In situ magnetic compensation for potassium spin-exchange relaxation-free

magnetometer considering probe beam pumping effect

Jiancheng Fang, Tao Wang, Wei Quan, Heng Yuan, Hong Zhang, Yang Li, and Sheng Zou
Citation: Review of Scientific Instruments 85, 063108 (2014); doi: 10.1063/1.4881685
View online: http://dx.doi.org/10.1063/1.4881685
View Table of Contents: http://scitation.aip.org/content/aip/journal/rsi/85/6?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Vectorial Kerr magnetometer for simultaneous and quantitative measurements of the in-plane magnetization
components
Rev. Sci. Instrum. 85, 053904 (2014); 10.1063/1.4871098
An optically modulated zero-field atomic magnetometer with suppressed spin-exchange broadening
Rev. Sci. Instrum. 85, 045124 (2014); 10.1063/1.4872075
In situ triaxial magnetic field compensation for the spin-exchange-relaxation-free atomic magnetometer
Rev. Sci. Instrum. 83, 103104 (2012); 10.1063/1.4756046
An all-optical scalar and vector spin-exchange relaxation-free magnetometer employing onoff pump modulation
J. Appl. Phys. 109, 07E507 (2011); 10.1063/1.3536673
Pumpprobe Faraday rotation magnetometer using two diode lasers
Rev. Sci. Instrum. 76, 056105 (2005); 10.1063/1.1912688

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47

REVIEW OF SCIENTIFIC INSTRUMENTS 85, 063108 (2014)

In situ magnetic compensation for potassium spin-exchange relaxation-free

magnetometer considering probe beam pumping effect
Jiancheng Fang,1 Tao Wang,1,a) Wei Quan,1 Heng Yuan,1 Hong Zhang,2 Yang Li,1
and Sheng Zou2
1
2

School of Instrument Science and Opto-Electronics Engineering, Beihang University, Beijing 100191, China
School of Instrument Science and Engineering, Southeast University, Nanjing 210096, China

(Received 6 March 2014; accepted 23 May 2014; published online 9 June 2014)
A novel method to compensate the residual magnetic field for an atomic magnetometer consisting
of two perpendicular beams of polarizations was demonstrated in this paper. The method can realize
magnetic compensation in the case where the pumping rate of the probe beam cannot be ignored.
In the experiment, the probe beam is always linearly polarized, whereas, the probe beam contains
a residual circular component due to the imperfection of the polarizer, which leads to the pumping
effect of the probe beam. A simulation of the probe beams optical rotation and pumping rate was
demonstrated. At the optimized points, the wavelength of the probe beam was optimized to achieve
the largest optical rotation. Although, there is a small circular component in the linearly polarized
probe beam, the pumping rate of the probe beam was non-negligible at the optimized wavelength
which if ignored would lead to inaccuracies in the magnetic field compensation. Therefore, the dynamic equation of spin evolution was solved by considering the pumping effect of the probe beam.
Based on the quasi-static solution, a novel magnetic compensation method was proposed, which contains two main steps: (1) the non-pumping compensation and (2) the sequence compensation with a
very specific sequence. After these two main steps, a three-axis in situ magnetic compensation was
achieved. The compensation method was suitable to design closed-loop spin-exchange relaxationfree magnetometer. By a combination of the magnetic compensation and the optimization, the magnetic field sensitivity was approximately 4 fT/Hz1/2 , which was mainly dominated by the noise of the
magnetic shield. 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4881685]
I. INTRODUCTION

The atomic magnetometer is an ultra-high sensitive magnetic field measuring instrument, and is widely used in many
fields, such as biomagnetism, paleomagnetism, and fundamental physics.13 The atomic magnetometer based on the
spin-exchange relaxation-free (SERF) theory has realized the
highest sensitivity with centimeter-sized volume,4 surpassing the sensitivity of superconducting quantum interference
devices (SQUID).5, 6 The SERF magnetometer benefits from
the total suppression of the spin exchange relaxation, whose
fundamental sensitivity is limited by the spin projection
noise. The fundamental sensitivity can reach atto-Tesla with a
several cubic centimeters volume.7
The magnetic signal of the SERF magnetometer is acquired by optical polarimetry techniques.8 Therefore, it is important to optimize the detuning of the probe beam to achieve
the largest optical rotation. First, a simulation of the optical
rotation and the pumping rate of the probe beam were demonstrated in this paper, based on which the probe beam was optimized. Although the probe beam is linearly polarized, due
to the imperfection of the polarizers, the probe beam contains a residual circular component of the order of approximately 103 104 , which leads to the pumping effect of the
probe beam in addition to the pump beam.9, 10 Although the
pumping effect of the probe beam is smaller than the pump
a) Electronic mail: wangtaowt@aspe.buaa.edu.cn

