Highly Sensitive Electromechanical Piezoresistive Pressure Sensors Based on Large-Area Layered PtSe2 Films.

Two-dimensional (2D) layered materials are ideal for micro- and nanoelectromechanical systems (MEMS/NEMS) due to their ultimate thinness. Platinum diselenide (PtSe2), an exciting and unexplored 2D transition metal dichalcogenide material, is particularly interesting because its low temperature growth process is scalable and compatible with silicon technology. Here, we report the potential of thin PtSe2 films as electromechanical piezoresistive sensors. All experiments have been conducted with semimetallic PtSe2 films grown by thermally assisted conversion of platinum at a complementary metal-oxide-semiconductor (CMOS)-compatible temperature of 400 °C. We report high negative gauge factors of up to -85 obtained experimentally from PtSe2 strain gauges in a bending cantilever beam setup. Integrated NEMS piezoresistive pressure sensors with freestanding PMMA/PtSe2 membranes confirm the negative gauge factor and exhibit very high sensitivity, outperforming previously reported values by orders of magnitude. We employ density functional theory calculations to understand the origin of the measured negative gauge factor. Our results suggest PtSe2 as a very promising candidate for future NEMS applications, including integration into CMOS production lines.

Layered two-dimensional (2D) materials have extraordinary electrical, optical, and mechanical properties that suggest high potential for a wide range of nanoelectronics applications.[1,2] 2D transition metal dichalcogenides (TMDs) have recently been intensively investigated due to their wide range of inherent electronic properties that complement graphene. Out of this family, platinum diselenide (PtSe2) is a relatively new and thus little explored TMD material. Monolayer PtSe2 is a semiconductor with a band gap of 1.2 eV,[3] while bulk PtSe2 becomes semimetallic with zero band gap.[4,5] It has been grown epitaxially[6] or synthesized on insulating substrates using thermally assisted conversion (TAC) of predeposited platinum (Pt) films by a vaporized solid selenium precursor.[4,7] The latter process can be carried out at temperatures between 400 and 450 °C, which is compatible with standard semiconductor back-end-of-line (BEOL) processing,[8] in contrast to other 2D materials.[9−11] We have demonstrated potential applications of TAC-grown PtSe2 in highly sensitive gas sensors and photodetectors with equal or even superior performance compared to graphene and other TMD-based devices.[7,12] Furthermore, top-down structured electrical devices with PtSe2 have been demonstrated.[13] Here, we explore the electromechanical properties of PtSe2 in nanoelectromechanical system (NEMS) pressure sensors. The devices utilize the piezoresistive effect in PtSe2, which gives rise to a change in its electrical resistance upon mechanical strain. We investigate PtSe2 membrane-based pressure sensors and find sensitivities that are orders of magnitude higher than those of nanomaterial-based devices,[14−17] including graphene.[18−20] Furthermore, we obtain an average negative gauge factor (GF) of −84.8 for PtSe2 from a bending beam setup. Detailed density functional theory (DFT) simulations reveal an increase in the density of states (DOS) at the Fermi level of bulk PtSe2, corroborating the experimentally observed piezoresistive properties.

Polycrystalline PtSe2 films with nanometer sized grains were grown on silicon/silicon dioxide (Si/SiO2) substrates by TAC at 400 °C of predeposited 1 nm thick Pt films (Figure S1 in the Supporting Information). Comprehensive characterization of layered PtSe2 thin films synthesized in this manner, including X-ray photoelectron spectroscopy and scanning transmission electron microscopy, is presented in our previous work.[4,7] The thickness and roughness of the layers were determined by atomic force microscopy (AFM, Figure a,b). A thickness of 4.5 nm was found for the selenized 1 nm initial Pt film. Assuming an interlayer distance of 0.6 nm for PtSe2,[21] this corresponds to 7–8 layers for the 4.5 nm thick PtSe2 film. The root-mean-square (RMS) roughness of the layer was 0.284 nm. The Raman spectrum of the PtSe2 film shows two sharp and prominent peaks corresponding to the Eg and A1g modes of the 1T phase, attesting the crystalline nature of the films (Figure c).[4] After the film growth, a polymethyl metacrylate (PMMA) layer was spin-coated on top of the PtSe2.

