SRF 910927-07

Basic Concepts of Measurements Made on Superconducting RF Cavities

Jens Knobloch, Laboratory of Nuclear Studies, Cornell University, August 1991

Presented here is an introduction to some of the concepts used in the tests performed on superconducting RF cavities in the SRF
group at Cornell. It is intended for the 'novice' to this field and by no means attempts a rigorous derivation. Assumed is some previous
exposure to the physics of waveguides and cavities. Explained is the concept of the Quality (Q) of a cavity, the coupling parameters and
how these quantities are determined by experiment. Furthermore a short overview of the results obtained for a typical cell is presented.

dU
1. THE THEORETICAL CONCEPTS Ptotal = - (4)
dt
1.1 THE QUALITY OF A CAVITY and we obtain U(t) by combining (4) with (2) to give:
The main task of the experiments is to determine
U(t) = Uo e-ωt/QL (5)
the performance of superconducting niobium cavities.
A possibility is to measure the so called unloaded
(5) tells us that U falls to 1/e in a time τ = Q L/ω which
Quality (Q o ) of the cavity as a function of the peak
can be measured experimentally to give the loaded Q.
electric field within at a given resonance frequency. Qo
We will return to this later on.
is defined by:
1.2 THE WIDTH OF A RESONANCE
Energy Stored in the Cavity
Q o = 2π
Energy dissipated in the walls per cycle
Usually, in the discussion of cavities and
waveguides one of the simplifications made is that the
If the angular frequency of the radiation is ω and we call
walls are perfectly conducting. In the case of isolated
the average dissipated power P diss then assuming an
cavities (ie. no ports) we have an infinite Q o and the
eiωt dependence for the fields we find
resonances of all modes are thus razor sharp δ-functions.
U However in a lossy cavity wall losses result in a
Qo = ω (1) finite Q o and energy transmission through any ports
Pdiss
means that Q L <>Q o which serve to broaden the
resonances. Excitation of a mode is hence possible
where U is the total energy stored in the cavity.
even if the frequency is not tuned perfectly, provided it
In practice though we somehow need to hook up
at least lies within the line width of the resonance.
the cavity to some source, to actually be able to 'inject'
We know that the energy density scales as the
the microwave radiation. As we will show later we also
electric field squared. Thus from (5) we can deduce that
need to measure the energy of radiation emitted through
the electric field in an loaded cavity decays as:
an output port in the cavity. Therefore input and output
probes are an experimental necessity, through which
E(t) = Eo e-ω o t/2Q L e-i(ωo+∆ω)t (6)
invariably energy will be lost. We denote the power
lost to the output probe by Pt and the power lost to the
where ωo is the resonance frequency of the equivalent
input probe by P e. We can define the loaded Quality
perfectly conducting cavity and ∆ω is included to allow
(QL) analogous to the unloaded Qo as:
for a possible (small) frequency shift in the resonance
U frequency due to any losses. If we wish to determine
QL = ω (2) the electric field as a function of frequency we simply
P total
need to Fourier transform (6):
where
P total = P diss + P e + P t (3) inf

is the total average Energy loss from the cavity per

cycle. We now are in a position to obtain an
E(ω) = Eο
0
∫e -ω ot/2Q L i(ω-ω ο-∆ω)t
e dt (7)

expression for the stored energy as a function of time which gives

[U(t)] once the input power to the cavity is shut off.
Since in this case all energy lost from the cavity Eo
E(ω) = (8)
must come from the energy stored in the within, we i (ω − ( ω o + ∆ ω )) - ω o /2Q o
have
SRF 910927-07

from which we see that the energy density per frequency If we identify the 'coupling parameters' as βe = Q o/Q e
interval scales as: and βt = Qo/Qt then (12) can be recast as:

1 1 1
U(ω) α |E(ω)|2 α (9) = (1 + βe + βt) (13)
(ω−ω o -∆ω) 2 + (ω o /2Q o ) 2 QL Qo

which is of the classic Breit-Wigner shape, shown in Clearly, if we can determine the coupling parameters
Figure 1. and the loaded Q then Qo can be calculated.

1.4 EQUIVALENT CIRCUITS FOR CAVITIES

To help us understand how to determine the

coupling parameters we turn to an equivalent circuit for
the cavity as shown in Figure 2.

