Fast sweep rate Faraday force magnetometer with simultaneous sample

temperature measurement
D. Slobinsky,1 R. A. Borzi,2 A. P. Mackenzie,3 and S. A. Grigera3
1)

SUPA, School of Physics and Astronomy, University of St Andrews, St Andrews KY16 9SS,
United Kingdom
2)
Instituto de Investigaciones Fisicoqumicas Te
oricas y Aplicadas (UNLP-CONICET),
c.c. 16, Suc. 4, and Departamento de Fsica, IFLP, UNLP, c.c. 67, 1900 La Plata,
Argentina.
3)
Instituto de Fsica de Lquidos y Sistemas Biol
ogicos, UNLP, La Plata 1900,
Argentina
(Dated: March 28, 2012)

Precise instantaneous sample temperature measurement were integrated into a simple and newly designed
Faraday force magnetometer based on the concept of a load-sensing variable capacitor. A plastic body
allows for fast sweep rates and precise temperature measurement. Moderate gradient fields of 1T/m yield
resolutions better than 1 105 emu. We report on the performance of this new design that can operate in
static and alternate mode down to below 100mK and up to at least 17T. Measurements on Dy2 Ti2 O7 and
Sr3 Ru2 O7 are presented as an example of its performance.
I.

INTRODUCTION

In experimental physics, magnetization, M, is usually

seen as the conjugate variable of the magnetic induction
field, B, in the free energy and is one of the most important variables that can be manipulated in the cryogenic
environment of a laboratory. M is a thermodynamic
variable of a system in equilibrium, that measures the
relative amount of up and down spins and the mean orbital angular momentum. This is the reason why, many
times, it can be used to quantify the behavior of systems
out-of-equilibrium, for instance, highly sensitive magnetization measurement are key to unravel the intricated
out-of-equilibrium dynamics of many magnetic insulating materials. (CITAS??)
There are a certain number of methods to measure
lab bulk magnetization. They can be separated into two
categories:
Inductively coupled methods, where the sample is inserted into a coil and magnetic induction is measured
by either modulating the magnetic field (Field Modulation) or moving relatively to each other the sample/coil
in an oscillatory fashion (Vibrating Sample Magnetometry/Vibrating Coil Magnetometry)1,2 . Although, both
these alternatives have been successfully applied over
large temperature regimes, they are less suitable for subKelvin temperatures where eddy current heating can
affect the measurement. These problems are usually
overcome at the expense of using dilution refrigerators
with large cooling power and/or lost of simplicity on the
probes. These measurements are inherently oscillatory
and there is a compromise with the sweep rate at which
the magnetic field is swept and the power that the mixing chamber can stand at a given temperature of the
setup. Another possibility is the use of extreme sensitive
SQUIDs for the read out. This, although suitable for
low temperatures, is relatively slow and very difficult to
implement for high magnetic fields.

The second category refers to Force methods, where the

magnetization is determined measuring the force a magnetic sample experiences in an inhomogeneous magnetic
field by means of a Faraday balance. The magnetization
is extracted using the equation
F = M B

(1)

where B is the external inhomogeneous field and M is

assumed to be homogeneous3 . The different alternatives
of the Faraday force method refer to the implementation
of the balance and of the gradient field.
In the high temperature regime, a balance with an arm
compensated at room temperature, and the other arm
in the cryogenic environment has been generally used.
On the other hand capacitive balances have been implemented for the sub-Kelvin regime4,5 . In this latter, the
whole experimental setup resides within the cryogenic environment, avoiding the heat leak implied by balancing at
room temperature. M is sensed by the change of the capacitance between two parallel plates, one fixed, and the
other movable. The force impelled by a sample attached
to the movable one, equation (1), is compensated by the
restoring force of a spring. This principle is based on the
assumption that at low temperatures the spring constant
generating the restoring force is temperature independient. The main drawback of any of the force methods is
the arguably mechanical stability of the movable platform against torque which affects the parallelism of the
capacitor plates.
Regarding the generation of the magnetic field gradient, several approaches have been implemented. Dedicated Helmholtz coils for gradient generation were reported in ref. 5 with linear gradients of up to 10T/m.
For the case where a second set of coils lacked, the stray
field of the main coil is used as gradient field by offcentering the sample. This method does not offer much
control given that the stray field increases linearly with
the applied field. Another variant was developed in ref.

