Articles

[Link]

Imaging the Néel vector switching in the

monolayer antiferromagnet MnPSe3 with
strain-controlled Ising order
Zhuoliang Ni1, A. V. Haglund 2, H. Wang3, B. Xu4, C. Bernhard 4
, D. G. Mandrus2,5, X. Qian 3
,
E. J. Mele1, C. L. Kane1 and Liang Wu 1 ✉

Antiferromagnets are interesting materials for spintronics because of their faster dynamics and robustness against perturba-
tions from magnetic fields. Control of the antiferromagnetic order constitutes an important step towards applications, but has
been limited to bulk materials so far. Here, using spatially resolved second-harmonic generation, we show direct evidence of
long-range antiferromagnetic order and Ising-type Néel vector switching in monolayer MnPSe3 with large XY anisotropy. In
additional to thermally induced switching, uniaxial strain can rotate the Néel vector, aligning it to a general in-plane direction
irrespective of the crystal axes. A change of the universality class of the phase transition in the XY model under uniaxial strain
causes this emergence of strain-controlled Ising order in the XY magnet MnPSe3. Our discovery is a further ingredient for com-
pact antiferromagnetic spintronic devices in the two-dimensional limit.

D
etection and control of the spin order in ferromagnetic both a Landau theory and a microscopic model that couples strain
materials is the main principle in current information tech- to nearest-neighbour exchange, we conclude that the phase transi-
nology. Despite the great advantages of antiferromagnetic tion of the XY model in the presence of strain falls into the Ising
(AFM) spintronics, controlling and detecting Néel vectors have been universality class instead of the XY class, which could explain the
limited in bulk materials1–5. The discovery of two-dimensional (2D) strain tunability. Finally, we find that the 180° AFM domain walls
ferromagnetic materials using the polar Kerr effect6,7 has triggered are highly mobile down to the monolayer after thermal cycles, pav-
tremendous interest in studying magnetism in the true 2D limit8–11 ing the way for future control of the antiferromagnetic domains by
and spintronic device applications in van der Waals heterostruc- strain or external fields on demand for ultra-compact 2D AFM tera-
ture materials12–18. Optical techniques are powerful tools to detect hertz spintronics.
magnetism19,20, but clear evidence for the AFM order in atomically
thin 2D crystals has not been identified due to the lack of sensi- Detecting AFM order and imaging AFM domains in thick
tive direct detection. For example, the polar Kerr effect is absent flakes
in AFM materials when the total magnetization is zero17,19,20. And, MnPSe3 belongs to the family of AFM transition metal phosphorous
although Raman spectroscopy is a powerful tool to study spin–pho- trichalcogenide MPX3 (M = Mn, Ni, Fe, Co; X = S, Se), among which
non coupling and collective magnons21–24, their identification often Mn compounds form inversion-breaking Néel order while others
does not provide unambiguous identification of the AFM order19,20. are in centrosymmetric zigzag ordered phases28,32–36. In contrast to
Second-harmonic generation (SHG) has been shown to be a sensi- MnPS3, which has dominant out-of-plane moments35,37, MnPSe3 has
tive tool to detect AFM orders due to inversion symmetry breaking in-plane spins with very large XY anisotropy according to the neu-
from the spin order in magneto-electric materials including bulk tron scattering measurement28 (Fig. 1a), which offers richer mag-
Cr2O3 (ref. 25), few-layer MnPS3 (ref. 26) and a synthetic bilayer CrI3 netic domain structures such as vortices and tunability31. Above the
(ref. 27). Nevertheless, the detection of intrinsic AFM in the mono- Néel temperature, TN, MnPSe3 belongs to the point group 3 (S6) and
layer has not been demonstrated yet. In this work, we develop a space group 148 (ref. 28), and has an inversion centre between two
sensitive SHG microscope and detect long-range Néel AFM order neighbouring Mn atoms but no mirror symmetry. The Mn atoms
and Néel vector switching down to the monolayer in MnPSe3 form a honeycomb lattice in one layer (Fig. 1a), and the honeycomb
(ref. 28). Temperature-dependent SHG measurement in repetitive layers form the rhombohedral (ABC) stacking along the c axis with
thermal cooling surprisingly collapses into two curves, which corre- a three-fold rotational symmetry. Different from FePS3 and NiPS3,
spond to the switching of an Ising-type Néel vector reversed by the which have a change from a two-fold rotational symmetry in the
time-reversal operation, instead of the six-state clock ground state bulk to a three-fold rotational symmetry in the monolayer21–23,38,
expected from the three-fold rotation symmetry in the structure29–31. MnPSe3 is always three-fold symmetric.
We image the spatial distribution of the Néel vectors across samples As shown in the inset of Fig. 1b, a small temperature-independent
and rotate the Néel vectors by an arbitrary angle irrespective of the SHG from∑ the electric quadruple (EQ) contribution,
lattice in the sample plane by applying uniaxial strain. By studying IEQ
i (2ω) ∝ |
EQ 2
jkl χ ijkl Ej (ω)∇k El (ω)| follows the lattice three-fold

Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA, USA. 2Department of Materials Science and Engineering, University
1

of Tennessee, Knoxville, TN, USA. 3Department of Materials Science and Engineering, Texas A&M University, College Station, TX, USA. 4Department
of Physics and Fribourg Center for Nanomaterials, University of Fribourg, Fribourg, Switzerland. 5Materials Science and Technology Division, Oak Ridge
National Laboratory, Oak Ridge, TN, USA. ✉e-mail: liangwu@[Link]

782 Nature Nanotechnology | VOL 16 | JulY 2021 | 782–787 | [Link]/naturenanotechnology

NAturE NAnOtECHnOlOgY Articles
a b TN c
80
Parallel Crossed 120° 60°
1.55 eV 3.10 eV 80 kc.p.s. Parallel
0.4 kc.p.s. 120° 60° Crossed
1.0
60

ISHG (kc.p.s.)

dM
dT (a.u.)
180° 0° 0.5
40
180° 0°
100 K
240° 300° 0
20
c
–0.5 5K
0 300°
b 50 100 150 200 250 300 240°
a
Temperature (K)

