Shimming a superconducting nuclearmagneticresonance imaging magnet with steel

D. I. Hoult and D. Lee

Citation: Review of Scientific Instruments 56, 131 (1985); doi: 10.1063/1.1138480

View online: [Link]
View Table of Contents: [Link]
Published by the AIP Publishing

Articles you may be interested in

Nuclear-magnetic-resonance diffusion simulations in porous media
J. Appl. Phys. 97, 083510 (2005); 10.1063/1.1871352

Variabletemperature solidstate nuclearmagneticresonance probe for superconducting magnets operating in the

range 3–350 K
Rev. Sci. Instrum. 55, 68 (1984); 10.1063/1.1137592

Mosfet highpower pulsed rf emitter—a design for nuclearmagneticresonance imaging

Rev. Sci. Instrum. 53, 1561 (1982); 10.1063/1.1136836

Gradient control system for nuclearmagneticresonance imaging

Rev. Sci. Instrum. 53, 1338 (1982); 10.1063/1.1137181

Small solenoid with a superconducting shield for nuclearmagneticresonance near 1 mK

Rev. Sci. Instrum. 53, 485 (1982); 10.1063/1.1136995

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: [Link] Downloaded to IP:
[Link] On: Fri, 05 Dec 2014 [Link]
 Shimming a superconducting nuc;ear-magnetic-resonance imaging magnet
with steel
D.1. Hoult
Biomedical Engineering and Instrumentation Branch. Division of Research Services, Building J3. Room
3WJ3, National Institutes of Health, Bethesda. Maryland 20205

D. lee
Picker International. Nuclear Magnetic Resonance Division. Highland Heights. Ohio 44143
(Received 25 June 1984; accepted for publication 15 August 1984)

Using a recently published paper as a basis, the magnetic field produced by steel bars inserted in a
superconducting NMR imaging magnet is analyzed in a spherically harmonic basis set. A
description is then given of how such bars were used to improve the homogeneity of the field
within the magnet's imaging volume from 1.2 parts per thousand to about lO ppm. The poor
homogeneity was caused by the magnet's being placed in a steel-laden environment, a situation
normally shunned by investigators, and it is the author's contention that the results obtained
abrogate the main objection to NMR equipment's being installed in an ordinary hospital building.
To facilitate the latter, the equations developed may also be used to estimate, prior to installation,
the effects of the proposed environment on field homogeneity.
INTRODUCTlON thereby substantially reducing electromagnetic interference.
A major dilemma facing the would-be purchaser of NMR Reserving the nearest section of the parking garage for liq-
imaging equipment is the selection of a site for the instru- uid-nitrogen storage would alleviate the problems caused by
ment. Amidst a host of minor problems, two major difficul- moving cars, and the garage had too Iowa ceiling for trucks
ties stand preeminent-the influence of the magnet on its to enter. Areas around the suite were within the Radiology
environment and the converse. Thus, there already exists a domain and, therefore, would pose no major problems, but
rich folklore concerning the influence of moving vehicles, the looming imponderable was what homogeneity the mag-
the necessity for new, iron-free buildings down to the last netic field would have. The most pessimistic estimate was
bolt and nail, and the horrible effects of stray fields upon 500 ppm, while the specification called for 10 ppm over a 30-
intravenous pumps, electrocardiogram monitors, etc. How- . cm-diam sphere. Ever optimistic, we proceeded with the
ever, while heeding the old adage about smoke and fire, it magnet installation, and a homogeneity of 1.2 parts in one
was felt that the clinical imaging equipment to be installed at thousand was obtained-well outside the range of the cor-
the National Institutes of Health (NIH) had to be in the Ra- recting coils.
diology department, close to other imaging instrumentation,
if the maximum efficiency and advantage were to be gained !. fIGHTING STEEL WITH STEEL
from it. That department lies in a new seven-story building In a recent review article, I Romeo and Roult present a
constructed of reinforced concrete, and it was immediately full mathematical basis for generating spherically harmonic
recognized that the room available posed severe problems. magnetic fields with the aid of current or dipoles. In particu-
The five tonne superconducting magnet would lie almost on lar, they discuss the uses of pieces of steel, placed on the
a major structural beam reinforced internally with four 4- surface of a sphere, to homogenize or "shim" magnets. Their
cm-side steel bars, and two important power distribution analysis is easily extended to steel pUaced on a cylindrical
points holding hundreds of kilograms of steel would be only surface (i.e., inside a superconducting magnet), the rationale
5 m away. To make matters worse, the beam would have to for its use lying in the fact that the field within the magnet
be reinforced with 13 X 13 X I-cm steel plates above and be- obeys Laplace's equation, and may therefore be resolved into
low spaced every 45 cm, a parking area would lie on the floor the orthogonal basis set which the spherically harmonic so-
beneath (but to one side), and an intensive care recovery lutions of Laplace's equation yield. Unwanted harmonics
ward would be directly above. The most charitable comment may then be removed by the appropriate placement of pieces
received about the site was "interesting"; most were unprint- of steel. Within (in contrast to outside) the magnet, the field
able. is overwhelmingly parallel to the axis (by definition, the z
Initial investigations using the electromagnet imaging direction). As we are trying to produce corrections of the
system at the NIH showed that shielding with l-cm steel order of parts per thousand, it follows that any analysis may
plates affixed to the 4-m-high ceiling of the imaging suite safely ignore fields in directions other than z. Consider then
would reduce the field above by a factor of between 3 and 4 to an element d V of steel within the bore of the magnet at polar
a point (500,uT) where the various monitors used in an inten- coordinates Q if,a,t/J) as shown in Fig. 1. At pointP (r,O,<,6 ) the
sive care unit would function satisfactorily. (That investiga- scalar potential is given br
tion has since been proven correct.) Further, such shielding,
if tied electrically via the wall studs to a copper ground plane -xHz
d<P= dVkgrad Q (l!v), (1)
on the floor of the suite, would create a crude Faraday shield, 417'

