Storage Reliability and Temperature Increment with Tilted Free Layer Magnetization in Nanopillars for Spin Torque Magnetic Memory

Transcription

1 490 Chiang Mai J. Sci. 2015; 42(2) Chiang Mai J. Sci. 2015; 42(2) : Contributed Paper Storage Reliability and Temperature Increment with Tilted Free Layer Magnetization in Nanopillars for Spin Torque Magnetic Memory Chayada Surawanitkun [a], Arkom Kaewrawang [b], Roong Sivaratana [c], Anan Kruesubthaworn [a] and Apirat Siritaratiwat*[b] [a] Faculty of Applied Science and Engineering, Nongkhai Campus, Khon Kaen University, Nongkhai 43000, Thailand. [b] KKU-Seagate Cooperation Research Laboratory, Faculty of Engineering, Khon Kaen University, Khon Kaen 40002, Thailand. [c] Seagate Technology, 1627, Teparak, Samutprakarn 10200, Thailand. *Author for correspondence; apirat@kku.ac.th Received: 12 December 2012 Accepted: 22 September 2014 ABSTRACT Recently, the temperature increment in magnetic tunnel junction (MTJ) nanopillars with relevant current induced magnetization switching for spin transfer torque magnetic random access memory (STT-MRAM) has become interesting because it affects the reliability of devices. In this work, the magnetic stability and the temperature rise in the MTJ nanopillar during the switching process at pulse durations of less than 1 ns were explored with tilted magnetization in the free layer (FL). The thermal simulation was performed by the 3D finite element method. The results indicate that the increase of the initial angle between the magnetization vector and the major axis in the FL, θ 0, can reduce the temperature increment in the device and the risk for the magnetic damage impacting the storage stability. In addition, it was found that the altered temperature during the switching process is proportional to the square of ln(π/2θ 0 )). Thus, the reliability of the high-density STT-MRAM can be improved by tilting the FL magnetization in MTJ devices. Keywords: current induced magnetization switching, STT magnetic random access memory, Joule heating, magnetic tunnel junction 1. INTRODUCTION Spin transfer torque magnetic random access memory (STT-MRAM) is a popular candidate for future memory technology due to several good points, such as high write/read speed, durability, non-volatility, high recording density, and energy-efficient writing [1-3]. Magnetic tunnel junction (MTJ) nanopillars consisting of an insulator layer located between two ferromagnetic (FM) layers, which are the free layer (FL) and the pinned layer (PL), are used as a bit cell of STT-MARM [1-3]. Also, the parallel (P) and

2 Chiang Mai J. Sci. 2015; 42(2) 491 anti-parallel (AP) orientations of the magnetization in the two FM layers relating to low and high resistance, respectively, are applied to write data in each bit cell [1, 3]. Because the high tunneling magnetoresistance ratio and the low resistance area product achieved by the MTJ structure with a thin MgO insulator barrier and the CoFeB ferromagnetic layers are significant to developing an effective MRAM technology, CoFeB/MgO/CoFeB MTJ stacks are widely adopted to produce a high density MARM [1]. The theory of current induced magnetization switching (CIMS) based on the STT effect can be utilized to change the direction of magnetization in the FL [1, 2, 4-6]. Meanwhile, the PL magnetization direction is fixed by the interaction between an antiferromagnetic (AFM) layer and an FM layer [1, 7]. The phenomenon, known as the exchange bias effect vanishes whenever the temperature in the AFM/FM bilayers exceeds the blocking temperature, T B [7-9]. This signifies that the magnetic properties in the device depend on temperature [10-13]. Moreover, the large current density of several MA/cm 2 required for CIMS leads to unavoidable Joule heating with high temperature increment [14, 15]. This might affect the magnetic properties, the thermal stability determined by magnetic anisotropy of the magnetic FL and its volume as a function of the film thickness and bit area [14, 15]. In particular, the switching current density increases drastically with decreasing pulse duration, τ p, of less than 10 ns and the thermal factor of the FL in the MTJ cell is an interesting issue in the τ p < 1 ns region [16]. Therefore, in order to maintain the stability of the device during read/write operations, the thermal stability factor,, associated with STT current writing should exceed about 40 [17]. Nowadays, there is much research focusing on the principal aspect of switching current density reduction together with increasing writing speed [18]. One is the tilt of the initial magnetization direction in the FM layers [18, 19]. In 2000, Sun presented an analytical solution for spin current driven magnetization dynamics with a homogenous (single-domain) magnetization state [20, 21]. It is important to estimate the critical current for magnetization reversal in the MTJ cell [21, 22]. Therefore, the aim of this work is to explore the effect of temperature increment, T, during the CIMS process for ultrafast switching time (less than 1 ns) with different initial magnetization directions in FL on the magnetic stability in the MTJ cell for STT-MRAM. 2. MATERIALS AND METHODS 2.1 Thermal Calculation The temperature profile of the MTJ nanostructure was found by numerical calculation based on the commercial finite element method (FEM) software (multiphysics finite-element method COMSOL). Three dimensional (3D) FEM was used for calculation of T undergoing Joule heating and heat conduction during the CIMS process. For the Joule heating principle, the device heated up by passing current through it can be described by: J 2 lat Q = I 2 Rt = σ, (1) where Q is the generated heat which is proportional to the power, I is the applied current, t is the duration of the applied current, J is the current density, R is the resistance, A is the area of current injection, l is the length and σ is the electrical conductivity. To analyze the effect of temperature in MTJ nanopillars during current injection for the