Figure 1
Figure 2
The curve falls to half its peak value at frequencies
ω o /2Q L to the right and left of the central frequency The two terminals represent the input to the cavity. Z1
(ω ο + ∆ω). We identify this with the line width at is meant to simulate the impedance of the circuitry
half maximum Γ = ω ο /Q L . In the case of the leading to the input.
superconducting rf cavities used in the SRF group, Qo In the absence of any energy losses in the cavity
is typically around 109 and the width of the curve is walls (characterized by R in the circuit) and the input
only a tiny fraction of the resonance frequency. circuitry we would have a simple LC-circuit with
resonance frequency:
1.3 RELATIONSHIP BETWEEN QL AND Qo
√
 1
ωo = (14)
Equation (5) is useful for determining the loaded Q LC
of a cavity hooked up to input and output lines.
However we usually wish to determine Q o so that we The energy stored in the resonant circuit alone is
can describe the characteristics of the cavity independent simply W stored = 1/2LI 2 where I is the peak current.
of the setup used to drive it. Similarly the average power dissipated is
To relate Q L to Q o we make two further Pdiss = 1/2RI2 so that the unloaded Qo follows from (1)
definitions which are similar to (1): as:

ωU 1/2 LI 2 ωL
Qe = (10) Qo = ω = (15)
Pe 1/2 RI 2 R
ωU
Qt = (11) Similarly, in the presence of the input line the power
Pt
dissipated in the load Z 1 is Pe = 1/2 Z 1 I 2 so that the
total power dissipated in the loaded circuit is P total =
combining (2) and (3) it is easy to see that:
1/2 RI 2 + 1/2 Z 1 I 2 . We see from (2) that Q L in (2)
1 1 1 1 takes on the form:
= + + (12)
QL Qo Qe Qt
ωL ωL/R
QL = = (16)
R + Z1 1 + Z 1 /R

-2-
SRF 910927-07

1-β 2
If we now identify β with Z1/R then using (15) we can Pr = ( 1 + β )
Pi (20)
see that
1 1
= (1 + β) (17) and by conservation of energy we have for the
QL Qo transmitted power (Pt):

which is of the same form as (13). We hence find that 4β

P t = Pi - Pr = Pi = Pdiss (21)
β in the case of the circuit is equivalent to βe defined for (1 + β) 2
the cavity.
which for continuous input power must be dissipated in
Consider now the equivalent circuit being driven at R. We call this P diss. Of course all these discussions
a frequency ω via a transmission line of impedance Z1. here apply to the case where power has been supplied to
It will thus present an impedance of the equivalent circuit for a sufficiently long time so that
any transients which may occur due to the build up of

√

fields in the inductor and capacitor have died down. In
Z2 = ( )
R 2- ωL -
1 2
ωC an actual cavity these transients would be due to the
cavity filling with energy when power is switched on.
This will be discussed later.

Imagine now, the input power were suddenly

switched off. Immediately after this happens a current
Iinst flows, given by:
to the transmission line. An electric wave travelling
I2 inst = P diss/R (22)
down the line is partially reflected at the interface in
order to satisfy the following boundary conditions:
using β = Z 1 /R it follows from (21) that the power
V i + V r = Vc emitted instantaneously (Pe) back into the transmission
I i - Ir = Ic line is:

where Vi and Vr (Ii and Ir) are the voltages (currents) of 4β2
P e = I2inst Z1 = Pi (23)
the incident and reflected waves respectively. Similarly (1 + β) 2
V c (I c ) represents the voltage (current) of the wave
transmitted across the boundary to our equivalent 2.1 DETERMINING Qo EXPERIMENTALLY
circuit. By definition the impedance is:
It is now up to us to put the results obtained for
Z 1 = Vr/Ir = V i/Ii, Z2 = V c/Ic the equivalent circuit to use in determining the Q o of
the cavity.
allowing us to solve the two equations to give: Comparing (17) with (13) we see that provided we
use β e = Q o /Q e for the coupling between the input
Vr 1 - Z 1 /Z 2 probe and the cavity and if no output probe is present,
= (18)
Vi 1 + Z 1/Z2 then the reflected power is simply given by (20), which
we repeat here:
If the equivalent circuit is driven at ωo (on resonance)
1 - βe 2
Pr =   P
then Z2 = R (the reactance is zero at ωo). In this case,
β  i (24)
thus:  1 + e

Vr 1-β
= (19) and in the same manner the instantaneously emitted
Vi 1 + β power is given by (23):