This article131 Rev. Sci.

is copyrighted as Instrum.
indicated56
in(1),
the January 1985 of AIP content is subject to the terms at: [Link] Downloaded
article. Reuse 131 to IP:
[Link] On: Fri, 05 Dec 2014 [Link]
 z aYnmlaz = - (n - m + llYn + I,m' (8)
Thus,

(9)

Writing dV = Adz, where A is the cross-sectional area of the

steel bar, we then have from Eq. (5) that
<I> = xHzA fiE (n - m - I)!
41/' n~Om~O m (n+ m - I)!

X[Pn_l.m(cosa)lr]~~:~Tn_I,m' (10)
where the ends of the steel rod have polar coordinates (h,r,t/J)
~~-....,...- - - - - - -. .y and 1$,/3,1/;) for the limits of integration. Remembering that
H = - grad <1>, and being only concerned with the field in
the z direction, we have, after changing the index n by 2,
H = - xHzA fiE (n - m + 1)1
41/' n~Om=O m (n + m)!
X [P" + I.m (cos a)lr + 2) :~~:~Ir"Pnm (cos B)
Xcos[m(~ - I/;)J. (11)
This, then, is the master equation for the effects of a steel bar
x lying inside the magnet, parallel to the z axis, and sufficiently
FIG. I. Coordinate system employed in the text. We are concerned with the short that its magnetization may be considered at all points
field at P(r,8,tb, produced by an element dVofsteel bar at Q(f,a,J/!,. to be in the z direction.

