- Email: [email protected]
Au bilayer thin films
Solid State Communications 307 (2020) 113811
Contents lists available at ScienceDirect
Solid State Communications journal homepage: http://www.elsevier.com/locate/ssc
Communication
Magneto-thermal transport behavior in freestanding Ni80Fe20/Au bilayer thin films Anand Katailiha a, Paul C. Lou a, Sandeep Kumar a, b, * a b
Department of Mechanical Engineering, University of California, Riverside, CA, 92521, USA Materials Science and Engineering Program, University of California, Riverside, CA, 92521, USA
A R T I C L E I N F O
A B S T R A C T
Communicated by E.V. Sampathkumaran
Spin polarization, when induced in a non-ferromagnetic material, can change the underlying material behavior especially if the spin diffusion length is of the same order as the sample dimension such as thickness. In this study, we experimentally demonstrate thermal hysteretic behavior induced by spin polarization in Ni80Fe20 (10 nm)/Au (100 nm) bilayer freestanding sample. The thermal hysteresis behavior is uncovered using magneto thermal characterization based on self-heating 3ω method. The third harmonic voltage shows diverging behavior and thermal hysteresis during cooling and heating of the sample under an applied magnetic field, which is attributed to the spin accumulation. The spin accumulation and thermal hysteresis in Au occurs due to ferro magnetic proximity polarization from Ni80Fe20 layer. The observed hysteresis behavior is also attributed to freestanding thin film structure and absence of substrate effects leading to longer spin diffusion length. This study demonstrates experimental evidence of ferromagnetic proximity polarization and resulting changes in thermal transport behavior in Au thin films.
Keywords: Au Ferromagnetic proximity effect Spin accumulation Thermal hysteresis
1. Introduction Spintronics is considered to be an energy efficient alternative to modern electronics. To realize spintronics devices, spin current with large spin polarization needs to be achieved. Researchers often assume that fundamental behavior of the spin current carrying material, usually non-ferromagnetic, does not change due to spin polarization. However, a normal metal in proximity to ferromagnetic metal can exhibit under lying material behavior unexpected of normal metal. Recently, There has been experimental studies that report spin dependent behavior in pSi [1,2] and n-Si [3]. We hypothesized that ferromagnetic proximity mediated transport behavior can be observed in any material if the spin diffusion length is of same order as the critical dimension. The spin diffusion length of Au at room temperature is reported to be 9.5 nm [4] ~35 nm [5] and can rise to ~ 98 nm [5,6] at low temperatures. For a non-local spin valve geometry, the critical dimension will be length. The critical dimension for a bilayer, having ferromagnetic spin source and normal metal layers, will be thickness of the sample. We propose that the ferromagnetic proximity polarization in Au thin films having thickness between 35 nm and 1 μm can induce large spin accumulation leading to changes in transport behavior. Instead of charge transport, the resulting
behavior due to spin polarization can be studied using thermal transport measurements. Thin films thermal transport behavior can be charac terized using self-heating 3ω method [7] as demonstrated recently [1–3]. The roots of the self-heating 3ω method [7,8] can be found in the solution of one-dimensional heat equation given by � ∂ ∂2 I 2 sin2 ωt ρCp Tðx; tÞ κ 2 Tðx; tÞ ¼ 0 (1) ðR0 þ R’ Δðx; tÞ ∂T LA ∂x where ρ, Cp , κ, L & A are mass density, specific heat, thermal conduc tivity, length of sample between the voltage terminals and crosssectional area of the sample respectively. R0 is the resistance calcu lated at current I0 and at temperature T0 . R’ is the derivative of R0 with � � 0 . Δðx; tÞ is the space (x) and time (t) respect to temperature at T0 dR dT T0
dependent change in temperature of the sample along the heat current. The equation for the 3ω response after solving eq. is given by V3 ω �
4I 3 RR’ L q0ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi π 4 Aκ 1 þ ð2ωγÞ2
(2)
where γ is the thermal time constant and is related with the heat capacity
* Corresponding author. University of California, 900 University Ave. Bourns Hall A305, Riverside, CA, 9252, USA. E-mail address: [email protected] (S. Kumar). https://doi.org/10.1016/j.ssc.2019.113811 Received 16 October 2019; Received in revised form 17 December 2019; Accepted 28 December 2019 Available online 3 January 2020 0038-1098/© 2020 Elsevier Ltd. All rights reserved.
