You are on page 1of 27

1 Supplementary information for

2 Electrically Switchable van der Waals Magnon Valves

3 Guangyi Chen1†, Shaomian Qi1†, Jianqiao Liu1, Di Chen1,2, Jiongjie Wang3, Shili Yan2, Yu
4 Zhang2, Shimin Cao1,2, Ming Lu1,2, Shibing Tian4, Kangyao Chen1, Peng Yu5, Zheng Liu6, X. C.
5 Xie1,2,7, Jiang Xiao3, Ryuichi Shindou1, Jian-Hao Chen1,2,8,9*
1
6 International Center of Quantum Materials, School of Physics, Peking University, Beijing, China
2
7 Beijing Academy of Quantum Information Sciences, Beijing, China
3
8 Department of Physics and State Key Laboratory of Surface Physics, Fudan University, Shanghai, China
4
9 Institute of Physics, Chinese Academy of Sciences, Beijing, China
5
10 State Key Laboratory of Optoelectronic Materials and Technologies, School of Materials Science and
11 Engineering, Sun Yat-sen University, Guangzhou, China
6
12 School of Materials Science and Engineering, Nanyang Technological University, Singapore, Singapore
7
13 CAS Center for Excellence in Topological Quantum Computation,University of Chinese Academy of
14 Sciences, Beijing 100190, China
8
15 Key Laboratory for the Physics and Chemistry of Nanodevices, Peking University, Beijing, China
9
16 Interdisciplinary Institute of Light-Element Quantum Materials and Research Center for Light-Element
17 Advanced Materials, Peking University, Beijing
18

19 These authors contributed equally to this work.
*
20 E-mail: Jian-Hao Chen (chenjianhao@pku.edu.cn)
21
22 Table of Contents:
23 S1. , ( ) of a couple of MnPS3 magnon valves at 2K
24 S2. , ( ) of a typical MnPS3 magnon valve at various temperatures
25 S3. Spin model and spin wave modes of MnPS3 under an in-plane magnetic field
26 S4. Spin Seebeck effect in MnPS3
27 S5. Quantitative analysis of the simulation
28 S6. Comparison between experimental and simulated and
29 S7. Comparison between our work and recent work on CrBr3
30 S8. The simulated crossing point I0 vs. injection current Iin
31 S9. Nonlocal magnon signal detection with different MnPS3 device geometries
32 S10. Absence of the anomalous Nernst effect in MnPS3 magnon valve
33 S11. The heater-detector distance-dependent signal and temperature in MnPS3 device
34 S12. Operation of a MnPS3 magnon valve with different device geometries
35 S13. R2ω vs. Iin for different MnPS3 devices with zero gate current
36 S14. V2ω vs. Iin for different MnPS3 devices with zero gate current
37 S15. Stability test of few-layer MnPS3 crystals and devices

1
38 S1. , ( ) of a couple of MnPS3 magnon valves at 2K

39
40 Fig. S1. , ( ) of a couple of MnPS3 magnon valves at 2K. , ( ) of two more
41 MnPS3 magnon valves at 2K and at different in-plane magnetic fields.

2
42 S2. , ( ) of a typical MnPS3 magnon valve at various temperatures.

43
44 Fig. S2. , ( ) of a typical MnPS3 magnon valve at various temperatures. It can be
45 seen that the higher the temperature, the weaker the , as well as its response to . For
46 temperature higher than 20 K, , as well as its dependence on completely disappear,
47 consistent with Fig. 2b in the main text.
48

49 S3. Spin model and spin wave modes of MnPS3 under an in-plane magnetic field
50
51 In this section, we propose a two-dimensional localized spin model with easy-axis single-ion
52 anisotropy to describe the antiferromagnetic insulator MnPS3. We carry out spin-wave analysis
53 of the Ising anti-ferromagnet under a transverse magnetic field to obtain low-energy magnon
54 excitations.
55
56 Magnetism for the antiferromagnetic insulator MnPS3 is described by a localized spin model
57 with an easy-axis single-ion anisotropy.