Fig. 1 | Characterization of bulk MnPSe3 samples. a, A schematic of the in-plane AFM Néel order and SHG measurement on MnPSe3. The Mn atoms in
each layer form a honeycomb lattice. The green arrow denotes the Néel vector. Laser pulses centred at 1.55 eV (red wavy arrow) are normally incident on
the a–b plane and reflected light at the second-harmonic frequency (blue wavy arrow) is detected. The parallel and crossed configurations correspond
to E(2ω) ∥ E(ω) and E(2ω) ⊥ E(ω), respectively, while we corotate E(ω) and E(2ω) by 360° in the a–b plane. b, Temperature dependence of SHG intensity
(ISHG, red) on a 100 nm sample exfoliated on SiO2/Si, and the temperature (T) derivative of the in-plane magnetic susceptibility (M) of a bulk crystal (blue);
kc.p.s. stands for one thousand counts per second. The transition temperature is marked by the black dashed line. Inset: polarization-resolved EQ SHG of the
100 nm flake at 100 K. The dots are the data and the solid lines are the fits. c, Polarization-resolved SHG polar patterns of the same 100-nm-thick MnPSe3
flake measured at 5 K (all of the lines are experimental data). The angle in b, c is defined between the polarization of the incidence pulse and the horizontal
direction in the lab.

rotational symmetry above TN, where E is the electric field, ω is the pattern. An optical image of a region with uniform thickness
fundamental frequency, χ is the susceptivility and i,j,k,l are Cartesian (~100 nm, a second exfoliated thick sample) and SHG maps at 5 K
indices (Methods). Below the Néel temperature, the formation of after two cooling processes across TN are shown in Fig. 2d–f. Sharp
the Néel AFM order with in-plane spins breaks the inversion sym- dark lines with very low SHG intensity are observed, with bright
metry ( P ), which allows
∑ anEDelectric dipole (ED) contribution to domains of high and nearly equal SHG intensity on both sides. In
the SHG, IED i (2ω) ∝ | jk χ ijk ( L) Ej ( ω ) Ek ( ω )| 2
(ref. 25
). χ ED
ijk (L) is one of the regions, we pick out a few points such as the blue dot in
proportional to the order parameter, the Néel vector L (ref. 39), and Fig. 2e and observe the same polar patterns shown as blue in the
changes sign when L flips by 180° (L = M1 − M2, where M1 and M2 top right of Fig. 2e. Crossing the dark line to a different region, we
are the magnetization of two neighbouring Mn atoms; Fig. 1a). In pick out a few points such as the orange dot and observe the crossed
the AFM state, the product of the inversion symmetry ( P ) and the polar patterns rotated by a small angle shown as orange on the top
time-reversal symmetry ( T ), the so-called PT symmetry, is still right of Fig. 2e. By keeping the laser spot at the orange point and
preserved40, even though both P and T are broken. This kind of performing a few thermal cycles, the polar pattern switches only
ED term is often called non-reciprocal or c-type SHG allowed by between the blue and orange dots shown in Fig. 2e.
the PT symmetry, while the EQ term is an i-type SHG, where ‘c’ We interpret the two regions with high SHG intensity as two
and ‘i’ mean changing and invariant under time-reversal symmetry, different AFM domains where the spins are reversed by 180°
respectively25. Figure 1b shows a typical SHG response as a func- under the time-reversal operation, and the dark lines are domain
tion of temperature on an ~100-nm-thick flake exfoliated on SiO2/ walls due to destructive interference25,41. The arrows in Fig. 2e,f
Si. A sharp turn-on of the ED SHG signal clearly indicates a phase indicate the opposite directions of the Néel vectors in different
transition at 67.9 ± 0.2 K, agreeing well with the TN (68 ± 0.5 K) regions determined by SHG polar pattern measurements at 5 K
determined from in-plane magnetization measurement. Below TN, (Supplementary Note 1). A second SHG map after a thermal cycling
as shown in Fig. 1c, a giant two-fold signal emerges in the crossed across TN in Fig. 2f shows that the domain wall is not pinned and
configuration, which clearly breaks the three-fold rotation symme- different regions still have only the two kinds of polar patterns
try. Another surprising observation is that the peak of the paral- shown in Fig. 2e. (The mapping on a 30-μm-thick naturally grown
lel polar pattern is only 1/20 of that in the crossed pattern, which sample is shown in Supplementary Figure 6.) The reason why we
was not observed in previous PT invariant van der Waals AFM could observe the domain wall between two AFM regions with a
materials26,27. The nodal direction of the two-fold crossed pattern is π phase shift by the destructive SHG interference is that the SHG
also shown to be close to the Néel vector direction (Supplementary has both ED and EQ contributions and only the ED term is sen-
Notes 1 and 2). sitive to the π phase shift. One could write the signal we observe
After confirming the detection of AFM order by SHG, we use ∑ ∑ EQ
as Ii (2ω) ∝ | jk ± χ ED
ijk (L)Ej (ω)Ek (ω) +
2
jkl χ ijkl Ej (ω)∇k El (ω)| ,
scanning SHG microscopy with 2 μm spatial resolution to study
the AFM domains. In MnPSe3 with in-plane spins and large where the ± signs indicate the sign change of the ED term under
XY anisotropy28,37, six energetically equal magnetic domains are time-reversal operation. As shown in Figs. 1b and 2a, the ED con-
expected due to the three-fold rotational crystalline anisotropy and tribution at 5 K is ~200 times larger than the EQ part, and therefore
the time-reversal operation29–31. Nevertheless, Fig. 2a shows that the two AFM domains have high and nearly equal SHG while the
temperature-dependent SHG intensity under ten consecutive cool- domain wall has very low SHG with the EQ contribution only.
ing runs across TN on the same spot of a 15-μm-thick sample col-
lapses into two curves instead of six. The crossed and parallel polar Detecting AFM order and imaging Néel vector switching
patterns that respond to these two AFM domains are shown in Fig. down to the monolayer
2b,c. To figure out the relation between these two domains, we per- Before we discuss the origin of the two-state Ising order instead
formed spatial scanning SHG microscopy at 5 K with the angles of of the six-state clock order, we investigate whether the AFM order
the two polarizers chosen near the maximum signal in the crossed exists and whether direct imaging of Néel vector switching could be

Nature Nanotechnology | VOL 16 | JulY 2021 | 782–787 | [Link]/naturenanotechnology 783

Articles NAturE NAnOtECHnOlOgY

a d
10
Thermal cycle

1 10
ISHG (kc.p.s.)
5

0
20 40 60 80 100
Temperature (K)

b c
120° 60° 120° 60°
10 kc.p.s. 0.8 kc.p.s. f

180° 0° 180° 0°

240° 300° 240° 300°

0 40 kc.p.s.
Crossed Parallel

Fig. 2 | Ising-type Néel vector switching in thick MnPSe3 samples. a, Temperature dependence of SHG intensity at the same spot on an ~15-μm-thick bulk
crystal for ten consecutive cooling runs across TN. b,c, Polarization-dependent SHG patterns (b, crossed and c, parallel) measured at 5 K after each cooling
in a. d, An optical image of a second ~100-nm-thick MnPSe3 flake exfoliated on SiO2/Si. Scale bar, 50 μm. e,f, The SHG spatial mapping at 5 K after the first
cooling (e) and the second cooling (f) on the thick sample. Light and dark green arrows show the direction nodes in the crossed polar pattern, which is
close to the Néel vector direction. Scale bars, 50 μm. Inset: crossed patterns on two sides of a domain wall with positions marked by the orange and blue
dots. The colour bar indicates the SHG intensity.