A. Katailiha et al.
�
Solid State Communications 307 (2020) 113811
� 2 Cp ¼ πρLγκ2 . The V3ω is a function of both thermal conductivity and heat
area using deep reactive ion etching (DRIE) as shown in Fig. 1 (a). The Pd (1 nm)/Ni80Fe20(10 nm)/Au (100 nm) sample is made freestanding by etching the thermal oxide underneath the sample using hydrofluoric acid vapor etching (HFVE). It needs to be clarified that while 10 nm Cr is deposited as adhesion layer but its final thickness may be significantly reduced to HFVE. The fabricated experimental setup is shown in Fig. 1 (a). The applied electric current causes a longitudinal temperature gradient T(x) and T0 is the far field temperature, which can be equated to the cryostat/substrate temperature. The current across the sample will lead to spin polarization of Au due to primarily non-local spin in jection [9] or ferromagnetic proximity polarization. The spin polarized Au layer will be equal to the spin diffusion length in Au as shown in Fig. 1 (b). The spin accumulation can also occur due to spin-Hall effect (SHE) as shown in Fig. 1 (b). The change in transport behavior due to spin polarization can be studied using the 3ω method as a function of magnetic field and temperature. We use Quantum Design’s physical property measurement system (PPMS) to undertake magneto-electro-thermal transport characteriza tion as a function of magnetic field as well temperature at high vacuum to reduce the convection heat loss. For magneto-electro-thermal char acterization, we first measured the V3ω response as a function of heating current from the 1 mA–5 mA. The V3ω response demonstrate a clear cubic relationship with the heating current as shown in Supplementary Fig. S1, which is essential for the thermal characterization. Once the cubic relationship is ascertained, R1ω and V3ω responses are acquired as a function of magnetic field from 2 T to 2 T. The applied magnetic field has no influence on the resistance of the sample at 300 K, which can be seen as flat magnetoresistance (MR) behavior in Fig. 1 (c) even when the field polarity is reversed. Since the sample is metallic, it is expected that it will obey Wiedemann Franz law (σ ∝κel , where σ is the electrical
capacity. The thermal conductivity can be expressed in terms of the third harmonic voltage V3ω in the low frequency limit by κ�
4I 3 Ro R’ L π4 V3ω S
(3)
The heat capacity and thermal conductivity can be considered as a function of resistance and V3ω response fðκ; Cp Þ ¼ VR3ω [1–3], which is
used to study changes in thermal transport as a function of magnetic field and temperature. In this study, we demonstrate ferromagnetic proximity polarization and resulting thermal hysteresis behavior in freestanding Ni80Fe20 (10 nm)/Au (100 nm) bilayer sample using magneto-thermal transport characterization. 2. Experimental setup and results
The self-heating 3ω method for characterization of thermal behavior needs a freestanding sample to eliminate conduction heat loss to the substrate. We developed the MEMS based platform with freestanding sample to study the spin dependent thermal transport behavior. The freestanding 3ω thin film structure was fabricated using commercially available 300 μm Si wafer as the substrate to begin with. We deposited 300 nm SiO2 using plasma enhanced chemical vapor deposition (PECVD), which acts as a sacrificial and electrical insulation layer. Using e-beam evaporation, we deposited 10 nm of Cr as an adhesion promoter followed by 100 nm of Au, 10 nm of Ni80Fe20 and 1 nm of Pd protective layer. We deposited 100 nm Al as electrode material. We then patterned the back side of the wafer and etched the Si right underneath the sample
Fig. 1. (a) The scanning electron micrograph showing the experimental setup with freestanding sample having Pd (1 nm)/Ni80Fe20 (10 nm)/Au (100 nm)/Cr (10 nm), (b) the schematic showing the non-local spin injection and spin-Hall effect, which will lead to spin accumulation in Au thin film; JC is charge current, JS is spin current and λSD is spin diffusion length of Au, (c) MR at 300 K and (d) the V3ω response as a function of magnetic field at 300 K. Arrows in (c)–(d) indicate the direction of field sweep. 2
A. Katailiha et al.
Solid State Communications 307 (2020) 113811
conductivity and κel is electronic thermal conductivity) thus, no change in electrical conductivity implies no change in thermal conductivity too, which means magnetic field dependent V3ω response should also exhibit no change (V3ω is related with κ according to eq. (3)). The V3ω response is observed to be weakly dependent on the applied magnetic field as shown in Fig. 1 (d). To uncover the temperature dependent thermal transport behavior, we measure the R1ω and V3ω response as a function of temperature from 300 K to 5 K on a second device. The measurement is carried out at 5 mA of heating current, no magnetic field and 0.3 K/min of heating and cooling rate. The R1ω measurement shows thermal hysteresis as shown in Fig. 2 (a), which can be attributed to the instrumental thermal drift. Similar thermal hysteretic behavior can also be seen in the V3ω response as shown in Fig. 2 (b). Then, we measured the R1ω and V3ω response as a function of temperature from 300 K to 50 K under an applied out of plane (z-direction) magnetic field of 1 T. We reduced the heating current to 3 mA and cooling/heating rate to 0.3 K/min to reduce the thermal drift. We do not observe significant thermal drift in the R1ω as shown in Fig. 2 (c). This measurement is carried out on a third device to ensure the repeatability of experimental results. The slope of the resistance as a function of temperature is measured to be 0.01357 Ω= K and the dif ference between the heating and cooling resistance is measured to be ~0.005 Ω. Using this information, we estimate the thermal drift to be ~0.4 K. Unlike the resistance measurement, the V3ω response shows a significant thermal hysteresis under applied magnetic field, which cannot be attributed to instrumental thermal drift since the V3ω value at the end of heating is lower than any value during cooling. We observe two diverging valleys during cooling (u, 238.2 K) and heating (v, 275.25 K). These valleys occur at different temperatures as shown in Fig. 2 (d). Under the influence of magnetic field (1 T along z-direction), the tem perature dependent behavior is similar to the field-sweep measurement (Fig. 1 (c) and (d)) i.e. resistance doesn’t change at the start and end of
the temperature cycle and V3ω response dropped significantly while coming back to room temperature. To understand the V3ω response, we analyzed equation (3). Using Fourier’s equation for thermal conduction, we can write� � � _ I 2 Ro L 4IR’ QL 4I 3 Ro R’ L κ� 4 (4) ¼ ¼ 4 S SδT π V3 ω S π V3ω ΔT ¼
π 4 V3 ω 4IR’
(5)
The ΔT is temperature gradient and it is proportional to the V3ω response. The equation for temperature gradient is similar in form to the equation for temperature gradient given in original 3ω method for cross thermal conductivity by D. Cahill [10]. From equation (5), we can state that the temperature gradient is reduced during one cycle of cooling and heating. These valleys can be interpreted as a magnetocaloric effect, which leads to the thermal hysteresis behavior similar to magnetocaloric effects [11,12]. This behavior is not a traditional thermal hysteresis behavior due to magnetocaloric effect [11,13–16] since only tempera ture gradient is reduced and not the actual temperature. But, we propose this is an emergent (not intrinsic) magnetocaloric effect due to combined effect of Ni80Fe20 and Au layers. As stated earlier, the thermal hysteresis can be considered a reduction in temperature gradient using equation (5). To further understand the emergent behavior as well as the valleys in the temperature dependent behavior, we measure the R1ω and V3ω responses cyclically under cooling and heating between 300 K and 200 K at cooling/heating rate of 1 K/min. This higher cooling/heating rate may induce thermal drift, which can be quantified. We undertake measurement for 12 cycles of cooling and heating at an applied out of plane magnetic field of 1 T. During this measurement, the V3ω response drops after every cycle of cooling and heating as shown in Fig. 3 (a). To uncover the effect of magnetic field, measurement is
Fig. 2. (a) resistance and (b) the V3ω response as a function of temperature between 300 K and 5 K at 5 mA of heating current applied across the sample and zero applied magnetic field showing hysteresis in cooling and heating thermal cycle attributed to the thermal drift, (c) resistance and (d) the V3ω response as a function of temperature between 300 K and 50 K at an applied magnetic field of 1 T. 3