3
, , , ,
58 =  ∑ ∑ , , ⋅ − ∑ + − ℎ∑ + (S1)

59
60 Here is an antiferromagnetic exchange coupling and is the easy-axis single-ion anisotropy.
61 denotes a honeycomb-lattice A-sublattice site, and ( = 1,2,3) connects a A-sublattice site
62 and its neighboring three B-sublattice site in the honeycomb lattice, with = (1,0),   =
√ √ , , ,
63 − , , = − ,− . ≡ , , is a localized spin of Mn atom (S=5/2)[1]

64 in the A-sublattice site ( ) and is a localized Mn spin at the B-sublattice site( + ). For
65 simplicity, we consider that the system is a two-dimensional magnet and ℎ is an in-plane field
66 along -direction.
67
68 Under the in-plane field, the antiferromagnetic moment will be deformed linearly in the field:
69 = (sin , 0 cos ), = (sin , 0, − cos ).
70 A canting angle is determined classically as a minimum of a classical magnetic energy:
71 = (3 (sin − cos )−2 cos − 2ℎ sin ),
72 where is a number of the A-sublattice sites. The minimum energy is given by:

73 sin = (S2)
( )

74 Magnetic collective excitations around the classical magnetic order are described by
75 Holstein-Primakoff bosons;
, , , , ,
= − , − = √2 , + = √2
, , , , ,
=− + , + = √2 , − = √2
, , , , , ,
76 where , , and , , are the spin operators in a rotated frame;
, , , ,
cos 0 − sin cos 0 sin
, , , ,
77 = 0 1 0 , = 0 1 0
, sin 0 cos , , −sin 0 cos ,

78 and are Holstein-Primakoff boson fields for A-sublattice Mn spin and B-sublattice Mn spin,
79 that represent fluctuations around the classical magnetic order. Around those that minimize the
80 classical magnetic energy, the Hamiltonian is stable against such small fluctuations:
≡ + + ( , )
81

4
82 That says, a spin-wave Hamiltonian is given by a quadratic form in the boson fields:

= cos 2 + + +
2
, ,

+ + − − − (
2 2
, ,

+ + + + + + + ) + (3

+2 ) +

83 Here equation (S2) is used for replacing ℎ by and . In the momentum space, the spin-wave
84 Hamiltonian reads,


=


= σ ⊗τ + σ ⊗ (cos φ τ − sin φ τ ) − σ ⊗τ + σ
⊗ (cos φ τ − sinφ τ )
85 with = + − sin , = | ( )|(−1 + cos 2 ), = sin ,

86 = | ( )|(1 + cos 2 ), | ( )| and φ are the modulus and phase of

87 ( )=∑ , , ≡ | ( )| respectively, with φ = −φ .


88
89 Under

1 /
0 1 −1
=σ ⊗ /
,
√2 0 1 1

cosh −sinh , cosh sinh ,


90 = , = ,
−sinh cosh , sinh cosh ,

91 The spin-wave Hamiltonian is diagonalized as,


92
93

5
94 = ∑ (ℏω ( )γ , γ , + ℏω ( )γ , γ , ).
95
96 Here cosh =ℏ ( )
, sinh = ℏ ( )
, cosh = ℏ ( )
, sinh =ℏ ( )
, ℏω ( ) =

97 ( + ) −( − ) , ℏω ( ) = ( − ) −( + ) , with ω ( ) < ω ( ).
98
99 The two spin-wave energy-momentum dispersions along high symmetric momentum line
100 are plotted in Fig. S3(A).

101
102 Fig. S3. Spin model and spin wave modes of MnPS3 under an in-plane magnetic field. (a)
103 The dispersions relation of spin waves along high symmetry directions of the MnPS3 crystal
104 when the in-plane nearest neighbor coupling J = 0.77meV, the magnetic anisotropy D =
105 0.0086meV and an in-plane magnetic field along the x direction of 4T. (b) Illustration of the high
106 symmetry directions in the Brillouin zone of Mn honeycomb lattice in MnPS3.
107

108 S4. Spin Seebeck effect in MnPS3


109
110 Based on a semi-classical Boltzmann transport theory of magnon in anti-ferromagnet2, we
111 will give an expression for spin Seebeck coefficient of MnPS3 at finite temperature and finite
112 magnetic field in this section.
113
114 The spin density along the field direction is given by magnon creation and annihilation
115 operators:
116

6
, , , , , ,
+ = cos + sin + cos − sin

, , , ,
117 = sin ∑ − + cos ∑ +

118 = sin ∑ (2 − − ) + cos ∑ ( + + + ). (S3)

119
120 The second terms in the last line are linear in the magnon creation or annihilation operators,
± ℏ ( )
121 so that they are time-dependent with a factor of , . Within the experimental
122 measurement resolution, these contributions must be averaged to the zero. Besides, the initial
123 phase in the time-dependent factor are equally distributed with 0,2 , because the magnon are
124 supposed to be thermally activated under the injector. After being averaged over the initial , the
125 contributions must be zero too. Thus, we consider the contribution of the first term in equation
126 (S3):
127
, , , ,
+ = sin − = sin (2 − − )

