Spin Accumulation based deep MOKE Microscopy

Abstract

Magnetic imaging techniques are widespread critical tools used in fields such as magnetism, spintronics or even superconductivity. Among them, one of the most versatile methods is the magneto-optical Kerr effect. However, as soon as light is blocked from interacting with the magnetic layer, such as in deeply buried layers, optical techniques become ineffective. In this work, we present a spin-accumulation based magneto-optical Kerr effect (SA-MOKE) microscopy technique that enables imaging of a magnetic thin-films covered by thick and opaque metallic layers. The technique is based on the generation and detection of transient spin-accumulations that propagate through the thick metallic layer. These spin-accumulation signals are directly triggered and detected optically on the same side, lifting any substrate transparency requirements. The spin-accumulation signals detected on a Cu layer decay with a characteristic length of 60 nm, much longer than the 12 nm optical penetration depth, allowing for detection of magnetic contrast with Cu capping layers up to hundreds of nm. This method should enable magnetic imaging in a wide-range of experiments where the surface of interest is covered by electrodes.

Figures (4)

• Figure 1: Conceptual schematic of SA-MOKE. The pump laser pulse (red wavy arrow) is incident on the Cu side and generates an electronic heat current ($\mathrm{J_{Q}}$) which flows up to the ferromagnetic (FM) layer and induces its demagnetization. The demagnetization in turn generates a spin current ($\mathrm{J_{S}}$) back into the Cu layer that flows to its surface. Finally, the probe laser pulse (blue wavy arrow) measures the magneto-optical Kerr rotation induced by the spin accumulation at the surface of the Cu layer. This signal is proportional to the magnetization of the deeply buried FM layer.
• Figure 2: Time-resolved magneto-optical Kerr responses in the [Co/Ni]/Cu structure in various configurations. a) Demagnetization of the FM layer due to hot electrons from the Cu layer. b) Spin accumulation in the Cu layer due to the optical heating of the FM layer. c) Spin accumulation in the Cu layer due to electronic heating of the FM layer. Schematics on the upper-right show the experimental configurations of pump (red pulse) and probe (blue pulse) in each case. We used incident pump fluences of 0.2 and 0.1 mJ/cm$^2$ for the experimental configurations a-b and c, respectively. The measurements were performed on a Cu-100 nm layer. The time zero is not comparable because the measurements were performed in different experimental conditions. Integration times and/or number of averaged scans are different in panels (a), (b) and (c). The standard error for measurements at each time delay is represented by the size of the data point markers (a-b) or error bars (c).
• Figure 3: Spin accumulation based MOKE imaging of magnetic domains through a thick Cu layer.a) Schematic representation of conventional polar wide-field (WF) MOKE microscopy, measuring the magnetic layer directly through the substrate. Magnetic domain images obtained for Cu thickness of b) 100 nm and c) 140 nm by WF-MOKE microscopy. d) Schematic representation of the SA-MOKE technique to image magnetic domains. Magnetic domain images obtained for Cu thickness of e) 100 nm and f) 140 nm by SA-MOKE. We used pump fluences of 0.1 and 0.2 mJ/cm$^2$ for the scanning TR-MOKE e and f, respectively. All measurements were carried out at room temperature.
• Figure 4: Cu thickness dependence on Kerr rotation and integration time for signal-to-noise ratio of 1.a) The peak of Kerr rotation as a function of Cu thickness for CoNi/Cu structure at fluence of 0.2 mJ/cm$^2$. The peaks of Kerr rotation, associated with demagnetization and spin current, are plotted using red open circles. The red and purple lines are exponential fits for lower and higher Cu thickness. The error bars are the standard error. b) Integration time to reach a signal-to-noise of 1, calculated from the peaks of Kerr rotation as a function of Cu thickness, is plotted using blue solid diamonds. The green and purple dash dotted lines correspond to integration time estimated from the optical and transport exponential fits, respectively.