Materials Science Forum Vols. 514-516 (2006) pp. 323-327 online at http://www.scientific.net © (2006) Trans Tech Publications, Switzerland
Heat Generation in Tunnel Junctions for Current-written Pinned Layer Switching J. Ventura1; a, A. Pereira1, J. M. Teixeira1, J. P. Araujo1, F. Carpinteiro1, J. B. Sousa1; b, Y. Liu2, Z. Zhang2 and P. P. Freitas2; c 1
DFFCUP, IFIMUP, Rua do Campo Alegre, 687, 4169-007, Porto, Portugal 2
INESC-MN, Rua Alves Redol, 9-1, 1000-029 Lisbon, Portugal
a
[email protected],
[email protected],
[email protected]
Keywords: Tunnel Junction, Heat, Heat Generation, Current written, MRAM Abstract. To commute between the different resistance states of a magnetic tunnel junction (TJ) one can use a thermally-induced pinned layer switching mechanism. When a sufficiently high electrical current flows through the insulating barrier, local temperatures inside the tunnel junction can increase above the blocking temperature of the antiferromagnetic layer used to pin the magnetization of the adjacent ferromagnet. Then, it is possible to switch the magnetization of the pinned layer with a small magnetic field H and thus revert the magnetic state of the TJ. Here we demonstrate thermally-induced pinned layer switching in thin magnetic tunnel junctions. We further present numerical results that suggest that heating is small when one takes into consideration the uniform current density flowing through the tunnel junction and that one must conclude that nanoconstrictions concentrate most of the current, increasing local current densities and temperature. Simulation of heating and cooling times demonstrates that current-induced pinned layer switching is a competitive mechanism for actual technological applications. Introduction Tunnel junctions (TJ) consisting of two ferromagnetic (FM) layers separated by an insulator (I) [1] show enormous potential for a multiplicity of applications such as read head [2], strain [3], current, position and speed [4] sensors or even to detect magnetically tagged biological specimens [5]. However, probably the most sought after application is high performance, low cost, non-volatile magnetic random access memories (MRAMs) [6]. In a tunnel junction, the magnetization of one of the FM layers (pinned layer) is fixed by an underlying antiferromagnetic (AFM) layer. The magnetization of the other FM layer (free layer) reverses almost freely when a small magnetic field is applied. Due to spin dependent tunneling one obtains two distinct resistance (R) states (the 0 and 1 bits of a magnetic memory) corresponding to pinned and free layer magnetizations parallel (low R) or antiparallel (high R). Consequently, the standard way to switch between R-states in MRAMs is to use magnetic fields generated by current lines. However, the undesirable switching of halfselected bits is still a concern for actual MRAM submicron devices. Furthermore, as the size of a memory cell decreases, the magnetic field needed to induce switching greatly increases. To overcome such limitations, a thermally-induced pinned layer switching mechanism was proposed [7, 8]. In fact, when a sufficiently high electrical current flows through the insulating barrier, local temperatures inside the tunnel junction can increase above the blocking temperature of the AFM layer (TB ~ 500 K). One is then able to switch the magnetization of the pinned layer with a small magnetic field H and, upon cooling (under H), a new exchange-bias pinning direction is impressed. Switching of the magnetic state of the TJ (from parallel to antiparallel or vice-versa) is then possible. Here we demonstrate the current-driven pinned layer switching effect in thin magnetic tunnel junctions (MnIr/CoFe/AlOx/CoFe/NiFe) with TB = 520 K [9] using a current density j = 0.75x106 A/cm2. However, numerical results on heat generation in tunnel junctions show that heating is small All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of the publisher: Trans Tech Publications Ltd, Switzerland, www.ttp.net. (ID: 193.137.24.76-02/10/06,12:16:44)
324
Advanced Materials Forum III
when such current densities are applied. In fact, only much larger current densities (~15x106 A/cm2) lead to heating above the blocking temperature of the AFM layer. One concludes that the experimental observation of thermally driven pinned layer reversal is due to localized heating in nanoconstrictions that concentrate most of the electrical current leading to high local current densities. Thus, to enhance device performance, the insulating barrier of the tunnel junction should have nanometric inhomogeneities where the local barrier thickness is smaller than average. Such hot-spots lead to a confinement of the electrical current and to an increase of local current densities. Furthermore, one shows that both heating and cooling (above and below TB