128
129 A deviation of the spin density from its classical value is given by the magnon density:
130
, ,
≡ + − 2 sin = −sin ( + )

= −sin +

= −sin sinh , , + cosh , , + sinh , ,


2 2 2

+ cosh , ,
2
131 = −sin ∑ sinh + sinh + cosh , , + cosh , , . (S4)

132
133 In the third line of equation (S4), we omitted and , which are time-dependent with a
± ℏ ( )
134 factor of , , for the same reason as in equation (S3). The omitted terms are time-

7
135 dependent with a factor of ± ⋯( )
, whose contributions must be averaged to zero in the
136 experiments. The first two terms in the last line represent the spin-wave correction to the
137 antiferromagnetic moment. We thus include them into 2SNsin in the left hand side, and
138 redefine the deviation of the spin density as:
139
, ,
140 ≡∑ + − 2 sin + sin ∑ sinh + sinh

= −sin cosh , , + cosh , ,

141
142 Note that the same signs in angular momenta of the two magnon modes; cosh > 0,
143 cosh > 0. Since the two magnon modes have different group velocity ( )=∇ ( )
144 ( = 1,2), the spin current density operator is:
145

= − sin ( ) cosh , , + ( ) cosh , ,

146
147 We follow the same argument as Ref.[2], to obtain the spin Seebeck coefficient2-5:
148
= ∙ T

ℏ ( )/
ℏ ( )
149 ( )= ∑ , ( ) ( )cosh
( ) ℏ ( )/
,

150
151 where , = 1⁄ , is the magnon relaxation time for the jth magnon branch and at magnon
152 momentum k and the dispersion relation of the two magnon branches are:
153
154 ℏ ( )= ( + ) −( − ) ,

155 ℏ ( )= ( − ) −( + ) ,
156
157 Here,

8
158 = + − sin , = | ( )|(−1 + cos 2 ),

159 = sin , = | ( )|(1 + cos 2 ),

160 ( )=∑ , , , ( )= ( ),

161 cosh = ℏ ( )
, cosh =ℏ ( )
, sin = ( )
,

√ √
162 = (1,0), = − , , = − ,− .

163 ( ) ( ) is generally in the tensor-form and so is the Seebeck coefficient . When considering

164 only small magnons around the point where ( )∝ , ( ) ( ) can be replaced by a
165 scalar quantity:

( ) ( )≈1 × ( )
(2π) 6π
166 Since

( )=
, ,

167 We can expand f(k) around the point:


( )
( )= = 1+ + +Ο
2
, , , ,

√ √
168 =3+ ( + + )− + − + + + +

169 +Ο

170 Because + + =0

171 So ( ) = 3 − + + + + +Ο ≈ 3−

172 With the above approximation, we are ready to simulate , as2,6,7:

, = ∗ + ∗ =2 + +

173 where, as stated in the main text, C is an overall constant, is the relative strength between
174 injector and gate electrode, is the conversion efficiency from electrical current to temperature
175 and … means taking the second harmonic component.

9
176
177 Fig. S4. Spin Seebeck effect in MnPS3. (a) the simulated time-averaged inverse spin Hall
178 voltage VISHE (black curve), the (1,1) component of the Seebeck coefficient tensor S (red curve),
179 and the temperature gradience (blue curve) as a function of the square of the input current I2

10
180 (proportional to the input power). (b) time dependent oscillations of AC injector current Iin (black
181 curve) and the corresponding time dependence of VISHE (red curve). (c) frequency distribution of
182 the AC injector current (black curve) and the VISHE (red curve). The values of the parameters ,
183 , used in producing these figures follow those found in Fig. 4 in the main text.
184
185 One can also get a better understanding of the physical process by simulating the functional
186 dependence of , ( ) and ∇ on a general input current I2 as can be found below in Fig.
187 S4a. Under an AC injector excitation Iin of frequency 18.07Hz (as used in the experiment) and
188 zero gate current, the temporal dependence and the frequency distribution of are shown in
189 supplementary Fig. S4b and S4c, respectively. We can see that in this case, the second harmonic
190 component , dominates the response, with a small fourth harmonic component.
191