A. Katailiha et al.
Solid State Communications 307 (2020) 113811
carried out for four cycles each at 2 T (red line) and at 0.5 T (blue line). These measurements also exhibit similar reduction in the V3ω response. During this measurement (20 cycles), the V3ω response drops from ~42.5 μV (point r) to ~32.2 μV (point s) as shown in Fig. 3 (a) while resistance values are stable as shown in Fig. 3 (b). Using the resistance data, we estimate the instrumental thermal drift to be ~10.12 K whereas the thermal hysteresis in V3ω response is much larger. The magnitude and direction of magnetic field does not have significant effect on the V3ω response behavior and the drop in the V3ω response after each cycle is primarily a temperature dependent phenomenon but in the presence of magnetic field. The reduction in the V3ω response craters after 20 cycles and further reduction during each thermal cycle of cooling and heating is insignificant. The cooling transition behavior is completely suppressed but heating transition can be observed after 20 thermal cy cles. The thermal hysteresis [11,13–16] behavior can arise due to magnetocaloric effect as stated earlier in addition to spin-crossover due to spin accumulation [17].
not affect the high temperature thermal hysteresis behavior. In addition, � the current density in our measurements is relatively small 2:7 � � 105 cmA2 . As a consequence, Rashba-Edelstein effect is not expected to give rise to significant spin accumulation in this study. This led us to hypothesize that ferromagnetic metal layer is the underlying cause of spin accumulation. The hybridization at Au and Ni80Fe20 interface may lead to proximity induced magnetization without the need of a current or thermal gradient. To ascertain it, we undertake a thermal cycling experiment where current is only applied to measure the V3ω response at the start and at the end. During the cooling (300 K–200 K) and heating (200 K–300 K) of the sample, the applied current is switched off. In this measurement, the V3ω response is same before and after thermal cycling unlike the thermal hysteresis behavior shown in Figs. 2 (d) and 3(a). This led us to conclude that interfacial hybridization is not the under lying cause of thermal hysteresis. The ferromagnetic layer can also influence the behavior due to ferromagnetic proximity polarization, which will include spin injection as shown in Fig. 1(b). In order to ascertain the contribution of Ni80Fe20 layer to the observed thermal hysteresis behavior, we fabricated a Ni80Fe20 thin film control sample on the suspended oxide layer. The control sample on suspended oxide removes the substrate effects. The sample could not be made freestanding since HF chemically etches the Ni80Fe20 thin film. We measured the resistance and the V3ω response as a function of temperature from 300 K to 50 K as shown in Supplementary Fig. S3. The V3ω response shows a diverging behavior at ~240 K similar to Au bilayer sample. This control experiment clearly show that the likely origin of the thermal hysteresis behavior is Ni80Fe20 thin film, which means that ferromagnetic proximity polarization due to spin in jection is the underlying mechanism as shown in Fig. 3 (c). The thermal hysteresis behavior can be considered a magnetocaloric effect. The magnetocaloric effect has been proposed in ferromagnetic/para magnetic heterostructures. This behavior arises due to interactions be tween ferromagnetic layer effecting the magnetic moment in paramagnetic metal [28]. The experimental results reported in this work support this mechanism. The magnetocaloric effect leads to successive reduction in V3ω response. This effect can be attributed to the free standing thin film structure as well since the substrate mediated ther malization of phonons is absent.
3. Discussion These experimental results are surprising and are not expected to arise from the Au thin films only. While nanoscale Au thin films (<15 nm) and Au nanoparticles have been reported to exhibit magnetism at low temperatures (below 4 K) [18–22], but there is no reported study of high temperature behavior similar to the one presented in this work. As proposed initially, the observed magneto-thermal transport behavior can arise in Au thin films due spin polarization since the spin diffusion length is of the same order as the thickness of the sample in this study. The enhancement in spin accumulation due to confinement has been reported for Permalloy/Au thin film structures [23], which has been used to study spin transport [24,25]. Spin accumulation mediated magnetization in Au thin films has also been observed [26]. There are multiple mechanisms that can lead to spin accumulation in the Au layer. These include Rashba-Edelstein effect, spin-Hall effect and ferromagnetic proximity polarization. Both Rashba-Edelstein effect and spin-Hall effect can be uncovered using angle dependent magnetore sistance (ADMR) [27]. The ADMR measurement in zy-plane at 300 K does not show any angle dependent behavior, which eliminates both Rashba-Edelstein effect and spin-Hall effect as an underlying mechanism for spin accumulation. It is noted that weak spin-Hall magnetoresistance (SMR) response is observed at 25 K (Supplementary Fig. S2), which will
Fig. 3. (a) the V3ω response as a function of temper ature cycled between 300 K and 200 K at applied magnetic field of 1 T (black) for 12 cycles, 2 T (red) for 4 cycles and 0.5 T (blue) for 4 cycles, (b) The resistance as a function of temperature cycled be tween 300 K and 200 K showing the thermal drift for 20 cycles between 300 K and 200 K. Resistance values during cooling and heating are same during thermal cycling indicating absence of thermal hysteresis and (c) schematic of the observed behavior showing the non-local spin injection in the Au layer and enhanced spin accumulation as the temperature is lowered leads to paramagnetic state, spin glass transition and ther mal hysteresis. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