3 492 Chiang Mai J. Sci. 2015; 42(2) switching process, the conduction theory in heat transfer considered together with Joule heating effect is given as follows: ρc T p -.(K T ) = Q, (2) t where ρ, c p, K and T are, respectively, the density, heat capacity, thermal conductivity and temperature in the device. The examined system comprises two Cu electrodes divided by SiO 2 insulator surrounding the lateral area of the MTJ pillar similar to the geometry reported by S.S. Ha et al. [23], as shown in Figure 1. The bottom and top rectangular electrodes have the optimal size of 1.43 μm 0.8 μm with thickness of 0.1 μm [24, 25]. The MTJ nanopillar structure in the STT-MRAM is assumed to be a rectangular cylinder, although its common feature is normally an elliptical cylinder. This is because the results of T in the local cross-sectional area of an elliptical pillar are similar to a rectangular one and the temperature is symmetrical in the direction of current flow along the z axis [23]. The MTJ multilayer structure (from the bottom electrode) considered here is PtMn (15)/CoFe (2.5)/Ru (0.85)/CoFeB (2.4)/MgO (0.83)/CoFeB (1.8) (thickness in nm). It has a resistance area at a parallel state of 4.3 Ω μm 2 and a lateral size of 130 nm 50 [26]. Figure 1. Diagram detailing the structure of the MTJ nanopillar in STT-MRAM. The values of the material parameters used to solve the heat conduction and Joule heating equations in this model are given in Table 1. Although these values correspond to bulk element properties, they are acceptable for realistic simulation of heat transfer in the MTJ device as similarly reported [11, 27-31]. This is because, whether a bulk material or a nanoscale material, physical and chemical properties of material depend on surface properties [32]. The difference between a bulk material and a nanomaterial is the increment of surface area to volume ratio of device, at the same volume, due to the size effect. However, for a thin film structure, the thickness is much smaller than the width and length so the properties are determined by the shape. The initial temperature inside and outside the sample is supposed to be room temperature. 2.2 Analysis of CIMS with Tile of FL Magnetization The model assumes that the current flows from bottom to top electrode or the electrons pass through the FL to the PL.