Figure 1

As-grown PtSe2 film characterization: (a) AFM measurement of the selenized 1 nm initial Pt film including height profile of the step between the Si/SiO2 substrate to the PtSe2 film. (b) High-resolution AFM measurement displaying the RMS roughness. (c) Raman spectrum of the as-grown 4.5 nm thick PtSe2 sample.

These centimeter-scale PtSe2 thin films were transferred onto the substrate comprising 79 independent pressure devices, using a typical polymer supported transfer process as described for graphene.[22] Each device consists of cavities covered by a patterned PtSe2 film located in between metal contacts (Figure a–c). The PMMA layer was not removed after the transfer in order to stabilize and protect the PtSe2 layer against environmental influences and to reduce possible cross sensitivity of the sensor. Details of the fabrication processes are described in the Methods section and in Figure S2a–d. A noninvasive Raman tomography cross-section, as demonstrated recently for graphene membranes,[23] confirms the presence of free-standing PtSe2 across the cavity (Figure d,e) where devices with suspended PtSe2 (Figure e, top) and collapsed membrane (Figure e, bottom) were measured and can clearly be distinguished with this method. The corresponding spectrum of the suspended PtSe2 shows the characteristic Eg and A1g modes without the Si(I) peak (Figure d, top), whereas the spectrum taken at the bottom of the trench shows only the characteristic Si(I) peak of the Si substrate (Figure d, bottom). This clear separation is feasible due to the 1.4 μm deep cavity, which defines the distance between the suspended PtSe2 and the Si substrate.

Figure 2

Pressure sensor fabrication and measurement. (a) Optical micrograph of one device with Au contacts and the PtSe2 channel across the cavity area. (b) Schematic cross-section of the cavity area with suspended PtSe2. (c) Schematic of one device including contacts and cavity area. (d) Raman point spectrum of the PtSe2 membrane area and the Si substrate. (e) Raman tomography cross sections of one device with suspended PtSe2 (device 3.2) and one with collapsed membrane (collapsed device). On the right are the corresponding optical micrographs of the cavity areas (device 3.2 and the collapsed device) as well as one other working device (device 3.3) and one device without cavity (device no cavity) as reference. (f) Resistance change in % for one pressure cycle (vented, pumped down to 200 mbar, vented) of the four devices shown in panel (e) (left y-axis) and pressure on the outside of the membrane (right y-axis).

The operation of the investigated pressure sensors can be described in the following way: the PtSe2/PMMA membranes seal the cavities and trap environmental pressure inside them.[24−26] A different pressure (lower or higher) outside the cavity leads to a deflection of the PtSe2 membrane. This deflection induces strain in the PtSe2 layer, resulting in a change of its resistance. The chip design used in the present experiments includes devices of similar size without cavities as control references. The pressure sensors, after wire-bonding and packaging into a chip socket, were placed in a home-built pressure chamber where pressure can be regulated from 1000 mbar (environmental pressure) down to 200 mbar (Figure S3). The pressure, humidity, and temperature inside the chamber are recorded and controlled with commercial sensors during the experiments. Argon (Ar) atmosphere was used to exclude cross-sensitivity effects with gases and/or humidity.[7] The current–voltage (I–V) characteristics of the PtSe2 device for environmental pressure shows good linearity, i.e., the transferred PtSe2 forms good Ohmic contacts with metal (gold) electrodes (Figure S4). The measured data is presented in Figure f along with the simultaneously measured pressure (dashed red line, right y-axis). The resistance across the PtSe2 sensors decreases with applied increasing strain on the membrane (caused by the decrease in pressure, Figure S5a). A pressure difference of 800 mbar between the outside and the inside of the cavity leads roughly to 7% of change in resistance.

The results of four devices are displayed in Figure f: two intact membranes on cavities (filled green squares and filled blue triangles), one with a collapsed membrane (empty purple diamond) and one device without cavity for reference (empty black circles). A slight decrease (increase) in humidity during evacuation (venting) of the chamber is commonly observed for all measurements (Figure S5b), but does not give rise to changes in the resistance. Cross-sensitivity toward humidity can thus be excluded from having any substantial influence on the device behavior. We conclude that the observed resistance change is primarily caused by pressure (i.e., strain) changes during the measurements.