0034-6748/2014/85(6)/063108/7/$30.00

beam, the power of the pump beam is always adjusted to a

small value to achieve best sensitivity. The pumping rate of
the probe beam was non-negligible compared with the pump
beam, which can cause inaccuracies in the precise magnetic
field compensation in the case where the pumping effect of
the probe beam is ignored.
Furthermore, to keep the atoms in the SERF regime
and suppress the spin-exchange relaxation, the condition that
the spin exchange rate far exceeds the Lamor procession
frequency should be satisfied.11 As the spin-exchange rate
is approximately quadratic in the magnetic field,12 a highpermeability magnetic shield and active compensation were
used to shield the ambient magnetic field.
Some in situ active compensation methods have previously been proposed in the literature for the SERF magnetometer.1315 Seltzer and Romalis proposed cross-modulation
to compensate the residual magnetic fields, which can realize compensating magnetic fields in three directions simultaneously.13 However, the compensation of the field in
the direction which was perpendicular to the pump beam
and the probe beam was not accurate, as it was extracted
from the dc-term which can drift with time. Furthermore,
the pumping effect of the probe beam was not considered,
and the compensation accuracy is reduced when the probe
beam frequency is near the resonance. This compensation
method is not suitable for use at the maximum optical rotation
point, because the optimized wavelength of the probe beam to
achieve the largest optical rotation is near the resonance. The

85, 063108-1

2014 AIP Publishing LLC

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47

063108-2

Fang et al.

cross-modulation process is repeated iteratively until the total field amplitude becomes as small as possible, which leads
to the cross-modulation not being suitable for designing a
high-speed closed-loop SERF magnetometer, and increasing
complexity of the compensation algorithm. Li et al. proposed parametric modulation, which can preserve the bandwidth of SERF magnetometers, and reduce the airflow noise.
However, resulting from the approximation in calculating
the modulation, the modulation reduces the magnetometers
sensitivity.15
In this paper, we propose a novel in situ magnetic compensation method to compensate the residual magnetic field
by considering the pumping effect of the probe beam, which
allows the active magnetic compensation near the resonance
of the potassium D2 line. This method allows the wavelength
of the probe beam to be optimized to achieve the largest optical rotation, which can greatly suppress the dependence on
the probe sensitivity. Additionally, considering the pumping
effect of the probe beam is important when the total relaxation rate approaches a few s1 , because the pumping rate of
the probe beam is non-negligible compared to the total relaxation rate. The compensation method includes two parts:
the non-pumping compensation was first employed to compensate magnetic field in three-axis. This step could easily
compensate a magnetic field in the direction which is perpendicular to the pump and probe beam, however, it is not operating in the SERF regime in the case where the pump beam
was blocked. Afterwards the pump beam was unblocked,
which caused light shift.16 In the second main step, sequence
compensation was employed in a certain order to precisely
compensate the magnetic field in the directions along the
pump beam and the probe beam. With the correct compensation sequence, it was not necessary to repeat the sequence
compensation process, which made the closed-loop magnetic
compensation easier to realize.

