be dissipated [25].

6 Motion of the Normal Fluid

at Given Superfluid Vortices
The kinematic approach can also be used to study the behaviour of the nor-
mal fluid given the superfluid vortices rather than viceversa. An example is the
calculation of Melotte and Barenghi [17] who studied the stability of normal
fluid motion induced by heat transfer in the presence of a uniform tangle of
superfluid vorticity of density L0 . The governing equation is a Navier - Stokes
equation modified by the introduction of a friction term
∂vn 1 ρs 1
+ (vn · ∇)vn = − ∇P − S∇T + νn ∇2 vn + F, (20)
∂t ρ ρn ρn
where
Bρs ρn
F= gωs (vs − vn ), (21)
2ρ
ωs = Γ L0 , and, assuming an isotropic tangle, g ≈ 2/3. Melotte and Barenghi’s
stability calculation determined the critical vortex line density at which the
normal fluid’s profile becomes turbulent. The results are in agreement with the
12 C.F. Barenghi

measurements of the transition from the weak T-1 state (in which the superfluid
is turbulent but the normal fluid is not) to the strong T-2 state (in which both
fluids are turbulent).
The same kinematic approach has been used to study how the superfluid
vortices affect the stability of normal fluid in channel flows. This modified Orr -
Sommerfeld problem[32] is described in the article by Godfrey.

7 Fully Coupled Motion of Superfluid Vortices

and Normal Fluid
The limitations of the kinematic approach can be overcome if one allows the
normal fluid and the superfluid vortices to determine each others self - consis-
tently during the evolution. Essentially one has to combine the (Lagrangian)
calculation of the superfluid vortices based on vortex dynamics (section 2) with
with the (Eulerian) calculation of the normal fluid based on a modified Navier
- Stokes equation (section 5). However, since the velocity field vn which is com-
puted is local, the equation of Schwarz (9) requires a modification[33]. At the
place of (9) one has

ds
= vs + h1 vsi + h2 s × (vn − vi ) − h3 s × s × vn , (22)
dt
where vi is given by (4). In the absence of friction we have h2 = h3 = 0 and
h1 = 1. The normal fluid is determined by a modified Navier - Stokes equation
like (20), but now F is obtained numerically by considering the friction force on
the normal fluid per unit length of superfluid vortex line

f = Ds × s × (vn − vL ) + Dt s × (vn − vL ), (23)

where D and Dt are mutual friction coefficients, and summing the contribution
of each segment of s that falls within the computational grid cell of the normal
fluid.
The drawback of the fully - coupled approch is the computational cost. A
useful compromize is to implement it in two dimensions x, y neglecting the z
dependence: in this way the vortex lines becomes vortex points (the intersection
of vortex lines with the plane z = 0) and the normal fluid can be obtained
easily using the stream function - vorticity formulation. It is found that a single
superfluid vortex point induces an elongates normal fluid jet [34] whose intensity
depends on the temperature, hence creating a dipolar vorticity structure in the
normal fluid. A similar calculation in three dimensions performed by Kivotides,
Barenghi and Samuels [35] showed that a superfluid vortex ring creates around
itself a normal fluid structure which consists of two coaxial vortex vortex rings
of opposite polarity.
Since the normal fluid is accelerated by the superfluid vortices by friction
which depends on the difference between vn and vL , the normal fluid speed
cannot exceed the vortex lines’ speed. If one considers the typical vortex line
 Superfluid Vortices and Turbulence 13

density L0 of experiments and estimates vL ≈ Γ/2πδ using the average inter-

−1/2
vortex spacing δ ≈ L0 , one finds an upper bound for vn , using which one
estimates the normal fluid Reynolds number to be in the range from 1 to 10, too
small for turbulence[36].
This conclusion refers only to the ability of vortex line to make the normal
fluid turbulent by directly stirring it. A second mechanism in which the vortex
line can make the normal fluid turbulent is by causing instabilities in a mean
flow, as showed in section 6. Finally normal fluid turbulence can also be induced
directly by the boundaries.

8 Discussion

I have mentioned only a selection of topics of current interest in the study of

quantized vorticity. Much work is in progress. It is clear that the NLSE will
be used more and more to understand the details of fundamental processes:
an example is vortex reconnections[37]. Similarly, the use of vortex dynamics
simulations in the absence of friction is proving useful to understand processes
such as the turbulence cascades of vortex waves [38]. At finite temperatures,
fully coupled numerical simulations are proving useful to understand the effects
of the reconnections on the normal fluid[39] and hopefully will be able in the
near future achieve contact between theory and experiment.

References
1. L.D. Landau and E.M. Lifshitz: Fluid Mechanics, 2nd edn. (Pergamon Press, Lon-
don 1987)
2. R.J. Donnelly:Quantized Vortices in Helium II, Cambridge University Press, Cam-
bridge (1991).
3. K.W. Schwarz: Phys. Rev. Lett. 49, 283 (1982); Phys. Rev. B 31 5782 (1985); Phys.
Rev. B 38 2398 (1988)
4. C.F. Barenghi, R,J, Donnelly and W.F. Vinen: J. Low Temp. Phys. 52 189 (1983)
5. D.C. Samuels and R.J. Donnelly: Phys. Rev. Lett. 65 187 (1990).
6. R.L. Ricca, D.C. Samuels and C.F. Barenghi: J. Fluid Mech. 391 29 (1999)
7. J. Koplik and H. Levine: Phys. Rev. Lett. 71, 1375 (1993).
8. H.E. Hall and W.F. Vinen: Proc. Roy. Soc. London A 238, 215 (1956).
9. R.N. Hills and P.H. Roberts: Arch. Rat. Mech. Anal. 66, 43 (1977).
10. C.F. Barenghi and C.A. Jones: J. Fluid Mech. 197 551 (1988)
11. C.F. Barenghi: Phys. Rev. B 45, 2290 (1992)
12. C.J. Swanson and R.J. Donnelly: Phys. Rev. Lett. 67 1578 (1991)
13. K.L. Henderson, C.F. Barenghi and C.A. Jones: J. Fluid Mech. 283 329 (1995).
14. K.L. Henderson and C.F. Barenghi: Phys. Lett. A 191 438 (1994).
15. W.F. Vinen: Proc. Roy. Soc. A 240, 114 (1957); ibidem 240, 128 (1957); ibidem
242, 493 (1957); ibidem 243 400 (1957).
16. J.T. Tough: ‘Superfluid turbulence’. In: Progress in Low Temperature Physics, vol.
VIII, ed. by D.F. Brewer (North Holland, Amsterdam 1987) p. 133.
17. D.J. Melotte and C.F. Barenghi: Phys. Rev. Lett. 80 4181 (1998).
14 C.F. Barenghi

18. P.L. Walstrom, J.G. Weisend, J.R. Maddocks and S.V. VanSciver: Cryogenics 28
101 (1988).
19. H. Borner, T. Schmeling and D.W. Schmidt: Phys. Fluids 26 1410 (1983)
20. F. Bielert and G. Stamm: Cryogenics 33 938 (1993)
21. M.R. Smith, R.J. Donnelly, N. Goldenfeld and W.F. Vinen: Phys. Rev. Lett. 71
2583 (1993)
22. J. Maurer and P. Tabeling: Europhysics Lett. 43 29 (1998)
23. M.R. Smith, D.K. Hilton and S. V. VanSciver: Phys. Fluids. 11 751 (1999)
24. W.F. Vinen: Phys. Rev. B 61 1410 (2000)
25. D.C. Samuels and D. Kivotides: Phys. Rev. Lett. 83 5306 (1999)
26. R.G.K.M. Aarts and A.T.A.M. deWaele: Phys. Rev. B 50 10069 (1994)
27. D.C. Samuels: Phys. Rev. B 46 11714 (1992)
28. D.C. Samuels: Phys. Rev. B 47 1107 (1993)
29. C.F. Barenghi, G. Bauer, D.C. Samuels and R.J. Donnelly: Phys. Fluids 9 2631
(1997)
30. D.K. Cheng, M.W. Cromar and R.J. Donnelly: Phys. Rev. Lett. 31 433 (1973)
31. R.M. Ostermeier and W.I. Glaberson: J. Low Temp. Phys. 21 191 (1975).
32. S.P. Godfrey, C.F. Barenghi and D.C. Samuels: Physica B 284 66 (2000); Phys.
Fluids 13 983 (2001).
33. O.C. Idowu, D. Kivotides, C.F. Barenghi and D.C. Samuels: J. Low Temp. Physics,
120 269 (2000).
34. O.C. Idowu, A. Willis, D.C. Samuels and C.F. Barenghi: Phys. Rev. B 62 3409
(2000)
35. D. Kivotides, C.F. Barenghi and D.C. Samuels: Science 290 777 (2000).
36. C.F. Barenghi, D. Kivotides, O. Idowu and D.C. Samuels: J. Low Temp. Physics,
121 377 (2000).
37. M. Leadbeater, T. Winiecki, D.C. Samuels, C.F. barenghi and C.S. Adams: Phys.
Rev. Lett. 86 1410 (2001).
38. D. Kivotides, C. Vassilicos, C.F. Barenghi and D.C. Samuels: Phys. Rev. Lett. 86
3080 (2001).
39. D. Kivotides, C.F. Barenghi and D.C. Samuels: to be published in Europhys. Lett.
(2001).
An Introduction to Experiments
on Superfluid Turbulence

Russell J. Donnelly

Cryogenic Helium Turbulence Laboratory, Department of Physics,

University of Oregon, Eugene, Oregon 97403

1 Introduction
A description of the experimental background of superfluid turbulence was as-
signed to me for this lecture. Superfluid turbulence, or as some call it, quantum
turbulence, has been an active field of physics since the 1950’s. The field was
pioneered experimentally and theoretically by Joe Vinen and as such is approach-
ing a half century in age. It is safe to say that with few exceptions the results
are unknown to those investigators who are interested in classical turbulence,
that is the kind of investigation which has been pioneered by Taylor, Landau,
Kolmogorov and others.
It has only recently been realized that liquid helium I, liquid helium II and
cryogenic (critical) helium gas are attractive candidates for investigating classical
turbulence problems, and in the process many have decided to look at the kinds
of challenges encountered in using helium II, that phase of liquid helium which
exhibits superfluidity.
In preparing this talk I had hoped to cover, however briefly, the entire corpus
of experimental work on the subject. It soon became evident that there was
too much material by far than could be covered in a single one-hour lecture.
Fortunately, I had recently written a review article in honor of Joe Vinen’s
retirement from the University of Birmingham [1] and Skrbek, Niemela and
myself had written a second paper for the same occasion [2]. These papers contain
virtually all the known results on cryogenic fluid mechanics. This talk will instead
concentrate mostly on future directions, and results obtained since the articles
cited above were written.

2 Update on Pipe Flow

Cryogenic helium is of significant value in generating, and studying the highest
possible Reynolds and Rayleigh number flows under controlled laboratory condi-
tions, primarily due to its extremely low value of kinematic viscosity. Pipe flow
is an example where the highest Reynolds numbers can be achieved by using
helium I. Our apparatus is shown in Fig.1. The pipe is a polished stainless steel
tube 4.67 mm in diameter and 25 cm long. The working fluid is drawn into the
bellows by moving the push rod at the top upward. At a preset time the rod
is pushed down at constant velocity and discharges a known mass of liquid per

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 17–35, 2001.

c Springer-Verlag Berlin Heidelberg 2001
18 R.J. Donnelly

Push Rod

Bellows 10 cm
Pump LHe

~
~
P
{

Smooth 25 cm
Pipe

Fig. 1. Bellows driven flow apparatus for measuring the friction factor in a smooth pipe.
The advantage of cryogenics can be appreciated by comparing the size of this apparatus
with the famous apparatus of Nikuradze built in the early 1930’s and weighing many
tons. The highest Reynolds numbers achieved are about the same in both Nikuradze’s
and our apparatus.

unit time. Flow velocity is limited by the power of the drive motor. A sensitive
capacitance manometer has been developed to measure the pressure gradient
in all working fluids and is located 38 diameters down stream of the inlet. The
in-situ differential pressure gauge was specifically designed to avoid mechanical
strains at low temperature, which can overwhelm the 0.05 Pascal resolution of
the device [3].
This experiment, primarily the effort of Chris Swanson of the University of
Oregon, with help from Gary Ihas, visiting from the University of Florida, has
revealed several subtleties in such measurements, including the need for high
purity of helium, fine stability of temperature and pressure, and care to maintain
a time-independent state. Agreement with data at room temperature using air
and water as a fluid is quite good.
The primary advantage of this apparatus is that it accommodates any non-
reactive fluid at a temperature between 1 K and 300 K, allowing unprecedented
ranges of Reynolds numbers to be obtained. For example, we have used room
temperature SF6 , air, and helium gas, as well as cryogenic helium I at 4.2 K
to measure the friction factor from Reynolds number 10 to 5,000,000, nearly
six decades of Reynolds numbers. The fact that the results agree with standard
engineering data from many sources attests to the utility of this new approach.
 An Introduction to Experiments on Superfluid Turbulence 19

Linear Servo Motor

Motor Control
and Counter

16 bit ADC

200MHz Pentium
GPIB, Labview
G
P
I
B
Function Vacuum Tight Sliding Seal
B Generator
U Vacuum
Pump
S

Lock-in
Amplifier

Amplifier
60 mF Grid Generated
Turbulence
Grid

Second Sound
Bias Voltage Receiver
100 Volts DC
Second Sound
Transmitter

Second Sound
LR-110 Standing Wave
Resistance
Bridge
Germanium
Thermometer 1 cm x 1 cm x 29 cm
Channel
Heater
LR-130
Temperature
Controller

Fig. 2. Schematic diagram of the University of Oregon towed grid apparatus in helium
II. Turbulence is generated by sweeping the grid upward, and is observed by measur-
ing the attenuation of second sound in the presence of the turbulence which can be
interpreted to give the vorticity in the superfluid.

It would be of interest to extend these results to helium II, where other

evidence suggests the results will be about the same [4]. However this is yet to
be accomplished in a satisfactory way.
20 R.J. Donnelly

3 Update on Towed Grid Experiments

Superfluid turbulence in counterflow turbulence takes the form of a tangle of
quantized vortex lines, a frictional drag between the moving normal fluid and
the cores of the lines (mutual friction) serving to maintain the turbulence. Coun-
terflow turbulence has no classical analogue. Other types of superfluid turbulence
have been studied over the years, but only recently has there been an experimen-
tal study of superfluid turbulence behind a steadily moving grid. This turbulence
is analogous to the much studied, and fundamentally important, case of homo-
geneous isotropic turbulence created by flow through a grid in a wind tunnel.
The towed grid apparatus, shown in Fig.2, consists primarily of a 1 cm x
1 cm square tube fitted with a sliding grid with about 65 % porosity. The grid
is swept upward at a preset time and creates a slug of turbulence of quite high
vorticity filling the channel. The probe is a pair of second sound transducers
which measure the attenuation of second sound owing to quantized vorticity in
the channel.
The results obtained with this apparatus have been fascinating and a series
of publications have come from this apparatus which show, among other things,
that the decay of the turbulence can most easily be understood treating the decay
as if it were from a classical fluid having the viscosity of the normal fluid, but the
density of the total fluid [5,6,7,8]. The physics underlying this phenomenon has
been carefully discussed by Vinen [9]. His point of view is also further developed
in a lecture at this conference [10].
Two new experiments have been performed which had not appeared in print
at the time of this conference. A brief description of both follows.

3.1 The Nature of Grid Turbulence in Helium II

A detailed experimental and theoretical study of superfluid grid turbulence was
carried out by Stalp et al [8]. As an observational technique they used the at-
tenuation of second sound, which serves to measure the length of vortex line
per unit volume, L, in the superfluid component. They showed that a classical
model could explain the observed time decay of line length. According to this
model superfluid helium behaves as a single fluid with a kinematic viscosity ν  .
The rate of dissipation of turbulent energy per unit mass has the classical form

ε = ν  ω 2  = ν  κ2 L2 (1)

where ω 2  is the mean square vorticity in the coupled fluids, and the identifi-
cation of ω 2  with κ2 L2 is discussed in [9]. Therefore according to the model
second sound measures ω 2 , as a function of time. After an initial time interval
the turbulent energy spectrum has the classical Kolmogorov form

E(k) = C2/3 k −5/3 (2)

within a range of wave numbers ke < k < kη . Energy is being continually

transferred in a cascade from lower to higher wave numbers by inertial forces,
 An Introduction to Experiments on Superfluid Turbulence 21

dissipation by viscosity taking place eventually at k ≥ kη = (ε/ν  )1/4 (the

3

Kolmogorov wave number). C is the Kolmogorov constant and is equal to about

1.5 for a classical fluid, and ke is the wave number of the energy-containing
eddies. ke (t) decreases until it saturates at a value equal to approximately the
inverse of the channel width.
The model made predictions that were in agreement with the experiments
reported in [5,6,7] in the temperature range studied, from the λ-temperature
down to about 1.3K; however, although ν  was shown to be of order μ/ρ, where
μ is the viscosity of the normal fluid and ρ is the total density of the helium, no
careful analysis of the temperature dependence of ν  was carried out.
It turns out that within the inertial range of wave numbers (ke < k < kη )the
relevant length scales are significantly larger than the expected spacing of quan-
tized vortex lines, which turns out to be close to kη −1 . Mutual friction is then
sufficient to keep the two fluids locked together on these length scales, the locked
fluids behaving as a single turbulent fluid in which there is negligible dissipa-
tion. Therefore, a Kolmogorov spectrum forms, as in a classical fluid, provided
of course that energy is dissipated by some means at or beyond the Kolmogorov
wave number kη . The model requires that this dissipation be described by equa-
tion (1). The fact that the vortex line spacing is of order kη −1 implies that in
reality the two fluids cannot remain locked together for wave numbers of order
or greater than kη . It follows that dissipation must be due to a combination of
viscous dissipation in the normal fluid and mutual friction, and it is therefore
not obvious that an equation of the form (1) should hold [9]. It was predicted
in [9] that the resulting rate of dissipation would be given by a formula similar
to (1), with a value of ν  that is weakly dependent on temperature and of order
μ/ρ as measured above 1 K.
We conclude that energy dissipation in grid turbulence is likely to be de-
scribed correctly by (1), in spite of the fact that the mechanism of energy dissi-
pation in the superfluid is quite different from that in a classical fluid. The precise
value of ν  depends on the dissipative mechanisms operating in the turbulence.
These mechanisms operate on a length scale of order the vortex line spacing,
and are therefore strongly influenced by quantum effects in the superfluid.
We see that the parameter ν  , which is analogous to the kinematic viscosity
in a classical fluid, is fundamentally important in superfluid turbulence. The
experimental technique was described in [5,6,7]. Data analysis is carried out for
that period of the decay during which the vortex line density falls as t−3/2 , when
the theory of the classical model gives
d
L(t) = √ (3C)3/2 t−3/2 (3)
2πκ ν 
where d is the width of the channel through which the grid is towed. At times
earlier than this period the scale of the energy-containing eddies has not yet sat-
urated at the channel width (see section 3.2 below) and the decay is slower.
At later times the inertial range of wave numbers becomes narrow and the
22 R.J. Donnelly

-4
2 .0 x 1 0

-4
1 .5 x 1 0
/s )
2
ν ' (c m

-4
1 .0 x 1 0

ν (T )

-5
5 .0 x 1 0
1 .0 0 1 .2 5 1 .5 0 1 .7 5 2 .0 0 2 .2 5

T e m p e ra tu re (K )

Fig. 3. The effective kinematic viscosity as a function of temperature, as measured

from experiment.

decay rate increases. The length of line is obtained from second sound resonance
measurements using the relation
16Δ0 A0
L(t) ≈ ( − 1) (4)
κB A(t)
where Δ0 is the full-width at half maximum for the second-sound (power) res-
onance curve in the absence of vortex lines (at the 50th harmonic), B is the
mutual friction constant obtained from measurements on uniformly rotating he-
lium [10], and A(t) and A0 are respectively the peak second sound amplitude
with and without vortex
 lines present. Using (4) and (5) we see that the ex-
periment measures C 3 /ν  . Note that Maurer and Tabeling [11] performed an
experiment with rotating disks in liquid helium in which they were able to mea-
sure the Kolmogorov spectrum by means of pressure fluctuations using a probe
inserted at a selected location in the flow and connected to a quartz pressure
transducer. Their results showed that the Kolmogorov spectrum could be ob-
served in both helium I and helium II (2.3K, 2.0K and 1.4K) and reached the
important conclusion that the Kolmogorov constant C is the same above and
below the lambda transition. With that information, we see that our experiment
determines ν  providing we know C. The Kolmogorov constant is taken to be
1.5 at all temperatures, which is the accepted classical value [12]. The resulting
values of ν  are shown as a function of temperature in Fig. 3. The error bars
 An Introduction to Experiments on Superfluid Turbulence 23

Fig. 4. The decaying helium II vorticity measured at T=1.3 K for the indicated ReM .

include contributions due to uncertainties in the values of B, and the frequency-

dependence of B was taken into account [13].
The experimental results reported here relate to temperatures greater than
1.1 K. There is a clear need to extend the temperature range downwards, es-
pecially because the theory then makes interesting predictions. The only exper-
iments so far reported on superfluid turbulence at these low temperatures are
those of Davis, Hendry and McClintock [14].

3.2 Four Regimes of Decaying Grid Turbulence in Helium II [15]

Attenuation of second sound in helium II has been used to observe up to six

orders of magnitude of decaying vorticity displaying four distinctly different
regimes of decaying grid turbulence in a finite channel. A purely classical spectral
model for homogeneous and isotropic turbulence describes most of the decay of
24 R.J. Donnelly

helium II vorticity in the temperature range 1.2K < T < 2K. The four regimes
switch successively as the energy containing and dissipative Kolmogorov length
scales gradually grow during the decay, finally both being saturated by the size of
the channel. In Fig.4 each curve represents an average of three individual decays.
As the decay curves tend to collapse on the universal curve, we shifted them for
clarity by a factor of two downwards, the uppermost remaining unchanged. The
early part of the vorticity decay displays a power law with exponent -11/10 (see
left inset, showing normalized data forReM = 104 ) and later −5/6 (see right
inset, showing normalized data forReM = 1.5 × 105 , 2.5 × 104 and 5 × 103 ). After
saturation, typically several orders of magnitude of decaying vorticity closely
follow the power law with exponent −3/2, represented by the thick solid line.
For the first time we report the fourth and last regime - a late exponential
decay - not shown separately in Fig.4. The nature of this last decay is not
understood at present.
This experiment explores over 8 orders of magnitude of decaying turbulent
energy, an impossible task for a wind tunnel, which would need to have a 1000
km test section to observe the same thing.

4 Agenda for the Future

4.1 The University of Oregon 6 cm Wind Tunnel

So far we have not had a continuous flow facility to work on. About two years ago
we decided to build a small wind tunnel using critical helium gas as the working
fluid. The advantage of critical helium gas is the enormous range of properties
which can be reached by adjusting the pressure and temperature. Flow velocities
available depend on the density. At low densities flow velocities up to 1m/s are
possible, at high densities velocities to 30cm/s are available. One of the optimal
operating points for high Re flow will be at 4 bar pressure and 6 K at which
the kinematic viscosity is 3.1 × 10−8 m2 /s. Mesh Reynolds numbers of 30,000 to
100,000 will be available, and corresponding microscale Reynolds numbers will
range from 150 to 280.
We have recently successfully operated this tunnel at 6.5 K and mesh Reynolds
numbers around 1500. The grid generated turbulence is probed with 10 micron
diameter cryogenic hot wire anemometer also developed in our laboratory. These
hot wires are observed to obey King’s law relating velocity and voltage, and pre-
liminary velocity time series show standard statistical features. Note that this is
the first cryogenic tunnel to operate below liquid nitrogen temperature.

4.2 Wind Tunnels for Model Testing

We have given considerable attention to the conceptual design of larger wind tun-
nels designed to be useful for model testing. These have been discussed recently
in [2] and do not need repeating in this article. The importance for superfluid
turbulence of these devices is the possibility of operating them in helium II.
 An Introduction to Experiments on Superfluid Turbulence 25

Fig. 5. Sketch of the University of Oregon 6 cm wind tunnel.

The acoustic characteristics of models being tested in such a tunnel is very

important, and unexplored at the present time. Acoustics is a very large field
having to do with the generation and interaction of sound with mechanical struc-
tures. It is of particular importance for submarine testing.

4.3 Tow Tanks

Liquid helium offers much promise for tow tank design, a subject also reported
in some detail in [2]. Again the importance for superfluid turbulence is the pos-
sibility of operating with helium II as the working fluid.
26 R.J. Donnelly

5 Challenges for the Future

5.1 The Challenge of Instrumentation

Although helium I is a Navier-Stokes fluid, the possibility of generating ever

higher Reynolds numbers carries with it ever decreasing Kolmogorov lengths. Of
primary concern is the smallness of the flow structures such as turbulent eddies
and boundary layers. Consider flow through a smooth pipe at high Reynolds
number. To be definite let us take a pipe 10 cm in diameter using liquid helium
having a kinematic viscosity ν = 2×10−4 (cm)2 flowing at a mean velocity u , and
density ρ = 0.146gm/(cm)3 . For turbulent flow a viscous sublayer is generated
at the walls
 which scales with the friction velocity . The friction velocity is given
by u∗ = τ /ρ , where τ , the wall stress, is τ = λρu2 /8 and λ is the empirical
friction factor [16]. The corresponding length scale is
ν
y= (5)
u∗
and the viscous sublayer has a thickness of a few times y. This y can be thought
of as the scale of the smallest eddies in the turbulent flow. Table 1 outlines
some typical values for helium I. Note that in helium I we can in principle use
velocities of flow up to the velocity of sound before encountering shock waves.
We cannot use such large velocities in helium II because second sound shock
waves will intrude at about an order of magnitude lower velocity. The speed of
sound in the gas is lower than in the liquid, hence the highest Reynolds number
generated with helium should be generated with helium I.
It can be seen that the length scales can range from microns to Angstroms.
For comparison with theories it is important to be able to probe the flow at the
smallest scales. The last entry for y in table 1 is 42 Angstroms. This should be
compared with the cube root of the atomic volume of liquid helium which is 3.6
Angstroms. Kolmogorov lengths in the Angstrom ranges, therefore, can approach
conditions where the continuum hypothesis for the Navier Stokes equation is
going to be in question. Probing such flows will be a major instrumentation
challenge. We must learn to build not only micron sized, but even nanometer
sized transducers in order to take advantage of the new range of high Reynolds
numbers afforded by cryogenic helium.
Another experimental challenge is to visualize the normal fluid. This problem
has always been with research with helium II, and not a great deal of progress has
been made except in specialized cases [18]. Perhaps the most promising avenue
for visualization will be PIV (mentioned below).
We now summarize briefly the diagnostic tools available for cryogenic fluid
mechanics. These are discussed in a recent review [17].
• The measurement of average flow velocity can be measured in many instances
by means of acoustic flowmeters or venturi tubes.
• Flow velocity at a point can be measured in convecting helium gas, for
example, by means of bolometer fluctuation correlations.
 An Introduction to Experiments on Superfluid Turbulence 27

• Hot wires, Laser Doppler Velocimeters (LDV) and Particle Image Velocime-
ters (PIV) should all work in principle in helium gas, helium I and helium
II.
• RMS vorticity can be measured in helium II by second sound attention in
both open flows and counterflows. There is some speculation that chemical
potential probes would give local information on vorticity fluctuations, but
this has yet to be implemented [17].
• Temperature gradients can be measured by means of standard germanium
thermometry and pressure gradients, for example, by means of capacitance
manometers [3].
• Lift and drag on models can best be measured using Magnetic Suspension
and Balance Systems (MSBS), discussed in [18] by Britcher. Of course in a
cryogenic environment the magnets can be superconducting.
• Wall stress gages can be fabricated for work in helium gas and helium I.
• Ion trapping can be used to measure vorticity in helium II [25].

Table 1. Viscous Sub Layer Thickness at Various Reynolds Numbers

u Re λ τ u* y
(cm/sec) erg/cm3 cm/sec (cm)
10 5 × 105 0.0132 0.0240 0.405 4.93 × 10−4
102 5 × 106 0.00898 1.64 3.35 5.97 × 10−5
103 5 × 107 0.00649 118.4 28.48 7.02 × 10−6
104 5 × 108 0.00489 8,940 247 8.08 × 10−7
2 × 104 1 × 109 0.00453 33,100 476 4.20 × 10−7

Hot wire anemometers are resistive self-heating devices which balance heat lost
to flow, which depends on the fluid velocity. Calibration provides the correlation
between fluid velocity and electrical power supplied. Standard hot wires have
d ∼ 5 microns and L ∼ 1000 microns. Standard materials (e.g. platinum) used at
room temperature are insensitive at low temperature. Our cryogenic hot wires are
made on a quartz fiber of diameter 10 microns. They consist of an evaporated Au-
Ge film of thickness of order 3000Å. A small sensitive region in middle of the fiber
is defined by masking the fiber and evaporating a metallic film over it as shown
in Fig.7. Masking is achieved by laying a small diameter fiber perpendicular to
the first fiber. The metal film provides electrical contact to the sensitive region.
Figure 8 shows how the small sensors are mounted. Sensor support dimen-
sions follow the “rule of 10”: they are placed ten times (more or less) their
characteristic dimension away from the sensitive region. Electrical contact to
the fiber is made through stainless steel wires, which are isolated from the brass
ring by epoxy. The fiber is epoxied to the wires using an electrically conducting
epoxy.
We have recently tested a few prototype 10 micron sensors in the wind tunnel
of Fig.5. A helium gas flow at 7 K was generated and we applied a relatively large
sinusoidal voltage to overheat one of the sensors and measured the temperature
28 R.J. Donnelly

Fig. 6. A hot wire anemometer.

Fig. 7. Principle of the cryogenic hot wire anemometer developed at the University of
Oregon.

of a nearby sensor. The characteristic heat signature at twice the input frequency
was clearly seen on the monitoring sensor. Furthermore, we found that the signal
to noise ratio to be the same as that predicted by our noise analysis. Thus we
feel confident that the sensors are working according to our expectations.
 An Introduction to Experiments on Superfluid Turbulence 29

Fig. 8. Mounting the 10 micron hot wire anemometer

We have made substantial strides in detector development. Our 10 micron

detectors are to be contrasted with the bolometers we use in our large convection
apparatus, which are 250 microns on side. But as we see from Table 1, we need
to push our detectors down in size by orders of magnitude. Such advances are
likely to depend on optical techniques.

5.2 Challenges for Understanding Counterflow Turbulence

Counterflow experiments have a history of many decades, but much remains to

be accomplished [1]. The brief outline below suggests some directions needing
exploration.

Comparing the decay of towed grid turbulence and counterflow tur-

bulence. We need to understand the differences between the decay of towed
grid turbulence and the decay of counterflow turbulence after the counterflow
is turned off. Fig.9 below shows decay of both types of turbulence in the same
apparatus, beginning at about the same vorticity and demonstrating nearly iden-
tical decay at long times. The experimental channel was identical in both cases
except the grid was removed for the counterflow experiment. The form of the
upper curve has been discussed by Schwarz and Rozen [19] and by Smith [5].
The decay curves coincide at long times and correspond exactly to the decay of
classical turbulence (see Section 2.2 above).

The first critical velocity. What accounts for the first critical velocity in
counterflow and its temperature dependence? This phenomenon is likely a case
of extrinsic nucleation, but lacks any quantitative explanation.
30 R.J. Donnelly

Fig. 9. An illustration of the dramatic differences in decay of two turbulent flows of the
same initial vorticity produced by a counterflow (upper curve) and towed grid (lower
curve).

The TI-TII transition. The TI-TII transition needs further attention. We

have Melotte and Barenghi’s insight which suggests that TI corresponds to tur-
bulence in the superfluid, and TII marks the onset of disturbances in the normal
fluid. Their linear stability analysis gives the right order of magnitude for the
transition, but does not capture the temperature dependence.

The Tough classification. Why do large and small aspect geometries in coun-
terflow exhibit such different behavior?

Combined rotation and heat flow. Combined rotation and heat flow is a
relatively new area of investigation. Prior to the investigation discussed in this
 An Introduction to Experiments on Superfluid Turbulence 31

section it was assumed (from earlier experiments) that the ordered array of vor-
tex lines produced by steady rotation and the disordered tangle produced by
counterflow preserved their identities in a combined experiment. Measurements
at Oregon with improved sensitivity by Barenghi, Swanson and Donnelly [20]
showed that the picture just described is far from true. The observations con-
sisted of measuring the amount of vortex line present owing to counterflow or
rotation alone using second sound attenuation, and comparing the observed line
density with what would be expected if the two sources of vorticity simply added.
The results are complicated, but appear to be relatively simple in two limits:
(i) Limit of large line densityLH due to heat, slow rotation. Here the effect of
rotation is not simply to add line densityLR = 2Ω/κ. Instead the tangle appears
to be polarized to accomplish the rotation. The effective polarization increases
with rotation Ω by analogy to a gas of magnetic dipoles in a magnetic field.
The results scale with LR /LH by analogy to μH/kT . Thus rotation appears to
produce alignment in the tangle, as does a magnetic field for dipoles, and LH
appears to play the role of disordering heat bath in the statistical mechanics of
superfluid turbulence. Indeed, it takes very little polarization of a dense tangle
to produce rotation at the relatively small angular velocities of the apparatus.
(ii) Limit of fast rotation and small axial heat flux. Any rotation eliminates
the critical velocity vc . In this limit two critical counterflow velocities appear,vc1
1
and vc2 , which scale as Ω 2 .
We might speculate here that the first critical velocity appears to correspond
to the Donnelly-Glaberson instability, excitation of helical waves by the coun-
terflow on the vortex lines induced by rotation [1]. The second appears to be
a transition to turbulence, with the rotation-induced array becoming a vortex
tangle. A more formal investigation of these effects is likely to be rewarding.

5.3 Challenges for Understanding Periodic

Boundary Layer Experiments
The decay of torsional oscillations of various pendulums was a veritable ”cottage
industry” in low temperature physics in the 1950’s. The overall situation was
reviewed by Donnelly and Hollis Hallett in 1958 [21]. The situation can be ap-
preciated by looking at the damping of oscillations of liquid helium in a U-Tube,
as reported by Donnelly and Penrose [22] and reviewed in [21].
Figure 10 summarizes the data. In helium I the damping is independent
of amplitude up to a critical amplitude (or velocity) called ht . In helium II,
however, the results show damping accounted for by the Landau two fluid model
below a temperature-dependent critical amplitude called hc , a steady increase
in damping up to an amplitude hn , and a further critical amplitude, also named
ht . The region of damping between hn and ht seems to be accounted for by the
Navier-Stokes equation. The authors assumed that the region between hn and ht
was characterized by an increasing coupling between the two fluids. The details
of these observations are unclear even today.
What causes the two fluid model to break down at a critical amplitude hc ?
What happens at ht ? What would happen if experiments (including the pendu-
32 R.J. Donnelly

Fig. 10. The variation with amplitude of the damping of gravity oscillations of liquid
helium in a U-tube at a period of 0.94 sec. After Donnelly and Penrose [22] and Donnelly
and Hollis Hallett [21].

lums) were ramped up in amplitude instead of starting suddenly at high ampli-

tude and allowed to decay. Oscillating cylinders and spheres are likely best for
such novel experiments.
 An Introduction to Experiments on Superfluid Turbulence 33

Fig. 11. Apparatus constructed by Davis, Hendry and McClintock [14] to measure the
decay of superfluid turbulence at 70 mK.

5.4 Instrumentation to Detect Vortices Below 1 K

Apparatus constructed by Davis, Hendry and McClintock [14] is the only one
to address the question of the nature of superfluid turbulence in the absence
of normal fluid experimentally. They arranged to generate turbulence by means
of an oscillating grid structure, and to examine it by trapping of negative ions
coming from a field emission tip on the vortex tangle. Their results indicate a
temperature independent decay process. The absolute line density of the vortices
remains unknown.
Second sound will not work when the temperature is too low because the
normal fluid has disappeared. Ion capture may be difficult because of lack of
knowledge of the dynamics of capture at very low temperatures, and experimen-
tal evidence that the capture cross section is decreasing at lower temperatures
[23]. Samuels and Barenghi [24] have suggested that there is enough heat energy
in a vortex tangle to be measured calorimetrically when the tangle has decayed.
Another possibility might be to design some sort of phonon (or roton)detector
fast enough to follow the decay of turbulence from a towed or oscillating grid.

5.5 The Normal Fluid and the Vortex Tangle

Understanding the relationship between the normal fluid and the vortex tangle
was an important topic of discussion at this conference. Theoretical progress on
34 R.J. Donnelly

the self-consistent interaction between the normal fluid and quantized vortices
will likely lead to the need to develop measurement techniques which will give
more information than just the RMS line density L.

5.6 Flow over Blunt Objects, Testing Models such as Submarines

Turbulence can easily be generated by high Reynolds number flows over sim-
ple objects like cylinders and spheres. Here the flows are shear flows and are not
homogeneous and isotropic. Indeed the whole subject of the flow over experimen-
tal models is mostly in the future. However Van Sciver has reported preliminary
measurements of flow over a sphere in a companion article in this volume [24].
We look forward to more complete measurements and interpretation as this work
progresses.

Acknowledgements
First of all I am grateful to Keith Moffatt, Director of the Newton Institute, for
the opportunity to have this remarkable gathering here. I am indebted to Joe Vi-
nen, Steve Stalp, Ladislav Skrbek, Carlo Barenghi, David Samuels, Renzo Ricca
and Peter McClintock for many useful discussions. My research is supported by
the National Science Foundation under grant DMR-9529609.