The absolute sensitivity (S) of a piezoresistive device is calculated using eq , where ΔR/R is the absolute value of the relative resistance change and ΔP is the pressure change during the sensor operation.[18]

A total of 19 working devices were measured on four different chips through 69 different measurements (Figure a). A sensitivity of 1.05 × 10–4 mbar–1 was extracted from Figure f. An average of 5.51 × 10–4 mbar–1 was calculated for all measured devices (best device not included). The results were compared to some other NEMS pressure sensors that operate in a similar pressure range (Figure b,c). These use gallium arsenide (GaAs),[27] Si nanowires (Si NWs),[15,16] single- and multiwalled carbon nanotubes (SWNT and MWNT)[14,17] and graphene[18−20] as the piezoresistive materials. However, for most of these (except Smith et al.[18] and Fung et al.[17]), a separate membrane is needed, which is treated, or where the piezoresistive materials are grown or applied in a separate fabrication step. This adds additional complexity to the fabrication process, which is not the case for the PtSe2-based pressure sensors because the material acts as membrane and sensor at the same time. The thickness of other material is limited due to their mechanical stability. The measured sensitivity outperforms the best reported 2D and low dimensional materials-based pressure sensors by a factor between 3 and 70 (Figure b). The SWNT-based pressure sensor[14] shows an equal sensitivity; however, the membrane areas are nearly 2 to 4 orders of magnitude larger than the PtSe2 membrane. To take this into account the sensitivity can be normalized to the membrane area to receive a direct comparison between the devices independent of their membrane size.[28,18] Here, the PtSe2-based pressure sensor shows nearly 2 to 5 orders of magnitude higher sensitivity than the other devices (Figure c; see also Table S6).

Figure 3

Comparison of all measured chips and devices as well as comparison of various pressure sensor devices using different materials for piezoresistive detection. (a) Box and whiskers plot of all measured chips, devices, and repeated measurements. (b) Comparison of the sensitivity in mbar–1 with other pressure sensors. (c) Comparison of the sensitivity normalized by membrane area in mbar–1 μm–2 with other pressure sensors. In plots (b) and (c), devices that require a separate membrane are indicated with a *, and the device with exceptionally high sensitivity is indicated by a black star.

We further point out that one of the devices showed an exceptionally high sensitivity of 1.64 × 10–3 mbar–1 (Figure S5c). Even though it could not be reproduced in other devices, this value was repeatedly measured for that particular device (Figure a, device 2.1). The sensitivity and normalized sensitivity of this device are plotted in Figure b,c (black star). When normalized to the membrane area, this device is over 3 orders of magnitude more sensitive than the highest performing devices shown in Figure c (a SWNT[14] and a graphene membrane[18] based device), and 4 to 6 orders of magnitude more sensitive than other reported devices.

A bending beam setup was used to measure the piezoresistive gauge factor of the 4.5 and 9 nm thick PtSe2 films, which consisted of 7 to 8 and approximately 15 layers, respectively. The bending beam setup enables measurements of well-defined strain fields in a very simple way (for details, see Supporting Information). The gauge factor (GF) can be extracted using eq .[29]

Here, ε is the strain in the bending beam at the position of the sensor (ε = 0.04%). A negative gauge factor (GF) of −84.8 was obtained (Figure a; Supporting Information sections S7 and S9). The measured resistance change is shown in Figure b and Figure S9b for two PtSe2 layer thicknesses and masses of 2 and 0.5 kg (Figure S9c). A measurement of multiple strain cycles with a mass of 2 kg is shown in Figure S9d. In some measurements, the resistance did not immediately return completely to its initial values after removing the mass (Figure b, 4.5 nm PtSe2 device; Figure S9b,d). This may be attributed to a slip of the device or the electrical contacts during measurements due to nonoptimal adhesion to the beam. Furthermore, a slight overshoot followed by oscillations was observed for some devices right after applying the mass. These may be interpreted as natural oscillations due to the manual application of the mass on the beam. Data from a commercially available metal strain gauge is used as a reference (Figure b). A decrease in resistance with increasing strain on the PtSe2 film corroborates the integrated pressure sensor measurements (Figure S9a). The GF of the PtSe2 films exceeds the GF of the metal strain gauge (GF of 2)[29] and of graphene (GF of 2–6).[24,30,31] While similar or higher GF have been found for doped silicon (GF of −100 to +200)[29] or mono- and trilayer MoS2 (GF of −148 and −43.5).[32]