where x is the susceptibility of the steel element, Hz is the

main magnetic field, k is the unit vector in the z direction, II. SHIMMING
and the Green's function v- I is the distance between P and Of prime importance in the locating of any imaging
Q. Now3 magnet is the arrangement of the magnet with respect to the
l/v = fiE (n - m)! Pnm(cos a)
steel in the floor. Under no circumstances should beams or
rods lie at an angle; they should be parallel or perpendicular
" = Om = 0 m (n + m)l In + I to the magnet axis, and preferably the former. Fortunately,
X r"Pnm (cos B)cos[m(~ - t/J))' (2) such an arrangement is normally dictated by the shape of the
where the Neumann factor Em = 1 if m = 0; otherwise room-the inner walls tend to lie parallel to the building
Em = 2. P"m are the associated Legendre functions. Now framework. Under such conditions, considerable symmetry
the only term of consequence in Eq. (1) is the z derivative of is displayed by the field within the magnet, making "shim-
Green's function, and Romeo and Hoult show that if the ming" much easier. Figure 2 shows the approximate loca-
spherical harmonic T"m is given by
(3)
then
aTnmlaz = (n + m)T" _ I,m' (4) Turret

Hence, from Eqs. (4) and (1), rearranging terms slightly, r

Steel Shield on CriHng

d<l> = - xHz
411'
f
"=
±
Om = 0
Em
(n
In -
+m -
m)l
1).
I
Marne. Sort'
Radiu8 • = 46cm

Xr"-IP"_I,m(COS B)Y"m' (5)

where Y"m is the inverse spherical harmonic
.,.,..-_+--2 Steel Bar..
Y"m =1 -i" + lIPnm (cos a)cos[m(t/J - ~ )]. (6) . . - - - Major Power
Dilltribution
Romeo and Hoult dealt only with points of steel, but such a
large correction was needed for the magnet that steel bars
had to be used. Hence, for bars lying parallel to thez axis, Eq.
(5) must be integrated, in so far as Y"m is concerned, alongz.
At first sight, this appears difficult. However, we note that4 Reinforcement

Pnm (cos a) = P _ in + I),m (COS a). (7) FW.2. Environment of the NIH magnet. The dominant influence on the
field was the supporting beam beneath the magnet which necessitated steel
Substitution of this identity in Eq. (6) with n' = - (n + 1), strips being placed at the azimuthal positions shown to annul first degree
converts Y nm into Tn' m' and from Eq. (4) we then have that harmonic components.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: [Link] Downloaded to IP:
132 Rev. [Link]., Vol. 56, No.1, January 1985 Imaging magnet 132
[Link] On: Fri, 05 Dec 2014 [Link]
 duced by the steel would introduce only S.6-ppm inhomoge-
Proton Offset neity. This figure was later confirmed by field plotting; how-
,"'requeney
2 in kHz ever, in combination with other high orders present, it was
sufficiently large to enable us to just meet the specification,
• r=5cm,no shims
and should therefore not have been introduced. The Appen-
o r = 15cm,wlth shim8
H=900 dix details what we should have used.
Summarizing, we now know ~he length and azimuthal
• positions of the steel needed to produce a relatively pure TIl
Ot--p__ ~~~.--+~~A~'~im~ut~h~in~n~.g~r~~.~~\~~.~ ••
harmonic; the one remaining question is that of how much to
use. We have found that the susceptibility of steel at a field of
90 /80 270 ~:o 0.25 T varies somewhat with composition. We used steel
• from a large number of sources as we searched for desirable

/
cross-sectional areas, so we assumed an average susceptibil-
-I
! 30 p.p,m,
ity of eight, based on a saturated induction of 2.1 T for cold