Since power is proportional to |V|2 we finally obtain

the reflected power (P r) as a function of the incident 4βe2
power (Pi): Pe = Pi (25)
(1 + β e ) 2

-3-
SRF 910927-07

computer codes. Once this factor has been obtained we

Solving these equations we obtain two expressions for are able to determine Epeak as follows.
βe: A second probe is added to the other end of the
cavity. Provided the coupling (βt) is <<1, ie. Qt >>Qo,

√
Pr only very little power will be dissipated via this output.
1 +/- In this limit the previous discussion and the expressions
Pi
βe = (26) for Pe and Pr are still valid. Since we know that Pt α U

√
Pr by (11) we obtain the relation:
1 -/+
Pi
E peak α √
 Pt (32)

and
1 A measurement of β and τ allows us to calculate Q o
βe = (27) using (29) and (30). This is done at low field levels for

√
Pi reasons which will become clear later. Furthermore we
2 -1
Pe know P diss from (21) which allows us to calculate U
using (1). It is then trivial to calculate Epeak from (31)
These allow us determine the coupling between a assuming we have previously determined κ e . The
transmission line and our cavity by simply measuring proportionality constant in (32) then follows from a
the incident power, the reflected power and, if we wish, measurement of Pt. At the same time this process also
the instantaneously emitted power P e as P i is turned yields the factor relating Pt to U.
off. One could argue that all this is not necessary since
Furthermore, (10) indicates that if the input power we can determine Qo and Epeak directly from (30) and
is turned off, the power emitted from the cavity as a (31). This is true. However, this relies critically on
function of time is proportional to the present energy U our ability to measure τ. At high field strengths
stored within. We know from (5) that U decays as: processes such as field emission and thermal breakdown
degrade the Qo of the cavity. If we were to measure τ
U(t) = Uo e-ωot/QL (28) from the decay of P e or P t as P i is pulsed we would
observe that τ varies. At the start of the decay τ is
small (small Qo) until the fields within the cavity drop
Note that it is QL which appears in this expression, not
below the threshold value for these processes. Qo then
Q o . Clearly the emitted power should therefore also
changes and correspondingly τ. To avoid the difficulties
decay with the same time constant
arising from this we determine the factors relating Pt to
QL U and E 2 peak at low field strengths, where field
τ= (29) emission etc. isn't possible and τ is constant
ωo throughout a decay. At the critical high fields our value
for P t then directly yields U and E peak . From the
which we can measure directly. Once we have former we then calculate Qo via (1) and (21).
determined βe from either (26) or (27) we finally obtain
the unloaded Qo:
2.2 EXPERIMENTAL TECHNIQUES
Qo = QL (1 + βe) = ωoτ (1 + βe) (30)
2.2.1 General Setup
To obtain the peak electric field (E peak ) in the
cavity we recall that energy density is related to the To couple the power into the cavity a setup as
electric field by: shown in Figure 3 is used. The power from the source
is carried to the cavity via a coaxial cable as a TEM
Epeak= κe 
√U (31) mode. The central conductor protrudes into the beam
tube of the cavity which forms the outer conductor.
where κ e is a constant dependent on the geometry of Since TEM modes cannot be supported in a
the cavity. For a pill box cavity this factor can be hollow waveguide, we obtain a mixing of TE and TM
calculated analytically, however for the geometries we modes once the end of the central conductor is reached.
are using it needs to be determined numerically using The diameter of the beam tube is arranged in such a
manner, that the lowest cutoff frequency lies above the
frequency we operate at.

-4-
SRF 910927-07

the cavity. A modulator controls the amplitude of the

signal via a Pin Attenuator (not shown here), allowing
us to pulse the signal. The resulting waveform is
amplified by a 100W amplifier and fed into the cavity
from the bottom. This is simply to avoid dirt falling
into the cavity from potential sources of dust, such as
the bellows and the coupler.