192 S5. Quantitative analysis of the simulation


193 As discussed in the main text and in supplementary section S3 and S4, the simulation of
194 , can be achieved via equation (4) with only three global parameters. Root means square
195 deviation between the simulated curves and the experimental curves is calculated, and
196 minimizing this deviation gives us the best values of the three global parameters: = 1.05 ×
197 10 V ∙ s/ℏ, = 0.25 and = 1.69 × 10 K/μA for device 2 shown in Figure 4 in the main
198 text and = 1.14 × 10 V ∙ s/ℏ , = 0.3 and = 1 × 10 K/μA for device 1 shown in
199 Figure 3 (see the simulation and experimental data in Figure S5).
200
201 All the three parameters are complex functions of multiple factors including device
202 dimensions, anisotropic heat conductance, specific heat of all the materials in the device,
203 interfacial heat conductance, magnon-phonon interaction, external magnetic field, etc., just to
204 name a few. It is beyond the main point of this article, e.g., the experimental realization of a van
205 der Waals magnon valve, to exhaust all the possible factors that influence these three parameters,
206 but we can still find that three parameters are highly reasonable and within the physical range
207 with the following short discussion.
208

11
209 First of all, α is the ratio of the effective strength of the injector electrode over the gate
210 electrode in terms of their influence to the detector-MnPS3 interface. This ratio is device
211 dependent, mainly depending on the device structure, the channel length and thickness, as well
212 as the material of the channel and the electrodes. The overall effect of these factors can be
213 experimentally measured in such an experiment: using the same detector, we can measure 1) the
214 , caused by an AC current through the injector (with the gate electrode floating) and 2) the
215 , caused by an AC current through the gate electrode (with the injector electrode floating).
216 We called the former signal , and the later signal , . Since gate is closer to the
217 detector than that of the injector, we can expect α < 1, which indeed is the case for all our
218 measurement using such device configurations. Quantitatively, we have carried out the before
219 mentioned experiment at device 1 (also shown in Figure 3), at 9T and 2K, with an AC current
220 100μA, and we obtained , =19.28μV and , = 29.92μV, so that , /
221 , =0.6442. Note that for device 1, we have = 1.05 × 10 V ∙ s/ℏ , = 0.3 and
222 = 1 × 10 K/μA . We can put into the model the two sets of current parameters: 1) =
223 100μA and = 0μA and 2) = 0μA and = 100μA ( ) . The simulated second
224 harmonic detector signal is , = 19.28μA and , = 30.06μA , leading to , /
225 , = 0.6414, which agrees to the experimental value of 0.6442 very well. This shows the
226 validity of the value of parameter α.
227
228 Now we turn to the discussion of parameter β. We can model β to be:

=

229 where R is the resistance of the electrode (injector and gate), is a characteristic time for the
230 system to achieve static conditions, is the specific heat of the MnPS3 crystal, m is the mass
231 of the crystal. Take the discussion on device 1 above, β = 1×10-4K/μA2, R is measured to be
232 1000Ω, the density and volume of MnPS3 crystal is = 2.916 × 10 kg/m and
233 ~20nm (thick) × 30μm(length) × 5μm(width) = 3 × 10 m , respectively, so the mass
234 = ∗ = 8.75 × 10 mg, = 0.5 J · mol K [8]. If we consider the thermal mass of
235 the Pt electrodes on the crystal, a rough estimation reveals that the = 3.22 × 10 mg, with
236 = 2.78 J · mol K [9]. We conclude that would fall in the range of 4 × 10 to

12
237 1.2 × 10 , much smaller than the time constant of ~0.06s with respect to the measurement
238 frequency 18.07Hz. This shows that the value of parameter β is within reasonable range.
239
240 The overall parameter C in equation (4) includes more factors in the physical process than
241 the parameters α and β, for example mixing conductance between MnPS3 and Pt. Note that the
242 value of C includes the dimensionless prefactor of the magnon relaxation time 1/ , (which is
243 of order 10-8) after summing contributions from all sublattice j and momentum k in the magnon
244 Brillouin Zone, which brings in various effects including crystal quality, temperature and
245 magnetic field. The fact that the C value from different MnPS3 devices are close to each other
246 shows that the device quality in our study is consistent.

247
248 Fig. S5. Quantitative analysis of the simulation. , ( ) of the MnPS3 magnon valve
249 device 1 at 2K and 9T, together with the simulation with the three parameters in equation (4) to
250 be = 1.14 × 10 V ∙ s/ℏ, = 0.3 and = 1 × 10 K/μA .
251
252 Finally, we discuss the possibility to further reduce the parameters in the simulation one by
253 one:

13
254 1) The overall parameter C in equation (4) mainly includes factors that relates the spin
255 current injected into the detector Pt electrode to the voltage generated in the electrode; it also
256 includes the dimensionless prefactor of the magnon relaxation time 1/ , after summing
257 contributions from all sublattice j and momentum k in the magnon Brillouin Zone, which brings
258 in various effects including crystal quality, temperature and magnetic field. Thus this parameter
259 is difficult to be removed yet very easily determined via fitting to an experimental curve.
260
261 2) The parameter β characterizes the effectiveness for the injection and gating current to be
262 converted to temperature gradience in the device. In principle it could be obtained from
263 parameters such as the resistivity of the Pt strips, the thermal conductivity of the Pt strips and the
264 Au electrodes that are connected to the Pt strips (the 200nm Pt strips are connected to Cr/Au
265 electrodes which extended into bonding pads for the devices), the thermal conductivity of the
266 MnPS3 crystal, as well as thermal resistivity of all the material interfaces. Considering the
267 complexity of all these external parameters and our goal to derive a predictive effective model,
268 the parameter β is best to be determined via fitting to an experimental curve.
269
270 3) The parameter α is the ratio of the effective strength of the injector electrode over the gate
271 electrode in terms of their influence to the detector-MnPS3 interface. This ratio is also device
272 dependent, mainly depending on the device structure, the channel length and thickness, as well
273 as the material of the channel and the electrodes. If equation (4) is a linear function of the
274 temperature gradience, parameter α could be the easiest to be obtained from additional tests, such
275 as the one we did above in this section (e.g., measure V2ω for = 100μA and = 0μA and
276 then measure V2ω again for = 0μA and , = 100μA). However, since the Seebeck
277 coefficient S is an integral function that contains the temperature, and it is highly non-linear as
278 shown in Supplementary Figure S4a, thus there is no simple relation between , / ,

279 and α. Accordingly, it is still the most straightforward way to use three parameters in the
280 simulation.
281
282
283
284

14
285 S6. Comparison between experimental and simulated and
286
287 For low AC current frequency, e.g. when the frequency of the injection current is
288 much lower than either the magnon frequency (~THz) and the response rate of the temperature of
289 the device from the application of current (~GHz, see section S5), we can consider the magnon
290 generation and detection process to be a non-equilibrium and static process. In such a process,
291 the application of AC and DC current, the local change of sample temperature and the
292 appearance of a detector signal can be considered instantaneous. Thus when and is
293 applied to the sample, the temperature T of MnPS3 below the detector Pt can be approximated as:
≈ + + ,∇ = +
294 So we have:
∝| (T) ∙ | = , , ∙ ( , , )
295
296 Thus the first harmonic signal and second harmonic signal can be express as the following
297 integral equations:

, ∝ , ∝ , , ∙ ( , , )

298

, ∝ , ∝ , , ∙ ( , , )

299 From the above integral equations, we found finite but vanishingly small , which agrees
300 with the physics of thermal magnon excitation as well as our experimental measurements. See
301 Figure S6 below for our experimentally measured first harmonic and second harmonic signal as a
302 function of the angle of the in-plane magnetic field.
303

15
304
305 Fig. S6. Comparison between experimental and simulated and . Magnetic field
306 angular dependence of and for the MnPS3 magnon valve devices. The magnitude of in-
307 plane magnetic field is 9T and the temperature of the sample is at 2K. Here no gate current is
308 applied and only an injection current of 100μA is applied to the injector. is a constant offset
309 voltage which accounts for the parasitic first harmonics signal that has no magnetic field
310 dependence. The likely source of such signal includes parasitic inductance or capacitance
311 between the measurement wires.
312
313 S7. Comparison between our work and recent work on CrBr3
314
315 First of all, the channel materials are very different. MnPS3 is a layered anti-ferromagnet
316 with Ising-type anti-ferromagnetic coupling in the sample plane, while CrBr3 is a 2D ferromagnet.
317 This means that there is zero magnetic moment in each layer of MnPS3, while each layer of
318 CrBr3 is magnetic. Such difference reflects strongly in the magnon spectra of the two materials
319 [PRX 8, 011010 (2018), PRB 103, 024424 (2021)], and also reflects strongly in the existence of
320 the magnetization at the Pt-CrBr3 interface and the absence of which at the Pt-MnPS3 interface,
321 resulting in the observation of a large anomalous Nernst signal in CrBr3 (PRB 101, 205407) and
322 the absence of which in our work.

