Magnetic diffusion, inductive shielding, and the Laplace transform

Alexander E. Krosney, Michael Lang, Jakob J. Weirathmueller, and Christopher P. Bidinosti

Citation: American Journal of Physics 89, 490 (2021); doi: 10.1119/10.0003508

View online: https://doi.org/10.1119/10.0003508
View Table of Contents: https://aapt.scitation.org/toc/ajp/89/5
Published by the American Association of Physics Teachers

ARTICLES YOU MAY BE INTERESTED IN

Schrödinger's equation as a diffusion equation

American Journal of Physics 89, 500 (2021); https://doi.org/10.1119/10.0002765

Space-time computation and visualization of the electromagnetic fields and potentials generated by moving point
charges
American Journal of Physics 89, 482 (2021); https://doi.org/10.1119/10.0003207

Particle velocity = group velocity: A common assumption in the different theories of Louis de Broglie and Erwin
Schrödinger
American Journal of Physics 89, 521 (2021); https://doi.org/10.1119/10.0003165

Simple derivation of the explicit forms of quantum-mechanical fundamental representations

American Journal of Physics 89, 535 (2021); https://doi.org/10.1119/10.0003900

Using mobile-device sensors to teach students error analysis

American Journal of Physics 89, 477 (2021); https://doi.org/10.1119/10.0002906

A guide to the literature of the finite rectangular well

American Journal of Physics 89, 529 (2021); https://doi.org/10.1119/10.0003327
Magnetic diffusion, inductive shielding, and the Laplace transform
Alexander E. Krosney,a) Michael Lang,b) Jakob J. Weirathmueller,c)
and Christopher P. Bidinostid)
Department of Physics, University of Winnipeg, 515 Portage Avenue, Winnipeg, Manitoba, Canada R3B 2E9
(Received 22 September 2020; accepted 25 January 2021)
In the quasistatic limit, a time-varying magnetic field inside a conductor is governed by the diffusion
equation. Despite the occurrence of this scenario in many popular physics demonstrations, the
concept of magnetic diffusion appears not to have garnered much attention itself as a subject for
teaching. We employ the model of a thin conducting tube in a time-varying axial field to introduce
magnetic diffusion and connect it to the related phenomenon of inductive shielding. We describe a
very simple apparatus utilizing a wide-band Hall-effect sensor to measure these effects with a variety
of samples. Quantitative results for diffusion time constants and shielding cutoff frequencies are
consistent with a single, sample-specific parameter given by the product of the tube radius, thickness,
and electrical conductivity. The Laplace transform arises naturally in regard to the time and
frequency domain solutions presented here, and the utility of the technique is highlighted in several
places. VC 2021 Published under an exclusive license by American Association of Physics Teachers.
https://doi.org/10.1119/10.0003508

I. INTRODUCTION with higher frequencies, thicker walls, and larger conductivi-

ties, all of which lead to increased inductive shielding. This
Demonstrations of the interaction between magnetic fields point of view also leads one to anticipate a phase lag
and non-magnetic conducting materials are very important— between the applied field and the field that has managed to
and popular. They provide strong, and often quite dramatic, diffuse into the enclosure. While this interpretation is cer-
visualizations of the Lorentz force, Faraday’s law, and tainly consistent with the physics at hand, one must concede
Lenz’s law, and are typically employed to launch a discourse that it is not as satisfactory as a more intuitive, standard
on the fact that electricity and magnetism, while seemingly description in terms of induced (eddy) currents and Lenz’s
disparate phenomena in static configurations, are truly one law, say. Moreover, for the steady-state sinusoidal case, with
and the same when time variation occurs. Particular manifes- Bðr; tÞ ¼ BðrÞeixt , the right hand sidepof Eq. (2) is readily
tations of such interactions, such as eddy currents or mag- ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
replaced with 2iBðrÞ=d2 (where d ¼ 2=l0 rx is the elec-
netic braking, are discussed to varying degrees in standard tromagnetic skin depth1–3), which only further obscures the
textbooks,1–3 while much greater variety and detail can be form and function of the diffusion equation. It appears then
found in the literature.4–18 that despite the generality and ubiquity of Eq. (2), the con-
To motivate our work on the related but seemingly lesser- cept of magnetic diffusion can indeed be easily overlooked.
known topic of magnetic diffusion, we begin by writing the One scenario where the notion of a magnetic field diffus-
differential equation for a time-varying magnetic field in a non- ing through a conductor is quite natural and intuitive is the
magnetic conductor (see, for example, Sec. 9.4.1 of Ref. 1) step response.3,19–21 Here, a switched magnetic field that
would be established (near) instantaneously in the absence of
@Bðr; tÞ @ 2 Bðr; tÞ
$2 Bðr; tÞ ¼ l0 r þ l0  ; (1) the conductor now takes time to diffuse through the bulk of
@t @t2 the material as induced eddy currents within it decay. This
where l0 is the permeability of free space, and r and  are approach is well established in the research literature,22–27
the electrical conductivity and permittivity of the conductor, but is largely unknown as a teaching demonstration, as far as
respectively. In the limit of slowly varying fields, indicative we can tell. In this paper, we present a simple experiment
of the scenarios in the many papers mentioned above, the that employs a wide-band Hall-effect sensor to directly mon-
second-order time derivative associated with the displace- itor the process of magnetic diffusion and determine associ-
ment current is negligible. As a result, in this quasistatic ated time constants. The same apparatus is used without
limit, Eq. (1) is safely replaced with modification to make ac measurements as a function of fre-
quency and quantify the onset of inductive shielding. The
@Bðr; tÞ two phenomena are linked, of course, and we also take this
$2 Bðr; tÞ ¼ l0 r ; (2) opportunity to highlight the utility of the Laplace transform
@t
for analyzing the response of the system in both the time and
which is indeed a form of the diffusion equation,3,19–21 frequency domains.
though it is rarely noted as such and few of the phenomena it This paper is organized as follows. First, we develop the
describes are ever regarded as diffusive processes. model of the thin conducting tube in a time-varying axial
In light of this circumstance, one is tempted to re-interpret field, providing both the step and steady-state sinusoidal
some such phenomena in terms of magnetic diffusion. For response and linking the two via the Laplace transform. This
example, an ac magnetic field impinging on a metallic enclo- model is intuitive, informative, and easy to realize experi-
sure can be viewed as having insufficient time each half- mentally. Next, we describe the apparatus and experimental
cycle to fully diffuse into the conductive structure before it procedure that we employ at our institution as an undergrad-
must start to diffuse out again. Such an effect is exacerbated uate laboratory exercise. We then present and discuss results