II. PRINCIPLE

The spin evolution of the atomic magnetometer can be

well described by the Bloch equation.10 In this paper, the
probe beam propagated along the x-direction, the pump beam
propagated along the z-direction, and the remaining axis
was the y-direction, as shown in Figure 3. By considering
the pumping effect of the probe beam, the spin polarization
evolves as
d 
1
P = [e B P + Rp (z P ) + Rm (x P ) RSD P ],
dt
q
(1)
where P is the electron spin polarization, B is the magnetic field, e is the electron gyromagnetic ratio, Rp is the
pumping rate of pump beam, Rm is the pumping rate caused
by the small circular component in the linearly polarized
probe beam, RSD is the spin-destruction rate, z is the unit
vector of the z-direction, and x is the unit vector of the
x-direction.
The atoms were pumped along the z-direction, and were
detected along the x-direction by the probe beam. In the case
of a slowly varying magnetic field, the steady-state solution

Rev. Sci. Instrum. 85, 063108 (2014)

can be considered as


Rm 2 + Bx2 + Rp (By + Bx Bz )


,
Px =
Rtot 2 + Bx2 + By2 + Bz2
Py =

Rm (Bx By + Bz ) + Rp (By Bz Bx )


,
Rtot 2 + Bx2 + By2 + Bz2



Rm (Bx Bz By ) + Rp 2 + Bz2


Pz =
,
Rtot 2 + Bx2 + By2 + Bz2

(2)

(3)

(4)

where the Rtot = Rp + RSD + Rm is the total relaxation rate,

and
Rtot
.
(5)
=
e
In the experiment, a 4-layer of mu-metal magnetic shield was
used to keep the residual magnetic field smaller than 10 nT.
When the pump beam was blocked, Eq. (2) becomes


Rm 2 + Bx2

.
Px =
(6)
Rtot 2 + Bx2 + By2 + Bz2
The pumping effect of probe beam was considered in this
paper. Usually, the pumping effect of the probe beam is ignored to simplify the quasi-static result of the Bloch equation. In this scenario, from Eq. (6) it is assumed that blocking
the pumping beam will not result in any output signal, but in
our experiments, the magnetometer was found to have a signal, which was symmetric with respect to zero magnetic field.
Based on this signal, a novel compensation method without a
pump beam was proposed. When the pump beam is blocked,
we compensate the residual magnetic field in the shield, a process that we call non-pumping compensation in this paper,
whose procedure is shown in Figure 2, the details of the steps
are described as follows:
1. The magnetic field in the z-direction is increased from
10 nT to 10 nT. Adjusting the offset of the z-direction
magnetic field in order to reach the maximum value of
the magnetometers output, the magnetic field in the zdirection is compensated;
2. In a similar fashion, the magnetic field in the y-direction
is increased from 10 nT to 10 nT. By adjusting the offset of the magnetic field in the y-direction in order to obtain the maximum value of the magnetometers output,
the magnetic field in the y-direction is compensated;
3. A bias magnetic field of 10 nT is added in the y-direction
and the magnetic field in the x-direction is increased
from 10 nT to 10 nT. By adjusting the offset of the
magnetic field in the x-direction to obtain the minimum
value of the magnetometers output, the magnetic field
in the x-direction is compensated. After the x-direction
compensation, the bias magnetic field in the y-direction
was removed which completes the non-pumping compensation procedure.
After the non-pumping compensation, the magnetic field
is small enough so that Bx 2 + By 2 + Bz 2 < 2 . However, this
compensation is not in itself sufficient. When the pump beam
illuminates the cell (previously the pump beam was blocked)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47

063108-3

Fang et al.

Rev. Sci. Instrum. 85, 063108 (2014)

the pump beam causes a light shift that can be treated as a

fictitious magnetic field.16, 17 Furthermore, the non-pumping
compensation is a rough compensation. A further compensation of the z- and x-direction residual magnetic shield was
necessary. A novel compensation process that we call sequence compensation is implemented in the second step. To
date in the literature the cross-modulation is the preferred
magnetic field compensation method, which does not consider the probe beam pumping effect.13 To make a comparison
with this method, an oscillating magnetic field modulation in
the x-direction and z-direction were applied as in the crossmodulation,
Bmx = Ax sin(x t),

(7)

Bmz = Az sin(z t).

(8)

Inserting the modulation into Eq. (2), we find

Px =


1   2
2
Rm +Bx0
+2Ax Bx0 sin(x t)+A2x sin2 (x t)
Rtot 2
+Rp [By + Bx0 Bz0
+Ax Bz0 sin(x t) + Az Bx0 sin(z t)

+Ax Az sin(x t)sin(z t)] ,

(9)

(10)
Demodulating the first harmonic x is unable to compensate the residual magnetic field in the x-direction Bx0 , due to
the first harmonic x being proportional to both Bx0 and Bz0 .
This problem can be solved by modifying the compensation
sequence. First of all, a magnetic modulation Az sin(z t) was
applied in the z-direction
1
Rtot

  2

2
Rm + Bx0


+Rp [By + Bx0 Bz0 + Az Bx0 sin(z t)] .

