TSTiffany S Santos

Tiffany S Santos

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Tiffany S Santos has a total of 75 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    The 70th annual conference on magnetism and magnetic materials

    by Tiffany S. Santos

  • Achieving low write error rates in perpendicular spin–orbit torque MRAM: A micromagnetic modeling feasibility study

    by Goran Mihajlović, Yabin Fan, Tiffany S. Santos, Michael K. Grobis, Neil Robertson

    We report a micromagnetic modeling study of switching perpendicularly magnetized free layers with spin–orbit torques (SOTs) generated by spin currents consisting of only orthogonal (conventional SOT) as well as both orthogonal and collinear (unconventional SOT) spin polarizations relative to the free layer (FL) magnetization. Focusing on the FL diameter of 50 nm, for conventional SOT, we find relatively high probabilities of FL back-hopping or incomplete switching at room temperature (T) for several bias configurations that deterministically switch the FL at T = 0 K, including in-plane bias field with and without Dzyaloshinskii–Moriya interaction (DMI) as well as the field-free case with DMI in the presence of field-like torque. For unconventional SOT, fast, field-free deterministic switching is found at T = 0 K for orthogonal y-spin SOT efficiency θy = 0.3 when collinear z-spin efficiency θz/θy > 0.25. Such large ratio is needed to change the nature of the magnetization switching from a non-uniform state with dominant component along y to an almost fully reversed and uniform state along -z, resulting in thermally robust and complete switching without back-hopping. In addition, the presence of z spins of such magnitude results in lowering of the switching current density by a factor of about 3, depending on the damping parameter α of the FL (more for lower α). Based on these results, we conclude that achieving low write error rate operation in perpendicular SOT MRAM is not feasible using approaches that exploit conventional SOT, but it should be straightforward to achieve with unconventional SOTs that meet parameter values determined by this study.

  • Open Access

    Leveraging a Passive MRAM Crossbar for Hardware-in-the-Loop and Continual learning

    by M. Drouhin, M. Grelier, M. Deroo, Y. Sassi, E. Monteblanco, N. Davila, G. Mihajlović, R. V. Chopdekar, T. S. Santos, J. A. Katine, et al.

    Artificial neural networks have enabled major advances in artificial intelligence, yet their growing computational and energy demands challenge conventional von Neumann architectures due to the costly separation of memory and processing. In-memory computing has emerged as a promising solution, particularly through memristive crossbar arrays capable of performing multiply-and-accumulate operations directly within memory. Among non-volatile memory technologies, magnetic tunnel junctions, the core elements of MRAM, offer high endurance, non-volatility, and CMOS compatibility, making them attractive candidates for neuromorphic systems. In this work, we report inference results on a 15 × 15 MRAM crossbar, benchmarking experimental measurements against numerical simulations. We further demonstrate hardware-in-the-loop inference and experimentally validate metaplasticity mechanisms enabling continual learning on simple tasks. These results highlight the potential of MRAM-based in-memory architectures for adaptive, energy-efficient, and scalable AI hardware.

  • Characterization of thermal stability of perpendicular spin transfer torque magnetoresistive random access memory cells with free layer diameter of around 20 nm

    by Goran Mihajlović, Wonjoon Jung, Noraica Dávila, Hans J Richter, Michael Tran, Tiffany S Santos, Jordan A Katine, Michael K Grobis

    We report an experimental evaluation of the thermal stability factor, Δ, for magnetoresistive random access memory bits with perpendicular magnetic anisotropy and a magnetic free layer diameter of about 20 nm, using magnetic field and spin transfer torque (STT) excitations to accelerate switching. We compare the results to the values obtained by thermal-only excitation. We find that fitting the magnetic switching field distributions to a macrospin model results in identical values of Δ mean and sigma to those determined by direct thermal excitation, also known as the retention bake method. On the other hand, fitting the pulse width dependence of the STT switching voltages in the thermal activation regime can either underestimate or overestimate Δ depending on the choice of exponent ξ that defines the scaling of Δ with STT. We find the best agreement for ξ = 1.45.

  • Open Access

    Enhanced write margin of perpendicular MRAM cells using thick MgO cap layer

    by G. Mihajlović, T. S. Santos, J. Li, J. A. Katine, M. K. Grobis

    Implementation of spin transfer torque magneto-resistive random access memory (STT-MRAM) in memory chips requires that the write margin of the MRAM cell, defined as the difference between breakdown voltage and write voltage for the specified endurance, write error rate, and write speed, is sufficiently large in order to accommodate resistance variations arising from external chip circuitry. We show that by increasing only the thickness of the MgO cap layer to make its resistance-area product close to that of the main MgO barrier, the write margin can be increased substantially without affecting the thermal stability of the cell.

  • Open Access

    Quantifying the topology of magnetic skyrmions in three dimensions

    by David Raftrey, Simone Finizio, Rajesh V. Chopdekar, Scott Dhuey, Temuujin Bayaraa, Paul Ashby, Jörg Raabe, Tiffany Santos, Sinéad Griffin, Peter Fischer

    Magnetic skyrmions have so far been treated as two-dimensional spin structures characterized by a topological winding number. However, in real systems with the finite thickness of the device material being larger than the magnetic exchange length, the skyrmion spin texture extends into the third dimension and cannot be assumed as homogeneous. Using soft x-ray laminography, we reconstruct with about 20-nanometer spatial (voxel) size the full three-dimensional spin texture of a skyrmion in an 800-nanometer-diameter and 95-nanometer-thin disk patterned into a 30× [iridium/cobalt/platinum] multilayered film. A quantitative analysis finds that the evolution of the radial profile of the topological skyrmion number is nonuniform across the thickness of the disk. Estimates of the micromagnetic energy densities suggest that the changes in topological profile are related to nonuniform competing energetic interactions. Our results provide a foundation for nanoscale metrology for spintronics devices using topology as a design parameter.