490 Am. J. Phys. 89 (5), May 2021 http://aapt.org/ajp C 2021 Published under an exclusive license by AAPT
V 490
for magnetic diffusion through conducting slabs and tubes, For the step field Bo ðtÞ ¼ Bo for t  0, the solution for ini-
as well as inductive shielding of the latter. In the appendices, tial condition Bi ð0Þ ¼ 0 is easily found to be19,20
we provide known solutions for the general case of a con-
ducting tube of arbitrary thickness, and also present the Bi ðtÞ ¼ Bo ð1  et=s Þ: (8)
design method and characterization of the coil we built to
generate a highly homogeneous magnetic field over the The magnitude of the current density in the tube is
length of the tube samples. therefore
Bo t=s
II. THE THIN CONDUCTING TUBE IN A TIME JðtÞ ¼ e : (9)
l0 h
VARYING AXIAL FIELD
Following the approach of Haus and Melcher,19 we con- Equations (8) and (9) show that the application of a dc step
sider a long, cylindrical, non-magnetic tube with inner radius field leads, by Lenz’s law, to an induced field that opposes
a, outer radius b, thickness h  b  a, permeability l0, and Bo. The overall interpretation is that the applied field diffuses
conductivity r, subject to an applied, uniform, axial mag- into the tube with the same characteristic time as the decay
netic field Bo ðtÞ. Assuming the tube to be thin (b=a  1), of the induced current. Since s depends not only on the tube
dimensions but also its composition, materials with high con-
Ampère’s law can be written as the boundary condition
ductivity, such as copper, will produce long-lasting induced
Bi ðtÞ  Bo ðtÞ ¼ l0 hJðtÞ; (3) currents (i.e., slow diffusion of Bo), whereas a poor conduc-
tor will produce short-lived induced currents (i.e., fast diffu-
where Bi ðtÞ is the total axial field inside the tube, J(t) is an sion of Bo.)
azimuthal (eddy) current density that is uniform over the We now consider an ac field of the form Bo ðtÞ ¼ Bo eixt ,
thickness of the tube, and the quantity hJ(t) is the instanta- with frequency-independent magnitude Bo. For the steady-
neous surface current that leads to the discontinuity of the state sinusoidal response, one anticipates a solution of the
field inside and outside the tube. This is, of course, just the form Bi ðtÞ ¼ Bi ðxÞeixt . By substituting Bo ðtÞ and Bi ðtÞ into
model of an infinitely long solenoid, which produces an axial Eq. (6), one finds the complex amplitude8,9,17,22
field of magnitude l0 hJðtÞ inside its volume and zero field Bo
outside. (Indeed, one could approach this entire problem as a Bi ðxÞ ¼ ; (10)
1  ixs
long, thin, single-turn inductor of length ‘, resistance
R ¼ 2pa=rh‘, and inductance L ¼ l0 pa2 =‘, as done by which reduces to Bi ðxÞ  Bo ð1 þ ixsÞ at low frequency9,10
I~
niguez et al.9) The assumption that the current density is and goes to zero at high frequency. From this result, the
uniform also implies that the tube is electromagnetically induced magnetic field generated by J(t) is easily
thin, i.e., h  d. determined
From Faraday’s law of induction, we also have
ixs
dU Bi ðtÞ  Bo ðtÞ ¼ Bo ðtÞ : (11)
E¼ ; (4) 1  ixs
dt
As expected, it is 90 out of phase with the applied field and
where E ¼ EðtÞ 2pa is the electromotive force associated tends to zero as x ! 0 (i.e., xL  R, poor shielding), while
with the induced electric field E(t) driving eddy currents it is 180 out of phase with the applied field and approaches
around the tube, and U ¼ Bi ðtÞ pa2 is the magnetic flux Bo ðtÞ as x ! 1 (i.e., xL  R, complete shielding).
through the tube. Using Ohm’s law J ¼ rE, Eq. (4) can be The relative magnitude of the net internal field can be
written as written as
2pa d jBi ðxÞj 1
JðtÞ ¼ pa2 Bi ðtÞ: (5) ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (12)
r dt Bo
1 þ ðxsÞ2
Combining Eqs. (3) and (5) and rearranging leads to the
differential equation 1
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; (13)
dBi ðtÞ Bi ðtÞ Bo ðtÞ 1 þ ðah=d2 Þ2
þ ¼ ; (6)
dt s s
and the onset of inductive shielding is seen to occur at a cut-
with time constant off frequency
1
s ¼ l0 rah; (7) fc  ð2psÞ1 (14)
2
containing the sample-specific parameter ðrahÞ. Equation ¼ ðpl0 rahÞ1 ; (15)
(6) is completely general and can be used to determine the which can also be cast as a critical thickness
total magnetic field inside a thin conducting tube for any uni-
form applied field Bo ðtÞ, as is done below for a step field and hc  d2 =a (16)
a sinusoidal field. Corresponding solutions for the general
case of a tube of any thickness are presented in Appendix A. ¼ ðpl0 raf Þ1 : (17)

