Role of Cu Layer on the Enhancement of Spin-to-Charge Conversion in Py/Cu/Bi2Se3

The enhancement of spin-to-charge conversion in Py/Cu/Bi 2 Se 3 is achieved when increasing the Cu layer thickness up to 7nm. The conversion rate is studied using spin pumping technique. The inverse IEE length λ IEE is found to increase up to ∼ 2.7nm when 7nm Cu layer is introduced. Interestingly, maximized λ IEE is obtained when the effective spin mixing conductance (and thus J s ) is decreased due to the Cu insertion. Monotonic increase of λ IEE with decreasing J s suggests that IEE relaxation time τ is enhanced due to additional tunnelling barrier (Cu) that limits the interface transmission rate. The results have shown the importance of interface engineering in Py/TI magnetic heterostructure, which is the key factor for optimizing spin-to-charge conversion eciency.


Introduction
Manipulation of spin current by electrical charges or voltages is one of the key subjects for new generation of spintronic devices. Three dimensional (3D) topological insulators (TI) are great candidate for spintronic applications, since their spin-momentum locked properties enable spin accumulation on the surface by passing electrical charge current through the TI channel. 1,2 In addition, the degree of spin polarization can be further manipulated by controlling the Fermi level. 3,4 Owing to the spin momentum locking, 3D spin current density J s injected onto TI surface will produce 2D charge density J c on TI surface states (SS), so-called inverse Edelstein effect (IEE). The inverse IEE length λ IEE is determined as J c /J s and can be experimentally probed using spin pumping technique. 5,6,7,8 Numerous studies have been reported for determining the spin-to-charge conversion e ciency in 3D TI. 5,9,10,11 Particularly, giant spin hall angle (SHA) as large as ∼0.43 was reported in Bi 2 Se 3 that attributed to the enhanced spin current by the surface states and then converted into dc-voltage due to the bulk inverse spin hall effect. 9 However, large variations of SHA was found, which is an order of magnitude difference and the authors ascribed such variation to the non-uniformity of interface quality. 9 On the other hand, Wang et al. observed the dominant role of surface states in spin-to-charge conversion, despite the unavoidable conducting bulk in Bi 2 Se 3 . 5 The effective spin mixing conductance was not increased monotonically although the thickness of Bi 2 Se 3 varied from 2 QL to 60 QL, suggesting surface states dominated mechanism. 5 Obviously, the spin pumping characteristic are important parameters for investigating the spin-to-charge conversion mechanism in 3D TI, where controlling the interfacial properties are necessary steps.
Cu is widely used to control the spin transmissivity in the multilayer devices. 12,13,14 Du et al. demonstrated that the insertion of Cu layer between Y 3 Fe 5 O 12 (YIG) and W substantially improved the spin current injection into W, while similar insertion between YIG and Pt degraded the spin current. 13 The authors reported the quantitative analysis and found that the spin transport e ciency in heterostructures depends on spin conductance of each constituents and their interfaces. 13 Similar results were also reported by Deorani et al. where the effect of Cu interlayer on spin mixing conductance indeed material dependent (Pt versus Ta). 14 Recently Cu layer was deposited on TI lm in order to eliminate the proximity induced ferromagnetism in the spin-orbit torque (SOT) devices. 1,15 Despite the fact that Cu is the most common used spacer layer in the spintronic devices, there is still lacking of quantitative studies about the role of Cu insertion on spin-to-charge conversion in TI that measured based on spin pumping mechanism.
In this work, we fabricated trilayer structure of Py/Cu/Bi 2 Se 3 ( Fig. 1) and studied the spin pumping characteristic by varying the Cu layer thickness. Cu layer is used to protect the TI surface from exchange interaction with Py. Our results imply that Cu also acts as the barrier to spin transmission into TI lm.
More importantly, spin-to-charge conversion e ciency was enhanced due to the introduction of Cu barrier.
The related mechanism is discussed in this work. The resistance of the multilayer samples R d were measured using four probe method. J c is determined as J c = I c /w = V s /wR d , where w and I c are the width of the sample and charge current as shown in Figure 3a.