The Bz0 can be eliminated by demodulating the first harmonic x , the offset of the magnetic field in the z-direction
was adjusted to decrease the first harmonic term, resulting in
magnetic field compensation in the z-direction, i.e., Bz0 0.
Furthermore, based on the modulation, the light shift of the
pump beam can be easily calculated by simply subtracting
the non-pumping z-direction compensated value from Bz0 .
The change of the residual magnetic field in the z-direction
was mainly caused by the light shift of the unblocked pumping beam.
III. EXPERIMENTS AND RESULTS

where Bx0 is the static residual magnetic field in the xdirection, and Bz0 is the static residual magnetic field in the
z-direction. If we set the Rm = 0 in Eq. (9), the equation is
the same with the cross-modulation, while lock-in amplifiers
referenced to x and z provide signals that are, respectively,
proportional to Bz0 and Bx0 .13 However, with the consideration of the probe beam pumping effect (Rm is not equal to
0), it is difficult to demodulate the Bx0 and the Bz0 simultaneously, and the cross-modulation will result in inaccuracies.
In this paper, the second step of the magnetic field compensation required a specific sequence. If a magnetic modulation
Ax sin(x t) was first applied in the x-direction


1
A2x
2
2

+ Rp By
Px =
+
B
+
R
m
x0
Rtot 2
2

Rm A2x
cos(2x t) .
+Ax (2Rm Bx0 +Rp Bz0 )sin(x t)
2

Px =

decrease the first harmonic term, resulting in magnetic field

compensation in the x-direction, i.e., Bx0 0. Then, a magnetic modulation of Ax sin(x t) was applied in the x-direction,
we find


A2x
1
A2x
2
Rm +

cos(2x t)
Px =
Rtot 2
2
2

(12)
+Rp [By + Ax Bz0 sin(x t)] .

(11)

The lock-in amplifier referenced to z provided a signal

which is only proportional to Bx0 . Therefore, the Bx0 can be
eliminated by demodulating the first harmonic z , the offset of the magnetic field in the x-direction was adjusted to

A. Optical rotation and probe beam pumping

rate simulation

A simulation of the optical rotation and the probe beam

pumping rate was implemented in order to optimize the
wavelength of the probe beam which leads to the maximum
optical rotation. A nominally linearly polarized beam which
contains a small circular polarized component can indeed lead
to a contribution to the pumping rate, which cannot be neglected. The magnetometer signal was obtained by optical
polarimetry techniques, and the optical rotation of the probe
beam is proportional to the x-direction polarization,8

= re [K]lcPx (fD1 Im[V ( D1 )]

2
1
+ fD2 Im[V ( D2 )]).
(13)
2
where was the optical rotation, V is the Voigt profile, Px is
the electron polarization in the x-direction, re is the classical
electron radius, [K] is the density of potassium, l is the length
of the probe beam pass through the cell, c is the light speed,
fD1 and fD2 are the oscillator strength of the D1 and D2 transitions, D1 and D2 are the resonance frequency of the D1 and
D2 transitions.
Inserting Eq. (2) into Eq. (13), and making Bx , Bz = 0,
which is reasonable when the residual magnetic field is well
compensated, we obtained

Rp By


 fD1 Im[V ( D1 )]
= re [K]lc
2
Rtot 2 + By2

1
+ fD2 Im[V ( D2 )] + off set.
(14)
2
The offset term represents the optical rotation due to the relaxation caused by the pumping effect of the probe beam, which
arises from the fact that the probe beam pumps the atoms in
the x-direction, and the spin polarization is not aligned perpendicular to the x-direction. As the By field is always an AC

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47

063108-4

Fang et al.

Rev. Sci. Instrum. 85, 063108 (2014)

signal, the offset was ignored in the simulation. Furthermore,

the optical depth of the probe beam should be considered,9
ODm = [K]lre c Re[V ( D2 )].