detected in the monolayer first. The ultra-thin flakes down to the bilayer and by 18 K in the monolayer due to ageing effects when
monolayer are exfoliated on SiO2/Si wafers. The number of layers is samples were exfoliated in air (Extended Data Fig. 5).
determined by a combination of atomic force microscopy and opti-
cal contrast measurements42 (Extended Data Fig. 1). To probe the Demonstration of strain-controlled Ising order
intrinsic properties of the materials, we exfoliate samples down to We observed the same two-state Ising order on all of the samples
the monolayer in a glove box. Figure 3a shows the layer-dependent with a thickness from a monolayer to ~30 μm on different sub-
square root of the SHG intensity measured as a function of the tem- strates. Also, all of the samples show two-fold crossed polar pat-
perature. All of the thin flakes show a clear phase transition down terns, regardless of whether they are prepared in air or in a glove box
to the monolayer with the layer-dependent transition temperature (Extended Data Figs. 2–6 and Supplementary Figs. 2, 4, 7 and 9).
shown in Fig. 3b. The transition temperature decreases from 66 K Since the flakes are exfoliated on SiO2 and the bulk micrometre-thick
in the nine-layer sample to 56 K in the bilayer sample, and it is 40 K crystals are glued on the metal platform directly, a certain amount of
measured in three different monolayer samples (Extended Data Figs. strain is inevitable. Therefore, we hypothesized that the Ising anisot-
2 and 3 for the other two monolayer samples). The decrease of the ropy is induced by the strain. In order to verify it, we deliberately
transition temperature in atomically thin samples occurs because of applied an ~2% uniaxial strain by exfoliating a 15 nm flake on the
the increasing fluctuation when approaching the 2D limit, which is polymer polydimethylsiloxane (PDMS)45,46 and then stretched the
also observed in other 2D magnets such as CrI3 (ref. 6) and Fe3GeTe2 PDMS as shown in Fig. 4a. The strain strength is determined by
(ref. 9). The intensity of SHG at 5 K from one layer to nine layers measuring the length of the optical image of the sample along the
follows a nearly quadratic dependence on the layer count (Fig. 3c), elongation direction shown in Fig. 4b before and after stretching
which is expected for breaking the inversion symmetry in all of (Supplementary Fig. 12 and Supplementary Note 4). The SHG map-
these samples43,44. This is different from synthetic layered AFM CrI3, ping with ~2% strain along the x direction, shown in Fig. 4c, is quite
which supports SHG signals only with even numbers of layers27. homogeneous, and the crossed polar patterns at different positions
Figure 3d shows an optical image of the monolayer MnPSe3 all point along the same direction (Fig. 4d.), which indicates that the
sample 1 (S1). We performed twelve thermal cycles at the green dot strain aligns the Néel vector.
shown in Fig. 3d with the temperature-dependent SHG collapsed We further applied the strain along 0, 45, 90 and 135 degrees with
on two curves, and we plot four of them as examples in Fig. 3i. The respect to the x axis defined in Fig. 4b and found that the crossed
corresponding polar patterns for the two AFM states are shown in polar pattern follows the rotation of the strain as shown in Fig. 4d–g
Fig. 3j,k. Note that the change of the orientation in the crossed pat- within the experimental accuracy of ±10°, which indicates that the
terns in the monolayer is larger than that in bulk samples because Néel vector is locked to the strain. We also demonstrated the Néel
of the reduced intensity ratio between the ED and EQ terms with vector rotation by strain in a three-layer (3L) sample (Extended
reduced thickness. Figure 3e–h shows four SHG maps at 5 K after Data Fig. 7). Because PDMS is transparent and reduces colour
thermal cycles, and the green dashed lines are mobile 180° AFM contrast, the 3L sample is almost invisible on PDMS. Instead of
domain walls. All of the maps exhibit a contrast with two domains, direct straining of a monolayer on PDMS, we exfoliated a mono-
with Fig. 3e showing a dominant bright domain. Measured polar layer sample S3 with a long wavy shape on a SiO2/Si substrate to
patters on selected dots marked by red and blue in Fig. 3e–h have induce different strain directions in different regions, and find that
the same red and blue patterns shown in Fig. 3j,k (Extended Data Néel vector direction is also locked to the local strain and points
Fig. 4 for more data). We observed a drop in TN by 2–4 K in the to different directions in different regions (Extended Data Fig. 3).

784 Nature Nanotechnology | VOL 16 | JulY 2021 | 782–787 | [Link]/naturenanotechnology

NAturE NAnOtECHnOlOgY Articles
a 20 b 80 c 1,000
1L
70
×0.75 2L

ISHG ( c.p.s.)
15 3L 60
100

ISHG (c.p.s.)
4L 50

TN (K)
10 9L 40
30
10
5 20
10
0 0 1
20 40 60 80 0 2 4 6 8 10 2 4 6 8 10
Temperature (K) Layer number Layer number

d e f g h
6

ISHG (c.p.s.)
4
Monolayer
2

i j k
4 5 c.p.s. 120° 60° 0.75 c.p.s. 120° 60°
Thermal cycle

3
ISHG (c.p.s.)

2 0
180° 0° 180°
0 0°

0 240° 300°
10 20 30 40 50 240° 300°
Temperature (K)

Fig. 3 | SHG and Néel vector switching of atomically thin MnPSe3 samples exfoliated on SiO2/Si in a glove box. a, Temperature dependence of SHG
intensity on samples with different layers. b, Layer-dependent Néel temperature. c, Layer-dependent SHG intensity in a log–log plot. The solid line is a fit for
I SHG =CN2, where N is the layer number and C is a constant. d, Optical image of a monolayer sample (S1). Scale bar, 10 μm. e–h, SHG intensity mapping over
the same area in d of the monolayer sample at 5 K after different thermal cycles. The dark and light regions represent two different domains where the Néel
vector switches by 180°. The domain walls are highlighted by the green dashed lines. The Néel vectors in two domains are denoted by the green arrows.
Scale bars in e–h, 10 μm. i, Temperature dependence of SHG intensity measured at the point marked by green in d. Curves from four different thermal cycles
are shown. j,k, SHG polar patterns (j, crossed and k, parallel) for the two domains marked by red and blue, respectively, in e–h. SHG polar patterns at the
selected red and blue dots in e–h are confirmed to have the same crossed pattern with the same colour in j,k. Raw data in Extended Data Fig. 4.