491 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 491
The latter result, noted by Fahy et al.,8 often comes as a sur- magnetic diffusion. The two experimental configurations
prise: It says that the onset of shielding for this geometry are shown in Fig. 1.
occurs not when the skin depth is comparable to the thick- The specific details of the various measurements, as well
ness of the tube (d h)—a commonly held misconception— as the dimensions and properties of the different samples, are
but when d2 ah.8,28 As a result, an electromagnetically presented further below. In general, however, the experimen-
thin tube with h  d can still provide efficient shielding if tal procedure is quite simple. An electromagnetic coil,
a  d. described in detail in Appendix B, is driven by a waveform
We end here by re-deriving the above results via the generator (Rigol DG1032Z) to generate the applied magnetic
Laplace transform approach. In doing so, our intention is not field. Field measurements are made by a Hall-effect sensor
to provide an overview of such methods—which are well (Sentron CSA-1VG)35 with its differential output connected
described in many standard textbooks29–32 and are further to two separate channels of an oscilloscope (Agilent DSO-X
discussed in the pedagogical literature33,34—but rather to 2014A). The internal math function of the oscilloscope
take the opportunity to highlight this elegant mathematical is used to determine the difference of the two channels,
tool in the context of a very interesting and simple electro- providing a final signal that is proportional to the magnetic
magnetic system. We will also use the Laplace transform in field at the location of the sensor. Signal averaging using the
two other instances in this paper, making use of standard built-in functionality of the oscilloscope is also employed.
transform-pair lookup tables throughout.29–32 The addition of a differential amplifier between the sensor
The Laplace transform of Eq. (6) with initial condition and oscilloscope could further improve performance,28,36 but
Bi ð0Þ ¼ 0 is was purposely omitted here in order to keep the apparatus as
simple as possible and hopefully further encourage the adop-
Bi ðsÞ Bo ðsÞ tion of magnetic diffusion and ac shielding studies in the
sBi ðsÞ þ ¼ ; (18)
s s undergraduate teaching lab. The 100-kHz bandwidth of
the sensor35,36 is suitable for the time and frequency scales
where B(s) denotes the transform of a magnetic field B(t), explored here.
and s is a complex frequency in general. Solving for Bi ðsÞ
gives the Laplace transform of Bi ðtÞ A. Step field measurements
Bo ðsÞ Step field measurements were performed by driving the
Bi ðsÞ ¼ : (19)
1 þ ss coil with a 50-Hz square-wave voltage alternating between
zero and 7.5 V. The input couplings of the oscilloscope were
For the dc step field turned on at t ¼ 0, Bo ðsÞ ¼ Bo =s and set to dc, and 256 averages were used. Because the analog
Eq. (19) becomes output channel of the Hall-effect sensor is referenced to its
Bo
Bi ðsÞ ¼ ; (20)
s þ s2 s
which has inverse transform Bi ðtÞ ¼ Bo ð1  et=s Þ, identical
to Eq. (8).
The steady-state sinusoidal response, on the other hand,
is given by Bi ðtÞ ¼ HðsÞ Bo eixt with transfer function
HðsÞ  Bi ðsÞ=Bo ðsÞ found from Eq. (19) and evaluated at
s ¼ ix.31 This gives
 
Bo
Bi ðtÞ ¼ eixt ; (21)
1  ixs

where the term in braces is the complex amplitude of

Eq. (10). It is also worth noting that the transfer function
found here is identical to that of a low-pass filter, which is
expected intuitively for inductive shielding (i.e., low fre-
quencies pass through the tube, high frequencies do not) and
also given that an equivalent LR circuit model can be used to
analyze this system.9

III. EXPERIMENTAL METHOD

We now describe the apparatus and experimental pro-
cedure that we use in an undergraduate laboratory exer- Fig. 1. Cutaway models of the two experimental configurations. The 3D-
cise to make quantitative studies of magnetic diffusion printed coil former (white) was made as a single piece with grooves for wire
windings, and square end supports that act as a stand. Top: the conductive
and inductive shielding in conducting tubes based on
slab is placed between the Hall-effect sensor and one end of the coil.
the preceding theoretical model. We also present an Bottom: the sensor is placed at the center of the coil and conductive tube.
ancillary experiment using conducting slabs that pro- Both the sensor and tube are held in position by additional 3D-printed parts,
vides a very simple and intuitive demonstration of which are not shown.

492 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 492
 !
common output channel held at 2.5 V (giving a full scale dif- sf et=sf  set=s
ferential output of 62.5 V),35 the gain of both scope channels SðtÞ ¼ So 1  ; (24)
was set to 500 mV/div. An offset of 2.5 V was applied to sf  s
both channels, and the sensitivity of the math waveform was  
set to 10 mV to capture the much smaller differential signal. t þ s t=s
¼ So 1  e when sf ¼ s; (25)
The time constant of our coil is calculated to be 2.2 ls s
(L ¼ 129 lH, R ¼ 8:13 X þ 50 X from the Rigol DG1032Z
output), which is less than both the scan time ( 3 ls) and ! So ð1  et=s Þ when sf  s: (26)
rise time ( 3 ls) of the sensor35,36 and does not limit the
overall bandwidth of the system. The signal of a step field One can also derive Eq. (24) from the Laplace convolution
measurement performed at the center of the coil in the of So ð1  et=s Þ with the impulse response of the low pass
absence of any conducting sample is shown in Fig. 2. filter ð1=sf Þet=sf .29–32
Given the finite bandwidth of the Hall-effect sensor, its Given that the magnetic diffusion time constants measured
output signal S(t) must also be determined by the differential in this work are around two orders of magnitude larger than
equation for a low-pass filter the rise time of the Hall-effect sensor (sf 3 ls), Eq. (26) is
appropriate here. As a result, we use as a fit function
dSðtÞ SðtÞ kBðtÞ
þ ¼ ; (22) SðtÞ ¼ So ð1  eðtts Þ=s Þ; (27)
dt sf sf