  • Open Access

    Thermal effects on damping determination of perpendicular MRAM devices by spin-torque ferromagnetic resonance

    by H. J. Richter, G. Mihajlović, R. V. Chopdekar, W. Jung, J. Gibbons, N. D. Melendez, M. K. Grobis, T. S. Santos

    We report device-level damping measurements using spin-torque driven ferromagnetic resonance on perpendicular magnetic random-access memory cells. It is shown that thermal agitation enhances the apparent damping for cells smaller than about 55 nm. The effect is fundamental and does not reflect a true damping increase. In addition to the thermal effect, it is still found that device-level damping is higher than film-level damping and increases with decreasing cell size. This is attributed to edge damage caused by device patterning.


  • Voltage Controlled Interlayer Exchange Coupling and Magnetic Anisotropy Effects in Perpendicular Magnetic Heterostructures

    by Akash Surampalli, Anup Kumar Bera, Rajesh Vilas Chopdekar, Alan Kalitsov, Lei Wan, Jordan Katine, Derek Stewart, Tiffany Santos, Bhagwati Prasad

    Spintronic devices that utilize spin transfer torque are promising for integrated memory applications. However, these devices face substantial energy consumption challenges due to the high current densities required for switching. Conversely, voltage‐driven spintronic devices, using capacitive displacement charge, can realize switching operations that are energy‐efficient (≈1–10 fJ bit−1). This work investigates switching based on voltage control of the interlayer exchange coupling in perpendicular magnetic anisotropy (PMA) multilayered heterostructures. Unlike previous works that utilized gating techniques that employ ionic transport mechanisms to control interlayer exchange coupling, this study employs electrostatic gating by using MgO, which is more compatible with modern spintronic‐based memories. These results suggest that the magnetization anisotropy, and the magnitude, and phase of the Ruderman‐Kittel‐Kasuya‐Yosida (RKKY) coupling function with spacer layer thickness can be controlled through electric gating, providing a promising avenue for the development of energy‐efficient magnetic data storage devices.


  • Measurement-driven neural-network training for integrated magnetic tunnel junction arrays

    by William A. Borders, Advait Madhavan, Matthew W. Daniels, Vasileia Georgiou, Martin Lueker-Boden, Tiffany S. Santos, Patrick M. Braganca, Mark D. Stiles, Jabez J. McClelland, Brian D. Hoskins

    The increasing scale of neural networks needed to support more complex applications has led to an increasing requirement for area- and energy-efficient hardware. One route to meeting the budget for these applications is to circumvent the von Neumann bottleneck by performing computation in or near memory. However, an inevitability of transferring neural networks onto hardware is the fact that nonidealities, such as device-to-device variations or poor device yield impact performance. Methods, such as hardware-aware training, where substrate nonidealities are incorporated during network training, are one way to recover performance at the cost of solution generality. In this work, we demonstrate inference on hardware-based neural networks consisting of 20 000 magnetic tunnel junction (MTJ) arrays integrated on CMOS chips in a form that closely resembles scalable and market-ready spin transfer-torque magnetoresistive random access memory (STT-MRAM) technology. Using 36 dies, each containing a MTJ-CMOS crossbar array with its own nonidealities, we show that even a small number of defects in physically mapped networks significantly degrades the performance of networks trained without defects and show that, at the cost of generality, hardware-aware training accounting for specific defects on each die can recover to comparable performance with ideal networks. We then demonstrate a robust training method that extends hardware-aware training to statistics-aware training, producing network weights that perform well on most defective dies regardless of their specific defect locations. When evaluated on the 36 physical dies, statistics-aware trained solutions can achieve a mean misclassification error on the MNIST dataset that differs from the software-baseline by only 2%. This statistics-aware training method could be generalized to networks with many layers that are mapped to hardware suited for industry-ready applications.


  • Open Access

    Intermixing of iron and cobalt with oxygen-rich magnesium oxide in CoFeB/MgO/CoFeB magnetic tunneling junctions

    by Michael A. Gribelyuk, Alan Kalitsov, Wonjoon Jung, Yudi Wang, Michael Tran, Xiaoyu Xu, Tiffany Santos

    Atomic-scale spectroscopic imaging of sputtered magnetic tunnel junction structures with a thick oxygen-rich MgO barrier reveals the diffusion of iron and cobalt into the MgO barrier from CoFeB electrodes. First principles calculations are performed to (1) confirm that Fe diffusion through Mg vacancies is energetically favorable, (2) quantify the reduction of interfacial perpendicular magnetic anisotropy due to Fe diffusion into MgO, and (3) predict that the presence of Fe impurities in MgO causes an increased leakage and a tunneling magnetoresistance decrease. Through the chemical shift of the Fe L3 edge and the peak ratio Fe L3/Fe L2 measured by electron energy loss spectroscopy, we suggest that, within MgO, iron with mixed oxidation state Fe2+ and Fe3+ or higher is found in the as-grown structure, which is reduced by annealing to Fe2+. These results indicate that the stoichiometry of as-deposited MgO barrier layers plays an important role in controlling the microstructure and optimizing the performance of magnetic tunnel junctions.