4
A. Katailiha et al.
Solid State Communications 307 (2020) 113811
4. Conclusion
[4] D. Qu, S.Y. Huang, B.F. Miao, S.X. Huang, C.L. Chien, Phys. Rev. B 89 (2014) 140407. [5] M. Isasa, E. Villamor, L.E. Hueso, M. Gradhand, F. Casanova, Phys. Rev. B 91 (2015), 024402. [6] M. Johnson, Phys. Rev. Lett. 70 (1993) 2142–2145. [7] L. Lu, W. Yi, D.L. Zhang, Rev. Sci. Instrum. 72 (2001) 2996–3003. [8] J. Taehee, M.T. Moneck, Z. Jian-Gang, IEEE Trans. Magn. 48 (2012) 3031–3034. [9] H. Idzuchi, Y. Fukuma, Y. Otani, Phys. E Low-dimens. Syst. Nanostruct. 68 (2015) 239–263. [10] D.G. Cahill, Rev. Sci. Instrum. 61 (1990) 802–808. [11] J. Liu, T. Gottschall, K.P. Skokov, J.D. Moore, O. Gutfleisch, Nat. Mater. 11 (2012) 620. [12] F.I.F. Nascimento, A.L. Dantas, L.L. Oliveira, V.D. Mello, R.E. Camley, A.S. Carriço, Phys. Rev. B 80 (2009) 144407. [13] O. Gutfleisch, T. Gottschall, M. Fries, D. Benke, I. Radulov, K.P. Skokov, H. Wende, M. Gruner, M. Acet, P. Entel, M. Farle, Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 374 (2016). [14] S.Y. Dan’kov, A.M. Tishin, V.K. Pecharsky, K.A. Gschneidner, Phys. Rev. B 57 (1998) 3478–3490. [15] A.M. Tishin, K.A. Gschneidner, V.K. Pecharsky, Phys. Rev. B 59 (1999) 503–511. [16] B.G. Shen, J.R. Sun, F.X. Hu, H.W. Zhang, Z.H. Cheng, Adv. Mater. 21 (2009) 4545–4564. [17] S. Brooker, Chem. Soc. Rev. 44 (2015) 2880–2892. [18] S. Reich, G. Leitus, Y. Feldman, Appl. Phys. Lett. 88 (2006) 222502. [19] H. Hori, Y. Yamamoto, T. Iwamoto, T. Miura, T. Teranishi, M. Miyake, Phys. Rev. B 69 (2004) 174411. [20] S. Prusty, V. Siva, N. Shukla, B. Satpati, K. Senapati, P.K. Sahoo, RSC Adv. 6 (2016) 106584–106590. [21] R. Singh, J. Magn. Magn. Mater. 346 (2013) 58–73. [22] V. Tuboltsev, A. Savin, A. Pirojenko, J. R€ ais€ anen, ACS Nano 7 (2013) 6691–6699. [23] P. Laczkowski, L. Vila, V.D. Nguyen, A. Marty, J.P. Attan� e, H. Jaffr�es, J.M. George, A. Fert, Phys. Rev. B 85 (2012) 220404. [24] O. Stejskal, J. Hamrle, J. Pi�stora, Y. Otani, J. Magn. Magn. Mater. 414 (2016) 132–143. [25] M. Emmanouil, A. Libe, M. Goran, V. Estitxu, L. Roger, C. F�elix, E.H. Luis, Nanotechnology 27 (2016), 095201. [26] M. Johnson, J. Appl. Phys. 75 (1994) 6714–6719. [27] Y. Lv, J. Kally, D. Zhang, J.S. Lee, M. Jamali, N. Samarth, J.-P. Wang, Nat. Commun. 9 (2018) 111. [28] S.N. Vdovichev, N.I. Polushkin, I.D. Rodionov, V.N. Prudnikov, J. Chang, A. A. Fraerman, Phys. Rev. B 98 (2018), 014428.
In conclusion, we report experimental measurement of magnetothermal transport behavior in the Ni80Fe20 (10 nm)/Au (100 nm) bilayer freestanding sample. The thermal transport behavior is uncov ered using self-heating 3ω-method based magneto thermal character ization. The third harmonic voltage shows diverging behavior and thermal hysteresis during cooling and heating under an applied mag netic field. The thermal hysteresis is attributed to ferromagnetic prox imity polarization and weak magnetocaloric effect in Ni80Fe20 thin films. The observed behavior can also be attributed to absence of sub strate effects leading to longer spin diffusion length and spin lifetimes. These experimental results in Au thin films may provide scientific di rection to study spin dependent behavior in widely studied diamagnetic and paramagnetic materials. In addition, layered thin films structures of normal metals/ferromagnetic metals can be used to achieve magneto caloric effect without using complex and rare earth magnetic materials. Acknowledgement AK and PCL have equal contribution to this work. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.ssc.2019.113811. References [1] P.C. Lou, S. Kumar, Phys. Status Solidi (b) (2017) 1700545. [2] P.C. Lou, W.P. Beyermann, S. Kumar, J. Appl. Phys. 122 (2017) 123905. [3] P.C. Lou, S. Kumar, J. Magn. Magn. Mater. 452 (2018) 129–133.