Figure 3: behavior of the fields in various parts of the

cavity

Hence the fields decay exponentially until they reach the

Figure 4: Outline of the experimental setup
actual cavity.
Using directional couplers (also not shown) part of
The Q o of a niobium cavity will generally be
the signal is diverted to an oscilloscope or power meter,
around 10 9 to 1010 at 1.5K. However, as field
before it reaches the cavity. Another directional coupler
emission, multipacting and other processes begin to set
also diverts part of the signal reflected at the cavity
in at higher electric fields, the Q o begins to degrade, input into the same scope, allowing us to measure P i
rapidly falling to 108 and lower. From (21) we see that
and P r . Similarly part of the signal from the output
the power transmitted into the cavity strongly depends
probe is routed into a power meter yielding Pt. The rest
on β e. If the match between Q e and Q o is bad (ie β e
either <<1 or >>1) then a lot of power is required to of the signal is amplified and fed into a mixer. A signal
achieve sizeable fields. It is hence desirable to maintain coming from the oscillator via a phase shifter forms the
unity coupling. Since Q o changes we have to be able second input for the mixer. The output dc signal is
to vary Q e as well. To facilitate this the cavity is dependent on the phase between the two inputs and is
mounted on bellows. A motor drive allows the operator transmitted to a frequency loop amplifier. It provides
to move the cavity up and down over the input probe the feedback for the oscillator which stabilizes the
which remains fixed. Since the fields between the tip of oscillator's frequency and ensures that the microwave
the input coupler and the iris of the cavity decay signal stays 'in step' with the oscillations in the cavity.
exponentially only small movements are required to As an analogy consider a parent pushing their kid on a
achieve changes in Qe of several orders in magnitude. swing. To be able to maintain the oscillations, pushes
A transmission probe is also added to the other end are needed at the resonant frequency of the swing. But
of the cavity. It is arranged in such that Q t ~ 1013, ie in addition to that they need to be in phase with the kid
βt <<1. As explained previously, this has a negligible on the swing.
effect on the cavity which hardly 'sees' the probe. The phase shifter in the line between the oscillator
and the mixer is required to compensate for a difference
The general outline of the experimental apparatus in distance travelled by the two input signals at the
used to test cavities is shown in Figure 4. For a more mixer and to compensate for any offsets the mixer
detailed description refer to Appendix A1 . might have. Note that at 1.5 GHz the wavelength in
An oscillator produces the initial 1.5 GHz signal, vacuo is 20 cm.
which is the resonant frequency of the TM010 mode in

1 For a detailed explanation of the individual microwave

components used in our setup refer to Kevin Green: A n
Introduction to Coaxial Microwave System, Cornell
University, June 1989

-5-
SRF 910927-07

the mixer in this case. When the phase is correctly set

the output should be at a minimum.
Once the phase is set the modulator is switched to
pulsed with a rep-rate of about 50 ms. Often it is
convenient to look at the reflected signal on the scope
which should look like one of the traces in Figure 6. If
the system is still not locked in on the resonance, we
expect all of the signal to be reflected, yielding square
wave pulses as in Figure 5a). In that case the frequency
and/or phase needs further adjustment. There is no set
recipe for the adjustment procedure and a bit of luck is
involved.
Once we do achieve a lock on, the trace will
resemble one of the other three in Figure 6. In general
the coupling won't be perfect (ie. β <> 1). Look at fig
6d) first. In this case the cavity is undercoupled, that is
β <1 and Q o < Q e. Initially, as the incident power is
turned on, no energy is stored in the cavity and we have
Figure 5
not yet established an equilibrium condition. This
results in a mismatch between the cavity and the
The cavity itself is housed in a dewar filled with
transmission line and hence all power is reflected
liquid helium, surrounded by a liquid nitrogen bath (see
leading to the large spike. As the cavity fills, the
Figure 5). The entire inner dewar is vacuum sealed so
matching improves and the reflected power drops2 .
that the interior can be evacuated using the large pumps
However, since we are undercoupled we still get
in the pump room. This lowers the vapor pressure
reflected power, even when the cavity is filled
allowing us to vary the temperature of the bath between
completely (ie the power dissipated equals the power
4.2K and 1.5K. A dip stick, which goes
entering the cavity). In such a case, we use the motor
superconducting at liquid helium temperatures is
drive to lower the cavity in the dewar thereby increasing
connected to a meter measuring its resistance. From
the distance by which the input probe
this one obtains a value for the amount of helium
One usually finds that at 4.2K the system is
remaining in the dewar.
undercoupled because the cavity Q is so low. Note also
2.2.2 MAKING THE MEASUREMENTS when the input power is cut, the reflected signal does
not drop to zero instantaneously. What we in effect are
First measurements are usually made at 4.2K after then viewing is the power emitted from the cavity
opening (Pe) as the energy in the cavity empties. The
the He transfer. At this temperature the surface
resistance of the cavity is still relatively high giving decay is exponential with the decay constant given by
rise to a large P diss . As a result Q o is low and we (29). Thus we can measure τ directly from the
expect a wide resonance, see (9). This makes it easier oscilloscope. At lower temperatures and small input
to tune the oscillator. Initially the oscillator is set to powers especially, one may find that the system is
about 1.5 GHz and the modulator is on cw-mode. The overcoupled, giving rise to a trace as in Figure 6c). In
input coupler needs to be fairly far in, so that Pe is high this case perfect coupling is achieved before the cavity
and matches P diss. This ensures near unity coupling. is completely filled and the reflected signal dips to zero
at that point. However as energy continues to 'pour' in,
Viewing the reflected signal on the oscilloscope just
we again have a mismatch and some energy is reflected,
gives a straight line which should be the same resulting in a rise in P r. In such a situation we would
amplitude as the input signal, assuming we aren't
raise the cavity to extract the input probe from the beam
locked in on the resonance by chance. The frequency is
tube in order to reduce the coupling.
varied slowly until the resonance is found. Barring
protrudes into the beam tube.Ideally we wish to
exceptional circumstances it should lie within a few
adjust the coupling until β = 1 and we obtain a trace
tens of MHz of 1.5GHz. The reflected signal will
just like Figure 6b). In practice that may not always be
oscillate violently once the resonance is found and the
system tries to lock on. Usually the phase shifter now 2 For a discussion of the reflected power during this
needs to be adjusted. It helps to look at the signal from
transient period (ie. while equilibrium has not yet been
achieved) refer to Appendix B.