References
1. R. J. Donnelly, J. Phys Condensed Matter, 11, 7783-7834 (1999).
2. L. Skrbek, J. J. Niemela and R. J. Donnelly, J. Phys Condensed Matter, 11, 7761-
7782 (1999).
3. Chris J. Swanson, Kris Johnson and Russell J. Donnelly, Cryogenics 38, 673-677
(1998).
4. P. L. Walstrom, J. G. Weisend II, J. R. Maddocks and S. W. Van Sciver, Cryogenics
28, 101 (1988).
5. M.R. Smith, R. J. Donnelly, N. Goldenfeld and W.F. Vinen, Physical Review
Letters 71, 2583 (1993).
6. M.R. Smith “Evolution and Propagation of Turbulence in Helium II”. PhD Dis-
sertation, University of Oregon (1992).
7. S. Stalp, “Decay of Grid Turbulence in Superfluid Helium”. PhD Dissertation,
Physics, University of Oregon (1998).
8. S.R. Stalp, L. Skrbek, R.J. Donnelly, Phys. Rev. Lett., 82, 4831 (1999)
9. W. F. Vinen, Phys. Rev. B61, 1410 (2000)
10. W. F. Vinen “An Introduction to the Theory of Superfluid Turbulence” Paper in
this volume
11. J. Maurer and P. Tabeling, Europhys. Lett., 43, 29 (1998).
12. K.R. Sreenivasan, Phys. Fluids, 7, 2778 (1995).
13. C.F. Barenghi, R.J. Donnelly, W.F. Vinen, J. Low Temp. Phys., 52, 189 (1983).
14. S. I. Davis, P. C. Hendry and P. V. E. McClintock, Physica B , 280, 43 (2000).
15. L. Skrbek, J. J. Niemela and R. J. Donnelly, Phys. Rev. Lett., 85, 2973 (2000).
16. H. Schlichting, Boundary Layer Theory, 7th ed, McGraw-Hill (1979).
 An Introduction to Experiments on Superfluid Turbulence 35

17. C. F. Barenghi C. E., Swanson, and R. J. Donnelly, J. Low Temp. Physics, 100,
385 (1995).
18. R. J. Donnelly (Editor), Liquid and Gaseous Helium as Test FluidsSpringer-
Verlag, 1991). In High Reynolds Number Flows Using Liquid and Gaseous He-
lium(Springer-Verlag, 1991). Also R. J. Donnelly and K. R. Sreenivasan (Editors)
Flow at Ultra-High Reynolds and Rayleigh Numbers(Springer-Verlag, 1998)
19. K. W. Schwarz, and J. R. Rozen, Phys. Rev. B , 44, 7563 (1991).
20. C. F. Barenghi C. E., Swanson, and R. J. Donnelly, Phys. Rev. Lett. , 50, 190
(1983).
21. R. J. Donnelly and A. C. Hollis Hallett, Annals of Physics, 3,320 (1958).
22. R. J. Donnelly and O. Penrose, Phys. Rev. , 103, 1137 (1956).
23. R. M. Ostermeier and W. I. Glaberson, Phys. Lett. 49A, 223 (1974).
24. D. Samuels and C. Barenghi, Phys. Rev. Lett. 81, 4381,1998.
25. R. J. Donnelly,Quantized Vortices in Helium II,(Cambridge University Press,
1991).
26. S. W. Van Sciver, “The Drag Crisis on a Sphere in Flowing HeII” Paper in this
volume.
The Experimental Evidence for Vortex
Nucleation in 4He

Éric Varoquaux1 , Olivier Avenel2 , Yury Mukharsky2 , and Pertti Hakonen3

1
CNRS–Université Paris-Sud, Laboratoire de Physique des Solides,
Bâtiment 510, F-91405 Orsay Cedex, France
2
Commissariat à l’Énergie Atomique, Service de Physique de l’État Condensé,
Bâtiment 772, Centre de Saclay, F-91191 Gif-sur-Yvette Cedex, France
3
Low Temperature Laboratory, Helsinki University of Technology,
02150 Espoo, Finland

Abstract. This update on the problem of vortex nucleation in superflows through

micro-apertures follows the recent reviews of the subject by Zimmermann [1] and one
of the authors [2]. Recent developments of the model of vortex nucleation involving a
vortex half-ring are assessed with an emphasis on the statistical properties of the critical
velocity transition. The occurrence of collapses and multiple slips is then discussed in
relation with the problem of vortex nucleation.

1 Single Vortex Nucleation

Thermally assisted vortex nucleation. When superflow is forced through
a micro-aperture and the critical velocity is reached, energy dissipation occurs
by phase slippage. Phase slips by 2π take place according to a scenario in which
1
/2 –rings nucleate at the wall of the aperture, at a site where the critical velocity
vc is exceeded [3,4,5,6]. A remarkable feature of the critical velocity for phase
slips is its dependence on temperature, which is very nearly a straight line from
about 2 K down to ∼ 0.15 K:
vc = v0 (1 − T /T0 ) . (1)
The value of T0 varies somewhat with the micro-aperture but lies in the vicinity
of 2.5 K. A sample of the experimental data obtained in various laboratories is
shown in Fig. 1.
Since superfluid 4 He is nearly fully in its ground state below 1 K, such a
large temperature dependence gives a strong indication that an Arrhenius-type
process must be taking place. The end products of this nucleation process are
vortices because phase slips by 2π are observed [7]. The point of view taken here
is that these vortices are nucleated (i.e., created ex nihilo). However, it has been
suggested that solitons [8] or bubbles [9] can appear as intermediate states in the
nucleation process. Other mechanisms based on pure fluid-mechanical motion
such as vortex mills [10,11], which will be discussed in Sect.2, are difficult to
reconcile with the large temperature dependence of vc below 1 K.
Thermal nucleation implies the existence of an energy barrier to the spon-
taneous formation of vortices which is overcome with the help of thermal fluc-
tuations. As it can be surmised that a critical velocity would be reached even

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 36–50, 2001.

c Springer-Verlag Berlin Heidelberg 2001
 The Experimental Evidence for Vortex Nucleation 37

Fig. 1. Critical velocity, normalised to the zero temperature linear extrapolation value
v0 , versus T , in Kelvin: () [5], () [1], for ultra-pure 4 He. The curves are computed
from the half-ring model for a0 = 2.2, 3.2, 4.5, 6.0 Å and are normalised to match the
experimental value at 0.5 K. The inset shows the influence of 3 He impurities on vc , (◦)
3 ppb, () 45 ppb.

in the absence of thermal (or quantum, as discussed below) fluctuations, this

energy barrier vanishes for a finite value vc0 of the critical velocity. The hydro-
dynamic instability threshold vc0 at which vortices appear spontaneously has
been shown to occur in numerical simulations of flows past an obstacle using the
Gross-Pitaevskii equation by Frisch et al. first [12] and by others [13,14]. Close
to the point where it vanishes, i.e. for v vc0 , a well-behaved energy barrier
takes the form [5]:
 3
2 v2 2
Ea = E J 1 − 2 , (2)
3 vc0
where EJ depends on temperature and pressure.
The nucleation rate for thermally activated process is given in terms of the
activation energy by Arrhenius’ law:

ω0 Ea
Γ = exp − , (3)
2π kB T

ω0 /2π being the attempt frequency. In experiments performed in a Helmholtz

resonator, such as those shown in Fig. 1, the velocity varies periodically at the
resonance frequency as vp cos(ωt) and the probability that a phase slip takes
38 É. Varoquaux et al.

place during the half-period ωti = −π/2 , ωtf = π/2 is

  tf
p = 1 − exp − Γ (P, T, vp cos(ωt ) dt
ti

ω0 −2πkB T Ea
= 1 − exp − exp − . (4)
2πω vp ∂Ea / ∂v|t=0 kB T

Eq.(4) stems from an asymptotic evaluation of the integral at the saddle point
t = 0. The accuracy of the asymptotic evaluation (4) becomes questionable as
T → 0 where the energy barrier vanishes. It has been checked by direct numerical
integration for typical cases and found to be quite satisfactory.
The critical velocity vc is defined as the velocity for which p = 1/2. This
definition is independent of the experimental setup, except for the occurrence in
(4) of the natural frequency of the Helmholtz resonator ω. The implicit equation
for vc reads:

ω0 −2πkB T Ea (P, T, vc )
exp − = ln 2 . (5)
2πω vc ∂Ea / ∂v|vc kB T

If Ea is given by (2), (5) can be solved analytically [5]:

  23 1
2
3 kB T
vc = vc0 1− γ , (6)
2 EJ

where, with logarithmic accuracy, γ ln(0.1ω0 /ω). It is immediately apparent

that the empirical linear dependence of vc on T , (1), is not well reproduced by
(6); this dependence could be satisfied by the following functional form for Ea ,

Ea = Ee (1 − v/vc0 ) , (7)

which is quite different from (2).

Thus, the temperature dependence of vc , (1), strongly suggests the existence
of a thermally activated nucleation process, but a well-behaved energy barrier
such as (2) leads to too fast an increase of vc as T → 0.

Statistical width of the critical transition. The velocity vc of each critical

event is a stochastic quantity. Its statistical spread can be characterised by the
‘width’ of the probability distribution defined [5,15] as the inverse of the slope of
−1
the distribution at vc , (∂p/∂v|vc ) . This critical width is found to be expressed
by:
    −1
2 1 1 ∂ 2 Ea  ∂Ea  1 ∂Ea 
Δvc = − + + . (8)
ln 2 2 vc ∂v 2 vc ∂v vc kB T ∂v vc
 The Experimental Evidence for Vortex Nucleation 39

Fig. 2. Vortex nucleation rate and phase slip probability in terms of the mean velocity
in the micro-aperture expressed as phase winding numbers [16] in 4 He at 12 mK, 0.6
bar, with 100 ppb of 3 He impurities: frame (a) top, time spent, in seconds (, left
scale), and number of phase slips (◦, right scale) per velocity bin of size 0.1; frame
(a) bottom, slip rate in s−1 as a function of flow velocity, in units of 2π; frame (b),
cumulative probability obtained from the histogram of the slip velocities. The plain
curve is a least square fit to the data with the functional form corresponding to (4).
The value p = 1/2, shown by the arrow, defines the critical velocity (here, vc = 59.68).

At low temperatures and large critical velocities, the quantity in curly brack-
ets in the right hand side of (8) is small with respect to the last term so that the
 −1
width is simply expressed as Δvc = −(2/ln 2) kB T ∂Ea /∂v|vc . Thus, the
statistical width is an approximate measure of the inverse of the slope of Ea in
terms of v.
This quantity is derived from p, itself obtained by integrating the histogram
of the number of nucleation events ordered in velocity bins. This procedure is
illustrated in Fig. 2: p shows an asymmetric-S shape characteristic of the double
exponential dependence of p on v, (4), a consequence of Arrhenius’ law, (3),
being plugged into a Poisson probability distribution. The observation of this
asymmetric-S probability distribution constitutes another experimental clue for
the existence of a nucleation process.
The critical transition of vc displays a measurable width, shown in Fig. 3,
which implies (as does the T -dependence of vc ) that the energy barrier is neither
very high compared to kB T nor very steep in terms of v. Fig. 3 contains data from
several groups. These data are somewhat scattered, especially above 0.5 K, but
they do show overall agreement within this scatter. They provide experimental
input on ∂Ea /∂v|vc and put tight limits on the theoretical models discussed
below.
It is also possible to measure the nucleation rate Γ directly as a function
of v. This quantity is the ratio, for a given velocity bin, of the number of slips
which have occurred at that velocity to the total time spent by the system at
the same velocity. The procedure is illustrated in Fig. 2. The slope of Γ (v) yields
∂Ea /∂v|vc ; the value of Γ at vc gives a combination of ω0 and Ea (vc ). Non-linear
40 É. Varoquaux et al.

Fig. 3. Statistical width of the critical velocity transition, normalised to the linear
extrapolation limit at T = 0, v0 , in terms of temperature: () [5], () [1], (×) [17] .

fits of p to (4) give estimates of ω0 , EJ and vc /vc0 which are independent of the
specific features of a given model. These estimates compare well with those of
the more precise analysis described below.

Quantum tunnelling. Below 0.15 K, vc ceases abruptly to vary with T , as

seen in Fig. 1. For ultra-pure 4 He (less than 1 part in 109 of 3 He impurities),
vc (T ) remains flat down to the lowest temperatures (∼ 10 mK). The crossover
from one régime to the other is very sharp. At the same crossover temperature
Tq , Δvc also levels off sharply. It is believed on experimental grounds that this
saturation is intrinsic and is not due to parasitic interferences [5].
If the nucleation barrier were undergoing an abrupt change at Tq , for instance
because of a bifurcation toward a vortex instability of a different nature [8], in all
likelihood Δvc would jump to a different value characteristic of the new process
(presumably small since vc reaches a plateau). Such a jump is not observed
in Fig. 3. Furthermore, vc levels off below Tq , which would imply through (5)
that Ea becomes a very steep function of v, but Δvc also levels off, which,
through (8), would imply the contrary. We are led to conclude [18] that, below
Tq , thermally-assisted escape over the barrier gives way to quantum tunnelling
under the barrier. In which case the attempt frequency is related to the crossover
temperature by
ω0 = 2πkB Tq . (9)
Relation (9) holds for barriers of the form (2) even in the presence of moderate
damping [19].
 The Experimental Evidence for Vortex Nucleation 41

In the quantum régime, the tunnelling rate for a cubic-plus-parabolic po-

tential well [20] can be used in place of (3) to compute vc and Δvc . The cor-
responding formulae are obtained by the following substitution in (3), (5) and
(8):
 1
ω0 ω0 Ea 2 Ea 36 Ea
a) =⇒ 864π , b) =⇒ .
2π 2π ω0 kB T 5 ω0
This procedure yields the smooth, if rapid, transition in vc and Δvc observed
at Tq . The value of ω0 given by (9) (ω0 /2π = 2 × 1010 Hz for Tq =0.147 K) is
consistent with the attempt frequency appropriate to the thermally-activated
régime [21] and that found directly from the fits to the probability p as shown
in Fig. 2. Furthermore, it agrees (for a0 = 5 Å) with the eigenfrequency of the
highest Kelvin mode that a vortex filament in 4 He can sustain, ω + = κ/πa20 =
ω0 where κ is the quantum of circulation. Thus, assuming vortex nucleation
by quantum tunnelling below Tq is fully consistent with the thermally-assisted
nucleation régime that prevails above Tq .

The 1/2 –ring model. Described in full in [5], this model has a long history,
following the work of Volovik [22], itself based on the theory of homogeneous
nucleation of vortices by Iordanski, and by Langer and Fischer [23]. It has been
very successful in accounting for vortex nucleation by ions [24]. In essence, the
model is based on the hydrodynamics of an Eulerian fluid assumed valid to scales
of the order of the vortex core diameter a0 , that is down to atomic size in 4 He.
Nucleation is assumed to occur on an asperity on the walls of the micro-aperture
where i) the flow velocity is largest, ii) the superfluid density is depressed. The
perturbed volume over which nucleation takes place is, on heuristic grounds,
of the order of a30 . The radius of the nucleated half-ring when it escapes to
the bulk fluid turns out of the order of 15 Å. As a consequence, the size and
shape of the asperity can be neglected in a first approximation; the asperity
only serves the purpose of breaking translational invariance. The bases of the
1
/2 –ring model have been re-examined critically by Sonin [25] and by Fischer
[26]. It can also be mentioned at this point that Nore et al. [27] have shown by
numerical simulations of 3D flows past an obstacle in the Gross-Pitaevskii model
that vortex filaments do evolve spontaneously into one, or possibly a few, half-
rings which thus constitute the preferred configuration of small vortices close to
walls.
The 1/2 –ring model is amenable to an expansion of the form (2) but the full
expression of Ea can straightforwardly be evaluated numerically and substituted
into the expressions of vc and Δvc , (5) and (8) [5]. The energy barrier impeding
vortex nucleation is found to vanish at
κ
vc0 = 0.432 . (10)
2πa0
This value, obtained for an Eulerian fluid, is to be compared with that obtained
by Rica for the Gross–Pitaevskii model, vc0 = 0.262κ/2πa0 [28], comparison
which points up quantitative inadequacies in the 1/2 –ring model outlined above.
42 É. Varoquaux et al.

Comparison with experiment and discussion. The outcome of the numer-

ical evaluation of vc (T ) and Δvc (T ) is plotted in Figs. 1 and 3 for various values
of the core parameter a0 which is the only free parameter in the problem. The
temperature dependence of a0 and ρs is taken from refs.[29] and [30] respectively.
In the bulk, the low temperature value of a0 1Å . At a wall boundary, it is
at least of the order of the static penetration depth (2.5 Å) [29], and probably
rather higher because of the large applied flow velocity [31].
The temperature dependence of the computed velocity shown in Fig. 1
matches that of the data above 0.5 K but markedly bends upward below that
temperature; the larger a0 , the more pronounced the bending. As noted before,
this behaviour comes from the form (2) of the energy barrier and is therefore
built into the model.
It is apparent in Fig. 3 that the width at high temperature is in reasonable
agreement with the smaller values of a0 (i.e. ∼ 3 Å), while the low temperature
end calls for larger values (∼ 4.5 Å). This trend is the same as for the critical
velocity in Fig. 1, and, as noted above, it can certainly be expected that a0
would increase close to the wall, hence with increasing velocities and decreasing
nucleated vortex size [3].
The critical velocities scale as the inverse of a0 (This scaling does not appear
in Fig. 1). An estimate of the absolute, local, value of vc , i.e., the value at the
nucleation site, has been made using 3 He impurities as local probes [32]. The
presence of a tiny concentration of 3 He quasiparticles strongly affects the critical
velocity, as shown in the insert of Fig. 1, because quasiparticles condense on the
vortex core and lower its energy. The local concentration of 3 He depends on
the local pressure, itself governed by the velocity according to the Bernoulli
equation. Detailed measurements of this effect have led to a value of ∼ 22 m/s
for the local nucleation velocity on the quantum plateau. This value is to be
compared with the mean flow velocity in the aperture which is measured in the
experiment and which is of the order of 5 m/s. Using the curves in Fig. 1 and
(10), the corresponding value of a0 is found to be ∼ 2.65 Å.
Thus the values of the vortex core parameter a0 which account for the various
experimentally measured quantities, the magnitude and the temperature depen-
dence of vc and of Δvc , agree with one another within a factor of about two
in spite of the blatant oversimplifications of the 1/2 –ring model. Hence, taking
account of the details of the nucleation site geometry and chemical environment,
of the exact nature of the boundary layer at the wall and of the properties of
the small nascent vortex cannot be expected to vastly improve the description
of vortex nucleation given above. On another hand, refinements are needed, es-
pecially for the conceptual grounds underlying the model since it is known that
very small vortices do not even exist [33].

Discussion of the 1/2 –ring model. There have been several attempts to
reformulate the 1/2 –ring model, both to put it on firmer theoretical grounds and
to possibly obtain a better description of the experimental results [1,3,25,26,34].
The 1/2 –ring model main inadequacies lie at the high velocity (small vortex)
 The Experimental Evidence for Vortex Nucleation 43

end. Burkhart et al. [3] have introduced corrections to the vortex energy and
momentum due to the proximity of a boundary as computed with the Gross-
Pitaevskii equation. These corrections go in the right direction but cannot be
considered as a full reformulation of the small-vortex-at-a-wall problem.
A reassessment of the model, due to Sonin [25] and aiming in particular at
a quasiclassical reformulation of vortex quantum tunnelling, yields the following
elegant analytical form for the energy barrier
  2
κ3 ρs κ
Ea = ln
32 π v 4π va0

in the thermally activated régime and a corresponding expression for quantum

tunnelling. This energy barrier does vanish for v ∗ = κ/4πa0 but quadratically in
1 − v/v ∗ . This functional dependence is further away from the empirical relation
(7) than the low temperature limit of the 1/2 –ring model (2). Consequently, and
as also shown by direct numerical computation, the curvature of vc (T ) at low T is
even more pronounced. An even more serious discrepancy is found for the width
Δvc : the computed value is much too large for any reasonable value of a0 . These
results may look surprising since the work of [25] starts from the same premises
as the conventional 1/2 –ring model. The difference has been checked numerically
to reflect the cost of the approximations made to obtain the analytical form of
Ea [35].
Sonin also discusses the influence of the geometry of the asperity but there
is no direct evidence from experiment that this geometry plays an important
rôle. On the contrary, the fact that the slope of vc (T ) with temperature, that is
T0 in (1), is found to be the same in quite different apertures shows that vor-
tex nucleation is not very sensitive to the details of the geometry. It is a little
more puzzling that the crossover temperature Tq from quantum tunnelling to
thermal activation has been found independent of cooldowns for two different
apertures at Saclay, although a somewhat different value of Tq has been reported
by the Berkeley group [36]. The overall agreement between various experiments
basically indicates that finer specific details are not very relevant and that the
simplifications of the 1/2 –ring model are reasonably well founded. It nonethe-
less remains that the nature and geometry of a typical nucleation site are quite
undetermined and that the enhancement factor between the mean aperture ve-
locity and the velocity at the nucleation site is not under control, as shown in
particular by Shifflett and Hess [37].
Theoretical fits comparable to those in Figs. 1 and 3 have been obtained by
Zimmermann et al. [1] with a variant of the 1/2 –ring model in which a0 is kept
fixed to the Hills-Roberts value but additional parameters are introduced to force
the vortex energy and momentum to go to zero for vanishing vortex sizes faster
than the classical vortex expressions. Also, the attempt frequency ω0 /2π and vc0
are treated as fitting parameters. These authors find that ln(ω0 /2π) should be
comprised between 15 and 25 (3.3 × 106 and 7.2 × 1010 Hz for ω0 /2π). The latter
value is compatible with vc0 = 26.5 m/s, which is in fair agreement with the
value of 22 m/s found in [32]. Thus, different approaches give converging results
44 É. Varoquaux et al.

and the 1/2 –ring model can be said to give a semi-quantitative description of
vortex nucleation when single phase slips are involved.

2 Multiple Slips and Collapses

Single phase slips are observed in experimental situations which may be loosely
characterised as ‘clean’, that is, for uncontaminated apertures of relatively small
sizes (a few micrometres at the most), with low background of mechanical and
acoustical interferences, etc ..., and with probing techniques which do not man-
handle the superfluid, namely, with low frequency Helmholtz resonators. When
these conditions are not met, flow dissipation occurs in a more or less erratic
manner by large events - giant slips or ‘collapses’ of the superflow. This last
situation is quite commonly met in practice, as discussed in [38].
Collapses constitute an apparent disruption of the vortex nucleation mech-
anism described in the previous section. Their properties have been studied in
phase slippage experiments and are reviewed in this section, together with possi-
ble mechanisms for their formation. It is likely that these events provide a bridge
between the ‘clean’ single phase slip case and the usual situation of critical ve-
locities which are temperature-independent below 1 K and which depend on the
channel size d according to vF ∼ (1/d) ln(d/a0 ) [39]. It is also possible that they
take part in the build-up of vortex tangles forming superfluid turbulence [40].
Neither problems are fully resolved at present.

The two types of large slips. Examples of multiple slips can be seen in Fig. 4
which shows the peak amplitude chart of a two-aperture resonator at 12.5 mK,
24 bars, in a 100 ppb 3 He in 4 He sample. The very large amplitude drop shown
in Fig. 4 and in the insert is rare (one in 104 to 105 slips) under the conditions of
this particular experiment. This type of events, called in [38] ‘singular’ collapses
and discussed further below, may occur at velocities much below the vortex
nucleation threshold (down to at least a third of vc ).
Besides the usual single slip pattern, there appears in Fig. 4 occasional double
slips (i.e. involving phase changes by 4π) and infrequent triple slips. Raising the
temperature to 80 mK, again for this particular cooldown, causes these multiple
slips to occur much more frequently and to involve more circulation quanta
on the mean. Lowering the pressure to 0 bar results in an almost complete
disappearance of multiple slips at all temperatures. These features are described
in detail in [38].
Some degree of understanding of the formation of multiple slips can be gained
by plotting the mean value of the phase slip sizes, expressed in number of quanta,
against the flow velocity at which the slips take place [41]. This flow velocity is
close to the critical velocity for single phase slips, i.e. the nucleation velocity; it is
varied by changing the temperature,
  the pressure, the resonator drive level. Such
a plot is shown in Fig. 5 for n+ , i.e. in flow direction conventionally chosen as
the + direction. Slips in the opposite (−) direction behave qualitatively in the
same manner but the phenomenon displays a clear quantitative asymmetry. As
 The Experimental Evidence for Vortex Nucleation 45

Fig. 4. Absolute peak amplitudes at successive half-cycles of the resonance, normalised

to the amplitude drop of a single slip versus time expressed in half-cycle number. The
half-period is 31.8 milliseconds. Individual measurements cannot all be resolved on
the scale of the figure. The top trace shows a succession of amplitude drops which
correspond, for the largest part, to succession of phase slips by 2π of opposite sign,
with occasional larger slips. When no slip occurs during the half-cycle, the resonance
amplitude grows under the action of the electrostatic drive continuously applied to the
resonator. The large feature around the 2000th half-cycle is a ‘singular’ collapse. The
insert shows the details of this particular collapse, • for positive peaks, ◦ for negative
peaks.

can be seen in Fig. 5, the mean slip size decreases on either side of the quantum
plateau, as does the nucleation velocity, but increases with pressure, contrarily
to the nucleation velocity which decreases with increasing pressure. It appears
clearly that the magnitude of the superflow velocity does not directly control,
by itself, the occurrence of multiple slips. This implies, as will be discussed fur-
ther below, that the phenomenon under study is not purely hydrodynamical in
the bulk of the fluid but involves some complex interplay with the boundaries.
As shown in Fig. 5, the velocity threshold for the appearance of multiple slips
depends
  on hydrostatic pressure; in fact, the P -dependence of the upturn of
n+ vs v exactly tracks that of the critical velocity for single phase slip nucle-
ation. This indicates that multiple slips appear because of an alteration of the
nucleation process itself.
The pattern of formation of multiple slips changes from cooldown of the cell
from room temperature to cooldown but remains stable for each given cooldown.
46 É. Varoquaux et al.

Fig. 5. Mean size of (positive) multiple slips vs velocity in phase winding number in
nominal purity 4 He (100 ppb 3 He): () pressure sweep from 0.4 to 24 bars at 81.5
mK (all even values of P , and 0.4, 1, 3, 5, 7 bars) - () temperature sweep at 16
bars - (◦) temperature sweep at 24 bars - (∗) drive level sweep at 24 bars, 81.5 mK
- () temperature sweep at 0 bar. For the temperature sweeps, from 14 to 200 mK
approximately, v first increases, reaches the quantum plateau and then decreases, as
shown in the insert of Fig. 1. Lines connect successive data points in the temperature
and pressure sweeps.

It seems to depend on the degree of contamination of the cell, degree which can-
not easily be controlled experimentally. The detailed microscopic configuration
of the aperture wall where nucleation takes place probably plays an major rôle
in multiple slip formation. Multiple slips are different from ‘singular’ collapses
and the underlying mechanisms responsible for both phenomena are bound to
be different as will be discussed below.

Remanent vorticity and vortex mills. Remanent vorticity in 4 He, which had
been long assumed, has been shown directly to exist by Awschalom and Schwarz
[42]. This trapped vorticity, according to Adams et al [43], either is quite loosely
bound to the substrate and disappears rapidly, or is strongly pinned. To account
for these observations, Schwarz has proposed the following formula, based on
numerical experiments, for the velocity at which vortices unpin,
κ b
vu  ln( ) ,
2πD a0
D being the size of the pinned vortex and b being a characteristic size of the
pinning asperity. Thus, vortices pinned on microscopic defects at the cell walls
can in principle exist under a wide range of superflow velocities.
 The Experimental Evidence for Vortex Nucleation 47

It has been suggested by Glaberson and Donnelly [10], in connection with the
critical velocity problem in an aperture, that imposing a flow to a vortex pinned
between opposite lips of the aperture would induce deformations such that the
vortex would twist on itself, undergo self-reconnections, and mill out free vortex
loops. As shown by numerical simulations of 3D flows involving few vortices
only [44], vortex loops and filaments are stable even against large deformations.
Vortices are not prone to twist on themselves and foster loops. It takes the
complex flow fields associated with fully developed vortex tangles to produce
small rings, as discussed at the Workshop [45,46]. And it takes some quite special
vortex pinning geometry to set up a vortex mill that actually works; Schwarz
has demonstrated by numerical simulations that a vortex pinned at one end
and floating along the flow streamlines with its other end free moving on the
wall develops a helical motion, a sort of driven Kelvin wave, and reconnects
sporadically to the wall when the amplitude of the helical motion grows large
enough [47]. This helical mill does churn out fresh vortices.
The above remarks on the stability of vortex loops or half-loops in their course
make it unlikely that the multiple slips be due to the production of small rings
by a vortex after having left the vicinity of the nucleation centre. Furthermore,
such a purely hydrodynamical process would depend on the velocity of the flow
only, contrary to the results shown in Fig. 5. What appears more likely is that
multiple slips are produced by a transient vortex mill of the helical type suggested
by Schwarz operating very close to the nucleation site. The pinning of the mill
vortex and its subsequent release would take place immediately after nucleation
when the velocity of the vortex relative to the boundary is still small and the
capture by a pinning site easy. This process depends on the precise details of the
pinning site configuration and of the primordial vortex trajectory, factors which
allow for the variableness of multiple slips on contamination and pressure (i.e.
single phase slip nucleation velocity) [48]. That pinning does take place close to
sites where vortices are nucleated is shown below.

In-situ contamination by atomic clusters: pinning and collapses. In a

series of experiments conducted at Saclay [49,50], it was found that heavy con-
tamination of the cell by atomic clusters of air or H2 caused numerous collapses of
the ‘singular’ type to occur. The peak amplitude charts of the resonator became
mostly impossible to interpret, except in a few instances where two apparent
critical velocities for single slips were observed. The higher critical velocity cor-
responded to the one observed in the absence of contamination. The lower critical
velocity is thought to reveal the influence of a vortex pinned in the immediate
vicinity of the nucleation site. This vortex induces a local velocity which adds to
that of the applied flow and causes an apparent decrease in the critical velocity
for phase slips. Because of this change, the presence of the pinned vortex could
be monitored, the lifetime in the pinned state and the unpinning velocity could
be measured, yielding precious information on the pinning process.
48 É. Varoquaux et al.

A observation about vortex nucleation which comes from these experiments is

rather straightforward: existing vortices, either pinned or moving, do contribute
to the nucleation of new vortices at the walls of the experimental cell.
Another observation is that a large number of unpinning events were tak-
ing place at an ‘anomalously low’ unpinning velocity. A parallel can be made
[50] with singular collapses, which may also occur at ‘subcritical’ velocities and
which were also quite numerous, suggesting that the two effects have a common
cause. Noting furthermore that pinning and unpinning processes were also quite
frequent in these experiments, releasing a fair amount of vagrant vorticity, it
appears quite plausible that both singular collapses and low velocity unpinning
events are caused by vagrant vortices hopping from pinning sites to pinning sites,
eventually passing by close to the pinning centre or the nucleation site, and giv-
ing a transient boost to the local velocity which pushes a pinned vortex off its
perch or causes a burst of vortices to be shed.
As shown in this section, the capability to observe the behaviour of single vor-
tices in resonator experiments, interpreted with the hindsight provided by finely
tuned numerical simulations, leads to a partial understanding of the phenomena
of pinning, multiple slips and collapses which are taking place on a nanometric
scale at the rims of the micro-aperture in strong superflows. Vagrant vortices
reentering the region where the vortex nucleation site lies, nanoscopic vortex
mills à la Schwarz, offer explanations for singular collapses and for multiple slips
respectively which account for the detailed signatures of these effects. Obtaining
a clearer and better grounded picture would require a knowledge of the mi-
crostructure of the nucleation site and its environment which is not available at
present.
A large number of results on topics relevant to this work were presented at
the QVD workshop to which the authors feel unable to refer appropriately here
and to which they hope that other contributions in this volume will give fuller
coverage. One of them (E.V.) wishes to acknowledge the hospitality of the Isaac
Newton Institute where part of the manuscript was prepared.

References
1. W. Zimmermann Jr., C.A. Lindensmith, J.A. Flaten, J. Low Temp. Phys. 110, 497
(1998), and references therein
2. E. Varoquaux, in: Topological Defects and the Non-Equilibrium Dynamics of Sym-
metry Breaking Phase Transitions, ed. by Y.M Bunkov, H. Godfrin (Kluwer Aca-
demic, The Netherlands 2000) p. 303 and references therein
3. S. Burkhart, M. Bernard, O. Avenel, E. Varoquaux: Phys. Rev. Lett. 72, 380 (1994)
4. W. Zimmermann, Jr: J. Low Temp. Phys. 93, 1003 (1993)
5. O. Avenel, G.G. Ihas, E. Varoquaux: J. Low Temp. Phys. 93 1031 (1993)
6. E. Varoquaux, O. Avenel: Physica B 197, 306 (1994) and references therein
7. W. Zimmermann, Jr.: Contemporary Phys. 37, 219 (1996)
8. C. Josserand, Y. Pomeau: Europhys. Lett. 30, 43 (1995)
9. C. Josserand, Y. Pomeau, S. Rica: Phys. Rev. Lett. 75, 3150 (1995)
10. W.I. Glaberson, R.J. Donnelly: Phys. Rev. 141, 208 (1966)
 The Experimental Evidence for Vortex Nucleation 49

11. K.W. Schwarz: Phys. Rev. Lett. 64, 1130 (1990)

12. T. Frisch, Y. Pomeau, S. Rica: Phys. Rev. Lett., 69,1644 (1992) C. Josserand, Y.
Pomeau, S. Rica: Physica D 134, 111 (1999) S. Rica: this Workshop
13. B. Jackson, T. Winiecki, M. Leadbeater, J.F. McCann, C.S. Adams: this Workshop,
and references therein
14. P.H. Roberts, N. Berloff: this Workshop, and references therein
15. W. Zimmermann Jr., O. Avenel, E. Varoquaux: Physica B, 165&166, 749 (1990)
16. Phase winding numbers are obtained from velocities in cm/s by multiplication by
lh /κ, the ‘hydraulic’ length lh characterising the geometry of the aperture. For a
phase slip by 2π, the phase winding number changes by one unit and the trapped
circulation in the resonator loop by one quantum.
17. J. Steinhauer, K. Schwab, Yu. Mukharsky, J.C. Davis, R.E. Packard: Phys. Rev.
Lett. 74, 5056 (1995) J. Low Temp. Phys. 100, 281 (1995). The criticism of the
1
/2 –ring model made in these references has been refuted by E. Varoquaux and O.
Avenel: Phys. Rev. Lett. 76, 1180 (1996)
18. G.G. Ihas, [Link], R. Aarts, R. Salmelin, E. Varoquaux: Phys. Rev. Lett. 69,
327 (1992)
19. V.I. Mel’nikov: Phys. Reports 209, 1 (1991)
20. A.O. Caldeira: PhD Thesis, quoted in A.O. Caldeira, A.J. Leggett: Ann. Phys.
(N.Y.) 149, 374 (1983)
21. E. Varoquaux, M.W. Meisel, O. Avenel: Phys. Rev. Lett. 57, 2291 (1986)
22. G.E. Volovik: Sov. Phys. JETP Lett. 15, 81 (1972)
23. J.S. Langer, J.D. Reppy: Prog. Low Temp. Phys., Vol. 6, ed. C. J. Gorter (North-
Holland , Amsterdam 1970)
24. C.M. Muirhead, W.F. Vinen, R.J. Donnelly: Phil. Trans. Roy. Soc. A 311, 433
(1984) Proc. R. Soc. London A 402, 225 (1985)
25. E.B. Sonin: Physica B 210, 234 (1995)
26. U.R. Fischer: Phys. Rev. B58, 105 (1998) Physica B 255, 41 (1998)
27. C. Nore, C. Huepe, M.E. Brachet: Phys. Rev. Lett. 84, 2191 (2000)
28. S. Rica: Physica D (to be published) and this Workshop
29. R.N. Hills and P.H. Roberts: J. Phys. C 11, 4485 (1978) P.H. Roberts, R.N. Hills
and R.J. Donnelly: Phys. Lett. 70A, 437 (1979)
30. J. Maynard: Phys. Rev. B 14, 3868 (1976)
31. Numerical simulations in the framework of the Gross-Pitaevskii equation presented
at the Workshop by C. Adams and N. Berloff illustrate the deformation of the
boundary layer with velocity quite vividly. It has also been shown in the same
framework that the vortex core radius diverges at a wall [3].
32. E. Varoquaux, G.G. Ihas. O. Avenel, R. Aarts: Phys. Rev. Lett. 70, 2114 (1993)
33. C.A. Jones, P.H. Roberts: J. Phys. A: Math. Gen. 15, 2599 (1982)
34. F.V. Kusmartsev: Phys. Rev. Lett. 76 1880 (1996)
35. The form of the logarithmic term in the expression for the energy barrier cannot
be considered as accurate (E. Sonin, private communication)
36. J.C. Davis, J. Steinhauer, K. Schwab, Yu. Mukharsky, A. Amar, Y. Sasaki, R.E.
Packard: Phys. Rev. Lett. 69, 323 (1992)
37. G. M. Shifflett, G. B. Hess: J. Low Temp. Phys. 98, 591 (1995)
38. O. Avenel, M. Bernard, S. Burkhart, E. Varoquaux: Physica B 210, 215 (1995)
39. [Link], W. Zimmermann Jr., O. Avenel: in Proc. NATO workshop on Exci-
tations in 2D and 3D Quantum Fluids, ed. by A.F.G. Wyatt, H.J. Lauter (Plenum
Press, NY 1991) p. 343 and references therein
40. K.W. Schwarz: Phys. Rev. Lett. 50, 364 (1983)
50 É. Varoquaux et al.