16
323 Second, the R2ω vs. Iin curves are very different. We have plotted the R2ω vs. Iin in Figure S13
324 for a couple of MnPS3 devices below. In order to compare with the work on CrBr3, the gate
325 electrodes in our devices are floating during the measurement. It can be seen that the shape of the
326 R2ω vs. Iin of MnPS3 device is very different from that of the CrBr3 device shown in PRB 101,
327 205407. Interestingly, the shape of the R2ω vs. Iin for CrBr3 device has some resemblance with
328 the V2ω vs. Iin for our MnPS3 devices, which would be a good topic for future works.
329
330 Third, the devices are very different. Our work realized the first diffusive magnon valves in
331 which a gate current controls whether the injected signal can be detected or not, which readily
332 enables digital logic operation; PRB 101, 205407 (2020) describe a non-local response curve for
333 the input signal without any external gate control.
334
335
336 S8. The simulated crossing point I0 vs. injection current Iin

337
338 Fig. S8. The simulated crossing point I0 (the first Igate value for V2ω = 0) vs. injection
339 current Iin. The simulation is based on parameters obtained from fitting to experimental data
340 shown in Figure 4 in the main text.
341
342
343
344
17
345 S9. Nonlocal magnon signal detection with different MnPS3 device geometries
346
347 Section S9 to S12 contain additional experimental evidences that excluded the possibility of
348 local spin Seebeck effect as well as anomalous Nernst effect in the MnPS3 magnon valves.
349
350 First of all, to clarify whether the heat is carried by phonon or by magnon, we have
351 fabricated a non-local device with a number of electrodes on MnPS3. As depicted in Fig. S9
352 below, we name four of the electrodes as Detector 1, Injector, oxidized Cu strip and Detector 2,
353 respectively. All electrodes are made from Pt except for the oxidized Cu strip. The oxidized Cu
354 strip is made from 10nm thick copper without any protection capping layer and then is exposed
355 to ambient condition for oxidation. We have made sure that the Cu strip was conductive right
356 after the deposition and not conductive after oxidation. The intention of the oxidation is to reduce
357 the thermal conductivity of the copper strip to 4 W/m*K[10], so that it is much lower than a Pt
358 electrode in terms of thermal conductivity (72W/m*K for Pt). The oxidized copper strip merely
359 acts as surface absorbates which only affect the top surface of MnPS3 and would not act as a
360 strong heat sink. In another word, the oxidized Cu strip should perturb magnon transport much
361 more than phonon transport in MnPS3, since the in-plane to out-of-plane ratio of magnetic
362 coupling strength is 405:1 [1] while the in-plane to out-of-plane ratio of thermal conductivity is
363 only 6:1 [11]. The strong in-plane versus out-of-plane anisotropy in the magnetic exchange
364 suggests that the Pt detectable magnon transport goes through only a few top layers of the sample,
365 while the phonon transport generally goes through the whole layers of the sample. Being only the
366 surface absorbates, the oxidized copper strip is expected to perturb the magnon transport
367 dramatically, while the phonon transport remains robust against such perturbation.
368
369 An AC signal is applied through the Injector electrode, and the signal is measured
370 simultaneously from Detector 1 and Detector 2. We found strong signal from Detector 1 (right
371 next to the Injector) and no signal from Detector 2 (the oxidized copper strip is between Detector
372 2 and the Injector). Since the temperature gradient between Injector and Detector 2 should be
373 finite as the case for Detector 1, which is confirmed by finite element analysis (see Fig. S9c).
374 The absence of the non-local inverse spin Hall signal from Detector 2 proves that phonon
375 transport is not the cause of the non-local signal from the Detector electrode.

18
376
377 Fig. S9. Nonlocal magnon signal detection with different MnPS3 device geometries. (a)
378 Schematic of nonlocal measurement on a MnPS3 device with oxidized Cu strip. (b) Temperature
379 dependence of V2ω at θ = 0 for the detector located at the left or right side of the oxidized Cu
380 strip. (c) Finite element analysis of the temperature distribution in MnPS3 device with oxidized
381 Cu strip.
382
383 S10. Absence of the anomalous Nernst effect in MnPS3 magnon valve
384
385 To quantify the effect of the anomalous Nernst effect in our experimental system, we have
386 measured the non-local second harmonic signal with an applied magnetic field of up to 14 T
387 rotated in the x-z plane (see inset in Fig. S10 below). Since there is finite temperature gradient
388 along the x axis from our device configuration as shown from the finite element analysis, the
389 temperature gradient along x also induces the Hall voltage along the y axis in the presence of the
390 magnetization along the z axis (ANEx). It’s considered that ANEx and ANEz are of similar