where B(t) is the magnetic field at the location of the sensor, where ts ¼ 3 ls is the scan time of the sensor, and time t is
sf is the time constant (or rise time) of the sensor, and k is its measured with respect to the function generator trigger cor-
sensitivity ( 280 V/T for the CSA-1VG).35 For a magnetic responding to the rising edge of the square-wave drive volt-
field of the form of Eq. (8), one can solve Eq. (22) directly age. If thinner tubes (i.e., shorter s) or narrower-band sensors
by standard methods for linear first-order equations.29,30 (i.e., longer sf) are employed, one may need to use Eq. (24)
Conversely, one can solve by Laplace transform. By analogy instead. Finally, we note that given the very similar time
with Eqs. (6) and (18), and by making use of Eq. (20) for the constants of our coil and sensor, the exponential fit in Fig. 2
Laplace transform of the field of Eq. (8), one quickly finds could be replaced by something akin to Eq. (25). This has lit-
the Laplace transform of S(t) tle bearing on our present study, however, and a more
detailed analysis is unwarranted here.
1 So
SðsÞ ¼ ; (23) B. Ac field measurements
1 þ ssf s þ s2 s
Here, the coil is driven by a 15 Vpp sine wave at 101 logarith-
where So ¼ kBo. The general solution for the sensor output, mically spaced frequencies over the range f ¼ 1–10,000 Hz.
along with two particularly useful limits, is thus The oscilloscope couplings and sensitivities are the same as in
Sec. III A. Over this frequency range, the magnitude of the drive
circuit impedance increases by only 1%; however, the phase
varies by about 8 , which is not insignificant. To account for
these changes, as well as any potential frequency dependence in
the receive chain, we also recorded the phase of the differential
signal (relative to the trigger) and repeat the same set of fre-
quency measurements with and without the conducting tubes.
The latter represents a background measurement that is used to
correct the tube data with respect to the phase and magnitude of
the applied field at each frequency. A similar process has been
described elsewhere.13 Custom PYTHON code was written to auto-
matically pass frequencies to the waveform generator and return
amplitude and phase measurements from the oscilloscope.
Signal averaging is set to 16 for the lowest frequencies and is
dynamically increased via the program for higher-frequency
measurements with tubes, which otherwise would suffer from
reduced signal-to-noise ratio (SNR) due to greater inductive
shielding. Overall, this strategy helps minimize run time. From
Eqs. (13) and (14), data for the corrected signal amplitude are fit
to the following function to extract the cutoff frequency fc:
So
jSðf Þj ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (28)
1 þ ðf =fc Þ2
Fig. 2. The differential signal from the oscilloscope for a step field measure-
ment using the configuration shown in the bottom of Fig. 1 with no tube.
The smooth curve is an exponential fit to the data assuming an offset of C. Samples
exactly 3 ls to account for the sensor scan time. The fit yields a time con-
stant of 2:1 l s, consistent with the quoted rise time of the sensor and previ- The response to a step field applied perpendicular to the
ous tests (Refs. 35 and 36). face of a slab was measured for three samples—one each of

493 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 493
Table I. Measured properties of the tube samples. The numbers in parenthe- and measured conductivities of the tubes, which are needed
ses are the uncertainty in the last digit(s) of each quantity, as determined by for a quantitative analysis of magnetic diffusion and induc-
standard techniques (Ref. 37). The tubes are seamless, and we assume their tive shielding via the models in Sec. II, are given in Table I.
electrical properties to be isotropic. A dc current of 5.000(5) A was used for
all resistivity measurements.
The length ‘, outer diameter 2b, and thickness h of the tubes
were measured with digital calipers at five different positions
h b ‘ V q r each to account for non-uniformities. The resistivity q of the
Tube (mm) (mm) (mm) (lV) (108 X m) (107 S/m) samples was determined via a standard 4-wire measurement
by driving a known current I through the tubes and measur-
Copper #1 0.804(7) 6.357(9) 153.0(5) 422(2) 1.66(3) 6.02(11) ing the voltage drop V across them with a digital multimeter
Aluminum 1.468(5) 12.706(4) 153.5(5) 313(3) 4.50(5) 2.22(2) (Agilent 34411A). From V ¼ IR and R ¼ q‘=A, where A is
Copper #2 1.672(15) 9.507(6) 152.0(5) 144.4(7) 1.73(2) 5.78(6) the cross-sectional area of the tube and a ¼ b  h is its inner
radius, one finds

copper, brass, and plastic. The nominal dimensions of the 1 V pðb2  a2 Þ V pð2bh  h2 Þ
slabs—which were in fact borrowed from our own sliding q¼ ¼ : (29)
r I ‘ I ‘
magnet demo—are 10 in. long, 2 in. wide, and 1/4 in. thick.
The response to both step and ac fields applied along the axis
of a tube was measured for three samples—two of copper IV. RESULTS AND DISCUSSION
and one of aluminum—each nominally 6 in. long, but with
different diameters and thicknesses. The precise dimensions A. Magnetic diffusion through slabs and tubes
The results for the slabs are shown in Fig. 3. The SNR is
poor for these measurements largely due to the much weaker
applied field at the location of the sensor in this configuration
(see Figs. 1 and 8.) This could be improved by using a small,
flat coil of many turns placed directly on the face of the slab.
Still, the results presented here clearly demonstrate that, as
expected, magnetic diffusion through copper is slower than
it is through brass, since the former is the better conductor.
Also, the step field is seen to pass through the non-
conducting plastic slab near-instantaneously (i.e., indistin-
guishable from the rise time in Fig. 2). Another result worth
noting in Fig. 3 is the near-instantaneous jump seen in the
field for copper and brass, when the sensor is placed directly
behind the slab but not on the center line (or axis) of the coil.
By symmetry, it is only at the central location where the net
field is expected to be uniquely zero just after the coil is
turned on. At any other location, the induced field does not

Fig. 3. The magnetic field measured at the face of the slabs opposite to the
coil. The legend refers to both graphs. Top: results for a measurement posi- Fig. 4. The magnetic field measured at the center of the tubes. For clarity,
tion that is on the center line of the coil. Bottom: results for a measurement only every 50th data point is shown. The solid lines are least-square fits to
position that is 2.5 mm away from the center line toward the top of the slab. Eq. (27) for all data at t  ts ¼ 3 ls.