Figure 4

(a) Bending beam setup (cantilever beam) with applied PtSe2 and commercially available metal strain gauges including stress simulation with the applied weight. (b) Electrical readout signal during the measurement with an absolute resistance change against time showing two PtSe2 devices (4.5 nm PtSe2, diamonds filled in blue; 9 nm PtSe2, squares filled in green) and a metal strain gauge (empty red circles) as reference with an enlarged view in the inset.

DFT calculations were conducted to gain insights into the observed resistance decrease with increased strain in PtSe2 films. The model was designed to represent the membrane covering the cavities of the PtSe2 pressure sensors. Here, the deformation can be thought of as biaxial strain acting on the PtSe2 membrane. The model was fully optimized in terms of the lattice vectors and atomic positions. In the bulk system, which is similar to the experimental conditions in this work, biaxial strain means that the layers are stretched or compressed in plane. As bulk PtSe2 is metallic, the density of states (DOS) at the Fermi level is found to increase under biaxial tensile strain (εa) and reduces under compression, as shown in Figure S10a. This result is in accordance with the experimental findings, where the resistivity decreases with increased εa.

Next, we have considered the effect of the interlayer strain and compression (εc) on the DOS. The results are shown in Figure S10b. The compression of the interlayer distance has the opposite influence on the system, and it slightly reduces the DOS at the Fermi level. Finally, we have considered the two effects acting in tandem in the following way: for in-plane stretching, there is compression out of the plane, and vice versa (Figure a). The results for the combined effect are shown in Figure b. Once again, the DOS at the Fermi level is increased for the in-plane stretching and compression in the out-of-plane direction. However, the effect is reduced compared to just εa, due to the opposite effect of the interlayer compression, as shown in Figure S10b. In comparison to the significant increase of the DOS at the Fermi level upon strain, the band structure shows only minor changes (Figure S10c) not indicating strong differences in the effective masses under strain.

Figure 5

(a) Top and side view of the PtSe2 bulk shown with the in-plane, a, and out-of-plane, c, lattice vectors, and the direction of the applied strain (εa and εc). (b) Density of states close to the Fermi level (shifted to zero) under applied tensile strain and compression.

An increase in DOS has also been reported for other single and few layer semiconducting TMDs, like MoS2,[33] MoSe2,[34] WSe2,[34] and also PtSe2.[35] The semiconducting materials exhibit a decrease in band gap and even a transition to metallic behavior with applied strain, resulting in a negative gauge factor. For more detailed future studies, the influence of many other parameters should be considered and investigated, including the adhesive interface between the steel beam and the polyimide foil, the thickness and mechanical properties of the foil, and the PMMA layer covering the PtSe2. Also, the crystallographic orientations, thickness, and other material properties of the TAC-grown PtSe2 film need to be investigated and considered in the simulations. Nevertheless, the results of the simulations show a clear indication of an increase in DOS, leading to a decrease in resistance under the applied strain, which is the most plausible cause of the experimentally extracted negative GF.

We have demonstrated PtSe2-based piezoresistive pressure sensors with high gauge factors and very high sensitivity. The experiments were conducted on large-area PtSe2 thin films grown by a scalable method of thermal conversion of metallic films at a low temperature of 400 °C. A negative gauge factor of approximately −85 was obtained from a PtSe2 strain gauge through bending beam experiments and can be understood in terms of a strain-induced increase in the DOS, as verified through DFT simulations. Furthermore, integrated piezoresistive pressure sensors were fabricated using PtSe2/PMMA membranes. The PtSe2 pressure sensors showed very high sensitivity, outperforming piezoresistive pressure sensors based on other materials by at least a factor of 82. When normalized to the active membrane area, the sensitivity of our devices is almost 2 to 5 orders of magnitude larger than values reported for any other technology to date. One exceptional device has an even higher sensitivity of 1.64 × 10–3 mbar–1 and outperforms existing technologies by an even greater margin. These results suggest that layered 2D PtSe2 has great potential for future piezoresistive NEMS applications, beyond the demonstrated pressure sensors. These devices have a footprint that is up to 4 orders of magnitude smaller than other pressure sensors while showing a similar or even higher sensitivity. Further down-scaling to smaller cavities will improve the stability and yield of the membranes and allow decreasing the membrane thickness to maintain the high sensitivity. The low growth temperature and general complementary metal–oxide–semiconductor (CMOS) material compatibility makes PtSe2 highly attractive for future silicon technology integration.