(15)

The final optical rotation expression can be given by

Rp By



f = exp(ODm ) re [K]lc
2
Rtot 2 + By2

1
fD1 Im[V ( D1 )] + fD2 Im[V ( D2 )] .
2
(16)
The pumping rate of the probe beam can be given by10
Rm = sm re cfD2 m Re[V ( D2 )],

(17)

where sm is the proportion of circular polarized light comprised in the probe beam, and m is the photon flux of the
probe beam.
The sm was set to 103 in the simulation, which was
matched in our observed experimental data. From Eqs. (16)
and (17), a simulation of the optical rotation and the pumping rate of the probe beam can be plotted as a function of the
wavelength of the probe beam and is shown in Figure 1.
The two optimization points near the D2 line of potassium are the points which result in the largest value of optical rotation, as shown in Figure 1. One is at a wavelength of
766.54 nm and the other is at 766.87 nm (chosen in our experiment). The power of the probe beam was approximately
4 mW, and the pumping rates were 2.4 s1 and 2.2 s1 respectfully at the optimization wavelengths, which were comparable to the spin destruction rate of approximately 10 s1
(well-made paraffin-coated cells). In our experiment, the spin
destruction rate of our cell was approximately 35 s1 . According to Eq. (10), if we use the cross-modulation scheme
(proposed by Seltzer et al.),13 the residual magnetic field in
the z-direction after compensation is equal to 2Rm Bx0 /Rp ,
which is proportional to the residual magnetic field in the xdirection, instead of finding exactly Bz0 = 0 point. The sensitivity of the magnetometer is optimized when Rp = Rm + RSD .
Then the residual magnetic field in the z-direction is approximately Bx0 /8.5 after compensation. If the residual magnetic
field in the x-direction is approximately 8.5 nT, then the resid-

FIG. 1. Calculations of optical rotation (solid line) and pumping rate of

probe beam (dashed line). Two vertical dashed-dotted lines indicate the two
optimized points.

ual magnetic field in the z-direction is approximately 1 nT,

which is a significant error if one ignores the probe beam
pumping effect. However, if the novel compensation method
proposed in this paper is applied, the residual magnetic field
can be accurately compensated, just as the experimental results show in Figure 5. The pumping rate of the probe beam
increased in proportion to the power of the probe beam and it
increases dramatically with the decease of the detuning of the
probe beam. The part of the curve where the optical rotation
appears constant near the D2 line of potassium is due to the
potassium atoms strongly absorbing the probe beam near the
resonance wavelength.
In the simulation of the non-pumping compensation
shown in Figure 2, the magnetic field in the z-direction is
increased from 10 nT to 10 nT. When the magnetometer output signal curve had a maximum, the residual magnetic field in the z-direction becomes zero. Subsequently, a
magnetic field from 10 nT to 10 nT was applied in the ydirection, which allowed the y-direction compensating point
to be found, which occurred when the magnetometer output
signal curve had a maximum. Finally, a 10 nT offset was
added in the y-direction to make the signal output vary as
the magnetic field is swept from 10 nT to 10 nT in the xdirection. When the magnetometer output signal curve had a
minimum, the residual magnetic field in the x-direction becomes zero.
B. Experimental setup

The experimental apparatus can be seen in Figure 3.

A 4-layer mu-metal magnetic shield was used to shield the
residual magnetic field, and the radius of the innermost layer
was approximately 0.2 m. Part of the magnetic shield was
composed of aluminum, which increased the Johnson noise.
Three-axis compensation coils placed into the magnetic shield
act to modulate and compensate the magnetic field. The shape

FIG. 2. Non-pumping compensation sequence, the signal output is produced

by simulation. The y- and z-direction residual magnetic fields are compensated when the response signals have maximums. While, an offset in the yor z-direction is necessary when compensating the residual magnetic field in
the x-direction, which was compensated when the response has a minimum.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47

063108-5

Fang et al.

FIG. 3. The layout of the SERF magnetometer experimental setup. The cell
was illuminated by pump beam and probe beam, the pump beam polarized
the atoms along the z-direction, and the linearly polarized probe beam propagated along the x-direction to detect the projection of the spin polarization
in the x-direction. The cell and the compensation coils were placed into a
4-layers mu-metal magnetic shield.

of the cell is roughly cubic, each side approximately 2.5 cm.

