Spin-independent transparency of pure spin current at normal/ferromagnetic metal interface
School of Physics and State Key Laboratory of Crystal Materials, Shandong University, Jinan 250100, China
† Corresponding author. E-mail:
skang@sdu.edu.cn
1. IntroductionIn recent years, a great deal of attention has been focused on the generation, transfer, and detection of pure spin current, which is fundamentally important and central to the development of the next generation of spintronics devices.[1,2] The interfacial spin transparency plays an important role on the transfer of spin current, and its research has attracted a lot of attention since the discovery of giant magnetoresistance (GMR).[3,4] In conventional methods of spin injection, such as spin pumping and the spin Seebeck effect (SSE), the polarization of spin current (
) is always parallel to the magnetization of the ferromagnetic mental (
).[5,6] In the spin valve structure,[7] the nonlocal voltage induced by spin-dependent electrochemical potentials due to interfacial spin accumulation can be observed. However, it is hard to directly observe the influence of the spin accumulation on the interfacial spin-dependent transparency at the interface of normal/ferromagnetic metal (NM/FM). As for the spin Hall magnetoresistance (SMR) measurement,[8,9] although the variation of local resistance induced by the inverse spin Hall effect (ISHE) related to the spin accumulation at the interface can be obtained with
parallel and antiparallel to
, there is no information about spin-dependent transparency. Recently, Avci et al.[10] found unidirectional spin Hall magnetoresistance via second-harmonic resistance measurement, and they claimed that this was induced by the difference of spin-dependent electrochemical potential due to spin accumulation at the interface. However, the quantitative influence of this spin accumulation on the interfacial spin transparency is still unclear. Generally, it is difficult to obtain the interfacial spin-dependent transparency of pure spin current with the conventional methods mentioned above. In the spin-valve-like trilayer structure with two thin Pt layers sandwiched by a thick yttrium iron garnet (YIG-Y3Fe5O12) layer,
is no long dependent on the applied magnetic field. If we replace the detector Pt layer with an FM layer, exploring the spin-dependent transparency of a pure spin current at the interface of NM/FM becomes possible.
In this paper, we study the transparency of a pure spin current at the Cu/FM interface in Pt/YIG/Cu/FM multilayers. Our nonlocal measurement is illustrated in Fig. 1(a). Via the spin Hall effect, an injected current (
in the Pt layer generates a spin accumulation at the interface of Pt/YIG with
aligned with the
axis. The s–d exchange interaction at the interface creates a non-equilibrium magnon population and spin accumulation in the YIG layer which drives magnon diffusion. The magnons are then converted to a spin current in the Cu layer by the reverse process. This spin current is further converted to a charge current again via the inverse spin Hall effect after flowing into the FM layer. For
along the −y axis in the Pt layer,
can be parallel or antiparallel to
with α = 0° or 180°, respectively, and vice versa for
along the +y axis. By comparing the nonlocal voltages in the FM layer with
parallel and antiparallel to
, we can obtain some information about spin-dependent transparency at the Cu/FM interface.
2. ExperimentFirst, (111)-oriented Gd3Ga5O12(GGG)/YIG (20) bilayers, Pt (5)/YIG (120) bilayer templates and Pt (5)/SiO2 (5)/YIG (120) trilayer templates (the numbers are the layer thicknesses in nm and below is the same) were deposited on naturally oxidized Si substrates. These templates were annealed at 800 °C in a tube furnace flowing oxygen gas for 1 hour. Finally, a series of Cu, Ni, and Fe20Ni80 (Py) were deposited on these annealed templates as follows:
(i) Pt (5)/YIG (120)/Cu (3)/Ni (10),
(ii) Pt (5)/YIG (120)/Cu (3)/Py (10),
(iii) Pt (5)/YIG (120)/Cu (10),
(iv) Pt (5)/YIG (120)/Py (20),
(v) Pt (5)/YIG (120)/Ni (20),
(vi) Pt (5)/SiO2 (5)/YIG (120)/Cu (3)/Ni (10).
All films were prepared with a magnetron sputtering system at room temperature, and the base pressure was better than 3×10−6 Pa. Shadow masks were used to pattern the films into devices. The width and length are 0.8 mm and 5 mm, respectively. To check the leakage current in the Pt/YIG/Cu/Ni sample, the charge current was applied between the Pt and Ni directly and the resistance was calculated to be larger than 30 MΩ between the Pt and Ni. The ferromagnetic resonance linewidth of the YIG in the SiO2/Pt/YIG is 87.6 Oe, which is much larger than that in the (111)-oriented GGG/YIG (8.3 Oe) (Fig. 1(b)). This is expected because the crystal quality of the YIG epitaxially grown on the (111)-oriented GGG single crystal is usually much better. Therefore, the YIG on SiO2/Pt can block charge current and transmit spin current via the magnon current. During different nonlocal measurements, different
were injected in the Pt and FM layer for Pt/YIG/Cu/FM and Pt/YIG/FM, respectively, and we applied a magnetic field of 7 kOe which was large enough to saturate
and
of the magnetic hysteresis loops measured by an alternating gradient magnetometer (AGM) (see Fig. 1(c)).