Methods

Materials Synthesis and Analysis

Pt layers with a thickness of 1 and 2 nm were sputter-deposited on 1.5 × 1.5 cm2 Si/SiO2 substrates using a Gatan coating system (Gatan 682 PECS) with a 19 mm diameter Pt target. A TAC process was used to selenize the predeposited Pt layers, as described in detail in refs (4) and (7). The cross-sectional Raman measurements were conducted using a WITec alpha 300R system with a 532 nm laser as the source of excitation. A laser power of 75 μW was used. For obtaining single Raman spectra, an 1800 g/mm grating was used, while the depth scan (cross-section) was conducted with a 300 g/mm grating. The system is limited to a lateral and vertical resolution of 300 and 900 nm, respectively.

Strain Gauge Fabrication

The strain gauges were fabricated by transferring as-grown PtSe2 films onto prepatterned polyimide foil substrates. A typical polymer-supported transfer method was used for the film transfer. PMMA (200 nm thick) (Allresist AR-P 649.04) was spun on the as-grown PtSe2 samples at a speed of 2000 rpm for 50 s. After spin-coating, the samples were annealed for 5 min at 80 °C to cure the PMMA. 2 M KOH was used to etch away the SiO2 layer and release the PMMA-covered PtSe2 films from the substrate. The PMMA-covered PtSe2 films were cleaned in DI-water, followed by a transfer to the prepatterned (500 nm thick copper contacts from evaporation) polyimide foil. After drying, the foil was glued (super glue Z70 by HBM) onto the steel beam (size of 300 × 30 mm2, 3 mm thick), 200 mm away from the loading point. A commercially available metal strain gauge was glued next to the PtSe2 strain gauge. Cables were then soldered onto the contact pads for electrical connections.

Pressure Sensor Fabrication

A Si chip (7 × 7 mm2) with 1.6 μm thick SiO2 was used as a device substrate (Figure S2a). The cavities have circular geometry (radii between 2 and 20 μm) with arrays of 1 to 15 cavities and rectangular geometries (length of 20 μm and width between 2 and 4 μm) in arrays between 1 and 3 cavities. They were structured by photolithography, followed by Ar- and CHF3-based reactive ion etching (RIE, 200 mW, 40 mTorr) to create a vertical profile with 1.4 μm of depth (Figure S2b). The contact regions were defined by photolithography, followed by a RIE process to embed the contacts and self-aligned metal evaporation (Cr/Au, 40 nm/400 nm, Figure S2c). The PtSe2 films grown from 1 nm initial Pt films were transferred onto the device substrate using a PMMA-supported transport method as described above (Figure S2d). Lastly, the chip was wire bonded into a 44-pin PLCC chip package using 25 μm diameter Au wire.

Simulation

All calculations were performed using the DFT method as implemented in the VASP plane wave code.[36,37] Ion–electron interactions were described by projector augmented wave approximation.[38] We have employed the Perdew–Burke–Ernzerhof (PBE)[39] functional under the generalized gradient approximation. van der Waals interaction corrections were included using the D3 approach proposed by Grimme.[40] Cutoff energy was set to 600 eV, and the convergence threshold for residual force was 0.01 eV Å–1. Brillouin zone integration was carried out at 6 × 6 × 6 Monkhorst–Pack k-grids. Spin–orbit coupling (SOC) was included in the electronic structure calculations. Bulk PtSe2 was fully optimized (lattice vectors and atomic positions) resulting in the following lattice vectors: a = b = 3.788 Å and c = 4.790 Å, and a distance between Se atoms in the Se–Pt–Se sandwich of 2.550 Å. This is in good agreement with the experimental data by Wang et al.,[6] with a = b = 3.70 Å and the distance between Se atoms of 2.53 Å.