Results And Discussion
We used the standard analysis of spin pumping on TI 5,6,7 to evaluate the spin-to-charge conversion J c /J s . Spin mixing conductance G eff ¯ which is used to illustrate the e ciency of generating spin current is extracted using Equation (1): where M s is saturation magnetization of Py, t Py is thickness of Py and g is Landé factor and u B is the Bohr magneton. M s is calculated using Kittel formula from f vs. H r (Figure 2c). Da = a Py/Cu/TI -a Py and is determined by analysing DH pp vs. f as shown in Figure 2d. For the spin current densities that injected across the interface due to spin pumping, Equation (2) is utilized as shown below: in which g is gyromagnetic ratio, w(=2pf) and h rf are frequency and amplitude of microwave magnetic eld respectively. The calculated J s is presented in Figure 3b. By dividing J c with J s , spin-to-charge conversion e ciency J c /J s can be determined as shown in Figure 3c.  Figure   3b). We further examined the J c vs. G eff ¯ as shown in Figure 4b. There is no enhancement of J c with increasing G eff ¯, revealing the spin-to-charge mechanism may not be dominated by the bulk spin hall effect (SHE). 14 Hence, we here suggest that the spin-to-charge conversion in our Py/Cu/TI system arises from inverse Edelstein effect, where the origin is the spin-momentum locked surface states of TI layer, same interpretation as other literatures. 5,11 Low G eff ¯ indicates strong spin back ow and spin memory loss (spin absorption) at the high SOC interface. 16 13 Here we refer FM to Py while NM to TI lm. One of the reason for the lower G eff ¯ compared with G Py/TI ¯ may be resulted from smaller spin conductance g Cu/TI of Cu/TI than that of G Py/TI ¯, similar to the case in Cu/Pt. 13,14 However, since G eff ¯ » G Py/TI ¯ at t Cu ³9nm, here we assume Cu/TI and Py/Cu present similar quality with G Py/Cu ¯ » g Cu/TI . Thus, by assuming degree of spin absorption at Cu/TI interfaces are similar in all cases, we suggest that the reason for lower G eff ¯ of 3nm and 7nm Cu-based trilayer samples could be due to the strong spin accumulation at this ultrathin regime. 13 Stronger spin back ow occurs, as compared to t Cu ³9nm, which eventually leads to reduction of G eff ¯.
The decrease of G eff ¯ seems have strong correlation to the spin-to-charge conversion e ciency. The next question is how such condition could increase the J c /J s ? Here we de ned J c /J s as l IEE = n f t where n f is Fermi velocity of TI surface states and t is IEE relaxation time. As shown in the Figure 4d, t is modi ed due to the tunnelling current into TI, which is determined by momentum relaxation time t p and interface tunnelling time t t as shown in Equation (3): 20 where l mf = n f t p = mean free path TI. According to this model, we proposed that the monotonic increase of l IEE with decreasing G eff ¯ attributed to the modi cation of IEE relaxation time t due to additional tunnelling barrier (Cu) that limits the interface transmission rate (1/t t ). 20,21 l IEE is always lower than l mf due to the correction factor of (1+2t p /t t ). It becomes clear that one can increase l IEE by reducing 1/t t , which can be done by introducing tunnelling barrier in between Py and TI layer. Using l IEE (t Cu =7nm) =2.7nm and based on our previous ARPES result, n f =5.7´10 5 m/s, 22 we nd t~4.7fs, which is same order of magnitude as Bi/Ag 23 and a-Sn/Ag 6 interfaces. Our extracted l IEE (=2.7nm) is higher than 0.1-0.4nm in Bi/Ag Rashba interface, 23 2.1nm and 2nm in TI SS of a-Sn/Ag 6 and HgTe/HgCdTe 7 respectively. We attribute the enhancement to the insertion of Cu tunnelling barrier. Although more theoretical calculation might be needed, our works indicate the importance of interface engineering to enhance the spin-tocharge conversion. This method could also be applied to other high SOC interfaces for obtaining high spin-to-charge conversion based on the inverse Edelstein effect, which is essential for spin current detector and other novel application such as broadband terahertz emitter. 24,25 Conclusion Page 6/11 In conclusion, we investigated the spin-to-charge conversion in Py/Cu/Bi 2 Se 3 using spin pumping technique. Enhancement of J c /J s with increasing t Cu is observed at room temperature, where J c /J s ∼2.7nm is achieved when 7nm of Cu layer is inserted. We proposed that the enhancement is attributed to the additional Cu interlayer as tunnelling barrier that modi ed the relaxation time at the interface. This work has provided a viable route for enhancing spin-to-charge conversion e ciency of TI, which is crucial for spin functional device applications.

Materials And Methods
Bi 2 Se 3 lms with 10nm thickness were synthesized using molecular beam epitaxy (MBE) method. The asgrown Bi 2 Se 3 were in situ capped with 2nm Se lm that used as protecting layer. The sample was then transferred to pulse laser deposition (PLD) chamber for depositing Cu and subsequently NiFe (Py) layer at room temperature. Before the depositions, Se layer was decapped in PLD chamber at 183°C for 1 hour.
A series of trilayer samples were prepared by varying Cu thickness, ranging from 3 to 11nm. Bilayer of Py/Bi 2 Se 3 was also prepared for comparison study. Py thickness was xed at 17 nm. 1 nm of Al lm was deposited on Py as capping layer. To evaluate spin-to-charge conversion, spin pumping technique was utilized (Fig. 1). Spin current was generated in Py via its ferromagnetic resonance (FMR) condition and injected into Bi 2 Se 3 , either with or without passing through the Cu layer (-z direction) (Fig. 1b). DC voltage was measured in x-direction and the produced charge current could be evaluated. All the measurements were done at room temperature.
Declarations Figure 1 (a) Tri-layer samples for spin pumping study, where nonmagnetic (NM) spacer was introduced in between FM and TI layer; (b) schematic illustrates the spin pumping experiment.