5
Contents lists available at ScienceDirect
Solid State Communications journal homepage: http://www.elsevier.com/locate/ssc
Communication
Magneto-thermal transport behavior in freestanding Ni80Fe20/Au bilayer thin films Anand Katailiha a, Paul C. Lou a, Sandeep Kumar a, b, * a b
Department of Mechanical Engineering, University of California, Riverside, CA, 92521, USA Materials Science and Engineering Program, University of California, Riverside, CA, 92521, USA
A R T I C L E I N F O
A B S T R A C T
Communicated by E.V. Sampathkumaran
Spin polarization, when induced in a non-ferromagnetic material, can change the underlying material behavior especially if the spin diffusion length is of the same order as the sample dimension such as thickness. In this study, we experimentally demonstrate thermal hysteretic behavior induced by spin polarization in Ni80Fe20 (10 nm)/Au (100 nm) bilayer freestanding sample. The thermal hysteresis behavior is uncovered using magneto thermal characterization based on self-heating 3ω method. The third harmonic voltage shows diverging behavior and thermal hysteresis during cooling and heating of the sample under an applied magnetic field, which is attributed to the spin accumulation. The spin accumulation and thermal hysteresis in Au occurs due to ferro magnetic proximity polarization from Ni80Fe20 layer. The observed hysteresis behavior is also attributed to freestanding thin film structure and absence of substrate effects leading to longer spin diffusion length. This study demonstrates experimental evidence of ferromagnetic proximity polarization and resulting changes in thermal transport behavior in Au thin films.
Keywords: Au Ferromagnetic proximity effect Spin accumulation Thermal hysteresis
1. Introduction Spintronics is considered to be an energy efficient alternative to modern electronics. To realize spintronics devices, spin current with large spin polarization needs to be achieved. Researchers often assume that fundamental behavior of the spin current carrying material, usually non-ferromagnetic, does not change due to spin polarization. However, a normal metal in proximity to ferromagnetic metal can exhibit under lying material behavior unexpected of normal metal. Recently, There has been experimental studies that report spin dependent behavior in pSi [1,2] and n-Si [3]. We hypothesized that ferromagnetic proximity mediated transport behavior can be observed in any material if the spin diffusion length is of same order as the critical dimension. The spin diffusion length of Au at room temperature is reported to be 9.5 nm [4] ~35 nm [5] and can rise to ~ 98 nm [5,6] at low temperatures. For a non-local spin valve geometry, the critical dimension will be length. The critical dimension for a bilayer, having ferromagnetic spin source and normal metal layers, will be thickness of the sample. We propose that the ferromagnetic proximity polarization in Au thin films having thickness between 35 nm and 1 μm can induce large spin accumulation leading to changes in transport behavior. Instead of charge transport, the resulting
behavior due to spin polarization can be studied using thermal transport measurements. Thin films thermal transport behavior can be charac terized using self-heating 3ω method [7] as demonstrated recently [1–3]. The roots of the self-heating 3ω method [7,8] can be found in the solution of one-dimensional heat equation given by � ∂ ∂2 I 2 sin2 ωt ρCp Tðx; tÞ κ 2 Tðx; tÞ ¼ 0 (1) ðR0 þ R’ Δðx; tÞ ∂T LA ∂x where ρ, Cp , κ, L & A are mass density, specific heat, thermal conduc tivity, length of sample between the voltage terminals and crosssectional area of the sample respectively. R0 is the resistance calcu lated at current I0 and at temperature T0 . R’ is the derivative of R0 with � � 0 . Δðx; tÞ is the space (x) and time (t) respect to temperature at T0 dR dT T0
dependent change in temperature of the sample along the heat current. The equation for the 3ω response after solving eq. is given by V3 ω �
4I 3 RR’ L q0ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi π 4 Aκ 1 þ ð2ωγÞ2
(2)
where γ is the thermal time constant and is related with the heat capacity
* Corresponding author. University of California, 900 University Ave. Bourns Hall A305, Riverside, CA, 9252, USA. E-mail address: [email protected] (S. Kumar). https://doi.org/10.1016/j.ssc.2019.113811 Received 16 October 2019; Received in revised form 17 December 2019; Accepted 28 December 2019 Available online 3 January 2020 0038-1098/© 2020 Elsevier Ltd. All rights reserved.