4 Chiang Mai J. Sci. 2015; 42(2) 493 In this case, the interaction of spin electron with the FL and PL magnetizations could result in a transition from P to AP magnetic state [1]. Because the switching energy for P to AP state is more than the opposite case [22, 26], it was considered in this report. Moreover, in principle, the FL magnetization in the MTJ pillar could be reversed when a current density is applied with a sufficient τ p. The critical current density, J C, depending on the initial angle of magnetization in the FL with different was estimated from the analytical solution described elsewhere [20]: (3) Here θ 0 represents the initial angle between the magnetization vector and the easy axis in the FL (angle in radians). The intrinsic critical current density, J C0, and the relaxation time, τ relax, are defined as follows: (4) (5) where e is the electron charge, α is the damping parameter, M S is the saturation magnetization, t free is the thickness of the FL, H eff is the effective field including exchange, anisotropy, demagnetization, and external fields, h = h/(2π), h is Planck s constant, η is the spin torque efficiency, P is the spin polarizing factor of the FL and γ is the gyromagnetic ratio ( m/(a s) for a free electron). Table 1. Material properties used in the 3D FEM for thermal analysis [11, 27-31, 33]. Material PtMn CoFe Ru CoFeB MgO SiO 2 Cu Electrical conductivity σ(ω m) Thermal conductivity K(W/m K) Heat capacity c p (J/(kg.K)) Density ρ(kg/m 3 ) The values of these parameters are M S of A/m, α of 0.01, P of 1, η of 1 and the exchange stiffness constant of [34, 35], based on the magnetic properties of CoFeB thin films as the FM layers in MTJ devices. For switching analysis, the simulation was performed by the Matlab based micromagnetic code M 3 [36]. Based on the considered magnetic and physical properties, the values of and were calculated to be and 1.11 ns, respectively. To consider the structure size affecting the critical current density, the value of is directly proportional to the FL thickness, as shown in equation (4). This indicates that reduction of the FL thickness in an MTJ device is an important factor to influence the decrement of the switching current density, which is an input for the thermal analysis. In order to estimate the thermal effect with the tilt of the initial magnetization

5 494 Chiang Mai J. Sci. 2015; 42(2) angle in the FM layers of the MgO-based MTJ device during the switching process, the was varied with pulse duration of less than 1 ns. The magnetic degradation due to the temperature increment with disappearance of exchange bias and failure in FL and PL was focused on in this work. 3. RESULTS AND DISCUSSION The profile of temperature in an MTJ nanopillar during CIMS is shown in Figure 2. The most significant temperature occurs inside the MTJ stack because of the high current density in the stack. Figure 3 displays the T of CoFeB-PL being higher than that of CoFeB-FL due to the PL being surrounded by the high temperature areas and the heat accumulation in the layer, which is similar to the results in [23]. Due to the maximum T arising at the end of the current period, the analytical results for maximum increase of the temperature, T max, in the MTJ stack were considered at this position. Furthermore, the results in Figure 4 show T max as a function of nanopillar thickness. It is clear that the highest T max is in the MgO barrier layer because of its lowest electrical conductivity. As mentioned above, the magnetic stability in the MTJ nanopillars is sensitive to the temperature rise caused by Joule heating during read/write operations and the initial magnetic degradations affect the thermal stability factor,, of the FL and the exchange interaction in the AFM/FM bilayers. For this work, the exchange interaction occurs in the PtMn/CoFe bilayers. Thus, it indicates that the magnetic properties in the MTJ stack can be degraded whenever the temperature in PtMn/CoFe bilayers exceeds the T B of 573 K [7, 38] and the of CoFeB-FL is less than 40 [17]. Figure 3. Temperature increase in the CoFeB-PL and the CoFeB-FL during CIMS for τ p of 1 ns with θ 0 of 1. Figure 2. Temperature distribution in the cross-section of the MgO-base MTJ structure during CIMS. Figure 4. Dependence of the maximum temperature increment in MTJ stack on nanopillar thickness for τ p of 1 ns with θ 0 of 1.