We also find that the parallel polar patterns are different between angle θ0, where θ0 is along the principal strain axis, and it provides
monolayers S1 and S2. Fittings of the patterns indicate that the rela- the term us L0 2 cos 2(θ − θ 0 ). Therefore, the θ dependent terms in
tive angles between the Néel vector and the crystal axis are different the Landau free energy then take the form,
between samples S1 and S2, indicating that the strain directions in
the two samples are different (Extended Data Fig. 2). The Néel vec- F(θ ) = bus L20 cos 2(θ − θ 0 ) + eL60 cos 6θ (1)
tor could be rotated to any direction by the strain in atomically thin
MnPSe3 due to the strain-locked Ising order, which is drastically where b and e are coefficients, us is the amplitude of the strain and
different from a non-XY system, where Néel vectors are switched L0 is the magnitude of the Néel vector at the minimum free energy
between principal crystal axes only47. We also noticed that the strain (Supplementary Note 5). As shown in Fig. 5, for us = 0, in the absence
does not change the transition temperature, most likely because the of strain, this describes a six-state clock model, whose critical behav-
strain-induced anisotropy is much smaller than the large XY anisot- iour is expected to be in the XY universality class. For us ≠ 0, the
ropy in this system37 (Extended Data Fig. 8, Supplementary Fig. 13 Ising anisotropy dominates, and the critical behaviour is in the Ising
and Supplementary Note 4). universality class29,30. In general, the direction of the Néel vector will
be determined by the competition between the strain and the crys-
Origin of the Ising order in an XY magnet under strain talline anisotropy (Supplementary Note 5–6). This model explains
In order to understand why strain leads to an Ising order instead why an Ising order is also induced in the ~15-μm-thick bulk sample
of a six-state clock order, we first employ a Landau expansion for mounted by glue, as the strain is quite small.
free energy as a function of the Néel vector L, which applies close to These conclusions also follow from a more microscopic model
the critical temperature when L is small. The lowest order term that of spins on a honeycomb lattice with interactions that reflect
accounts for the three-fold crystalline anisotropy, along with PT the symmetries of the crystal. We consider an XY model with
symmetry, is L6 cos 6θ , where θ is the polar angle of the Néel vector nearest-neighbour couplings that are modified by strain and have
measured from the a axis in Fig. 1a, and L is the magnitude of the the form,
Néel vector. The strain is described by a second-rank tensor whose
principal axes designate the directions of tensile and compressive ( )
strain. It can be diagonalized with a rotation about the c axis by an H = J||,ij si,|| sj,|| + J⊥,ij si,⊥ sj,⊥ + Jc si,|| sj,⊥ + si,⊥ sj,|| (2)

Nature Nanotechnology | VOL 16 | JulY 2021 | 782–787 | [Link]/naturenanotechnology 785

Articles NAturE NAnOtECHnOlOgY

a b c

Strain direction

PDMS
Strained sample

Metal substrate x
0 2,500
ISHG (c.p.s.)

d 90° e 90° f 90° g 90°

135° 45° 135° 45° 135° 45° 135° 45°

180° 0° 180° 0° 180° 0° 180° 0°

225° 315° 225° 315° 225° 315° 225° 315°

270° 270° 270° 270°

Fig. 4 | Strain-tunable Néel vector in MnPSe3. a, A schematic of the method to control the in-plane strain. A MnPSe3 flake is first exfoliated on a PDMS
substrate. A tunable stretching force is then applied by a micro-manipulator on the PDMS to transfer strain to the sample, which is maintained by
attaching the PDMS to a gold-coated sample holder. b, An optical image of a stretched 15-nm-thick MnPSe3 flake with a uniaxial strain ~2% along the x
axis. Scale bar, 10 μm. c, SHG spatial mapping with a 2% strain applied along the x axis at nominally 5 K. Scale bar, 10 μm. d–g, Crossed polar patterns with
strain direction along 0° (d), 45° (e), 90° (f) and 135° (g) with respect to the x axis. The red arrow represents the strain direction. The measured data are
shown in blue open circles, with fits shown in blue lines.

Six-state clock MnPSe3 also provides a platform to study the Kosterlitz–Thouless

e order transition in a truly 2D XY magnet with six-state clock order if the
strain could be tuned to zero by a voltage-controlled piezo-stage49.
Additionally, our imaging and strain-tuning methods are generi-
T cally applicable to other van der Waals AFM materials including the
intrinsic AFM topological insulator50. Looking forward, we hope
3D XY
bus XY order
critical point that the discovery of strain-tunable AFM order in atomically thin
MnPSe3 will stimulate further investigations in designing spatial
Disordered
Ising
strain distribution to create spin patterns on demand for magnon
order propagation in low-dissipation terahertz spintronic devices.
3D Ising
critical point Online content
Any methods, additional references, Nature Research report-
Fig. 5 | Renormalization group flow of the Landau theory. When μs = 0, ing summaries, source data, extended data, supplementary infor-
the system flows into the three-dimensional (3D) XY critical point and mation, acknowledgements, peer review information; details of
develops the six-state clock order at low temperature. When μs ≠ 0, the author contributions and competing interests; and statements of
system flows into the 3D Ising critical point and develops the two-state data and code availability are available at [Link]
Ising order at low temperature. s41565-021-00885-5.

Received: 6 December 2020; Accepted: 2 March 2021;

where we distinguish the coupling for spin polarizations along Published online: 19 April 2021
(J∥,ij) and perpendicular to (J⊥,ij) the ij-th Mn–Mn bond; Jc is a
symmetry-allowed cross coupling between orthogonal spin compo-
References
nents s∥,ij and s⊥,ij in each bond. The total energy is proportional to 1. Wadley, P. et al. Electrical switching of an antiferromagnet. Science 351,
L2 cos(2(θ − θ o − α/2)), where α/2 defines a misalignment angle of 587–590 (2016).
the spin orientation from the principal strain axis. As long as the 2. Wadley, P. et al. Current polarity-dependent manipulation of
response is linear, decreasing the amount of strain while keeping the antiferromagnetic domains. Nat. Nanotechnol. 13, 362–365 (2018).
same direction will not change the tipping angle (Supplementary 3. Němec, P., Fiebig, M., Kampfrath, T. & Kimel, A. V. Antiferromagnetic
opto-spintronics. Nat. Phys. 14, 229–241 (2018).
Note 6 for more details). 4. Cheong, S.-W., Fiebig, M., Wu, W., Chapon, L. & Kiryukhin, V. Seeing is
believing: visualization of antiferromagnetic domains. npj Quantum Mater. 5,
Conclusions 3 (2020).
The direct imaging of Néel vector switching in the monolayer goes 5. Nair, N. L. et al. Electrical switching in a magnetically intercalated transition
metal dichalcogenide. Nat. Mater. 19, 153–157 (2020).
beyond previous works of domain imaging in bulk AFM samples2,25
6. Huang, B. et al. Layer-dependent ferromagnetism in a van der Waals crystal
and opens the possibility of ultra-compact AFM spintronics. Our down to the monolayer limit. Nature 546, 270–273 (2017).
work also creates a method to control the 2D antiferromagntism 7. Gong, C. et al. Discovery of intrinsic ferromagnetism in two-dimensional
other than electric gating15,16 and magnetic fields27,48. The monolayer van der Waals crystals. Nature 546, 265–269 (2017).

