Sie sind auf Seite 1von 3


1, JANUARY 2010

Spintronic Memristor Temperature Sensor

Xiaobin Wang, Member, IEEE, Yiran Chen, Member, IEEE, Ying Gu, and Hai Li, Member, IEEE

Abstract—Thermal fluctuation effects on the electric behavior

of a spintronic memristor based upon the spin-torque-induced
domain-wall motion are explored. Depending upon material,
geometry, and electric excitation strength, the device electric be-
havior can be either sensitive or insensitive to temperature change.
We present temperature sensor designs that operate at a temper-
ature sensitive region. The sensitivity is achieved through a com-
bination of the temperature-dependent domain-wall mobility and
the positive feedback between memristor resistance and driving
Index Terms—Domain wall, spintronic memristor, temperature
sensor, thermal fluctuations.


T HE MEMRISTOR concept [1] recently received signif-

icant attention due to the demonstration of a practical
nanoscale memristor device based upon ionic transport in a
resistive memory stack [2]. Nanoscale spintronic memristor has
been proposed based upon spin-torque-induced magnetization Fig. 1. Spin-valve magnetic domain-wall memristor.
motion [3] and spin transport at semiconductor/ferromagnet
junction [4]. where RH and RL are the high and low resistances of the spin
Thermal fluctuation at finite temperature has important im- valve. D is the spin-valve length, and X is the domain-wall
plications to nanoscale magnetic devices [5], [6]. In this letter, position.
we propose a spintronic memristor temperature sensor explor- Domain-wall position is moved through current-induced spin
ing thermal fluctuation effects on domain-wall motion in a spin- torque excitation. For a ferromagnet with magnetization satura-
valve structure. tion Ms , exchange strength A, easy z-axis anisotropy Hk , and
hard y-axis anisotropy Hp , the spin-torque-induced domain-
II. D EVICE T HEORY wall motion at finite temperature is described through stochastic
Landau–Lifshitz–Gilbert equation [7] with a spin torque term
The spin-valve domain-wall memristor structure is shown in [8], [9]. Using rigid wall approximation [10], [11], the domain-
Fig. 1. The spintronic memristor consists of a long spin-valve wall motion is expressed in terms of magnetization spherical
strip which includes two ferromagnetic layers: reference and angles Ms (sin θ cos φ, sin θ sin φ, cos θ) as
free layers. The magnetization direction of the reference layer
is fixed by coupling to a pinned magnetic layer. The free layer is θ(x, t) = θ0 (x − X(t)) φ(x, t) = φ0 (t)
divided by a domain wall into two segments that have opposite
magnetization directions to each other. The device time domain where θ0 (x) = arccos[tanh(x/w)] is the function of the
resistance depends upon domain-wall position 
domain-wall shape. w = 2A/Ms Hk is the domain-wall
thickness. X(t) is the domain-wall position. Domain-wall
R(t) = RH − (RH − RL )X(t)/D (1) speed is dX(t)/dt. The domain-wall position X(t) satisfies the
following stochastic differential equations [10], [11]:

dφ α dX
+ = ηφ
Manuscript received October 9, 2009. First published December 1, 2009; dt w dt
current version published December 23, 2009.
1 dX dφ vs
X. Wang and Y. Chen are with Seagate Technology, Bloomington, MN 55435 −α = ω0 sin(2φ) + + ηX (2)
USA (e-mail:; w dt dt w
Y. Gu is with the St. Cloud State University, St. Cloud, MN 56301 USA
H. Li is with the Department of Electrical and Computer Engineering, where ω0 = γHp /2 (γ being the gyro-magnetic ratio), α is the
Polytechnic Institute of New York University, Brooklyn, NY 11201 USA (e- damping parameter, and vs = P JuB /eMs is the spin torque
mail: excitation strength. P is the polarization efficiency, uB is the
Color versions of one or more of the figures in this letter are available online
at Bohr magneton, and e is the elementary electron charge. ηφ (t)
Digital Object Identifier 10.1109/LED.2009.2035643 and ηX (t) are the φ and X component thermal fluctuation

0741-3106/$26.00 © 2009 IEEE

Authorized licensed use limited to: Seagate Technology. Downloaded on January 9, 2010 at 21:38 from IEEE Xplore. Restrictions apply.

fields, respectively. Their magnitudes are determined through

the fluctuation dissipation condition
2αkB T
ηφ (t)ηφ (t ) = ηX (t)ηX (t ) = δ(t − t ) (3)
where kB is the Boltzman constant, T is the temperature,  is
the Planck constant, and N = 2wS/a3 is the number of spins
in the domain wall with the cross-sectional surface area S and
domain-wall thickness w. a is the lattice constant.