-6-
SRF 910927-07

possible, especially when the cavity Q has degraded a height of the square pulses when the system is taken
lot due to field emission, multipacting etc. out of lock. Again one should note that these powers
are measured at the directional coupler. However if we
determine both Pi and Pe from the reflected power trace
both signals will suffer the same attenuation on their
way from the cavity to the coupler, so that the ratio Pi
to the true P e will be the same as that of P f to the
measured value of Pe . Furthermore since only the ratio
of the two is required we don't need to calibrate the
scope to obtain absolute units.
Finally we also measure the transmitted power, Pt
and P i, this time in proper units. Note that this needs
to be done with the modulator set to cw. The powers
are measured on calibrated power meters.
Once these preliminary measurements have been
completed we can calibrate the system as explained in
section 2.1. We are now in a position to measure Q
versus Epeak as the input power is increased. Once the
calibration has been done measurements of τ are no
longer required.

For moderately high fields the procedure remains

unchanged w.r.t. to the one described above. However
Figure 6 as E peak is increased to perhaps 15 - 20 MV/m it
Once things are optimized we are in a position to becomes increasingly likely that breakdown and other
measure Q o . As mentioned earlier, τ is measured processes occur.
Often one first observes multipacting. This is a
directly from the decay of the emitted power after the
result of stray electrons in the cavity, which can for
input power is switched off.
For our future measurements we require P i , this example be a result of cosmic rays or emission from the
surface. At certain field strengths these electrons will
being the power incident on the cavity. However note
follow a closed orbit trajectory, and may impact on the
that the forward power is measured at a directional
surface. Near the surface of the cavity the magnetic
coupler located outside of the dewar. We must remember
fields are strong and electrons in this region will follow
that the cables in the dewar leading to the cavity do have
cyclotron orbits of frequency
some attenuation α. We need to therefore first calibrate
this. Denoting the forward power measured at the eB
directional coupler by P f we have P i = α P f . ωc = (32)
m
Furthermore if the cavity is out of lock then all power
will be reflected. Thus the reflected power we measure at where e is the electron’s charge, B is the applied field
the directional coupler (Pr) will be given by Pr = α Pi.. strength and m is the electron’s mass.
Hence we can solve these two equations to give: Thus if nω c (n = integer) is equal to the driving

√

Pr frequency ω the electrons will return to the same point
α= (Cavity out of lock) in phase with the field and impact on the surface.
Pf
If the secondary emission coefficient (δ) of that surface
is greater than 1, further electrons are emitted usually
so that in future we can always obtain Pi by measuring
with a few electron volts of energy. They too follow
Pf once α has been determined. the closed orbit, gaining energy from the perpendicular
Measurement of P i and the instantaneously electric field and impact at the same point. We therefore
emitted power P e yields β directly from (27). P e is arrive at an avalanche situation. If the input power is
obtained from the height of the peak produced as the increased it no longer serves to increase the stored
incident power is switched off (see Figure 6). On the energy in the cavity, but simply enhances the
other hand Pf is given by the height of the peak as the multipacting process. It is as if we have reached a
input power is switched on or, alternatively, by the barrier beyond which Epeak no longer can be increased.