786 Nature Nanotechnology | VOL 16 | JulY 2021 | 782–787 | [Link]/naturenanotechnology

NAturE NAnOtECHnOlOgY Articles
8. Deng, Y. et al. Gate-tunable room-temperature ferromagnetism in 30. Lou, J., Sandvik, A. W. & Balents, L. Emergence of U(1) symmetry in the 3D
two-dimensional Fe3GeTe2. Nature 563, 94–99 (2018). XY model with Zq anisotropy. Phys. Rev. Lett. 99, 207203 (2007).
9. Fei, Z. et al. Two-dimensional itinerant ferromagnetism in atomically thin 31. Cheong, S.-W. SOS: symmetry-operational similarity. npj Quantum Mater. 4,
Fe3GeTe2. Nat. Mater. 17, 778–782 (2018). 53 (2019).
10. Thiel, L. et al. Probing magnetism in 2D materials at the nanoscale with 32. Wildes, A., Roessli, B., Lebech, B. & Godfrey, K. Spin waves and the critical
single-spin microscopy. Science 364, 973–976 (2019). behaviour of the magnetization in MnPS3. J. Phys. Condens. Matter. 10,
11. Chen, W. et al. Direct observation of van der Waals stacking dependent 6417 (1998).
interlayer magnetism. Science 366, 983–987 (2019). 33. Wildes, A., Rule, K. C., Bewley, R., Enderle, M. & Hicks, T. J. The magnon
12. Song, T. et al. Giant tunneling magnetoresistance in spin-filter van der Waals dynamics and spin exchange parameters of FePS3. J. Phys. Condens. Matter.
heterostructures. Science 360, 1214–1218 (2018). 24, 416004 (2012).
13. Klein, D. R. et al. Probing magnetism in 2D van der Waals crystalline 34. Wildes, A. R. et al. Magnetic structure of the quasi-two-dimensional
insulators via electron tunneling. Science 360, 1218–1222 (2018). antiferromagnet NiPS3. Phys. Rev. B. 92, 224408 (2015).
14. Wang, Z. et al. Very large tunneling magnetoresistance in layered magnetic 35. Ressouche, E. et al. Magnetoelectric MnPS3 as a candidate for ferrotoroidicity.
semiconductor CrI3. Nat. Commun. 9, 2516 (2018). Phys. Rev. B. 82, 100408 (2010).
15. Huang, B. et al. Electrical control of 2D magnetism in bilayer CrI3. Nat. 36. Lançon, D. et al. Magnetic structure and magnon dynamics of the
Nanotechnol. 13, 544–548 (2018). quasi-two-dimensional antiferromagnet FePS3. Phys. Rev. B. 94, 214407 (2016).
16. Jiang, S., Li, L., Wang, Z., Mak, K. F. & Shan, J. Controlling magnetism in 2D 37. Jeevanandam, P. & Vasudevan, S. Magnetism in MnPSe3: a layered 3d5
CrI3 by electrostatic doping. Nat. Nanotechnol. 13, 549–553 (2018). antiferromagnet with unusually large XY anisotropy. J. Phys. Condens. Matter.
17. Gibertini, M., Koperski, M., Morpurgo, A. & Novoselov, K. Magnetic 2D 11, 3563 (1999).
materials and heterostructures. Nat. Nanotechnol. 14, 408–419 (2019). 38. Kang, S. et al. Coherent many-body exciton in van der Waals antiferromagnet
18. Long, G. et al. Persistence of magnetism in atomically thin MnPS3 crystals. NiPS3. Nature 583, 785–789 (2020).
Nano Lett. 20, 2452–2459 (2020). 39. Sa, D., Valenti, R. & Gros, C. A generalized Ginzburg-Landau approach to
19. Mak, K. F., Shan, J. & Ralph, D. C. Probing and controlling magnetic states in second harmonic generation. Eur. Phys. J. B 14, 301–305 (2000).
2D layered magnetic materials. Nat. Rev. Phys. 1, 646–661 (2019). 40. Li, X., Cao, T., Niu, Q., Shi, J. & Feng, J. Coupling the valley degree of
20. Huang, B. et al. Emergent phenomena and proximity effects in freedom to antiferromagnetic order. Proc. Natl Acad. Sci. USA 110,
two-dimensional magnets and heterostructures. Nat. Mater. 19, 3738–3742 (2013).
1276–1289 (2020). 41. Yin, X. et al. Edge nonlinear optics on a MoS2 atomic monolayer. Science 344,
21. Lee, J.-U. et al. Ising-type magnetic ordering in atomically thin FePS3. Nano 488–490 (2014).
Lett. 16, 7433–7438 (2016). 42. Casiraghi, C. et al. Rayleigh imaging of graphene and graphene layers. Nano
22. Wang, X. et al. Raman spectroscopy of atomically thin two-dimensional Lett. 7, 2711–2717 (2007).
magnetic iron phosphorus trisulfide (FePS3) crystals. 2D Mater. 3, 43. Zhao, M. et al. Atomically phase-matched second-harmonic generation in a
031009 (2016). 2D crystal. Light Sci. Appl. 5, e16131–e16131 (2016).
23. Kim, K. et al. Suppression of magnetic ordering in XXZ-type 44. Liu, F. et al. Disassembling 2D van der Waals crystals into macroscopic
antiferromagnetic monolayer NiPS3. Nat. Commun. 10, 345 (2019). monolayers and reassembling into artificial lattices. Science 367, 903–906 (2020).
24. Vaclavkova, D. et al. Magnetoelastic interaction in the two-dimensional 45. Liu, Z. et al. Strain and structure heterogeneity in MoS2 atomic layers grown
magnetic material MnPS3 studied by first principles calculations and Raman by chemical vapour deposition. Nat. Commun. 5, 5246 (2014).
experiments. 2D Mater. 7, 035030 (2020). 46. Zhang, Q. et al. Strain relaxation of monolayer WS2 on plastic substrate. Adv.
25. Fiebig, M., Pavlov, V. V. & Pisarev, R. V. Second-harmonic generation as a Funct. Mater. 26, 8707–8714 (2016).
tool for studying electronic and magnetic structures of crystals. J. Opt. Soc. 47. Chen, X. et al. Electric field control of Néel spin–orbit torque in an
Am. B 22, 96–118 (2005). antiferromagnet. Nat. Mater. 18, 931–935 (2019).
26. Chu, H. et al. Linear magnetoelectric phase in ultrathin MnPS3 probed by 48. Fiebig, M. Revival of the magnetoelectric effect. J. Phys. D Appl. Phys. 38,
optical second harmonic generation. Phys. Rev. Lett. 124, 027601 (2020). R123 (2005).
27. Sun, Z. et al. Giant nonreciprocal second-harmonic generation from 49. Mutch, J. et al. Evidence for a strain-tuned topological phase transition in
antiferromagnetic bilayer CrI3. Nature 572, 497–501 (2019). ZrTe5. Sci. Adv. 5, eaav9771 (2019).
28. Wiedenmann, A., Rossat-Mignod, J., Louisy, A., Brec, R. & Rouxel, J. Neutron 50. Otrokov, M. M. et al. Prediction and observation of an antiferromagnetic
diffraction study of the layered compounds MnPSe3 and FePSe3. Solid State topological insulator. Nature 576, 416–422 (2019).
Commun. 40, 1067–1072 (1981).
29. Oshikawa, M. Ordered phase and scaling in Zn models and the three- Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
state antiferromagnetic Potts model in three dimensions. Phys. Rev. B 61, published maps and institutional affiliations.
3430 (2000). © The Author(s), under exclusive licence to Springer Nature Limited 2021