The cell contained a drop of potassium, 300 Torr Helium-4
and 50 Torr of N2 . The cell was placed into a vacuum oven,
which was made of G-10 fiberglass and continuously pumped
and maintains a pressure of 0.1 Pa. The walls of the oven
were coated with low emissivity paint to reduce the heat radiation. A 100 kHz AC electrical heater was used to heat
the cell to 170 C. The Faraday modulator was used to reduce the probes 1/f noise, whose modulation frequency was
approximately 5.1 kHz. The pump beam was circular polarized and polarizes the potassium atoms along the z-direction.
The pump beam was approximately 2.5 cm in diameter and
originated from a Toptica DL pro external cavity diode laser
system. The probe beams were linearly polarized (the fraction of the light which contains a circular polarization component was approximately 103 ) and propagated along the
x-direction, and were approximately 2 mm in diameter and
originated from a Toptica DL DFB laser system.
The orthogonality of the probe beam and the pump beam
was adjusted and the gradiometric result was obtained by
imaging two probe beams onto two adjacent photodiodes.
C. Experimental results

The non-pumping experimental data were plotted in

Figure 4. The power of the probe beam was approximately
4 mW. The experimental results were in good agreement with
the predicted values obtained from the simulation. Scanning
the magnetic field in the z- and y-direction, respectively, resulted in an output signal that displays a maximum, and scanning the magnetic field in x-direction with a bias magnetic
field in the y-direction produced an output signal curve with a
minimum.
After the non-pumping compensation, the pump beam
was unblocked, and the sequence compensation was imple-

Rev. Sci. Instrum. 85, 063108 (2014)

FIG. 4. Experimental data of non-pumping compensation. Plot of the experimental data of the magnetometer signal response with scanned magnetic field
in x-direction (solid square), y-direction (solid circle), and z-direction (solid
triangle). And the x-direction response curve was gained with 10 nT offset in
y-direction. For comparison, the theoretical data that were based on Eq. (6)
were also plotted in the figure. (Solid line for the theoretical result of the xdirection, dashed line for the theoretical result of the y-direction, and dotted
line for the theoretical result of the z-direction).

mented. A 160 pT oscillating magnetic field was applied in

the z-direction to compensate the magnetic field in the xdirection. Then, a 160 pT oscillating magnetic field was applied in x-direction modulation to compensate the residual
magnetic field in the z-direction. To keep the process quasistatic, the modulation frequency was chosen to be 1 Hz. The
sequence in which the oscillatory magnetic fields are applied
to eliminate the residual field in the z- and x-direction is vitally important. If the residual magnetic field in the z-direction
was first compensated, while the residual magnetic field in the
x-direction is non-zero, it would result in an incorrectly compensated magnetic field in the z-direction.
The cross-modulation method is frequently used elsewhere in the literature but it does not consider the pump
effect of the probe beam,13 consequently, the compensation
sequence started with the z-direction is equivalent to the xdirection in the cross-modulation. Whereas, from Eq. (10)
we can find that if the cross-modulation compensation starts
with adding modulation in the x-direction, then the residual magnetic field in the z-direction after compensation is
equal to 2Rm Bx0 /Rp , which is proportional to the residual magnetic field in the x-direction instead of finding exactly
Bz0 = 0 point. When the residual magnetic field in the zdirection and the x-direction is not well compensated, and the
calibrating magnetic field By is much less than the (By  ),
from Eq. (2) we can find
Rp
Rm 2 + Rp Bx0 Bz0



.
+
B
y
2
2
2
2
Rtot 2 + Bz0
+ Bx0
Rtot 2 + Bz0
+ Bx0
(18)
The scale factor of the magnetometer is reduced due to the
residual magnetic field in the z-direction and the x-direction.
If the magnetic field is well compensated, and Eq. (18)
becomes
Px =

Px =

Rp
Rm
By +
.
Rtot
Rtot

(19)

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47

063108-6

Fang et al.