respectively) occur over very small time scales (≤ ns), making current-written pinned layer switching a competitive mechanism for technological implementation. Experimental details The complete structure of the Ion Beam Deposited tunnel junction series [10, 11] used in this work is glass/bottom lead/Ta (90 Å)/NiFe (50 Å)/MnIr (90 Å)/CoFe (80 Å)/AlOx (3 Å+ 4 Å)/CoFe (30 Å)/NiFe (40 Å)/Ta (30 Å)/TiW(N) (150 Å)/top lead. The AlOx barrier was formed by two-step deposition and oxidation processes. NiFe, CoFe, MnIr and TiW(N) stand for Ni80Fe20, Co80Fe20, Mn78Ir22 and Ti10W90(N). The bottom and top leads are made of Al 98.5% Si 1% Cu 0.5%, 600 Å and 3000 Å thick respectively, and are 26 µm and 10 µm wide. The junctions were patterned to a rectangular shape with areas (A) ranging from 1x1 µm2 to 4x2 µm2 by a self-aligned microfabrication process. The samples were annealed at 550 K under an external magnetic field to impress an exchange bias direction between the AFM and FM pinned layers, taken here as the positive direction. Magnetoresistance (MR) was measured with a four-point d.c. method and an automatic control and data acquisition system. Current pulses of 1s duration were applied between consecutive MR measurements to induce current-driven pinned layer switching. Experimental results Figure 1i) shows a standard magnetoresistance MR(H) cycle, giving TMR = 14% under a small and constant bias current (I = 1 mA). The strongly pinned layer is fixed in the direction of positive fields. The initial state (A) under H = -200 Oe then corresponds to antiparallel free/pinned layer magnetizations (the free layer is easily oriented into the H-direction), giving a high-R state. When the magnetic field is increased towards positive values, this state essentially remains until H again reverses the magnetization of the free layer (resulting in parallel free/pinned layer magnetizations), to produce a lower R state (B). The reverse sequence occurs when H is decreased from +200 Oe to 200 Oe, closing the MR(H) cycle in the initial antiparallel state (A). Keeping H = -200 Oe (state A), one applied a current pulse of I = 40 mA (j = 1.3 MA/cm2). One notices that the following MR(H) cycle appears inverted (Fig. 1ii). This means that while applying the current pulse under H = -200 Oe, the tunnel junction has reverted to its parallel state under sufficiently large current pulses. Such pinned layer magnetic switch was made possible through heating under high current pulses, which rise the local temperature inside the TJ above the blocking temperature of the 90 Å thick MnIr AFM layer (TB = 520 K; [9]). The corresponding magnetization then easily aligns with the applied negative magnetic field, opposite to the initial exchange bias direction. Subsequent cooling below TB (when the current is removed) leaves the TJ in such inverted exchange bias direction. This effect has already been considered for pinned-layer writing under current pulses [7]. If a current pulse of equal magnitude is now applied under a magnetic field in the opposite direction (H = +200 Oe), the tunnel junction reverts to its original behavior (Fig. 1iii). One can then current-reverse the magnetic state of the tunnel junction using large enough current densities and magnetic fields of opposite signs.
Materials Science Forum Vols. 514-516
9
After a current pulse
After a current pulse
9
under H=-200 Oe
325
under H=+200 Oe
R (
R (
)
)
A
8
B
(i)
H (Oe)
8
-100
0
100
(ii)
H (Oe)
-100
0
100
(iii)
H (Oe)
-100
0
100
Fig. 1: i) Standard initial MR(H) cycle. ii) MR(H) cycle performed after applying a large current pulse I = 40 mA (j = 1.3 MA/cm2) under H = -200 Oe. iii) MR(H) cycle performed after applying a large current pulse under H = +200 Oe. The thicker (thinner) arrows refer to the pinned (free) layer magnetization. Numerical Results Heat generation in tunnel junctions arises from two processes [12]: usual Joule heating in the metallic layers and electron inelastic scattering upon ballistic tunneling. The steady-state heat equation can then be written as [12]: cpd
∂T ∂ 2T jV − x lin − K 2 = ρj 2 + e , ∂t lin ∂x
(1)
where cp is the heat capacity, d is the mass density, K is the heat conductivity, T is the temperature, x is the stack position, ρ is the electrical resistivity, j = V/(RA) is the current density, V is the bias voltage and lin is the inelastic scattering mean free path. Numerical results were obtained assuming that the current density is constant throughout the junction stack. Two different simulations were performed. In the first (second), one disregards (includes) the top and bottom leads and the temperature at the bottom and top of the tunnel junction stack (leads) is assumed fixed at 300 K. Other boundary conditions do not alter much the results obtained by the simulation [12]. Values of the parameters used can be found in Ref. [12].