41. E. Varoquaux, O. Avenel, M. Bernard, S. Burkhart: J. Low Temp. Phys. 101, 821
(1995)
42. D.D. Awschalom, K.W. Schwarz: Phys. Rev. Lett. 52, 49 (1984)
43. P.W. Adams, M. Cieplak, W.J. Glaberson: Phys. Rev. Lett. 78, 3602 (1985)
44. K.W. Schwarz: private communication to E.V.; the 3D simulations presented at
the QVD Workshop by C. Adams also show that vortex loops can undergo severe
deformations and not break apart (to appear in Euro. Phys. Lett.)
45. B.V. Svistunov: Phys. Rev. 52, 3647 (1995)
46. M. Tsubota, T. Araki, S.K. Nemirovskii: Phys. Rev. B (in press) and this Workshop
47. K.W. Schwarz: Phys. Rev. Lett. 64, 1130 (1990)
48. W. Zimmermann: Jr., J. Low T. Phys. 91, 219 (1993) J. Flaten: PhD Thesis,
Univ. of Minnesota (1997, unpublished). The large dissipation events reported by
these authors apparently fall in yet another category than the multiple slips and
the singular collapses discussed here as they persist over a number of resonator
periods.
49. P. Hakonen, O. Avenel, E. Varoquaux: Phys. Rev. Lett. 81, 3451 (1998)
50. E. Varoquaux, O. Avenel, P. Hakonen,Yu. Mukharsky: Physica B 255, 55 (1998)
Applications of Superfluid Helium
in Large-Scale Superconducting Systems

Steven W. Van Sciver

National High Magnetic Field Laboratory, Florida State University, USA

Abstract. The application of superfluid helium (He II) in large-scale superconducting

systems in reviewed. For most of these systems, He II is the coolant of choice, com-
pared to alternative low-temperature helium cooling methods, for its combination of
excellent thermal transport properties at reduced temperature. These advantages come
at some cost to system designers as the hardware associated with delivering He II is
more complex. The paper provides overviews of several major superconducting systems
that utilize He II cooling in terms of the relevant thermal fluid properties. Included
are issues of transient and steady-state heat transport, high Reynolds number flows,
and two-phase phenomena. An effort is made to show the connection between the fun-
damental superfluid properties and the need of the application. Areas where further
understanding of the superfluid state could benefit applications are also discussed.

1 Introduction
Liquid helium has a wide range of uses in low temperature technology today. The
properties of liquid helium that make it particularly valuable are its low temper-
ature, persistence of the liquid state to the lowest achievable temperatures and
the existence of the superfluid state with its associate unique transport phenom-
ena. These properties combined with an ever-improving cryogenic engineering
infrastructure have allowed the development of a number of large-scale systems
that use liquid helium cooling.
There are primarily two classes of large-scale applications for liquid helium:
1. Large superconducting systems such as magnets or RF cavities for high-
energy physics accelerators, fusion systems and other magnet facilities.
2. Large space-based instruments for infrared astronomy and other fundamen-
tal studies.

Each of these applications has a unique set of requirements that in turn place
requirement on the coolant, which determines the preferred state for the liquid
helium. In many cases, the best choice for the coolant is superfluid helium (or
He II) typically in the temperature range 1.4 K to 2.1 K. The decision to select
He II as a coolant is usually driven by a combination of the lower temperature
and beneficial heat transport characteristics.
The present paper begins with an overview of the basic design features and
operating characteristics employed in several large-scale He II cooled applica-
tions. The motivation for He II cooling, the unique design features, and the

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 51–65, 2001.

c Springer-Verlag Berlin Heidelberg 2001
52 S.W. Van Sciver

required engineering database are reviewed in the context of these applications.

In the subsequent section, the current understanding of He II as an engineer-
ing fluid is presented followed by comments about areas where further work is
needed.
By far the largest and most varied application for He II cooling is in the area
of large-scale superconducting magnet systems. These include magnet systems
for high energy physics, fusion experiments as well as other specialized magnet
facilities. In addition, there are several large-scale electron accelerators that uti-
lize superconducting RF cavities and are also cooled by He II. Space does not
permit a detailed discussion of all these applications, so the present discussion
will concentrate on three such systems, which represent the range of technical
demands placed on He II cryogenics:

1. The dipole magnet system for the Large Hadron Collider (LHC) which
is under construction at the European Organization for Nuclear Research
(CERN) laboratory in Geneva, Switzerland.
2. The recently completed outsert superconducting magnet of the 45-T Hybrid
Magnet at the National High Magnetic Field Laboratory in Tallahassee,
Florida.
3. The RF cavity system for the Teravolt Electron Synchrotron Linear Ac-
celerator (TESLA) under design at the Deutsches Elektronen-Synchrotron
(DESY) laboratory in Hamburg, Germany.

There are a number of advantages and some disadvantages to the use of

He II for superconducting systems. The first and most obvious advantage is
simply that the lower temperature provides improved superconductor properties.
For example, NbTi superconductors, which are used in more than 95% of the
operating superconducting magnets today, have substantially improved current
densities with operation near 1.8 K as compared to the normal 4.2 K operation in
atmospheric He I. This improved performance allows the achievement of higher
magnetic fields. For example, a standard 8 T NbTi laboratory magnet will often
achieve 10 T when operated at 1.8 K. Similarly, niobium RF cavity used in
electron accelerators have losses which decrease significantly with temperature
justifying operation near 2 K.
The other, somewhat more subtle, advantage to the use of He II cooling in
superconducting systems is in the uniquely high effective thermal conductivity
of the fluid; a property which provides improved thermal stability to the super-
conducting system. In application, superconductors dissipate negligible amounts
of heat during steady current operation. However, there are number of special
conditions that can lead to power dissipation in the superconductor. To mini-
mize this effect, the superconductor is co-processed with a normal metal (usually
copper) that acts as a shunt for current should conditions prevent the supercon-
ductor from carrying the full current. The heat necessary to bring about the
momentary normalization of the superconductor can come from a variety of
sources (dB/dt, mechanical motion) and must be guarded against. Efficient heat
removal from the conductor is therefore a critical aspect of good reliable design.
 Applications of Superfluid Helium 53

Since He II can transport large heat fluxes (compared to that of He I), its sta-
bilizing effect on magnet systems can be very beneficial. The understanding of
the relevant thermal processes in He II has been a subject of numerous recent
experimental investigations.
The principal disadvantage to the use of He II cooling in large superconduct-
ing systems originates from the increased complexity and cost of the cryogenic
system. In very simple terms, the cost of refrigeration scales with the Carnot
factor, which at low temperatures goes as T−1 . Therefore, a unit of power de-
posited in a He II cryogenic system operating at 1.8 K will require a refrigeration
system 2.3 times larger than the equivalent system operating at 4.2 K. Since the
cost of refrigeration systems scales monotonically with power, the advantages of
superconductor performance and stability must outweigh this cost issue. Thus,
the selection of a He II cooling system is only warranted in cases where the
overall system design is optimized by this approach. What follows are three such
examples.

2 Superconducting Systems That Use He II Cooling

2.1 Accelerator Magnet System for LHC

The LHC accelerator currently under construction at CERN will, when complete,
operate the world’s largest He II cryogenic system [1]. In this case, all the main
ring magnets (dipole and quadrupoles) are made with NbTi superconductor and
cooled to 1.9 K in a He II bath pressurized to atmospheric pressure. The dipole
magnets, shown in cross section in Fig. 1, operate at 8.3 T, which is higher than
previously built accelerator dipoles such as for the SSC and therefore require the
lower temperature and improved thermal stability that He II operation affords.
There are eight refrigeration plants around the 26.7 km circumference of the
accelerator, each cooling one 45◦ sector. The 1.8 K refrigeration capacity for
each sector is between 2.1 and 2.4 kW.
The system uses pressurized He II to cool the magnets, which reduces the pos-
sibility of trapped vapor within the windings of the coil. The use of pressurized
He II to cool magnets was first demonstrated on a large scale in a development
program for the Tore Supra tokamak plasma experiment in France [2]. The ap-
proach requires the use of a low pressure of He II heat exchanger as part of
the installation. In the case of LHC, the heat exchanger is a corrugated pipe
within the magnet vessel and immersed in the He II reservoir. A schematic of
this system is shown in Fig. 2. The heat exchanger is partially filled with satu-
rated He II. The two phase He II heat exchanger provides a much more efficient
heat removal method than can be obtained through counterflow heat transport
in the bulk fluid. Although heat transport is very efficient in bulk He II, the long
distances between refrigeration stations would demand very large He II cross
sections to minimize the temperature gradient.
54 S.W. Van Sciver

Fig. 1. Schematic of LHC dipole magnet

Fig. 2. Cooling system for LHC dipole magnets with two-phase heat exchanger

2.2 High Field Solenoid for the NHMFL 45-T Hybrid

The 45-T hybrid is a large combined superconducting and resistive magnet re-
cently put in service at the NHMFL [3]. The superconducting portion of this
magnet is a 14 T solenoid with a 710 mm cold bore operating at 10 kA. The
outer coil of this magnet is made with NbTi conductor, while the high field
inner sections use Nb3 Sn for its superior high field characteristics. The conduc-
tor in this magnet is a Cable-in-Conduit type (CICC), cf. Fig. 3, consisting of
a superconducting stranded cable in a rectangular steel jacket. This is an ad-
vanced conductor design developed for leading edge magnet systems. The liquid
helium coolant, which is contained within the jacketed conductor, absorbs the
transient heat generation from the conductor and conducts it to the surrounding
 Applications of Superfluid Helium 55

path. The NHMFL hybrid outsert is the first CICC magnet that employs He II
cooling. As in the case of LHC, He II cooling is applied to achieve the highest
current densities in the conductor, while providing a good thermal stabilizing
environment.

Fig. 3. Cable in Conduit conductor for hybrid magnet

A schematic cross section of the 45-T hybrid magnet system is shown in

Fig. 4. The coils are located in the magnet vessel, which is separate from the
He II refrigeration system and cooled by counterflow through a horizontal duct.
Heat extraction occurs at a set of saturated heat exchangers located in the
supply cryostat. These heat exchangers are supplied from the upper reservoir
of the cryostat with the vapor flow going through room temperature vacuum
pumps. The total steady state heat load on this system is less than 10 W at
1.7 K [4]. Thus, the scale of the refrigeration system is modest compared to the
LHC.

2.3 RF Cavity Systems for the TESLA Electron Collider

The TESLA electron collider is one of the proposed major accelerator systems
that will follow the completion of the LHC [5]. This system is a linear electron
collider as compared to a circular LHC machine, which accelerates protons. Lin-
ear machines do not require the large number of dipole bending magnets that
are the signature of the hadron colliders. Rather the electron colliders depend
on high gradient RF cavities for accelerating the beam and use a relatively small
number of magnets for beam steering and focusing. The RF cavities can be either
56 S.W. Van Sciver

Fig. 4. Schematic of 45-T hybrid magnet

made from copper and operate at room temperature or from superconducting

niobium operating near 2 K. This latter approach has been demonstrated in
the Continuous Beam Accelerating Facility (CEBAF) at the Thomas Jefferson
Laboratory in Virginia. The use of niobium cavities provides higher accelerating
voltages and lower overall losses compared to operation at higher temperature
(e.g. 4.2 K).

Fig. 5. Schematic of TESLA cooling system

A schematic of the cryogenic system for TESLA [6] is shown in Fig. 5. It

consists of a long cryostat, which contains the RF cavities, filled with saturated
He II typically at 2 K. Unlike superconducting magnets, large local disturbances
are not expected to occur in RF cavities, so the complexity of a pressurized He II
system is not required. The RF cavity cryostats, cross section shown in Fig. 6,
are maintained at constant temperature over a long length by linking them to
a common vacuum line, which may be partially filled with liquid. Thus, the
 Applications of Superfluid Helium 57

principal technical issue associated with the design of the He II system pertains
to the behavior of two-phase He II/vapor in near horizontal channels of relatively
large diameter.

Fig. 6. Schematic of TESLA cavity cryostat

3 Application Relevant He II Properties

The application of He II cooling to large systems requires an understanding of

the fundamental properties of this unique fluid. Of greatest interest are those
properties that affect engineering design of the system. In some cases, the strong
demands are placed on the He II coolant and it is only through practical inves-
tigation combined with basic understanding that one can apply He II cooling
in an optimal way. The present section reviews current practical knowledge of
He II as it affects applications. Some discussion of the need for further work is
also given.

3.1 Second Sound Pulse Transport

The transport of heat pulses in He II is primarily of relevance to the stability of

superconducting systems (mostly conductors) in He II. This is a very complex
problem, the solution of which continues to elude scientists and engineers. The
primary reason for this problem comes from an incomplete understanding of the
source and characteristics of the heat deposition spectrum that originates from
58 S.W. Van Sciver

a working superconducting magnet. Such disturbances come from a variety of

sources (AC losses in the conductor or mechanical motion of the winding) and
can vary considerably from magnet to magnet. Thus, most of the effort is placed
on understanding the transient thermal response of the He II coolant. One would
like to understand the development of the fully turbulent state in He II and the
limits of heat transfer. Nemirovskii and Tsoi [7] have presented this problem in
the form of a regime map, which qualitatively defines the transition of He II
from the ideal superfluid state through the development of the fully developed
turbulent state to the limit of heat transfer film boiling. For short times, the heat
transport is by a burst of second sound pulse, which leads to the development of
turbulence. This second sound pulse can be of benefit to the operation of very
high current density magnets, which sometimes experience very short duration
intense disturbances.
Several experiments have reported on intense heat pulse propagation in He II
[7,8]. Shimazaki et al. [8] measured the shape and heat content of thermal shock
pulses in a counterflow He II channel. Fluxes as high as 40 W/cm2 for up to
1 ms duration were applied and shown to propagate at second sound velocities.
Turbulence generated in the He II followed the pulse an effect that limited the
maximum energy transported to about 10 mJ/cm2 . This is a large energy de-
position for high field compact superconducting magnets and thus could affect
design and performance. However, one would like to better understand how this
process transforms into the fully developed turbulent state, which is discussed
in the next section.

3.2 Transient and Steady Transport in the Mutual Friction Regime

In many large-scale superconducting systems, the He II can be assumed to be
fully turbulent and the heat transport governed by the mutual friction regime.
There are several features of these systems that support this assumption. First,
most of these systems have relatively large channels and He II reservoirs, so
that the He II already contains a considerable amount of turbulence. Also, large
magnets have slower thermal response times, typically 10 ms to 1 s, so that
propagation of second sound pulses can only be a small portion of the total flux.
This assumption is the design basis of most large-scale superconducting systems
that are cooled with He II.
With the assumption of fully developed turbulence, one can then apply the
mutual friction form of the two fluid equations to the general problem of heat
transport in the He II [9]. Thus, the temperature gradient in the fluid is given
by the relationship,
 1/3
−1 dT
q=− f (1)
dx
where f = ρ3Aρ n
4 3 and A is the usual Gorter Mellink parameter. The interesting
ss T
development related to this process is the realization that (1) can be applied
much like Fourier’s Law for conduction in materials, although the form is pro-
portional to the cube root of the temperature gradient. Thus, slowly varying
 Applications of Superfluid Helium 59

heat transport processes should have a diffusion-like character and be governed

by the He II diffusion equation,
 1/3
∂T ∂ −1 ∂T
ρCp = f . (2)
∂t ∂x ∂x
This formulation has been successful at treating a wide variety of transient
heat transport problems related to superconducting magnet systems such as the
thermal stability of a composite superconductor.
In a superconducting composite conductor, the power generated in the con-
ductor is a function of time. For short times, the power usually peaks as the
current distributes within the conductor and the initial heat deposited diffuses
away. After the peak, the power usually levels off and corresponds to the steady-
state joule heat in the copper stabilizer. A simplified form for the heat generated
in the conductor is shown by the cross-hatched portion in Fig. 7. To ensure that
the conductor returns to the superconducting state, it is necessary to have the
heat transfer to the He II exceed the heat generated. The highest level of reli-
ability is obtained by requiring that the Joule heat never exceed the maximum
steady-state heat flux in the He II. However, to optimize the design, it is really
only necessary to have the integrated average heat transfer exceed the genera-
tion. This condition, which was first proposed by P. Seyfert et al. [10], is shown in
Fig. 7. The curve represents the maximum heat transport by the He II governed
by the diffusion process, see (2). As long as the excess heat generation, area A,
does not exceed the excess heat tranfer, area B, the magnet should be stable.

A H e a t T ra n s fe rre d to H e II

o w e r (W )

B
I2R
H e a t G e n e ra tio n

T im e
Fig. 7. Stability criterion for He II cooled conductor

3.3 The He II Energy Equation

Some applications of He II require forced circulation through a closed loop. This
approach adds complexity to the system, but can provide significantly enhanced
60 S.W. Van Sciver

heat removal particularly in distributed or remote systems. For example, such a

forced flow system was considered for the 45-T hybrid outsert magnet, but not
accepted in the final design due to the additional complexity and cost. Adding
a net fluid velocity to He II provides some advantage in design as well as intro-
ducing a new variable into the analysis of the fluid behavior.
As in the case of static He II applications, if one assumes that the He II is
governed by the fully developed turbulent mutual friction process, then forced
flow He II heat transport should obey a modified energy equation. This equation
has the form [11],
 1/3
∂ −1 ∂T 1 ∂p ∂T ∂ρ ∂T
f − − ρuC −u = ρC (3)
∂x ∂x ρs ∂x ∂x ∂x ∂t

where the pressure gradient terms take into account the frictional losses and
change of internatl energy with pressure. As before, this relationship has success-
fully modeled the behavior of heat transport in He II for a variety of experimental
systems. An interesting observation about this formulation is the fact that He II
will display the Joule Thomson effect when experiencing a pressure drop through
an insulated tube [12]. The result will be an increase in temperature, which will
generally have a negative impact on applications.
An example of heat transport in forced flow He II is displayed in Fig. 8.
The flow is from left to right and a heat pulse is deposited at x = 0. The
plot displays calculated and experimental time-dependent temperature profiles
at different locations along a channel containing He II. Note that the peak in
the temperature profile broadens due to thermal diffusion as time progresses
and the pulse propagates. Also, the location of the peak moves at approximately
the speed of the fluid, suggesting a rough method of measuring fluid velocity in
such systems. At these high velocities, the Joule Thomson effect contributes by
producing a gradual increase of the fluid temperature as seen by the increase in
the baseline temperature as the fluid moves through the tube.
From the application viewpoint, the steady-state and transient heat transport
characteristics a fully turbulent He II can be understood in terms of the He II
diffusion and energy equations. There are limits to this representation and these
limits need to be explored. Work is continuing on the development of turbulence
and heat transport at very high velocities. This seems to be an area where some
fundamental work would be able to contribute.

3.4 Fluid Dynamics of Forced Flow He II

Technical interest in forced flow He II has grown out of a number of applica-

tions. As mentioned previously, the LHC accelerator magnet system and the
NHMFL hybrid magnet both had early designs that utilized forced flow He II
cooling. In addition, several space-based infrared telescopes, including the cur-
rently underway Space Infrared Telescope Facility (SIRTF), have evaluated ways
of circulating He II. The fundamental question that arose out of these studies
 Applications of Superfluid Helium 61

Fig. 8. Pulsed heat transport in forced flow He II (from [17])

was whether the He II pressure drop and associated fluid dynamic properties
would be unique. Since that time, a considerable body of research has come to
support the notion that the dynamic behavior of forced flow He II is essentially
similar to that of classical fluids. Thus, He II when flowing through tubing dis-
plays a pressure drop that can be described by classical correlation [13]. This
result has been supported by theoretical investigations that suggest the two flu-
ids are coupled through turbulent interactions and thus flow together. Figure 9
displays recent measurements of the friction factor for He II flowing in a 10 mm
ID tube at Re>107 (u> 10 m/s). These data continue to support the previous
observations.
Forced flow He II at high Reynolds number has more recently become inter-
esting as a test fluid for basic fluid dynamic investigations [14]. The fluid has a
very small kinematic viscosity allowing high Reynolds numbers to be achieved
in sub-sonic flows. A question that continues to be raised is to what extent can
one ignore superfluid effects. This is still an open issue and a subject for further
investigation. In a companion paper at this conference, we report on drag coeffi-
cient measurements for a sphere in flowing He II [15]. The measurements suggest
a temperature dependence to CD not seen in normal fluids and this dependence
appears to correlate with the normal fluid density, ρn . Clearly, this topic needs
further experiment supported by theoretical analysis.
62 S.W. Van Sciver

4 .1 m s t r a ig h t
4 .6 m lo o p
f 2 m c o p p e r lo o p
1 .2 m s t r a ig h t
V o n K a r m a n - N ik u r a d s e
C o le b ro o k
0 .0 1

0 .0 0 1
5 6 7 8
1 0 1 0 1 0 1 0
R e
Fig. 9. He II friction factor for pipe flow at high Reynolds number

3.5 He II/Vapor Two Phase Flow

An area that has received recent investigation driven by the needs of large accel-
erators such as LHC or TESLA is two phase He II/vapor flow. In many of these
large systems, there are tubing sections partially filled with He II and in near
horizontal configuration. This operating condition leads to some interesting phe-
nomena due to the existence of the free surface. As is in the case of single phase
He II systems, the heat transfer and fluid dynamics behavior are of interest to
designers. Further, because of the free surface and relative velocity between the
liquid and vapor phases, there are a variety of flow stability issues that need to
be addressed. These are topics of current investigation [16,17,18]. Modeling these
systems is complex due to the interaction between the two phases [19,20]. To
date, most studies have treated the He II as a classical fluid with heat transport
character governed by the He II energy equation. Clearly, this fluid system is
complex and further experimental and theoretical work is needed.

3.6 Fountain Effect (Fluid Management)

The thermomechanical fountain effect in He II has been considered during the
design of a variety large scale applications. The most successful application of this
effect occurred with the Superfluid Helium On Orbit Transfer (SHOOT) experi-
ment which flew on the Space shuttle in 1993 [21]. This experiment demonstrated
the ability of a large-scale fountain pump to tranfer He II in micro-gravity condi-
tions. Also, this experiment as well as most other space-based He II systems used
a porous plug phase separator to contain the liquid helium within the dewar and
vent the vapor. The porous plug phase separator uses the heat of evaporation to
remove the small internal heat generated in the dewar. The temperature differ-
 Applications of Superfluid Helium 63

ence between the He II bath and exciting vapor provides the thermomechanical
pressure head to hold back the liquid.

Fig. 10. Schematic of method to use porous plugs to manage He II fluid level

The use of the thermomechanical fountain effect is much more limited in

large-scale superconducting systems. It has been considered in the design of
space-based magnet systems for similar reasons to its use in infrared telescope
technology, i.e. mostly as a phase separator. The fountain effect may also be
useful as part of the fluid management system for ground-based accelerator sys-
tems containing horizontal He II/vapor two phase flows. One concern in these
systems is “dry out”of the piping due to the pressure and temperature gradient
along the channel. Higher temperature means higher vapor pressure, which will
depress the liquid level compared to that down stream. Thus, the liquid in the
channel will slope upward in the downstream direction if there is no net liquid
flow. One possible method to overcome this concern is to use the fountain effect
to return and circulate He II to the upstream side of the channel. A schematic
of how this process can be achieved is shown in Fig. 10. The porous plugs are
part of a parallel fluid handling system. The increase in temperature on the up-
stream side would have the effect to force liquid through the porous plug thus
raising the level and preventing dry-out. There is no current application for this
concept, but some technical sub-components demonstrated the approach [22].

4 Conclusions
Superfluid helium has become an engineering fluid for a number of technical
applications in superconductivity and space-based instrumentation. A consider-
able volume of practical data has been accumulated through the development of
these systems. Steady-state and slowly varying transient thermal processes can
be described to be a diffusive process much like conduction. Forced flow at rela-
tively high velocities appears to obey clssical correlations. There are a number of
areas where further study is needed. We do not have adequate understanding of
the development of turbulence particularly at high heat fluxes. Also, two-phase
64 S.W. Van Sciver

flow is an entire subject that has only recently been under investigation. Future
applications will no doubt require additional investigation.

Acknowledgements
The National High Magnetic Field Laboratory is jointly funded by the National
Science Foundation and the State of Florida.

References
1. P. Lebrun, Cryogenics for the Large Hadron Collider. IEEE Trans. On Applied
Super., Vol 10, 1500 (2000)
2. G. Claudet and R. Aymar, Tore Supra and He II Cooling of Large High Field
Magnets, Adv. Cryog. Engn. 35A, 55 (1990)
3. J.R. Miller, et al., An Overview of the 45-T Hybrid Magnet System for the
NHMFL, IEEE Trans. On Magnetics, Vol 30, 1563 (1994)
4. S.W. Van Sciver, et al., Design, Development and Testing of the Cryogenic System
for the 45-T Hybrid, Adv. Cryog. Engn. Vol 41, 1273 (1996)
5. R. Brinkmann, et al., Conceptual Design of a 500 Gev e+e- Linear Collider with
Integrated X-Ray Laser Facility, DESY 1997-048 ECFA 1997-182, May 1997
6. G. Horlitz, et al., The TESLA 500 Cryogenic System and He II Two Phase Flow,
Cryogenics Vol 37, 719 (1997)
7. S.K. Nemirovskii and A.N. Tsoi, Transient Thermal and Hydrodynamic Processes
in Superfluid Helium, Cryogenics Vol 29, 985 (1989)
8. T. Shimazaki, M, Murakami and T. Iida, Temperature measurement in Transient
Heat Transport Phenomena Though a Thermal Boundary Layer in High Vortex
Density, Adv. Cryog. Engn. Vol 41, 265 (1996)
9. S.W. Van Sciver, Chap. 10: Helium II (Superfluid Helium), in: Handbook of Cryo-
genic Engineering, J.G. Weisend II (ed.). Taylor & Francis (1998)
10. P. Seyfert, Practical Results on Heat Transfer in Superfluid Helium, in: Stability
of Superconductors in He I and He II, IRR Commission A 1/2 (1981), pp. 53-62
11. B. Rousset, Pressure Drop and Transient Heat Transport in Forced Flow Single
Phase He II at High Reynolds Number, Cryogenics Vol 34 supplement, 317 (1994)
12. P.L. Walstrom, Joule Thomson Effect and Internal convection Heat Transfer in
Turbulent He II Flow, Cryogenic Vol 28, 151 (1988)
13. P.L. Walstrom, et al., Turbulent Flow Pressure Drop in Various He II Transfer
System Components, Cryogenics Vol 28, 101 (1988)
14. R.J. Donnelly, Ultra High Reynolds Number Flows Using Cryogenic Helium: An
Overview, in: FLow at High Reynolds and Rayleigh Numbers, R.J. Donnelly and
K. Sreenivasan (eds.), Springer (1998)
15. M.R. Smith, Y.S. Choi and S.W. Van Sciver, The Temperature Dependent Drag
Crisis on a Sphere in Flowing He II, (this publication)
16. P. Lebrun, et al., Cooling Strings of Superconducting Devices Below 2 K: The
He II Bayonet Heat Exchanger, Adv. Cryog. Engn., Vol 43, 419 (1998)
17. B. Rousset, et al. Behavior of He II in Stratified Counter-Current Two Phase
FLow, in: Proceedings of the ICEC17, Institute of Physics Publishing (1998) pp.
671-674
18. J. Panek and S.W. Van Sciver, Heat Transfer in a Horizontal Channel Containing
Two Phase He II, Cryogenics Vol 39 (1999)
 Applications of Superfluid Helium 65

19. L. Grimaud, et al., Stratified Two Phase Superfluid Helium Flow, Cryogenics Vol
37 (1997)
20. Y. Xiang, et al. Numerical Study of Two Phase He II Stratified Channel with
Inclination, IEEE Trans on Applied Super. Vol 10, 1530 (2000)
21. M. DiPirro and P.J. Shirron, The SHOOT Orbital Operations, Cryogenics Vol 32
(1992)
22. J. Panek, Y Zhao and S.W. Van Sciver, Liquid Level Control Using a Porous Plug
in a Two Phase He II System, Adv. Cryog. Engn., Vol 43, 1401 (1998)
The Temperature Dependent Drag Crisis
on a Sphere in Flowing Helium II

Yeon Suk Choi2 , Michael R. Smith1 , and Steven W. Van Sciver1,2

1
National High Magnetic Field Laboratory, Florida State University, Tallahassee,
FL 32310, USA
2
Mechanical Engineering Department, FAMU-FSU College of Engineering,
Tallahassee, FL 32310, USA

Abstract. In a previous paper, we reported observing a drag crisis on a sphere in

flowing He I and He II. Data in He II suggested a possible temperature dependence
to the critical Reynolds number, as well as the magnitude of the crisis. In this paper,
we explore temperature dependence more completely. Dynamical similarity arguments,
which lead to Reynolds number scaling in the case of the Navier–Stokes equations, are
applied to the two-fluid equations. The result is a modified Reynolds number involving
the factor 1-δ, where δ ≡ ρs /ρ. The ramifications of this argument, together with other
possible scaling relationships, are discussed. Data and critical Reynolds numbers are
plotted for several temperatures between 1.6 K and 2.0 K. Results appear to agree well
with the proposed scaling for He II.

1 Introduction

The low kinematic viscosity (ν = η/ρ, where η and ρ are dynamic viscosity
and total density, respectively) of liquid helium makes it an attractive fluid
for modern dynamical similarity studies, where one wishes high Reynolds num-
bers (Re=Ud/ν, where U and d are the characteristic velocity and dimension
of the flow field) without transonic effects. While research suggests that helium
above 2.176 K (He I) behaves as a classical fluid, He II (the liquid phase below
2.176 K) is a quantum fluid with a wide range of non-classical macroscopic prop-
erties. Still, studies have found that classically generated turbulence in He II may
behave classically in certain experiments [1,2,3]. Uncertainties about the micro-
scopic character of classically generated turbulence in He I and He II motivated
the previous work [4,5] in which the form drag on a sphere in flowing He I and
He II was calculated from the observed pressure distribution over the surface.
If the critical Reynolds number for the drag crisis in He II is temperature
dependent, then the dimensionless equations of motion for the two fluid system
must scale with other dimensionless parameters, in addition to or instead of the
Reynolds number. In the classical fluid dynamics of ordinary fluids (including
He I), dynamical similarity and scaling arguments for expressing experimental
data in terms of Reynolds number, coefficients of drag, lift and so on, spring
rigorously from non-dimensionalizing the Navier–Stokes equations. Expressing
He II data in terms of an effective Reynolds number however, has been more
of an empirical convenience. Perhaps one reason for this is the empirical nature

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 66–72, 2001.

c Springer-Verlag Berlin Heidelberg 2001
 The Temperature Dependent Drag Crisis 67

of the two-fluid equations themselves, and the sometimes unspoken question

of whether the current set of equations accurately represent all aspects of the
physics. In the spirit of investigation however, we present one such analysis as a
framework for discussing our data. [6]
The generally accepted form for the two-fluid equations in one dimension [7]
is shown in (1) and (2), where we have restricted the problem to one dimension
for simplicity. The entropy terms describe the exceptionally high effective ther-
mal conductivity of the system through a mechanism called counterflow. The
equations for the superfluid and normal fluid components are coupled through
the velocity difference vns = vn − vs , and the mutual friction term Fns . One
expression for the mutual friction is shown in (3), where α is a temperature
dependent parameter. The local line length per unit volume of quantized vortex
lines existing in the superfluid component is given by L, and K is the quantum
of circulation around a vortex line.
2
dνs ρs dP dT ρs ρn dνns
ρs =− + ρs S + − Fns (1)
dt ρ dx dx 2ρ dx

2
dνn ρn dP dT ρs ρn dνns d 2 νn
ρn =− − ρs S − + Fns + η 2 (2)
dt ρ dx dx 2ρ dx dx
2
Fns = − αkρs Lνns (3)
3
Now, we define non-dimensional quantities in terms of representative veloci-
ties, distances and temperatures as follows;

x0 = x/D t0 = U · t/D
L0 = L · D2 T0 = (Tλ − T )/Tλ
νn0 = νn /U S0 = Tλ S0 /U 2
νs0 = νs /U P0 = P/ρU 2

where the zero subscript refers to the dimensionless quantity and Tλ = 2.176 K is
the lambda point. Upon substituting these quantities into (1-3), and rearranging
terms a bit, we arrive at;
2   2
dνs0 dP0 dT0 αk 1 − δ dνns0
=− − S0 + 3 · L0 νns0 + (4)
dt0 dx0 dx0 UD 2 dx0

   2    2
dνn0 dP0 δ dT0 3 αk δ dνns0
=− + · S0 − · L0 vns0 − (5)
dt0 dx0 1−δ dx0 UD 2 dx0
1 1 d2 νn0
+ ·
1 − δ Re dx20

The terms involving the temperature gradient drive counterflow, in which the
two fluids flow in opposite directions, and heat is transported within the system.
68 Y.S. Choi, M.R. Smith, and S.W. Van Sciver

Note that the effect of a thermal gradient upon the normal fluid component
(Eq. 6) becomes larger than 1 − δ → 0 where δ ≡ ρs /ρ. This is comforting,
since it agrees well with our expectations. Furthermore, the classical equations
of motion are recovered in the limit δ → 0 . From a more general perspective,
Eqs. (4) and (6) describe a wide spectrum of physics, which goes beyond fully
developed co-flowing turbulent fields. The driving terms for such non-classical
phenomena as counterflow however (we include for consideration at this point
terms involving vns0 ), must affect the dynamics of all flows at suitably small
length scales. Since the co-flowing condition (vns0 = 0) generally depends upon
mutual friction, it must depend upon vns0 = 0 across some range of length scales
within the flow.
In addition to the many factors of (1-δ) in (4) and (6), which underscore
the strong role of temperature in the dynamics of He II, there is one additional
dimensionless quantity, UD/(2/3)αk, which is tied into the mutual friction. Al-
though we will be discussing co-flowing turbulence, where vns0 may zero, L0 may
still be quite large. Thus, the term on the whole may fluctuate dramatically on
a local scale.
For the sake of the argument at hand however, consider the case where vns0
is strictly zero (fully coupled superfluid turbulence), with zero temperature gra-
dient. Then at suitably large length scales, we are left with
dvs0 dP0
=− (6)
dt0 dx0

dvs0 dP0 1 1 d2 vn0

=− + · (7)
dt0 dx0 1 − δ Re dx20
The most important point to notice is that we never truly recover classical
equations of motion, except in the limit as δ → 0. The viscous term is driven
by (1-δ) · Re, which provides a mechanism through which the effective critical
Reynolds number associated with the drag crisis might scale with temperature.
At first glance, this appears to pose a dilemma. A large body of data characteriz-
ing skin friction in pipes scales nicely with Reynolds number, alongside classical
data, yet without any apparent temperature dependence. One possible explana-
tion for this may lie in the fundamental difference between pipe flow, and flow
over a bluff body, such as a sphere. Pipe flow is a fully developed turbulent field,
where the two fluids largely flow together [1], except at very small length scales
and within a very narrow region close to the wall. Here, the non-slip condition
on the normal fluid provides the observed pipe friction and associated pressure
drop. Flow over a sphere however, possesses a stagnation point on the leading
edge, from which the two fluids accelerate under different boundary conditions,
with a non-zero velocity difference. This brings into play the other terms in (4)
and (6). Thus, we expect δ, and thereby temperature, to play a greater role in
flows where the two fluids are allowed to have substantially different velocity
fields (vns = 0).
As a final preliminary speculation, note that the viscous term in (7) becomes
quite large as 1-δ → 0, similar to inviscid flow (Re → 0). Thus, it may be that
 The Temperature Dependent Drag Crisis 69

the drag crisis itself changes character and magnitude as the normal component
vanishes (ρn → 0). Ultimately in this limit, one expects to recover some sort of
potential flow.

2 Experimental Apparatus and Protocol

The apparatus was described in a previous publication [4]. A 10 mm diameter
sphere is suspended upon a strut oriented perpendicular to the oncoming flow.
A single pressure tap located on the surface of the sphere is used to map out the
pressure distribution on the azimuth connecting the upstream and downstream
points by rotating the sphere/strut. This pressure distribution is then integrated
to yield the form drag.
The principal difference from the previous work is a matter of protocol. Drag
versus Reynolds number were taken at several intermediate temperatures be-
tween 1.6 K and 2.0 K in order to determine the critical Reynolds number for
the drag crisis as a function of temperature. Additionally, drag and pressure
distribution were observed at different temperatures for fixed Reynolds number.

3 Results and Discussion

The measured pressure distribution in He II at 2.0 K, expressed in terms of the
coefficient of pressure, CP = (P(Θ) − P0 )/ 12 ρU2 , is shown in Fig. 1. Points which
comprise the curve are the result of averaging a series of individual measure-
ments. Since many of the important dynamics are dependent upon the equatorial
velocity (Θ = 90◦ ), the Reynolds number was calculated based upon the mean
velocity in this smallest cross-section. In Fig. 1, the Reynolds number spans the
range from 1.1 × 105 to 7.8 × 105 . Profiles for Reynolds number at or above
1.8 × 105 show supercritical behavior with the boundary layer separation occur-
ring at approximately 100 degrees. The data corresponding to the two lowest
Reynolds numbers exhibit variations which may be due to transition, together
with very low signal levels.
Assuming azimuthal symmetry of the pressure distribution with respect to
the oncoming flow, together with the spherical shape of the surface, we may
integrate the coefficient of pressure to calculate the coefficient of drag directly.
For the discrete data here, this is easily performed by a summation. Figure 2
shows drag coefficients for our experiment, together with published results for
the smooth sphere, shown as the solid line. Error bars are derived from the
statistical scatter in the individual measurements which comprise the points on
the curve in Fig. 1. The lower Reynolds number data have larger error bars since
these measurements correspond to lower signal level. Although He II exhibits a
drag crisis at approximately the same Reynolds number as He I, the variability
in the He II data led us to speculate about a temperature dependence to the
turbulent transition within the boundary layer.
The results displayed in Fig. 2 suggest that the drag coefficient in He II in-
creases with decreasing temperature. For fixed Reynolds number, the coefficient
70 Y.S. Choi, M.R. Smith, and S.W. Van Sciver

Fig. 1. Pressure distribution for various Reynolds number at 2.0 K

Fig. 2. Drag coefficient vs. Reynolds number

is the largest at 1.6 K decreasing monotonically to the value in He I. Further,

the minimum in the drag coefficient occurring just above the transition also in-
creases with decreasing temperature. These two observations can be evaluated
in more detail by careful study of the data.
Figure 3 is a plot of the drag coefficient versus temperature for fixed Reynolds
number. Below the lambda transition, the coefficients increase approximately
linearly although not with the same slope for each case. It is also interesting
 The Temperature Dependent Drag Crisis 71

to note that the extrapolation of the measured He II drag coefficients to Tλ

appear to coincide with the approximately constant values measured in He I.
Although these data need further confirmation, it certainly appears that the
drag coefficient is temperature dependent in He II.

Fig. 3. Drag coefficient vs. temperature

To test the two fluid equation scaling arguments, we plot in Fig. 4 the drag
coefficients versus Re (1-δ), as suggested by [5]. The open symbols correspond
to He I data and the closed symbols are for He II.
Comparing Fig. 4 to Fig. 2, we note that the scaling shifts the minimum in
the drag coefficient to approximately the same value of the modified Reynolds
number for each temperature. That value, Re (1-δ) ≈ 2.3 × 105 , is in reasonable
agreement with the minimum in the classical drag coefficient curve at approxi-
mately Re = 3 × 105 .