3. Results and discussionFigure 2(b) presents nonlocal voltage
in Pt/YIG/Cu/Ni with
as a function of α, β, and γ. It is clear that, for
along the −y axis,
exhibits maxima at α, β = 0° and 180° (
collinear with
, and minima at α, β = 90°, 270° and for the applied field along the γ direction (
perpendicular to
. For the Pt/YIG/Cu/Ni samples, the angular dependence of
fits excellently with the data. Figure 2(c) presents nonlocal voltage
at α =180° in the Ni layer for Pt/YIG/Cu/Ni with injected currents
, −4 mA, 0, +4 mA, and +8 mA in the Pt layer. As illustrated in Fig. 2(b), the magnitude of the nonlocal voltage almost scales linearly with the injected current, that is,
, which is consistent with the spin accumulation density induced by the SHE at the Pt/YIG interface. Once
is reversed, the measured
changes the sign. This is because the sign of
changes when
is reversed. The angular dependence of the nonlocal voltages in the Ni layer for Pt/YIG/Cu/Ni is consistent with that of the nonlocal voltages in the Pt layer for Pt/YIG/Pt.[11,12]
Compared to the
in Pt/YIG/Cu/Ni, there is no measureable
with
in Pt/YIG/Cu and Pt/SiO2/YIG/Cu/Ni samples as shown in Fig. 3(a). Obviously, the large ISHE voltage is not from the Cu layer in the Pt/YIG/Cu/Ni sample. In order to identify the possible influence of thermal gradients and related thermoelectric effects, a 5-nm SiO2 layer was inserted at the interface of Pt/YIG in the Pt/YIG/Cu/Ni sample to block the spin current induced by the spin accumulation at the Pt/YIG interface. Nevertheless, both SSE and ANE in our samples can be ruled out, since there is a negligible voltage obtained with the insertion of SiO2. Furthermore, the thermal voltages induced by anomalous Nernst effect (ANE)[13] and spin Seebeck effect (SSE) are proportional to
, and exhibit different angular dependence of
.[14] Hence, the nonlocal voltage is almost induced by the ISHE in the Ni layer where the spin current is converted from the magnons current in YIG due to the spin accumulation at the Pt/YIG interface.
More interestingly, for the angular dependence of
in the Pt/YIG/Cu/Ni sample illustrated in Fig. 2, there is almost no difference for
parallel and antiparallel to
. When we pass a current in FM and measure the
in NM for Pt/YIG/FM (see Fig. 3(b)), we can clearly see the asymmetry with a maximum value when
is antiparallel to
. The magnitude of the nonlocal voltage scales linearly with the spin accumulation density at the YIG/FM interface, and therefore the interfacial spin accumulation is different for α = 0° and 180°. This is because the spin electrons are spin-polarized in FM. When we pass a current in Pt which is not spin-polarized, the spin accumulation at the Pt/YIG interface induced by SHE is the same for α = 0° and 180°. Pure spin currents will be induced in the Cu layer via the magnon current in YIG, and flow across the Cu/FM interface into the FM layer. Unlike NM metal, there should be a strong spin-dependent transparency at the Cu/Ni interface due to the spin polarization in the FM layer. At the Cu/FM interface, there will be a spin accumulation whose spin polarization is consistent with that of the spin current. With the spin accumulation at the interface, a conductivity mismatch of majority and minority electrons in the two materials will be induced. With
parallel to
, the transparency of majority or minority electrons at the interface is enhanced or reduced, respectively, and vice versa with
antiparallel to
. Therefore, different interfacial spin transparency of spin current should be obtained with
parallel and antiparallel to
.
However, our experimental results show that almost the same nonlocal voltage is obtained for
parallel and antiparallel to
, indicating a spin-independent transparency of pure spin current at the NM/FM interface and flowing into the FM layer. This means that the transparency of majority or minority electrons at the Cu/FM interface is almost the same for
parallel and antiparallel to
. Furthermore, our results demonstrate that the influence of the conductivity mismatch induced by spin accumulation at the interface on the interfacial spin transparency observed in an early experiment[10] may be very weak and could be negligible. Therefore, there is a spin-independent transparency of spin current at the Cu/FM interface.
We should mention that in the Pt/YIG/Cu/FM samples, the intervening Cu layer provides a shunting effect on the voltage while ISHE occurs in the FM layer. If V0 and Vnl stand for the ISHE voltage generated in the FM layer and the measured voltage, respectively, we will have a relationship
.[15] RCu and RFM are the resistances of Cu and FM layer, respectively. Since the resistance of 3-nm Cu layer is much larger than that of 10-nm FM layer,[16] the Vnl is almost equal to the V generated in the FM layer.
In general the Vnl in Pt/YIG/Cu/FM samples can be expressed by
a product of three terms. The first term includes the spin injection coefficient
A, the sample dimension
L, the injected current density
in the Pt layer, and the angular dependence
, where
φ is the angle between the applied magnetic field and
. The second term includes the electrical resistivity
ρ and the spin Hall angle of FM. The third term describes the decay of the pure spin current due to spin diffusion length
λsf and layer thickness
t of the FM layer. We assume that the spin current injection efficiency remains the same among samples, and there is no significant decay of spin current in the Cu layer. From the
in Fig.
4, with
λsf = 3.2 nm and 1.7 nm for Ni and Py, respectively,
[5] we can obtain
, indicating a relative larger one for Py.
4. ConclusionIn summary, we studied the transparency of pure spin current with
parallel and antiparallel to
at the Cu/FM interface in Pt/YIG/Cu/FM multilayers, and the spin-independent nonlocal voltage was obtained from nonlocal measurement. Our result indicates a spin-independent transparency at the Cu/FM interface, which demonstrates that the influence of spin-dependent electrochemical potential due to spin accumulation on the interfacial spin transparency is very weak and can be negligible. Furthermore, a larger spin Hall angle of Py than that of Ni was obtained by comparing the nonlocal voltages in Py for Pt/YIG/Cu/Py with that in Ni for Pt/YIG/Cu/Ni.