The Spin Dynamics and Magnetic Microscopy Project develops metrology for magnetodynamic effects such as ferromagnetic resonance, switching, and damping. We apply our measurement tools and expertise to materials and structures commonly used in nanomagnetic devices.
Collaborations with industry leaders have led to new understanding of magnetic damping in advanced materials and replication of our measurement tools. We investigate fundamental aspects of spin transfer in materials and structures that offer improved performance in future devices. These studies include spin-orbit torques, spin pumping, spin-Hall effect at ferromagnet/nonmagnet interfaces, and graphene interfaces, all of which play a role in proposed magnetic memory and logic devices with low energy consumption.
The magnetic damping parameter in magnetic materials is a measure of how quickly energy dissipates in magnetic excitations. The development of materials with ultra-low damping is critical for the reduction of the power required to switch magnetic bits in future nonvolatile magnetic random-access memory. Such magnetic memory is expected to play a vital role in the future growth of the “Internet of Things,” where the ubiquity of extremely low power processors in virtually everything will usher-in a new era of consumer electronics.
We have discovered that the electron band structure at the Fermi energy controls damping in metallic and half-metallic alloy systems. This insight led us to the discovery of ultra-low damping in a simple cobalt-iron binary alloy. The damping parameter for the alloy is almost an order of magnitude lower than the smallest reported value for any other metal. The discovery was facilitated by the use of a ferromagnetic resonance spectrometer with world-record sensitivity that we developed. Prior to this result, it was generally believed that ultra-low damping was not possible for conductors because of coupling between magnetic spin waves (magnons) and the momentum of conduction electrons. Such spin-momentum coupling tends to draw energy more readily out of the magnetization. However, along with collaborators at Uppsala University in Sweden, we found that the newly discovered alloy has a particular band-structure feature that leads to ultra-low damping in spite of the presence of conduction electrons. The identification of this key feature in the band-structure provides a framework for the future discovery of new alloys and compounds with even lower values of damping.
Along with collaborators at the University of Colorado/JILA and Uppsala University in Sweden, we use coherent, high-harmonic light sources to probe the dynamics in magnetic materials at time scales that span the femtosecond to the nanosecond. Recently, we demonstrated that spectroscopic analysis of time-resolved magneto-optic reflected light can be used to test fundamental theories of ferromagnetism in metals under extreme conditions of ultrafast, optically driven demagnetization. These measurements will help inform researchers as to the ultimate scalability of magnetic memory as a replacement for dynamic random-access memory (DRAM) in computers.
We used a coherent source of ultrafast, extreme-ultraviolet (EUV) light based on high-harmonic generation (HHG). The EUV pulses were diffracted from lithographically patterned stripes of cobalt (see figure) and the magnetization-dependent spectrum of reflected EUV light was monitored as it changed over time. We are able to probe how the optical “pump” excitation affects the electrons near the Fermi level and hence the magnetization. In this experiment, we discovered about half of the resultant spin disorder could be explained in terms of optical modification of the electron band structure, as expected from Stoner electron-hole-pair theory, but the other half stems from the ultrafast formation of magnons (spin waves), as would be expected from quantum-mechanical Heisenberg theory.
Future computer hard disk drives are likely to require patterned media composed of uniform, perpendicularly magnetized nanodots instead of present-day continuous magnetic films. Similarly, spin-transfer torque magnetic random-access memory (STT-MRAM) utilizes the magnetization state of a "free layer" nanodot to encode a digital bit. The damping of gyromagnetic precession is a critical figure of merit for the performance of both patterned media and STT-MRAM. Determination of how the damping scales with nanomagnet size is essential to predict future device performance. We have developed a new, highly sensitive, magneto-optic instrument to measure the dynamics of magnetic nanodots as a function of excitation frequency, with the goal of evaluating nanodot quality and homogeneity via a rapid spectroscopic analysis of individual nanodots. The same measurement techniques can be applied to characterize the dynamics of other nanoscale magnetic devices, such as the "free layer" in spin torque oscillators.
We determined that the damping for in-plane oriented nanomagnets in the 100-400 nm size range depends strongly on the particular spin-dynamics eigenmode that is excited: the more uniform the eigenmode, the lower the damping. The measurements were accomplished in 2013 with a unique, NIST-built, optical microscope system that utilizes heterodyne mixing of continuous-wave laser beams to both generate microwaves to excite spin dynamics and to detect the resultant dynamics via the magneto-optic Kerr effect. Since then, we have repeated our measurements, but now with nanostructures patterned from films ranging in thickness from 3 to 15 nm. Here, the mode-dependent damping is strongly dependent on film thickness, with the strongest effect in the thinnest films. These results suggest that the fundamental mechanism that causes the mode-dependent damping is driven by spin relaxation at the ferromagnetic layer interfaces. These results have significant implications for the write-current requirements for STT-MRAM. Specifically, it appears that mode-uniformity is an important figure of merit for minimization of the write current.