4 Conclusion

We have measured the pressure distribution over the surface of a sphere in flowing
He II as a function of Reynolds number. The He II data shows clear evidence
of a drag crisis at approximately the same Reynolds number. The coefficients of
drag have a minimum value at same modified Reynolds number based on scaling
the He II two fluid equations, Re ×(1 − δ) ≈ 2.3 × 105 . The variability in the
He II data confirms a temperature dependence to the turbulent transition within
boundary layer.
72 Y.S. Choi, M.R. Smith, and S.W. Van Sciver

Fig. 4. Drag coefficient vs. modified Reynolds number

Acknowledgments
We wish to thank the National High Magnetic Field Laboratory and the De-
partment of Energy, Division of High Energy Physics for their financial support.
Thanks to David K. Hilton for useful conversation and Scott Maier for technical
assistance.

References
1. P.L. Walstrom, J.G. Weisend II, J.R. Maddocks and S.W. Van Sciver, Turbulent
pressure drop in various He II transfer system components, Cryogenics 28, 101,
1988.
2. D.C. Samuels, Velocity matching and Poiseuille pipe flow of superfluid helium,
Phys. Rev. B 46, 11714, 1992.
3. C.F. Barenghi, D.C. Samuels, G.H. Bauer and R.J. Donnelly, Numerical evidence
for vortex-coupled superfluidity: Quantized vortex lines in an ABC model of tur-
bulence, Phys. Fluids 9, 2631, 1997.
4. M.R. Smith and S.W. Van Sciver, Measurement of the pressure distribution and
drag on a sphere in flowing He I and He II, Advances in Cryogenic Engineering,
Vol 43, 1473, 1998.
5. M.R. Smith, D.K. Hilton and S.W. Van Sciver, Observed drag crisis on a sphere in
flowing He I and He II, Physics of Fluids, Vol. 11, No.4, 1999.
6. A similar analysis was originally presented in Evolution and Propagation of Tur-
bulence in Helium II, Ph.D. Thesis, M.R. Smith, 1992, University of Oregon.
7. R.J. Donnelly, Quantized Vortices in Helium II, Cambridge, New York, 1991.
Experiments on Quantized Turbulence
at mK Temperatures

S.I. Davis, P.C. Hendry, P.V.E. McClintock, and H. Nichol

Department of Physics, Lancaster University, Lancaster LA1 4YB, UK.

Abstract. An experiment to investigate the free decay of quantized turbulence in iso-

topically pure superfluid 4 He at millikelvin temperatures is discussed. The vortices are
created by a vibrating grid, and detected by their trapping of negative ions. Preliminary
results suggest the existence of a temperature-independent vortex decay mechanism be-
low T ∼ 70 mK.

1 Background
The renaissance of interest in the turbulent hydrodynamics of HeII has led to
the realisation that, in many respects, it exhibits unexpected similarities to anal-
ogous flows in classical fluids at high Reynolds number [1,2,3]. Unlike a classical
fluid, HeII is well described by a two-fluid model, with a normal (dissipative)
component mutually interpenetrating with a superfluid (inviscid) component
with quantized circulation
  
h
κ = vs .dl = n (1)
m4
where the integral is taken around a loop enclosing the vortex, vs is the superfluid
velocity, m4 is the 4 He atomic mass and the quantum number n is an integer.
The flow properties of HeII in an open geometry are dominated by singly-
quantized vortex lines [4], linear singularities around which the superfluid flows
at tangential velocity vs . Unless velocities are kept extremely small, the liquid
flowing through a tube becomes filled with a tangled mass of such vortex lines.
Because of their quantization, they represent a particularly simple form of tur-
bulence. In that the vortex cores can be considered as part of the normal fluid
component, but the encircling superflow field in accordance with (1) is of super-
fluid component, vortices provide a weak coupling (mutual friction) between the
two components. So it is not at all clear, at first sight, why this complex liquid
system should ever behave like a single-component classical fluid. The question
has recently been discussed in considerable detail by Vinen [3]. One of the aims
of the present project is to establish the properties of the turbulent liquid when
it really does consist of just a single component, i.e. in the low temperature limit
where the normal fluid density is negligible.
In the conventional scenario at higher temperatures 1 < T < Tλ , a vortex
tangle can be maintained by the work done by the driving force, which could be
e.g. a pressure or temperature gradient for bulk flow, or for thermal counterflow

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 73–79, 2001.

c Springer-Verlag Berlin Heidelberg 2001
74 S.I. Davis et al.

(in which the normal and superfluid components move in opposite directions),
respectively. On removal of the driving force, the tangle decays according to the
Vinen [5] equation
dL  2
= −χ2 L (2)
dt m4
where L is the length of vortex line per unit volume and χ2 is a (weakly
temperature-dependent) dimensionless constant. The physical mechanism driv-
ing the decay has been discussed by Schwarz [6], who concluded that it involves
crossings and consequent reconnections of lines. The rapid self-induced motion
of the resultant sharp cusps through the normal fluid is dissipative, and causes
the rounding off of cusps with consequent line shrinkage. The presence of the
normal fluid component is thus a key component of the decay mechanism.
Details of the decay process, e.g. the existence of two distinct decay rates
[5,7], are not at all well understood, partly because it sometimes seems to in-
volve turbulence in the normal fluid component as well as in the superfluid. The
experiment that we describe below avoids this complication. Vortex decay is in-
vestigated at sufficiently low temperatures (in the mK range) that normal fluid
component as such is absent. Under these conditions, it is far from obvious how
the tangle will decay, but it appears that there are two possibilities –

1. Essentially the same decay process occurs as above 1K, but now driven by the
dilute phonon gas rather than by normal fluid, or
2. A new decay mechanism comes into play.

If the former applied, decay times would be expected to become exceedingly

long as T decreased, given that the vortex-phonon cross section is tiny [8] and
that the phonon density is falling as T 3 . The question is difficult to settle on
theoretical grounds, and best resolved by experiment.

2 Creation and Detection of Vortices

Above 1K vortices can conveniently be created [4] by thermal counterflow in
which a thermal gradient causes the two fluids to flow in opposite directions
above a critical velocity. Vortices are usually detected [4] by their attenuation
of second-sound, an entropy-temperature wave in which the two fluids move in
antiphase. Neither of these techniques can be used at mK temperatures, because
there is no normal fluid component. New techniques have therefore been required.
To create vortices we have employed a resonantly-excited circular grid [9].
The expectation was that, in the absence of normal fluid, the amplitude of
vibration would grow until a critical velocity was attained, after which the energy
drawn from the exciting field would be converted to vorticity.
To detect the vortices we have observed the attenuation of ion signals caused
by trapping of the ions on vortex cores as the ion clouds pass through the vor-
tex tangle. An essential requirement of the experimental design is the necessity
of preventing the ions from themselves creating vortex rings – converting to
 Quantized Turbulence at mK Temperatures 75

F ie ld
e m is s io n tip
T o p e le c tro d e
O s c illa tin g
H V g rid
B o tto m e le c tro d e

F ris c h g rid C o lle c to r

Fig. 1. The experimental arrangement (schematic). Some of the ions created by the
field emission tip may get trapped on vortices created previously by the oscillating grid,
thereby reducing the signal arriving at the collector. The perforated top and bottom
electrodes complete the double capacitor needed to excite oscillations of the grid, and
the Frisch grid screens the collector from the approaching charge.

large slowly-moving charged-vortex-ring complexes, and thus being lost from

the signal. An ion in HeII at mK temperatures, in the absence of normal fluid
component, will accelerate steadily under the influence of any electric fluid, how-
ever small, until it attains a critical velocity. It is necessary to ensure that this is
the Landau critical velocity vL , and not the critical velocity for vortex creation.
This can be accomplished by choosing negative (rather than positive) ions [10],
by applying pressures P > 11 bar [10], by using isotopically pure 4 He [11], and
by ensuring that the electric field is not too large [12]. Under these conditions
the ions do not create vortex rings but, rather, travel freely through the liquid
with an average speed slightly in excess of vL [13].
The mechanism by which a rapidly moving ion can get trapped on a vortex
under these conditions is unclear. The vortex presents an effective potential [14]
and, above 1K, normal fluid dissipation enables the ions to lose enough energy
to get trapped on the vortex core. In the absence of normal fluid, and with the
ion moving almost ballistically, like a free particle in a vacuum, we hoped that
the ions would lose enough energy to be trapped by exciting vortex waves.

3 The Experiment
The electrode structure used for the experiment is shown schematically in Fig. 1.
The operating procedure was performed in two stages. First, a high constant
voltage (usually 500 V) was applied to the vortex-generating grid and a periodic
driving voltage of ±270 V was applied to one of the adjacent plates. The drive
76 S.I. Davis et al.

was maintained for several seconds, to build up a tangle of vorticity. Secondly,

the potentials on the electrodes were adjusted so that the electric field would
draw ions from the field-emitter to the collector. The field-emitter was then
pulsed to create an ion cloud, which travelled down the cell. As it passed the
Frisch screen-grid it induced a signal in the collector which was amplified and
recorded using a Nicolet NIC-80 data processor. The sequence was then repeated,
ensemble-averaging the collector signals to enhance the signal/noise ratio.

4 Preliminary Results
Figure 2 shows a typical sequence of ion signals. The first signal is a reference,
recorded before the grid had been vibrated, and the others show how the signal
gradually recovered after the grid vibration had been completed. It is evident
that there is significant attenuation as a result of the grid oscillation – demonstat-
ing immediately that the technique described in the preceding sections enables
us both to create and to detect vorticity in the mK temperature range.

4 0 0 0

3 0 0 0

2 0 0 0

1 0 0 0

0 1 0 0 0 2 0 0 0 3 0 0 0 4 0 0 0

Fig. 2. A set of collector signals. The duration of each of them is ∼ 200 μs, and a
period of 1.5 s in real time separates each signal from its neighbour. The first signal is
for reference, recorded before the grid was vibrated.

In Fig. 3, the signal heights are plotted as a function of time for several
temperatures T . It appears that for T < 70 mK nothing changes, within ex-
perimental error: the decay mechanism is apparently temperature-independent
within this range.

5 Discussion
Our preliminary data are too scattered for us to be able to draw definite con-
clusions about the form of the decay and there is, in any case, no theoretical
 Quantized Turbulence at mK Temperatures 77

4 6 8 1 0 1 2 1 4 1 6 1 8

Fig. 3. Signal amplitudes, showing their evolutions as a function of time t, for several
temperatures.

form with which to compare. Nonetheless a plot of ln(S0 − S) against time t,

where S0 and S are repectively the unattenuated and attenuated signals, pro-
duces what is definitely a curve, showing that the decay is non-exponential. It
is also interesting to compare with the form of decay that occurs above 1 K in
the presence of normal fluid component. By integration of the Vinen equation
(2), it is straightforward to demonstate that
  −1
S0 A  κ 
ln = χ2 t + L−1i (3)
S κd 2π
where A is a constant containing the ion-vortex trapping cross-sections, Li is
the initial vortex line density at t = 0, and d is the length of the vorticity-
filled region. Thus a plot of [(ln(S0 /S)]−1 against t would be expected to yield
a straight line. It does so, within experimental error, as shown in Fig. 4.
It is interesting to speculate on the physical nature of the low temperature
T –independent decay mechanism inferred from the results of Fig. 3. The fact
that the decay is non-exponential would appear to rule out spontaneous decay
processes, e.g. where there was a constant probability per unit time that any
given element of vortex might emit a phonon and become incrementally shorter.
But the T –independence strongly suggests that the phonon gas plays no role,
given that the phonon density is falling as T 3 within the range of interest. One
possibility, perhaps, is that phonons are emitted during reconnections, leading
to line-shrinkage. Simulations by Tsubota et al [15] suggest that reconnections
are indeed the key to the problem, but that the resultant decay arises because
78 S.I. Davis et al.

1 2

1 0

0
4 6 8 1 0

Fig. 4. Plot of a typical set of data to test the applicability of equation (3). S and S0
are respectively the amplitudes of a signal and of the reference signal.

of the strongly kinked lines that are produced, resulting [3] in Kelvin waves and
a cascade of energy towards smaller and smaller length scales until it is radiated
as sound.
A difficulty in interpreting the results is that the absolute vortex line densities
are unknown, because the ion-vortex trapping cross-section is unknown under
the conditions of the experiment. We can obtain a very approximate estimate
from the measured linewith of the grid resonance which, for typical electrode
potentials and driving amplitudes implies that the energy dissipation of the grid
is (0.4±0.2)μW; insertion of this value in (2), on the assumptions that all of the
dissipation goes into vortex creation, χ takes the same value as above 1K and
that the vortex tangle remains mostly between the electrodes, yields a steady
state line density of ∼1010 m−2 . Any or all of the assumptions could be in error,
however, and the only unambiguous way to clarify the situation will be through
direct measurement of the ion-line trapping cross-section in a rotating cryostat
where the line density is known precisely.

6 Conclusions
In conclusion, we would emphasize the preliminary character of these results,
and the large number of unknowns. As pointed out above, we know nothing
about the spatial distribution of the vorticity, although we imagine that it stays
mostly between the top and bottom electrodes of the triple capacitor in Fig. 1.
 Quantized Turbulence at mK Temperatures 79

Nor do we know anything about the nature of (possible) temperature depen-

dences of the vortex-generation and ion-trapping mechanisms. Our ignorance of
vortex/ion trapping cross-section represents an even more serious lacuna, be-
cause it means that we cannot calculate absolute vortex line densities from the
ion signal attenuation.
Nonetheless, we can tentatively conclude, first, that it is possible to generate
and detect vorticity in HeII at mK temperatures and, secondly, that the decay
of quantized vorticity becomes temperature-independent below T ∼ 70 mK.

Acknowledgements

It is a pleasure to acknowledge helpful discussions with C F Barenghi, R J

Donnelly, L Skrbek and W F Vinen. The work was supported by the Engineering
and Physical Sciences Research Council (UK).

References
1. S.R. Stalp, L. Skrbek, R.J. Donnelly: Phys. Rev. Lett. 82, 4831 (1999)
2. C.F. Barenghi: J. Phys.: Condens. Matter 11, 7751 (1999)
3. W.F. Vinen: Phys. Rev. B 61, 1410 (2000)
4. R.J. Donnelly: Quantized Vortices in He II (Cambridge University Press, Cam-
bridge 1991)
5. W.F. Vinen: Proc. R. Soc. A 242, 493 (1957)
6. K.W. Schwarz: Phys. Rev. B 18, 245 (1978); 31, 5782 (1985); 38, 2398 (1988)
7. K.W. Schwarz, J.R. Rozen: Phys. Rev. Lett. 66, 1898 (1991)
8. A.L. Fetter: Phys. Rev. A 136, 1488 (1964)
9. M.I. Morell, M. Sahraoui-Tahar, P.V.E. McClintock: J. Phys. E: Sci. Instrum. 13,
350 (1980)
10. L. Meyer, F. Reif: Phys. Rev. 123, 727 (1961)
11. R.M. Bowley, P.V.E. McClintock, F.E. Moss, P.C.E. Stamp: Phys. Rev. Lett. 44
161 (1980)
12. R.M. Bowley, P.V.E. McClintock, F.E. Moss, G.G. Nancolas, P.C.E. Stamp: Phil.
Trans. R. Soc. Lond. A 307, 201 (1982)
13. D.R. Allum, P.V.E. McClintock, A. Phillips, R.M. Bowley: Phil. Trans. R. Soc.
Lond. A 284, 179 (1977)
14. R.J. Donnelly, P.H. Roberts: Proc. R. Soc. Lond. A 312, 519 (1969)
15. M. Tsubota, T. Araki, S.K. Nemirovskii, J. Low Temperature Phys. 119, 337
(2000)
Grid-Generated He II Turbulence
in a Finite Channel – Experiment

J.J. Niemela, L. Skrbek, and S.R. Stalp

Cryogenic Helium Turbulence Laboratory, Department of Physics,

University of Oregon, Eugene, OR 97403, USA

Abstract. We present experimental data on decaying turbulence, generated by tow-

ing a grid through a stationary sample of He II. We describe in detail the experimental
apparatus and physical principles that allow observation of up to six orders of magni-
tude of decaying vortex line density over three orders of magnitude in time using the
second sound attenuation technique.

1 Introduction
Superfluid turbulence has long been an area of study, with an emphasis largely on
flows created by applying a heat current in He II; i.e. on thermal counterflow[1].
It is an advantage that, experimentally, counterflow turbulence requires no mov-
ing parts for its generation. The connection of this type of flow with classical
turbulence, however, is not obvious. More recently, it has been of some interest to
explore this connection by generating turbulent flows in He II in a similar manner
as for classical fluids. In particular, turbulence created in the wake of a grid can
create nearly homogeneous and isotropic turbulence (HIT)[2], and application
of this procedure to He II has led to new insights and a few surprises[3,4,5]. In
particular, a deep similarity appears to exist between grid turbulence in classical
fluids and in He II, a quantum fluid.
In this article, we focus on the experimental apparatus and techniques used
to generate grid turbulence in He II, and discuss the observed decay of the vor-
tex line density behind a towed grid. Measurements of second sound attenuation
allow detection of up to six orders of magnitude of decaying vortex line density
L, which can be converted into roughly eight orders of magnitude of decaying
turbulent energy[6] - at present hardly a feasible goal for any laboratory experi-
ment on classical turbulence. In a companion article in this book, we will further
interpret the data in terms of classical hydrodynamics. The underlying quantum
nature of this experiment has been recently discussed by Vinen [7].

2 Experimental Setup

The schematic of the experimental apparatus is shown in Fig.1. The turbulence

is generated by towing a grid through a stationary sample of He II. We use a
65% open brass monoplanar grid of rectangular tines, 1.5 mm thick, with a mesh
size, M (tine spacing) of 0.167 cm[3,4]. The grid is attached to a central stainless

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 80–86, 2001.

c Springer-Verlag Berlin Heidelberg 2001
 Grid-Generated He II Turbulence 81

Linear Servo Motor

Motor Control
and Counter

16 bit ADC

200MHz Pentium
GPIB, Labview
G
P
I
B
Function Vacuum Tight Sliding Seal
B Generator
U Vacuum
Pump
S

Lock-in
Amplifier

Amplifier
60 mF Grid Generated
Turbulence
Grid

Second Sound
Bias Voltage Receiver
100 Volts DC
Second Sound
Transmitter

Second Sound
LR-110 Standing Wave
Resistance
Bridge
Germanium
Thermometer 1 cm x 1 cm x 29 cm
Channel
Heater
LR-130
Temperature
Controller

Fig. 1. Schematic of the experimental apparatus

steel pulling rod of diameter 0.24 cm which exits the cryostat via a pair of tight
sliding seals. The space between the seals is continually evacuated to prevent the
introduction of impurities inside the cryostat during the experiment. Above the
cryostat, the rod is attached to a computer controlled linear servo motor that
positions the grid with 0.01 cm accuracy and provides the towing velocity, vg ,
up to 2.5 m/s. This enables the exploration of a wide range of mesh Reynolds
numbers 2 × 103 ≤ ReM = vg M ρ/μ ≤ 2 × 105 , where μ is the dynamic viscosity
82 J.J. Niemela, L. Skrbek, and S.R. Stalp

of the normal fluid and ρ the total density. It is worth noting that the linear
servo motors, developed at the University of Oregon, accomplish this precise
positioning without the problems associated with electrical switching noise and
mechanical resonances that are characteristic of the more commonly used stepper
motors.
The channel of square cross section is 29 cm long and has a width d = 1 cm.
It was manufactured by an electroforming process with a tolerance of 25 μm and
a surface roughness less than 0.5 μm. The channel is suspended vertically in the
helium cryostat and during the measurement is totally submersed in superfluid
helium, which enters the channel via 16 holes, 500 μm in diameter, placed near
the top end. Except for these, the channel is leak-tight. There is no free surface of
the liquid inside the channel during the experiment, and second sound coupling
to the free surface of the bath is minimized by the use of the small diameter
holes for mass transfer between the channel and the bath.
The temperature is measured and controlled via a germanium resistance
thermometer and heater placed in the He II bath. The thermometer is calibrated
against the saturated vapor pressure with a temperature accuracy better than 1
mK and a resolution within 10 μK[4]. Details of possible temperature fluctuations
inside the channel that might occur during and just after pulling the grid are
unknown, as the thermometer is located outside the channel and its response
time is about 100 ms. For maximum cooling, the main bath is pumped via a 20
cm diameter line connected to a 300 CFM vacuum pump and a 1400 CFM roots
blower. This enables bath temperatures down to 1 K to be reached easily. It
is difficult, however, to assure temperature stability below about 1.1 K, so this
marks the lowest reliable temperature for the experiment. Experiments close to
the lambda temperature (above 2 K) require special care (see below) mainly due
to the strong temperature dependence of the second sound velocity, u2 .
To probe the quantized vortex line density resulting from pulling the grid,
we excite and detect second sound using vibrating nuclepore membranes 9 mm
in diameter mounted flush on opposing walls of the channel. These 6 μm thick
polycarbonate membranes have a dense distribution of 0.1 μm holes and on
one side is evaporated an approximately 95 Å thick layer of gold, which is then
pressed against the channel wall. This gold layer forms one electrode of a ca-
pacitor transducer, with the other being a brass electrode pressed towards the
opposite side of the membrane from outside the channel. Applying an ac signal
(∼ 0.3 − 1 VP P ) in addition to a dc bias (∼ 100 V) results in an oscillatory
motion of the membrane. In He II, the normal fluid is clamped by viscosity in-
side the small holes of the membrane, while the superfluid component passes
freely, thereby exciting second sound, an entropy wave in He II. Directly across
the channel the second sound wave produces a corresponding oscillation of the
other membrane (i.e., the receiver) and the induced signal is input to a lock-in
amplifier referenced to the transmitter frequency. The channel acts as a second
sound resonator. The excitation amplitude is adjusted to be the upper half of
the linear response range and the n-th harmonic- with n about 50- of the funda-
mental frequency is used: typically 30 − 40 kHz. A Lorentzian resonance peak is
 Grid-Generated He II Turbulence 83

obtained, having a full width at half maximum that is temperature dependent

and typically reaches values of Δ0 = 20 − 500 Hz without quantized vortices in
the channel.
To understand the use of second sound in detecting the vortex line density
and its relation to averaged rms vorticity, we consider the seminal work of Hall
and Vinen[8]. In experiments with a rotating container of He II, they observed
an excess attenuation of second sound in a direction perpendicular to the rota-
tion axis due to the presence of quantized vortices, αL = BΩ/2u2 . This extra
attenuation resulted from the scattering of the elementary excitations - normal
fluid - by the vortex lines and was absent for second sound propagating parallel
to the rotation axis. Here B is the dimensionless mutual friction parameter (B
generally depends both on temperature and frequency[9]) and Ω denotes the
angular velocity of rotation. It is now well known that the rotating bucket of He
II displays, on average, the same shape of the surface meniscus as any classical
fluid, since the superfluid mimics solid body rotation by creation of a lattice of
rectilinear quantized vortices aligned in the direction of the rotation axis. In this
case, the vorticity ω = 2Ω = κL, where κ is the circulation quantum (κ = h/m4 ,
where h is Planck’s constant and m4 the mass of the helium atom) and L is the
total length of the vortex line per unit volume. It is often assumed that this rela-
tion between ω and L holds in general. By considering a second sound resonance
as an infinite series of reflected waves in a rotating cavity, the extra attenuation
(in a limit of small attenuation) due to quantized vortices becomes [4]
 
BκL πΔ0 A0
αL = = −1 (1)
4u2 u2 A
where A and A0 are the amplitudes of the second sound standing wave resonance
with and without vortices present, respectively. We can extend this formula
to the case of a homogeneous vortex tangle, taking into account that vortices
oriented parallel to the second sound propagation do not contribute to the excess
attenuation. Then we have[3]
 
16Δ0 A0
L= −1 (2)
Bκ A

It can be shown[4] that for arbitrary attenuation one has to use the more
general formula
 
8u2 1 + p2 P + 2p2 P + p4 P 2
L= ln √ (3)
πBκd 1 + P + 2P + P 2

where p = A0 /A and P = 1 − cos(2πdΔ0 /u2 ) that for small dΔ0 /u2 reduces
to (2). It is essential to use formula (3), as using the approximate formula (2)
at some experimental conditions leads to results that are more than an order of
magnitude off!
It is important to consider the time response of the measuring system with
regard to the finite velocity of second sound. Our calculations show that in most
84 J.J. Niemela, L. Skrbek, and S.R. Stalp

1 0
6
T = 1 .7 5 K

) 5
-2

1 0 p o w e r -3 /2
v o r te x lin e d e n s ity ( c m

4
1 0

3
1 0

2
1 0

1
1 0
0 .1 1 1 0 1 0 0
tim e ( s )
Fig. 2. The log-log plot of the decaying vortex line density versus time after grid passes
2 mm above the measuring volume. Each decay curve represents an average of three
identical pulls. The decay curves, in order, correspond to ReM = 2×105 (the uppermost
one), 1.5 × 105 , 105 , 5 × 104 , 2.5 × 104 , 104 , 5 × 103 and 2 × 103 . For each ReM , the
decaying vortex density displays an inertial range with power law exponent -3/2.

cases (except close to the lambda temperature) there is negligible error intro-
duced into the deduced vortex line density, for the following reasons. Immediately
after the grid is towed through the measuring volume the quality factor is very
low, of order unity, and the detecting system can be described rather as a second
sound pulse technique with the characteristic time response given by the time
of flight d/u2 ∼ 10−3 s, where u2 ∼ 20 m/s [10]. As the turbulence decays, the
characteristic time constant increases with the (temperature dependent) qual-
ity factor. Without the vortex tangle, the typical linewidth of the second sound
resonance is 20-500 Hz, the typical frequency used is 30-40 kHz, so the quality
factor reaches 60-2000 and the time response gradually rises to about 0.1-1 sec
at the very end of the decay, where it constitutes an error of less than 1%.
There are also other time restrictions than the finite velocity of second sound.
As it takes a time τg ∼= d/vg to tow the grid through measuring volume, we use
only that data obtained on the time scale longer than τg and also exceeding 8τLI ,
where τLI is the time constant of the lock-in amplifier used for detection of the
amplitude of the second sound signal. Another time scale restriction involves
consideration of how soon the flow can be assumed as nearly HIT [6]. It was
discussed in[11] that a physical criterion to estimate this time leads to about
1-2 turnover times of the largest eddies present in the flow, i.e., those of the size
of the channel. It follows that the minimum time needed to assume nearly HIT
conditions is about 1-2 widths downstream from the grid, or using the Taylor
 Grid-Generated He II Turbulence 85
-4
2 .0 x 1 0

-4
1 .5 x 1 0
/s )
2
ν ' (c m

-4
1 .0 x 1 0

-5
5 .0 x 1 0
1 .1 1 .2 1 .3 1 .4 1 .5 1 .6 1 .7 1 .8 1 .9 2 .0 2 .1 2 .2
T e m p e ra tu re (K )
Fig. 3. The effective kinematic viscosity of the superfluid turbulence ν  deduced from
vorticity decay data at various temperatures, and assuming a value of the Kolmogorov
constant consistent with classical experiments. The dashed line represents ν = η/ρ
which has a dissimilar temperature dependence indicative of quantization effects, where
η is the normal fluid viscosity and ρ is the total density of helium II.

frozen hypothesis, the time it takes the towed grid to pass 1-2 widths of the
channel.
The data acquisition process can be briefly described as follows. The cryostat
is filled with liquid helium, the bath being pumped and the temperature con-
trolled to a desired value. With the grid “parked” towards the top of the channel,
the second sound resonance curve is measured and fitted to a Lorentzian, giving
the value of Δ0 . The grid is then lowered to the bottom and after a necessary
waiting time (typically 2 minutes) pulled through the channel such that its veloc-
ity is constant and equal vg for at least 5 cm below and above the experimental
volume. It is then slowed down and smoothly ”parked” against the top again.
The data acquisition is triggered when the grid passes a predetermined position
2 mm above the measuring volume, as determined by a signal from a photodiode.
Typically 100 s of data are recorded at a rate of 100 Hz from the output signal
of the lock-in amplifier, representing the recovering second sound standing wave
amplitude, A(t). After another waiting period to ensure that the turbulence has
decayed down to a negligibly low level, the reference amplitude A0 is read. The
decaying vortex line density is then calculated using formula (3). It implicitly
assumes that there is no additional attenuation of the second sound that might
be caused by a turbulence in the normal fluid.
Since the second sound is transmitted and detected via membranes on oppos-
ing sides of the channel, we obtain information from a measuring volume of order
d3 ∼
= 1 cm3 . It is this natural integration that bypasses tedious statistical analysis
involved in any local velocity measurements in conventional turbulence, provides
86 J.J. Niemela, L. Skrbek, and S.R. Stalp

enormous sensitivity and unprecedented dynamical range of the method, making

it very useful and complementary to classical turbulence studies.
Examples of decay data and the effective kinematic viscosity of the superfluid
turbulence ν  , are shown in Figs. 2 and 3 respectively. The decay curves represent
up to six orders of magnitude of decaying vortex line density over three decades
of time. The overall form of the decay does not change with temperature. At any
temperature, after some initial period depending on ReM , the decay curves tend
to collapse and display the power law decay L ∝ t−3/2 [5,12]. We note if vorticity
and L are assumed to be related in a similar manner as discussed above for the
rotating bucket experiments, then this feature and others are understandable in
terms of a classical spectral decay model[6,11,12], further discussed in a compan-
ion article in this book. Quantum effects show up in the measured temperature
dependence of the effective kinematic viscosity (see Fig. 3). Finally, we have had
many useful discussions at this workshop concerning complementary measure-
ments to probe the normal fluid turbulence and/or energy dissipation directly,
and also extending the protocol to near zero temperature to complement the
work of McClintock’s group reported here. These ideas are presently receiving
attention.

Acknowledgements

The towed grid He II experiment has been developed at the University of Oregon
over many years. We would like to acknowledge R.J. Donnelly, W.F. Vinen and
M.R. Smith for their valuable contributions to the conception and design of this
experiment. This research was supported by NSF under grant DMR-9529609.

References
1. R.J. Donnelly: Quantized vortices in helium II. Cambridge University Press (1991)
2. G. Comte-Bellot, S. Corrsin: J. Fluid Mech. 25, 657 (1966); 48, 273 (1971)
3. M.R. Smith: Evolution and propagation of turbulence in helium II. PhD Thesis,
University of Oregon, Eugene (1992)
4. S.R. Stalp: Decay of grid turbulence in superfluid helium. PhD Thesis, University
of Oregon, Eugene (1998)
5. M.R. Smith, R.J. Donnelly, N. Goldenfeld, W.F. Vinen: Phys. Rev. Lett. 71, 2583
(1993)
6. L. Skrbek, J.J. Niemela, R.J. Donnelly: Phys. Rev. Lett. 85, 2973, (2000)
7. W.F. Vinen: Phys. Rev. B 61, 1410 (2000)
8. [Link], W.F. Vinen: Proc. Roy. Soc. London A238, 204 (1954); 238,215(1954)
9. C.F. Barenghi, R.J. Donnelly, W.F. Vinen: J. Low Temp. Phys. 52, 189 (1983)
10. R. J. Donnelly, C. F. Barenghi: [Link]. Chem. Data 27, 1217 (1998).
11. L. Skrbek, S.R. Stalp: Phys. Fluids 12, 1997 (2000)
12. S.R. Stalp,L. Skrbek, R.J. Donnelly: Phys. Rev. Lett. 82, 4831(1999)
Intermittent Switching Between Turbulent
and Potential Flow Around a Sphere
in He II at mK Temperatures

Michael Niemetz, Hubert Kerscher, and Wilfried Schoepe

Institut für Experimentelle und Angewandte Physik, Universität Regensburg,

D-93040 Regensburg, Germany

Abstract. Intermittent switching between potential flow and turbulence is observed

with an oscillating sphere in HeII below 0.5 K, where there is no normal fluid component
and no viscosity. The remaining dilute phonon gas is in the ballistic regime and therefore
turbulence in the pure superfluid can be investigated. The amplitude of the driven
oscillations is a measure of the damping which in case of potential flow is due to
residual ballistic phonon scattering or, when the flow is turbulent, is due to a large
nonlinear turbulent drag. In an intermediate range of driving forces the flow is observed
to be unstable, intermittently switching between both patterns. We have investigated
this phenomenon down to 25 mK and have made a statistical analysis of the time series
measured at various constant driving forces and temperatures. We obtain a temperature
independent probability density for switching, different for both directions. We find a
regime of metastable laminar flow whose lifetime is limited by natural radioactivity
and cosmic rays.

1 Experiment
The experimental setup used in our investigations consists of a ferromagnetic mi-
crosphere (radius r = 124 μ m, m = 27 μ g) suspended between two niobium elec-
trodes by superconducting levitation. While cooling the horizontally arranged
electrodes forming a parallel plate capacitor (distance d = 1 mm, diameter 2 mm)
a dc-voltage of several hundred volts is applied to the capacitor giving rise to
an electric charge at the surface of the electrodes and the sphere. As the elec-
trodes become superconducting, the magnetic sphere is repelled and levitates at
an equilibrium position between the two electrodes. Horizontal stability is pro-
vided by trapped flux in the electrodes. As the sphere carries an electric charge
q vertical oscillations (f ≈ 150 Hz) can be excited by applying a resonant ac
electric field to the capacitor corresponding to a force F = q · Uac /d on the
sphere. The oscillations can be detected by measuring the current I = v · q/d
induced in the electrodes of the capacitor by the moving charge. The space be-
tween the superconducting electrodes is filled with pure 4 He (3 He concentration
≤1 ppb). This setup provides a simple geometry without disturbance by mechan-
ical suspension elements and a low background dissipation (Q-factors above 106
are achieved when the cell is evacuated). We measure the velocity amplitude of
the oscillating sphere for different driving forces at different temperatures. For
a more detailed description of the experimental technique and its applications,
see [1–4].

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 87–94, 2001.

c Springer-Verlag Berlin Heidelberg 2001
88 M. Niemetz, H. Kerscher, and W. Schoepe

2 8 3 5
3 0

2 5

2 6 2 0
s ta b le 1 5
la m in a r 1 0

2 4 flo w 5
0
v (m m /s )

in te r m itte n t 0 2 0 0 4 0 0 6 0 0 8 0 0

s w itc h in g
2 2

v c 2
2 0
v c 1
F c
s ta b le tu r b u le n c e
1 8
0 5 0 1 0 0 1 5 0
T = 3 0 0 m K F (p N )

Fig. 1. Three regimes of flow around the sphere: For low driving forces there is stable
laminar flow (•), resulting in a linear dependence v(F ) = F/λ. For large driving forces
there is stable turbulent flow () resulting in large dissipation and low oscillation
amplitudes. (The solid line is a fit of the turbulent v(F ) dependence derived in the text.)
Between these two regions there is a regime where both flow patterns are unstable and
which extends from vc1 over an interval Fc of driving forces to vc2 where turbulence
becomes stable. In this regime the system switches between laminar and turbulent
flow intermittently. The inset shows a larger range of driving forces and velocities. The
enlarged region is marked by a box

2 Results
The inset of Fig. 1 shows a typical result of the experiment. For small driving
forces the flow around the sphere remains laminar corresponding to a linear drag
force λv, which is given by ballistic scattering of phonons (left regime in Fig. 1).
For large driving forces there is stable turbulent flow instead, accompanied by
large nonlinear dissipation. Between these two regimes we find an interval of
driving forces, where neither laminar nor turbulent flow is stable but where
instead the fluid switches between these states intermittently.

2.1 Stable Turbulent Flow

Fully developed turbulence in a classical fluid causes a drag force
|Fd | = γv 2 , γ = Cd πr2 /2 (1)
(Cd ≈ 0.4) on a sphere moving through the fluid. As we perform an ac experiment
where v(t) = v sin(2πf t) we employ the energy balance between the energy input
per half period T /2 at resonance and the dissipation


T /2 
T /2

F sin(ωt) v sin(ωt) dt = Fd (v(t)) v sin(ωt) dt (2)

0 0
 Intermittent Switching Between Turbulent and Potential Flow 89

1 0 0
Fig. 2. Turbulent flow around
8 0 the sphere: Data points taken
up to one hundred times larger
v (m m /s )

6 0 driving forces than in Fig. 1

are well described by the tur-
4 0 bulent drag force introduced
in the text. The shift of the
2 0 T = 1 0 0 m K apex of the parabola towards
negative F values can clearly
0 be observed
0 2 4 6 8 1 0 1 2
F (n N )

to find the corresponding F (v) dependence F (v) = 8γv 2 /3π. In Fig. 2 a larger
range of driving forces is shown, providing a more complete view of the turbulent
regime. The solid line is a fit, using a similar drag force as in (1) but reduced by
a constant value:
|Fd | = γ(v 2 − v02 ) = γv 2 − F0 . (3)
By applying the energy balance (2) this drag force corresponds to
 
8γ 3 2
F (v) = v − v0 = γ  (v 2 − vc1
2 2
) (4)
3π 2

which perfectly fits the data points obtained in the superfluid (Fig. 2). A similar
drag force has been calculated for turbulent flow of a two dimensional dilute
Bose-Einstein condensate around a cylinder [5]. Even if these calculations are
not directly applicable to liquid helium, there might be a similar mechanism in
both cases for the reduced drag force. In our case turbulent drag exists only
above a critical velocity vc1 = 19.4 mm/s (s. Fig. 1).

