Introduction

CuInP2S6 (CIPS) is one of the most promising two dimensional (2D) ferroelectric materials for microelectronic applications1,2,3. The van der Waals (vdW) layered structure provides interfaces without dangling bonds to integrate with other materials for functional heterostructures such as field effect transistors. An intriguing aspect of CIPS is the existence of multiple polar and antiferroelectric (AFE) phases that are close in energy and can be interconverted by external stimuli such as temperature, strain, electric fields, and interfacial conditions4,5,6,7,8. Four polar phases, underpinned by the position of the Cu ions with respect to the P2S6 layers, are possible according to density functional theory (DFT)4. If Cu is situated within the layers (“in-layer phase”), the spontaneous polarization is ~ 5 μC/cm2. If Cu displaces into the vdW gap (“in-gap phase”), the polarization increases to ~11.3 μC/cm2. Previously, the in-layer and in-gap phases were referred to as low-polarization and high-polarization phases. The in-layer polarization (Pl) and in-gap polarization (Pg) polarization vectors can point upwards (+) or downwards (-) along the z-direction if Cu is situated at the top or bottom of the P2S6 layers, respectively. The calculated piezoelectric coefficients for the Pl phase are −15.6 pm/V and +2.5 pm/V for the Pg phase. Since PFM is sensitive to the direction of the polarization, the polarization states +Pl, -Pl, +Pg, and -Pg are expected to be measured at −15.6 pm/V, 15.6 pm/V, +2.5 pm/V, and −2.5 pm/V, respectively, which allows us to assign the polar phase and polarization orientation to experimental PFM data. For the in-layer polar phase, d is negative due to a negative electrostrictive coefficient9,10. The high ionic conductivity in CIPS11,12,13 underpins unusual transitions between polar states that lead to the polarization aligning against the electric field as Cu ions move across vdW gaps5,12,13,14. An AFE surface phase characterized by zero piezoresponse has been reported on top of the +Pl phase15. DFT calculations indicated that in this case, an interlayer AFE phase in which the in-layer polarization orientation alternates between adjacent CIPS layers is the most energetically favorable. However, intra-layer AFE phases in which the polarization alternates within the layers are energetically close and might be stabilized through strain. DFT calculations indicate that formation of energetically close polar and AFE phases can further be impacted by metal interfaces and also depend on the number of layers, strain and contact doping7. These theoretically calculated metal-CIPS interactions are particularly interesting with regard to applications in microelectronics, in which metal contacts are used to read, write, and transport signals. If the metal contact can impact the polar properties of CIPS in electronic circuits, it can be used as a (static) tuning parameter. Moreover, metal contacts are unavoidable in many applications, and it is therefore important to know if and how they are affecting the material functionality.

In this work, we investigate the ferroelectric behavior of CIPS interfaced with Ag and Cu electrodes. We discover that in the absence of a direct current (DC) electric field, the Ag electrode enhances the piezoresponse, whereas Cu interfaces suppress it. While Ag electrodes show polarization switching similar to the surrounding CIPS, on Cu electrodes, the electromechanical response to an alternating current (AC) voltage is oscillating in an opposite phase. The differences can mainly be ascribed to the stabilization of an in-gap polar phase under the Cu electrode that exhibits positive, instead of negative, electrostriction that characterizes the more prevalent in-layer polar phase. In addition, AFE states can be stabilized under both types of electrodes, but are most commonly found under Cu electrodes, especially before applying DC voltages (Fig. S1).

Results

Piezoresponse force microscopy (PFM) measurements on several Ag and Cu electrodes of varying thickness reveal remarkable effects of the type of metal on piezoresponse and polarization switching. Consistently, the piezoresponse was increased on Ag electrodes compared to the surrounding CIPS. In contrast, piezoresponse was strongly suppressed at Cu electrodes and showed no or very low response. We applied voltage spectroscopy16 (VS) in grids across the electrodes and the surrounding CIPS without metal contacts, as indicated in the height images of the Supporting Information (SI) Fig. S1. At each grid location, a triangular-square voltage waveform with maximum amplitudes of +/−16 V was applied in three cycles (s S2). Simultaneously, an alternating voltage (AC) bias was applied to measure the piezoresponse and study polarization switching and map the piezoresponse amplitude, phase and piezoelectric coefficient d in the unpoled state as well as after poling with positive and negative DC voltages. Height and piezoresponse images for representative Ag and Cu electrodes before and after VS are shown in the SI Fig. S1. The PFM amplitude is related to the strength of the piezoelectric material response, and the phase indicates whether the piezoelectric deformation oscillates in-phase (0) or out-of-phase (л) with the applied AC voltage, which gives information about the polarization orientation. The piezoelectric response d is calculated from the amplitude and phase signal17.