Nature Nanotechnology | VOL 16 | JulY 2021 | 782–787 | [Link]/naturenanotechnology 787

Articles NAturE NAnOtECHnOlOgY

Methods 2
Icrossed (2ω, ϕ) ∝ χ EQ EQ

Sample preparation. Single crystals of MnPSe3 were grown by the chemical xxzx sin 3ϕ − χ yxzx cos 3ϕ . (4)
 
vapour transport method. Elemental powders of high purity Mn, P and Se were
pressed into a pellet and sealed inside a quartz tube under vacuum. The tube Here, ϕ is the angle of the incident linear polarization with respect to the a axis
was then annealed for a week at 730°C to form polycrystalline MnPSe3 powder, of the crystal. Note that in the fit, there is also a constant angle shift in ϕ, which is
the composition of which was verified with powder X-ray diffraction. Crystals the angle between the horizontal axis in the lab and the crystal a axis. For the SHG
were then grown using the chemical vapour transport method with iodine as the patterns below TN, we fit the crossed polar patten by,
transport agent: 2 g of the powder and 0.4 g of iodine crystals were placed at the
end of a quartz tube, which was sealed off at 13 cm length under vacuum. The Icrossed (2ω, ϕ, θ ) ∝ L2 |sin(ϕ − θ )|2 , (5)
sealed tube was then set in a temperature gradient of 650/525 °C for four days to
transport the starting materials placed at the hot end to the cold end. The Mn/P/ and denote the node direction in the polar pattern as the Néel vector direction.
Se ratio was measured to be 1.00(1):0.96(1):3.07(1) with energy-dispersive X-ray Here, θ is the angle of the direction of the Néel vector with respect to the crystal a
spectroscopy. The ultra-thin samples were prepared by a standard mechanical axis. Supplementary Note 1 contains more details.
exfoliation process on Si substrates with 90-nm-thick SiO2 from a few MnPSe3
bulk crystals. Due to the small interlayer van der Waals coupling, one could easily
exfoliate MnPSe3 crystal down to atomically thin layers. The total exposure time Data availability
in air for samples exfoliated in a glove box was less than one minute before the All data needed to evaluate the conclusions in the paper are present in the paper
samples were loaded into the cryostat in vacuum. The samples studied in this and the Supplementary Information. Additional data related to this paper could be
work were as follows: (1) 15-μm-thick and 30-μm-thick bulk crystals mounted on requested from the authors.
the metal platform of the cryostats by glue; (2) one ~50-nm-thick flake and four
~100-nm-thick flakes exfoliated on SiO2/Si in air; (3) an 8L sample and at least Acknowledgements
three samples of each thickness from monolayer to five-layer exfoliated on SiO2/Si We thank S. W. Cheong and O. Tchernyshyov for helpful discussions. The project design,
in air; (4) a 9L sample and at least three samples of each thickness from monolayer data collection and analysis, and preparation of the manuscript are supported by L.W.’s
to four-layer exfoliated on SiO2/Si in a glove box; and (5) a 3L sample, an 8L startup package at the University of Pennsylvania. The development of the SHG photon
sample and more than five thick (10–20 nm) samples exfoliated on PDMS for the counter is supported by the ARO YIP award under grant W911NF1910342 to L.W. The
strain-tuning experiment. The data on a 10-nm-thick sample and a 15-nm-thick measurement by atomic force microscopy is supported by the ARO MURI under grant
sample exfoliated on PMDS are shown in the paper. W911NF2020166 to L.W. The acquisition of the oscillator laser for the SHG experiment
is supported by the National Science Foundation through Penn MRSEC (DMR-
SHG microscopy. The sample was loaded on a metal platform in a closed-cycle 1720530). E.J.M. acknowledges support from National Science Foundation EAGER
cryostat, and the temperature of the metal platform was controlled by a local heater, 1838456. C.L.K is supported by a Simons Investigator grant from the Simons Foundation.
which induces only a submicrometre shift of the sample position between 5 K and D.G.M acknowledges support from the Gordon and Betty Moore Foundation’s EPiQS
100 K. An ultrafast 800 nm Ti–sapphire laser pulse with a duration of ~50 fs at a Initiative, grant GBMF9069. H.W. and X.Q. acknowledge support from National Science
repetition rate of 80 MHz was focused onto a 2 μm beam spot on the sample by a Foundation DMR-1753054 and the Texas A&M University President’s Excellence Fund
×50 objective under normal incidence. A typical laser power of 200 μW was used X-Grants Program. B.X. and C.B. are supported by the Schweizerische Nationalfonds
except for the following cases: 500 μW was used for thick flakes and bulk crystals. by grant no. 200020-172611. The density functional theory calculations were conducted
In the bilayer sample in Extended Data Fig. 6, 400 μW was used. No sample damage with the advanced computing resources provided by Texas A&M High Performance
was observed during the measurement. The reflected SHG light was collected by Research Computing.
the same ×50 objective and reflected by a dichroic mirror into a photomultiplier
tube connected with a lock-in amplifier or a photon counter. The photon counter is
locked to 80 MHz in order to reduce the dark count below 0.2 c.p.s. without cooling Author contributions
the photomultiplier by cryogen. The detection sensitivity in the experiments is L.W. conceived the project and coordinated the experiments and theoretical work.
0.2 c.p.s. Because the parallel signal is one order of magnitude smaller than the L.W. designed the SHG imaging set-up and built it with Z.N.; Z.N. performed the
crossed signal, the polarization extinction ratio is important when measuring experiments and analysed the data under the supervision of L.W.; L.W., Z.N., E.J.M.
parallel patterns. The polarization of the fundamental light is controlled by a and C.L.K. discussed and interpreted the data. E.J.M. performed the spin model
half-wave plate as well as a linear polarizer. The polarization of second-harmonic calculation. C.L.K. performed the Landau theory calculation. A.V.H. and D.G.M. grew
light is analysed by a linear polarizer. The SHG imaging microscopy is achieved by the crystals and performed the magnetization measurements. H.W. and X.Q. performed
moving the sample with three Attocube nano-positioners. the first-principles calculation. B.X. and C.B. performed the optical conductivity
measurement. L.W. and Z.N. wrote the manuscript with the input of all authors. All
Strain tuning. We exfoliated MnPSe3 on PDMS with a square shape and applied authors edited the manuscript.
tensile strain on two sides of the PDMS by a micro-manipulator. The stretched
PDMS was then attached to a gold-coated sample platform. The strain was estimated
by measuring the length change along the stretching direction in the optical image.
Competing interests
The authors declare no competing interests.
A low transfer ratio (~13%) from PDMS to the sample was observed. To change
the sample’s strain direction, we peeled off the PDMS from the sample platform
and then stretched it in another direction. To apply strain along the 45° and 135°
directions, we cut the four corners of the PDMS to form a smaller square shape in
Additional information
Extended data is available for this paper at [Link]
order to reduce twisting while applying strain. The error bar of the estimation of
strain strength was ±0.5%, and the error bar of the strain direction was ±10°. Supplementary information The online version contains supplementary material
available at [Link]
Symmetry analysis for SHG polar patterns. For the SHG patterns above TN, the Correspondence and requests for materials should be addressed to L.W.
angle dependences are described by,
Peer review information Nature Nanotechnology thanks the anonymous reviewers for
2 their contribution to the peer review of this work.
Iparallel (2ω, ϕ) ∝  χ EQ EQ

