Electrical detection of spin pumping: dc voltage generated by ferromagnetic resonance at ferromagnet/nonmagnet contact

We describe electrical detection of spin pumping in metallic nanostructures. In the spin pumping effect, a precessing ferromagnet attached to a normal metal acts as a pump of spin-polarized current, giving rise to a spin accumulation. The resulting spin accumulation induces a backflow of spin current into the ferromagnet and generates a dc voltage due to the spin dependent conductivities of the ferromagnet. The magnitude of such voltage is proportional to the spin-relaxation properties of the normal metal. By using platinum as a contact material we observe, in agreement with theory, that the voltage is significantly reduced as compared to the case when aluminum was used. Furthermore, the effects of rectification between the circulating rf currents and the magnetization precession of the ferromagnet are examined. Most significantly, we show that using an improved layout device geometry, these effects can be minimized.

Figure 1

Simplified picture of the spin pumping process. (a) Population of spin-up and spin-down bands in equilibrium. (b) Situation after sudden reversal of the magnetization direction. The arrows denote spin flow from one-spin population to another one. (c) Equilibrium situation again but with magnetization in opposite direction (Adapted from Ref. 25).

Figure 2

The $F/N$ structure in which the resonant precession of the magnetization direction $\mathbf{m}$ pumps a spin current ${\mathit{I}}_{\mathit{s}}^{\text{pump}}$ into $N$. The spin pumping builds up a spin accumulation ${\mathit{\mu }}_{\mathit{s}}^N$ in $N$ that drives a spin current ${\mathit{I}}_{\mathit{s}}^{\text{back}}$ back into the $F$. The component of the ${\mathit{I}}_{\mathit{s}}^{\text{back}}$ parallel to $\mathbf{m}$ can enter into $F$. Since the interface and the bulk conductances of $F$ are spin dependent, this can result in a dc voltage across the interface.

Figure 3

(a) Schematic of the device. On the lower side, through the shorted end of a coplanar strip a current ${\mathit{I}}_{\mathit{rf}}$ generates an rf magnetic field, denote by the arrows. The Py strip in the center produces a dc voltage $\Delta V=V^{+}-V^{-}$. H denotes the static magnetic field applied along the strip. (b) Scanning electron microscope pictures of the central part of the devices.

Figure 4

Schematic of the experimental setup and of the microwave frequency modulation method. (a) A TTL signal at a reference frequency $f_{\text{ref}}$ (17 Hz), generated by a lock-in amplifier (master device), is first fed into a frequency doubler. Then, the TTL at $2\times f_{\text{ref}}$ is fed into a CW microwave generator. At each TTL input, the CW generator provides frequency hopping of the rf current switching between $f_{\text{high}}$ and $f_{\text{low}}$ at $f_{\text{ref}}$. The dc voltages produced by the device are amplified and detected by the lock-in amplifier as a difference $\Delta V=V\left(f_{\text{high}}\right)-V\left(f_{\text{low}}\right)$. (b) At the bottom, the resonant frequency dependence on the static magnetic field is shown. Next, the diagrams of the dc voltage vs static magnetic field corresponding to the resonance at high and low frequencies. On top, the measured difference in dc voltage between the two frequencies, $\Delta V=V\left(f_{\text{high}}\right)-V\left(f_{\text{low}}\right)$ is plotted.

Figure 5

The dc voltage $\Delta V$ generated by a Pt/Py/Al device in response to the rf magnetic field plotted as a function of the static magnetic field. The frequencies of the rf field are as shown. The peaks (dips) correspond to resonance at $f_{\text{high}}\left(f_{\text{low}}\right)$. The data are offset vertically for clarity (Ref. 24).

Figure 6

(a) Gray scale plot of the dc voltage $\Delta V$, measured function of static field for different high (low) frequencies of the rf field from the Pt/Py/Al device (Ref. 41). The dark (light) curves denote resonance at $f_{\text{low}}\left(f_{\text{high}}\right)$. (b) The static magnetic-field dependence of the resonance frequency of the Py strip (dots). The curve is a fit to Eq. (11). (c) The amplitude of the dc voltage from a Al/Py/Al device as a function of the square of the rf current, at 13 GHz and 139 mT (dots). The line shows a linear fit (Ref. 24).

Figure 7

The dc voltage $\Delta V$ generated across the (a) Al/Py/Al and (b) Pt/Py/Pt devices as a function of the static magnetic field. The frequencies of the rf field are as shown (Ref. 24).

Figure 8

Overall distribution of the amplitude of the dc voltages as a function of different contact materials. Different symbols represent different batches of samples. This includes four devices with longitudinal electrode device geometry, indicated by symbol (○) and discussed in Sec. 4B.

Figure 9

The dc voltage generated by (a) Al/Py/Pt and (b) Pt/Py/Pt devices for a device geometry shown in part (d). (c) The comparison of the signals for the two Pt/Py/Al devices, one with longitudinal electrode device geometry (bold line) and other with transverse electrode device geometry. (d) SEM picture of an longitudinal electrode geometry, Pt/Py/Al device.

Figure 10

The dc voltage at $f_{\text{high}}=18\text{GHz}$, $f_{\text{low}}=13\text{GHz}$ for (a) Pt/Py/Al and (b) Pt/Py/Pt devices for different angles between the static field and the long axis of the strip.

Figure 11

To isolate the contribution from different effects we performed: (a) The sum between the voltage at offset angle of $10°$ and $-10°$, $V\left(10°\right)+V\left(-10°\right)$, vs static field for Pt/Py/Al and Pt/Py/Pt devices. The result represents two times contribution from spin pumping effect. (b) The difference between the voltages, $V\left(10°\right)-V\left(-10°\right)$, which represents two times contribution from bulk rectification effect, as explained in the text.