-2
.~ . rolled stee1. 5 From Eq. (11), allowing for the eight pieces at
the various azimuthal angles needed for a first degree correc-
'-~
0
tion/ and using steel bars subtending 81.8 (r = 49.1°,
f3 = 130.9°),
F1G.3. First azimuthal NMR field plot at a radius rof 5 em at (} = 9(1', i.e., in
the center xy plane. The ordinate is in proton Hz. Also shown is a similar
(14)
plot, but at a radius of 15 em, after the addition of steel bars.
Substituting H II ==2180 Hz and r = 5 cm from Fig. 3, and
tion of steel in the NIH imaging suite, and it is clear that the x = 8, Hz = 10.6 MHz for protons, a = 46.2 cm, e = 90°,
overriding inhomogeneity in the field is caused by the steel c/J - t,b = 0, we have that A = 1.1 cm 2 • Using 3/8-in.-square-
beam and is predominantly described by the spherical har- bar (0.9 cm 2 l, Fig. 3 shows the results obtained at a radius of
monic T II-the field is a maximum in the quadrant of the 15 cm. The correction is excellent, showing residual second
beam. The azimuthal field plot of Fig. 3 bears out this state- and third degree components which, of course, were also
ment. A Fourier analysis (and inspection) shows an over- annulled using similar techniques. The particular steel used
whelming first degree (m = 1) harmonic [cf., Eq. (3)]. Subse- had a susceptibility of9.7 rather than 8.
quent plots at larger radii confirmed (by virtue of a linear Crude shimming with steel on the outside of the magnet
dependence of first degree Fourier coefficient upon radius) has been used for many years on a trial and error basis as a
that the dominant order was n = 1. method of eliminating gross inhomogeneity. It cannot, how-
Romeo and Hoult dictate that to produce a pure first ever, compare to the technique described above, for there are
degree correction, steel should be placed at ± 172.5", several factors beyond the investigator's control. First, the
± 142.5°, ± 127.5", and ± 82.5" azimuthal (t,b) positions rel- "turret" on the magnet, used for the entry of cryogens, re-
ative to the azimuth of the desired Til harmonic, as shown in stricts the placement of the steel. It is highly likely that one
Fig. 2. However, the length ofthe steel still has to be chosen. would wish to place steel in the region covered by the turret,
Steel paranel to the z axis and centered at z = 0 produces no for the building engineer may want the magnet above a sup-
even order, first degree harmonics. Hence, the arrangement porting beam because of its weight. Second, any steel on the
of Fig. 2 produces TI]o T 31 , T 51 , etc. One then questions how outside is at a radius more than double that of steel inside
important the higher order harmonics are, given that the and so at least ten times the quantity is needed. Third, be-
ratio of r to f in Eq. (11) is no more than 1:3. The easiest cause the field outside the magnet is not totally in thez direc-
approach, which we adopted and now regret a little, is to set tion, it is difficult to predict the effects of external steel and to
the coefficient of T31 in Eq. (11) equal to zero, viz calculate desirable lengths. Our magnet manufacturers at-
tempted an initial shimming with external steel and used at
P4 I {cos a) = ~ sin a(7 cos3 a -
3 cos a) = O. (12) least 40 times more than we did. Further, it is easy to see that
2 if an arbitrary length of steel is used, considerable higher
Thus, rand /3, the limits of integration, are 49.1° and 130.9°, order harmonics may be introduced. For example, pieces of
respectively, and h = g = f = a/sin 49.1°, where a is the cy- steel placed at if,a,t,b) to annul TIl' introduce T31 as
lindrical radius at which the steel was placed. Use ofEg. (11)
to assess the importance of the T51 harmonic introduced by
the steel shows that for the desired TIl correction, the
H31 = sin 2 a P5 (cos a) (..!...-)2P
I 31 (COS B), (15)
H\I P 31 (COS a) a P\I(cos B)
amount of T51 at a radius r is
where we have differentiated Eq. (5) with respect to z. Thus,
HSI =
4
sin a P61(COS a) (..!...-)4 PSI(cos B) (13) while correcting (at say, a = 90°) for an inhomogeneity of 1.2
HII 3 P 21 (COS a) a PII(cos B) parts per thousand caused by Til' we introduce at r = 15 cm,
for a = 100 cm, an inhomogeneity of 50 ppm as T 31 , which is
e
Hence for a = 49.1°, = 90°, a = 46.2 cm, and r = 15 cm, unacceptable. In our opinion external steel is good only for
the ratio is 4.7X 10- 3 • As the original inhomogeneity was small final adjustments, where its use may be more conven-
1.2 parts per thousand, it was estimated that the T51 pro- ient.