Piezoresponse maps

Figure 1 shows the PFM maps across Ag and Cu electrodes (Fig. 1a, b, respectively) extracted from three different voltage steps of the triangular-square waveform. Initially, in the unpoled state at 0 V, the amplitude is higher on the Ag electrode than on the surrounding CIPS, and the phase contrast between the electrode and CIPS is homogeneous (Fig. 1a). The amplitude after the + 16 V step, however, is lower on the Ag electrode compared to CIPS and about the same after applying the −16 V pulse. The PFM phase after each DC voltage shows the expected phase change indicative of polarization reversal. There is no clear PFM phase contrast between the Ag electrode and CIPS after the +/−16 V pulses. In contrast, the PFM amplitude measured on the Cu electrode is lower than on the surrounding CIPS, and the phase is opposite for CIPS and Cu (Fig. 1b) before and after poling, leading to seemingly opposite polarities. The VS maps of the piezoelectric coefficient at 0 V correspond to the piezoresponse images measured before VS that are shown in SI Fig. S1.

Fig. 1: VS maps of off-field PFM amplitude, phase and d response.
figure 1

The maps were extracted at 0 V (before applying a bias), after +16 V, and after –16 V pulses of the 3rd cycle of the voltage waveform and were measured across a an Ag electrode and b a Cu electrode. The dashed lines in the amplitude images at 0 V indicate the approximate positions of the electrodes. The image sizes are 8 × 8 μm2.

Full size image

Hierarchical clustering was used to further analyze the ferroelectric behavior of the obtained piezoelectric hysteresis loops (Fig. S3). The dendrogram indicates how similar the data of different clusters are. The cluster maps show the distribution of the first two clusters that represent the two main branches in the dendrograms. For Ag, cluster 0 is not exclusively located at the electrode area, and some pixels in the electrode are assigned to cluster 1 (Fig. 2a). However, there is a clear cluster contrast between the Cu electrode and CIPS (Fig. 2b). The on-field d loops that were acquired during applying electric fields are similar for Ag and CIPS whereas for Cu and CIPS, there are distinctive features such as the slope and shape of the hysteresis. The off-field loops measured after applying electric field pulses show a strong overlap for Clusters 0 and 1 across the Ag electrode, with the response in the negative voltage branch appearing nearly identical. However, at the positive branch, Cluster 0 has a near-zero response, whereas the response of Cluster 1 is higher. The similarities between the two clusters are also evident from the dendrogram, which shows shorter vertical lines representing the Euclidean distance than the dendrogram for the data acquired across the Cu electrode. The stronger differences between the clusters evident in the dendrogram are highlighted by the opposite orientations of the off-field d loops across the Cu electrode and the surrounding CIPS. As also observed on the Ag electrode, the cluster 1 associated with the CIPS surface shows a loop orientation facing to the right, whereas the orientation of the loop measured on the Cu electrode is facing to the left. The difference in loop orientation is consistent with the observed phase and d contrast in the maps shown in Fig. 1.

Fig. 2: Distribution of measured piezoelectric coefficients.
figure 2

Histograms of the off-field piezoelectric coefficient d(pm/V) measured on the electrode (top) and surrounding CIPS (bottom) for a Ag and b Cu electrodes. The dashed lines indicate the DFT calculated d values for the two in-gap and in-layer polar phases, as well as the AFE phase.