A. Katailiha et al.
�
Solid State Communications 307 (2020) 113811
� 2 Cp ¼ πρLγκ2 . The V3ω is a function of both thermal conductivity and heat
area using deep reactive ion etching (DRIE) as shown in Fig. 1 (a). The Pd (1 nm)/Ni80Fe20(10 nm)/Au (100 nm) sample is made freestanding by etching the thermal oxide underneath the sample using hydrofluoric acid vapor etching (HFVE). It needs to be clarified that while 10 nm Cr is deposited as adhesion layer but its final thickness may be significantly reduced to HFVE. The fabricated experimental setup is shown in Fig. 1 (a). The applied electric current causes a longitudinal temperature gradient T(x) and T0 is the far field temperature, which can be equated to the cryostat/substrate temperature. The current across the sample will lead to spin polarization of Au due to primarily non-local spin in jection [9] or ferromagnetic proximity polarization. The spin polarized Au layer will be equal to the spin diffusion length in Au as shown in Fig. 1 (b). The spin accumulation can also occur due to spin-Hall effect (SHE) as shown in Fig. 1 (b). The change in transport behavior due to spin polarization can be studied using the 3ω method as a function of magnetic field and temperature. We use Quantum Design’s physical property measurement system (PPMS) to undertake magneto-electro-thermal transport characteriza tion as a function of magnetic field as well temperature at high vacuum to reduce the convection heat loss. For magneto-electro-thermal char acterization, we first measured the V3ω response as a function of heating current from the 1 mA–5 mA. The V3ω response demonstrate a clear cubic relationship with the heating current as shown in Supplementary Fig. S1, which is essential for the thermal characterization. Once the cubic relationship is ascertained, R1ω and V3ω responses are acquired as a function of magnetic field from 2 T to 2 T. The applied magnetic field has no influence on the resistance of the sample at 300 K, which can be seen as flat magnetoresistance (MR) behavior in Fig. 1 (c) even when the field polarity is reversed. Since the sample is metallic, it is expected that it will obey Wiedemann Franz law (σ ∝κel , where σ is the electrical
capacity. The thermal conductivity can be expressed in terms of the third harmonic voltage V3ω in the low frequency limit by κ�
4I 3 Ro R’ L π4 V3ω S
(3)
The heat capacity and thermal conductivity can be considered as a function of resistance and V3ω response fðκ; Cp Þ ¼ VR3ω [1–3], which is
used to study changes in thermal transport as a function of magnetic field and temperature. In this study, we demonstrate ferromagnetic proximity polarization and resulting thermal hysteresis behavior in freestanding Ni80Fe20 (10 nm)/Au (100 nm) bilayer sample using magneto-thermal transport characterization. 2. Experimental setup and results
The self-heating 3ω method for characterization of thermal behavior needs a freestanding sample to eliminate conduction heat loss to the substrate. We developed the MEMS based platform with freestanding sample to study the spin dependent thermal transport behavior. The freestanding 3ω thin film structure was fabricated using commercially available 300 μm Si wafer as the substrate to begin with. We deposited 300 nm SiO2 using plasma enhanced chemical vapor deposition (PECVD), which acts as a sacrificial and electrical insulation layer. Using e-beam evaporation, we deposited 10 nm of Cr as an adhesion promoter followed by 100 nm of Au, 10 nm of Ni80Fe20 and 1 nm of Pd protective layer. We deposited 100 nm Al as electrode material. We then patterned the back side of the wafer and etched the Si right underneath the sample
Fig. 1. (a) The scanning electron micrograph showing the experimental setup with freestanding sample having Pd (1 nm)/Ni80Fe20 (10 nm)/Au (100 nm)/Cr (10 nm), (b) the schematic showing the non-local spin injection and spin-Hall effect, which will lead to spin accumulation in Au thin film; JC is charge current, JS is spin current and λSD is spin diffusion length of Au, (c) MR at 300 K and (d) the V3ω response as a function of magnetic field at 300 K. Arrows in (c)–(d) indicate the direction of field sweep. 2
A. Katailiha et al.
Solid State Communications 307 (2020) 113811
conductivity and κel is electronic thermal conductivity) thus, no change in electrical conductivity implies no change in thermal conductivity too, which means magnetic field dependent V3ω response should also exhibit no change (V3ω is related with κ according to eq. (3)). The V3ω response is observed to be weakly dependent on the applied magnetic field as shown in Fig. 1 (d). To uncover the temperature dependent thermal transport behavior, we measure the R1ω and V3ω response as a function of temperature from 300 K to 5 K on a second device. The measurement is carried out at 5 mA of heating current, no magnetic field and 0.3 K/min of heating and cooling rate. The R1ω measurement shows thermal hysteresis as shown in Fig. 2 (a), which can be attributed to the instrumental thermal drift. Similar thermal hysteretic behavior can also be seen in the V3ω response as shown in Fig. 2 (b). Then, we measured the R1ω and V3ω response as a function of temperature from 300 K to 50 K under an applied out of plane (z-direction) magnetic field of 1 T. We reduced the heating current to 3 mA and cooling/heating rate to 0.3 K/min to reduce the thermal drift. We do not observe significant thermal drift in the R1ω as shown in Fig. 2 (c). This measurement is carried out on a third device to ensure the repeatability of experimental results. The slope of the resistance as a function of temperature is measured to be 0.01357 Ω= K and the dif ference between the heating and cooling resistance is measured to be ~0.005 Ω. Using this information, we estimate the thermal drift to be ~0.4 K. Unlike the resistance measurement, the V3ω response shows a significant thermal hysteresis under applied magnetic field, which cannot be attributed to instrumental thermal drift since the V3ω value at the end of heating is lower than any value during cooling. We observe two diverging valleys during cooling (u, 238.2 K) and heating (v, 275.25 K). These valleys occur at different temperatures as shown in Fig. 2 (d). Under the influence of magnetic field (1 T along z-direction), the tem perature dependent behavior is similar to the field-sweep measurement (Fig. 1 (c) and (d)) i.e. resistance doesn’t change at the start and end of
the temperature cycle and V3ω response dropped significantly while coming back to room temperature. To understand the V3ω response, we analyzed equation (3). Using Fourier’s equation for thermal conduction, we can write� � � _ I 2 Ro L 4IR’ QL 4I 3 Ro R’ L κ� 4 (4) ¼ ¼ 4 S SδT π V3 ω S π V3ω ΔT ¼
π 4 V3 ω 4IR’
(5)
The ΔT is temperature gradient and it is proportional to the V3ω response. The equation for temperature gradient is similar in form to the equation for temperature gradient given in original 3ω method for cross thermal conductivity by D. Cahill [10]. From equation (5), we can state that the temperature gradient is reduced during one cycle of cooling and heating. These valleys can be interpreted as a magnetocaloric effect, which leads to the thermal hysteresis behavior similar to magnetocaloric effects [11,12]. This behavior is not a traditional thermal hysteresis behavior due to magnetocaloric effect [11,13–16] since only tempera ture gradient is reduced and not the actual temperature. But, we propose this is an emergent (not intrinsic) magnetocaloric effect due to combined effect of Ni80Fe20 and Au layers. As stated earlier, the thermal hysteresis can be considered a reduction in temperature gradient using equation (5). To further understand the emergent behavior as well as the valleys in the temperature dependent behavior, we measure the R1ω and V3ω responses cyclically under cooling and heating between 300 K and 200 K at cooling/heating rate of 1 K/min. This higher cooling/heating rate may induce thermal drift, which can be quantified. We undertake measurement for 12 cycles of cooling and heating at an applied out of plane magnetic field of 1 T. During this measurement, the V3ω response drops after every cycle of cooling and heating as shown in Fig. 3 (a). To uncover the effect of magnetic field, measurement is
Fig. 2. (a) resistance and (b) the V3ω response as a function of temperature between 300 K and 5 K at 5 mA of heating current applied across the sample and zero applied magnetic field showing hysteresis in cooling and heating thermal cycle attributed to the thermal drift, (c) resistance and (d) the V3ω response as a function of temperature between 300 K and 50 K at an applied magnetic field of 1 T. 3