2
6 where a film of Nickel was deposited a few millimeters below the magnetometer. This gives a gradient field
peaked at 5T/m that varies rapidly with distance, only
allowing for measurement of small samples.
One of the advantages of the Faraday method is that
the magnetization measurement can be integrated with
other measurements7 . This is because the balance can be
used as the platform for a non-magnetic implementation
of a second simultaneous experiment. With this concept
we have developed a simple Faraday force magnetometer
for the specific purpose of measuring the magnetization
and the instantaneous temperature. Below, an example
of the relevance of these simultaneous measurements is
illustrated in the particular case of a frustrated insulating
magnet.

II.

DESIGN FEATURES

The magnetometer was designed to fit into the 35mm

vacuum can of a dilution refrigerator with a cooling power
of 25W at 100mK (Kelvinox25, Oxford Instruments).
The main a magnet (Oxford Instruments) generates a
magnetic field up to 17T, while a couple of superconducting modulation coils in gradient configurations can
produce < 1T/m. A sketch of the magnetic system and
the capacitive magnetometer is shown in figure 1.
Because of the small cooling power produced by our
refrigerator and aiming for the possibility of fast sweep
rates it was essential to avoid eddy current heating. In
order to achieve this, and also with the purpose of thermally isolate the sample environment for precise sample
temperature measurements, we opted for a plastic magnetometer made on Tufset (Tufnol Composites Limited).
For maximum simplicity, it consists of two assemblies,
each of them containing one plate of the capacitor to be

sensed. The first one (pieces a to d in figure 2), which

hosts the fixed capacitance plate, includes the rod leading from the mixing chamber of the dilution refrigerator
to the center of the magnet. The second one includes the
magnetometer assembly containing two pieces: a body
(piece f) that screws into the first assembly, and a sample holder (which includes the movable capacitor plate)
suspended by wires (piece e). For this small design we inverted the geometry from the one of ref. 5 by placing the
movable capacitor plate at the top of the sample holder
as depicted in figure 2. In this way, the magnetometer
assembly screws into the probe rod allowing control over
the initial capacitance.
The magnetometer body consists of a 20mm outer and
18mm inner diameter cylinder machined out of a single piece of Tufset, with four windows for easy access to
the sample holder, and eight cone-shaped holes on the
columns between the windows, for the wire suspending
the sample holder.
Athough not a critical constraint in our case, the small
outer dimension avoids the use of a centering mechanism,
but reduces the theoretical sensitivity with respect to ref.
5. This, for the simple model of the deflection of a beam,
is proportional to L3 , with L the length of the wires from
the inner diameter to the sample holder at the center8 .
In order to compensate for this we decided to change the
wire material, eventually settling into 0.19mm ...(fishing line) with the further advantage of reducing metallic
components.
One ...(fishing line) wire threading the eight holes of
the magnetometer body and sample holder is used to
hang the latter from the former. The holes are subsequentially filled with epoxy (#1266, Stycast).
The sample holder is also machined out of a single
piece of Tufset. We painted the capacitor plate on it

z
capacitor

b
c

d
e

main coil

gradient coils
f

sample
Figure 1: Schematic view of the magnetic system and load
cell. The capacitively sensed load cell is in the homogeneous
magnetic field of a 17T main coil and in the center of a set of
Helmholtz coils in gradient configuration.

Figure 2: (Right) Plastic magnetometer. (Left) The two

assemblies that form the magnetometer: the rod assembly
includes a) stainless steel rod from the mixing chamber to
the center of the magnet, b) oxygen free copper adapter, c)
threaded Tufset piece, d) top plate of the capacitor made on
brass. The magnetometer assembly is entirely made of Tufset
and includes e) sample holder + movable capacitive plate and
f) body

3
Mixing Chamber
R2
Sample
R1
RuO2

Figure 3: (Left) Sample holder with sapphire plate and sample. (Center) Detail of the sapphire plate. Gold wires at the
back of it are used to thermalize the sample that is greased
to the front of the plate. The sample is connected by another
gold wire to a RuO2 thermometer hanging from the magnetometer body by its contacts. (Right) Thermal resistance
diagram, R1 is much lower than R2 .