FIG. 5. The z-direction residual magnetic field (Bz0 ) after sequence compensation (solid circle) and cross-modulation (solid square) are plotted in the
figure. The Bz0 after the sequence compensation is approximately zero with
the increasing of residual magnetic field in the x-direction. However, the Bz0
after the cross-compensation shows a slight fluctuation and this deviation is
approximately proportional to the residual magnetic field in the x-direction.

When the pumping rate of the probe beam is not especially

large, the scale factor is almost the same regardless if there
is no probe beam pumping effect, and the pumping effect of
the probe beam only causes an offset term instead of reducing the sensitivity. So the sequence compensation can mostly
preserve the sensitivity even with the pumping effect of the
probe beam.
In order to clearly demonstrate the comparison of the sequence compensation (proposed in this paper) with the crossmodulation (proposed by Seltzer)13 the power of the pump
beam was slightly decreased to reduce the pumping rate of the
pump beam. First of all, the magnetic field in three directions
was carefully compensated to zero. Then, a bias magnetic
field of 0.5 nT was added in the z-direction to simulate the
residual magnetic field in the z-direction, i.e., Bz0 0.5 nT.
Similarly, a bias magnetic field of 0.5 nT was added in the
x-direction to simulate the residual magnetic field in the xdirection, i.e., Bx0 0.5 nT. Then, cross-modulation began by
compensation in the z-direction. A modulation with 160 pT
amplitude and 1 Hz frequency was applied in the x-direction
to compensate the residual magnetic field Bz0 , and the Bz0
compensation value of the cross-modulation was recorded.
After each compensation sequence, the three direction residual magnetic field was restored to zero, and the experiment
was repeated by increasing the Bx0 in steps of 0.56 nT and
set Bz0 0.5 nT, and recorded the Bz0 compensation value of
cross-modulation and sequence compensation.
In the same way, the magnetic field in three directions
was carefully compensated to zero, and we added bias magnetic fields to simulate the residual magnetic field in the z- and
x-directions, i.e., Bz0 0.5 nT, Bx0 0.5 nT. Then, the sequence compensation was implemented. The sequence compensation had a specific sequence, the modulation in the zdirection was first added and the residual magnetic field in
the x-direction was compensated. Afterwards, the modulation
in the x-direction was applied to compensate the Bz0 , the Bz0
compensation value was recorded. After each compensation
sequence, the residual magnetic field in the three directions
was restored to zero, and the experiment was repeated with increasing the Bx0 in steps of 0.56 nT and set Bz0 0.5 nT, and

Rev. Sci. Instrum. 85, 063108 (2014)

FIG. 6. SERF magnetometer spectra with 22 pT calibration signal at

22.5 Hz, a sensitivity of 4 fT/Hz1/2 was achieved with gradiometer. (Inset)
frequency response of the SERF magnetometer to a calibrated applied signal.

recorded the Bz0 compensation value of the cross-modulation

and sequence compensation.
The experimental result of the two compensation methods is shown in Figure 5. In the experiment, we found that the
z-direction residual magneic field compensation value (Bz0 )
of the cross-modulation was proportional to the residual magnetic field in the x-direction (Bx0 ). In contrast, compensation
value of the sequence compensation was insensitive to the
residual magnetic field in the x-direction.
After optimizing the wavelength and power of the probe
and pump beam, and with the residual magnetic field well
compensated, a 22 pT calibration signal at 22.5 Hz was applied in the y-direction. The output signal of the SERF magnetometer was acquired in an acquisition time of 100 s. The
Welchs power spectral density estimate was used to evaluate the magnetometers sensitivity. As shown in Figure 6, the
magnetometer signal was divided by a normalized frequencyresponse function in order to obtain the absolute field sensitivity, and the
noise level of the 2-channel gradiometer was
divided by 2 to obtain intrinsic sensitivity. The single channel magnetic field sensitivity reached 20 fT/Hz1/2 and using
a 2-channel gradiometer an intrinsic sensitivity of 4 fT/Hz1/2
was achieved. The sensitivity of the gradiometer reduced with
increasing frequency, which was due to the narrow bandwidth
of the magnetometer. The sensitivity of the gradiometer reduced dramatically with increasing frequency as shown by the
frequency response in Figure 6 (Inset).