2.2 Intermittent Switching

In contrast to classical fluids there is no smooth transition from laminar flow to
fully developed turbulence in the superfluid, as reported earlier [2]. Instead, the
fluid switches between laminar and turbulent flow intermittently. We recorded
the velocity amplitude of the oscillations as a function of time while holding tem-
perature and driving force constant. Figure 3 shows three sections of 13 minutes
out of 4 h time series obtained for different driving forces at 300 mK. At the
lowest driving force (series (a)) we observe quite long laminar phases, where
the velocity amplitude exponentially approaches the equilibrium value given by
phonon drag and driving force. The laminar phases are interrupted by turbulent
phases intermittently and the velocity amplitude drops to the lower value vt
given by the large nonlinear turbulent drag. When using a larger driving force
(series (b)) we find a steeper increase of the velocity amplitude corresponding
to a higher equilibrium value that is not reached any more, as the lifetimes of
90 M. Niemetz, H. Kerscher, and W. Schoepe

3 0 a ) 4 7 p N
2 8
2 6
2 4
2 2
2 0
1 8

3 0 b ) 5 5 p N
2 8
v (m m /s )

2 6
2 4
2 2
2 0
1 8

3 0 c ) 7 5 p N
2 8
2 6
2 4
2 2
2 0
1 8

tim e ( 1 0 0 s / d iv is io n )
Fig. 3. Random sections of 800 s out of a 4 h time series of the oscillation velocity
amplitude for three different drives at 300 mK. Series (a) was taken at the lowest
driving force, showing the velocity increase during laminar phases and its exponential
saturation at a level given by phonon drag and driving force. Laminar phases are
interrupted intermittently by turbulent phases accompanied by a sharp drop of the
oscillation velocity amplitude to an equilibrium value given by the large turbulent drag
and driving force. Series (b) was taken using a larger driving force, resulting in a higher
saturation value (that is not reached any more) and slightly longer turbulent phases.
At the largest drive (series (c)) there is turbulent flow most of the time, interrupted
by laminar phases

the laminar phases are shorter than in series a. Taking a closer look at the life-
times of the turbulent phases shows that they in turn are longer than in series
a. Increasing the driving force further results in turbulent flow most of the time,
interrupted by short laminar phases. As the switching between both flow pat-
terns is intermittent, the results are discussed in terms of a statistical analysis
of the velocity amplitudes reached during laminar phases and their lifetimes as
well as the lifetimes of the turbulent phases. We apply reliability theory [6] for
the evaluation of the time series.

2.3 Turbulent Phases

A typical distribution of lifetimes of turbulent phases is shown in Fig. 4a, where

the number P of turbulent phases exceeding a certain lifetime t is plotted versus
t. The data points are described very well by a straight line in the semilogarithmic
plot, corresponding to exponentially distributed lifetimes exp(−t/μ). The slope
of the straight line gives the mean lifetime μ. We have performed this analysis for
many time series at different temperatures ranging from 28 mK up to 400 mK and
 Intermittent Switching Between Turbulent and Potential Flow 91

1 0 0 0
a ) b )
1 0 0
n u m b e r o f tu r b u le n t p h a s e s

1 0 0

μ (s )
1 0 1 0 3 2 m K
1 0 0 m K
T = 3 0 0 m K 2 0 0 m K
F = 5 9 p N 3 0 0 m K
4 0 3 m K
1
1
0 1 0 2 0 3 0 4 0 5 0 2 0 3 0 4 0 5 0
t (s ) F -λ v t (p N )

Fig. 4. Analysis of the lifetimes of the turbulent phases. (a) The number of turbu-
lent phases living longer than a certain time t shows an exponential decay (solid line)
corresponding to exponentially distributed lifetimes. The reciprocal slope of the distri-
bution in the plot gives the mean lifetime μ. (b) Mean lifetimes of turbulent phases for
different driving forces and temperatures. There is no temperature dependence, but a
strong dependence on the driving force. The solid line is a fit of a fourth-power law
divergence of the mean lifetime μ at a critical driving force value

in a wide range of driving forces and have always found exponentially distributed
lifetimes. A comparison of the mean lifetimes obtained is shown in Fig. 4b. The
values of μ are independent of temperature and collapse to an universal drive
dependence if the strongly temperature dependent laminar drag λv (where v = vt
is the velocity amplitude of the turbulent phase) is subtracted from the external
driving force. The mean lifetime of turbulent phases increases with the driving
force and diverges at a critical value Fc = 54 pN approximately with a fourth-
power law and stays infinite at larger drives corresponding to turbulent velocities
above vc2 , see Fig. 1. The power dissipated at Fc is 0.6 pW, corresponding to the
production of ≈ 1 mm vortex lines per half period. This is equivalent to producing
a vortex ring with a diameter of ≈ 1.4 times the diameter of the sphere.

2.4 Laminar Phases

A similar analysis can be performed for the laminar phases, but now it is conve-
nient to analyze first the amplitudes reached during laminar phases. Figure 5a
shows the number of laminar phases which exceed a given amplitude Δ v = v −vt .
The data are perfectly described by a parabola in the semilogarithmic plot,
P (Δ v) = P (0) exp −(Δ v/vw )2 , which means that the probability density func-
tion of the amplitudes corresponds to a Weibull distribution [6]. This result
holds for all driving forces and temperatures and the fitting parameter vw is
constant (Fig. 5b). The results for a given experiment fit well into a 10% band-
width around a mean value. In order to extract the probability of breakdown
of the metastable laminar flow it is necessary to analyze the lifetimes t of the
laminar phases, i.e. P (t) = P (Δ v(t)). The failure rate Λ(t) (i.e. the probability
92 M. Niemetz, H. Kerscher, and W. Schoepe

7
2 8 m K
a ) b ) 1 0 0 m K 2 0 0 m K
3 0 0 m K 4 0 3 m K
1 0 0 6
n u m b e r o f la m in a r p h a s e s

+ 1 0 %

(m m /s )
5
4 .7 9
1 0

w
v
-1 0 %
4
T = 3 0 0 m K
F = 5 5 p N

1 3
0 2 4 6 8 1 0 1 2 0 1 0 2 0 3 0 4 0 5 0
Δ v (m m /s ) F -λ v t (p N )

Fig. 5. Analysis of velocity amplitudes reached during laminar phases. (a) The number
of laminar phases having an amplitude larger than Δ v shows a quadratic dependence
on Δ v in this semilogarithmic plot (solid line), corresponding to a Weibull probability
density of velocities. (b) Fitting parameter vw for different driving forces and temper-
atures. We find no systematic dependence on these parameters. The values vary in a
10% bandwidth around a mean value of 4.8 mm/s

of breakdown after the laminar phase has survived for a time t) is given by [6]
 2
d d Δ v(t) 2
Λ(t) = − ln P (Δ v(t)) = = 2 Δ v Δ v̇ . (5)
dt dt vw vw
Because the velocity amplitude is exponentially approaching an equilibrium
value with a time constant τ = 2m/λ, this result has several implications: First,
as Δ v̇ is constant for small values of Δ v (or t  τ ), the failure rate is increasing
with the velocity amplitude of the sphere. Second, with the velocity reaching the
equilibrium value (or t  τ ), Δ v̇ goes to zero, and so does Λ(t). This means that
the laminar phase, having survived for many τ and having reached its stationary
velocity amplitude, will live forever although the velocity is clearly above vc1 . Ex-
perimentally, however, this is not exactly true. This can be seen in Fig. 6, where
the probability distribution for large lifetimes is shown. The decrease slows con-
siderably down for lifetimes greater than approximately 100 s (corresponding to
3τ at 300 mK), but the distribution does not approach a constant value. Instead,
a constant failure rate is found, leading to a mean lifetime of 25 min (indicated by
the straight line in Fig. 6). Obviously, there must be another mechanism causing
the breakdown of those long lived laminar phases. We can exclude mechanical
vibrations or acoustic noise to be the origin, as we tried to destroy such long
laminar phases by slamming the door, jumping on the floor, or even refilling
helium into the cryostat. But placing a small radioactive source (60 Co, 74 kBq)
outside the dewar, had a dramatic effect on the lifetimes. As can be seen from
Fig. 6, the mean lifetime changes by a factor 8.3 from 25 minutes to 3.0 minutes.
We have measured the dose rate of the source at the position of the measuring
cell inside the cryostat (taking into account a measured 20% loss in the dewar
 Intermittent Switching Between Turbulent and Potential Flow 93

walls) to be 440 nGy/h (± 5%). Comparing this value with a measured dose rate
due to natural background radiation in our laboratory of 50 nGy/h (± 10%),
which is typical for our area, we obtain an increase of the dose rate due to the
source by a factor of (440 + 50)/50 = 9.8 This compares well with the ratio of
lifetimes of metastable laminar phases obatined in the experiment. Therefore, it
is obvious that natural background radioactivity limits the lifetime of metastable
laminar phases above the critical velocity vc1 . This effect may be attributed to
local vorticity generated by ions produced by radiation, inducing the breakdown
of the metastable laminar flow around the sphere.

3 Conclusion
The turbulent flow of the superfluid causes a drag force on the sphere which
is very similar to turbulence in classical fluids except for a constant offset in
analogy with a dilute Bose-Einstein condensate. Furthermore, we have found
that the transition from potential flow to turbulent flow around a sphere in
superfluid 4 He at mK temperatures occurs by intermittent switching between
both flow patterns instead of the gradual transition observed in viscous fluids. A
statistical analysis of this switching phenomenon has shown that the lifetimes of
turbulent phases diverge at a critical driving force. The probablity for breakdown
of the laminar phases has been obtained. Finally, there exist metastable laminar
phases above the critical velocity vc1 . Their lifetime is limited only by natural
background radioactivity [7].
We have repeated our experiment several times by heating the measuring cell
above Tc of niobium in order to prepare a new levitating state of the sphere. All
our observations were reproducible. The three quantities vc1 , vw and Fc which
we expect to be affected by the properties of the sphere have standard deviations
of 8%, 13% and 15%, respectively, probably due to asymmetries of the surface
of the sphere.
We understand now why above 0.5 K the intermittent switching changes into
the hysteretic behavior observed earlier [1,2]: the velocity increases Δ v become

Fig. 6. Influence of radioac-

1
d r iv in g fo r c e = 4 7 p N tivity on the lifetime of lam-
T = 3 0 0 m K inar phases. By adding a ra-
n o r m a liz e d n u m b e r o f la m in a r p h a s e s

dioactive source the mean life-

time for long living lami-
nar phases (t τ = 31 s
n a tu r a l b a c k g r o u n d r a d ia tio n
at 300 mK) is reduced from
0 ,1 25 minutes to 3.0 minutes.
The mean lifetimes were ob-
tained by analyzing the ap-
6 0
a d d itio n a l C o s o u rc e proximately exponential tail
at large lifetimes (straight
0 2 0 0 4 0 0 6 0 0 8 0 0 1 0 0 0 1 2 0 0 lines)
t (s )
94 M. Niemetz, H. Kerscher, and W. Schoepe

very small at higher phonon drag which implies a very low failure rate of the
laminar phase. But if it fails (i.e. when the driving force F is largely increased)
the following turbulent phase is stable because the critical drive is exceeded.
With our present results in superfluid 4 He it appears extremely promising to
extend these experiments to superfluid 3 He in order to investigate the transition
to turbulence in this very different quantum fluid.

References
1. J. Jäger, B. Schuderer, and W. Schoepe: Physical Review Letters 74, 566 (1995).
2. J. Jäger, B. Schuderer, and W. Schoepe: Physica B 210, 201 (1995)
3. P. Eizinger, W. Schoepe, K. Gloos, J.T. Simola, and J.T. Tuoriniemi: Physica B
178, 340 (1992)
4. M. Niemetz, W. Schoepe, J.T. Simola, and J.T. Tuoriniemi: Physica B 280, 559
(2000)
5. T. Winiecki, J.F. McCann, and C.S. Adams: Physical Review Letters 82, 5186
(1999)
6. B.V. Gnedenko, Yu.K. Belayev, and A.D. Solovyev: Mathematical Methods of
Reliability Theory (Academic Press, New York, 1969)
7. M. Niemetz, H. Kerscher, and W. Schoepe: [Link]
mat/0009299 and to be published
Vortex Filament Methods for Superfluids

David C. Samuels

Dept. of Mathematics, Univ. of Newcastle, Newcastle upon Tyne, NE1 7RU, UK

1 Introduction
Vortex filaments are an idealized form of rotational flow where the vorticity
is confined to a small core region, of radius a, around a one dimensional line
embedded in the three dimensional flow. Outside of this core region the flow is
potential. When the dynamics of the core size are not important these objects
may also be referred to as vortex lines.
In classical fluid mechanics, by which I mean solutions of the Navier-Stokes
or Euler equations, vortex filaments are a useful tool for understanding the
geometry and dynamics of a flow. But after an initial popularity in the early
1980’s [1] [2] [3] [4] [5], the use of vortex filament methods fell out of favor
in classical fluid mechanics. Though there has been some slight resurgence in
this method recently [6] [7] due to the rapidly increasing computational power
available and the development of new computational algorithms, direct numerical
simulations and large eddy simulations have become the methods of choice for
calculating the motion of fluids. One reason for the decreased use of vortex
filament methods is that while they give a clear and intuitive understanding of
a flow through the easy visualization of the vortex filaments, this representation
was often just a rough cartoon of the true flow. Vortex filaments are only a
convenient idealization in a classical flow. The vorticity in a realistic classical
flow rarely takes the form of clearly discrete vorticity filaments.
But in superfluids like helium II vortex filaments are real [8]. Due to the
quantization of circulation, vorticity in a helium II flow can only exist within
vortex filaments with a core size of a. Since this core size is very small in helium
II, about 1 Angstrom, the thin-core vortex filament idealization is actually a
very accurate description of the true superfluid flow.
There are a few special qualities of the superfluid vortex filament [8] that
make it even simpler than the standard model of a vortex filament in a classi-
cal fluid. The circulation around each superfluid vortex filament in helium II is
set by quantum mechanics to be an integral multiple of κ = 9.97x10−4 cm2 /sec.
Since an n quantum vortex filament contains more energy than n single quan-
tum vortices, it is generally assumed that only single quantum vortex filaments
will be commonly observed. Thus all helium II superfluid vortices have identical
circulation. The core size a is also determined by quantum mechanics and is
closely related to the concept of a healing length, the length scale required for a
wave function to change from its bulk value to zero. In the classical fluid vortex
filament the core size is an important variable and much of the complications,

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 97–113, 2001.

c Springer-Verlag Berlin Heidelberg 2001
98 D.C. Samuels

and the interesting dynamics, of classical vortex filaments are due to changes in
the core size along a filament [9]. Without a variable core size, the dynamics of
a quantized vortex filament are much simpler than those of its classical counter-
part. Of particular importance is the effect of the constant core size on vortex
stretching. When a classical vortex filament is stretched, the core size shrinks
and the local vorticity rises along the filament. When a quantum vortex filament
is stretched, it can only elongate with no change in core size. Due to this dif-
ference, many prefer to reserve the phrase “vortex stretching” for the classical
process, and will say that vortex stretching does not occur for individual quan-
tized vortex lines. One should note, however, that a bundle of parallel quantized
vortex filaments can undergo vortex stretching, when the core size is interpreted
as the radial size of the bundle of filaments.
Many of the complications of classical vortex filament calculations have to
do with vortex stretching and with other core dynamics which are assumed
negligible in helium II quantized vortex filaments. We thus have a much simpler
system to deal with, both in terms of physical behaviour and in terms of the
computation of the motion of the filaments. My discussion in this paper will be
concerned directly with the superfluid vortex filaments of helium II, and when
necessary I will take the values of constants to be those of helium II.

2 Vortex Filament Motion

We represent the superfluid vortex filament by a series of mesh points distributed
along the centerline of the vortex filament. The motion of the vortex filaments
is calculated as the motion of each of these mesh points.
Since the vortex core size is on the atomic scale, it is usually assumed that
the effective mass of the vortex is negligible and the motion of the vortex can
be found by setting the sum of the forces at each point on the vortex filament
to be zero and solving the resulting equation for the velocity of the vortex line
[8]. From these procedure we get the basic equation of motion for the superfluid
vortex filament,

dS/dt = V S +V I +αS  ⊗(V N −V S −V I )−α S  ⊗[S  ⊗(V N −V S −V I )] (1)

where S is a position of a point on the vortex line, V S is the local superfluid

flow due to all non-vorticity sources, V I is the superfluid velocity due to the
presence of the vortex filaments (the induced velocity). S  is the derivative of
S with respect to arclength along the vortex line. V N is the local normal fluid
velocity and α and α are temperature dependent mutual friction parameters. It
is common to claim that α > α and then neglect the last term in the equation
of motion. However, this inequality is not true for temperatures very near the
lambda transition and there should also be the more general concern that since
these α and α terms in (1) are at right angles to each other it does not strictly
matter that one term may be of smaller magnitude than the other. The α term
produces a qualitatively different type of motion of the vortex line than does the
α term, and it should therefore not be discarded in an offhand manner.
 Vortex Filament Methods for Superfluids 99

The two fluid model presents the motion of helium II as a superposition

of two fluids, the normal fluid and the superfluid. The vortex filaments that
concern us here are vorticity structures of the superfluid. They interact with
the normal fluid through the mutual friction force, and the motion of these
vortex filaments can be strongly affected by the flow in the normal fluid. The
reverse is also true; the presence and motion of the superfluid vortex filaments
can strongly affect the flow of the normal fluid. This coupled interaction between
the two fluids is the subject of many of the articles in this volume. Calculating
the fully coupled motion of the entire two-fluid system is a very complicated
problem (which is discussed more fully in the article by Idowu et al). While
work on the fully coupled calculations is just beginning, most superfluid vortex
calculations up to this time have only been concerned with the motion of the
superfluid vortices without considering the response of the normal fluid flow
to the mutual friction force. This kinematic approach requires us to define a
normal fluid flow, with spatial and possibly even time dependence, and use (1) to
calculate the motion of the vortex line due to its self induced motion, advection
in the superfluid potential flow VS and the response to mutual friction with
the normal fluid flow VN . This is the problem that we will consider in this
article. The vortex filament techniques discussed here are also needed for the
fully coupled calculations described elsewhere in this volume.

2.1 The Biot–Savart Law and the Local Induction Approximation

The first difficulties in the vortex filament method come from the definition of
the term V I in the equation of motion. This term represents the advection of
the point on the vortex filament by the velocity field due to all the superfluid
vortex filaments present in the flow. This advection velocity at the point r is
given by the Biot-Savart law.

κ (S − r) ⊗ dS
V I (r) = (2)
4π |S − r|3

The line integral is taken over the entire length of vortex filament present in
the fluid. The Biot-Savart law can be used to find the flow at any point r in
the fluid, but the immediate difficulty with this equation comes from its use
in calculating the advection velocity in (1). In this case, the point r lies on
the vortex line and the Biot-Savart law contains a singularity at that point.
To heal this singularity we must include some aspect of the core structure of
the vortex filament. Many different methods of de-singularizing the Biot-Savart
law have been used in classical fluid dynamics [10] [11] [12] [13], where the core
dynamics of the filaments can often be quite complicated and important, but in
superfluid vortex methods I have only seen one method used [14]. This method
breaks the line integral into two parts; an integral over a local neighborhood
around the point r (this contains the singularity), and the rest of the Biot-Savart
integral (this part is non-singular). Then the local integral must be replaced by
an algebraic approximation that takes into account the core size of the filament.
100 D.C. Samuels

It is convenient to split up the Biot-Savart integral using the local meshing of

the vortex line. In calculating the motion of the J’th mesh point on the filament,
the local section of the integral is taken to be section containing the mesh points
J − 1, J, and J + 1. By calculating the Biot-Savart integral over a circular vortex
ring excluding the section of line between the mesh point J − 1 and J + 1, and
then subtracting this value from an analytic expression for the velocity of an
ideal vortex ring, the remainder can be identified as the Biot-Savart integral
over the section of the vortex line from J − 1 to J + 1. Thus this section of the
Biot-Savart integral (containing the singularity) is replaced by the term
  
κ   2 l+ l −
V local = (S ⊗ S ) ln . (3)
4π aef f

S  is the unit tangent vector to the vortex line at the J’th mesh point and S  ,
the second derivative of the line with respect to arclength, is the local curvature
vector. The length scales l+ and l− are distances between the J’th mesh point
and the J + 1 and J − 1 points respectively. The parameter aef f is an effective
core size since numerical constants of order one have been absorbed into this
parameter. This choice for the local part of the Biot-Savart integral gives the
correct velocity for a planar vortex ring. Defining the velocity of a planar vortex
ring requires that you specify some details about the core structure of the vortex
(hollow core vs solid body rotation, for example). Different choices of the core
structure will alter the log term in (3), but unless you are calculating flows on
very small length scale, on the order of the core size, then these details make
negligible changes to the value of (3).
Once you prevent the singularity by splitting the Biot-Savart law into two
sections you are still left with the difficulty of calculating the integral over the
rest of the vortex line. In a simulation, we only know the positions of the discrete
mesh points along the line, so the mesh points must be joined in some numerically
convenienent manner to calculate the integral between the mesh points. Any
interpolation method through the mesh points will do the job. I prefer to use
the simplest method of integrating over piece-wise linear vortex line segments
between the mesh points. The velocity at a point r (which may be a mesh point
on the vortex line or may be any point in the volume of the fluid) due to the
Biot-Savart integral over the segment of vortex line between the J and J + 1
mesh points is
 SJ+1
κ (S − r) ⊗ dS
V segment (r) = . (4)
4π SJ |S − r|3
Taking this integral over a straight line between the two mesh points gives

κ (|RJ | + |RJ+1 |)(RJ ⊗ RJ+1 )

V segment (r) = (5)
4π |RJ ||RJ+1 |(|RJ ||RJ+1 | + RJ · RJ+1 )

where RJ = SJ − r and RJ+1 = SJ+1 − r. Other interpolation methods will

have more complicated results for the Biot-Savart integral.
 Vortex Filament Methods for Superfluids 101

One of the reasons that vortex filament methods have fallen out of favor in
classical fluid dynamics is the fact that the calculation of the Biot-Savart integral
requires order N 2 operations over a mesh of N points. The calculation of this
integral is computational expensive whenever a large number of mesh points are
needed to represent the vortex lines. This calculation is often not practical, even
with the speed of today’s computers, if the amount of vortex line is large, or if
the fine detail of the vortex line shape needs to be captured using a very fine
meshing. And these are often exactly the cases that are of the most interest to
us.
One way around this difficulty is the Local Induction Approximation, (LIA)
[15]. In this method we keep the local term in the induced motion (3) and neglect
completely the non-local Biot-Savart integral. This is typically done with some
minor adjustments to the log term so that we have
κ  2R 
V I,LIA = (S  ⊗ S  ) ln , (6)
4π aef f
where R is a length scale of the filament. R may be taken as a constant, such as
the length scale of the computational box, or it may be taken as the local radius
of curvature of the vortex line. In the first case, the log term is a constant, and
it is often absorbed into a non-dimensional timescale. In the second case, the log
term will vary along the filament and with time, but unless you are capturing a
very wide range of curvatures with your mesh, or the curvatures are only about
a factor of 10 larger than the effective core size, the log term is nearly constant.
LIA is a very convenient approximation. It is simple and easy to calculate.
The time required to calculate (6) increases only linearly with the number of
mesh points. The interpretation of (6) is simple: the induced velocity is in the
direction of the local binormal S  ⊗ S  and is inversely proportional to the
local radius of curvature. It is commonly used in analytic investigations of the
properties of vortex filaments. And LIA correctly describes the motion calculated
by the Biot-Savart law of simple vortex line geometries such as the planar vortex
ring and low amplitude vortex waves. But it is a very severe approximation to
the true equation of motion!
LIA clearly works very well for calculating the motion of single vortex lines
which do not loop around so that sections of the filament far apart in arclength
are actually close to each other. Also it probably works well in a random vortex
tangle, where the non-local part of the Biot-Savart law may tend to cancel out
when integrated over the random vortex lines. LIA will not work well for vortex
configurations which tend to have parallel, or antiparallel, sections of vortex
filament near one another (for example, vortex knots [16]). LIA will not work well
for flows which tend to develop any alignment of vorticity. Unfortunately, aligned
superfluid vorticity does tend to be formed by flows at non-zero temperatures
where the normal fluid has some local vorticity structure (the superfluid vortices
tend to align with the normal fluid vorticity [17]). In the case of turbulent flows, I
personally doubt that the LIA captures enough of the physics of the interactions
between the vortex filaments to give much useful information on these flows,
though others would debate this point.
102 D.C. Samuels

Some of the qualitative differences between the motion of vortices under LIA
and the full Biot-Savart law should be pointed out. A superfluid flow calculated
by LIA can never develop any rotational flow, aside from the rotation around a
single filament. In an LIA calculation, the energy of the flow is just an energy
per unit length of the vortex lines and when there is no mutual friction (at zero
temperature) the length of vortex line does not change with time (without further
assumptions to the model, such as phonon emission). Under the Biot-Savart law,
the kinetic energy in the flow can be quite difficult to calculate, and even at zero
temperature the length of vortex line can change while the kinetic energy remains
constant. In a LIA calculation, the physical effect of vortex stretching cannot
occur while it can occur in a Biot-Savart calculation whenever there are even
just two approximately parallel vortex lines.
With the vortex filaments represented by N mesh points, the vortex equation
of motion (1) becomes N coupled, first order ordinary differential equations. This
coupled set of ODEs can be solved by any general method. I prefer to use a fifth
order Runge-Kutta method with an adaptive stepsize. An adaptive method like
this calculates the result of one time step in two different ways (one fourth order
method and one fifth order method) and compares the results at each of the N
mesh points to estimate the error. An allowed error range is defined and if the
estimated error lies below this allowed error range for all N points, the time step
is increased, while if the estimated error is above the allowed range for any of the
N mesh points, the time step is decreased and that timestep is recalculated. This
is a sturdy method, capable of automatically handling the rapid changes in vortex
line velocities which can occur, often as a result of vortex line reconnections. It is
not a very efficient method however. Whenever any region of very high curvature
(or a close approach of two vortex filaments under the Biot-Savart law) occurs
then the timestep for the entire vortex tangle can drop significantly. Some type of
‘multi time step’ method, where the motion of fast moving sections of the vortex
line could be calculated using more time steps than the slow moving sections of
the vortex, would would be far more efficient, particularly when the equation of
motion is as expensive to evaluate as is the Biot-Savart law.

2.2 Boundary Conditions

There are three basic types of boundary conditions used in vortex filament cal-
culations. In order of increasing complexity these are: an infinite fluid, periodic
boundary conditions, and solid boundaries (no penetration of the superfluid
through the boundary). In an infinite fluid calculation no boundary conditions
are required in the calculations, as long as you set up the initial conditions as
closed vortex loops. Many flow quantities, such as the total kinetic energy, have
simple line integral definitions that only work in the infinite fluid simulations,
where the vorticity is zero at spatial infinity. This is a good choice of bound-
ary conditions for studying simple closed loop vortex structures, such as vortex
knots, but it is not a good choice for complicated vortex tangles.
Periodic boundary conditions work well for vortex tangles since they allow a
homogeneous flow to develop and allow simple definitions of averaged quantities
 Vortex Filament Methods for Superfluids 103

such as the vortex line density. Periodic boundary conditions are easily pro-
grammed for LIA simulations, though care must be taken at the places where
neighboring mesh points along a filament lie on opposite sides of your periodic
boundary. But these problems are easily handled. The main difficulty with peri-
odic boundary conditions comes from the Biot-Savart calculations. In principle,
periodic BCs make the Biot-Savart integral infinitely long. In practice, this inte-
gral must be cut off at some arbitrary point. The costs of periodic BC in Biot-
Savart calculations can be staggering. For a three dimensional periodic cube,
including the first layer of periodic vortices around the central cubic volume will
increase the number of vortex mesh points by a factor of 27. Including even this
level of periodicity is usually prohibitively costly. To deal with problem I will
sometimes define a buffer layer around the central cubic volumn, with a width
of 1/2 to 1/4 of the width of the computational volume, and I will only include
in the Biot-Savart calculation the periodic vortex filaments that fall within this
fairly thin buffer zone. This type of method actually only ensures that vortex
filements moving through one side of the periodic volume and re-entering on the
other side, experience a more smoothly varying velocity field during the transi-
tion. Ideally, this type of problem could be removed by an analytic method to
calculate the velocity field due to this infinitely repeating vorticity distribution.
While this has been done in two dimensions, I know of no such method in 3D.
In LIA, solid boundary conditions are quite easy to implement [14]. In this
case, for any vortex filaments which end on the solid surface we must set the
normal vector at the ends of the vortex filament equal to the local normal vector
of the solid surface. This is the only condition needed no matter how complex the
boundary geometry. In Biot-Savart calculations we must use image vortices for all
the vortex filaments in the flow. For a single flat boundary, these image vortices
are simply vortices in the mirror image positions, and with their orientation
reversed. For a single flat surface this doubles the length of the Biot-Savart
calculations, which is not a terribly onerous increase. But for multiple solid
surfaces, you must then include the images of images, ad infinitum (just as
you must do in electrostatic calculations). This naturally introduces the same
computational difficulties as in the periodic BC case, and it must be solved in
the same manner, either by cutting off the image calculations at some arbitrary
number of images or by developing an analytic summation method to collapse
the infinite images to a reasonable calculation. I know of no one who has done
the latter for a 3D vortex flow, though the tools may exist in the electrostatics
literature for us to use.

2.3 Meshing of the Filaments

Since the CPU time cost of a vortex simulation rises rapidly as the number of
mesh points increases it is important to manage this number carefully. As the
vortex filaments are locally compressed or elongated the spacing between the
mesh points will change with time. New mesh points will need to be inserted on
some sections on the filament to keep a certain mesh resolution, and to keep the
number of mesh points at a minimum some mesh points will need to be removed
104 D.C. Samuels

from other sections of the filament as they move too close to one another (close
in terms of arclength). The criteria that you choose for remeshing depends on
the behaviour that you are trying to capture with a specific simulation.
For this discussion, let the arclength distance between two neighboring mesh
points, J and J + 1 be represented by δJ,J+1 . The simplest remeshing criteria is
to attempt to keep an approximately uniform distance between the mesh points.
In this method, the mesh point J would be removed if δJ−1,J + δJ,J+1 < δ and a
mesh point would be added between the J and J + 1 points if δJ,J+1 > 2δ. This
will keep all the distances along the mesh roughly in the range δ/2 to 2δ. This
is the most straightforward meshing method, but it suffers from the limitation
that it can only resolve structures on the vortex filament of scale δ or larger.
This method is best if you are only interested in the vortex behaviour at one
particular length scale.
If you want to calculate the development of vortex filament structures over
a wide range of length scales then you must use a more complicated remeshing
method. We would like to have more mesh points in the parts of the vortex
filament with high curvature and fewer in regions of low curvature. This can be
done by interpreting the meshing length scale δ as a variable, proportional to the
local radius of curvature. Typically, I will take δ = 2πR/32, where R is the local
radius of curvature of the filament. This will mesh a planar ring, at any length
scale, with approxamately 32 mesh points. Of course, there must be a limit on
the range of this meshing. This is done by setting a range of R over which this
remeshing will be used, with minimum and maximum mesh lengths to be used
outside this range. With this method it is possible for the mesh to represent a
range of two or three orders of magnitude in radius of curvature without having
an unreasonably large number of mesh points. It may be possible to extend
this range by another order of magnitude by setting the meshing length δ to
decrease slower than linearly with R. Techniques like this are used in 2D vortex
simulations in the meshing of the surface of vorticity blobs [18].
However you choose to remesh, removing mesh points poses no problems but
in adding mesh points you must make a choice on the position of the new mesh
point. Once again, the simplest choice is to use a piece-wise linear interpolation,
and simply place the new point at the center of the straight line between the J
and J + 1 point. But if you do this, then you are likely to change the first and
second derivatives, S  and S  at the points J and J +1. Worse than this, the new
point would be introduced with zero local curvature, almost certainly causing a
rapid (and artificial) change in curvature along the filament. For adding mesh
points a better interpolation must be used, and almost any interpolation will
perform well as long as it is not piece-wise linear. I use a method that inserts
new mesh points at a position where the curvature vector, S  , at the new mesh
point is the average of the curvature vectors at the two neighboring points, J
and J + 1. This prevents the intruduction of sudden jumps in the vortex line
curvature, errors which would radiate vortex waves and affect more than just
the local segment of the vortex line.
 Vortex Filament Methods for Superfluids 105

For ease of explanation in this paper I have been labelling neighboring mesh
points with consecutive number, J and J + 1 for example. Any remeshing natu-
rally will destroy this order, and consecutive numbering is not practical. Instead,
a data structure should be used that has for each mesh point, the labels of the
two neighboring mesh points. Then a remeshing involves only the addition or
deletion of a mesh point and the necessary changes in this data structure for the
neighboring mesh points.
One further remeshing routine should be used. Occasionally, very small vortex
rings may develop. These may be formed by the decay through mutual friction
of larger vortex loops. Or they may be generated by the pinching off of a small
loop in a reconnection event. However they are formed, these small loops should
be removed from the calculation. By definition, the curvature of these loops will
be very large, and thus their motion will be very fast. With an adaptive stepsize
technique, the code will automatically drop the time step to a very small value
to attempt to accurately calculate the rapid motion of these small loops. If you
are not using an adaptive stepsize, then the motion of these small loops will not
be accurately calculated, possibly leading to the sudden and artificial expansion
of the small loop. In either method these small loops must be removed from the
calculation. It is a good idea to keep a record at least of the length of vortex line
removed from the simulation by this routine.

3 Reconnections of Filaments
When two classical Navier-Stokes vortex filaments cross, viscous effects will re-
connect the vortex filaments. Lacking these viscous effects, the crossing of two
classical Euler vortex filaments is believed to lead to a singularity and the break-
down of the Euler equations [19]. This singularity formation is not observed in
calculations of the crossing of superfluid vortex filaments, where vortex recon-
nection occurs without the need for viscosity [20]. In this way, superfluid vortices
are clearly not Euler vortex filaments. The difference lies in the core structure
of the superfluid vortices. At the centerline of each vortex filament, quantum
mechanics requires the density of the superfluid (given by the amplitude of the
ground state wave function) to go to zero and the vorticity of the flow is singular
along this centerline.
Studies of superfluid vortex filament reconnections must be done with a quan-
tum theory calculating the evolution of the wavefunction of the superfluid ground
state. A more detailed discussion of these calculations can be found in the articles
by Roberts and by Adams in this volume. These studies have traditionally been
done using a Non-Linear Schrodinger Equation (NLSE). It is recognized that
the NLSE is probably a poor representation of helium II superfluid because the
dispersion curve calculated from the NLSE contains only phonons and no rotons.
The NLSE does, however, seem to be a very good representation of the new Bose
condensed alkali atom superfluids. Without a fundamental microscopic theory
of helium II superfluidity, we must consider the results of NLSE calculations,
keeping in mind the possible errors of these calculations. The NLSE calculations
106 D.C. Samuels

clearly show that when two superfluid vortex lines cross, a reconnection event
occurs and the topology of the vortex lines changes [20]. This behavior must be
included as a basic assumption in the superfluid vortex filament model since it
is not a consequence of the equation of motion. This discussion will be made in
terms of the crossing to vortex filaments in the bulk of the fluid. In simulations
with boundaries, reconnections of filaments with the boundaries must also be
considered, but these reconnections may be easily detected as vortex segments
approach and eventually hit the boundaries so we need not discuss them further
here.
The reconnection assumption leaves us with the problem of detecting the
crossings of 1-D lines moving through 3-D space, when we only know the posi-
tions of a finite number of mesh points along these lines at discrete time intervals.
Though the assumption can be quite simply stated, it has proved very difficult
to develop any satisfactory algorithms to implement this. The first reconnec-
tion algorithm, due to Schwarz [21], was a simple, intuitive one: vortex lines are
reconnected whenever two mesh points come within a pre-defined distance Δ
of each other. While this algorithm is very simple to use, the objections to it
are many and strong. This algorithm will obviously trigger a reconnection event
early, since a true reconnection event does not occur until the distance between
the lines is on the order of a core radius (practically zero in filament simulations).
Sometimes, a large section of vortex filament, including many mesh points, will
move within the reconnection distance triggering a series of reconnection events
resulting in the formation of a number of small vortex loops where only one
crossing event would actually have occured. Even worse, in some cases this al-
gorithm will trigger a reconnection where no crossing of the vortex lines would
actually occur. This error happens when vortex lines pass by each other, without
intersection but within the distance Δ. This may at first seem to be a rare and
thus negligible event, but it is not. We now know that vorticity concentrations
in the normal fluid generate bundles of well-aligned superfluid vortex filaments
at the center of the normal fluid rotation [17]. These superfluid vortex filaments
may be closer to each other than the defined reconnection distance Δ but they
cross only rarely. With the Δ reconnection algorithm, these vorticity bundles
will reconnect wildly.
The Δ reconnection model also introduces a new and completely artificial
length scale to the simulation; the reconnection length scale Δ. If you set Δ too
large, you will have many spurious reconnections. If you set it too low, you will
miss many real reconnections, as vortex lines jump past one another in the finite
time step, plus you will still have some spurious reconnections of nearly aligned
vortices. Some of the problems of this algorithm can be helped (though not
eliminated) by making the reconnection length scale variable along the vortex
filament and setting it equal to the local mesh size (a natural choice). This
refinement to the algorithm still suffers from the separate problems of premature
and also spurious reconnection.
A different reconnection detection algorithm, the one that I currently use,
is based on the idea that a vortex filament reconnection event is a dissipative
 Vortex Filament Methods for Superfluids 107

event, even in a superfluid. The reconnection is assumed to involve the emission

of phonons and possibly rotons, thus converting superfluid energy to heat. With
this in mind, we assume that reconnections can only occur if they are accom-
panied by a loss of energy in the superfluid. Identifying vortex line energy with
vortex line length (an exact proportionality in LIA, but only approximate for
the Biot-Savart law), we reconnect the vortex lines when the reconnection will
directly lower the line length. The reconnections that occur in this algorithm
still happen when the filaments are slightly less than a local mesh length apart,
so it still suffers from the premature reconnection problem. But the generation
of additional small vortex loops in a reconnection rarely occurs in this algorithm
and spurious reconnections from near misses and between aligned vortices are
almost eliminated.
It is worth taking care with the reconnection algorithm. Differences in the
implementation of reconnections [22] [23] [24] were initially blamed for inconsis-
tancies between the vortex tangle simulations of Schwarz and Buttke in the late
80’s, though the more serious error of the introduction of an unphysical ’mixing
step’ by Schwarz was eventually determined to be the prime cause of the dis-
agreement [25]. We must admit that the reconnection algorithms in use today
still leave much to be desired and can be improved. A reconnection is a major
event, drastically changing the vortex filament motion, curvature distribution,
and topology. The close approach and eventual crossing of two vortex filaments
create superfluid flow structures on the full range of length scales, down to the
core size scale. And the motion of a large section of the vortex filament can
be altered by a reconnection as the initially localized, but extreme, change in
curvature radiates out from the reconnection point as vortex waves. It is still
worthwhile to carry out detailed investigations of individual vortex reconnec-
tions [26] (see Lipniacki’s contribution to this volume) in order to develop better
reconnection models for use in both the filament simulations and in analytic
theories of the averaged quantities in superfluid turbulence.