-7-
SRF 910927-07

Correspondingly, if we look at the transmitted power Pt fully understood and the reader is referred to literature for
which is a measure of the energy stored in the cavity, the latest theory on these. Here we simply present a
we see that once the fields within are high enough for short overview.
multipacting, clipping occurs. An increased P i no Field emission3 occurs when the probability of
longer increases U and thus Pt remains constant. Hence tunnelling through the potential barrier presented to the
we never achieve unity coupling, and Pr remains finite. electrons in the metal (the work function) becomes large
enough to produce a significant current. This in turn
draws energy from the incident power thereby spoiling
the cavity Q. Typical of barrier penetration is the
electron transmission probability's exponential
dependence on field strength, which serves to lower the
barrier height. Fowler and Nordheim treated this
problem quantum mechanically and found that4:

AE2 φ 3/2
j(E) = exp -B (33)
φ  E 

where j is the current, φ is the workfunction of the

metal, A and B are constants given below and E is the
electric field strength. We have
Figure 7 A eV
A = 1.54 x10-6
It is possible to break through this multipacting Vcm4
by increasing the magnetic field levels so that ω c = ω
and
cannot be satisfied. Furthermore, δ is a function of
energy, being greater than one in only a narrow energy
V eV-3/2
interval (E 1 to E 2 ) and hence it is only necessary to B = 6.83 x107
cm
break through a finite number barriers in this energy
range. One also finds that it is possible to alter the provided j(E) is given in A/cm2 , E in V/cm and φ in
surface characteristics due to the severe electron eV. Field strengths in typical cavities are of the order
bombardment and lower the secondary emission of 10 MV/m. Using such a field in (33) we find that
coefficient to values below one by allowing the current is utterly negligible. However it is often
multipacting to proceed for some time (perhaps 20 observed that field emission does limit the cavity's
minutes or so). This method is very effective and is performance and one finds empirically that field
termed 'processing'. Multipacting barriers which can be emission still displays Fowler-Nordheim behavior
processed away are referred to as 'soft barriers'. provided we scale the electric field in (33) by a factor of
In fact the cavity design is in part such that βFN. It is as if locally the field is enhanced, sometimes
multipacting is reduced. For the best shapes used,
by factors up several hundred. The reasons for this are
stable electron trajectories are not possible and any
by no means understood properly, yet.
orbiting electrons quickly drift to the equator of the This observation implies that (33) needs to be
cavity where the perpendicular electric field is zero. modified to give:
Thus multipacting electrons no longer can gain energy
and the process is halted. It is therefore fairly simple to A(βFNE)2 φ 3/2 
process through any multipacting barriers which may j(E) = exp B (34)
still exist. Cavity cleanliness is also important in φ  βFNE
avoiding multipacting. One finds that δ depends on
surface composition, clean surfaces in general The field emitted electrons are accelerated by the
displaying a lower δ and a narrower window (E1 - E2) electric field and smash into the cavity walls at other
for which δ >1. 3 Field emission, for example, is discussed in H. Padamsee:
At higher field strengths it is common to run into Superconducting RF, AIP Conference Proceedings: The
a combination of field emission and thermal breakdown. Physics of Particle Accelerators.
At present, the mechanisms leading to these are still not