494 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 494
Table II. Ratios of tube radii; the time constants predicted from the thin-tube B. Inductive shielding by tubes
model using the values given in Table I; the measured time constants
extracted from fits to the data in Fig. 4; and the cutoff frequencies calculated Example data of the ac signals measured at the center of
from sfit via Eq. (14). the tubes are shown in Fig. 5 for a drive frequency of 1 kHz.
At this frequency, all samples exhibit a clear phase shift with
Tube a/b sthin ðlsÞ sfit ðlsÞ fc (Hz) respect to the voltage trigger, as well as a reduction in ampli-
Copper #1 0.874(2) 169(4) 179.3(2) 887.8(7)
tude with respect to the background value. These are both
Aluminum 0.8845(6) 230(3) 244.9(2) 649.9(5)
hallmarks of the presence of eddy currents and thus the onset
Copper #2 0.824(2) 476(7) 544.5(6) 292.3(3)
of inductive shielding. (Indeed, it is very informative for stu-
dents to simply observe how the signal on the oscilloscope
changes for a given sample as they increase the drive fre-
quency on the function generator.) From the comparative
necessarily cancel the applied field. Equivalently, one can
results shown in Fig. 5, it is also clear that the degree of
think of this result as being a consequence of the magnetic
shielding for the different tubes is consistent with their
field lines initially wrapping around the exterior of the slab
respective cutoff frequencies predicted from the diffusion
while its interior is still fully shielded by the induced cur-
measurements above (Table II) or by what one would esti-
rents. Again, because of symmetry, there can be no magnetic
mate from their properties in Table I via Eq. (15).
field at the central point located directly on the back (or
It is also possible to discern a small phase shift in the
front) face of the slab, since the approaching field lines must
background signal in Fig. 5. This is due to the complex
spread out in opposite directions about this point.
The results for the tubes are shown in Fig. 4 along with a
background measurement (i.e., no tube) for comparison.
Time constants are determined from fits to Eq. (27) with
ts ¼ 3 ls and are compiled in Table II along with predicted
values from the thin-tube model. Our results largely agree
with the rule-of-thumb that the thin-tube model should be
good to within 10% for a=b  2=3.20,25 The greatest dis-
crepancy is seen with our thickest sample, which perhaps
suggests that one should use tubes with a/b closer to 0.9, say,
if the goal is to provide teaching demonstrations that agree
very closely with the thin-tube model. We also point out that
time constants predicted from the general model (Eq. (A6))
are closer to, but still do not agree within error, with our
measured values. Still, these results provide an excellent
demonstration of magnetic diffusion, consistent with the
trends predicted from measured sample properties. With a
good degree of confidence, then, we use our values of sfit
here to predict the cutoff frequencies for the ac measure-
ments of Sec. IV B. These are also listed in Table II.

Fig. 6. Complex components of the internal magnetic field for the copper
tubes. The legend refers to both graphs. The solid and dashed-dotted lines
are the functional form for the general model (Eq. (A7)) and the thin-tube
Fig. 5. Signal waveforms at a drive frequency of 1 kHz. Amplitudes are nor- model (Eqs. (13) and (7)), respectively. The dashed lines are the low-
malized with respect to the background value (21.9 mV) measured here. frequency limit (Refs. 9 and 10) of the latter as discussed in Sec. II.

495 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 495
impedance of the drive circuit arising from the inductance of Table III. Cutoff frequencies predicted for the thin-tube model using the val-
the coil. To correctly extract the complex components of ues given in Table I; the measured cutoff frequencies extracted from fits to
the field inside the tubes, the raw data of the type shown in the data in Fig. 7; and the ratios h=d and ah=d2 calculated from fc;fit .
Fig. 5 must be corrected with respect to the phase and ampli-
Tube fc;thin (Hz) fc;fit (Hz) h=d ah=d2
tude of the background signal at each frequency as discussed
in Sec. III B. The results of this process are shown in Fig. 6 Copper #1 942(20) 946(2) 0.381(8) 1.00(4)
for the two copper samples. Overlaid on top of these data are Aluminum 691(8) 710(1) 0.366(4) 1.03(2)
curves for the general model (Appendix A), the thin-tube Copper #2 335(5) 327(1) 0.457(7) 0.98(2)
model, and the low-frequency limit of the latter generated
using the sample properties given in Table I. The low-
frequency model proposed by I~niguez et al.9,10 is suitable to seen in the background measurement at high frequencies.
a few hundred hertz or less for these samples, while the thin- This again is due to the small increase in coil impedance; all
tube model can extend the range of study by perhaps another sample data have been corrected for this by normalizing to
order of magnitude. The general model, on the other hand, the background amplitude at each frequency value as men-
provides excellent agreement over the full frequency range tioned above. With regard to the results in Table III, one can
studied here. The limitation of the thin-tube model is easily see that all measured values of cutoff frequency agree to
understood from the well-known rule-of-thumb that the skin within a few percent or less with the values predicted from
depth of copper is roughly 1 cm at 60 Hz, which translates to the thin-tube model. Also, as shown in the last two columns
3 mm at 670 Hz or 1 mm at 6 kHz. Looking at the copper of the table, the onset of inductive shielding does indeed
tube thicknesses given in Table I, then, one sees that the con- occur when d2 ah and not when d h.8,28
dition of being electromagnetically thin (i.e., h  d) will
certainly break down over the frequency range studied here,
V. CONCLUSION
and deviations from the thin-tube model are to be expected
at the higher frequencies. A review of the literature reveals that the concept of mag-
To better highlight the inductive shielding of the tubes, as netic diffusion is rarely considered for the purposes of peda-
well as their behavior as low-pass filters, the magnitude of gogy.38 The thin conducting tube in a uniform, time-varying
the internal field is plotted versus frequency in Fig. 7 for all axial field provides a complete and very accessible model for
tube samples. The low-frequency end of the data are fit to exploring magnetic diffusion as well as the related phenome-
Eq. (28), and cutoff frequencies are compiled in Table III non of inductive shielding. The product of the tube radius,
along with predicted values from the thin-tube model. We thickness, and electrical conductivity provides a single,
chose to limit the fitting range to data with normalized sample-specific parameter that sets both the time constant
amplitude greater than 0.5, which from Fig. 6 still show for stepped dc fields to diffuse through the tube and the cut-
good agreement with the thin-tube model. The predicted cut- off frequency for ac fields to penetrate the interior of the
off frequencies from Table II are also presented in Fig. 7 and tube. While not required, the use of the Laplace transform to
show good agreement with the ac measurements. A final fea- solve for and link the time and frequency domain solutions
ture of interest in Fig. 7 is the slight decrease in amplitude of this system further broadens the educational experience
here.
A simple apparatus utilizing a wide-band Hall-effect sen-
sor allows either stepped or ac measurements without any
configurational changes. The addition of a differential ampli-
fier following the Hall-effect sensor could further improve
performance. As it stands, the present setup provides more
than sufficient SNR to make meaningful qualitative and
quantitative tests on a variety of samples. Time constants
and cutoff frequencies extracted from the two types of mea-
surement for conducting tubes show good agreement with
each other as well as with predicted values.
Through a judicious choice of frequency range and tube
thickness, one can design a student laboratory experiment
that resides fully within the limits of the thin-tube model
(h  a; d). This could be desirable from the point of view of
minimizing the amount of information needed to understand
the experiment. Still, many students will likely wonder what
happens as the skin depth becomes smaller than the tube
thickness, say. The general solutions (also provided here)
answer such questions and should be well within reach for
an advanced undergraduate student.