This article133
is copyrighted as indicatedVol.
Rev. [Link]., in the
56,article.
NO.1, Reuse of 1985
January AIP content is subject to the terms
Imaging at: [Link] Downloaded
magnet 133 to IP:
[Link] On: Fri, 05 Dec 2014 [Link]
 nents of Eq. (18) somewhat. The result of these efforts is
Proton orr8et shown in Fig. 4. Plotting the field over the surface of a sphere
"~requency
in kH2
15 em in radius revealed a peak-to-peak variation offield of

\ \
3
.With Tenere' Steel
OWith Zonal & Tesserlll Steel,
&. Current ~hiDUI

--T" Steel', Cont<ibulio. /

about 10 ppm-the desired specification. Outside this
sphere, however, the homogeneity degraded rapidly due to
the T51 harmonic introduced by the steel strips.
A final point of concern was whether or not the rapid
switching of the field gradients necessary for imaging would
.\ ! 30 .».m. j. be marred by the steel. However, while some change was
.", ./ noted, it was clear that far more steel could have been toler-
ated. It is worth noting the Eq. (II) may be used to predict
., -/
" '-
o_o-o-o-o-o~Q::::--0-~
"
"'.
"............ ~.~)-O-O-O-O_
/'
~/
.~ approximately the amount needed, prior to installation. For
example, if we consider that the steel beam in Fig. 2 is 1.3 m
from the magnet axis, that its effective magnetic length is
-10
o Axial POliilion in cm&. 10 about the same value, and that it is saturated, then Eq. (11)
shows that, within the magnet,
FIG.4. Axial NMR field plot (8 = DO) following the insertion of all the re-
quired steel bars. The ordinate is in proton Hz. the broken line shows the Hll ;:::5.4X IOZTll (28.4 kHz/mI. (19)
contribution of those bars used to cancel Til' Also shown is the final plot.
after the addition of extra steel to reduce second and fourth order compo- From Fig. 3, the measured gradient is 43.6 kHz/m which,
nents and adjustment of the shim coil currents. considering the uncertainties inherent in the assumptions,
and the neglect of other steel in the suite, is in reasonable
agreement with Eq. (19). When it is more appropriate to con-
III. ZONAL EFFECTS sider the approximation that the steel is at a single point,
then Eq. (35) of Ref. 1 should be used. In this case, the effect
The above description has concentrated on the removal
on first order homogeneity decreases as the fourth power of
of the harmonic Til' as it was the dominant inhomogeneity.
distance from magnet, which is why emphasis in this writing
Steel strips, as has been mentioned, were also used to remove
has been upon steel in the floor. The authors feel, in conclu-
T22 and T 33 , and pieces of steel were employed to annul T32 ,
sion, that the main objection to placing NMR imaging
T 43 , and TS3 in the manner prescribed by Romeo and Hoult.
equipment in an ordinary hospital department has been
Field plotting involved Fourier analysis of azimuthal plots at
overturned. The NIH environment is very bad, and the mag-
various radii and z positions, followed by curve fitting of
net was considered (by some) unshimmable. This was not so,
each degree's Fourier components to polynomials in r and/
and provided a magnet is aligned with steel beams in the
or z to find the orders (n) comprising each degree (m). How-
floor, should not be so in other locations too.
ever, azimuthal plotting followed by Fourier analysis conve-
niently ignores zonal (m = 0) harmonics, I and of course, we
APPENDIX: THE GENERATION OF A PURE T11
may expect that all the steel positioned symmetrically about
HARMONIC
the center (z = 0) plane will produce substantial even order
variations of Hz with z. From Eq. (11), the zonal contribu- From Eq. (11), the coefficient of TSI in the expansion of
tions of a single strip are the field produced by a steel bar parallel to the z axis is pro-
portional to P61 (cos a). As l
H=
- xH A [
~
z 3P3(cos asI + 5Ps(cos
Z2
a~
Z4