(figure 3) with room temperature air-cured silver paste

(4929N, Dupont) which we connected to a connector on
the magnetometer body by a 25m gold wire. The sample space on this piece is of 4mm by 3mm in the plane
containing z, and 8mm depth (although is designed to
accommodate mainly very thin samples). This space is
divided into two by a sapphire plate of 0.2mm that slides
into a perpenpendicular slot and is fixed with vacuum
grease (Apiezon N), leaving effectively a sample space of
4mm in depth (left of figure 3). The sapphire acts as an
orientation plane for the sample that is greased to the
front of it with Apiezon grease, as shown in the center of
figure 3.
Two 25m gold wires of negligible spring constant,
coming from a connector on the magnetometer body are
bonded to the the back of the orientation plate with silver paste (#6838, Dupont). Finally, a thick 0.5mm silver
wire screwed into the mixing chamber, thermally grounds
the connector of the body and then the sapphire and sample. There are connectors for all wires coming from the
rod to the magnetometer body, namely, the thermometry manganin wires, the miniature coaxial cable (type
SS, Lakeshore) to the lower plate and the thermal link to
the mixing chamber. In this manner, the body assembly
can be screwed into the rod assembly to tune the initial
capacitance at room temperature to a convenient value
without twisting the wires. The sapphire plate is the
main thermal link between the sample and the mixing
chamber.
In order to measure the temperature of the sample,
we attached with silver paste a 25m gold wire from
the sample to the back of a commercial 1k RuO2 chip
(crg0805 smt, Tyco Electronics) used as a thermometer
which hangs by its leads from the magnetometer body.
In this way we obtained a much lower thermal resistance
between the sample and the thermometer, R1 , than between the sample and the mixing chamber R2 .

III.

PERFORMANCE

A. Magnetization and temperature measurements of

Dy2 Ti2 O7

We tested the performance of our magnetometer assembly on the frustrated magnet Dy2 Ti2 O7 . This cooperative paramagnet is of the spin-ice family9 and undergoes a yet not completely understood freezing transition at about 650mK, where the system develops out-ofequilibrium dynamics for any experimentally achievable
time. A detailed study of the phenomena can be found
in ref. 10.
A single crystal of Dy2 Ti2 O7 was grown by floating
zone method in St Andrews University, Laue oriented
and cut into a prism of ...mg and 2 0.7 0.5 mm3
with the long axis along [111]. The sample was mounted
on the sapphire plate as explained in figure 3. The magnetic field and temperature are controlled from a PC
which in turns records the capacitance measured with
a capacitance bridge operating at 1kHz (2500A, AndeenHagerling). This bridge and a resistance bridge (SIM
921AC, Stanford Research Systems) which measures the
RuO2 thermometer on the sample are set to give a data
point per second. External vibrations are damped placing the cryostat dewar on the platform of an active vibration isolation system (vario 45/60/90, Halcyonics).
The main magnet field is swept at constant speed while
the gradient coils are fixed with a constant current at
< 1T/m. The capacitance for no applied gradient in this
experiment was around 7.2pF. Figure 4a) shows the raw
data for the low field part of the magnetization curve
at base temperature ( 80mK) for different sweep rates.
The resolution on these curves is better than 1105 emu.
The out-of-equilibrium behavior of this cooperative paramagnet can be inferred from the dependence of the magnetization on the sweep rate. For slow sweep rates M
follows an increasingly undulating curve before becoming
discontinuous with a very sharp jump of about 50% of the
saturation value (that is 5B /Dy for this orientation11 )
for the fastest sweep rate. Figure 4b) shows the temperature reading from the calibrated RuO2 thermometer
connected to the sample for the same set of fields that
4a). The spikes on this figure are a spurious effect given
by tiny spikes of the magnetic field on the main coil that
are picked up by the thermometer (due to eddy current
heating) and not by the magnetometer. Finally, figure
4c) shows the derivative of the raw data with respect to
time, i.e., dM/dt = dM/dH sweep rate, multiplied by
the magnetic field. This magnitud is proportional to the
heat produced in the process (see equation (2) below).
Note the similitude between 4b) and 4c) this is due to
the sample dynamics and is only seen due to the experimental arrangement. The mixing chamber is only weakly
thermally coupled to the magnetometer head, but very
well connected to the sample, giving a very efficient path
for the heat generated by the sample to diffuse towards
it.