Fig. 7. Normalized magnitude of the internal field as a function of fre- ACKNOWLEDGMENTS

quency. The solid lines are least-square fits of Eq. (28) to all data points
with ordinate value greater
The authors gratefully acknowledge the support of the
pffiffithan
ffi 0.5. The horizontal dashed line indicates the
Natural Sciences and Engineering Research Council of
half-power amplitude 1= 2 that defines the cutoff frequency of a low-pass
filter. The vertical dashed lines indicate the cutoff frequency predicted for Canada, especially the Undergraduate Student Research
each tube from the preceding step field measurements (see Table II). Awards for A.E.K. and J.J.W.

496 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 496
 !
APPENDIX A: SOLUTIONS FOR THE CONDUCTING X1
cn sn
TUBE OF ARBITRARY THICKNESS Bi ðsÞ ¼ Bo 1s ; (A8)
n¼1
1 þ sn s
For the general problem of a uniform, axial field Bo ðtÞ
¼ Bo ðtÞ ^z applied to a non-magnetic, conducting tube of any which evaluated at s ¼ ix gives an alternative form of
thickness, one must solve the diffusion equation in the bulk Eq. (A7). We have shown the two solutions to be numeri-
of the tube, subject to boundary conditions. For this geome- cally equivalent for the tube parameters and frequency range
try, Eq. (2) becomes studied here. We did not attempt to prove mathematical
equivalence, although it appears the necessary details can be
@ 2 Bz ðq; tÞ 1 @Bz ðq; tÞ @Bz ðq; tÞ gleaned from the work of Jaeger.23
þ ¼ l0 r : (A1) The solutions presented in this appendix are also valid for a
@q2 q @q @t
non-magnetic, conducting tube in a uniform, transverse mag-
netic field.22,23 As a result, they also hold for a uniform field
Solutions are provided below; derivations are found in the applied at any angle to the axis of the tube. This is not the case
accompanying references. for the more general scenario of a magnetic tube, however.21–23
For a step field Bo ðtÞ ¼ Bo for t  0, the solution for the
internal field subject to initial condition Bi ð0Þ ¼ 0 can be
written as20,21,23 APPENDIX B: NUMERICAL OPTIMIZATION OF
! THE DRIVE COIL FOR IMPROVED
X 1
HOMOGENEITY
t=sn
Bi ðtÞ ¼ Bo 1  cn e ; (A2)
n¼1 We originally built a standard solenoid comprising two layers
of 100 evenly spaced windings to serve as a drive coil. The sole-
with coefficients noid was wound with #32 AWG enameled magnet wire on a
3D-printed former. We found the time constant of this coil was
4 J0 ðbcn ÞJ2 ðacn Þ greater than that of the Hall-effect sensor (see Fig. 2), so we
cn ¼ (A3)
a cn ðJ02 ðbcn Þ  J22 ðacn ÞÞ
2 2
decided to re-make it with the same dimensions and wire but
using only 50 windings per layer. We also took this opportunity
and time constants to optimize the winding pattern to provide greater field homoge-
l0 r neity over the length of the tube samples. While this does offer
sn ¼ ; (A4) greater fidelity with our theoretical models, it is not critical for
c2n obtaining satisfactory results, and preliminary tests with our
original solenoid did yield nearly identical time constants and
where cn is the n-th root of the equation cutoff frequencies to those reported above.
The details of the optimization algorithm are given below.
J0 ðbcÞY2 ðacÞ  Y0 ðbcÞJ2 ðacÞ ¼ 0; (A5) The final winding pattern and 3D former can be seen in Fig. 1.
The specific locations of the current loops are given in
and J and Y are Bessel functions of the first and second Table IV, allowing one to easily duplicate our coil or scale it to
kind of order . After sufficient time has passed any desired radius. The calculated magnetic field profiles of the
ðt > l0 r=c21 Þ, the field within the conductor volume can optimized coil and the original solenoid are shown in Fig. 8,
described by the n ¼ 1 term only. In this case, the field is along with those of the well-known Lee-Whiting and Helmholtz
once again given by Eq. (8) with time constant20 designs39 for comparison. The parameters of the various coils
are summarized in Table V. For the Lee-Whiting and
l0 r Helmholtz coils, we considered two designs: fixing either their
s¼ : (A6)
c21 length or their radius to equal those of our solenoid and opti-
mized coil. Measurements of the field profile along the axis of
For an ac field Bo ðtÞ ¼ Bo eixt , the solution for the com- the optimized coil are also shown in Fig. 8 and confirm the
plex amplitude of the internal field is8,22 expected improvement in homogeneity. Accurate measurements
of the off-axis field are more challenging to achieve and were
2
Bi ðxÞ ¼ Bo ½I0 ðzo ÞK2 ðzi Þ  K0 ðzo ÞI2 ðzi Þ 1 ; (A7)
z2i Table IV. The axial positions of the current loops comprising the optimized
coil in ascending order by column. The values, normalized to the coil radius
where zi ¼ ð1  iÞa=d; zo ¼ ð1  iÞb=d, and I and K are R and rounded to the third decimal point, give the distance zi to the i-th loop
modified Bessel functions of the first and second kind of on either side of the central plane of the coil (z ¼ 0). For our coil
order . One can show that this is equivalent to the result R ¼ 2.28 cm, which is the average radius for the two layers of #32 AWG
given by I~ niguez et al.,11 keeping in mind that the latter wire (thickness 0.2 mm).
ixt
employs e for the temporal dependence of the applied
field, which leads to a conjugate solution. 6zi =R
Finally, based on the discussion in Sec. II, the Laplace 0.163 0.887 1.744 2.350 3.129
transform of Bi ðtÞ for the step response divided by the 0.234 1.195 1.933 2.997 3.140
Laplace transform of the step function (1=s) should lead to 0.426 1.250 2.098 3.052 3.396
the complex amplitude Bi ðxÞ for the steady-state sinusoidal 0.525 1.261 2.201 3.063 3.451
response. The procedure is trivial here and starting from 0.745 1.555 2.339 3.118 3.505
Eq. (A2) one quickly arrives at