A. Katailiha et al.
Solid State Communications 307 (2020) 113811
carried out for four cycles each at 2 T (red line) and at 0.5 T (blue line). These measurements also exhibit similar reduction in the V3ω response. During this measurement (20 cycles), the V3ω response drops from ~42.5 μV (point r) to ~32.2 μV (point s) as shown in Fig. 3 (a) while resistance values are stable as shown in Fig. 3 (b). Using the resistance data, we estimate the instrumental thermal drift to be ~10.12 K whereas the thermal hysteresis in V3ω response is much larger. The magnitude and direction of magnetic field does not have significant effect on the V3ω response behavior and the drop in the V3ω response after each cycle is primarily a temperature dependent phenomenon but in the presence of magnetic field. The reduction in the V3ω response craters after 20 cycles and further reduction during each thermal cycle of cooling and heating is insignificant. The cooling transition behavior is completely suppressed but heating transition can be observed after 20 thermal cy cles. The thermal hysteresis [11,13–16] behavior can arise due to magnetocaloric effect as stated earlier in addition to spin-crossover due to spin accumulation [17].
not affect the high temperature thermal hysteresis behavior. In addition, � the current density in our measurements is relatively small 2:7 � � 105 cmA2 . As a consequence, Rashba-Edelstein effect is not expected to give rise to significant spin accumulation in this study. This led us to hypothesize that ferromagnetic metal layer is the underlying cause of spin accumulation. The hybridization at Au and Ni80Fe20 interface may lead to proximity induced magnetization without the need of a current or thermal gradient. To ascertain it, we undertake a thermal cycling experiment where current is only applied to measure the V3ω response at the start and at the end. During the cooling (300 K–200 K) and heating (200 K–300 K) of the sample, the applied current is switched off. In this measurement, the V3ω response is same before and after thermal cycling unlike the thermal hysteresis behavior shown in Figs. 2 (d) and 3(a). This led us to conclude that interfacial hybridization is not the under lying cause of thermal hysteresis. The ferromagnetic layer can also influence the behavior due to ferromagnetic proximity polarization, which will include spin injection as shown in Fig. 1(b). In order to ascertain the contribution of Ni80Fe20 layer to the observed thermal hysteresis behavior, we fabricated a Ni80Fe20 thin film control sample on the suspended oxide layer. The control sample on suspended oxide removes the substrate effects. The sample could not be made freestanding since HF chemically etches the Ni80Fe20 thin film. We measured the resistance and the V3ω response as a function of temperature from 300 K to 50 K as shown in Supplementary Fig. S3. The V3ω response shows a diverging behavior at ~240 K similar to Au bilayer sample. This control experiment clearly show that the likely origin of the thermal hysteresis behavior is Ni80Fe20 thin film, which means that ferromagnetic proximity polarization due to spin in jection is the underlying mechanism as shown in Fig. 3 (c). The thermal hysteresis behavior can be considered a magnetocaloric effect. The magnetocaloric effect has been proposed in ferromagnetic/para magnetic heterostructures. This behavior arises due to interactions be tween ferromagnetic layer effecting the magnetic moment in paramagnetic metal [28]. The experimental results reported in this work support this mechanism. The magnetocaloric effect leads to successive reduction in V3ω response. This effect can be attributed to the free standing thin film structure as well since the substrate mediated ther malization of phonons is absent.
3. Discussion These experimental results are surprising and are not expected to arise from the Au thin films only. While nanoscale Au thin films (<15 nm) and Au nanoparticles have been reported to exhibit magnetism at low temperatures (below 4 K) [18–22], but there is no reported study of high temperature behavior similar to the one presented in this work. As proposed initially, the observed magneto-thermal transport behavior can arise in Au thin films due spin polarization since the spin diffusion length is of the same order as the thickness of the sample in this study. The enhancement in spin accumulation due to confinement has been reported for Permalloy/Au thin film structures [23], which has been used to study spin transport [24,25]. Spin accumulation mediated magnetization in Au thin films has also been observed [26]. There are multiple mechanisms that can lead to spin accumulation in the Au layer. These include Rashba-Edelstein effect, spin-Hall effect and ferromagnetic proximity polarization. Both Rashba-Edelstein effect and spin-Hall effect can be uncovered using angle dependent magnetore sistance (ADMR) [27]. The ADMR measurement in zy-plane at 300 K does not show any angle dependent behavior, which eliminates both Rashba-Edelstein effect and spin-Hall effect as an underlying mechanism for spin accumulation. It is noted that weak spin-Hall magnetoresistance (SMR) response is observed at 25 K (Supplementary Fig. S2), which will
Fig. 3. (a) the V3ω response as a function of temper ature cycled between 300 K and 200 K at applied magnetic field of 1 T (black) for 12 cycles, 2 T (red) for 4 cycles and 0.5 T (blue) for 4 cycles, (b) The resistance as a function of temperature cycled be tween 300 K and 200 K showing the thermal drift for 20 cycles between 300 K and 200 K. Resistance values during cooling and heating are same during thermal cycling indicating absence of thermal hysteresis and (c) schematic of the observed behavior showing the non-local spin injection in the Au layer and enhanced spin accumulation as the temperature is lowered leads to paramagnetic state, spin glass transition and ther mal hysteresis. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