-8-
SRF 910927-07

sites. Due to Bremsstrahlung we thus get very of the surface. This causes heating of the surrounding
energetic x-rays, which are a signature for field areas, driving these normal as well once the temperature
emission. These x-rays, sometimes, are even capable of of the defect is above 9.2K (T c for Nb). Because the
penetrating the ~2 foot sheetrock blocks used for normal state is far more resistive (by many orders of
shielding. The number of emission sites increases magnitude) a substantial amount of power dissipation
exponentially with the field strength and we observe a will occur in this region. Q o drops dramatically and
steep decrease in the Q o versus E peak curve obtained hence τ too, typically from about 0.5s (Q o ~ 1010) to
once field emission occurs (see Figure 9). only a ~ milliseconds. A trace of P t looks somewhat
Field emission currents are spread out over areas as like Figure 8.
small as 10-9 cm2 and typically current densities are
around 100 A/cm2 although they can reach densities as As the cavity starts filling the fields inside cause
high as 109 A/cm2. Due to the finite resistance of the thermal breakdown. The Q then suddenly drops yielding
niobium and the fast time scales involved we therefore a tiny τ and resulting in a mismatch between the input
can have regions of severe heating. These drive the coupler and the cavity. No more energy is input in the
niobium into the normal state and ultimately are cavity. The remaining fields within decay over a period
capable of melting the material. Sometimes emitters ~ τ, yielding the spikes in fig. 8. During the decay the
will process away leading to what is believed are small normal regions are cooled by the helium bath and go
'explosions'. Subsequent examination of emission superconducting again. The original Q thus is restored,
sites under an electron microscope frequently reveals and we once again observe transmitted power and the
characteristic 'star bursts', so called because of the star process described above repeats once more. Increasing
like area (perhaps 50 - 100µm across) of reduced the input power thus only serves to increase the
secondary emission coefficient around the emission site. frequency of these breakdowns.
At the center usually lies a lump of molten niobium. To reduce thermal breakdown and field emission it
Different methods have been developed to reduce is imperative to go through extensive clean room
field emission and to push the fields at which it occurs procedures in order to minimize any foreign particles.
to higher limits. They include a) extended rinsing, b) Very effective in improving field strengths at which
heat treatment c) He processing and d) RF processing. thermal breakdown first occurs, is an increased thermal
For a discussion of these the reader is referred to conductivity of niobium. Any heat produced by the
reference 3. The last one on the list (d) is currently tiny impurities is quickly carried away by the
being studied extensively4 and involves 'blasting' the surrounding regions, thereby reducing the likelihood
cavity at high power levels (often pulsed with Powers that surrounding niobium is also driven normal. The
~10's of kW and pulse times ~100 µs). In certain thermal conductivity is strongly related to the purity of
instances, following this procedure one finds that the the niobium. Often quoted is the RRR or the 'residual
emitter has processed away. resistance ratio' which is the ratio of the resistance at
room temperature to that at 4.2K with the niobium in
the normal state. (At 4.2K this is done by driving the
material normal with a sufficiently high magnetic field,
H > H c = critical field strength). Since the RRR is
directly related to the impurity concentration, one finds
that high RRR materials are not as susceptible to
Figure 8: Trace of Pt with thermal breakdown occurring breakdown as low RRR ones. Typically the cavities we
use have a RRR of the order of 100 - 500 at present.
Field emission as well as imperfections and However we still need to contend with field emission
impurities as small as 0.1 mm across can also lead to a which proves to be the limiting process.
breakdown behavior called thermal breakdown. Tiny
regions on the surface of the cavity which are not 3. TYPICAL RESULTS
superconducting, or only become superconducting at
temperatures lower than the bulk superconductor will Figure 9 presents a Q ο versus E peak curve
dissipate far more energy from the rf-field than the rest somewhat typical of the results usually obtained.
Initially the Q o is high (1010) and remains constant as
4 Joel Graber 'An Apparatus for High Power Processing of
the power is increased. In this region multipacting is
Field Emitters in Superconducting RF Cavities' (MS Thesis),
Cornell University, May 1990 frequently observed but as mentioned previously it
usually is not difficult to process through this.

-9-
SRF 910927-07

At higher field levels (~ 15 MV/m) the first X- expected from the extrapolation of the curve. This is
rays being produced in the cavity were detected, an probably an example of an emitter being processed
indication that field emission was occurring. We see away, yielding an improved quality. However, as the
that from that point on the Qo rapidly falls off reaching field is increased to even greater values, new emitters
the low 109 at field levels around 35 MV/m. Note become active and the Qo drops further.
however the point at about 31 MV/m. During the For other cavities other features in such a plot
experiment we found that the X-rays suddenly were may exist. The constant Qo at low field levels and the
reduced as that point was being measured. We also see rapid fall off at higher ones, though, is typical for all
that its Q o is significantly higher than would be results.