497 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 497
 Table V. Coil parameters for the optimized (Opt), solenoid (Sol), Lee-
Whiting (LW), and Helmholtz (H) coils. The coil efficiency v is defined
here as the calculated central field strength Bz ð0; 0Þ per unit current divided
by the total number of turns N comprising each coil type. The coil radius R
is taken to be the average of the two layers for the Opt and Sol designs;
while all turns are assumed to exist on the same radius for the LW and H
designs. The half-length L is the axial location of the outermost current loop
for all designs.

Coil v ðlT/A/turn) N R (cm) L (cm)

Opt 6.55 100 2.28 7.98

Sol 7.51 200 2.28 7.98
LW-1 4.07 26 8.48 7.98
H-1 2.82 2 15.95 7.98
LW-2 15.18 26 2.28 2.14
H-2 19.76 2 2.28 1.14

where zi is the distance to the i-th loop on either side of the

central plane of the coil, and the current I was set to unity.
To begin, all loops are evenly spaced as per a regular sole-
noid. Optimizing the axial field homogeneity over a distance
zopt requires minimizing the integral
ð zopt  
 BðzÞ  Bð0Þ 
  dz: (B2)
 Bð0Þ 
0

This was done by randomly selecting a pair of symmetric

loops and displacing them a distance dz away from and
towards z ¼ 0. If either displacement reduces Eq. (B2), the
changes are saved and another pair is randomly selected;
otherwise, the changes are discarded. Random selection con-
tinues until further displacement of all pairs does not result
in an improvement, in which case the value of dz is
decreased. Once dz is reduced beyond a set minimum thresh-
old value (typically given by the resolution of the 3D
printer), the program exits and saves the final zi values.
To prevent overlap and wire grooves with a separation
wall smaller than printing capabilities, additional constraints
are placed on the current loop locations. If moving a pair pla-
ces their wires within a minimum threshold distance relative
to another pair, two calculations are performed. The first
bundles the neighboring wires such that they form adjacent
windings within a single groove. Alternatively, the wires are
spaced exactly by the minimum threshold value. If either
Fig. 8. Calculated profiles of Bz ð0; zÞ for the various coils listed in Table V. scenario improves field homogeneity, the changes are saved;
The vertical dashed lines indicate the optimization region of z ¼ 67:5 cm. otherwise, they are discarded.
Measurements of the optimized coil made with the Hall-effect sensor (circles,
top panel) confirm a field homogeneity within 1% over almost the full length
of the tubes used here. The Lee-Whiting and Helmholtz coils would require a a)
Electronic mail: [email protected]
large radius (middle panel) to achieve a comparable homogeneity. b)
Electronic mail: [email protected]
c)
Electronic mail: [email protected]; Present address: Max Rady
College of Medicine, University of Manitoba, 750 Bannatyne Avenue,
not pursued here. However, for an axisymmetric coil such as Winnipeg, Manitoba, Canada R3E 0W2.
d)
this, the constraint of Maxwell’s equations ensures that the Author to whom correspondence should be addressed. Electronic mail:
homogeneity of the field away from the central axis must simi- [email protected]
1
larly improve.40 D. J. Griffiths, Introduction to Electrodynamics, 4th ed. (Pearson, Boston,
2013).
Our design optimization was performed by considering 2
G. L. Pollock and D. R. Stump, Electromagnetism (Addison Wesley, San
the net axial field produced by the sum of contributions from Francisco, 2002).
the symmetric pairs of current loops1–3 comprising the coil 3
A. Garg, Classical Electromagnetism in a Nutshell (Princeton U. P.,
Princeton, 2012).
4
W. M. Saslow, “Maxwell’s theory of eddy currents in thin conducting
X
N=2
l0 R2 I=2 l0 R2 I=2 sheets, and applications to electromagnetic shielding and MAGLEV,” Am.
BðzÞ ¼ þ ;
i ðR2 þ ðz  zi Þ2 Þ 3=2
ðR2 þ ðz þ zi Þ2 Þ3=2 5
J. Phys. 60, 693–711 (1992).
M. A. Nurge et al., “Drag and lift forces between a rotating conductive
(B1) sphere and a cylindrical magnet,” Am. J. Phys. 86, 443–452 (2018).

