Size and voltage dependence of effective anisotropy in sub-100-nm perpendicular magnetic tunnel junctions

Stephan K. Piotrowski, Mukund Bapna, Samuel D. Oberdick, Sara A. Majetich, Mingen Li, C. L. Chien, Rizvi Ahmed, R. H. Victora
2016 Physical review B  
Magnetic tunnel junctions with perpendicular magnetic anisotropy are investigated using a conductive atomic force microscope. The 1.23 nm Co 40 Fe 40 B 20 recording layer coercivity exhibits a size dependence which suggests single domain behavior for diameters ≤ 100 nm. Focusing on devices with diameters smaller than 100 nm, we determine the effect of voltage and size on the effective device anisotropy K eff using two different techniques. K eff is extracted both from distributions of the
more » ... utions of the switching fields of the recording and reference layers, and from measurement of thermal fluctuations of the recording layer magnetization when a field close to the switching field is applied. The results from both sets of measurements reveal that K eff increases monotonically with decreasing junction diameter, consistent with the size dependence of the demagnetization energy density. We demonstrate that K eff can be controlled with a voltage down to the smallest size measured, 64 nm. PACS numbers: 85.75.-d,73.40.Gk,75.78.-n,75.70.-i 1 I. INTRODUCTION Magnetic tunnel junctions (MTJs) with perpendicular magnetic anisotropy (PMA) are an attractive building block for non-volatile memories. PMA MTJs (p-MTJs) show promise in terms of the key requirements for implementation into products competitive with current data storage and memory technologies: large tunnel magnetoresistance (TMR), low writing energy cost, non-volatility over ∼ 10 years, and scalability of these properties toward ∼ 1 Tbit/inch 2 densities. Room temperature TMR ratios greater than 100% have long existed in in-plane MTJs 1,2 . In state of the art in-plane MTJs, TMR in excess of 600% is achieved by controlling the diffusion of Ta in the film stack through the addition of boron to the magnetic electrodes 3 . Despite these achievements, in-plane MTJs suffer from scalability issues due to their dependence on shape anisotropy for thermal stability and the high energy cost of switching the magnetization by the spin transfer torque (STT) effect 4,5 . For in-plane MTJs, switching energies E sw = I 2 c Rt, where I c is the critical switching current, R is the resistance, and t is the length of the pulse, of approximately 10 µJ/bit were achieved for current pulses on the order of 10 ms 6 . This value was drastically reduced using nanosecond pulses, yielding E sw on the order of single pJ/bit in purely in-plane MTJs 7 . In high TMR p-MTJs the large out-of-plane demagnetization energy does not contribute to E sw 8 . Recently, TMR ratios up to 162% were obtained in p-MTJs by further controlling interlayer diffusion through the substitution of Ta with Mo in the film stack 9 . PMA is achieved when the CoFeB thickness is less than about 1.5 nm, so that the effective anisotropy K eff is dominated by the interfacial anisotropy between Fe in the CoFeB and oxygen in the MgO 10 . In such p-MTJs, switching energies of hundreds of fJ/bit were achieved in 60 nm × 170 nm ellipses 11 . One of the most promising aspects of p-MTJs is that the interface anisotropy can be controlled by applying a bias across the tunnel junction 12-15 . This voltage controlled magnetic anisotropy (VCMA) effect is independent of STT. It can be combined with a magnetic field to switch p-MTJs at current densities on the order of 10 2 A/cm 214 , about 4 orders of magnitude smaller than what is needed for STT switching. As a result, the magnetization state of p-MTJs can be reversed with very low energies. Grèzes et al. achieved switching energies of approximately 6 fJ for junction diameters of about 50 nm by exciting precessional motion using a 600 Oe 2 in-plane field combined with voltage pulses 16 . This is comparable to the energy associated with writing CMOS bits 17 . The scaling behavior of p-MTJs below the single domain (SD) limit is of particular interest. The coercivity is largest at the SD limit and scales with volume rather than domain wall energy below this limit. Although a large range of sizes has been studied in the literature, very few studies address the critical size for SD reversal 18, 19 . In this paper, we present a study of devices with K eff between roughly 1 − 2 × 10 5 erg/cc in the SD regime. We present transport measurements made with a conductive atomic force microscope (C-AFM) on Co 40 Fe 40 B 20 /MgO/Co 40 Fe 40 B 20 MTJs with interfacial PMA patterned with reactive ion etch (RIE) processes. This builds on our previous work, where we have used C-AFM to measure the magnetoresistance of isolated magnetite nanoparticles on an FePt film 20 , in-plane MTJs 21 , and MTJs with interfacial PMA 22,23 . We focus on the effective recording layer anisotropy K eff and the VCMA effect as a function of size in the SD regime. In one set of measurements, we extract K eff by studying the switching field distributions (SFDs) obtained from minor TMR hysteresis loops. In a complimentary set of experiments, we determine K eff from the statistics of random magnetization reversal as a function of time under thermally unstable conditions imposed by applying a magnetic field. These measurements are repeated for different bias voltages to determine K eff (V ). II. METHODS A. Sample Preparation The MTJ film stack was grown on thermally oxidized Si with the structure Si SiO 2 /Ta (7) / Ru(26)/Ta(7)/Co 40 Fe 40 B 20 (0.82)/MgO(2.1)/Co 40 Fe 40 B 20 (1.23)/Ta(10)/Ru(20) where the numbers in parentheses are thicknesses in nanometers. The thin film deposition was carried out as in Ref. 14 . The coupon which was patterned came from a sample where the MgO thickness was varied linearly across the wafer. The thickness of 2.1 nm was chosen to minimize STT effects while maintaining an appreciable TMR. To pattern the film stack, a 30 nm thick Al mask was sputtered on top of the Ru layer and an additional 20 nm SiN x mask was
doi:10.1103/physrevb.94.014404 fatcat:75hhx5avsbechej5huam3i7teq