Full size image

Piezoresponse histograms

To understand the observed differences between Ag and Cu electrodes, we must understand polar switching on the atomic level. As discussed above, prior work showed that the piezoresponse in CIPS is closely linked to in-gap and in-layer phases or even AFE states, dependent on the position of Cu ions. However, the loops shown in the SI Fig. S3 are averaged over the clusters, and the error bars are too large to clearly establish a link to different polar phases. Therefore, we analyze histograms of the measured piezoelectric coefficients at 0 V, +16 V and −16 V for electrode and CIPS areas (Fig. 2) to identify the different polar phases (Pl phase, Pg phase, AFE phase) and polarization orientation ( + or – for Pl and Pg) for a total of 5 different polar states. Due to the correlation between piezoelectric response and crystal structure, the 5 states can be identified by the measured electromechanical response. However, typically one of the Pl states shows a lower piezoresponse closer to ~−10 pm/V rather than the theoretical prediction of ~−15 pm/V, possibly due to strains that affect one polarization orientation more than the other18. Similarly, the Pg states have a wider distribution of values around the DFT calculated 2.5 pm/V4,18. The absence of piezoresponse could indicate AFE, non-polar, or disordered phases within the probing volume. However, zero response is also observed before voltage pulses are applied, making disorder or non-polar phases due to displaced ions unlikely. Therefore, we conclude that the observed zero-response originates from an AFE phase. According to previous calculations, for freestanding CIPS, interlayer AFE ordering underpinned by layers of alternating polarization is energetically more favorable than alternating dipole moments within the layers7,15. Since the AFE phase also occurs on CIPS surrounding both electrodes, we assume interlayer AFE.

As evident from the histogram in Fig. 2a, the Ag electrode shows a high negative response corresponding to +Pl at 0 V. At +16 V, mostly AFE and + Pg is observed. At −16 V, the + Pl phase is restored. The CIPS surrounding the Ag electrode starts out with a mixture of +Pl and −Pg phases with some AFE, +Pg and −Pl pixels. At +16 V, the response changes to d associated with AFE, +Pg and −Pl phases and turns into mainly +Pl at −16 V. The response at high positive and negative voltages is therefore similar for the Ag electrode and the surrounding CIPS areas. The Cu electrode shows mostly zero or low positive response at 0 V, corresponding to AFE and +Pg phases (Fig. 2b). At +16 V, the response is predominantly around −5 pm/V and −12 pm/V, corresponding to the −Pg and +Pl phases, respectively. At −16 V, the measured d values exhibited AFE (0 pm/V), +Pg (a few pm/V) and −Pl (~ 10 pm/V). CIPS surrounding the Cu electrode shows mostly −Pg and +Pl values as well as some AFE and +Pg pixels. At +16 V, the response is quite uniform, −Pl, whereas at −16 V, the d values are similar to the initial ones at 0 V. It is evident that the response from the Cu electrode and the surrounding CIPS are of opposite polarity at all extracted voltage steps. In contrast, the response on and off the electrode is similar for the Ag electrode. The CIPS surrounding the electrodes shows variations that are common for different flakes, even from the same sample batch. The CIPS adjacent to the Ag electrode exhibits a higher response for the +Pl state than the CIPS around the Cu electrode, and vice versa for −Pl. Moreover, during voltage cycling, the +Pg phase is more common than −Pg in CIPS surrounding Ag, whereas the CIPS surface adjacent to the Cu electrode shows more −Pg during switching. However, multiple Cu and Ag electrodes were tested on different flakes, and the impact of Ag and Cu interfaces on the polarization and phase switching showed the same characteristics as discussed.

Piezoresponse loops

The histograms provide information on the voltage- and area-dependent dominant phases; however, to understand switching pathways, we extract loops from specific pixels that exhibit the most common switching type for their area (Fig. 3). The loop number indicates how often this type of switching occurs, with Loop 1 representing the most observed transitions. The loops directly relate to the histograms in Fig. 2. Response levels at high positive and negative voltages indicate the type of polar or AFE phase. The fact that these measured d levels are nearly constant within a wide voltage range corroborates that the observed piezoresponse results from transitions between phases. In contrast, the averaging effect resulting from different polarities of domains within the PFM excitation volume would result in changes of the response as the ratio between the domains strongly depends on the applied voltage. Moreover, a shift in PFM loops due to background signals (e.g., originating from electrostatic interactions) can be excluded because the signal levels of the polar states are in good agreement with the theoretically predicted values. In addition, the surface potential of CIPS is small (~−140 to −170 mV, see SI Fig. S4) and would therefore lead to only negligible electrostatic signal contributions19.