4 Analysis of the Superfluid Flow

Once you have developed and tested your vortex filament code you are faced
with the analysis problem common to all 3-D fluid dynamics simulations: how to
make sense of all that mass of data. Though the problem is the same, the solution
must be different for vortex filaments due to the very different nature of the data.
Many quantities that can be calculated easily from the 3-D velocity fields of a
classical fluid dynamics simulation are actually very difficult to calculate with
vortex filament methods. One example of this is kinetic energy. Conversely, there
are at least a few things (such as flow topology) that are easier to calculate in a
vortex filament simulation than in a standard fluid dynamics simulation.
The simplest measurable quantity in a vortex filament simulation is the total
length of vortex line. In a finite
 volume V this is measurement gives you directly
the vortex line density L = ( dξ)/V , the length of vortex line per unit volume
where the integral is taken over the arclength ξ of the vortex filaments. This is
108 D.C. Samuels

the fundamental averaged quantity of a vortex tangle and many of the analytic
model of superfluid vortex turbulence have been based on this quantity alone.
These models are differential equations for the evolution of the line density as a
function of the average normal fluid velocity and some temperature dependent
parameters. The most basic of these models is the Vinen equation [27]

dL χ2 κ 2
= χ1 αVns L3/2 − L , (7)
dt 2π
where χ1 and χ2 are temperature dependent parameters and Vns is the average
relative velocity between the normal fluid and the superfluid. The standard in-
terpretation of the Vinen equation is that the first term represents the growth of
L due to mutual friction and the second term represents the decay of the tangle
due to reconnection events, though other interpretations of these terms are have
been made [28].
Now let us define some measure of the isotropy of the tangle. Very little work
has been done on measures of isotropy in vortex tangles. One set of measures
defined by Schwarz [28] defines three quantities which measure the isotropy of a
tangle relative to two perpendicular unit vectors, r̂ and r̂⊥ .

1
I = [1 − (S  · r̂ )2 ]dξ (8)
VL

1
I⊥ = [1 − (S  · r̂⊥ )2 ]dξ (9)
VL

1
I = (S  ⊗ S  ) · r̂ dξ (10)
V L3/2
In flows with a unidirectional normal fluid flow the unit vectors r̂ and r̂⊥ are
usually taken to be parallel and perpendicular to the direction of the normal
fluid flow. The interpretation of these three isotropy measures is discussed in
[28]. Other isotropy measures can be defined, such as the length of line in a
projection along a vector r̂
 
1
J(r̂) = 1 − (S  · r̂)2 dξ. (11)
VL

It is possible that more refined versions of (7) could be defined using these
isotropy measures as well as L.
The local rate of extension (or compression if negative) of the vortex line is
defined by
r(ξ) = −S  · VL (12)
where VL is the vortex line velocity at the arclength position ξ. Integrating
r(ξ) over the arclength gives the instantaneous rate of change of the vortex line
density. 
dL 1
= rdξ (13)
dt V
 Vortex Filament Methods for Superfluids 109

Simulations show that even in a steady state vortex tangle the line length density
has strong fluctuations, so one must be careful in interpreting instantaneous
measures such as (13).
It is tempting to define a whole range of quantities averaged over the vortex
tangle. One that must be mentioned is the averaged curvature

1
C= |S  |dξ, (14)


where  = dξ is the total vortex filament length. This quantity can be used
to define an average length scale of the tangle, 1/C, and an average velocity
scale κC. Following the evolution of this quantity with time can detect if the
vortices are ‘crinkling’ (increasing C) or smoothing (decreasing C). But I urge
against an over-reliance on averaged quantaties. It is easy enough to measure the
full distribution of values of quantities such as the curvature in the tangle, thus
measuring the full range of behavior of the vortices and not just some average
value. Again, little has been done with this approach.
But what about the more traditional fluid mechanics measures such as kinetic
energy and linear and angular impulse? While these are quite easily determined
in a simulation where the velocity is defined on a grid extending throughout the
fluid, they are actually quite difficult to measure in a vortex filament simulation.
Purely for diagnostic purposes, one can define a 3D grid and calculating the
velocity on the grid by the Biot-Savart law (2) and then such quantities can be
easily measured. But it is important when doing this to check the convergence
of your results with grid resolution. The 1/r velocity field of the vortex lines can
make resolution of the energy and momentum difficult. It is possible to avoid
this problem by using definitions of these quantities in terms of the line integrals
over the vortex filaments [29]. Some useful integrals are: kinetic energy

T = κρs V s · (S ⊗ dS), (15)

linear impulse 
1
I= κρs S ⊗ dS, (16)
2
angular impulse 
1
A = κρs S ⊗ S ⊗ dS, (17)
3
and helicity 
J =κ V s · dS. (18)

These integrals are limited to flows with no vorticity at infinity, and the integrals
must be taken over closed vortex loops.
While simple things such as kinetic energy can become quite difficult to mea-
sure in vortex filament methods, there are some qualities of the flow which are
simpler to define and measure in filament methods. There has been much re-
cent interest in the topology of vorticity in Euler flows, interest which is mainly
110 D.C. Samuels

generated by the possibility of producing singularities in these flows in finite

time [19]. Remembering that the vorticity in a classical Euler fluid may be dis-
tributed thoughout space (not just as vortex filaments!), the topology of the
vorticity field can be difficult to define, or picture, easily. But since the vorticity
of the superfluid is confined to discrete vortex filaments, vortex topology is a
natural and simple quality in a superfluid [16]. In an Euler flow, the topology
of the vorticity is conserved as the flow develops because filament reconnections
are not allowed, and the conservation of topology can lead to strong constraints
on the development of the flow [30]. But since superfluid vortices can reconnect,
the topology of a superfluid vortex is not conserved, and may change with every
reconnection. Despite this complication, it is possible that by determining the
topology (knotted and linked loops) of a vortex tangle we can develop a mea-
sure of the complexity of a tangle, a quality that is missing from all averaged
measurements on the vortex tangle.

5 Alternative Approaches
The computational cost of calculating the full Biot-Savart law is high, due to
its non-locality. The usefulness of local induction approximation calculations is
suspect because of its complete locality. Is there not a middle way that contains
some, but not all, of the non-locality of the Biot-Savart integral? The recent
development of fast multipole methods seems to fill this need [31] [6]. Originally
developed for simulations of large numbers of gravitationally interacting parti-
cles, these methods can lower the computational time of pairwise interactions
between particles from order N 2 to order N log(N ). In these methods the inter-
acting particles are grouped into clusters, and then into clusters of clusters, et
cetera, in a hierarchical tree structure. The interactions between the clusters are
calculated by a finite multipole series expansion, greatly decreasing the number
of calculations needed while retaining the primary non-local effects. The general-
ization from the original gravitational interaction to a vortex motion calculation
is straightforward and these new methods have fueled much of the recent resur-
gence (however slight) in vortex methods for classical fluid dynamics. These
methods work best in cases where the fluid vorticity has some strong spatial
structure. The drawback to these methods is the high overhead cost associated
with the determination of the hierarchical tree data structure, and the necessity
of rebuilding this data structure as the vortices move, and groups of vortices
break up or coalesce. For a moderate number of vortex mesh nodes, up to a few
thousand in my experience, the Biot-Savart calculations are still faster than the
fast multipole methods, though this limitation may become outdated soon as
new implementation algorithms are being developed. For simulations af actual
superfluid turbulence, where we must reproduce a wide range of spatial scales in
the superfluid vorticity (requiring a large number of mesh points) and include
the non-local vorticity interactions of turbulence (ruling out the LIA approach)
we must use some non-local approximate method in our simulations, and the
fast multipole methods are currently our best hope.
 Vortex Filament Methods for Superfluids 111

Yet another approach is to drop the idea of vortex filaments completely and
use a locally averaged set of equations for the superfluid vorticity field, ωs . These
approaches use the Hall-Vinen-Bakharevich-Khalatnikov (HVBK) equations (or
modifications of these equations) for the coupled evolution of continuous vorticity
fields in the superfluid and the normal fluid. Since this approach does not use
vortex filaments it lies somewhat out of the scope of this article, but is dealt
with in the articles by Henderson and Holme in this volume.

6 Conclusions: What Needs to Be Done

The primary research challenge before us today is to develop a detailed under-
standing of the interaction between the normal fluid and the superfluid [32][33],
particularly in turbulent flows. In terms of simulations, this presents us with the
double task of calculating the turbulent flow of the normal fluid (including the
forcing on the normal fluid due to mutual friction, see the paper by Idowu et
al, this volume), and simultaneously calculating the motion of a large length of
superfluid vortex lines. The simulation method must allow for the possibility of
the formation of organized structures in the superfluid vorticity, so non-local vor-
tex interactions must be included. The solution of the full superfluid turbulence
problem will be difficult and will take some time, both CPU time and time to
develop an understanding of the problem. But there is much to be learned from
simpler studies, short of the full turbulence solution. In kinematic simulations
(simulations where the normal fluid velocity is given and only the superfluid mo-
tion is calculated) the most complex normal fluid flow that has been used so far is
an ABC flow [34]. We could learn many things from kinematic simulations with
more complicated normal fluid flow structures. We should take a hint from the
closely related field of magnetohydrodynamics and investigate the development
of superfluid vorticity driven by a turbulence model to represent the normal fluid
velocity. These types of studies would give us valuable information on the reac-
tion of the superfluid to complex normal fluid flows. Does the superfluid vortex
tangle lack structure (as it does under a uniform VN ) or do structures develop
in the superfluid vorticity? If so, are there any relationships between the flow
structures in the normal fluid flow and those developing in the superfluid? What
are the scaling laws for kinetic energy in the turbulent superfluid? How do the
velocity correlation functions in the superfluid compare to those in classical tur-
bulence? We could have at least preliminary and approximate answers to these
questions from kinematic simulations. In fully coupled simulations (solving for
the motion of both VN and the superfluid vorticity) we must investigate the
possible changes in the normal fluid turbulence itself due to the mutual friction
forcing from the superfluid. This problem bears a resemblance to turbulence
with polymer additives, where the Navier-Stokes turbulence is clearly affected
by the 1-D forcing from the polymers.
A not-entirely separate issue is the problem of superfluid turbulence at zero
temperature [35] (see the articles by Tsubota and McClintock, this volume). In
studies of this type we need a firmer understanding of the loss of energy by
112 D.C. Samuels

phonon and roton emission from reconnecting vortices [36] and possibly even
from just the motion of vortices. Does this ‘zero temperature dissipation’ clearly
affect the properties of the vortex tangle? Is it enough to produce an inertial
range in the pure superfluid turbulence? Can we develop an understanding of
the energy cascade in a pure superfluid vortex tangle? I say that this is a not en-
tirely separate issue from that of the fully coupled helium II turbulence because
we need to understand the behaviour of the fully coupled turbulence in both the
high and low temperature limits. In the high temperature limit, helium turbu-
lence must surely be classical Navier-Stokes turbulence after the lambda tran-
sition temperature is passed and some experiments show no detectable change
in the measured turbulence quantities as this transition temperature is passed.
Pure superfluid turbulence is the low temperature limit of fully coupled helium
II turbulence and there is no reason not to expect this transition to also be
very smooth. Helium II turbulence experiments show little, if any, temperature
dependence down to approximately 1.4 Kelvin [37], where the normal fluid rel-
ative density and the mutual friction interaction both become quite small. An
understanding of both classical Navier-Stokes turbulence and pure superfluid
turbulence will be helpful in developing our understanding of coupled helium II
turbulence.

References
1. A. J. Chorin: SIAM J. Sci, Stat. Comput. 1, 1 (1980)
2. A. Leonard: J. Comput. Phys. 37 289, (1980)
3. B. Couet, O. Buneman, A. Leonard: J. Comput. Phys. 39 305, (1981)
4. A. Leonard: Ann. Rev. Fluid Mech. 17 523 (1985)
5. W. T. Ashurst, E. Meiburg: J. Fluid Mech. 189 87 (1988)
6. G-H. Cottet, P. D. Koumoutsakos: Vortex Methods (Cambridge University Press,
Cambridge, 2000)
7. R. Cortez: J. Comput. Phys. 160 385 (2000)
8. R. J. Donnelly: Quantized Vortices in Helium II (Cambridge University Press,
Cambridge, 1991)
9. V. M. Fernandez, N. J. Zabusky, V. M. Gryanik: J. Fluid. Mech. 299 289 (1995)
10. T. Y. Hou: Math. of Comp. 58 103 (1992)
11. R. Klein, O. Knio: J. Fluid Mech. 284 275 (1994)
12. M. F. Lough: Phys. Fluids 6 1745 (1994)
13. R. Klein, L. Ting: Appl. Math. Lett. 8 45 (1995)
14. K. W. Schwarz: Phys. Rev. B 31 5782 (1985)
15. R. L. Ricca: Fluid Dyn. Res. 18 245 (1996)
16. R. L. Ricca, D. C. Samuels, C. F. Barenghi: J. Fluid Mech. 391 29 (1999)
17. D. C. Samuels: Phys. Rev. B 47 1106 (1993)
18. D. G. Dritschel, N. J. Zabusky: Phys. of Fluids 8 1252 (1996)
19. R. B. Pelz: Phys. Rev. E 55 1617 (1997)
20. J. Koplick, H. Levine: Phys. Rev. Lett. 71 1375 (1993)
21. K. W. Schwarz: Phys. Rev. Lett. 49 283 (1982)
22. T. F. Buttke: Phys. Rev. Lett. 59 2117 (1987)
23. K. W. Schwarz: Phys. Rev. Lett. 59 2118 (1987)
 Vortex Filament Methods for Superfluids 113

24. T. F. Buttke: J. Comp. Phys. 76 301 (1988)

25. K. W. Schwarz: J. Comp. Phys. 87 237 (1990)
26. T. Lipniacki: Eur. J. Mech. B - Fluids 19 361 (2000)
27. W. F. Vinen: Proc. Roy. Soc. Lond. A 242 493 (1957)
28. K. W. Schwarz: Phys. Rev. B 38 2398 (1988)
29. P. G. Saffman: Vortex Dynamics (Cambridge University Press, Cambridge, 1992)
30. H. K. Moffatt, A. Tsinobar: Topological Fluid Mechanics (Cambridge University
Press, Cambridge, 1990)
31. L. Greengard, V. Rokhlin: J. Comp. Phys. 73 325 (1987)
32. D. Kivotides, C. F. Barenghi, D. C. Samuels: Science 290 777 (2000)
33. O. C. Idowu, A. Willis, C. F. Barenghi, D. C. Samuels: Phys. Rev. B 62 3409
(2000)
34. C. F. Barenghi, D. C. Samuels, G. H. Bauer, R. J. Donnelly: Phys. Fluids 9 2631
(1997)
35. B. V. Svistunov: Phys. Rev. B 52 3647 (1995)
36. M. Leadbeater, T. Winiecki, D. C. Samuels, C. F. Barenghi, C. S. Adams: Phys.
Rev. Lett. 86 1410 (2001)
37. J. Maurer, P. Tabeling: Europhys. Lett. 43 29 (1998)
Introduction to HVBK Dynamics

Darryl D. Holm

Theoretical Division and Center for Nonlinear Studies, Los Alamos National
Laboratory, MS B284, Los Alamos, NM 87545

Abstract. We review the Hall-Vinen-Bekarevich-Khalatnikov (HVBK) equations for

superfluid Helium turbulence and discuss their implications for recent measurements
of superfluid turbulence decay.
A new Hamiltonian formulation of these equations renormalizes the vortex line
velocity to incorporate finite temperature effects. These effects also renormalize the
coupling constant in the mutual friction force between the superfluid and normal fluid
components by a factor of ρs /ρ (the superfluid mass fraction) but they leave the vortex
line tension unaffected. Thus, the original HVBK form is recovered at zero tempera-
ture and its mutual friction coefficients are renormalized at nonzero temperature. The
HVBK equations keep their form and no new parameters are added. However, a tem-
perature dependent trade-off does arise between the mutual friction coupling and the
vortex line tension.
The renormalized HVBK equations obtained via this new Hamiltonian approach
imply a dynamical equation for the space-integrated vortex tangle length, which is the
quantity measured by second sound attenuation experiments in superfluid turbulence.
A Taylor-Proudman theorem also emerges for the superfluid vortices that shows the
steady vortex line velocity becomes columnar under rapid rotation.

1 HVBK Equations
Recent experiments establish the Hall-Vinen-Bekarevich-Khalatnikov (HVBK)
equations as a leading model for describing superfluid Helium turbulence. See
Nemirovskii and Fiszdon [1995] and Donnelly [1999] for authoritative reviews.
See Henderson and Barenghi [2000] for a recent fluid mechanics study of steady
cylindrical Couette flow using computer simulations of the incompressible HVBK
equations.
In the Galilean frame of the normal fluid with velocity vn , the HVBK equa-
tions may be expressed as follows, upon ignoring thermal diffusivity and viscosity,
∂t ρ = − divJ , ∂t Ji = − ∂k Tik , ∂t S = − div(Svn ) + R/T ,
 1 
ρs ∂t vs + ρs (vs · ∇)vs = − ρs ∇ μ − |vs − vn |2 + ρs f , (1)
2
and summing over pair of upper and lower repeated indices. One may consult,
e.g., Bekarevich and Khalatnikov (BK) [1961] and Donnelly [1999] to compare
these equations with the form they take in the reference frame of the superfluid.1

1
In making this comparison it is useful to recall the Galilean transformation of the
chemical potential, μ  = μ − 12 |vs − vn |2 , where μ  is evaluated in the superfluid

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 114–130, 2001.

c Springer-Verlag Berlin Heidelberg 2001
 Introduction to HVBK Dynamics 115

Notation. Here ρ and ρs denote the total and superfluid mass densities, respec-
tively. The entropy density of the normal fluid is S, its temperature is denoted
T , and ρn = ρ − ρs denotes its mass density. The superfluid velocity is denoted
vs and
J = ρs vs + ρn vn
is the total momentum density. In the entropy equation R is the rate that heat is
produced by the phenomenological friction and reactive forces in f , which must
be specified to close the theory. We also denote

Stress tensor: Tik = ρs vs i vsk + ρn vn i vnk + (P + λ·ω) δik − λi ω k ,

Euler’s pressure law: P = − ε0 + T S + ρμ ,
Superfluid First Law: dε0 = μ dρ + T dS + (J − ρvn )·d(vs − vn ) + λ·dω . (2)

In the superfluid First Law, ω = curl vs is the superfluid vorticity with magnitude
|ω| = ω̂ ·ω, where ω̂ = ω/|ω| is its unit vector. BK [1961] takes the energy density
ε0 to depend on the magnitude of the superfluid vorticity, |ω|, as

ρs κ |ω| R
ε0 = ln .
4π a
This is the energy per unit length of a superfluid vortex line, ρs (κ2 /4π)ln(R/a),
with quantum of circulation κ = h/m 10−3 (cm2 /sec) and ratio R/a of mean
distance between vortices R to effective vortex radius a, times the vortex length
per unit volume, |ω|/κ. Hence, we find

∂ε0 ∂|ω| κ R
λ= = |λ| ω̂ with |λ|/ρs ≈ ln = λ0 ,
∂|ω| ∂ω 4π a

where one ignores the derivative of R κ/|ω| inside the logarithm. The ap-
pearance of λ in the stress tensor Tik shifts the pressure P , and div T introduces
an additional force − ω · ∇λ ≡ −ρs T into the motion equation. The quantity T
is called the “vortex line tension.”
BK [1961] assigned the following form to the phenomenological coupling force
f appearing in the superfluid velocity equation in (1),

f = (vL − vs ) × ω , vL = vn − ρs (α s0 + β ω̂ × s0 ) , (5)
where s0 = v0 − vn , v0 = vs + ρ−1
s curl λ , (6)

with “vortex velocity” vL and “slip velocity” s0 introduced as auxiliary quanti-

ties. The HVBK equations written as (1) in the normal-fluid frame conserve the
energy,  
1 
E= ρ|vn |2 + (J − ρvn )·vn + ε0 d 3 x . (7)
2
frame and μ in the normal-fluid frame. See Putterman [1974] for a clear discussion
of the role of Galilean transformations in superfluid hydrodynamics.
116 Darryl D. Holm

The BK [1961] form of the phenomenological force f also implies the dissipative
heating rate,
R = (J − ρvn + curl λ) · ω × (vL − vn ) ,
which is Galilean invariant and positive. Substituting this form of f into the
superfluid motion equation in (1) and taking its curl provides the following
equation for the superfluid vorticity, ω = curl vs ,

∂t ω = curl (vL × ω) .

This vorticity equation implies the HVBK superfluid Kelvin circulation

theorem,  
d d
ω · dS = vs · dx = 0 .
dt S dt ∂S(vL )
The Kelvin formula (10) expresses conservation of the flux of superfluid vorticity
through any surface S whose boundary ∂S moves with the velocity vL , so vL
may be regarded as the local velocity of a vortex line. Equivalently, the Kelvin
formula expresses conservation of superfluid velocity circulation around any loop
that moves with the vortex line velocity vL .
In a key phenomenological step that closed the theory, BK [1961] assigned
the undetermined functions α and β in the force f and its auxiliary vortex line
velocity vL as
B  ρn Bρn
1 + αρs = , βρs = . (11)
2ρ 2ρ
The dimensionless coefficients B and B  were introduced earlier in Hall and
Vinen (HV) [1956] to parameterize the Gorter-Mellink [1949] “mutual friction”
force (B) and its reactive component (B  ). Hence the name, HVBK equations
for this closure.
The assignments in BK [1961] of the undetermined functions α, β in (11), as
well as the vortex slip velocity s0 in (6) are designed to reproduce the phenomena
observed in HV [1956] and yet still conserve mass, momentum and energy. This
phenomenological approach used in BK [1961] is indeterminate, however, in the
sense that some freedom still remains in making these assignments. The HVBK
equations (1) that result from this approach do possess the desired conservation
laws for mass, momentum and energy. And they also possess a Kelvin theorem
for the circulation of superfluid velocity. However, because of the indeterminacy
inherent in the phenomenological approach, the HVBK equations (1) are not
unique in possessing these properties. An alternative assignment of the vortex
slip velocity is

s = v − vn with v = v̄ + ρ−1 curl λ , with v̄ = ρ−1 (ρs vs + ρn vn ) . (12)

The mean velocity v̄ also figures prominantly in Hills and Roberts [1977] discus-
sion of the HVBK equations. As we shall see, the alternative expression (12) for
the auxiliary vortex slip velocity in terms of the mean velocity v̄ arises naturally
in the Hamiltonian derivation of a slightly modified set of HVBK equations.
 Introduction to HVBK Dynamics 117

These equations possess the same formal conservation and circulation properties
as HVBK, modulo redefining the vortex slip velocity as s rather than s0 . The
vortex slip velocity s in (12) is defined relative to the Galilean frame of the nor-
mal fluid, which is present only at finite temperature. The HVBK s0 in (6) is the
limit of the vortex slip velocity s for zero temperature, at which no normal fluid
remains. The vortex slip velocity s in (12) is a slight modification of s0 in (6)
necessary to incorporate finite temperature effects, without changing the form of
the HVBK theory, obtained from a Hamiltonian derivation of these equations in
the normal fluid frame. In the Hamiltonian framework, the energy-momentum
conservation laws and Kelvin circulation theorem are all natual consequences.
Moreover, the velocities v and vn are identified as being dual to the momenta
given by ρvs and ρn (vn − vs ), respectively.
Outline. We shall use a Hamiltonian approach with Lie-Poisson brackets to de-
rive the expression (12) for the vortex line velocity v at finite temperature from
first principles by using the energy E in (7) as the Hamiltonian. The momenta
conjugate to the velocities v and vn shall be our basic dynamical variables. The
finite temperature vortex line velocity v and slip velocity s determined this way
turn out to be
ρ 
v = ρ−1 (J + curl λ) , and s = v − vn
s
(v0 − vn ) . (13)
ρ

At zero temperature, ρ → ρs and these reduce to the BK [1961] phenomenolog-

ical expressions with v0 given by (6). Thus, the finite temperature corrections
found by using the Hamiltonian approach renormalize the HVBK slip velocity
in the mutual friction force f by the factor ρs /ρ (the superfluid mass fraction).
Aside from this renormalization, the vortex line tension is left unaffected by
this renormalization, the superfluid vortex equation keeps its form and no new
parameters are added.
Technical details of deriving this renormalized theory from its Hamiltonian
and Lie-Poisson brackets are given in the Appendix.
Main results. We shall use the superfluid vortex dynamics for the renormalized
HVBK equations obtained via this Hamiltonian approach to write a dynamical
equation for the space-integrated total vortex tangle length, which is the
quantity measured in the Oregon experiments on superfluid turbulence reported
in Skrbek, Niemela and Donnelly [1999].
We shall also study the restriction of the renormalized HVBK equations for
the incompressible case, in which ρn and ρs are constants and one takes
∇ · vn = 0 and ∇ · vs = 0. Finally, we shall demonstrate the invariance of the
forms of these equations upon transforming into a rotating frame. The Coriolis
force in such a rotating frame couples to the vortex line velocity vL , which of
course differs from both the superfluid velocity and the normal velocity. We shall
derive a Taylor-Proudman theorem for steady superfluid vortices under rapid
rotation. According to this superfluid Taylor-Proudman theorem, the vortex line
velocity becomes columnar under sufficiently rapid rotation. That is, the lateral
vortex line velocity is nondivergent and independent of the axial coordinate, and
118 Darryl D. Holm

the axial velocity decouples from the lateral motion. Therefore, under sufficiently
rapid rotation, the superfluid vortex filaments will straighten and become parallel
to the axis of rotation as they approach a steady state.
Numerical implications. This renormalization of the vortex line element slip
velocity in the HVBK equations from s0 → s s0 ρs /ρ is sensitive to temper-
ature, but it does not affect the vortex line tension. Therefore, a temperature
sensitive trade-off arises between mutual friction and vortex line tension that
may be worth testing in numerical simulations such as those reported in Hen-
derson and Barenghi [2000]. The HVBK equations are thought to break down in
the presence of strong counterflow. However, as general conservation laws there
is no mechanism in the equations that would signal this breakdown. A rotating
Rayleigh-Besnard experiment might be useful in testing the range of validity of
the HVBK equations (Barenghi, private communication). Such an experiment
might also indicate how these equations should be modified in the presence of
strong counterflow.
Experimental implications. Temperature sensitivity of the coupling between
the superfluid vortices and the normal fluid component is an area of intense
current investigation in superfluid Helium turbulence, see Donnelly [1999]. One
would like to know whether the ρs /ρ renormalization of the mutual friction forces
relative to the vortex line tension would matter significantly in comparisons of
the predictions of the HVBK equations with modern experiments in Helium
turbulence at low, but finite temperatures.
Superfluid vortex dynamics. To begin addressing this issue, we may use the
superfluid vorticity equation for the renormalized HVBK equations obtained in
the Appendix via the Hamiltonian approach to write an explicit equation for
the dynamics of Vinen’s vortex length density L = |ω|/κ. In the superfluid
turbulence decay experiments reported by Skrbek, Niemela and Donnelly [1999]
the spatial integral of this quantity is measured as a function of time to decrease
over six decades as t−3/2 . The integrated vortex length measured in these ex-
periments is predicted by the renormalized HVBK equations to be governed by
the superfluid vorticity dynamics alone.
Upon including mutual friction, the superfluid vortex dynamics for the renor-
malized HVBK equations is expressed as, cf. equation (9),
∂t ω = curl(vL × ω) ,
in which the renormalized total vortex line velocity given by, cf. equation (5),
B  ρn Bρn
vL = v − s− ω̂ × s , where s = v − vn , (15)
2ρ 2ρ
and its Hamiltonian limit is found to be
v = v̄ + ρ−1 curl λ , with v̄ = ρ−1 J and λ = λ ω̂ . (16)
Thus, relative to the Hamiltonian approach, the terms in B and B  are additional
velocities introduced by phenomenology, while v is the vortex line velocity in
the absence of mutual friction.
 Introduction to HVBK Dynamics 119

The HVBK superfluid vorticity equation implies the following dynamics for
the integrated vortex length measured in the turbulence decay experiments,
 
d 3
Ld x = ω̂ · ∂t ω/κ d 3 x
dt
!" # !" #
length vorticity dynamics

= L ṽ · (ω̂ × curl ω̂) d 3 x
!" #
transport · curvature

Bρn
− L (ω̂ × curl λ ω̂) · (ω̂ × curl ω̂) d 3 x
2ρ2
!" #
damping by curvature

   Bρn 
+ L n̂ × ω̂ · (ω̂ × v ) + (v − vs ) dS . (17)
2ρ
!" #
creation and destruction at the boundary

Here ω̂ is the unit vector tangent to a superfluid vortex filament, so κ = (ω̂ ×

curl ω̂) is its local curvature. The transport and damping of the vortex tangle
length is proportional to this local curvature. The effective velocity ṽ in the
transport term is given by

B  ρn Bρn
ṽ = v − (v − vn ) − ω̂ × (v − vn ) . (18)
2ρ 2ρ

According to the last term in (17), vortex length is created or destroyed at the
boundary, unless the vortex filaments approach it in the normal direction, so
that n̂ × ω̂ = 0.
Formula (17) for the evolution of the total superfluid vortex length presents a
trade-off between the mass-weighted velocity v and the local induction velocity
(or filament curvature) ω̂ × curl ω̂, in the interior of the domain. This trade-
off in the interior competes with the process of creation and destruction at the
boundary. For example, in counterflow turbulence, the superfluid moves toward
the heater at the boundary, so the term in v would tend to be nonzero. In
contrast, for grid turbulence, v is small, so this term would tend to contribute
less.
 This formula governs the dynamics of the experimentally measured quantity
L d 3 x. However, it does not yet show how to obtain the t−3/2 decrease seen in
this quantity by Skrbek, Niemela and Donnelly [1999] in their experiments on
decay of turbulence.
Suppose the main source of decay were the term labeled “damping by cur-
vature” in formula (17) and the flow were isothermal and incompressible. This
would imply

1 d L/R2  3
L = − c0 (T ) λ0 =− , (19)
L dt L 2 (t + t0 )
120 Darryl D. Holm

where λ0 = λ/ρs = (κ/4π)ln(b/a) is the quantum vortex constant, t0 is a time

shift in the experimental analysis, c0 (T ) ≡ Bρn ρs /(2ρ2 ) and angle brackets ·
denote spatial integral over the measurement domain. In particular,
 
L ≡ L d x , L/R  ≡ L |ω̂ × curl ω̂|2 d 3 x .
3 2
(20)

The measured t−3/2 decrease in L implies via formula (19) that the length-
weighted mean curvature of the vortex tangle L/R2 /L decays due to mutual
friction as t−1 . Thus, on the average as the vortex length decays, the vor-
tices tend to straighten, under the effects of mutual friction damping.
Preservation of helicity versus preservation of vortex length. The he-
licity, or linkage number for the superfluid vorticity is defined as

Λ = (vs · ω) d 3 x .

The helicity satisfies an evolution equation obtained from the superfluid vortex
dynamics,
   
dΛ 1
= − (n̂ · ω) μ − vn2 − vs · (vL − vn ) dS − (n̂ · vL )(vs · ω) dS .
dt 2
Therefore, even with mutual friction, helicity is created and destroyed only on
the boundary. Moreover, helicity will be preserved, provided both ω and vL are
tangential at the boundary. The former condition, however, is the opposite of
that required for the creation and destruction of vortex length at the boundary
to cease. Therefore, no equilibrium should be expected that preserves both the
helicity and the vortex length in a superfluid.
Superfluid vortex equilibria are not ABC flows. The steady equilibrium
solutions of the superfluid vorticity dynamics satisfy
curl (vL × ω) = 0 . (23)
For example, a steady equilibrium exists when ω and vL are parallel. Note that
these “super-Beltrami flows” are not eigenfunctions of the curl. Therefore, they
are not Arnold-Beltrami-Childress (ABC) flows, as occur for the Euler equations.

2 Incompressible Renormalized HVBK Flows

To express the renormalized HVBK equations in the incompressible limit, we
begin by recollecting the compressible equations and abbreviating |vs − vn |2 =
vs2 n ,
∂t S = − div(Svn ) + R/T ,
∂t ρ = − div(ρs vs + ρn vn ) ,
   1 
ρs ∂t vs + (vs · ∇)vs = − ρs ∇ μ − vs2 n + ρs (vL − vs ) × ω ,
   2 
∂t ρs vs i + ρn vn i = − ∂k ρn vn i vnk + ρs vs i vsk − ∂i P − ∂k τik ,
 
τik = klm ∂ l vs i λm − λi ω k + δik λ · ω .
 Introduction to HVBK Dynamics 121

As we have seen, finite temperature effects renormalize the total vortex line
velocity as
 
ρn B B
vL = v − ω̂ × s + s , where s = v − vn , (24)
ρ 2 2

and the Hamiltonian part of the line velocity (with corrections for finite temper-
ature) is defined as

v = v̄ + ρ−1 curl λ , with v̄ = ρ−1 (ρs vs + ρn vn ) and λ = λ ω̂ . (25)

To the extent that ρ, ρs , ρn and S all may be taken as constants for a given
temperature and the heating rate R is negligible, then the velocities vn and vs
are incompressible, i.e.,

∇ · vn = 0 and ∇ · vs = 0 .

In this situation, the pressure P may be obtained from the Poisson equation,
   
−∇2 P + λ · ω = div ρs (vs · ∇) vs + ρn (vn · ∇) vn − ω · ∇λ , (27)

found from the divergence of the total momentum equation. Combining the su-
perfluid motion equation with total momentum conservation results in an equa-
tion for the normal fluid velocity in the incompressible case
 ρs 
ρn ∂t vn + ρn (vn · ∇)vn = − ∇ P  − ρs μ + vs2 n − ρs (vL − vs ) × ω + ω · ∇λ ,
2
where vL is given in equation (24). We set P  ≡ P + λ · ω and take it as the
total pressure. (One also could have absorbed λ · ω into P earlier, by including
it in Euler’s pressure law.) Since λ = |λ|ω̂ and ω̂ is a unit vector, we find for
constant ρs the standard relation for the vortex line tension denoted as T.
Namely,
ω · ∇λ = − λ0 ρs ω × curl ω̂ ≡ ρs T ,
where λ0 = λ/ρs = (κ/4π)ln(b/a) is a constant.
Remark. We note that the quantity T known as the vortex line tension first
appears in the normal fluid equation, as a reaction to the presence of the super-
fluid. The standard convention for introducing the mutual friction force has the
effect of shifting T into the superfluid equation. By action and reaction, though,
T could appear in either equation.
These equations of motion must be completed by providing an equation of
state relation for the quantity μ − 12 vs2 n . BK [1961] assumes a law of partial
pressures,
ρn  ρs  1
Pn = P = P  − Ps and Ps = P = ρs μ − ρs vs2 n .
ρ ρ 2
122 Darryl D. Holm

In this case, the renormalized HVBK motion equations for incompressible flow
reduce to
1 ρn
∂t vs + (vs · ∇)vs = − ∇P  − Fn s + T , (31)
ρ ρ
1 ρs
∂t vn + (vn · ∇)vn = − ∇P  + Fn s . (32)
ρ ρ
In these superfluid motion equations, the renormalized mutual friction force
Fn s is defined as the sum (with ω = curl vs )
ρ  B B  0
s
Fn s = (s 0 × ω) + Fn0 s , where Fn0 s = ω̂ × s 0 + s × ω . (33)
ρ 2 2
Here Fn0 s is the HVBK mutual friction force without any finite temperature cor-
rections. To acquire these formulas, we used the relations for the incompressible
case,
ρs   ρs  0  ρs 0
s = v − vn = vs + λ0 curl ω̂ − vn = v  − vn = s , (34)
ρ ρ ρ
with s 0 = vs + λ0 curl ω̂ − vn , and we eliminated vL by using the relation
 
  ρ s ρn B B
− ρs vL − vs − λ0 curl ω̂ = ρn s + ω̂ × s + s . (35)
ρ 2 2

In equation (33) for Fn s , the quantity (s 0 × ω) is the Hamiltonian reactive

force (which could be naturally absorbed into Vinen’s B  parameter) and Fn0 s
is the phenomenological mutual friction force defined according to the standard
convention as in BK [1961] and Donnelly [1999]. The finite-temperature correc-
tions contribute an overall factor of ρs /ρ to the standard zero-temperature
expression Fn0 s for the phenomenological mutual friction force. No new parame-
ters are added, but a temperature dependent trade-off is identified between
the renormalized mutual friction coupling and the vortex line tension, since the
vortex line tension remains unaffected by the finite-temperature corrections.
In the isothermal case, the motion equations are closed by the Poisson equa-
tion for P , since the other coefficients (B, B  , ρn /ρ, etc.) are specified functions
of temperature and they may be taken as constants, for an isothermal incom-
pressible superfluid flow.
Note that equations (31-32) may be rewritten with ωn = curl vn as
ρs 0 ρn ρs 1
∂t vs + ∇μs = vn × ω + s × ω − 2 Fn0 s , μs = P /ρ + vs2 , (36)
ρ ρ 2
ρs 0 ρ2s 0 1
∂t vn + ∇μn = vn × ωn + s × ω + 2 Fn s , μn = P /ρ + vn2 . (37)
ρ ρ 2
These equations imply an equation for the velocity difference,
1 ρs
∂t (vs −vn )+ ∇(vs2 −vn2 ) = vn ×(ω −ωn )− Fn0 s with ωn = curl vn . (38)
2 ρ
 Introduction to HVBK Dynamics 123

and there is no tendency for mutual friction to cause any alignment in the
vorticities of the superfluid and its normal component. Instead, the curl ω̂ part
of Fn0 s = 0 would break any such alignment, if it were to form spontaneously.
Indeed, alignments sufficient for steady solutions are

s 0 × (∇μs × ∇μn ) = 0 , s0 × ω = 0 and s 0 × vn = 0 , (39)

with s ≡ vs + λ0 curl ω̂ − vn ,
0

provided μs and μn are functionally unrelated. Thus, the steady equilibrium

alignments imposed by mutual friction involve vn , vs and curl ω̂, as well as the
independent gradients of μs and μn . For example, one class of equilibria has s 0 ,
vn , ω all aligned tangent to intersections of level surfaces of μn and μs .