Domain-wall mobility at finite temperature depends upon
both spin torque excitation strength and thermal fluctuation
magnitude. Equation (2) predicts that the domain-wall velocity
is a function of the normalized spin torque excitation strength
and the normalized thermal fluctuation magnitude.
Spin torque excitation strength is proportional to cur-
rent density. The normalized current density is defined as
P JuB /eMs wω0 . At zero temperature, the domain wall starts
to move only when the current density is above the crit-
ical value, P JuB /eMs wω0 = 1. At finite temperature, the
domain wall can move even when current density is below
the critical value. The average domain-wall velocity (normal-
ized by vs ) as a function of the normalized current density
(P JuB /eMs wω0 ) and the normalized thermal fluctuation
magnitude (4kB T /N ω0 ) can be obtained by solving the sto-
chastic differential equation (2). Detailed mathematical tech-
niques can be referred to [12]. Fig. 2 shows the normalized
domain-wall velocity as a function of the normalized current
density for different normalized thermal fluctuation magni- Fig. 2. Average domain-wall velocity as a function of the normalized current
tudes. Temperature sensitive and insensitive regions can be density for different thermal fluctuation strengths. (a) and (b) are the same figure
observed. Curves with kneeling shapes are around the critical plotted at different normalized current density scales.
current density, where the domain-wall velocity is sensitive to
thermal fluctuation magnitude.
The positive feedback between resistance and driving strength
The spin-valve low and high resistances scale as RL /Rh0 =
is a unique property of the memristor. Memristor’s resistance
[D/Z(h/h0)] and RH = RL (1 + GM R)RL , where Rh0 is the
depends upon the integration of current/voltage excitation.
sheet film resistance with a thickness h0 , giant magnetoresis-
For a constant voltage pulse driving, a higher temperature
tance ratio (GMR) is the magnetoresistance ratio, and D, Z,
results to an increased domain-wall moving distance. The
and h are the ferromagnet length, width, and thickness.
increased domain-wall moving distance results to a smaller
resistance. The smaller resistance results to a higher driving
current density, thus providing a positive feedback to further
For temperature sensing, a biasing voltage pulse with con- increase domain-wall distance. This positive feedback acceler-
stant magnitude is applied to the spintronic memristor. Resis- ates domain-wall speed and reduces device resistance further
tance difference before and after voltage pulse is measured. for constant voltage pulse driving.
This resistance difference is calibrated to the sense temperature We illustrate the temperature sensing working principle
magnitude. Fig. 2 shows that domain-wall mobility increases as through a practical device design example. The material
temperature increases. Equation (1) shows that memristor resis- properties of the device are magnetization saturation Ms =
tance dropping is proportional to domain-wall moving distance. 1010 emu/cc, exchange strength A = 1.8 · 10−11 J/m, easy
Thus, after applying a constant magnitude voltage pulse, device z-axis anisotropy Hk = 100 Oe, and hard y-axis anisotropy
resistance dropping is bigger at a higher temperature. Hp = 5000 Oe. The resistance of the sheet film is 50 Ω for a
The temperature sensing memristor is operated at a re- 70-A thickness square thin film, and GMR is 12%. The geome-
gion where its electric behavior is sensitive to temperature try of the device is 268 nm long, 17 nm wide, and 10 nm thick.
change. This is achieved through a combination of temperature- The damping parameter is α = 0.02, and polarization efficiency
dependent domain-wall mobility and the positive feedback is P = 0.3. The number of spins in the domain wall is N =
between resistance and driving strength in memristor. Fig. 2 2.2 · 107 . Using the aforementioned parameters, for Fig. 2,
shows the sensitive dependence of domain-wall velocity upon the normalized thermal magnitude is 4kB T /N ω0 = 0.008 at
temperature at kneeling region around the critical current value. room temperature T = 300 K. The critical current density is

Authorized licensed use limited to: Seagate Technology. Downloaded on January 9, 2010 at 21:38 from IEEE Xplore. Restrictions apply.