Figure 9

-10-
SRF 910927-07

4. OTHER CAVITIES Due to different sizes and geometries the apparatus

used for the multicell and mushroom cavities is
In addition to the single cell cavities described somewhat different to the one described here. In
here, the SRF group also tests multicell cavities (in addition to this the multicell cavities require some
effect several single cell cavities welded together) and additional 'attention', for example the individual cells
mushroom cavities. The latter is actually only half a need to be tuned (by varying their size) so that all
cavity with a flat baseplate. It is constructed in such a resonant frequencies are the same. This process is called
way that the plate can be removed after a test allowing 'bead pulling'. Despite these differences though, the
us to examine the surface under an scanning electron methods used to obtain the final Qo versus Epeak curve
microscope (SEM). A dimple about 3/4 cm in diameter are very similar to the ones described here and the
ensures that the highest field strengths within the cavity theoretical discussion still applies.
are in this region thereby reducing the area that needs to
be searched in the SEM.

-11-
SRF 910927-07

APPENDIX A)
Detailed schematic diagram of the setup used for single cell, 1.5 GHz Niobium cavities.

-12-
SRF 910927-07

APPENDIX B)5 (B5) one finds that the following analysis yields
unphysical results.
Presented here is a short discussion of the behavior Now back to our original problem. By (1) we
of reflected power as a function of time during the know that the power dissipated in the cavity [P d (t)]
transient periods just after the input power is switched scales linearly with U(t), ie:
on. In this case the the cavity is still filling with
energy, ie. dU/dt is not zero, as would be the case in the ωU(t)
Pd(t) = = Pdiss (1 - e-t/2τ)2 (B6)
steady state. By conservation of energy we can write Qo
down the following relation:
which we can recast using (21) as:
dU
Pr = Pi - Pdiss - (B1)
dt 4β
Pd(t) = (1 - e-t/2τ)2 Pi (B7)
(1 + β) 2
where Pdiss is given by (21) for the steady state. Now,
from (5) we recall that U decays exponentially as e-t/τ Furthermore we recall that from (1)
once the input power is turned off. This implies that
the electric field decays as e-t/2τ. Similarly, when the PdissQo Pi Qo
4β
input power is initially switched on, we expect the Uo = = (B8)
ω (1 + β) 2 ω
electric field to rise as:

E(t) = Eo (1 - e-t/2τ) (B2)

where Eo is the steady state E-field as t tends towards

infinity. We can thus conclude that the energy density
rises as:

U(t) = Uo (1 - e-t/2τ)2 (B3)

One could raise the objection that we should expect U(t)

to increase according to:

U(t) = Uo (1 - e-t/τ) (B4)

which would lead us to believe that E(t) behaves as:

E(t) = Eo √
 
1 - e -t/τ (B5)

which differs from (B2). To resolve this problem, we Figure B1: Plot of the reflected power for a) perfect
appeal to the equivalent circuit again. Recall that when coupling, b) overcoupling and c) undercoupling.
calculating the resonant frequency, as well as the time
dependence of the fields in for example the capacitor, we We can therefore rewrite (B1) as:
use the fields as the fundamental quantity, not the
energy, in which case we obtain a (1 - e-t/τ) dependence
Pr(t) =Pi  1 - (1-e-t/2τ)2 
4β
for rising fields. In an analogous manner, if we
consider the electric fields in the cavity it is actually  (1 + β) 2 
quite straight forward to derive expression (B2) for E(t). x
We will thus use (B2) and (B3) in the following 4β Q ο d
analysis. If indeed one instead proceeds to use (B4) and -P
i ω ( 1 + β) 2 dt
(
(1 - e -t/2τ ) 2 ) (B9)

5 From private communications with Joel Graber, Cornell

University.

-13-
SRF 910927-07 - 14 -

which can easily be simplified, noting thatQ 0/ (ω(1 +

β)) = τ, giving:

Pr(t) =Pi x

x 1 - (1-e-t/2τ)e-t/2τ
4β 4β
(1-e -t/2τ ) 2 -
 (1 + β) 2 (1 + β) 

This equation describes the traces in Figure 6 for

various couplings very well. Some typical plots are
given in Figure B1. Note that the third term in the
above equation goes to zero as t goes to infinity. It
represents the transients. On the other hand, the second
term asymptotically approaches 4β/(1 + β)2 , so that
the steady state value for Pr(t) is given by:

2
Pr = (11 +- ββ ) P i

in agreement with (20).

-14-