4
A. Katailiha et al.
Solid State Communications 307 (2020) 113811
4. Conclusion
[4] D. Qu, S.Y. Huang, B.F. Miao, S.X. Huang, C.L. Chien, Phys. Rev. B 89 (2014) 140407. [5] M. Isasa, E. Villamor, L.E. Hueso, M. Gradhand, F. Casanova, Phys. Rev. B 91 (2015), 024402. [6] M. Johnson, Phys. Rev. Lett. 70 (1993) 2142–2145. [7] L. Lu, W. Yi, D.L. Zhang, Rev. Sci. Instrum. 72 (2001) 2996–3003. [8] J. Taehee, M.T. Moneck, Z. Jian-Gang, IEEE Trans. Magn. 48 (2012) 3031–3034. [9] H. Idzuchi, Y. Fukuma, Y. Otani, Phys. E Low-dimens. Syst. Nanostruct. 68 (2015) 239–263. [10] D.G. Cahill, Rev. Sci. Instrum. 61 (1990) 802–808. [11] J. Liu, T. Gottschall, K.P. Skokov, J.D. Moore, O. Gutfleisch, Nat. Mater. 11 (2012) 620. [12] F.I.F. Nascimento, A.L. Dantas, L.L. Oliveira, V.D. Mello, R.E. Camley, A.S. Carriço, Phys. Rev. B 80 (2009) 144407. [13] O. Gutfleisch, T. Gottschall, M. Fries, D. Benke, I. Radulov, K.P. Skokov, H. Wende, M. Gruner, M. Acet, P. Entel, M. Farle, Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 374 (2016). [14] S.Y. Dan’kov, A.M. Tishin, V.K. Pecharsky, K.A. Gschneidner, Phys. Rev. B 57 (1998) 3478–3490. [15] A.M. Tishin, K.A. Gschneidner, V.K. Pecharsky, Phys. Rev. B 59 (1999) 503–511. [16] B.G. Shen, J.R. Sun, F.X. Hu, H.W. Zhang, Z.H. Cheng, Adv. Mater. 21 (2009) 4545–4564. [17] S. Brooker, Chem. Soc. Rev. 44 (2015) 2880–2892. [18] S. Reich, G. Leitus, Y. Feldman, Appl. Phys. Lett. 88 (2006) 222502. [19] H. Hori, Y. Yamamoto, T. Iwamoto, T. Miura, T. Teranishi, M. Miyake, Phys. Rev. B 69 (2004) 174411. [20] S. Prusty, V. Siva, N. Shukla, B. Satpati, K. Senapati, P.K. Sahoo, RSC Adv. 6 (2016) 106584–106590. [21] R. Singh, J. Magn. Magn. Mater. 346 (2013) 58–73. [22] V. Tuboltsev, A. Savin, A. Pirojenko, J. R€ ais€ anen, ACS Nano 7 (2013) 6691–6699. [23] P. Laczkowski, L. Vila, V.D. Nguyen, A. Marty, J.P. Attan� e, H. Jaffr�es, J.M. George, A. Fert, Phys. Rev. B 85 (2012) 220404. [24] O. Stejskal, J. Hamrle, J. Pi�stora, Y. Otani, J. Magn. Magn. Mater. 414 (2016) 132–143. [25] M. Emmanouil, A. Libe, M. Goran, V. Estitxu, L. Roger, C. F�elix, E.H. Luis, Nanotechnology 27 (2016), 095201. [26] M. Johnson, J. Appl. Phys. 75 (1994) 6714–6719. [27] Y. Lv, J. Kally, D. Zhang, J.S. Lee, M. Jamali, N. Samarth, J.-P. Wang, Nat. Commun. 9 (2018) 111. [28] S.N. Vdovichev, N.I. Polushkin, I.D. Rodionov, V.N. Prudnikov, J. Chang, A. A. Fraerman, Phys. Rev. B 98 (2018), 014428.
In conclusion, we report experimental measurement of magnetothermal transport behavior in the Ni80Fe20 (10 nm)/Au (100 nm) bilayer freestanding sample. The thermal transport behavior is uncov ered using self-heating 3ω-method based magneto thermal character ization. The third harmonic voltage shows diverging behavior and thermal hysteresis during cooling and heating under an applied mag netic field. The thermal hysteresis is attributed to ferromagnetic prox imity polarization and weak magnetocaloric effect in Ni80Fe20 thin films. The observed behavior can also be attributed to absence of sub strate effects leading to longer spin diffusion length and spin lifetimes. These experimental results in Au thin films may provide scientific di rection to study spin dependent behavior in widely studied diamagnetic and paramagnetic materials. In addition, layered thin films structures of normal metals/ferromagnetic metals can be used to achieve magneto caloric effect without using complex and rare earth magnetic materials. Acknowledgement AK and PCL have equal contribution to this work. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.ssc.2019.113811. References [1] P.C. Lou, S. Kumar, Phys. Status Solidi (b) (2017) 1700545. [2] P.C. Lou, W.P. Beyermann, S. Kumar, J. Appl. Phys. 122 (2017) 123905. [3] P.C. Lou, S. Kumar, J. Magn. Magn. Mater. 452 (2018) 129–133.
5