4
3.5

0.5

0.003T/min
0.0125T/min
0.02T/min
0.025T/min

1.5

dM/dt x H (a.u.) Temperature(mK)

0.5

7e-03

0.3
0.2

5e-03

0.1

3e-03

0
0.6
0.5
0.4
0.3
0.2
0.1
0

1e-03
5e-02
4e-02
3e-02
2e-02
1e-02
0e+00
0

10

12

Capacitance(pF)

M(B/Ru)

2.5

300

9e-03

0.4

Torque(a.u.)

Magnetization(B/Dy)

a)

14

0H(T)

b)

200

Figure 5: (Top) Antisymmetrize part of the raw signal as

regard to magnetic gradient. The magnetization is extracted
from the difference of the data with positive and with negative gradient. (Bottom) Symmetric part giving the torque.
The right axis is proportional to the raw signal. Note that
torque signal is almost an order of magnitud bigger that magnetization for this starting capacitance of 26pF.

100
c)
200
100
0
0.15

0.2

0.25

0.3 0.35
0H(T)

0.4

0.45

B.

Figure 4: a) Magnetization vs. magnetic field for different

sweep rates at base temperature ( 80mK). b) Temperature
vs. magnetic field as measured from the RuO2 thermometer
on the sample for the same parameters of a). c) Derivative of
the magnetization respect to time multiply by the magnetic
field (i.e., rate of heat generation) as a function of magnetic
field. The similitude between b) and c) illustrates the integration of magnetization and sample temperature measurement.

The diffusion of heat from the sample to the mixing

chamber can be modelled by the continuity equation:
d
dt (E) + J = 0, with the energy E = CT M H,
where C is the heat capacity of the sample (given the
heat capacity and thermal conductivity of the plastic at
these temperatures the contribution of the sample holder
can be neglected), T the temperature, and J the flux of
heat. This equation can be transformed into
C

dT
dM
= G(T Tmch ) +
H
dt
dt

(2)

where G is a heat conduction constant and Tmch is the

temperature of the mixing chamber. The specific heat
of Dy2 Ti2 O7 is very small in the region of parameters
of figure 412 , and the thermalization very good, which
means that the left hand side of (2) can be neglected with
respect to the right hand side giving the proportionality
measured in 4c): T = Tmch + G1 dM
dt H.

Magnetization measurement of Sr3 Ru2 O7

The ruthenate Sr3 Ru2 O7 was used to test the performance of the probe at high fields and to test the method
of gradient field modulation to measure susceptibility.
At low temperatures, Sr3 Ru2 O7 is a paramagnetic Fermi
liquid that undergoes a set of meta-magnetic first order
transitions as a function of magnetic field. These transitions occur below 1.2K and at 8T depending on
the angle the c-axis of the sample makes with the magnetic field. This material provides a valuable test for the
mechanical stability of the magnetometer given that for
slight misalignments from the c-axis, the magnetization
as a function of field lacks a feature around 12T that is
present in torque13 .
We used the same experimental setup than in the case
Dy2 Ti2 O7 and measured a single crystal of ...mg and dimensions ... . With an initial capacitance of 26pF, and
the sample between 5 to 10 degrees off the c-axis (the
alignment of the sample is inferred from the known dependence of Hc with angle14,15 ); a very large quadratic
signal (even parity as a function of magnetic field) was
observed. This behavior was identified with a torque
component coupling in given that the material was known
to behave as a paramagnet (odd parity). The magnetization was extracted by subtracting the data for a field
sweep with a positive gradient field from that with a negative or null gradient field (divided by two). The torque
contribution was extracting from the sum of the signals
(divided by two). The torque and magnetization extraction are shown in figure 5 where the absence of the torque
feature at 12T is in agreement with the experiments of

5
In summary, we presented a newly designed plastic
magnetometer with resolution better than 105 emu. The
magnetometer was successfully integrated with a sample
thermometer that measures the instantaneous change in
temperature of the sample with high accuracy down to
temperatures below 100mK. We demonstrate the possibility of extracting the torque signal from the raw signal
and confirm that this signal is bigger the higher the initial equilibrium capacitance. Furthermore, we used the
same probe, to measure the magnetic susceptibility by
modulating the gradient field for low frequencies.

0.07
0.07

Susceptibility(a.u.)

