Observation of the Planar Nernst Effect in Permalloy and Nickel Thin Films with In-Plane Thermal Gradients

We present experimental evidence of a transverse thermopower, or planar Nernst effect, in ferromagnetic metal thin films driven by thermal gradients applied in the plane of the films. Samples of 20 nm thick Ni and ${\mathrm{Ni}}_{80}{\mathrm{Fe}}_{20}$ were deposited on 500 nm thick suspended Si-N thermal isolation platforms with integrated platinum strips designed originally to allow measurement of thermally generated spin currents (the spin Seebeck effect). The low thermal conductivity of the thin supporting Si-N structure results in an essentially 2D geometry that approaches the zero substrate limit, dramatically reducing the contribution of thermal gradients perpendicular to the sample plane typically found in similar experiments on bulk substrates. The voltage on the platinum strips generated transverse to the applied thermal gradient ($V_T$) is linear with increasing $△T$ and exhibits a sign reversal on hot and cold sides of the sample. However, $V_T$ is always even in applied magnetic field and shows a $\sin \theta \cos \theta$ angular dependence, both key indicators of the planar Nernst effect. Within the 5 nV estimated error of our experiment there is no evidence of a signal from the spin Seebeck effect, which would have $\cos \theta$ angular dependence, suggesting a reduced spin Seebeck coefficient in a planar, entirely thin-film geometry.

Figure 1

(a) Schematic of the thermal isolation platform with $\nabla T$ and $H$ applied in the negative $x$ direction, (b) SEM micrograph of the platform, and (c) plot of thermal conductance vs $T$ for a Si-N bridge and a 20 nm Ni-Fe film deposited on a Si-N bridge. (d) SEM micrograph of one island shows a triangle-shaped lead for measuring film resistance and thermopower, false color outlines show a 20 nm thick Ni-Fe film and a 10 nm thick Pt spin detector.

Figure 2

(a) Example raw data for $V_T$ across a Pt detector. (b) Averaged data when hot (${\nabla }_{x}T=62.5\mathrm{K}/\mathrm{mm}$, and (c) when both sides of the film heated equally leaving only the background thermovoltage generated by $\nabla T_y$. (d) Hot and cold measurements (${\nabla }_x=\pm 62.5\mathrm{K}/\mathrm{mm}$) after subtracting the $\nabla T_y$ contribution. For all measurements, the heater is cycled on at 6 seconds and off at 11 seconds.

Figure 3

$V_T$ versus $\Delta T$ for Ni-Fe ($H=100\mathrm{Oe}$, $\theta =0$). (a) $V_T$ for one detector when hot and cold with a linear fit. Additional points at $\Delta T=50\mathrm{K}$ are for the same detector on four additional Ni-Fe films. The same trends occur for the other Pt detector (not shown). (b) $V_T$ versus $H$ (applied in the $\pm \widehat{x}$ direction) for one detector on a Ni-Fe film when hot and cold for up and down $H$ sweeps.

Figure 4

(a) AMR data for a Ni-a. (b) $V_T$ versus $H$ for one detector on Ni-b when hot and cold for up and down $H$ sweeps. (c) $V_T$ versus $H$ for Ni-b when cold for one sweep direction at various $\theta$ (the angle between $M$ and $\nabla T$) and (d) when hot. (e) $V_T$ vs $\theta$ for Ni-b when the detector is hot and cold with $\sin \theta \cos \theta$ fit. Data when hot appear reversed due to the sign change introduced by reversing $\nabla T$ Insets: Schematics show the sign change for $V_T$ when $\nabla T$ is reversed.