I
+ ...
] (h.y)

~61
. (16)
P61 (COS a) = (2l/8)sin a
Thus, for the eight strips used to produce Til' X (33cos 5a - 30 cos 3 a + 5 cos a), (AI)
Z2 Z4)
H = xHzA (0.3494 + 0.334=- + .... (17) it is clear that no Ts, harmonic is introduced by bars whose
6
a a
ends are at declinations 33.88° and 146.12°, and at 62.04° and
Using the value for A obtained from Eq. (14) for x = 8, we 117.96°. Now the coefficient of T31 in the expansion is pro-
have that the TIl steel correction shims introduce a zonal portional to P41 (cos a)lls and for the two sets of [Link]
field given by above, P41 (cos a) has opposite signs. Thus, by using the ap-
H = 1.33 X lQ3zZ + 5.98 X 1.03z4 + .... (18) propriate amounts of steel in two bars subtending 112.24°
Figure 4 shows an axial plot offield versus distance which is and 55.92°, T31 may be cancelled, leaving T ll , T7l , etc.
in good agreement with calculation when the total quantity P4 I (cos a)/Is = sinS aP41 {cos alias,
of steel used for other corrections is taken into account. and at 33.88°
Thus, we see that by introducing steel to cancel tesseraJ
(m #0) harmonics, we have substituted instead zonal har- sinSaP41 (cos a) = + 0.114,
monies, for which shim coil systems are highly efficient. Un- while at 62.04°, its value is - 0.813. It foHows that the bar
fortunately, not quite enough current was available to subtending 112.24° should have a cross-sectional area 7.13
smooth this plot, and so, folJowng Romeo and Hoult, 24 greater than that subtending 55.92°. Let the area of the latter
small pieces of steel were placed in two rings at a = 43° and beA. Then, from Eq. (11). the combined Til harmonic pro-
13i in order to reduce the second and fourth order compo- duced by the two bars is

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: [Link] Downloaded to IP:
134 Rev. [Link]., Vol. 56, No.1, January 1985 Imaging magnet 134
[Link] On: Fri, 05 Dec 2014 [Link]
 ACKNOWLEDGMENTS
HII = - xHzA{ . 3 6204'
41Ta 3 [P21 (cosa)sm a] 117.96'
The authors gratefully acknowledge the help of J. Ross
+. . 3a ]3388'}T
7 13[P21 (COS a )sm (A2)
146.12' 11 and R. Dienesch, and computational assistance from Dr. V.
or Sank and Dr. C-N. Chen.
Hll = - 0.410 xHzA (r/a 3 )sin (J cos(tP - I/J). (A3)
Allowing for the eight sets of bar pairs at the various azi-
muthal (I/J) positions needed to produce a pure first degree
harmonic, i.e., ± 172.5°, ± 142.5", ± 127.5", and ± 82.5" IF. Romeo and D.l. Hoult, Magn. Reson. Med. 1,44 (1984).
relative to the azimuth needed, we have that 2J. D. Jackson, Classical Electrodynamics (Wiley, New York, 1975).
3p. M. Morse and H. Feshbach, Methods of Theoretical Physics (McGraw-
HII = 1.86xHz A (r/a 3 )sin (J cos(tP - I/J). (A4) Hill, New York, 1953).
"Handbook ofMathematical Functions, edited by M. Abramowitz and I.A.
Thus, we see, by comparison with Eq. (14), that removal of Segun (Dover, New York, 1968).
T51 when generating a given Til harmonic requires about 3.2 sCRC Handbook of Chemistry and Physics, edited by R. C. Weast (CRC,
times more steel by volume-not too high a price to pay. Cleveland, 1978).

This article is135

copyrighted as indicated inVol.
Rev. [Link]., the56,
article.
No.1, Reuse of AIP
January content is subject to Imaging
1985 the termsmagnet
at: [Link] Downloaded
135 to IP:
[Link] On: Fri, 05 Dec 2014 [Link]