Fig. 3: Piezoresponse loops and corresponding switching paths.
figure 3

Piezoresponse loops of the most common types of switching and schematic drawings of the polar and AFE phases during switching for a the Ag electrode, b CIPS surrounding the Ag electrode, c the Cu electrode and d CIPS surrounding the Cu electrode. The loop number refers to how common the type of switching is for a specific area, with Loop 1 indicating the most common switching path. The schematic drawings depict the position of the Cu atoms for the different polar or AFE phases at +16 V and −16 V (see label above the tip for voltage polarity). The blue, dashed arrows indicate the movement of Cu ions. The black arrows refer to the polarization vectors.

Full size image

The most common switching paths for the Ag electrode and the surrounding CIPS are between +Pg and +Pl (Loops 1 in Fig. 3a, b, respectively). This can be explained by assuming the positively charged Cu ions move in the direction of the electric field across the vdW gap and a layer in Loop 1, as depicted in the schematic drawing. This mechanism, based on ionic motion, allows for the polarization +Pg to align against the electric field at +16 V, which is not possible in conventional ferroelectric switching. The second most common transition on the Ag electrode is between AFE and +Pl states, where alignment against the polarization at +16 V is avoided through AFE ordering (Loop 2 in Fig. 3a). On CIPS, conventional ferroelectric switching between −Pl and +Pl states is observed (Loop 2 in Fig. 3b). The Cu electrode exhibits predominantly ferroelectric switching between the in-gap phases with −Pg at +16 V and +Pg at −16 V (Fig. 3c, loop 1). Loop 2 switches between −Pg and an AFE state that is stabilized at negative voltages. Loop 3 shows that transitions between +Pl (+16 V) and +Pg (−16 V) states are underpinned by a small, reversible ion displacement from the layer into the vdW gap. The most common switching path on CIPS surrounding the Cu electrode is between −Pl and +Pl at positive and negative voltages, respectively (Fig. 3d). Loop 2 shows transitions between −Pl and −Pg during which Cu moves across a layer and a gap.

Discussion

The results show that conventional ferroelectric switching between −Pl and +Pl states, which is underpinned by a polarization reversal without a phase change, exclusively occurs on CIPS surfaces without electrodes. The other common transition in CIPS is characterized by a bias-induced phase change without changing the orientation of the polarization vector. These ionic mechanisms lead to an alignment of the polarization against the electric field for both CIPS areas, although opposite in-gap and in-layer polarities are preferred. This discrepancy could arise from strain differences in the flakes on which Cu and Ag were deposited. Although the increase of polarization under opposite field directions would be fundamentally inaccessible to classical ferroelectrics, CIPS is known for field-induced ionic motion across vdW gaps and layers that contribute to unusual polarization control5,20,21,22,23. Neumayer et al. reported the alignment of the polarization against the electric field in CIPS capacitors5 as well as the CIPS surfaces14 based on this polar–ionic coupling. Moreover, Li et al. observed multiple polarization states and switching against the electric field due to Cu ion migration12. Jian et al. studied the transition from ferroelectric to ionic switching mechanism based on Cu ion migration22 and demonstrated bidirectional polarization switching under a unidirectional electric field based on Cu crossing the vdW gap21. Liang et al. investigated polar and ionic switching kinetics of CIPS using photovoltaic short currents to demonstrate anomalous switching behavior23. The observed switching paths are, therefore, a manifestation of a phenomenon that, while unusual from the perspective of classical ferroelectrics, is increasingly discussed in literature and supported by a growing body of experimental and theoretical studies.

AFE states were only commonly observed on electrodes, especially Cu, before applying voltage pulses and at negative voltages. On the Ag electrode, the AFE state is stabilized at positive voltages and at negative voltages for Cu. The positive, rather than negative, electrostriction of the in-gap phase compared to the in-layer phase leads to opposite piezoresponse polarities between the electrode and CIPS. The Cu electrode stabilizes +Pg and AFE states, which are less common on CIPS. Notably, in the case of Cu electrodes, there is no ionic-based switching that includes Cu transport across gaps and layers, unlike on CIPS or the Ag electrode. Additional transitional AFE states during switching between polar states appear in Loop 1 of Fig. 3a, b, c, Loop 2 in Fig. 3d and Loop 3 in Fig. 3c.