3 Rotating Frame Renormalized HVBK Equations

We transform to a rotating frame with relative velocities denoted with an as-

terisk as vs∗ = vs − R(x), etc., and curl R = 2Ω. After a calculation involving
Legendre transformations, we obtain the Hamiltonian for the relative motion,
cf. the Hamiltonian in (61) of the Appendix,
 
∗ ∗ 1
h(M , ρ, S, u, A , n) = ρ |vn∗ + R(x)|2 + (M∗ − ρA∗ − ρvn∗ ) · vn∗ (40)
2
 
+ ε0 (ρ, S, vs∗ − vn∗ , ω ∗ + 2Ω) − R · ρ(vn∗ + R) + (ρ − n)(A∗ − R) d 3 x .

Here M∗ − ρA∗ = J∗ = J − ρR and vs = u − (A∗ − R). The equations resulting

from the Lie-Poisson bracket (55) of the Appendix in these relative variables
keep their forms and the condition n = ρ is still preserved. We conclude with
the following three remarks.
Superfluid Coriolis force couples to the vortex line velocity. The Hamil-
tonian evolution equation for the superfluid velocity in the rotating frame is
expressed as
 1 1 
∂t vs∗ + (vs∗ · ∇)vs∗ = − ∇ μ − |vs∗ − vn∗ |2 − |R|2 + (v∗ − vs∗ ) × ω ∗ + v∗ × 2Ω .
2 2
The last term is the Coriolis force and it involves the relative vortex line velocity.
The curl of this equation yields
 
∂t (ω ∗ + 2Ω) = curl v∗ × (ω ∗ + 2Ω) .

The form of the vortex dynamics equation is invariant under passing to a

steadily rotating frame, and the superfluid Coriolis force contains the renormal-
ized vortex line velocity, rather than the superfluid velocity. Therefore, this is
not merely a kinematic force. The vortex line velocity appearing in the superfluid
Coriolis force includes the interaction between the vortex lines and the superfluid
124 Darryl D. Holm

component. It also includes the interaction with the normal component, since
v depends on the relative momentum density and contains finite temperature
effects. The superfluid Coriolis force is essential in the spin up problem in He-II,
see, e.g., Reissenegger [1993].
Superfluid Taylor-Proudman theorem. For steady, or slow motions and
rapid rotation we have  
0 = curl v∗ × 2Ω .
If the rotation is uniform (∇Ω = 0) and oriented vertically (Ω = |Ω|ẑ) this
becomes
   T
0 = 2|Ω| ∂z v∗ − ẑ div v∗ = 2|Ω| ∂z v∗x , ∂z v∗y , −∂x v∗x − ∂y v∗y ,

where ( )T denotes transpose of a row vector into a column vector. Thus, for
steady, or slow motions and rapid uniform rotation, we find that vortex line
motion is columnar. That is, the lateral vortex line velocity is nondivergent
and independent of the axial coordinate, and the axial velocity decouples from
the lateral motion. Therefore, under sufficiently rapid rotation, the superfluid
vortex filaments will straighten and become parallel to the axis of rotation as
they approach a steady state. However, they may still undergo nondivergent
motion in the lateral plane. This superfluid Taylor-Proudman theorem explains
why steady superfluid vortices tend to be aligned with the rotation axis under
rapid uniform rotation. The same conclusion applies, if the velocity v∗ in the
Hamiltonian formulation is replaced by the phenomenological relative velocity
∗
vL = vL − R. Similar considerations are discussed in Sonin [1987] from a more
microscopic viewpoint.
Relative total momentum is not conserved for rotating compressible
flows. Since the Hamiltonian depends explicitly on spatial position, instead of
conserving relative total momentum, we have the balance
∂ h  ρ
∂t Ji∗ + ∂j Ti∗ j = − i  = ∂i |R|2 ,
∂x explicit 2
where h is the Hamiltonian density in equation (40). This relative momentum
balance is the effect of centrifugal force. Here we have dropped terms proportional
to ρ − n, since ρ = n is still preserved in a rotating frame. Consequently, the
stress tensor in the relative momentum equation also keeps its form in passing to
a rotating frame, although the total relative momentum is no longer conserved
if the flow is compressible.

Acknowledgments
I am grateful to H. R. Brand, A. Brandenburg, P. Constantin, R. Donnelly, V.
V. Lebedev, F. Lund, J. E. Marsden, J. Niemela, A. Reisenegger, L. Skrbek, K.
Sreenivasan and W. F. Vinen for stimulating discussions and encouragement. I
am also grateful for hospitality at the UC Santa Barbara Institute for Theoretical
Physics where this work was initiated during their Hydrodynamic Turbulence
 Introduction to HVBK Dynamics 125

program in spring 2000. This research was supported by the U.S. Department of
Energy under contracts W-7405-ENG-36 and the Applied Mathematical Sciences
Program KC-07-01-01.

Appendix: Lie-Poisson Hamiltonian Formulation

Conservation of the number of quantum vortices moving through superfluid 4 He
(and across the streamlines of the normal fluid component) is expressed by

d
ω · n̂ dS = 0 , (46)
dt S

where the superfluid vorticity ω is the areal density of vortices and n̂ is the unit
vector normal to the surface S whose boundary ∂S moves with the vortex line
velocity v . When ω = curl vs this is equivalent to a vortex Kelvin theorem

d
vs · dx = 0 , (47)
dt ∂S(v )

which in turn implies the fundamental relation

∂t vs − v × ω = ∇μ . (48)

The superfluid velocity naturally splits into vs = u − A, where u = ∇φ

and (minus) the curl of A yields the superfluid vorticity ω. The phase φ is
then a regular function without singularities. This splitting will reveal that the
Hamiltonian dynamics of superfluid 4 He with vortices may be expressed as an
invariant subsystem of a larger Hamiltonian system in which u and A have
independent evolution equations.
We begin by defining a phase frequency in the normal velocity frame as

∂t φ + vn · ∇φ = ν . (49)

The mass density ρ and the phase φ are canonically conjugate in the Hamiltonian
formulation. Therefore, one may set ν = − δh/δρ for a given Hamiltonian h and
the phase gradient u = ∇φ satisfies
δh
∂t u + vn · ∇u + (∇vn )T · u = − ∇ , (50)
δρ

where ( )T denotes transpose, so that (∇vn )T · u = uj ∇vnj . The mass density ρ

satisfies the dual equation
δh
∂t ρ + ∇ · (ρvn ) = − ∇ · . (51)
δu
Perhaps not surprisingly, the rotational and potential components of the super-
fluid velocity vs = u−A satisfy similar equations, but the rotational component
126 Darryl D. Holm

is advected by the vortex line velocity v , instead of the normal velocity vn . Ab-
sorbing all gradients into u yields

∂t A + v × ω = 0 . (52)

Taking the difference of the equations for u and A then recovers equation (48)
as  δh 
∂t vs − v × ω = − ∇ vn · u + with vs = u − A , (53)
δρ
in which one uses regularity of the phase φ to set curl u = 0. It remains to
determine v from the Hamiltonian formulation. Including the additional degree
of freedom A associated with the vortex lines allows them to move relative to
both the normal and super components of the fluid, and thereby introduces
an additional reactive force without introducing any additional inertia. This
Hamiltonian approach thus yields renormalized HVBK equations.
Proposition: Upon splitting the superfluid velocity into vs = u − A (with u =
∇φ so that ω = − curl A) the (renormalized) HVBK equations in the Galilean
frame of the normal fluid form an invariant subsystem of a Lie-Poisson
Hamiltonian system,
∂f
= {f, h} with f, h ∈ (M, ρ, S, u, A, n) ,
∂t
and Lie-Poisson bracket given by

{f, h} =
   
δf δh δh δh δh
− (Mk ∂j + ∂k Mj ) + ρ ∂j + S∂j + (∂k uj − uk , j )
δMj δMk δρ δS δuk
   
δf δf δf δh δf δh δf δh
+ ∂k ρ + ∂k S + (uk ∂j + uj , k ) + ∂k + ∂j
δρ δS δuj δMk δρ δuk δuj δρ
 
δf δh Aj , k − Ak , j δh δf δh
− ∂j + − ∂k d 3x . (55)
δAj δn n δAk δn δAk

Remarks: Here M is the total momentum density, the total mass density is ρ
and the entropy density is S. We shall interpret the density n later, after we
develop the Hamiltonian equations of motion. It shall emerge that n = ρ is an
invariant condition and, hence,

M − nA = J = p + ρvn ,

for n = ρ, where p = J − ρvn = ρs (vs − vn ) is the relative momentum density of

the superfluid in the Galilean frame of the normal fluid. The momentum density
associated with the vortex fluid will be N = −nA. The Hamiltonian will be the
energy E in (7).
The Lie-Poisson bracket in the Proposition appeared first in Holm and Ku-
pershmidt [1987] in a study of various approximate equations for the dynamics of
 Introduction to HVBK Dynamics 127

multicomponent superfluids with charged condensates. The mathematical struc-

ture of this Lie-Poisson bracket and its association with the dual of a certain Lie
algebra is discussed in Holm and Kupershmidt [1987]. Our re-interpretation of
this Poisson bracket introduced and studied earlier shall now yield a extension of
the HVBK equations that enables the vortex line velocity v and hence the vor-
tex reactive force and mutual friction force to be expressed at finite temperature.
Identifying this Poisson bracket as being dual to a Lie algebra establishes that
it satisfies the Jacobi identity, ijk {fi , {fj , fk }} = 0. The term in the Poisson
bracket responsible for the reactive vortex force will turn out to be {Ai , Aj } = 0.
The Poisson bracket {vs i , vs j } would vanish (as does {ui , uj } = 0) and thus
the reactive vortex force would be absent, in any Hamiltonian formulation for
which A and n were not independent degrees of freedom from M, ρ, S. Volovik
and Dotsenko [1979, 1980] obtain a different result and provide no Lie-algebraic
justification.
A Lagrangian formulation of these equations is also available. However, it
involves an equation for δl/δν about which nothing is known physically.
Corollary #1: The Lie-Poisson bracket is equivalent to the following separate
Hamiltonian matrix forms for the dynamical equations
⎡ ⎤ ⎡ ⎤⎡ ⎤
Mi Mj ∂i + ∂j Mi S∂i ρ∂i ∂j ui − uj , i δh/δMj
∂ ⎢ ⎥
⎢ S ⎥ = −⎢
⎢ ∂j S 0 0 0 ⎥ ⎢ δh/δS ⎥
⎥⎢ ⎥
⎣
∂t ρ ⎦ ⎣ ∂j ρ 0 0 ∂j ⎦ ⎣ δh/δρ ⎦ , (57)
ui uj ∂i + ui , j 0 ∂i 0 δh/δu j
and, upon defining N = −nA,
    
∂ Ni Nj ∂i + ∂j Ni n∂i δh/δNj
=− . (58)
∂t n ∂j n 0 δh/δn
These are individually expressed as
∂t S = {S, h} = − div(S δh/δM) ,
∂t n = {n, h} = − div(n δh/δN) ,
∂t ρ = {ρ, h} = − div(ρ δh/δM + δh/δu) ,
 
∂t u = {u, h} = − ∇ δh/δρ + (δh/δM) · u + (δh/δM) × curl u ,
∂t (N/n) = {(N/n), h}
   
= − ∇ δh/δn + (δh/δN) · (N/n) + δh/δN × curl (N/n) ,
 
∂t Mj + Nj = {Mj + Nj , h} = − ∂k Tjk .

Corollary #2: Consider a translation invariant Hamiltonian density with de-

pendence
h(M, ρ, S, n, vs , ω, A),
where vs = u−A, A = −N/n and ω = curl vs . The stress tensor Tjk is expressed
in terms of this Hamiltonian as
  
∂h ∂h  ∂h  ∂h ∂h 
k
T j = Mj +vs j + curl −vs l, j mlk + δj P −Aj
k
.
∂Mk ∂vs k ∂ω k ∂ωm ∂Ak vs
 128 Darryl D. Holm

where
∂h ∂h ∂h ∂h
P = Ml +ρ +S +n − h,
∂Ml ∂ρ ∂S ∂n
as in the Euler relation for pressure.
Remark. One notes many parallels and correspondences among these equations.
Note especially the expected similarities in the equations for u and N/n. Recall
that A = − N/n, so that the superfluid velocity is given by vs = u − A =
u + N/n. The evolution of the superfluid velocity is consistently composed as
the sum of these two separate dynamical pieces.
Proof of the Proposition: The following Hamiltonian h (and conserved en-
ergy) will yield the HVBK equations in the frame of the normal fluid upon using
this Lie-Poisson bracket
  
1 2
h = d x − ρ vn + (M − ρA) · vn + ε0 (ρ, S, vs − vn , ω) .
3
(61)
2
The variational derivatives of the Hamiltonian h are computed in this reference
frame by using the thermodynamic first law (2). Namely,
  1 
δh = d 3 x μ − vn2 − A · vn δρ + T δS + vn · δM + (p + curl λ) · δu
2
    
− p + curl λ + ρvn · δA + M − p − ρvn − ρA · δvn .

Here we have used the velocity split δvs = δu − δA and assumed the boundary
condition n̂ · ω × λ = 0 when integrating by parts. This boundary condition
is satisfied identically, since λ = λ ω̂ in the HVBK theory. Upon substituting
these variational derivatives into the Lie-Poisson bracket, Corollary #1 yields
the following equations expressed in the normal fluid reference frame,

∂t S = {S, h} = − div(Svn ) ,
∂t n = {n, h} = − div(ρvn + p + curl λ) ,
∂t ρ = {ρ, h} = − div(ρvn + p + curl λ) ,
( Hence, the condition n = ρ is preserved.)
1
∂t u = {u, h} = − ∇(μ − vn2 + vn · vs ) + vn × curl u ,
2
( Hence, curl u = 0 is preserved.)
 
∂t A = {A, h} = n−1 ρvn + p + curl λ × curl A ,
( Hence, v = vn + ρ−1 (p + curl λ) when n = ρ is used.)
 
∂t Mj − nAj = {Mj − nAj , h} = − ∂k Tjk

Remarks:
(1.) Preservation of the condition n = ρ by these equations allows the introduc-
tion of the momentum-carrying field A as an independent degree of freedom
 Introduction to HVBK Dynamics 129

without introducing additional material inertia, provided the dynamically pre-

served condition n = ρ holds initially. This is reminiscent of the preservation
of Gauss’s Law by the continuity equation for mass conservation in a fluid
plasma.
(2.) The curl of the dynamical equation for the field A implies the vortex line
velocity
1 δh
v = − = v̄ + ρ−1 curl λ , where v̄ = vn + ρ−1 p = ρ−1 J .
n δA
The velocity v̄ is the mass averaged velocity. The vortex slip velocity s corre-
sponding to the vortex line velocity v is the basis for the phenomenological
reactive and mutual friction forces f and Rayleigh dissipation function R in
the HVBK system. Namely,
s = v − vn = ρ−1 (p + curlλ) , with λ = λ ω̂ .
As expected, this expression agrees with BK [1961] at zero temperature. Note
that the renormalized HVBK equations introduce no new parame-
ters.
(3.) The corresponding equation for vs = u − A is then obtained as
1 2
∂t vs = − ∇(μ − v + vn · vs ) + v × ω , where v = v̄ + ρ−1 curl λ ω̂ .
2 n
This may be expressed equivalently in manifestly Galilean invariant form as
 1 
∂t vs + (vs · ∇)vs = − ∇ μ − |vs − vn |2 + f  , where f  = (v − vs ) × ω .
2
The term f  is the Hamiltonian contribution to the reactive vortex
force. This contribution would vanish if the vortex lines moved with the
superfluid velocity.
(4.) The stress tensor Tjk = πjk + τjk for total momentum conservation is given by
summing
 
πjk = ρs vs j vsk + ρn vn j vnk + P δjk ,
 
τjk = ∂l k l m vs j λm − λj ω k + ω · λ δjk .

The divergence of τjk defines the vortex line tension T as

∂k τjk = − ω · ∇λ + ∇(ω · λ) = − ρs T + ∇(ω · λ) ,

In the stress tensor πjk the pressure P is defined by the Euler relation,
P = − ε0 + μρ + T S ,
so that in the normal-fluid frame the pressure satisfies
dP = ρdμ + SdT − p · d(vs − vn ) − λ · d ω .
The stress tensor Tjk = πjk + τjk may be derived by using Corollary #2 for
the Hamiltonian formulation.
130 Darryl D. Holm

Implications of the HVBK vortex dynamics.

The new Hamiltonian formulation of the renormalized HVBK equations pre-
sented in the Proposition provides a formula for the slip velocity of a vortex
line element in a turbulent superfluid at finite temperature. Namely, for the
Hamiltonian h in equation (61), one finds
 
s = v − vn = ρ−1 p + curl λ . (69)
This formula for v recovers the HVBK expression in BK [1961] at zero tem-
perature. Otherwise, it provides an extension to finite temperature of the
HVBK vortex force
f = (vL − vs ) × ω , with vL = vn − ρs (α s + β ω̂ × s) , (70)
where the renormalized vortex slip velocity is given by
ρs 0
s = v  − vn = (v − vn ) , for constant ρs . (71)
ρ 
The corresponding heating rate R is given by
R = (J − ρvn + curl λ) · ω × (vL − vn ) = ρρs β ω |s × ω̂|2 ,
which is positive, as it must be.

References
1. Bekarevich, I. L. and I. M. Khalatnikov [1961] Phenomenological derivation of the
equations of vortex motion in He II, Sov. Phys. JETP 13 643-646.
2. Donnelly, R. J. [1999] Cryrogenic fluid dynamics, J. Phys.: Condens. Matter 11
7783-7834.
3. Gorter, C. J. and J. H. Mellink [1949] Physica 15 285.
4. Henderson, K. L. and C. F. Barenghi [2000] The anomalous motion of superfluid
helium in a rotating cavity, J. Fluid Mech. 406 199-219.
5. Hills, R. N. and P. H. Roberts [1977] Superfluid mechanics for a high density of
vortex lines, Arch. Rat. Mech. Anal. 66 43-71.
6. Holm, D. D. and B. Kupershmidt [1987] Superfluid plasmas: multivelocity non-
linear hydrodynamics of superfluid solutions with charged condensates coupled
electromagnetically, Phys. Rev. A 36 3947-3956.
7. Nemirovskii, S. K. and W. Fiszdon [1995] Chaotic quantized vortices and hydro-
dynamic processes in superfluid helium, Rev. Mod. Phys. 67 37-84.
8. Putterman, S. J. [1974] Superfluid Hydrodynamics, North Holland, Amsterdam.
9. Reissenegger, A. [1993] The spin up problem in Helium-II J. Low Temp. Phys. 92
77-106.
10. Skrbek, L., J. J. Niemela and R. J. Donnelly [1999] Turbulent flows at cryogenic
temperatures: a new frontier, J. Phys.: Condens. Matter 11 7761-7783.
11. Sonin, E. B. [1987] Vortex oscillations and hydrodynamics of rotating superfluids,
Rev. Mod. Phys. 59 87-155. 1987
12. Volovik, G. E. and V. S. Dotsenko [1979] Poisson brackets and continuous dynam-
ics of the vortex lattice in rotating He-II, JETP Lett. 29 576-579.
13. Volovik, G. E. and V. S. Dotsenko [1980] Hydrodynamics of defects in condensed
media in the concrete cases of vortices in rotating Helium-II and of disclinations
in planar magnetic substances, Sov. Phys. JETP 58 65-80.
Magnus Force, Aharonov–Bohm Effect,
and Berry Phase in Superfluids

Edouard Sonin

Racah Institute of Physics, Hebrew University of Jerusalem

1 Introduction
If the vortex moves with respect to a liquid, classical [1] or quantum, there
is a force on the vortex normal to the relative vortex velocity with respect to
the liquid. This is the Magnus force, which plays an important role in mod-
ern condensed-matter physics. In particular, it determines the mutual fiction in
superfluids [2,3,4] and the Hall effect in superconductors [5].
An obvious generalization of the classical Magnus force in the superfluid
seemed to be a force proportional to the superfluid density ρs [2]:

ρs [(v L − v s ) × κ] = F , (1)

where v L is the vortex velocity, v s is the superfluid velocity, and κ is a vector

along the vortex line with κ = h/m being the circulation of the superfluid
velocity around the vortex. In absence of the external force F on a superfluid,
the vortex moves with the superfluid velocity. However, quasiparticles (rotons [6]
and phonons [7]) produce an additional force transverse to the vortex velocity
and
F = −D(v L − v n ) − D [ẑ × (v L − v n )] . (2)
Here v n is the normal velocity. It was shown [8] that the additional transverse
force ∝ D = −κρn (the Iordanskii force) is connected with the analogue of the
Aharonov–Bohm effect [9] in superfluids. A similar force on the magnetic-flux
tube exists in the original Aharonov–Bohm effect for electrons [10].
Ao and Thouless [11] have pointed out that the Magnus force is connected
with the Berry phase [12], which is the phase variation of the quantum-mechanical
wave function of a quantum liquid generated by the adiabatic transport of the
vortex round a close loop. From the Berry-phase analysis Ao and Thouless [11,13]
concluded that the only force transverse to v L is the force proportional to ρs ,
i.e., D = 0 in Eq. (2). This conclusion disagreed with the previous calculations
and therefore was vividly discussed [14,15].
The present paper is an attempt to bring together two points of view in
order to find a source of disagreement [16]. I restrict myself with the problem of
the Galilean invariant quantum Bose-liquid described by the Gross–Pitaevskii
theory [17]. At large scales the theory yields equations of the hydrodynamics of
an ideal inviscous liquid. In presence of an ensemble of sound waves (phonons)
with the Planck distribution, which is characterized by a locally defined normal
velocity, one obtains the two-fluid hydrodynamics. The momentum balance in the

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 131–137, 2001.

c Springer-Verlag Berlin Heidelberg 2001
132 Edouard Sonin

area around a moving vortex demonstrates the existence of the Iordanskii force. I
also discuss the Berry phase. According to Refs. [11,13], the Berry phase and the
Magnus forces are proportional to the total current circulation at large distances.
But the total current circulation contains a normal-fluid contribution, which is
proportional to the Iordanski force. Taking this contribution into account, the
Berry-phase analysis agrees with the momentum-balance approach.

2 Gross–Pitaevskii Theory
and Two-Fluid Hydrodynamics
In the Gross–Pitaevskii theory [17] the ground state and weakly excited states
of a Bose-gas are described by the nonlinear Schrödinger equation
∂ψ 2 2
i =− ∇ ψ + V |ψ|2 ψ (3)
∂t 2m
for the condensate wave function ψ = a exp(iφ). The nonlinear Schrödinger
equation is the Euler–Lagrange equation for the Lagrangian
 
i ∂ψ ∂ψ ∗ 2 V
L= ψ∗ −ψ − |∇ψ|2 − |ψ|4 . (4)
2 ∂t ∂t 2m 2

The Noether theorem yields the momentum conservation law ∂ji /∂t+∇j Πij = 0
where j = Im{ψ ∗ ∇ψ} is the mass current, and the momentum-flux tensor is
 
2 ∗ ∗ V 2 2 2
Πij = (∇i ψ∇j ψ + ∇i ψ ∇j ψ) + δij |ψ| −
4
∇ |ψ| . (5)
2m 2 4m

Using the Madelung transformation [3], one obtains from complex Eq. (3)
two real equations for the liquid density ρ = ma2 and the liquid velocity v =
(/m)∇φ = (κ/2π)∇φ. Far from the vortex line these equations are hydrody-
namic equations for an ideal inviscous liquid:
∂ρ
+ ∇(ρv) = 0 , (6)
∂t

∂v
+ (v · ∇)v = −∇μ . (7)
∂t
Here μ = V a2 /m is the chemical potential. Equation (5) becomes the hydrody-
namic momentum-flux tensor Πij = P δij + ρvi (r)vj (r).
A plane sound wave propagating in the liquid generates the phase variation
φ(r, t) = φ0 exp(ik · r − iωt). Then ρ(r, t) = ρ0 + ρ(1) (r, t) and v(r, t) = v 0 +
v (1) (r, t), where ρ0 and v 0 are the average density and velocity in the liquid,
whereas ρ(1) (r, t) and v (1) (r, t) = (κ/2π)∇φ are periodical variations of the
density and the velocity due to the sound wave (ρ(1)  = 0, v (1)  = 0). Equations
(6) and (7) linearized with respect to ρ(1) and v (1) yield the sound equation for

φ with the sound velocity cs = V a2 /m and the spectrum ω = cs k + k · v 0 .
 Magnus Force, Aharonov–Bohm Effect, and Berry Phase 133

The total mass current expanded up to the terms of the second order with
respect the wave amplitude and averaged over time is

κ2 k
j = ρ0 v 0 + ρ(1) v (1)  = ρ0 v 0 + ρ0 φ20 k. (8)
8π 2 cs
If there is an ensemble of phonons with the Planck distribution
 −1  −1
E(p) − p · v n cs p + p · (v 0 − v n )
n0 (E, v n ) = exp −1 = exp −1 ,
T T
(9)
the total mass current linearized with respect to v 0 − v n is

1
j = ρ0 v 0 + 3 n0 (p)p d3 p = ρ0 v 0 + ρn (v n − v 0 ) . (10)
h
Here p = k is the phonon momentum, E = cs p+p·v 0 is the phonon energy, and
v n is the drift velocity of phonons. This expression is equivalent to the two-fluid
expression assuming that ρ = ρ0 = ρs + ρn , v 0 = v s , and  the normal density is
given by the usual two-fluid expression ρn = −(1/3h3 ) [∂n0 (ε, 0)/∂E]p2 d3 p.
Expanding the momentum-flux tensor up to terms of the second-order with
respect to the sound wave amplitude one obtains:
 
c2s ρ(1)  v(1) 
2 2
Πij = P0 δij + ρ0 v0i v0j + − ρ0 δij
ρ0 2 2
+ρ(1) (v(1) )i v0j + ρ(1) (v(1) )j v0i + ρ0 (v(1) )i (v(1) )j  . (11)

For the Planck distribution this yields the two-fluid momentum flux tensor

Πij = P δij + ρs vsi vsj + ρn vni vnj . (12)

Taking into account phonon-phonon interaction, which establishes the local

Planck distribution of phonons, we can derive all equations of the two-fluid
hydrodynamics, as shown in Ref. [18]. Thus the two-fluid hydrodynamics can
be derived from the hydrodynamics of an ideal inviscous fluid, and the latter
follows from the Gross–Pitaevskii theory.

3 Interaction of Phonons with a Vortex in Hydrodynamics

In presence of a vortex the sound equation is (see Refs. [8,14] for more details)

∂2φ ∂φ
2
− c2s ∇2 φ = −2v v (r) · ∇ , (13)
∂t ∂t
where
κ×r
v v (r) = (14)
2πr2
134 Edouard Sonin

is the circular velocity field induced by a vortex line. Here r is the position vector
in the plane xy. The sound wave produces the density variation
 
ρ0 κ ∂φ
ρ(1) = − 2 + v v · ∇φ(r) . (15)
cs 2π ∂t
One can calculate the differential cross-section in the Born approximation,
but since it is quadratic in κ this does not yield a transverse force [7,8,14]. Instead
we consider a quasiclassical solution of the sound equation:
   
iδS iκk
φ = φ0 exp −iωt + ik · r + = φ0 exp(−iωt + ik · r) 1 + θ , (16)
 2πcs
r
where δS = −(k/cs ) v v · dl = θκk/2πcs is the variation of the action due
to interaction with the circular velocity around the vortex. The angle θ is an
azimuth angle for the position vector r measured from the direction opposite
to the wave vector k. This choice provides that the quasiclassical correction
vanishes for the incident wave far from the vortex. One can check directly that
Eq. (16) satisfies the sound equation (13) in the first order of the parameter
κk/cs . The velocity generated by the sound wave around the vortex is
 
κ κ ik
v (1) = ∇φ = φ0 exp(−iωt + ik · r) ik − v v . (17)
2π 2π cs

According to Eq. (16) the phase φ is multivalued, and one must choose a cut for
an angle θ at the direction k, where θ = ±π . The jump of the phase on the cut
line behind the vortex is a manifestation of the Aharonov–Bohm effect [9]: the
sound wave after its interaction with the vortex has different phases on the left
and on the right of the vortex line. This results in an interference [8,14]. The
width of the interference region is dint ∼ r/k. 
Now we consider the momentum balance using the condition that dSj Π⊥j =
0 for a cylindrical surface around the vortex line. The subscript ⊥ points a com-
ponent normal to the wave vector k of the incident wave. The total momentum-
flux tensor can be obtained from Eq. (11) assuming v 0 (r) = v v (r) + v s :

Πij = −ρ0 (v s − v L ) · v v δij + ρ0 v0i v0j

+ρ(1) (v(1) )i vvj + ρ(1) (v(1) )j vvi + ρ0 (v(1) )i (v(1) )j  . (18)

The first two terms in this expression yield the momentum flux without phonons,
which produces the Magnus force for a liquid with the density ρ0 and the ve-
locity v s . The rest terms cancel except for the contribution from the term
ρ0 (v(1) )i (v(1) )j  in the interference region where v(1)⊥ = (κ/2πr)∂φ/∂θ. The
contribution depends on the phase jump and for a single sound wave is:

1 κ2 k 1 κ3 k 2
ρ0 (v(1) )⊥ (v(1) )r rdθ = ρ0 φ20 [δS(−) − δS(+)] = ρ0 φ20 ,
8π 2  8π 2 cs
(19)
where δS(±) = ∓κk/2cs are the action variations at θ → ∓π.
 Magnus Force, Aharonov–Bohm Effect, and Berry Phase 135

For the Planck phonon distribution the condition dSj Π⊥j = 0 yields:

ρ0 [(v L −v s )×κ]−[ρ(1) v (1) ×κ] = ρs [(v L −v s )×κ]+ρn [(v L −v n )×κ] = 0 . (20)

The term ∝ (v L − v n ) is the Iordanskii force, which corresponds to D = −κρn

in Eq. (2). Uniting in Eq. (20) terms linear in v L , we see that the total transverse
force (effective Magnus force) is proportional to the total density ρ0 = ρs + ρn .

4 Momentum Balance in the Two-Fluid Hydrodynamics

Up to now we analyzed spatial scales much less than the mean-free-path lph of
phonons (ballistic region). Now we shall see what is going on at scales much
larger than lph where the two-fluid hydrodynamics is valid.
Interaction between phonons and the vortex in the ballistic region produces
a force concentrated along the vortex line in the hydrodynamic region. The re-
sponse of the normal fluid to this force is described by the Navier-Stokes equation
with the dynamic viscosity ηn :
∂v n ∇P ρs S
+ (v n · ∇)v n = νn Δv n − − ∇T , (21)
∂t ρ ρn ρ
where νn = ηn /ρn is the kinematic viscosity, S is the entropy per unit volume,
and T is the temperature. At r < rm ∼ νn /|v n − v L | one may neglect * the
nonlinear inertial (convection) term (v n · ∇)v n . The line force −Fi = τij dSj
on the normal fluid produces a divergent velocity field (the Stokes paradox [1]):
F r
v n (r) = v n + ln , (22)
4πηn lph
where v n is the normal velocity at distances r ∼ lph , which separate the ballistic
and the hydrodynamic regions. Here τij = −ηn (∇i vnj + ∇j vni ) is the viscous
stress tensor. Due to viscosity the normal velocities v n∞ and v n at large (r ∼ rm )
and small (r ∼ lph ) distances from the vortex line are different (the viscous drag
[2]):
4πηn
F = (v n∞ − v n ) . (23)
ln(rm /lph )
The momentum
* balance
* in the two-fluid region must include the viscous
stress tensor: Πij dSj + τij dSj = 0, where Πij is given by Eq. (12). Since
* the
normal velocity field does not contain the circular velocity v v , the flux Πij dSj
yields the superfluid Magnus force, i.e., the force F satisfies Eq. (1).
However, at very large distances r  rm the nonlinear convection term is
more important than the viscous term. Thus the scale rm separates the viscous
and convection subregions. In the convection subregion the viscosity becomes
ineffective and the momentum flux related to the linear force on the normal
fluid should be connected with the normal part of the momentum flux tensor
Πij given by Eq. (12). The normal momentum transmission requires a circula-
tion of the normal velocity [19] with magnitude determined by the transverse
136 Edouard Sonin

force on the normal fluid. This is confirmed by the solution of the Navier-Stokes
equation obtained by Thouless et al. [20]. However, separation on longitudinal
and transverse components of a force should be done with respect to the normal
velocity v n∞ − v L , but not v n − v L . Using Eq. (23), Eq. (2) for the force from
the normal fluid can be rewritten (neglecting the longitudinal force ∝ D) as
 
1 D ln(rm /lph )
F =−   2 (v L − v n∞ ) + D [ẑ × (v L − v n∞ )] .
D ln(rm /lph ) 4πηn
1+ 4πηn
(24)
The transverse component of this force determines the normal circulation at very
large distances:

D κ
κn = dl · v n = −    2 =  2 , (25)
D ln(rm /lph ) κρn ln(rm /lph )
ρn 1 + 4πηn
1 + 4πηn

where we used the value D = −κρn for the Iordanskii force. In the limit of
a strong viscous drag κρn ln(rm /l)/4πηn  1 the transverse force and related
normal circulation are suppressed [20]. But the effect of the superfluid Magnus
force and the longitudinal force ∝ D is also suppressed in this limit.

5 Magnus Force and the Berry Phase

Let us consider now the Berry phase in the hydrodynamic description using the
Lagrangian obtained from Eq. (4) with the Madelung transformation:
κρ ∂φ κ2 ρ V
L= − ∇φ2 − ρ2 . (26)
2π ∂t 8π 2
The first term with the first time derivative of the phase φ (Wess-Zumino term)
is responsible for the Berry phase Θ = ΔSB /, which is the variation of the
phase of the quantum-mechanical wave function for an adiabatic motion of the
vortex around a closed loop [12]. Here
 
κρ ∂φ κρ
ΔSB = dr dt = − dr dt (v L · ∇L )φ . (27)
2π ∂t 2π
is the classical action variation around the loop and ∇L φ is the gradient of the
phase φ[r − r L (t)] with respect to the vortex position vector r L (t). However,
*∇L φ = −∇φ, where ∇φ is the gradient with respect to r. Then the loop integral
dl yields the circulation of the total current j = (κ/2π) < ρ∇φ > for points
inside the loop, but vanishes for points outside. As a result [13],

ΔSB κ
Θ= =V (dl · j) . (28)
 2π
where V is the volume inside the loop (a product of the loop area and the liquid
height along a vortex). Contrary to Eq. (27), the integral in Eq. (28) is related
with the variation of r, the vortex position vector r L being fixed.
 Magnus Force, Aharonov–Bohm Effect, and Berry Phase 137

If the *circulation of the normal velocity vanished, the current circulation

would be (dl · j) = ρs κ, and the Berry phase and the transverse force would
be proportional to ρs (see Geller et al. [13]). However, according to* Sec. 4, at
very large distances the normal circulation κn does not vanish and (dl · j) =
ρs κ + ρn κn . Using a proper value of κn given by Eq. (25), the Berry-phase
analysis yields the same transverse force as the momentum balance. But one
cannot find κn from the Berry-phase analysis, since the latter deals only with
very large distances, whereas κn is determined by the transverse force, which
arises at small distances (in the ballistic region) and appears in the small-distance
boundary condition for the Navier-Stokes equation in the two-fluid region. This
conclusion agrees with the recent analysis by Thouless et al. [20].
Discussions with Lev Pitaevskii and David Thouless during the present work-
shop had a great impact on my view presented in this article. I appreciate also
interesting discussions with Andrei Shelankov. The work was supported by the
grant of the Israel Academy of Sciences and Humanities.

References
1. H. Lamb, Hydrodynamics (Cambridge University Press, New York, 1975).
2. H.E. Hall and W.F. Vinen, Proc. Roy. Soc. A238, 204 (1956).
3. R.J. Donnelly, Quantized vortices in helium II (Cambridge University Press, Cam-
bridge, 1991), Sec. 2.8.3.
4. E.B. Sonin, Rev. Mod. Phys. 59, 87 (1987).
5. P. Nozières and W.F. Vinen, Phil. Mag. 14, 667 (1966).
6. E.M. Lifshitz and L.P. Pitaevskii, Zh. Eksp. Teor. Fiz. 33, 535 (1957) [Sov. Phys.-
JETP 6, 418 (1958)].
7. S.V. Iordanskii, Zh. Eksp. Teor. Fiz. 49, 225 (1965) [Sov. Phys.-JETP 22, 160
(1966)].
8. E.B. Sonin, Zh. Eksp. Teor. Fiz. 69, 921 (1975) [Sov. Phys.-JETP 42, 469 (1976)].
9. Y. Aharonov and D. Bohm, Phys. Rev. 115, 485 (1959).
10. A.L. Shelankov, Europhys. Lett., 43, 623 (1998).
11. P. Ao and D.J. Thouless, Phys. Rev. Lett. 70, 2158 (1993).
12. M.V. Berry, Proc. R. Soc. London A 392, 45 (1984).
13. M.R. Geller, C. Wexler, and D.J. Thouless, Phys. Rev. B 57, R8119 (1998).
14. E.B. Sonin, Phys. Rev. B 55, 485 (1997).
15. H.E. Hall and J.R. Hook, Phys. Rev. Lett., 80, 4356 (1998); E.B. Sonin, ibid. 81,
4276 (1998); C. Wexler et al., ibid. 80, 4357 (1998); 81, 4277 (1998).
16. A more detailed report on the present analysis is to be published in Proceedings
of the the workshop “Microscopic structure and dynamics of vortices in uncon-
ventional superconductors and superfluids” Dresden, Germany, March 2000, cond-
mat/0104221.
17. E.P. Gross, Nuovo Cimento 20, 454 (1961); L.P. Pitaevskii, Zh. Eksp. Teor. Fiz.
40, 646 (1961) [Sov. Phys.-JETP 13, 451 (1961)].
18. S.J. Putterman and P.H. Roberts, Physica 117 A, 369 (1983).
19. Existence of the normal circulation at large distances in the presence of the trans-
verse force on the vortex was pointed out by Pitaevskii (unpublished).
20. D.J. Thouless, M.R. Geller, W.F. Vinen, J-Y. Fortin, and S.W. Rhee, cond-
mat/0101297.
Using the HVBK Model
to Investigate the Couette Flow of Helium II

Karen L. Henderson

Faculty of Computer Studies & Mathematics, University of the West of England,

Bristol, BS16 1QY, U.K.