498 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 498
6 24
M. H. Partovi and E. J. Morris, “Electrodynamics of a magnet moving C. P. Bean et al., “Eddy-current method for measuring the resistivity of
through a conducting pipe,” Can. J. Phys. 84, 253–271 (2006). metals,” J. Appl. Phys. 30, 1976–1980 (1959).
7 25
B. Irvine et al., “Magnet traveling through a conducting pipe: A variation M. A. Weinstein, “Magnetic decay in a hollow circular cylinder,” J. Appl.
on the analytical approach,” Am. J. Phys. 82, 273–279 (2014). Phys. 33, 762 (1962).
8 26
S. Fahy et al., “Electromagnetic screening by metals,” Am. J. Phys. 56, K. Lee and G. Bedrosian, “Diffusive electromagnetic penetration into
989–992 (1988). metallic-enclosures,” IEEE Trans. Antennas Propag. 27, 194–198 (1979).
9  27
J. I~niguez et al., “Measurement of the electrical conductivity of metallic M. J. Ramos et al., “The phase angle method for electrical resistivity
tubes by studying magnetic screening at low frequency,” Am. J. Phys. 73, applied to the hollow circular cylinder geometry,” J. Appl. Phys. 67,
206–210 (2005). 1167–1169 (1990).
10  28
J. I~niguez et al., “Measurement of electrical conductivity in nonferromag- C. P. Bidinosti and M. E. Hayden, “Selective passive shielding and the
netic tubes and rods at low frequencies,” Am. J. Phys. 77, 949–953 (2009). Faraday bracelet,” Appl. Phys. Lett. 93, 174102 (2008).
11  29
J. I~niguez et al., “The electromagnetic field in conductive slabs and cylin- M. L. Boas, Mathematical Methods in the Physical Sciences, 3rd ed. (John
ders submitted to a harmonic longitudinal magnetic field,” Am. J. Phys. Wiley & Sons, Hoboken, 2006), Chap. 8.
30
77, 1074–1081 (2009). K. F. Riley, M. P. Hobson, and S. J. Bence, Mathematical Methods for
12
C. P. Bidinosti et al., “The sphere in a uniform rf field—Revisited,” Physics and Engineering, 3rd ed. (Cambridge U. P., Cambridge, 2006),
Concepts Magn. Resonance 31B, 191–202 (2007). Chap. 13.
13 31
M. L. Honke and C. P. Bidinosti, “The metallic sphere in a uniform ac C. A. Desoer and E. S. Kuh, Basic Circuit Theory (Mcgraw-Hill, New
magnetic field: A simple and precise experiment for exploring eddy cur- York, 1969), Chap. 13.
32
rents and non-destructive testing,” Am. J. Phys. 86, 430–438 (2018). I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
14
J. R. Nagel, “Induced eddy currents in simple conductive geometries: Products, 7th ed. (Elsevier, Amsterdam, 2007), Chap. 17.
33
mathematical formalism describes the excitation of electrical eddy cur- K. Riess, “Some applications of the Laplace transform,” Am. J. Phys. 15,
rents in a time-varying magnetic field,” IEEE Antennas Propag. Mag. 45–48 (1947).
34
60(1), 81–88 (2018). See also “Correction,” IEEE Antennas Propag. Mag. C. L. Bohn and R. W. Flynn, “Real variable inversion of Laplace transforms:
60(4), 83 (2018). An application in plasma physics,” Am. J. Phys. 46, 1250–1254 (1978).
15 35
P. J. H. Tjossem and E. C. Brost, “Optimizing Thomson’s jumping ring,” See datasheet and application notes—in particular, Current Sensing with
Am. J. Phys. 79, 353–358 (2011). the CSA-1V—at the distributor website <https://gmw.com/product/csa-
16
C. L. Ladera and G. Donoso, “Unveiling the physics of the Thomson 1vg-so/>. The CSA-1V comes in a standard, surface mount SOIC-8 pack-
jumping ring,” Am. J. Phys. 83, 341–348 (2015). age, for which small breakout PCBs can be purchased from many vendors.
17 36
R. W. Latham and K. S. H. Lee, “Theory of inductive shielding,” Can. J. C. P. Bidinosti et al., “A simple wide-band gradiometer for operation in
Phys. 46, 1745–1752 (1968). very low background field,” Concepts Magn. Resonance 37B, 1–6 (2010).
18 37
J. R. Reitz, “Forces on moving magnets due to eddy currents,” J. Appl. J. R. Taylor, An Introduction to Error Analysis: The Study of
Phys. 41, 2067–2071 (1970). Uncertainties in Physical Measurements, 2nd ed. (University Science
19
H. A. Haus and J. R. Melcher, Electromagnetic Fields and Energy Books, Sausalito, 1996), Chaps. 3 and 4.
38
(Prentice-Hall, Englewood Cliffs, 1989), Chap. 10. See supplementary material at https://www.scitation.org/doi/suppl/
20
H. E. Knoepfel, Magnetic Fields: A Comprehensive Theoretical Treatise 10.1119/10.0003508 for a more extensive list of papers on the interaction
for Practical Use (John Wiley & Sons, New York, 2000), Chap. 4. of magnetic fields and conducting materials.
21 39
W. R. Smythe, Static and Dynamic Electricity, 2nd ed. (McGraw-Hill, J. L. Kirschvink, “Uniform magnetic fields and double-wrapped coil sys-
New York, 1950), Chap. XI. tems: Improved techniques for the design of bioelectromagnetic
22
L. V. King, “XXI. Electromagnetic shielding at radio frequencies,” experiments,” Bioelectromagnetics 13, 401–411 (1992).
40
London, Edinburgh, Dublin Philos. Mag. J. Sci. 15(97), 201–223 (1933). S. R. Muniz and V. S. Bagnato, “Analysis of off-axis solenoid fields using
23
J. C. Jaeger, “III. Magnetic screening by hollow circular cylinders,” the magnetic scalar potential: An application to a Zeeman-slower for cold
London, Edinburgh, Dublin Philos. Mag. J. Sci. 29(192), 18–31 (1940). atoms,” Am. J. Phys. 83, 513–517 (2015).

499 Am. J. Phys., Vol. 89, No. 5, May 2021 Krosney et al. 499