much lower power supply voltage (which can be even used

in the subthreshold voltage region) and power consumption
(compared to some on-chip oscillation-ring-based temperature
sensor that consumes huge dynamic power), nanoscale feature
size, low cost, and the mature integration technology with the
CMOS process. Please note that our device is targeting the
highly integrated on-chip thermal detection applications (e.g.,
cell size < 1 μm2 ). Therefore, it may not be fair to compare our
device to any standalone or bulk thermal sensors.
We note that our main contribution in this letter is to demon-
strate the theory and feasibility of a temperature sensor based
on the domain-wall motion in magnetic strip in a spin-valve
structure. The operating range and device properties can be
further tuned according to practical needs and material capabil-
ities (e.g., the thermal effects in the device scale as temperature
Fig. 3. Spintronic memristor resistance as a function of time at different
temperatures (300, 350, and 400 K) for a constant magnitude voltage pulse
over domain-wall thickness, as shown in Section III). Varying
driving. domain-wall thickness through the tuning magnetic mater-
ial properties could change the device temperature operating
range. Linearity or resolution of the device can be improved
Jsc = eMs wω0 /P uB = 3 · 108 A/cm2 . The normalized veloc-
by utilizing the magnetic stack with high GMR or even the
ity is P Jsc uB /eMs = 52 m/s at critical current density.
tunneling magnetoresistance ratio stack. Also, it is well known
We use a single square wave pulse with a duration of 80 ns
that memristor has a rich dynamic behavior when excited with
and a magnitude of 0.3 V to excite the device. Fig. 3 shows the
a dynamic current/voltage profile. Thus, device behavior can
device resistance dropping as a function of time for different
be further optimized by examining the combined effects of
temperatures (300, 350, and 400 K). Device resistance eventu-
current/voltage excitation and domain-wall motion.
ally flattens out when domain wall settles to a fixed position
after voltage pulse excitation. Solid curves on the figure are
the resistance changes for the proposed spintronic memristor at R EFERENCES
different temperatures. Dash lines are the resistance dropping [1] L. O. Chua, “Memristor—The missing circuit element,” IEEE Trans.
Circuit Theory, vol. CT-18, no. 5, pp. 507–519, Sep. 1971.
for a nonmemristive device without positive feedback between [2] D. B. Strukov, G. S. Snider, D. R. Stewart, and R. S. Williams, “The
the resistance and the integration of the driving strength. The missing memristor found,” Nature, vol. 453, no. 7191, pp. 80–83,
dash lines are equivalent to the simulations with fixed driving May 2008.
[3] X. Wang, Y. Chen, H. Xi, H. Li, and D. Dimitrov, “Spintronic memris-
current strength of 0.3 V /RH S. It can be seen that a positive tor through spin torque induced magnetization motion,” IEEE Electron
feedback between resistance and driving strength in memristor Device Lett., vol. 30, no. 3, pp. 294–297, Mar. 2009.
significantly increases the temperature sensing margin. [4] Y. V. Pershin and M. Di Ventra, “Spin memristive systems: Spin memory
effects in semiconductor spintronics,” Phys. Rev. B, Condens. Matter,
The material of our proposed spintronic memristor tempera- vol. 78, no. 11, pp. 113 309-1–113 309-3, Sep. 2008.
ture sensor is similar to the commercial recording head mater- [5] N. Smith and P. Arnett, “White-noise magnetization fluctuations in mag-
ial. However, we believe that our proposed device structure is netoresistive heads,” Appl. Phys. Lett., vol. 78, no. 10, pp. 1448–1450,
Mar. 2001.
novel: we are not aware of any applications of domain-wall mo- [6] N. H. Bertram, X. Wang, and V. L. Safonov, “Dynamic-thermal effects
tion in a spin-valve strip structure in any practical devices. The in thin film media,” IEEE Trans. Magn., vol. 37, no. 4, pp. 1521–1527,
positive feedback between domain-wall motion and accumu- Jul. 2001.
[7] W. F. Brown, “Thermal fluctuations of a single-domain particle,” Phys.
lated resistance change is the key ingredient for this memristive Rev., vol. 130, no. 5, pp. 1677–1686, Jun. 1963.
device. The sensing temperature range of our proposed device [8] J. C. Slonczewski, “Current driven excitation of magnetic multilayers,”
is from 300 to 400 K, which covers the normal operation range J. Magn. Magn. Mater., vol. 159, no. 1/2, pp. L1–L7, Jun. 1996.
[9] L. Berger, “Emission of spin waves by magnetic multilayer traversed by a
of semiconductor chip. The required power supply voltage is current,” Phys. Rev. B, Condens. Matter, vol. 54, no. 13, pp. 9353–9358,
0.3 V, with a typical power consumption of 150 μW. The linear- Oct. 1996.
ity is 0.5 Ω/◦, and the surface area is 300 nm × 20 nm. Because [10] G. Tatara and H. Kohno, “Theory of current-driven domain wall motion:
Spin transfer versus momentum transfer,” Phys. Rev. Lett., vol. 92, no. 8,
the working principle of the device is based upon integration p. 086 601, Feb. 2004.
of domain-wall motion under thermal fluctuation, the intrinsic [11] R. A. Duine, A. S. Nunez, and A. H. MacDonald, “Thermally assisted
random noise variations are averaged and suppressed. The de- current-driven domain-wall motion,” Phys. Rev. Lett., vol. 98, no. 5,
p. 056 605, Feb. 2007.
vice provides reproducible temperature readings. Compared to [12] H. Risken, The Fokker-Planck Equation. New York: Springer-Verlag,
the available on-chip temperature sensors, our device requires 1984.

Authorized licensed use limited to: Seagate Technology. Downloaded on January 9, 2010 at 21:38 from IEEE Xplore. Restrictions apply.