Abstract. We review the application of the two-fluid HVBK equations to helium II in

Couette geometry, that is flow between concentric, rotating cylinders. This application
is particularly interesting as a large number of experiments have been carried out in this
geometry and also because Couette flow is an exact solution of the HVBK equations
for both the normal fluid and superfluid.

1 Introduction
When the temperature of liquid helium drops below the transition temperature
of Tλ = 2.172 k a phase transition occurs and it becomes a quantum liquid
called helium II. Helium II can be described macroscopically by Landau’s two-
fluid model in which it is considered to be a mixture of a viscous normal fluid
and an inviscid superfluid. In addition, vortex lines appear in the superfluid
component when helium II rotates or when it moves along a tube faster than
a small critical velocity. Feynman [1] showed that the circulation about each
individual vortex line is quantised, taking the value of Γ = 9.97 × 10−4 cm2 /sec.
The most generally accepted equations for modelling the macroscopic flow of
helium II are the Hall-Vinen-Bekharevich-Khalatnikov (HVBK) equations which
were derived by a number of people over the years [2,3,4,5,6]. These equations
extend Landau’s two-fluid model to take into account the presence of quantized
vortex lines in the flow. The derivation of the equations is based on a continuum
approximation, assuming a high density of vortex lines, all aligned roughly in
the same direction.
The incompressible isothermal HVBK equations of motion of the two fluids
are

∂vn ρs
+ (vn ·∇)vn = −∇pn + νn ∇2 vn + F , (1a)
∂t ρ
∂vs ρ n
+ (vs ·∇)vs = −∇ps − νs T − F, (1b)
∂t ρ
∇· vn = 0, ∇· vs = 0. (1c)

where vn and vs are the velocity profiles of the normal fluid and superfluid
respectively, ρn and ρs the normal fluid and superfluid densities, ρ = ρn + ρs
helium’s total density, pn and ps effective pressures and νn the normal fluid

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 138–145, 2001.

c Springer-Verlag Berlin Heidelberg 2001
 Using the HVBK Model to Investigate the Couette Flow of Helium II 139

kinematic viscosity. The relative amount of normal fluid and superfluid present
in the flow depends on the temperature T of the liquid: if T → Tλ then ρs /ρ → 0
and ρn /ρ → 1; if T → 0 then ρs /ρ → 1 and ρn /ρ → 0.
The mutual friction force, F , describes the interaction between the normal
fluid and the vortex lines and is given by

F = 12 B ω
 × (ωs × (vn − vs − νs ∇× ω
 )) + 12 B  ωs × (vn − vs − νs ∇× ω
s s s
 ), (2)

where ωs represents the superfluid vorticity and is a measure of the number

and direction of vortex lines contained in a given small region of fluid. The term
ω s = ωs /|ωs | represents the unit vector in the direction of superfluid vorticity.
B and B  are temperature dependent coefficients which describe the interaction
between the normal fluid and the vortices [7,8].
The vortex tension force, −νs T , describes the energy in the vortex lines and
is such that
T = ωs × (∇× ω  s ), (3)
where ν s = (Γ/4π) log(b0 /a0 ) is the vortex tension parameter, a0 is the vortex
core radius and b0 = (|ωs |/Γ )−1/2 is the intervortex spacing.
In this paper we review the application of the HVBK equations to Taylor-
Couette flow, that is flow between two concentric rotating cylinders. Taylor-
Couette flow has been used as a bench-mark for fluid mechanics since Taylor’s [9]
pioneering work to investigate the transition from Couette flow to Taylor vor-
tices, which established a firm ground for using the Navier-Stokes equations and
the no-slip boundary conditions. Progress in helium II has been slower than for
classical fluids due in part to problems of flow visualization at such low temper-
atures. In considering a classical fluid, introduction of flakes or other small par-
ticles into the working fluid (usually oil or water), results in the Taylor vortices
being clearly evident. In contrast there are only limited visualisation techniques
available to the experimentalist at temperatures close to absolute zero. Recent
attempts have been made to reveal the flow pattern of helium II by adding small
particles [10]. However this was only successful at high rotation rates (40 times
the critical angular velocity at which linear stability analysis predicts Couette
flow becomes unstable).
Experiments on helium II between concentric cylinders were first performed
by Kapitza [11] in 1941 and Donnelly & LaMar [12] have written a review of
experiments involving helium II in Couette apparatus. We shall expand on two
types of experiments performed, which have been used to compare theoretical
predictions with. Early Taylor-Couette experiments were concerned with deter-
mining the viscosity of helium II by measuring the torque exerted by the flow
on the stationary cylinder. A break in the linear dependence of the torque with
the angular velocity of the rotating cylinder is taken to denote a transition from
one solution to another. The second experimental technique, that of measuring
the extra attenuation of a second sound wave, can be used to probe the super-
fluid vorticity. Second sound waves occur when there is a periodic counterflow
between the normal fluid and superfluid, which corresponds to a wave of heat.
Angular velocity is plotted against the attenuation factor and breaks in the curve
140 K.L. Henderson

are interpreted as transitions in the flow. By measuring the extra attenuation of

second sound waves in the axial, azimuthal and radial directions it is theoreti-
cally possible to get an idea of the number and direction of the quantized vortex
lines. In practice the information obtained is less complete than this.
In this paper we restrict our attention to rotation of the inner cylinder only,
keeping the outer cylinder fixed. The usual simplifying assumption of infinite
cylinders is adopted in Sects. 2 and 3.1 where we consider the stability of he-
lium II Couette flow and the nonlinear flow of helium II beyond this transition
respectively. In Sect. 3.2 we consider the basic flow of helium II between finite
cylinders with stationary endcaps.

2 Linear Theory

In the case of flow between infinite cylinders with inner radius R1 rotating with
angular velocity Ω and stationary outer radius R2 , Couette flow, whose velocity
profile is given by:
v c = (A + C/r) er (4)
is an exact solution of the HVBK equations for both the normal fluid and su-
perfluid. This is provided that Ω is greater than a small critical value at which
vortex lines first appear in the gap. A and C are constants depending on R1 ,
R2 and Ω determined by the no-slip condition imposed on the normal fluid at
the cylinder walls. The superfluid, being inviscid is not required to satisfy such
boundary conditions. The only restriction is that there is no penetration through
the boundary. From (4) it can be seen that the superfluid vorticity is purely ax-
ial and has magnitude 2|A|, thus the vortex lines are aligned in the direction of
rotation.
Early attempts to theoretically examine the stability of Couette flow theo-
retically were made by Chandrasekhar & Donnelly [13], however the issue has
only recently been resolved fully by Barenghi & Jones [14] and Barenghi [15].
They performed a linear stability analysis on the HVBK equations. The Couette
state was linearly perturbed and they numerically calculated the critical angular
velocity, Ωc , and corresponding critical axial wavenumber at which Couette flow
becomes linearly unstable. We summarize their findings below:

• the axisymmetric mode onsets first,

• the stability of helium II Couette flow is more stable than classical Couette
flow in the high temperature region but less stable in the low temperature
one,
• the critical axial wavenumber decreases as the temperature decreases, becom-
ing zero at relatively high temperatures.

For a classical fluid the critical axial wavelength at which Couette flow becomes
unstable and Taylor vortices form occurs at kc ≈ 3.1, in other words the result-
ing Taylor vortices are approximately square. This is not the case for helium II.
As the temperature decreases below the lambda temperature of Tλ = 2.172k the
 Using the HVBK Model to Investigate the Couette Flow of Helium II 141

critical axial wavenumber decreases resulting in an elongation of the Taylor cells.

This effect is strongly temperature dependent and the critical axial wavenum-
ber becomes zero at a relatively high temperature for certain parameter ranges
(e.g. T = 2 k).
Compared to the classical case T = 2.172k, the stability of helium II is
initially enhanced as the temperature decreases below the transition tempera-
ture, due to the tension in the vortex lines. However the stability is dramatically
reduced as the temperature drops further. These results prompted further exper-
iments which were performed by Swanson & Donnelly [16]. They carried out a
series of second sound experiments at temperatures close to Tλ . Comparison be-
tween theory [15] and experiments [16] gave excellent agreement, particularly for
temperatures close to the lambda temperature. Although the qualitative picture
was correct at lower temperatures it is not realistic to expect such good quanti-
tative agreement here, due to the breakdown of the infinite cylinder assumption.
For T < 2.07 k linear theory predicts that the critical axial wave number is zero.
In such a region, the Taylor cells would be so elongated that only a few would be
present in the apparatus and end effects would undoubtedly become important.
It was only at this stage that the validity of the HVBK equations was con-
firmed. This led the way for further research in this area, namely numerically
solving the nonlinear two-fluid HVBK equations of motion of helium II.

3 Nonlinear Solutions
3.1 Infinite Cylinder Assumption
Using the infinite cylinder assumption, Henderson, Barenghi & Jones [17] nu-
merically solved the HVBK equations for the first time to obtain the nonlinear
flow of helium II between infinite cylinders. The aim of the work was to ob-
tain solutions for helium II corresponding to what would correspond to Taylor
vortices in a classical flow and to investigate what happens to the vortex lines.
Axisymmetric solutions were considered since linear theory predicts that the ax-
isymmetric mode onsets first. To solve equations (1a-1c), boundary conditions
are required. The nonlinear problem is 6th order in both the normal fluid and
superfluid, however the linear problem is only 2nd order in the superfluid. Thus
two further boundary conditions for the superfluid are needed in addition to the
no penetration of the boundary used successfully in the linear stability analysis.
The extra boundary conditions employed were

ωφs = 0 at r = R1 and R2 (5a)

vφs = Ωr at r = R1 and R2 . (5b)

Equations (5a,5b) force the superfluid vorticity to be purely axial at the cylinder
walls. This is consistent with Couette flow, in which the vortex lines are purely
axial throughout the flow and results in the mutual friction being small at the
boundaries which is an advantage numerically. The normal fluid satisfies the
standard no slip boundary conditions as for the linear model.
142 K.L. Henderson

The HVBK equations were solved numerically using a pseudospectral method,

based on expansions in Chebychev polynomials in the radial direction and trigono-
metric functions in the axial direction [18]. Results were obtained for angular
velocities of up to 15% above the critical angular velocity at which linear theory
predicts that Couette flow becomes unstable. Apart from the axial stretching of
the Taylor cells, the normal fluid displays a velocity profile similar to that of a
classical fluid, However the superfluid velocity profile is markedly different to the
classical case; instead of a meridional flow consisting of single pair of cells in each
period, we find a more complex pattern of eddies and counter-eddies. Perhaps
what is of most interest is the orientation of the vortex lines. The numerical
results show that the superfluid vorticity is still predominantly axial and the
deflection in the azimuthal direction is smaller than that in the radial direction.
Considering the (r, z) plane the vortex lines are most densely situated near the
inner cylinder at positions of maximum inflow and are deflected towards the
outer cylinder at the centre of the cell, where there is outflow.
There are two ways of comparing the numerical results with experiments.
Firstly by comparing the additional attenuation of a second sound wave due to
the vortex lines and secondly by measuring the torque exerted on the outer cylin-
der. We compared the relative change in the azimuthal attenuation coefficient
at an angular velocity of 5.4% above the onset of Taylor vortices and found an
order of magnitude correspondence between the experimental value of Swanson
& Donnelly [16]. Our predicted value was lower than the observed value which
could be due to the following points:
• The observed attenuation will depend on the spatial structure of the mode
used to probe the flow.
• The observed attenuation does not discriminate the sense in which the vor-
tex lines point, however the HVBK equations are derived using an averaged
approach to the vorticity of the superfluid.
Although many experiments have been carried out to measure the torque in
helium II, few experiments have been carried out in a parameter range such that
the stability curve has a minimum at non-zero axial wavenumber. We compare
our torque measurements with those of the experiment of Donnelly [19] in which
values of the torque above the transition are reported. At small angular velocity,
in the Couette flow regime, the torque is proportional to the viscosity μ of the
fluid. At higher angular velocities there is a break in the curve, corresponding to
the onset of Taylor vortices, the torque exerted on the outer cylinder increases
as the axisymmetric taylor vortices are more efficient in transferring angular mo-
mentum than the azimuthal Couette motion. Excellent agreement was found [20]
between the calculation and the experimental data in the nonlinear regime for
Ω > Ωc .
These two results validated the HVBK equations for the first time in the
nonlinear regime. Although comparisons with available experiments in the non-
linear regime are encouraging, it is clear that end effects become important, even
at relatively high temperatures.
 Using the HVBK Model to Investigate the Couette Flow of Helium II 143

3.2 Unit Aspect Ratio

In order to be able to compare with further experimental data, end effects need
to be included in the model. Henderson & Barenghi [21] considered helium II
contained within a cylindrical annulus of inner radius R1 , outer radius R2 , height
H where the gap between the cylinders has been chosen such that H = R2 −
R1 . Thus the Couette annulus has unit aspect ratio, in that the gap between
the cylinders is equal to the height between the endcaps. The inner cylinder
rotates with constant angular velocity Ω, whilst the outer cylinder and two end
plates are stationary. This simple flow configuration enabled us to study how
the vortex lines respond to a shear in the presence of boundaries which are both
parallel and perpendicular to the natural axial direction of the vortex lines. The
axisymmetric form of the HVBK equations (1a-1c) were solved using a finite
difference approach taking a regular grid in both the r and z direction. The
boundary conditions on the curved cylinder walls were taken to be the same as
for the infinite cylinder case. However extra boundary conditions are also needed
on the two endcaps, z = 0, H. For the normal fluid standard no slip boundary
conditions were imposed. Whilst for the superfluid the following were used

vzs (r, 0) = vzs (r, H) = 0, (6a)

ωrs (r, 0) = ωrs (r, H) = 0, (6b)
ωφs (r, 0) = ωφs (r, H) = 0. (6c)

The first condition (6a) ensures that there is no penetration of the superfluid
through the boundary, whilst the last two conditions (6b,6c) correspond to per-
fect sliding of the vortex lines as discussed by Khalatnikov [5].
The main result of this investigation is the anomalous motion of helium II
when compared to the motion of a classical fluid. The velocity profile obtained
is a superposition of an azimuthal motion vφ around the inner cylinder and a
toroidal motion vr and vz in the vertical plane. The latter motion is in the form
of a pair of cells similar to a Taylor vortex pair, but being caused by boundaries
rather than a centrifugal instability, it is hereafter referred to an Ekman cell
pair. The first interesting finding is that vφs is almost z-independent, that is
the superfluid moves around the cylinders in a column-like fashion, which is
due to the tension in the vortex lines. This effect becomes more pronounced at
lower temperatures when the superfluid component is higher as is illustrated in
Fig. 1a,b. Each figure extends over the whole computational domain with the
inner cylinder and outer cylinder on the left and right respectively. In contrast,
vφn exhibits strong z-dependence due to the no-slip boundary conditions imposed
on the normal fluid at the ends and walls of the cylinders and has a similar profile
to that of a classical fluid, see Fig. 1c.
144 K.L. Henderson

Fig. 1. Azimuthal motion vφ of (a) the superfluid at T =1.8 K, (b) the superfluid at
T =2.17 K, (c) the normal fluid at T =2.17 K. Lighter regions correspond to larger
magnitude.

Fig. 2. Motion of helium II compared to a classical fluid. Classical: (a) radial veloc-
ity vr ; (b) axial velocity vz . Helium II at T =2.11 K: (c) vrn ; (d) vzn ; (e)vrs ; (f) vzs .
Lighter/darker regions correspond to positive/negative velocities.

The second interesting finding comes from looking at the Ekman cells in both
the normal fluid and superfluid. In a classical fluid the two Ekman cells form with
outflow at the centre and inflow at the ends of the cylinder, as in Fig. 2a,b. The
results for helium II are quite different. We find that the superfluid Ekman cells
 Using the HVBK Model to Investigate the Couette Flow of Helium II 145

always rotate in a counter-classical way due to the mutual friction force, that is
outflow occurs at the ends of the cylinder with inflow at the centre, see Fig. 2c,d.
It is also seen that the normal fluid Ekman cells rotate in a counter-classical way
at lower temperatures, see Fig. 2e,f, but revert to a classical direction close to the
transition temperature Tλ = 2.172k as one would expect. We also investigated
the magnitude and direction of the superfluid vorticity in order to gain a picture
of how the vortex lines are situated in the flow. As for the infinite cylinder case,
the superfluid vorticity is primarily axial with small deflections in both the r
and φ direction. However the vorticity is concentrated near the inner rotating
cylinder, which is a measurable result.

4 Discussion
The excellent agreement between the linear stability analysis [15] and exper-
imental data [16] was a rigorous test of the validity of the HVBK equations
at least in the linear regime. The HVBK model has been validated further by
the good agreement between the nonlinear calculation [18] and experimental
data [19]. Obtaining nonlinear solutions for the flow of helium II between infi-
nite and finite cylinders has allowed us to gain more insight into the flow which,
because of the low temperature environment, cannot be observed directly like
a classical fluid. It has also enabled us to explore the boundary conditions for
the superfluid. A possible future direction of the work would be to extend the
aspect ratio in order to investigate the transition to Taylor cells.

References
1. R.P. Feynman: ‘Application of quantum mechanics to liquid helium.’ In P rogress
in Low Temperature Physics 1. (C.J. Gorter, North Holland 1955)
2. H.E. Hall, W.F. Vinen: Proc. Roy. Soc. London A 238, 215 (1956)
3. H.E. Hall: Phil. Mag. Suppl. 9, 89 (1960)
4. I.L. Bekharevich and I.M. Khalatnikov: Sov. Phys. JETP 13 , 643, (1961)
5. I.M. Khalatnikov: An Introduction to the Theory Superfluidity. (Benjamin 1965)
6. R.N. Hills, P.H. Roberts: Arch. Rat. Mech. Anal. 66, 43 (1977)
7. C.F. Barenghi, R.J. Donnelly, W.F. Vinen: J. Low Temp. Phys. 52, 189 (1983)
8. D.C. Samuels & R.J. Donnelly: Phys. Rev. Lett. 65, 187 (1990)
9. G.I. Taylor: Phil. Trans. Roy. Soc. Lond. A 223, 289 (1923)
10. F. Bielert, G. Stamm: Cryogenics 33, 938 (1993)
11. P.L. Kapitza: J. Phys. USSR 4, 181 (1941)
12. R.J. Donnelly, M.M. Lamar: J. Fluid Mech. 186, 163 (1988)
13. S. Chandrasekhar, R.J. Donnelly: Proc. Roy. Soc. London A 241, 9 (1957)
14. C.F. Barenghi, C.A. Jones: J. Fluid Mech. 197, 551 (1988)
15. C.F. Barenghi: Phys. Rev. B 45, 2290 (1992)
16. C.J. Swanson, R.J. Donnelly: Phys. Rev. Lett. 67, 1578 (1991)
17. K.L Henderson, C.F. Barenghi, C.A. Jones: J. Fluid Mech. 283, 329 (1995)
18. K.L. Henderson, C.F. Barenghi, J. Low Temp. Phys. 98, 351 (1995)
19. R.J. Donnelly: Phys. Rev. Lett. 3, 507 (1959)
20. K.L. Henderson, C.F. Barenghi: Phys. Lett. A 191, 438 (1994)
21. K.L. Henderson and C.F. Barenghi: J. Fluid Mech. 406, 199 (2000)
An Introduction
to the Theory of Superfluid Turbulence

W.F. Vinen

School of Physics and Astronomy, University of Birmingham, Birmingham B15 2TT,

UK, and
Cryogenic Helium Turbulence Laboratory, Department of Physics, University of
Oregon, Eugene, Oregon 97403, USA

1 Introduction
In one sense superfluid turbulence is an old subject: it was mentioned as a theo-
retical possibility by Feynman in 1955[1]; and it has been known experimentally
since the early 1950s that flow of the superfluid component of helium II can
become turbulent when there is a steady counterflow of the two fluids, such as
occurs in a steady heat current[2]. The original experimental discovery was ac-
companied by the beginnings of a theory[3], and this theory has been developed
steadily, especially by Schwarz[4,5], so that many aspects of this type of turbu-
lence are now well understood. However, counterflow turbulence has no classical
analogue, and it has attracted little interest from those who study classical fluid
mechanics. Types of flow for which classical analogues do exist were observed
by low temperature physicists for many years, but the presence of the two fluids
were thought to make them very complicated, and they were not therefore stud-
ied in detail. More recently experiments have been reported on the analogue of a
rather simple case of classical turbulence, namely that produced by steady flow
through a grid[6,7]. In the classical analogue the turbulence is approximately ho-
mogeneous and isotropic, and its study has been important in the development
an understanding of classical turbulence[8]. The superfluid analogue promises to
be equally important.
In this paper I shall first describe some aspects of the theory of counterflow
turbulence. But I shall then devote most of the paper to grid turbulence, where
the theory is less well developed, although I shall make use of an important
experimental result obtained with a more complicated type of flow generated by
two counter-rotating discs[9]. My aim is to stimulate interest in the theory of
superfluid turbulence, particularly, at this stage, in the simple case of grid tur-
bulence, among both low temperature physicists and those with a background in
classical fluid mechanics. I shall focus on open questions and unsolved problems,
questions and problems that are clearly seen in grid turbulence, but which are
more widely relevant. My own background is in experimental quantum fluids,
and certainly not in theoretical fluid mechanics. I shall tend to speculate about
what I see as the physics of superfluid turbulence, and others will tell me where
my physical intuition is unreliable or, hopefully, where it can be developed along
more rigorous lines.

C.F. Barenghi, R.J. Donnelly, and W.F. Vinen (Eds.): LNP 571, pp. 149–161, 2001.

c Springer-Verlag Berlin Heidelberg 2001
150 W.F. Vinen

Superfluid grid turbulence illustrates in perhaps its simplest form what is

sometimes known as “vortex coupled superfluidity”; a turbulent state in which
the superfluid and normal components seem to be coupled together and behave
like a single classical fluid[10]. It has gradually become apparent that this type
of superfluid turbulence is quite common in cases where the two fluids are not
forced to move with different velocities. The ideas that we develop here in the
context of grid turbulence are likely therefore to be widely applicable.
A turbulent superfluid velocity field must take the form of a tangled array of
quantized vortex lines (circulation κ = h/m4 ). Motion of the lines is governed
by the classical Magnus effect. A force of mutual friction, f per unit length of
line, acting effectively on the core of the line normal to its length, arises from
the scattering of thermal excitations[11]. We shall assume that f = γ (vn − vL ),
where vn is the component of the velocity of the normal fluid perpendicular to
the length of the vortex and vL is the velocity of the line; for simplicity we ignore
any transverse component of f . Numerical values of γ are given in reference[12].
A number of different approaches have been used to develop our understand-
ing of superfluid turbulence. One relies on simple physical arguments backed
by dimensional analysis and ideas of dynamical similarity; another relies on
computer simulations. Perhaps there will in future be one that is based on rig-
orously established general principles; the fact that the turbulent velocity field
in the superfluid component is due entirely to discrete quantized vortex lines
may facilitate such a development (see the contribution of Gary Williams to
this workshop). The simulations are of two types: those pioneered by Schwarz[4]
and based on the idea that, except on very short length scales, quantized vor-
tices can be regarded as vortex filaments moving according to classical fluid
mechanics, with inclusion of the force of mutual friction f , which modifies the
motion through the Magnus effect; and those based on the Gross-Pitaevskii or
non-linear Schrodinger equation (NLSE), which includes a quantum description
of the vortex core, of the thermal excitations, and of their interaction, albeit one
that cannot be expected to be quantitatively applicable to liquid helium. For
many purposes the classical vortex filament approach is probably sufficient, but
it cannot give a fully satisfactory description of reconnections, which play an
important role in superfluid turbulence.
We shall confine our discussion to turbulence in superfluid 4 He; the case for
experiments on turbulence in superfluid 3 He is mentioned briefly in reference[13].
Experiments in which superfluid turbulence has probably been observed in 3 He
are reported by Fisher at this Workshop.

2 Counterflow Turbulence
In counterflow turbulence the vortex tangle is believed to be at least approxi-
mately homogeneous, provided that the average velocities of the two fluids, Vn
and Vs , are spatially uniform. The turbulence is maintained by the mutual fric-
tion. If the vortex lines move on average with the superfluid, the total average
force of mutual friction per unit volume, Fns , is equal to γL (Vn − Vs ), where L
 The Theory of Superfluid Turbulence 151

is the length of line per unit volume, and where we have ignored factors of order
unity arising from the random orientation of the lines. We can illustrate one
approach to superfluid turbulence by deriving the dependence of L on (Vn − Vs )
from a principle of dynamical similarity[3,14,15].
Let us assume that the vortex tangle is characterized by a single length,
 = L−1/2 , characteristic of both the vortex line spacing and the vortex radius
of curvature. Taking into account the Magnus effect and the force f , we can
easily show that the velocity with which any element of line moves is given by

vL = Vs + vs + ακ̂ × (Vn − Vs − vs ) , (1)

where vs is the superfluid velocity at the element due to the rest of tangle,
κ̂ is the unit vector parallel to the element, and α ≈ γ/ρs κ. There are two
contributions to the magnitude of vs : vs1 ≈ κ/2π, which is due to neighbouring
lines at the distance  from the element; and vs2 ≈ (κ/4π) ln (/ξ0 ), which is
due to the local curvature of the line. (We note in passing that computations
based on the“localized induction approximation”take into account only the latter
contribution.)
Suppose that we change the length scale in the vortex tangle by a factor g, so
that  → g, and let us ignore the logarithmic factor in vs2 . Then vs changes by
the factor g −1 . If we change Vs and Vn by the same factor, we see from equation
(1) that vL is also changed by the same factor (formally, we change the length
scale by g and the time scale by g 2 ). Therefore the whole tangle evolves in the
same way as it would before scaling, except for the scaling factor g, suggesting
that there is a principle of dynamical scaling. Application of this principle shows
−1
easily that  is proportional to (Vs − Vn ) , so that the mutual friction per
3
unit volume is proportional to (Vs − Vn ) , as is observed to be approximately
true. Taking into account the logarithmic term in vs2 introduces logarithmic
corrections, which are indeed probably observed.
As we have mentioned, the superfluid turbulence is maintained by the mutual
friction acting on individual elements of line. That the length of line can in
principle increase as a result of the mutual friction is clear from the fact[3]
that a vortex ring can grow if the self induced velocity of the ring is in the
direction of, but less in magnitude than, (Vn − Vs ); otherwise it will shrink.
Appropriately oriented parts of the tangle with low curvature can behave in a
similar way, although it is not obvious that a steady turbulent state (a finite
L) can be maintained. A detailed understanding of counterflow turbulence came
only from the simulations of Schwarz[4,5], which showed that a steady turbulent
state can be achieved through the effect of reconnections, which generate vortex
configurations that favour the growth of line.
This theory of counterflow turbulence is based on the assumption that the
flow of the normal fluid is laminar. The turbulent flow of the superfluid must
then occur on length scales not significantly larger than ; flow on larger length
scales would be damped out by mutual friction. Recently it has been suggested by
Melotte and Barenghi[16] that the laminar flow of the normal fluid in counterflow
may not always be stable, so that both fluids may become turbulent, probably
152 W.F. Vinen

on length scales significantly larger than . The theory for such a regime presents
us with a major challenge, which we mention again in Sect. 3.5.

3 Grid Turbulence in Superfluid Helium

Experiments on grid turbulence in superfluid helium at temperatures above 1K
have been described by Niemela at this workshop; the original measurements
were reported in reference[7]. I shall first remind you of the experimental results
and of the idea that they are consistent with a quasi-classical model, as discussed
by Skrbek at this workshop. Then I shall discuss why this model may work.
Finally, I shall extend the discussion to lower temperatures, where there are so
far no really satisfactory experimental results, but where new and potentially
interesting problems arise.

3.1 Measurements of the Decay of Vortex Lines,

and the Quasi-classical Model

In the experiments a grid is towed through the helium, and a measurement is

made of the time dependence of the excess attenuation of second sound in a small
fixed region in the helium behind the grid. The excess attenuation is caused by
mutual friction associated with vortex lines in the turbulent superfluid behind
the grid, and the measurements lead to data showing the decay of the line density,
L, with time. The grid moves through the helium at a velocity exceeding both
that required to create vortex lines (the mechanism need not concern us), and
that required to create turbulence in the normal fluid. Turbulence is therefore
produced in both fluids. The average velocity of each fluid must vanish (there
is nothing to maintain a counterflow, and any transient counterflow would be
damped rapidly by mutual friction), so the situation is quite different from that
discussed in Sect. 2.
As explained by Skrbek, the observed decay in L is consistent with a quasi-
classical model, based on the following two assumptions. First, on length scales
larger than the vortex line spacing, , which turns out to be close to the length
scale at which viscous dissipation is expected to occur in the turbulent normal
fluid, the two fluids are coupled together in the sense that the two velocity fields
are the same; the coupled fluids behave like a single classical fluid, the turbu-
lence exhibiting an inertial range of wavenumbers in which there is negligible
dissipation and in which the energy spectrum has the classical Kolmogorov form

E(k) = C2/3 k −5/3 , (2)

where C is a constant of order unity, and  is the rate of energy dissipation
per unit mass of helium at a high wavenumber, presumably of order or greater
than −1 . (Roughly speaking, energy is injected from the grid into eddies with
size of order the mesh size (wavenumbers of order the reciprocal of the mesh
size); non-linear terms in the equation of motion cause the energy to flow to
 The Theory of Superfluid Turbulence 153

smaller eddies (higher wavenumbers) in a cascade, until it can be dissipated by

viscosity.) Secondly, this rate of dissipation is given by
 = ν  κ2 L2 , (3)
where ν  has the dimensions of kinematic viscosity. If we interpret κ2 L2 as an ef-
fective mean square vorticity in the superfluid, which is of doubtful validity[17,13],
equation (3) is reminiscent of the rate of viscous dissipation in a turbulent clas-
sical fluid. Of course, as in a classical fluid, the Kolmogorov spectrum may rep-
resent an oversimplification, which fails to take account of, for example, inter-
mittency (see, for example, reference[18]); but it seems reasonable to ignore this
point at the present stage in the development of our understanding of superfluid
turbulence.
It is important to appreciate that we can say only that the experimental data
on grid turbulence are consistent with this quasi-classical model. Other models
might also be consistent: for example, one in which both the dissipation and
the energy spectrum take different forms (the energy spectrum could involve the
quantum of circulation). But the quasi-classical model is perhaps the simplest
model that will account for the experiments, and, as we shall argue, it can be
given some theoretical backing and is consistent with other experiments.
There is clearly a need to find direct experimental evidence for the valid-
ity of the Kolmogorov spectrum (2) in superfluid grid turbulence; i.e. a direct
measurement of the turbulent energy spectrum. We know of no such evidence
for grid turbulence, but evidence does exist for a more complicated type of flow
generated by two counter-rotating discs. This is from the important work of
Maurer and Tabeling[9], who measured pressure fluctuations in this type of flow
(on a rather course length scale), both above and below the superfluid phase
transition. They find that over a certain range of frequencies these fluctuations
do have a Kolmogorov spectrum, the spectrum being the same above and be-
low the transition. It seems likely therefore that the value of the Kolmogorov
constant, C, is the same above and below the phase transition and equal to the
value expected for a classical fluid. We assume that a similar result will hold for
grid turbulence. In that case the quasi-classical model of grid turbulence would
account unambiguously for the experimental results.

3.2 Superfluid Turbulence on Length Scales

Larger than the Vortex Line Spacing
Consider flow of the superfluid component on length scales large compared with
, and suppose that this flow can be achieved by a relatively small re-arrangement
of the vortex tangle, as turns out to be the case in grid turbulence[17]. Such a flow
can be characterized by a velocity field that is similar to that found in a classical
fluid, and we suggest that the dynamical behaviour of the superfluid is then sim-
ilar to that of a classical fluid with the same density at high Reynolds number.
An example, not involving turbulent flow, where the truth of this suggestion can
be verified is provided by wave motion in an otherwise uniformly rotating super-
fluid, which contains a uniform array of lines. Waves with wavevectors directed
154 W.F. Vinen

parallel to the axis of rotation have a character that depends on the magnitude
(q) of the wavevector relative to the line spacing . If q  1 the waves are Kelvin
waves on the individual vortices; if q  1 the waves become indistinguishable
from the classical “inertia waves”found in a classical rotating liquid (see, for
example, references[19,20]. In the case of turbulent flow we must remember that
the non-linear term in the Navier-Stokes equation couples motion on different
length scales, so that the validity of our suggestion depends on the hypothesis
of the “independence of Fourier components for distant wavevectors”[8]. But we
know of no formal proof of this validity.
We emphasize that this similarity between superfluid and classical flow can-
not extend to wavenumbers of order or greater than −1 , where the discrete
nature of the vorticity cannot be ignored.

3.3 The Turbulent Energy Spectra in Superfluid Grid Turbulence

We can now start to see some theoretical justification for the quasi-classical
model. We see that when the superfluid component flows with sufficient speed
through a grid it could lead to turbulence in that component, characterized by
the Kolmogorov spectrum for k  −1 , as in a classical fluid. When both fluids
flow through the grid two such turbulent flows could be generated. However, the
presence of vortex lines in the superfluid component gives rise to mutual friction,
which must tend to couple the two fluids. It can be shown[17] that, provided the
normal fluid flow is not significantly affected by viscosity, and provided that
k  −1 , this friction is sufficient to ensure that the two fluids have associated
with them the same velocity fields. (The demonstration involves a proof that the
time required for mutual friction to eliminate relative motion in the two fluids on
a length scale k −1 is considerably less than the (“turnover”) time for a turbulent
eddy of size k −1 to lose its energy by inertial transfer to other eddies.) Given, as
we have seen, that viscous dissipation is expected to occur in the normal fluid
only for k ≥ −1 , we can begin to understand the success of the quasi-classical
model.
The type of coupled motion of the two fluids that we are discussing here
relates of course to the vortex coupled superfluidity that we mentioned in Sect. 1.
The basic idea of such coupling is presumably more widely applicable than to the
simple case of grid turbulence; examples are provided by the flow investigated by
Maurer and Tabeling[9], to which we have already referred, and to that round a
sphere investigated recently by Smith et al[21]. The theory advanced here relies
on very general arguments; support from the theory of specific types of flow,
albeit very idealised, has been provided by Barenghi and his colleagues[33,23].
We emphasize that on length scales comparable with or less than , where
the flow of the superfluid component is strongly constrained by the fact that
vorticity can be associated only with discrete quantized vortex lines, the two
velocity fields cannot be the same, even in the absence of viscous dissipation in
the normal fluid.
The quasi-classical model requires not only the coupled motion of the two
fluids for k  1 but also the validity of equation (3) for the total rate of
 The Theory of Superfluid Turbulence 155

dissipation of turbulent energy, both fluids contributing to this energy. It is far

from obvious that equation (3) is correct.
There is also the question of the value of ν  . It turns out[7] that the existing
experiments on grid turbulence yield only the ratio C 3 /ν  . There has been much
private discussion about the value of the Kolmogorov constant C that ought to
be used. In the case of turbulence in a classical fluid C seems to be universal
(within fairly large experimental error and at reasonably high Reynolds number)
and equal to about 1.6[24]. The suggestion has been made that in the case of
superfluid grid turbulence we are dealing with a quantum liquid, so that C might
take a different value, which is perhaps temperature dependent. However, if we
accept evidence from the experiments of Maurer and Tabeling[9], to which we
have already referred, then C has its classical, temperature-independent, value.
We conclude therefore, at least for the present, that it is sensible to use the
value C = 1.6 to deduce the value of ν  from experiment, as is done by Niemela
and Skrbek at this Workshop. We remark that this value has the same order
of magnitude as ηn /ρ, where ηn is the viscosity of the normal fluid and ρ is
the total density of the helium, but that its temperature dependence is quite
different. Further development of our theoretical discussion requires therefore
both a demonstration that equation (3) is at least reasonable and some discussion
of the observed value and temperature dependence of ν  . It turns out that this
discussion is best postponed until we have discussed superfluid turbulence at
very low temperatures.
It may be relevant to add that the classical Kolmogorov spectrum does not
depend for its validity on the Navier-Stokes equation; it depends only on the prin-
ciple that the statistical properties of the turbulence are determined uniquely
and universally by the length scale involved and the rate of energy dissipation at
high wavenumbers. But it does depend on the absence of quantum effects, which
would introduce Planck’s constant, probably in the form of the quantum of cir-
culation, κ, as a relevant parameter. If quantum effects were important, the spec-
trum of the turbulent energy might take the form E(k) = 2/3 k −5/3 H(κ−3 k −4 ),
allowed by dimensional analysis, where H is some unknown function. Although
this may not be ruled out by the experiments on superfluid grid turbulence, it
does seem to be inconsistent with the experiments of Maurer and Tabeling[9].

3.4 Superfluid Turbulence at Very Low Temperatures

The experiments on grid turbulence that we have discussed so far were carried
out at temperatures above 1K, where there is a significant fraction of normal
fluid. We focus next on lower temperatures, including those so low that the frac-
tion of normal fluid can be neglected. No experiments have yet been reported on
superfluid turbulence produced by steady flow through a grid at low tempera-
tures. Preliminary study of turbulence produced by an oscillating grid has been
reported by Davis et al[25], and the latest results will be reported at this Work-
shop by McClintock; but this type of turbulence is not homogeneous and may
be difficult to interpret at this stage. Two simulations relevant to low tempera-
tures have been reported. Tsubota et al[26] have studied the decay of turbulence
156 W.F. Vinen

by the Schwarz technique. The turbulence is produced initially by simulated

counterflow at a high temperature: the counterflow is then turned off; the tem-
perature is changed to the value required; and the decay is then simulated. Nore
et al[27] perform a simulation relevant to zero temperature based on a solution
of the NLSE, with a Taylor-Green vortex as an initial flow. Our own approach
to low-temperature grid turbulence[17], which is quite specula