6 Chiang Mai J. Sci. 2015; 42(2) 495 To examine the effect of the increased temperature during CIMS on the vanishing of the exchange bias, the T max results are considered at the PtMn/CoFe interface, as defined by T max, PtMn/CoFe. It also is the hottest location in the PtMn layer, as shown in Figure 4. Figure 5 shows the T max, PtMn/CoFe depending on the various θ 0 with τ p of ns. For different τ p values, the T max, PtMn/CoFe decreases with increasing θ 0 due to the increase of the spin torque leading to considerable reduction in critical current. Consequently, the magnetic stability concerning the exchange bias between AFM/FM bilayers in MTJ devices can be improved. Additionally, it is found that the results of the temperature at PtMn/CoFe bilayers for the τ p 0.3 ns with θ 0 1 are less than its T B. Therefore, the magnetic degradation in the bilayers during the switching mechanism does not occur at this condition. Although the temperature at PtMn/CoFe bilayers is over the T B at some θ 0 angles for τ p 0.2, the increasing θ 0 can be used to reduce the risk of damage. Figure 5. The T max, PtMn/CoFe depending on the initial angle θ 0 with different values of τ p. To explore the of the FL impacted by the altered temperature during the magnetization reversal process, the analytical results are reported with the highest temperature increase in the FL, T max,fl. The value of as a function of temperature can be calculated by [26]: K u V = k, (6) B T where K u, V and k B are the magnetic anisotropy, volume of FL and Boltzmann constant, respectively. For the size structure and the material properties considered in this study, the value of K u is J/m 3. The T max,fl depending on the initial angle θ 0 at a τ p of ns is presented in Figure 6. The increasing temperatures in the FL become lower at larger θ 0 for each τ p value. This is because of the reducing J C with increasing θ 0 value. The variance of with altered temperature in the FL for the various θ 0 at a τ p of ns is displayed in Figure 7. It is found that the FL magnetic instability presented with of less than 40 occurs at T max,fl of over 144 K. The results in Figure 6 and 7 also indicate that the magnetic degradation with respect to the thermal stability is significant at a τ p of 0.1 ns with θ 0 15 and a τ p of 0.2 ns with θ 0 3. Moreover, the results show the convergence of the T max of about 0 K at the of for the different, although the approaches the same value at of, as can be seen from Figures 5 and 6. Hence, based on the considered structure, the increase of temperature in the MTJ pillar with the CIMS process at a tilted FL magnetization of over will not affect the storage stability.

7 496 Chiang Mai J. Sci. 2015; 42(2) (7) Figure 6. The results for T max,fl as a function of θ 0 for different values of τ p. Figure 8. Dependence of the maximum temperature, T max, in the free layer and the PtMn/CoFe bilayers on ln 2 (π/2θ 0 )) at τ p of 0.5 ns. Figure 7. Thermal stability factor as a function of θ 0 for different values of τ p. In addition, the variation of temperature increment in an MTJ device is clearly correlated with the initial angle θ 0, as shown in Figure 8. This is reasonable because the temperature increment is proportional to the square of the input current density and, from equation (3), the current density varies with the value of ln(π/2θ 0 )). Figure 9 shows the increment of temperature in the free layer varying in direct proportion to changes in the MTJ resistance, R MTJ. Consequently, estimation of the maximum temperature increment as a function of R MTJ, and is performed by the least squares regression method. Therefore, the altered temperature in the free layer during the switching process can be approximated as equation (7). Figure 9. Dependence of the maximum temperature, T max, in the free layer on the MTJ resistance, R MTJ, at τ p of 0.1 ns with θ 0 of 20. The prediction of the T max,fl and the thermal stability at the different angles for the pulse duration of ns by equation (7) is presented in Figure 10 and Figure 11, respectively. When comparing the results from the simulation model and the calculation in equation (7), the temperature values in the free layer are similar. This indicates that the

8 Chiang Mai J. Sci. 2015; 42(2) 497 thermal analysis during the switching process in an MTJ device can be predicted by calculating the temperature increment. Figure 10. The T max,fl with variance of θ 0 for τ p of ns. Figure 11. Thermal stability factor with variance of θ 0 for τ p of ns. All of this leads to a decrease in temperature of the MTJ nanopillars during the switching process under the tilted magnetization in the FL for ultrafast switching. Thus, the magnetic instability in the device from the temperature factor can be improved. This indicates that, for development of the future MRAM based on an MTJ device, the initial angle of magnetization in the ferromagnetic layers is an interesting factor to estimate the minimum energy for the switching process. 4. CONCLUSIONS The effect of the magnetic instability caused by the temperature increase during the switching process with respect to the tilted magnetization in the FL for ultrafast switching (τ p 1 ns) in MTJ devices was investigated and analyzed by 3D FEM. The results show that the increase of the initial angle between the magnetization vector and the major axis in the FL can reduce the temperature rise in the device. Thus, the thermal stability and the magnetic reliability can be improved by increasing the θ 0 angle. In addition, the temperature increment for the different τ p values is proportional to the square of ln(π/2θ 0 )). Hence, the tilt of the initial magnetization angle in the FM layers of an MTJ cell is the important factor to be considered when developing future STT-MRAM technology. ACKNOWLEDGEMENTS This work is financially supported by The Royal Golden Jubilee Ph.D Program under The Thailand Research Fund (TRF), Grant No. PHD/0049/2551. The authors would like to thank T. Mewes and C.K.A. Mewes for the Matlab based micromagnetic code M 3 progarm. REFERENCES [1] Kawahara T., Ito K., Takemura R. and Ohno H., Spin-transfer torque RAM technology: Review and prospect, Microelectron. Reliab., 2012; 52: DOI /j.microrel [2] Katine J.A. and Fullerton E.E., Device implications of spin-transfer torques, J. Magn. Magn. Mater., 2008; 320: DOI j.jmmm [3] Zabel H., Progress in spintronics, Superlattice Microst., 2009; 46: DOI /j.spmi