xxzx cos 3ϕ + χ yxzx sin 3ϕ , (3)
 
Reprints and permissions information is available at [Link]/reprints.

Nature Nanotechnology | [Link]/naturenanotechnology

NAturE NAnOtECHnOlOgY Articles
a b c 0.0

Optical contrast
1L -0.1
2L
0.78 nm
-0.2

3L 0 1 2 3 4
Layer number

Extended Data Fig. 1 | Thickness characterization of atomically thin MnPSe3 samples. a, Optical image of a multilayer MnPSe3 exfoliated on the SiO2/Si
substrate. Scale bar: 10 μm. b, Atomic force microscopy image of the same sample in a. The step between monolayer and bilayer is around 0.78 nm. Scale
bar: 10 μm. c, Optical contrast of samples with different layer numbers. Green circles are data extracted from sample shown in a and orange circles are
data from other samples. The black line is a linear fit.

Nature Nanotechnology | [Link]/naturenanotechnology

 Articles NAturE NAnOtECHnOlOgY

a 3L b
20

15

I SHG (c.p.s.)
10

2L 1L 5

c d e
120° 60° Domain 1 120° 60°
4 Domain 2

3
I SHG (c.p.s.)

2 180°
0 0° 180°
0 0°

1
5 c.p.s. 0.75 c.p.s.
0
10 20 30 40 50 60 240° 300° 240° 300°
Temperature (K) Crossed Parallel

Extended Data Fig. 2 | SHG data for monolayer S2 exfoliated on SiO2/Si in a glove box. a, Optical image of the monolayer S2. There are some small
bilayer/trilayer islands inside the monolayer sample. Scale bar: 10 μm. The dashed region is the area for SHG mapping. b, SHG intensity mapping
of the monolayer S2 at 5 K. c, Temperature dependence of SHG intensity measured at the green point marked in a for one thermal cycle. d-e,
Polarization-dependent SHG patterns measured at the green point in (d) crossed and (e) parallel configuration after two different thermal cycles and at 5
K. Data from both domains are shown. The dots are experimental data and the solid lines are the best fit. The patterns are different from the monolayer S1,
indicating the angles between the Néel vector and the crystalline axis are different in the two samples.

Nature Nanotechnology | [Link]/naturenanotechnology

NAturE NAnOtECHnOlOgY Articles
a c 120° 60° Domain 1 e 5
Domain 2
4

SHG (c.p.s.)
3
1L 180°
0 0° 2

5 c.p.s. 1

0
10 20 30 40 50
240° 300°
Temperature (K)

b d 120° 60° Domain 1 f 5

6 Domain 2
4

SHG (c.p.s.)
5
4 3
I SHG (c.p.s.)

3 180°
0 0° 2
2
1
1 5 c.p.s.
0 0
10 20 30 40 50
240° 300° Temperature (K)

Extended Data Fig. 3 | SHG data for monolayer S3 exfoliated on SiO2/Si in a glove box. a, Optical image of the monolayer S3. Scale bar: 10 μm. The dashed
box denotes the region of SHG mapping. b, SHG intensity mapping of the monolayer S3. Scale bar: 10 μm. c-d, Polar patterns at the two points denoted by
purple and green dots are measured after thermal cycles. c, Crossed patterns of two domains of the purple point measured at 5 K after two different thermal
cycles. d, Crossed patterns of two domains of the green point measured at 5 K after two different thermal cycles. Different crossed polar patterns also
with different orientations at the purple and the green points indicate that their Néel vectors have different directions. e, Temperature dependence of SHG
intensity at the purple point for one thermal cycle. f, Temperature dependence of SHG intensity at the green point for one thermal cycle.

Nature Nanotechnology | [Link]/naturenanotechnology

Articles NAturE NAnOtECHnOlOgY

a b c
6
5
4

I SHG (c.p.s)
3
2
1
0

d e f
120° 60° P1 120° 60° P2 120° 60° P3
5 5 5

180°
0 0° 180°
0 0° 180°
0 0°

240° 300° 240° 300° 240° 300°

g 120° 60°
h i 120° 60°
P4 120° 60° P5 P6
5 5 5

180°
0 0° 180°
0 0° 180°
0 0°

240° 300° 240° 300° 240° 300°

Extended Data Fig. 4 | More data of the monolayer S1 shown in the main text. a, Optical image of the sample S1. Scale bar: 10 μm. The dashed region
is the area for SHG mapping. b, SHG intensity mapping at ϕ = 120∘ (the peak of domain 1) in the crossed pattern. c, SHG intensity mapping at ϕ = 160∘
(the peak of domain 2) in the crossed pattern. Maps in b,c are after the same thermal cycle. The domain walls are highlighted by the green dashed lines.
d-i, Crossed patterns of six different points marked by different colors in b after the same thermal cycle. P1-P5 are the same points in Fig. 3f in the main
text. P6 is very close to sample corner and displays smaller SHG intensity with a slightly different orientation, which indicates the Néel vector at P6 has a
slightly different direction. All the SHG measurements are performed at 5 K.