IV. CONCLUSION

A novel compensation method was demonstrated to compensate the residual magnetic field by considering the pumping rate of the probe beam. This method was important for
optimization of the SERF magnetometer, especially when desiring that the sensitivity of the SERF magnetometer reach
its shot noise limit. In addition this method was useful in
the case where the probe laser could not be largely detuned
from the D2 line of the potassium or less detuned to use saturated absorption frequency stabilization. Furthermore, this
method can be easily used to calculate the light shift produced by the pump beam, which can be obtained simply by

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47

063108-7

Fang et al.

subtracting the non-pumping z-direction compensation value

from the final z-direction compensation value. We optimized
the wavelength of the probe beam to obtain the largest optical
rotation, which can reduce the dependence on the sensitivity of the optical polarimetry. A magnetic field sensitivity of
4 fT/Hz1/2 was realized, which is mainly dominated by the
magnetic shield noise. The Mn-Zn ferrite has a low electrical conductivity and high magnetic permeability, which is an
ideal low-noise magnetic shield material.18, 19 Future work,
will examine using a ferrite magnetic shield as the innermost
layer to reduce the Johnson noise. In addition, a hybrid optical pumping technology will be employed to improve the
sensitivity, whose advantages include having a thinner optical depth, and the reduction of the light shift effect due to the
pump beam.20, 21
ACKNOWLEDGMENTS

The authors acknowledge Patrick J. Clarke for his advice and discussion. This work is supported by the National
Science Foundation of China under Grant Nos. (61227902,
61121003, 61374210, 61074171, and 61273067) and Beijing
Nature Science Foundation (3122025).

Rev. Sci. Instrum. 85, 063108 (2014)

1 I.

K. Kominis, T. W. Kornack, J. C. Allred, and M. V. Romalis, Nature

(London) 422, 596 (2003).
2 R. Wyllie, M. Kauer, R. T. Wakai, and T. G. Walker, Opt. Lett. 37, 2247
(2012).
3 T. H. Sander, J. Preusser, R. Mhaskar, J. Kitching, L. Trahms, and S.
Knappe, Biomed. Opt. Express 3, 981 (2012).
4 H. B. Dang, A. C. Maloof, and M. V. Romalis, Appl. Phys. Lett. 97, 151110
(2010).
5 D. Drung, Physica C 368, 134 (2002).
6 H. Xia, A. B.-A. Baranga, D. Hoffman, and M. V. Romalis, Appl. Phys.
Lett. 89, 211104 (2006).
7 D. Budker and M. Romalis, Nat. Phys. 3, 227 (2007).
8 S. J. Seltzer, Ph.D. dissertation, Princeton University, 2008.
9 T. W. Kornack, Ph.D. dissertation, Princeton University, 2005.
10 J. M. Brown, Ph.D. dissertation, Princeton University, 2011.
11 J. C. Allred and R. N. Lyman, Phys. Rev. Lett. 89, 130801 (2002).
12 I. M. Savukov and M. V. Romalis, Phys. Rev. A 71, 023405 (2005).
13 S. J. Seltzer and M. V. Romalis, Appl. Phys. Lett. 85, 4804 (2004).
14 J. Fang and J. Qin, Rev. Sci. Instrum. 83, 103104 (2012).
15 Z. Li, R. T. Wakai, and T. G. Walker, Appl. Phys. Lett. 89, 134105 (2006).
16 I. A. Sulai, R. Wyllie, M. Kauer, G. S. Smetana, R. T. Wakai, and T. G.
Walker, Opt. Lett. 38, 974 (2013).
17 J. Fang, S. Wan, Y. Chen, and R. Li, Appl. Opt. 51, 7714 (2012).
18 T. W. Kornack, S. J. Smullin, S.-K. Lee, and M. V. Romalis, Appl. Phys.
Lett. 90, 223501 (2007).
19 S.-K. Lee and M. V. Romalis, J. Appl. Phys. 103, 084904 (2008).
20 M. V. Romalis, Phys. Rev. Lett. 105, 243001 (2010).
21 Y. Ito, H. Ohnishi, K. Kamada, and T. Kobayashi, IEEE Trans. Magn. 47,
3550 (2011).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitationnew.aip.org/termsconditions. Downloaded to IP:
193.231.11.233 On: Tue, 07 Oct 2014 15:43:47