150 MA/cm
1500
T (K)
Advanced Materials Forum III
T (K)
326
2
600
750 2
j (MA/cm )
15
75 MA/cm
30
2
450
4 MA/cm
2
37.5 MA/cm
2
300
Ta (30 Å)
NiFe (40 Å)
AlO
x
40 CoFe (30 Å)
30 CoFe (80 Å)
20 MnIr (90 Å)
10 NiFe (50 Å)
x (nm) Ta (90 Å)
0
Fig. 2: Simulation of heating processes inside the studied tunnel junctions, under different electrical current densities (1 MA= 106 A). Inset: maximum temperature increase as a function of current density passing through the junction. The results shown in the figure (inset) were obtained disregarding (taking into consideration) the top and bottom leads. Steady-state regime Our numerical results (Fig. 2) indicate that large heating can occur near the insulating barrier for high current densities. The results presented in the figure were obtained disregarding the top and bottom leads but allow us to have a more detailed knowledge on the temperature increase behavior in the region of interest. More accurate results are shown in the inset of Fig. 2, where both leads were included. Notice that the simulated maximum temperature increase is (for the same current density) higher when the leads are present. The expected maximum temperature increase arising from an uniform current density j = I/A =1.3x106 A/cm2 is small (~30 K; inset of Fig. 2), and to reach TB = 520 K our numerical results suggest jnum~15x106 A/cm2. This corresponds to an effective area through which current flows Aeff = I/jnum≈0.3 µm2, i.e., about 7% of the total tunnel junction area. These results then suggest that the considered uniform current density j = 1.3x106 A/cm2 is only an average value and that nanoconstrictions concentrate most of the current flowing through the barrier. Such hot-spots have been revealed by atomic force microscopy [12] and correlate well with results presented here. Dynamic regime For MRAM application, write operations should occur in the ns-time frame. For that reason we performed dynamical simulations of heating and cooling (above and below TB respectively) to predict if time was a constrainment when using the current-written pinned layer switching scheme. Figure 3 shows the obtained results using j = 112.5 MA/cm2 and disregarding the top and bottom leads. One observes that both heating up to TB and cooling back to room temperature occur in a very small time frame of less than 1 ns, making current-written pinned layer switching a competitive mechanism for actual technological applications.
Materials Science Forum Vols. 514-516
327
525
T (K)
525
i)
350 0
25
50
75
t (ps)
T (K)
41 ps
40 ps
44 ps
ii) 4 ps
50 ps
20 ps
350
60 ps
1 ps
100 ps
0
20
x (nm)
40
0
20
x (nm)
40
Fig. 3: i) Simulation of dynamical heating processes as a function of time for j = 112.5 MA/cm2. ii) Cooling under zero applied current. Inset: maximum temperature inside the tunnel junction as a function during heating and cooling. All simulations were performed disregarding the top and bottom leads. Acknowledgement Work supported in part by POCTI/0155, POCTI/CTM/36489/2000, IST-2001-37334 NEXT MRAM, POCTI/CTM/45252/02 and POCTI/CTM/59318/2004 projects. J. Ventura is thankful for a FCT doctoral grant (SFRH/BD/7028/2001). Z. Zhang and Y. Liu are thankful for FCT post-doctoral grants (SFRH/BPD/1520/2000 and SFRH/BPD/9942/2002). References [1] J. S. Moodera, L. R. Kinder, T. M. Wong and R. Meservey, Phys. Rev. Lett. 74, 3273 (1995). [2] D. Song, J. Nowak, R. Larson, P. Kolbo and R.Chellew, IEEE Trans. Magn. 36, 2545 (2000). [3] M. Lohndorf, T. Duenas, M. Tewes, E. Quandt, M. Ruhrig and J. Wecker, Appl. Phys. Lett. 81, 313 (2002). [4] P. P. Freitas, F. Silva, N. J. Oliveira, L. V. Melo, L. Costa and N. Almeida, Sensors and Actuators A 81, 2 (2000). [5] H. A. Ferreira, D. L. Graham, P. P. Freitas and J. M. S. Cabral, J. Appl. Phys. 93, 7281 (2003). [6] B. N. Engel, N. D. Rizzo, J. Janesky, J. M. Slaughter, R. Dave, M. DeHerrera, M. Durlam and S. Tehrani, IEEE Trans. Nanotechnol. 1, 32 (2002). [7] J. Zhang and P. P. Freitas, Appl. Phys. Lett. 84, 945 (2004). [8] I. L. Prejbeanu, W. Kula, K. Ounadjela, R. C. Sousa, O. Redon, B. Dieny and J. P. Nozieres, IEEE Trans. Magn. 40, 2625 (2004). [9] H. Li, P. P. Freitas, Z. Wang, J. B. Sousa, P. Gogol and J. Chapman, J. Appl. Phys. 89, 6904 (2001). [10] J. Ventura, J. B. Sousa, Y. Liu, Z. Zhang and P. P. Freitas, Phys. Rev. B 72, 094432 (2005). [11] J. B. Sousa, J. Ventura, M. A. S. da Silva, P. P. Freitas and A. Veloso, J. Appl. Phys. 91, 5321 (2002). [12] R. C. Sousa, I. L. Prejbeanu, D. Stanescu, B. Rodmacq, O. Redon, B. Dieny, J.Wang and P. P. Freitas, J. Appl. Phys. 95, 6783 (2004).