19
391 magnitude12, where ANEz refers to the Hall voltage along y induced by the temperature gradient
392 along z in the presence of the magnetization along the x axis.
393
394 The angle of the magnetic field with respect to the z axis is marked as φ. An injection
395 current of 100 μA is applied to the injector of our MnPS3 device. We can see from Fig. S10a that
396 the data fits well to a sinφ function, in which the signal is zero when the magnetic field is along
397 the z axis (perpendicular to the sample plane). From Fig. S10a one can also see that only the
398 magnetic field component along the x axis could produce non-zero non-local second harmonic
399 signal. Figure S10b shows minimal magnetic-field dependence of the non-local signal with the
400 magnetic field along the z axis (i.e., φ=0). This data proves unambiguously the absence of
401 anomalous Nernst effect with magnetic field perpendicular to the MnPS3/Pt interface (ANEx),
402 because the finite element calculation shows a finite temperature gradient along x. The absence
403 of ANEx points to the absence of ANEz12.
404
405 In fact, MnPS3 is a layer antiferromagnet where the spin within one Mn atomic layer is
406 aligned antiferromagnetically with a coupling constant J that amount to about 106T of magnetic
407 field, which far exceeds the magnetic field applied in the experiment. It is natural that the
408 MnPS3/Pt interface remains non-magnetized. We have added this data in Supplementary Figure
409 S10, with a short description.

410
411 Fig. S10. The absence of the anomalous Nernst effect in MnPS3 magnon valve. (a) The non-
412 local second harmonic signal as a function of angle φ between the external magnetic field (B=9T)
413 and the z direction, angle φ is determined the same as shown in the inset.(b) The absence of
414 magnetic field dependence of V2ω at φ = 0.

20
415 S11. The heater-detector distance-dependent signal and temperature in MnPS3 device.
416
417 We have measured the distance dependence of V2ω,0 (black dots in Fig. S11) and the
( / )
418 experimental data is fitted to = ∗ (
[13], where is a factor characterizing the
/ )

419 magnitude of the second harmonic signal, λ is the decay length of the diffusive magnons. The
420 fitting gives =−2.4±0.2μV·μm and λ = 3300nm±200nm, which is consistent with previous
421 report in the literature (e.g. λ~2800 nm for 16-nm MnPS3, 1100 nm for 8-nm MnPS3 [14]). As
422 shown in Fig. S11, this decay length (red curve) is much longer than a decay length of the
423 temperature gradient from the finite element calculation (blue curve) that represents how far the
424 phonon carries the heat in space. Thus, the longer decay length in the heater-detector distance
425 dependence suggests that the signal cannot be explained by the phonon transport.
426

427
428 Fig. S11. The heater-detector distance-dependent signal and temperature in MnPS3 device.
429 Left axis: The distance-dependent non-local second harmonic signal V2ω,0 and relaxation length
430 fitting13 at 2K and 9T; Right axis: The finite element simulation of distance-dependent
431 temperature of MnPS3 under the same device configuration. The decay length of the second
432 harmonic thermal magnon signal is measured to be 3300nm±200nm.
433
434
435

21
436 S12. Operation of a MnPS3 magnon valve with different device geometries
437
438 Two different device geometries, namely, the injector-gate- detector configuration as well as
439 the injector-detector-gate configuration is tested. We found that in the injector-detector-gate
440 configuration, the general behavior is similar (see Fig. S12b). Finite element analysis shows that
441 the variation of the temperature gradient is also similar for the gate located at the right or left side
442 of the detector electrode (see Fig. S12c&d), while the temperature of the MnPS3 below the
443 detector electrode is slightly higher for the case of the injector-detector-gate configuration, which
444 may be the cause of the slight difference in V2ω vs. Igate observed experimentally.
445
446 The parameters used in the finite element analysis are listed below:
COMSOL Material
Conductivity 8.9E6[S/m]
database
Pt
COMSOL Material
thermal conductivity 71.6[W/(m*K)]
database
in-plane thermal ACS Nano,14,
6.3[W/(m*K)]
conductivity 2424−2435(2020)
MnPS3
through-plane thermal ACS Nano,14,
1.1[W/(m*K)]
conductivity 2424−2435(2020)
CRC Handbook of
SiO2 thermal conductivity 1.38[W/(m*K)] Chemistry and Physics
(92nd ed.).p12.213
COMSOL Material
Si thermal conductivity 130[W/(m*K)]
database
through-plane thermal Computational Materials
MnPS3/SiO2 5E-7[K*m2/W]#
resistance Science, 142, 1–6 (2018)
through-plane thermal PHYSICAL REVIEW B
Pt/MnPS3 1.4E-7[K*m2/W]⸹
resistance 101, 205407 (2020)
447
448 Table S1. The parameters used in the finite element analysis. #There is no data found for
449 MnPS3/SiO2 in the literature, we used value from through-plane thermal resistance between
450 MoS2/SiO2 instead. ⸹There is no data found for Pt/MnPS3, we used estimated value for CrBr3/Pt
451 in the literature.
452