Spin-orbit torques at ferromagnet/normal metal (FM/NM) interfaces promise to be much more efficient as a means of switching magnetic memory elements than spin-polarized charge currents. The reason is intrinsic to the geometry employed for spin-orbit torques: an electric field applied parallel to the FM/NM interface results in a flow of angular momentum across the interface. Hence, even if the NM layer is extremely thin, the flux of angular momentum remains constant. Since the ohmic heating scales as the square of the charge current, this becomes a powerful means of minimizing the energy required to switch a FM memory element.
We successfully measured spin-orbit torques in FM/NM bilayers, where the FM is Permalloy (Ni80Fe20) and the NM layers are Pt, Cu/Pt, Pd, Ta, Nb, and Cu/Au. We specifically measured the inverse forms of the spin-Hall effect (iSHE) and the Rashba-Edelstein effect (iREE) in these systems, where microwave stimulation of the FM layer gives rise to an ac voltage in the NM layer.
With the iSHE, the spin chemical potential, generated by spin precession at the FM/NM interface, is converted to a charge current. On the other hand, the iREE directly converts the spin precession into a charge current, and with a phase that is shifted by 90 degrees relative to that of the iSHE.
Microwave magnetic fields were applied with a proximate coplanar waveguide assembly, and the generated electrical signals in the multilayer samples were extracted by use of an electrically isolated, terminated waveguide structure. Our measurement method employs phase-sensitive detection to separate the two spin-orbit contributions to the measured microwave signal. Furthermore, a calibration reference component was included in the device design to permit separation of the spin-orbit torque contributions from any inductive signals.
Our measurement results indicate that the Rashba-Edelstein contribution to the net spin-orbit torque is comparable to the spin-Hall contribution. These results will have significant implications for the utility of employing spin-orbit torques in three-terminal magnetic memory designs. Such designs have been proposed as a means of greatly improving the switching efficiency of spin-transfer torque magnetic random-access memory, a promising nonvolatile alternative to conventional CMOS cache memory in system-on-chip applications.
Thermal fluctuations are the main limitation to the scalability to nonvolatile magnetic memory. To prevent thermal erasure of a bit in a magnetic memory element, it is imperative that the magnetization be in a uniform "single-domain" state. However, a novel physical mechanism at play in certain types of FM/NM multilayers might compromise the uniformity of magnetization: the Dzyaloshinskii-Moriya interaction (DMI). The DMI favors chiral magnetization orientation rather than a uniform state, and a sufficiently large DMI is expected to greatly reduce the thermal stability of magnetic memory. Hence, a quantitative measurement technique is required to characterize the DMI at different FM/NM interfaces to ensure that the particular FM/NM combination does not result in an excessively large DMI.
With the DMI, the usual exchange interaction is modified to favor chiral spin orientation rather than either uniform (ferromagnetic) or antiparallel (antiferromagnetic) configurations. The DMI requires a combination of spatial symmetry breaking and spin-orbit coupling. These requirements are satisfied at ferromagnet/normal metal interfaces, such as that between Permalloy and Pt. We demonstrated the ability to accurately measure the DMI at the Permalloy/Pt interface by use of Brillouin light scattering (BLS). With BLS, we use inelastic photon scattering to measure the asymmetry in the dispersion of spin waves propagating perpendicular to the Permalloy magnetization. Spin waves have an inherent chirality associated with the precession of the spin orientation. For the case where the spin wave chirality is favored by the DMI, the waves propagate faster.
It had been theoretically predicted that BLS could be used to measure the DMI, but the effect is rather subtle, and the measurement required several stages of data processing and instrument calibration to avoid measurement artifacts. Nevertheless, we were successful in measuring the DMI for films with varying Permalloy thickness, and we were able to correlate the DMI with the exchange energy that gives rise to ferromagnetic order in the bulk of the Permalloy film. A direct proportionality between exchange and the DMI was originally predicted by Moriya, but had never been confirmed until now. This result is of immediate technological relevance for those working to develop STT-MRAM as a nonvolatile alternative to DRAM.
For optimal thermal stability, it is required that the magnetization within the data storage layer be entirely uniform. However, a sufficiently strong DMI can cant the magnetization at the device edges, making it much more susceptible to thermal fluctuations that compromise data integrity.