Nature Nanotechnology | [Link]/naturenanotechnology

NAturE NAnOtECHnOlOgY Articles
a 14 b
bulk×0.1
12 4L
3L 100
10 2L

ISHG (c.p.s.)
1L
√I SHG (a.u.)

8
6 10
4
2 1
0 1 2 3 4 5 6 78
20 40 60 80 100 Layer
Temperature (K)

c 70 d e
1L
60
50
40
TN (K)

30
20 2 c.p.s.
10
0
0 2 4 6 8 10 12 14 16 18 20 0 1 2 3
Layer SHG (c.p.s.)

Extended Data Fig. 5 | SHG data of atomically thin samples exfoliated on SiO2/Si in air. a, Temperature dependence of square root of the SHG intensity
in samples with different thickness. All of the samples in this figure are exfoliated in air. b, SHG intensity at 5 K as a function of layer numbers in a log-log
plot. Data are shown in yellow dots. The solid line is a fit for I ∝ N2. c, Néel temperature as a function of layer numbers. Note that there is a large reduction
of both SHG signal and Néel temperature of monolayer compared to those exfoliated in the glove box due to the aging effect. See more data on the aging
effect on a bilayer sample in Supplementary Figure 10. d, SHG intensity mapping of a monolayer MnPSe3 exfoliated in air at 5 K. Scale bar: 3 μm. Inset:
Optical image of the monolayer sample. f, Crossed polar pattern measured at the center of the monolayer sample shown in d at 5 K.

Nature Nanotechnology | [Link]/naturenanotechnology

Articles NAturE NAnOtECHnOlOgY

a b
250

200

150

I SHG (c.p.s.)
100
Bilayer
50

c 100 d 120° 60°

e 120° 60°
Thermal cycle
75
I SHG (c.p.s.)

1 6
50 180° 0° 180° 0°
25
100 c.p.s. 10 c.p.s.
0
20 40 60 80 100 240° 300° 240° 300°
Temperture (K) Parallel Crossed

Extended Data Fig. 6 | Néel vector switching of a bilayer sample exfoliated on SiO2/Si in air. a, Optical image of the bilayer sample. Scale bar: 10 μm.
b, SHG intensity mapping of the bilayer sample at 5 K. Note that the SHG intensity of this sample is quite uniform, which indicates that the Néel vector
direction is nearly aligned. c, SHG intensity of 6 consecutive cooling runs across TN. The curves collapse into two, indicating the existence of two domains.
d, Crossed and e, parallel polar patterns measured after each cooling at 5 K. The measured data are shown in dots. The blue and red shaded regions are
guides for the eye, corresponding to the two different AFM domains.

Nature Nanotechnology | [Link]/naturenanotechnology

NAturE NAnOtECHnOlOgY Articles
a b c d ϕ=0 ϕ=90° e 120° 60°

180°
0 0°

20
15 c.p.s.

15
240° 300°

I SHG (c.p.s.)
10
f 120° 60°

0
180°
0 0°

3L 15 c.p.s.

240° 300°

Extended Data Fig. 7 | Strain tunability of Néel vector of a trilayer sample on a PDMS substrate. The strained sample is prepared using the same method
as in the main text. The measured thickness is around 3.2 nm, which is a typical thickness for a trilayer sample. The measured Néel temperature is also
consistent with a typical trilayer sample. a, Optical image of the sample exfoliated on the PDMS before adding strains. The trilayer sample is marked by the
black box. Scale bar: 10 μm. b-c, Optical images of the sample when vertical and horizontal strain are added (marked by the red arrows), respectively. The
sample and PDMS are mounted on the metal platform and the contrast of the sample is low. A 5% strain is added on the PDMS by a micro-manipulator.
The directions of the cracks of the ~ 60-nm thick flake on the top also indicate the strain directions. d, SHG intensity mapping of the dashed area in a under
different strain directions (marked by the red arrows) and ϕ in the crossed pattern. The dark intensity maps in d indicate that the Néel vectors are mainly
along the strain direction, and rotated by around 90∘ when the strain directoin is swithced from vertical to horizontal. e, Crossed pattern at the center of
the trilayer sample with vertical strain direction. f, Crossed pattern at the center of the trilayer sample with horizontal strain direction. All of the above SHG
measurements are operated at 5 K.

Nature Nanotechnology | [Link]/naturenanotechnology

Articles NAturE NAnOtECHnOlOgY

a b c

Thick PDMS

Sample
Thin PDMS
Metal platform

d e
ϕ=0 ϕ=90°
30
As-exfoliated

SHG (c.p.s.)
20

As-exfoliated 10

0
55 60 65 70
Temperature (K)

f
40 Horizontally

SHG (c.p.s.)
Horizontally 30 strained
strained 20
10
0
55 60 65 70
Temperature (K)
0 100 200
I SHG (c.p.s.)

Extended Data Fig. 8 | Strain dependence on the Néel temperature of an 8-L sample on a PDMS substrate. The MnPSe3 samples in the strain-tuning
experiment (Fig. 4 in the main text, extended Fig. 7 and supplementary Fig. 13) are exfoliated on a ~ 30 μm PDMS, which is too thick to have the best
thermal conductance. One needs to calibrate the sample temperature by measuring a thick ( > 50 nm) flake on the same PDMS each time. To measure
the sample temperature without calibration, we use thinner home-made ( ~ 5 μm) PDMS to increase the thermal conductance and encapsulate the sample
after exfoliation and attach the ~ 5 μm thick PDMS to the metal platform after applying strain (shown in a). b, Optical image of an as-exfoliated sample ( ~ 8
L) on the PDMS substrate before being put onto the metal platform. Scale bar: 10 μm. c, Optical image of the same sample under a horizontal strain. A 5%
strain on PDMS is added by a micromanipulator and then the PDMS is attached to the metal platform. Scale bar: 10 μm. d, SHG intensity mappings of the
as-exfoliated and strained sample measured at ϕ = 0∘ and ϕ = 90∘ in the crossed polar pattern. The as-exfoliated sample shows that the Néel vectors are
not oriented horizontally while the strained sample favors a horizontal Néel vector orientation. The data is collected at 5 K. e-f, Temperature dependence
of SHG intensity of the (e) as-exfoliated and the (f) strained sample. The difference of the Néel temperature is within 1 K.

Nature Nanotechnology | [Link]/naturenanotechnology