9 498 Chiang Mai J. Sci. 2015; 42(2) [4] Berger L., Emission of spin waves by a magnetic multilayer traversed by a current, Phys. Rev. B, 1996; 54: DOI /PhysRevB [5] Slonczewski J.C., Current-driven excitation of magnetic multilayers, J. Magn. Magn. Mater., 1996; 159: L1-L7. DOI / (96) [6] Ralph D.C. and Stiles M.D., Spin transfer torques, J. Magn. Magn. Mater., 2008; 320: DOI /j.jmmm [7] Nogu s J. and Schuller I.K., Exchange bias, J. Magn. Magn. Mater., 1999; 192: DOI /S (98) [8] Lombard L., Gapihan E., Sousa R.C., Dahmane Y., Conraux Y., Portemont C., Ducruet C., Papusoi C., Prejbeanu I.L., Nozi res J.P., Dieny B. and Schuhl A., IrMn and FeMn blocking temperature dependence on heating pulse width, J. Appl. Phys., 2010; 107: 09D728. DOI / [9] Nozi res J.P., Jaren S., Zhang Y.B., Zeltser A., Pentek K. and Speriosu V.S., Blocking temperature distribution and long-term stability of spin-valve structures with Mn-based antiferromagnets, J. Appl. Phys., 2000; 87: DOI / [10] Lee K. and Kang S.H., Design consideration of magnetic tunnel junctions for reliable High-temperature operation of STT-MRAM, IEEE Tran. Magn., 2010; 46(6): DOI /TMAG [11] Surawanitkun C., Kaewrawang A., Siritaratiwat A., Kruesubthaworn A., Sivaratana R., Jutong N., Mewes C.K.A. and Mewes T., Magnetic instability in tunneling magnetoresistive heads due to temperature increase during electrostatic discharge, IEEE Trans. Device Mater. Rel., 2012; 12(3): DOI /TDMR [12] Laosiritaworn Y., Punya A., Ananta S. and Yimnirun R., Mean-field analysis of the Ising hysteresis relaxation time, Chiang Mai J. Sci., 2009; 36(3): [13] Laosiritaworn Y., Ananta S. and Yimnirun R., Monte Carlo investigation of ferromagnetic properties under compressive stress, Chiang Mai J. Sci., 2010; 37(2): [14] Lee D.H. and Lim S.H., Increase of temperature due to Joule heating during current-induced magnetization switching of an MgO-based magnetic tunnel junction, Appl. Phys. Lett., 2008; 92: DOI / [15] NamKoong J.H. and Lim S.H., Temperature increase in nanostructured cells of a magnetic tunnel junction during current-induced magnetization switching, J. Phys. D: Appl. Phys., 2009; 42: DOI / / 42/22/ [16] Aoki T., Ando Y., Watanabe D., Oogane M. and Miyazaki T., Spin transfer switching in the nanosecond regime for CoFeB/MgO/CoFeB ferromagnetic tunnel junctions, J. Appl. Phys., 2008; 103: DOI / [17] Chen E., Apalkov D., Diao Z., Driskill-Smith A., Druist D., Lottis D., Nikitin V., Tang X., Watts S., Wang S., Wolf S.A., Ghosh A. W., Lu J.W., Poon S.J., Stan M., Butler W.H., Gupta S., Mewes C.K.A., Mewes T. and Visscher P.B., Advances and future prospects of spin-transfer torque random access memory, IEEE Tran. Magn., 2010; 46: DOI /TMAG