22
453
454 Fig. S12. Operation of a MnPS3 magnon valve with different device geometries. (a)
455 Schematics of nonlocal measurement on a MnPS3 device with different gates. (b) V2ω,0 versus
456 DC gate current Igate at B = 9T and temperature of 2K with different geometries. (c)(d) Finite
457 element analysis of the temperature distribution in MnPS3 device for the gate located at the left
458 (c) or right (d) side of the detector electrode.
459
460
461
462
463

23
464 S13. R2ω vs. Iin for different MnPS3 devices with zero gate current

465
466 Fig. S13. The R2ω vs. Iin for different MnPS3 devices with zero gate current.
467
468
469
470
471
472

24
473 S14. V2ω vs. Iin for different MnPS3 devices with zero gate current

474
475 Fig. S14. The V2ω vs. Iin curves for the same set of MnPS3 devices as shown in Fig. S13 with
476 zero gate current.
477
478
479
480
481

25
482 S15. Stability test of few-layer MnPS3 crystals and devices

483
484 Fig. S15. Stability test of few-layer MnPS3 crystals and devices. (a) the optical micrograph of
485 few-layer MnPS3 on SiO2 substrate right after exfoliation and 8 months after exfoliation. (b)
486 Optical micrographs of few-layer MnPS3 on SiO2 substrate before heating and after heating to
487 150, 250 and 350℃ for 10 minutes in air. (c) The device performance of our MnPS3 magnon
488 device right after fabrication and after 8 months.

26
489 Reference
490 1. Wildes, A. R., Roessli, B., Lebech, B. & Godfrey, K. W. Spin waves and the critical
491 behaviour of the magnetization in MnPS3. J. Phys.: Condens. Matter 10, 6417-6428
492 (1998).
493 2. Rezende, S. M., Rodriguez-Suarez, R. L. & Azevedo, A. Theory of the spin Seebeck
494 effect in antiferromagnets. Phys. Rev. B 93, 014425 (2016).
495 3. Poonja, S., Patel, S., Henry, L. & Roorda, A. Dynamic visual stimulus presentation in an
496 adaptive optics scanning laser ophthalmoscope. Journal Of Refractive Surgery 21, S575-
497 S580 (2005).
498 4. Cornelissen, L. J. et al. Magnon spin transport driven by the magnon chemical potential
499 in a magnetic insulator. Phys. Rev. B 94, 014412 (2016).
500 5. Duine, R. A., Brataas, A., Bender, S. A. & Tserkovnyak, Y. in Universal Themes of Bose-
501 Einstein Condensation Ch. 26, 505-524 (2017).
502 6. Sanders, D. J. & Walton, D. Effect of magnon-phonon thermal relaxation on heat
503 transport by magnons. Phys. Rev. B 15, 1489-1494 (1977).
504 7. Agrawal, M. et al. Direct measurement of magnon temperature: new insight into
505 magnon-phonon coupling in magnetic insulators. Phys. Rev. Lett. 111, 107204 (2013).
506 8. Takano, Y. et al. Magnetic properties and specific heat of MPS3 (M=Mn, Fe, Zn). J.
507 Magn. Magn. Mater. 272, E593-E595 (2004).
508 9. Boerstoel, B. M., Zwart, J. J. & Hansen, J. Specific Heat of Palladium, Platinum,Gold
509 and Copper Below 30k. Physica 54, 442-458 (1971).
510 10. Kusiak, A. et al. CuO thin films thermal conductivity and interfacial thermal resistance
511 estimation. European Physical Journal-Applied Physics 35, 17-27 (2006).
512 11. Kargar, F. et al. Phonon and Thermal Properties of Quasi-Two-Dimensional FePS3 and
513 MnPS3 Antiferromagnetic Semiconductors. ACS Nano 14, 2424-2435 (2020).
514 12. Liu, T. et al. Spin caloritronics in a CrBr3-based magnetic van der Waals heterostructure.
515 Phys. Rev. B 101 (2020).
516 13. Cornelissen, L. J. et al. Long-distance transport of magnon spin information in a
517 magnetic insulator at room temperature. Nat. Phys. 11, 1022-1026 (2015).
518 14. Xing, W. Y. et al. Magnon Transport in Quasi-Two-Dimensional van der Waals
519 Antiferromagnets. Phys. Rev. X 9, 011026 (2019).
520

27

You might also like