The exact mechanism for the observed impact of metal interfaces on polar states and switching behavior is unknown. DFT studies have shown that metal electrodes strongly affect the Cu position in CIPS at the interface and thus the polar phase, with the extent of this effect varying by metal7. This can also modify the switching behavior underneath the metal electrodes. However, a direct link to the DFT studies cannot be drawn since the calculations focused on few-layer flakes at 0 K in the absence of electric fields7. Alternatively, both Ag and Cu contain reactive, mobile ions that would facilitate interdiffusion as well as filament formation. However, the reported strong interactions between the CIPS sulfur atoms and the Cu atoms of the electrode7 may prevent strong field-induced movement of Cu in CIPS. Further differences can arise from the larger lattice spacing in Ag than in Cu, which might change the interfacial strain. The work function of Cu is higher than that of Ag, which typically leads to differences in band bending and may alter the charge transport within CIPS and at the interfaces. In addition, the polarization magnitude and orientation are expected to impact band bending. Vacancies and charge accumulation that may arise from field-induced ion migrations further complicate switching dynamics. Future studies are needed to decipher the origin of the experimental observations.

In conclusion, the type of metal can strongly impact the ferroelectric behavior of CIPS. In particular, Cu electrodes are able to stabilize AFE and switching within the in-gap phases, leading to lower piezoresponse amplitudes and an electromechanical response that is oscillating at an opposite phase compared to CIPS only. The in-gap phase exhibits a more than two times higher polarization than the more common in-layer phase, which can be exploited in electro-optical applications or in electronic elements that require higher switching currents to detect a polarization change. In addition, long-range Cu motion appears to be suppressed, which avoids disorder and defects. In the absence of DC electric fields, Ag electrodes show higher piezoresponse than the surrounding CIPS, which is advantageous for electromechanical sensors and actuators. This work highlights the importance of carefully choosing the metal used for electrodes and contacts to obtain the desired electromechanical and polar properties of devices that contain CIPS. Moreover, these findings might also extend to other metal thio- and selenophosphates and provide an additional tuning knob for functional properties in a variety of material systems.

Methods

Sample fabrication

CIPS flakes were grown as described in Susner et al.24. For this study, bulk CIPS flakes of several tens of μm thickness were used.

Metal electrodes of ~5 × 5 μm2 were fabricated on CIPS flakes using a combination of electron beam lithography (EBL) and a lift-off process. Electron beam resist, 950 PMMA A4 (Kayaku, Westborough, MA) was spincoated on the substrate at 6000 rpm for 45 s, and baked on a hotplate for 2 min at 180 oC. Electron beam lithography was performed with a JEOL JBX8100-FS system operating at 100 kV acceleration voltage. A built-in optical microscope in the EBL system was used to find the position of each flake, where the electrode array was to be patterned, without exposing the resist. A 40 nA beam current and 1000 µC/cm2 dose were used to define the geometry of the ~5 × 5 µm2 electrodes. The patterned substrate was developed in 1:3 methyl isobutyl ketone: isopropyl alcohol (IPA) for 1 min, rinsed with IPA, and dried with nitrogen. A lift-off process, using acetone and IPA, was carried out following deposition of Cu by electron beam evaporation. This fabrication process was repeated for the Ag electrode array.

PFM measurements

The PFM measurements were conducted in band excitation mode on an Asylum Research Cypher AFM equipped with National Instruments electronics. Budgetsensor ElectriMulti75-G probes with a nominal force constant of 3 N/m were used. PFM voltage spectroscopy was performed in grids of 20 × 20 pixels over an area of 8 × 8 μm2 in a pulsed triangular waveform between +16 V and −16 V consisting of three cycles. The grid area included the electrode and surrounding CIPS with no metal contacts. The piezoresponse amplitude, phase and d were calibrated as outlined in reference17. Data analysis was performed with Python. The frequency of transitions between polar and AFE phases for the different areas was calculated by identifying the polarization state for each pixel at +16 V and −16 V based on the measured d level.