0.06

0.06
0.05

0.05

0.04
0.03

0.04
0.03
0.02
0.01
0

0.02
0.01
0
7.6
50mK
150mK
300mK
500mK
700mK

6.6

6.8

7.65

7.7

7.75

7.2
7.4
0H(T)

7.8

7.6

7.8

Figure 6: Gradient modulation susceptibility at 3Hz for different temperatures around the meta-magnetic transtions.
Three main peaks can be seen with a noise comparable to
the one on ref. 15. Inset: blow up of the main peak

ref. 13. In another experiment with an initial capacitance

of about 2.6pF the raw data still showed a quadratic
shape but with a much less pronounced behavior16 . In
any case, the subtraction and addition of the signals with
a positive and negative gradient fields have proven to give
the right magnetization and torque in all cases. However,
the substraction and addition reduce the sensitivity up
to an order of magnitud.
Using the same sample of Sr3 Ru2 O7 we tested the idea
suggested in ref. 5 of alternating the gradient field to
measure susceptibility. For this, we replaced the automatic capacitance bridge for a manual capacitance bridge
(1616, General Radio) to which we connected two lock-in
amplifiers in series (SR830, Stanford Research Systems),
the first operating at 1kHz and the second driving the
gradient coils through a hi-fi amplifier at the excitation
frequency. Figure 6 shows the real part of the susceptibility measured at 3Hz around the meta-magnetic transitions. This data has a noise comparable to the previously
reported field modulated susceptibility data15 , where low
temperature transformers had been used to boost the signal. The main impediment to increase the frequency was
the increase in temperature due to eddy current heating.
We believe that the frequency range can be increased
by keeping the excitation power constant, decreasing the
current amplitude while increasing the frequency, or by
making a feedback in order to measure keeping the capacitor plates static which, in principle, could overcome
the problem of a coupling into the mechanical resonance
mode of the setup.

ACKNOWLEDGMENTS

We wish to thank to E. Yelland, J. Bruin and A. Rost

for helpful discussions and for their help in the lab, and to
R. Perry and A. Gibbs for providing the crystals for the
experiments. This work was done with the financial support of EPRSC and Royal Society (UK) and CONICET
(Argentina).

REFERENCES
1 S.

Foner, Rev. Sci. Instrum. 30, 548 (1959)

Legl, C. Pfleiderer and K. Kr
amer, Rev. Sci. Instrum. 81,
043911 (2010).
3 Strictly, F = (M B)
4 A. G. Swanson, Y. P. Ma, J. S. Brooks, R. M. Markiewicz, and
N. Miura, Rev. Sci. Instrum. 61, 848 (1990).
5 T. Sakakibara, H. Mitamura, T. Tayama and H. Amitsuka, Jpn.
J. Appl. Phys. 33, 5067 (1994).
6 T. Sakona and M. Motokawa, Rev. Sci. Instrum. 71, 3474 (2000).
7 J. S. Brooks, M. J. Naughton, Y. P. Ma, P. M. Chaikin and R.
V. Chamberlin, Rev. Sci. Instrum. 58, 117 (1987).
8 See for example L. D. Landau and E. M. Lifshitz, Theory of
Elasticity (Course of Theoretical Physics), Pergamon Pr, 2nd
edition (June 1981)
9 For a review on spin-ice see S.T. Bramwell and M.J.P. Gingras,
Science 294, 1495 (2001).
10 D. Slobinsky, C. Castelnovo, R. A. Borzi, A. S. Gibbs, A. P.
Mackenzie, R. Moessner and S. A. Grigera, arXiv:1010.4143v1.
11 T. Sakakibara, T. Tayama, Z. Hiroi, K. Matsuhira and S. Takagi,
Phys. Rev. Lett. 90, 207205 (2003).
12 R. Higashinaka, H. Fukazawa, K. Deguchi and Y. Maeno, J. Phys.
Soc. Jpn. 73, 2845 (2004).
13 E. Ohmichi, Y. Yoshida, S. I. Ikeda, N. V. Mushunikov, T. Goto,
and T. Osada, Phys. Rev. B 67 024432 (2003).
14 S. A. Grigera, R. A. Borzi, A.P. Mackenzie, S. R. Julian, R. S.
Perry and Y. Maeno, Phys. Rev. B 67 214427 (2003).
15 S. A. Grigera, P. Gegenwart, R. A. Borzi, F. Weickert, A. J.
Schofield, R. S. Perry, T. Tayama, T. Sakakibara, Y. Maeno, A.
G. Green and A. P. Mackenzie, Science 306, 1154 (2004).
16 It should be note that higher initial capacitances amplified torque
signals quadratically with inverse initial equilibrium distance.
2 S.