10 Chiang Mai J. Sci. 2015; 42(2) 499 [18] Nikonov D.E., Bourianoff G.I., Rowlands G. and Krivorotov I.N., Strategies and tolerances of spin transfer torque switching, J. Appl. Phys., 2010; 107: DOI / [19] Zhou Y., Zha C.L., Bonetti S., Persson J. and Akerman J., Spin-torque oscillator with tilted fixed layer magnetization, Appl. Phys. Lett., 2008; 92: DOI / [20] Sun J.Z., Spin-current interaction with a monodomain magnetic body: A model study, Phys. Rev. B, 2000; 62: DOI /PhysRevB [21] Berkov D.V. and Miltat J., Spin-torque driven magnetization dynamics: Micromagnetic modeling, J. Magn. Magn. Mater., 2008; 320: DOI /j.jmmm [22] Lee J.M., Lee C.M., Ye L.X., Su J.P. and Wu T.H., Switching properties for MgO-based magnetic tunnel junction devices driven by spin-transfer torque in the nanosecond regime, IEEE Tran. Magn., 2011; 47(3): DOI /TMAG [23] Ha S.S., Lee K.J. and You C.Y., Effect of the resistance-area product on the temperature increase of nanopillar for spin torque magnetic memory, Curr. Appl. Phys., 2010; 10: DOI /j.cap [24] James D., MRAM puts new spin on process, fab strategy, EE Times-Asia, [25] Leuschner R., Klostermann U. and Ferrant R., U.S. Pat. No (2010). [26] Zhao H., Lyle A., Zhang Y., Amiri P.K., Rowlands G., Zeng Z., Katine J., Jiang H., Galatsis K., Wang K.L., Krivorotov I.N. and Wang J.P., Low writing energy and sub nanosecond spin torque transfer switching of in-plane magnetic tunnel junction for spin torque transfer random access memory, J. Appl. Phys., 2011; 109: 07C720. DOI / [27] Papusoi C., Sousa R., Herault J., Prejbeanu I.L. and Dieny B., Probing fast heating in magnetic tunnel junction structures with exchange bias, New J. Phys., 2008; 10: DOI / /10/10/ [28] Sousa R.C., Prejbeanu I.L., Stanescu D., Rodmacq B., Redon O., Dieny B., Wang J. and Freitas P.P., Tunneling hot spots and heating in magnetic tunnel junctions, J. Appl. Phys., 2004; 95(11): DOI / [29] Lee S.M. and Cahill D.G., Thermal conductivity of sputtered oxide films, Phys. Rev. B,1995; 52(1): DOI /PhysRevB [30] Carey M.J., Childress J.R. and Maat S., U.S. Pat. No (2009). [31] Idris I. and Sugiura O., Film characteristics of low-temperature plasma-enhanced chemical vapor deposition silicon dioxide using tetraisocyanatesilane and oxygen, Jpn. J. Appl. Phys., 1998; 37: DOI /JJAP [32] Luisa F. and Duncan S., Nanotechnologies: Principles, Applications, Implications and Hands-on Activities, European Union, 2012: [33] MatWeb, online materials information resource. [Online]. Available at: Accessed 17 January [34] Wiese N., Dimopoulos T., R hrig M. and Wecker J., Switching of submicronsized, antiferromagnetically coupled CoFeB / Ru/ CoFeB trilayers, J. Appl. Phys., 2005; 98: DOI /

11 500 Chiang Mai J. Sci. 2015; 42(2) [35] Hou Z., Liu Y., Cardoso S., Freitas P.P., Chen H. and Chang C.R., Dynamics of the reference layer driven by spin-transfer torque: Analytical versus simulation model, J. Appl. Phys., 2011; 109: DOI / [36] Mewes C.K.A., Mewes T., Matlab based micromagnetics code M 3, [Online]. Available at: Mcube/Mcube.shtml. Accessed 17 January [37] Surawanitkun C., Kaewrawang A., Imtawil V., Mewes C.K.A., Mewes T. and Siritaratiwat A., Instability of storage and temperature increment in nanopillars due to human body model electrostatic discharge, J. Electrostat., 2011; 69: DOI / j.elstat [38] Dieny B., Sousa R.C., Herault J., Papusoi C., Prenat G., Ebels U., Houssameddine D., Rodmacq B., Auffret S., Prejbeanu-Buda L., Cyrille M.C., Delaet B., Redon O., Ducruet C., Nozieres J.P. and Prejbeanu L., Spintronic Devices for Memory and Logic Applications; in Buschow K.H.J., ed., Handbook of Magnetic Materials, Elsevier, North-Holland, 2011: