Interactions at heterointerfaces influence actuation in wet cast 1T-MoS₂ and V₂O₅·0.5H₂O thin films

Abstract

Interfacial interaction strengths are often invoked as determining factors in the chemomechanical coupling across actuating lamellar structures. However, electrochemical layered actuators of 100 nanometers to a few microns in thickness are often well described with classical models which depend only on bulk elastic moduli and relative thicknesses. We report a set of electrochemical systems composed of flexible working electrodes based on sub-micron thin films of 1T-MoS₂ and V₂O₅·0.5H₂O deposited onto metallic Au and Ni surfaces. Changes in electrode curvature were measured as a function of applied potential from which induced strains and stresses were calculated using a Timoshenko multi-layer beam bending model. The 1T-MoS₂ system achieved a maximum actuation strain of 0.57(5)% and 1.29(13)% while the V₂O₅·0.5H₂O system achieved 1.17(8)% and 1.2(2)% on Ni and Au respectively. Based on these results, small differences in interfacial interactions, such as in the case of the V₂O₅·0.5H₂O, were not distinguishable, whereas for very thin films of 1T-MoS₂, where strong differences between Au–S and Ni–S were present, the strong Au–S interaction resulted in greater actuation strains.

---

Michael L. Aubrey

Michael Aubrey is an Assistant Professor of Chemistry at the University of Texas at Austin who specializes in the electrochemistry of materials. Prof. Aubrey graduated from UNC Chapel Hill in 2011 with a BS in Chemistry while developing ceramic materials for the purification of aluminum, earned his PhD at UC Berkeley under the supervision of Jeffrey Long on the electrochemical and conductive properties of metal–organic frameworks, and completed post-doctoral research under Hemamala Karunadasa at Stanford University on layered perovskite layered heterostructures. Since 2021, his research group has focused on 2D inorganic materials and the solid-state electrochemistry of di- and trivalent metals.

---

Introduction

Actuators that translate chemical diffusion, thermal expansion, and applied electric fields into mechanical forces are sought for their application in artificial muscle, robotics, and micromechanical systems.^1–4 Such devices should possess low weight, high strain amplitudes, high flexibility, low operating voltages, and fast response rates. As such, materials including dielectric elastomers, ionic hydrogels, shape-memory alloys and piezoceramics have been considered for use as actuators.^5–10

Electrochemical actuators are multilayer heterostructures that operate via a capacitive or half-reaction-induced strain in an electrode material, which can subsequently deform the device as a whole.^11–16 Actuators of this type are composed of the electrochemically active layer bonded to an electrochemically inert substrate (Fig. 1a). During reduction, the active layer expands due to processes including expansion of the crystal lattice, insertion of charge balancing cations, and solvent swelling. This expansion may be limited by strong intermolecular interactions between the active and substrate layers, resulting in traction forces at the interface and out-of-plane deformation of the entire structure. This is then reversed by oxidation of the active layer in which the triggering process is reversed allowing for multiple drive cycles. Electrochemical actuators have previously demonstrated relatively large strain amplitudes (>1%),^17–19 generated stresses comparable or greater than biological muscle (0.35 MPa),^12,20–22 and operate within sub-volt potential ranges.^14,23 Furthermore, electrochemical actuators allow for precise control by tuning the extent of the reaction, and the ability to be operated either in solution or as self-contained solid-state devices.^24,25

Fig. 1 (a) Heterointerface between a redox-active material and a conductive substrate. Following electrochemical reduction, the active layer expands and traction forces induce bulk deformation of the electrode. (b) Schematic and dimensions of multilayer electrodes fabricated. (c) Interfacial structures of MoS₂ with either Au or Ni and (d) V₂O₅·0.5H₂O with either Au or Ni. (e) Schematic of a three-electrode cell used to determined changes in curvature, κ, as function of electrochemical potential.

Previous electrochemical actuators have been constructed from conducting polymers with high strains (>10%),^26–28 or stiff materials such as carbon nanotubes and nanoporous metals which typically exhibit lower strains (0.15–1.0%).^29–33 More recently, transition metal oxides and chalcogenides (NiO,³⁴ MnO₂,¹⁷ Ni(OH)₂,¹⁵ V₂O₅,¹² W₁₈O₄₉,³⁵ MoS₂ ²¹) have been used in electrochemical actuators. These materials are highly crystalline, possess high elastic moduli (5–35 GPa),^12,20 and are capable of relatively larger strains (1.38–5.3%).^17,19 Many have also been prepared as nanosheets and nanoribbons to facilitate film uniformity and ion diffusion.^21,36

The physical interface between active and substrate layers has been an important design factor, with increased transduction of strains across the interface attributed to the entanglement between layers or their geometric frustration.^12,20,24 The strength of chemical interactions across interfaces, such as strong Au–thiol interactions, have also been proposed to correlate strongly with induced stresses in molecular machines and monolayers.^37,38 It has also been shown that increased adhesion between thin films and substrates is achievable through functionalization of polymers to prevent delamination or fracture.^39,40 Based on these observations, under certain conditions stabilizing chemical interactions likely increase strain transduction across an actuating interface. Similar to biological muscle fibers in which myosin head groups physically attach to actin filaments and drag along the fibers to cause muscle contractions strong intermolecular interactions will drag along the substrate transducing the expansion of the active layer across the interface.

However, classical models of multilayer actuators do not account for such interatomic interfacial interactions, despite accurately describing a wide range of previously reported systems.^14,19,41,42 For example, Timoshenko's beam bending models for multi-layer systems models actuation strains in electrochemical actuators with micron scale layers using only layer thickness and bulk elastic moduli.⁴³ Other models have attempted to account for electrostatic interactions in molecular actuators,³⁷ however a holistic understanding of if, or to what degree other chemical bonding interactions influence bulk actuation remains under developed. Deviating from classical models that do not account for electrostatic interatomic interactions, we expect increased interaction strengths to manifest as an apparent increase in actuation strain when layer thicknesses and elastic moduli are accounted for.

To compare the strength of interfacial attractions and their impact on actuation performance, herein we prepared a series of flexible trilayer structures composed of a 25–100 μm Kapton films, a thin layer of Au or Ni metal, and a wet cast layer of redox active material: either MoS₂ or V₂O₅·0.5H₂O nanocrystals (Fig. 1b). MoS₂ and V₂O₅·0.5H₂O, were chosen owing to their significantly different surface chemistries when in contact with the metallic substrates. MoS₂ is expected to display a stronger Au–S interaction^44,45 than Ni–S interaction (Fig. 1c), aiding in strain transduction through the layers. For similar reasons, V₂O₅·0.5H₂O paired with Ni may participate in hydrogen bonding between surface hydroxides and vanadyls, while vanadyl oxide donors interact more weakly with noble Au via van der Waals contacts (Fig. 1d).^46,47 As such, we expected to observe greater transduction of strain in MoS₂|Au and V₂O₅·0.5H₂O|Ni systems. Experimentally, this difference in actuation strain was observed as a change in curvature of a multilayer electrode (Fig. 1e) during potential sweep and step measurements.

Experimental

Materials and chemicals

All reagents were purchased from commercial vendors and used without further modification unless otherwise noted. Kapton films were obtained from Dupont. Vanadium dioxide (97%, STREM) was stored in a nitrogen-filled glovebox until used. Propylene carbonate was degassed using the freeze–pump–thaw method and then packed into a nitrogen-filled glovebox and stored over dry 4 Å sieves. Sieves were prepared by heating at 210 °C while under vacuum for 24 hours at 50 mTorr.

Synthesis of MoS₂ nanosheets

Exfoliated MoS₂ nanosheets were prepared as previously reported with slight modifications.⁴⁸ MoS₂ (300 mg, ThermoFisher Scientific, 98%), 3 mL 1.6 M n-butyllithium in hexanes (Sigma-Aldrich), and 10 mL of hexanes were added to a Schlenk flask under dinitrogen atmosphere then stirred for 3 days. The solid product was collected by cannula filtration and washed three times with 3 mL of hexanes. The product was then washed twice with 3 mL portions of isopropyl alcohol added dropwise and the filtrate was quenched with excess isopropyl alcohol. Exfoliation of the lithiated material was carried out by resuspending the powder at a concentration of 1 mg mL⁻¹ and sonicating for 1 h in a bath sonicator. The product was then centrifuged at 8000 rpm (8659 RCF) for 12 hours to yield 1T-MoS₂ nanosheets suspended in the supernatant.

Fabrication of 1T-MoS₂ electrodes

25 μm thick Kapton film was cut into 25 × 2 mm strips. Then 100 nm of Au or 100 nm of Ni was deposited on one side of the Kapton film by PVD using a Kurt J Lesker PVD75 E-beam and Sputtering system. Films were then transferred into Teflon troughs and a few drops of the suspended nanosheet material was deposited on top of the film and dried at ambient temperature under a constant flow of dinitrogen gas. Once dry, the film was rinsed in deionized water and dried. This was repeated until the total volume of suspension deposited was between 0.5 and 1 mL. Electrochemical experiments were carried out within one day of fabrication.

Synthesis of V₂O₅·0.5H₂O

A 1 : 4 mixture of VO₂ (8 mg, STREM) and V₂O₅ (32 mg, Sigma-Aldrich) was dispersed in 28.5 mL of deionized water by bath sonication for 30 min. The mixture was then heated for 48 hours at 85 °C. The resulting greenish black suspension of V₂O₅·0.5H₂O nanosheets was centrifuged at 8000 rpm (8659 RCF) for 20 hours and produced a light brown-orange supernatant and dark green-black pellet. The supernatant was deposited onto the substrate the same day. Using solution promptly was necessary to avoid settling of the dispersed solid.

Fabrication of V₂O₅·0.5H₂O electrodes

25 μm thick Kapton film was cut into 25 × 2 mm strips. Then 100 nm of Au or 25 nm of Ni was deposited on each side of the Kapton film by PVD using a Kurt J Lesker PVD75 E-beam and Sputtering system. The resulting bilayer films were adhered to glass slides with Kapton tape and heated to about 200 °C on a hotplate. About 20 μL of V₂O₅·0.5H₂O supernatant solution was spray coated onto films using a nebulizer. Films were then transferred into Teflon troughs and dried at 110 °C in an oven. Once dried, troughs were removed from the oven and additional solution was dropcast onto the films which were then loaded back into the oven until dried. This process was repeated until 130 μL of solution had been deposited and dried. Electrodes were stored in an N₂-filled glovebox.

Determination of electrode active material mass

The mass of active material deposited on electrodes was determined by inductively coupled plasma optical emission spectroscopy (ICP-OES) using an Agilent 5800 and Agilent SPS 4 Autosampler. The deposited layers of 1T-MoS₂ or V₂O₅·0.5H₂O were dissolved in nitric acid. The nitric acid solution was then filtered through 0.1 μm Watman nylon syringe filter and diluted to about 2% in water. Concentrations were determined by fitting vanadium and molybdenum emission lines to a standard calibration curve.

Material characterization

Electron micrographs of deposited V₂O₅·0.5H₂O and 1T-MoS₂ were collected using an Apreo 2C loVac and FEI Quanta 650. Powder X-ray diffraction (PXRD) was collected using Rigaku II Miniflex G6 equipped with a 600 W Cu Kα source, a 0.5° soller slit and Hypix-400MF hybrid pixel array detector. Fourier transform infrared (FTIR) spectroscopy was collected using a PerkinElmer Spectrum III IR equipped with a Pike GladiATR attenuated total reflectance accessory and diamond window. Diffuse reflectance UV-vis spectroscopy was collected using a PerkinElmer Lambda 365. UV-vis diffuse reflectance was converted to a pseudo-absorbance using the Kubelka–Munk transformation and scaled by the Jacobian to plot as a function of photon energy.⁴⁹ Water content was determined by thermogravimetric analysis using a TA instruments Q50 at a scan rate of 1 °C min⁻¹ between 25 and 500 °C under flow of nitrogen. Layer thicknesses of V₂O₅·0.5H₂O and 1T-MoS₂ were determined by optical profilometry using a Keyence VK-1100 optical profilometer with a 404 nm laser and 96 000× magnification.

Mechanical characterization

Young's moduli of the V₂O₅·0.5H₂O and 1T-MoS₂ layers were determined by nanoindentation using a Bruker Hysitron TI950 to measure the out of plane moduli of the materials which was then converted to the in-plane modulus using their respective Poisson ratios. Samples were fabricated as in analogous fashion to their respective electrode except that, for nanoindentation, the active layers were prepared to be >1 μm thick, as determined by optical profilometry. The films were then superglued onto magnetic sample pucks. To avoid substrate influence, the indentation depth was set to 100 nm and averaged over 200 points across the sample.

Electrochemical measurements

All electrochemical characterization was performed using a Biologic SP-200 potentiostat.

For V₂O₅·0.5H₂O electrodes, three-electrode cells were constructed using an Ag/AgNO₃ reference electrode and platinum wire counter electrode (Fig. 1e). The V₂O₅·0.5H₂O working electrodes were submerged in 0.5 M solutions of NaClO₄ in propylene carbonate inside an N₂-filled glovebox. The cell was then sealed prior to cyclic voltammetry (CV) and chronoamperometry (CA) measurements. CV measurements were performed between −0.8 and +1.0 V vs. Ag/AgNO₃ with scan rates ranging from 10 mV s⁻¹ to 100 mV s⁻¹. CA measurements were conducted following a potential step from the open circuit voltage to −0.8 V vs. Ag/AgNO₃ for four minutes followed by a potential step to +1.0 V vs. Ag/AgNO₃ for four minutes.

All 1T-MoS₂ electrode measurements were collected against an Ag/AgCl reference electrode and Pt counter in a 0.5 M aqueous solution of sodium sulfate. CV was carried out at a scan rate ranging from 100 mV s⁻¹ to 1000 mV s⁻¹ over a potential window of −0.3 to +0.3 V vs. Ag/AgCl. CA was conducted following a potential step from +0.3 to −0.3 V vs. Ag/AgCl with a 30 seconds hold at each voltage.

Actuation measurements

The custom three-electrode cells were designed with transparent windows through which the electrode was imaged during electrochemical measurements using a digital microscope (Amscope UTP200X020 MP). During actuation measurements the cell was stood vertically to avoid twisting of the working electrode. Changes in electrode curvature were quantified by fitting the image of the electrode to a circle for each frame. Curvatures along with each layers' thicknesses and moduli were then used to find actuation strain according to a multilayer beam bending model:⁴³

κ = c_κa (1)

where κ is curvature (m⁻¹), a is actuation strain, m_i is the thickness of layer h_i divided by thickness of the Kapton (h₁), n_i is the ratio of Young's moduli between each layer E_i and Kapton (E₁), and

Y₂ = 4m₂n₂ + 6m₂²n₂ + 4m₂³n₂ + m₂⁴n₂² (3)

Y₃ = 4m₃n₃ + 6m₃²n₃ + 4m₃³n₃ + m₃⁴n₃² (4)

Y₂₃ = m₂m₃n₃[(4m₂² + 6m₂m₃ + 4m₃²)n₂ + 12(1 + m₂ + m₃)] (5)

Results and discussion

MoS₂ active layers

Dispersions of 1T-MoS₂ nanosheets were synthesized by reductive intercalation of 2H-MoS₂ with 1.6 M n-butyllithium in hexanes, followed by exfoliation via forced hydration and sonication.⁴⁸ The nanosheet suspension was wet-cast onto a bilayer film of 100 nm Au or Ni atop 50 μm Kapton. The solution was allowed to dry at room temperature under N₂ flow to avoid reversion to 2H-MoS₂ at higher temperatures.⁵⁰ Surface morphology of the restacked material was imaged using SEM revealing a polycrystalline morphology of 1 μm sheets (Fig. 2a).

Fig. 2 (a) SEM image of active layer of 1T-MoS₂ on a Au substrate. (b) PXRD pattern of lithiated MoS₂ before and after exfoliation. Diffractogram was collected by depositing 1T-MoS₂ onto a glass substrate.

PXRD was performed before and after exfoliation of the Li_xMoS₂ phase (Fig. 2b). PXRD patterns of the deposited material displayed significant broadening of peaks, consistent with the expected reduced crystallite size and increased intracrystalline strain in the exfoliated product. The (001) and (002) reflections were observed in both the as-synthesized Li_xMoS₂ powder and liquid-phase exfoliated material. The shift to lower angle is characteristic of restacking of exfoliated sheets and water intercalation between MoS₂ monolayers following sonication in water.^51,52

Electrochemical characterization of 1T-MoS₂ films

CVs were performed in a three-electrode cell vs. a Pt counter and Ag/AgCl reference electrode in aqueous 0.5 M sodium sulfate electrolyte. An aqueous system was chosen for its faster rate kinetics, which allowed for more rapid data collection, and for the smaller voltage window (−0.3 to 0.3 V) needed to perform actuation. The metallic substrates Au and Ni were both stable at the neutral pH within the electrochemical stability window of this electrolyte. The qualitative shape of the voltammogram was consistent with a capacitive process and consistent with previous observations made for metallic 1T-MoS₂ electrodes (Fig. 3a).⁵³ Films with similar masses had thicknesses ranging from 110 to 220 nm, and were found to have specific capacitances between 81 and 56 F g⁻¹ for the positive charging current and between −78 and −46 F g⁻¹ for the negative charging current respectively. This indicates that thicker films likely possess lower surface areas and/or lower porosity. The capacitance of the electrode was constant with sweep rates between 100 and 1000 mV s⁻¹ (Fig. 3b). The larger negative current density was consistent with the n-type character of 1T-MoS₂.

Fig. 3 (a) Cyclic voltammetry of films composed of either 1T-MoS₂|Au|Kapton or 1T-MoS₂|Ni|Kapton using a 0.5 M Na₂SO₄ electrolyte at 100 mV s⁻¹, a Ag/AgCl reference electrode and Pt wire counter electrode. (b) Sweep rate dependence of 1T-MoS₂|Au|Kapton at rates of 100, 200, 300, 400, 500, 600, and 1000 mV s⁻¹.

Actuation performance

Actuation of 1T-MoS₂ on Au and Ni. Chemomechanical coupling between the charge state of the 1T-MoS₂ and the trilayer was observed as a change in the curvature of the electrode during electrochemical measurements. Curvature was determined as a single value for the entire length of the working electrode by fitting its image to a circle by a combination of random sample consensus and least-squares minimization. The change in curvature was expected to be proportional to the actuation strain (eqn (1)).

For the electrode shown in Fig. 4a and b, 1T-MoS₂ was deposited on the inner radius of the electrode. As potential was swept from 0.3 to −0.3 V, the curvature of the electrode decreased, consistent with the expected expansion of 1T-MoS₂ resulting from both intercalation of Na⁺ and added electron density on the 1T-MoS₂ sheets (ESI Movie S1†). The change in strain of all 1T-MoS₂ films during actuation measurements was approximately linear with respect to potential, consistent with the linear increase in charge vs. potential for a purely capacitive electrode (Fig. 4c). The mechanism of actuation is attributed to reconfigurations of bond lengths and interatomic distances at the surface of the electrode during formation of the electrical double layer as well as swelling of the material due to the influx of electrolyte. The hysteresis observed on the Ni system above 0 V was attributed to the lower conductivity of the Ni substrate at a scan rate of 100 mV s⁻¹.

Fig. 4 Images of a 1T-MoS₂|Au|Kapton working electrode at (a) +0.3 V and (b) −0.3 V using a 0.5 M Na₂SO₄ electrolyte at 100 mV s⁻¹, an Ag/AgCl reference electrode and Pt wire counter electrode. (c) Change in curvature as a function of potential during CV over one cycle for of either 1T-MoS₂|Au|Kapton or 1T-MoS₂|Ni|Kapton at scan rates of 100, 500 and 1000 mV s⁻¹ (additional cycles available in ESI Fig. S5†). (d) Scan rate dependence of curvature for a 1T-MoS₂|Au|Kapton working electrode. (e) Current response for a potential step from −0.3 V to +0.3 V vs. Ag/AgCl using either an Au or Ni substrate. (f) Curvature response following the potential step from −0.3 V to +0.3 V vs. Ag/AgCl shown in (e).

On both gold and nickel, the change in curvature over the potential window for an electrode is constant for scan rates up to 100 mV s⁻¹, after which the maximum observed strain begins to decrease. At scan rates of 500 and 1000 mV s⁻¹, 89% and 86% of the maximal observed stain, respectively, was retained (Fig. 4d and ESI S1†).

To further demonstrate the kinetic response, a potential step experiment was performed via chronoamperometry between potential of 0.3 and −0.3 V. Both the Au and Ni systems demonstrated rapid electrochemical equilibration (Fig. 4e). On gold, the current decayed to 5% its initial value within 5 s, whereas on Ni, the same occured within 8 s. The actuation response equilibrated more quickly. The change in strain of the Au system reached steady state in under a second, at a rate too rapid to be quantified. The change in strain of the Ni system, in contrast, took about 2 s to equilibrate following the potential step (Fig. 4f). The slower response on Ni was attributed to the lower conductivity of the Ni substrate.

Electrochemical actuation. The Young's modulus for the 1T-MoS₂ deposit was determined using nanoindentation (ESI Fig S2†). The modulus of 1T-MoS₂ layer was found to be 1.50 GPa, assuming a Poisson's ratio of 0.3.⁵⁴ This value is similar to, but lower than the previously reported modulus of 2.5 GPa for this material likely due to differences in deposition conditions.²¹ Wet casting likely formed a less compact, and therefore less stiff, active layer than the vacuum filtration-based deposition method previously reported. Actuation strain was determined using the trilayer beam bending model (eqn (1)–(5)). Average actuation strains and stresses over a set of 7 electrodes were found to be 1.29(0.13)% and 19(2) MPa on a Au (Table 1). A comparison with previously reported MoS₂ actuators is presented in Table S5.†

Table 1 Calculated curvature (κ), strain% (ε), stress (σ), and number of samples (n) for 1T-MoS₂|Au|Kapton and 1T-MoS₂|Ni|Kapton electrodes with average thicknesses of 160(10) nm and 130(8) nm respectively

	1T-MoS2|Au|Kapton		1T-MoS2|Ni|Kapton
	CV	CA	CV	CA
κ (m^−1)	2.5(0.2)	2.37(0.18)	0.63(0.04)	0.54(0.04)
ε (%)	1.29(0.13)	1.23(0.07)	0.57(0.05)	0.49(0.04)
σ (MPa)	19(2)	15(1)	8.6(0.1)	7.8(0.6)
n	7	7	7	7

---

There was a significant decrease in actuation performance on the Ni substrates compared to Au. In the case of nickel actuation strains and stresses over a set of 7 electrodes were found to be 0.57(0.05)% and 8.6(0.1) MPa. This observed difference in calculated strains accounts for the higher stiffness on Ni, and yet it persists.

Interatomic interactions at the 1T-MoS₂|Au(Ni) interface

The particularly strong interaction between Au and the sulfur on the topmost layer of 1T-MoS₂ of the restacked nanosheets may be one possible reason for the higher induced strain in the gold system when compared to the nickel system.^55–57 Au–S interactions being stronger than Ni–S interactions is supported by comparing the heats of adsorption for sulfur onto Au and Ni. Reported enthalpy of adsorption are approximately 240 kJ onto Au and 135–189 kJ onto Ni.^58,59 As the active material is the same for both systems, the resulting actuation strains would be expected to be the same should these interfacial bonding interactions prove insignificant. The thinner the film, the smaller the number of nanosheet–nanosheet interactions and therefore it is likely that the effects of the 1T-MoS₂|Au interface become more pronounced. In films much thinner than a micron we find that the strength of interfacial interactions or other factors not accounted for in classical models may significantly impact actuation.

V₂O₅·0.5H₂O active layers

The compound V₂O₅·0.5H₂O was synthesized using a slightly modified version of a previously reported synthesis.³⁶ VO₂ and V₂O₅ were combined in 1 : 4 by mass ratio in DI water and heated to 85 °C for 5 days. Solutions were then centrifuged at 8000 rpm (8659 RCF) for 20 h to obtain a suspension of nanofibers. Working electrodes were prepared by first depositing 100 nm (25 nm) Au (Ni) onto both sides of a Kapton film, then on only one side spray coating a thin layer of the suspended V₂O₅·0.5H₂O supernatant before drop casting an additional 130 μL at 110°C (ESI Fig. S7†). The average mass and thickness of V₂O₅·0.5H₂O deposited across 5 and 13 electrodes respectively were determined by ICP-OES and optical profilometry to be 0.066(4) mg and 570(40) nm respectively. Metal thicknesses of Au (100 nm) and Ni (25 nm) were chosen to account for the greater elastic modulus of Ni, such that equivalent actuation strains would produce approximately equivalent changes in electrode curvature.

Surface morphology of the active layer was imaged using SEM, after heating the deposited V₂O₅·0.5H₂O films an intergrown nanofiber morphology was observed (Fig. 5a). The bulk V₂O₅·0.5H₂O powder was further characterized by PXRD (Fig. 5b), FTIR spectroscopy, and UV-vis diffuse reflectance spectroscopy. The PXRD pattern showed only the (0 0 l) family of peaks, which is attributed to strong preferential orientation of the V₂O₅·0.5H₂O layers parallel to the electrode plane.

Fig. 5 (a) SEM of the top face of a V₂O₅·0.5H₂O|Au|Kapton electrode. (b) PXRD pattern for V₂O₅·0.5H₂O deposited onto a borosilicate glass slide and dried at either 25 °C or 110 °C.

In the FTIR spectrum of V₂O₅·0.5H₂O (ESI Fig. S8†), peaks at 3000 and 1600 cm⁻¹ are assigned to the O–H stretching and bending modes of water, while a peak at 992 cm⁻¹ matched the expected stretching mode of the vanadyl. Other peaks at 738 and 472 cm⁻¹ originate from bridging V–O–V and V–O stretches respectively.⁶⁰

Diffuse reflectance spectroscopy revealed stronger absorption in the sub-2 eV region in V₂O₅·0.5H₂O that was not present in V₂O₅ starting material and the hemihydrate appeared visually as a very dark green of the hydrate compared to the characteristic orange of V₂O₅ (ESI Fig. S9†). V₂O₅·0.5H₂O was also observed to exhibit electrochromic behavior during electrochemical measurements, transitioning from a dark green color to a light orange upon oxidation in accordance with the transition between V⁴⁺ and V⁵⁺ in accordance with previous studies (ESI Fig. S10†).²⁴

Electrochemical characterization of V₂O₅·0.5H₂O films

CV was performed on V₂O₅·0.5H₂O|Au|Kapton and V₂O₅·0.5H₂O|Ni|Kapton electrodes using a 0.5 M NaClO₄ propylene carbonate electrolyte, an Ag/AgNO₃ reference electrode, a platinum wire counter electrode, and a scan rate of 10 mV s⁻¹ (Fig. 6a). The propylene carbonate electrolyte was chosen to access a wider potential window than was available under aqueous conditions in previously demonstrated electrochemical actuation using V₂O₅ nanofibers.^12,24 The broad redox peaks in Fig. 6a are indicative of a reversible pseudocapacitive process involving a combination of reductive intercalation of sodium into V₂O₅·0.5H₂O and capacitive charging of the crystallites' surfaces.^61,62 Collecting CV as a function of scan rate between 10–100 mV s⁻¹ (Fig. 6b) revealed a more resistive response at faster scan rates, compared to 1T-MoS₂. This was consistent with the lower electronic conductivity of the films estimated by 2-contact impedance spectroscopy to be 2 × 10⁻⁶ S cm⁻¹. Current density was found to increase in proportion to the square root of the scan rate as expected for a diffusion limited process (ESI Fig. S11†).

Fig. 6 (a) Cyclic voltammetry of films composed of either V₂O₅·0.5H₂O|Ni|Kapton or V₂O₅·0.5H₂O|Au|Kapton using a 0.5 M NaClO₄ electrolyte at 10 mV s⁻¹, a Ag/AgNO₃ reference electrode and Pt wire counter electrode. (b) Sweep rate dependence of V₂O₅·0.5H₂O|Ni|Kapton at rates of 10, 15, 20, 40, 80, and 100 mV s⁻¹ (V₂O₅·0.5H₂O|Au|Kapton data available in ESI Fig. S13†).

CA was performed in the same cell using potentials steps from the open circuit potential to −0.8 V followed by alternating steps between −0.8 V and 1.0 V, holding at each potential for 4 minutes. Current decay during CA measurements also aligned with diffusion limited Cottrell behavior (ESI Fig. S12†).

Actuation performance

Actuation of V₂O₅·0.5H₂O on Au and Ni. Each electrodes strain was monitored during CV and CA as changes in strain relative to the open circuit state. In Fig. 7a, at the open circuit potential of 0.3 V vs. Ag/AgNO₃, V₂O₅·0.5H₂O is on the inner radius for a V₂O₅·0.5H₂O|Au|Kapton electrode. As the potential was swept in the negative direction V⁵⁺ is formally reduced to V⁴⁺, causing the active layer's lattice to expand with the increase in ionic radii and intercalation of charge balancing sodium cations. The expansion applies tensile traction forces across the substrate surface. Because this only occurs at the active-substrate layer interface these forces decrease the curvature of the electrode (Fig. 7b) as the active layer expands. Once the sweep direction was reversed, vanadium electrodes remained at sufficiently reducing potentials and continued to expand (Fig. 7c). Once sufficiently oxidizing potentials were reached vanadium V⁴⁺ oxidizes to V⁵⁺ and the lattice contracts. The electrode's strain correspondingly increased up to 0.8 V and ultimately returned to its initial state once the potential was swept back to 0.3 V, which is consistent with a chemomechanically reversible process.

Fig. 7 Images of (a) V₂O₅·0.5H₂O|Au|Kapton and (b) V₂O₅·0.5H₂O|Ni|Kapton working electrodes at 1.0 and −0.8 V potentials vs. Ag/AgNO₃ reference electrode with Pt counter electrode in 0.5 M NaClO₄ PC electrolyte at 10 mV s⁻¹ scan rate. (c) Change in strain% as a function of potential during CV over one cycle for of either V₂O₅·0.5H₂O|Au|Kapton or V₂O₅·0.5H₂O|Ni|Kapton with 744 and 601 nm thick active layers respectively. (d) Scan rate dependence of curvature for V₂O₅·0.5H₂O|Ni|Kapton working electrode (additional cycles available in S14 & S15†). (e) Current response for a potential step from 0.0 V to −0.8 V then from −0.8 to +1 V vs. Ag/AgNO₃ using either an Au or Ni substrate with peak reductive currents of −1.22 and −2.66 mA respectively. (f) Strain% response following the potential steps from shown in (e).

Fig. 7c plots the change in strain during a single CV sweep for both V₂O₅·0.5H₂O|Au|Kapton and V₂O₅·0.5H₂O|Ni|Kapton systems (additional cycles available in S14 & S15†). At the upper limits of the potential window, the change in strain plateaus as all the accessible V⁴⁺ has been oxidized and the corresponding CV's current diminishes towards zero. The non-linear change in strain distinguishes V₂O₅·0.5H₂O from 1T-MoS₂ and originates from the concomitant faradaic reaction of a limited species.^15,23 1T-MoS₂ in contrast demonstrated a largely capacitive steady state current response resulting in a linear strain response, Fig. 3b and 4d.^21,63,64

During initial cycling, electrodes start and end points shifted in curvature and strain% values (Fig. S14†). This was due to a rearrangement of the active material during cycling previously labeled as “creep”^37,65 and diminished with further cycling appearing to stabilize overtime. Associated with this rearrangement there was an approximate loss of 0.04% in the actuation strain between the first and ninth recorded cycles.

V₂O₅·0.5H₂O|Ni|Kapton electrodes deformation response was slower compared to corresponding Au electrodes (Fig. 7c). This contributed to lower overall average changes in curvature and strain% during swept potentials for V₂O₅·0.5H₂O|Ni|Kapton at 10 mV s⁻¹ as reported in Table 2. We attributed this to the greater charge transfer resistance from Ni to V₂O₅·0.5H₂O relative to the Au electrodes.

Table 2 Calculated curvature (κ), strain% (ε), stress (σ), and number of samples (n) for V₂O₅·0.5H₂O|Au|Kapton and V₂O₅·0.5H₂O|Ni|Kapton electrodes with average thicknesses of 500(50) nm and 560(20) nm respectively

	V2O5·0.5H2O|Au|Kapton		V2O5·0.5H2O|Ni|Kapton
	CV	CA	CV	CA
κ (m^−1)	18(3)	20(3)	16(1)	22(2)
ε (%)	1.12(0.16)	1.24(0.20)	0.81(0.04)	1.17(0.08)
σ (MPa)	16(2)	18(3)	12(0.6)	17(1)
n	6	6	7	7

---

The effect of increasing scan rates on actuation response was observed in Fig. 7d. Due to the pseudo-capacitive nature of V₂O₅·0.5H₂O, greater scan rates did not provide sufficient time for the intercalation of cations into the Na_xV₂O₅·0.5H₂O lattice during reduction, leading to the maximum observed strain decreasing by 50% between sweep rate of 10 mV s⁻¹ and 100 mV s⁻¹ (ESI Fig. S16 and S17†). This behavior is consistent with previously reported electrochemical actuators.^20,23,63,66

Because scan rates kinetically limited the degree of reduction, CA measurements were conducted (Fig. 7e) to provide a longer time for equilibration at the potential limits. For both the Au and Ni electrodes a rapid change in strain was observed followed by a slower increase in strain over longer time spans (Fig. 7f). In contrast, the current was observed to decay and equilibrate more rapidly than the strain following the potential step (ESI Movie 2†). The faster process was assigned to capacitive charging of the crystallite surfaces. Following this the electrode would continue to curve and strain due to the much slower faradaic process and intercalation of sodium. This supports the diffusion of sodium during reduction as the kinetically limiting process to the actuation response. The strain% of V₂O₅·0.5H₂O|Au|Kapton appeared to saturate more quickly relative to Ni. This was found to contrast with the responses in MoS₂ systems which mechanically saturated much faster following a potential step, Fig. 3f.

Quantitatively, V₂O₅·0.5H₂O|Au|Kapton electrodes demonstrated average displacements (Δκ) of 18(3) and 19(3) m⁻¹ for CV and CA respectively, while V₂O₅·0.5H₂O|Ni|Kapton reached 16(1) and 22(2) m⁻¹ displacements during CV and CA measurements respectively (Table 2). Actuation strains were calculated via the same multilayer beam bending model as used for 1T-MoS₂. A Young's modulus for V₂O₅·0.5H₂O of 1.46 GPa was determined by nanoindentation (ESI Fig. S18†) and the Poisson ratio of the similarly structured anhydrous V₂O₅.⁶⁷ Based on these findings it was found that due to the slower pseudocapacitive mechanism of V₂O₅·0.5H₂O CA provided better quantitation of steady state changes in electrode curvature and strain%. In the case of actuation strains determined by CA, no significant difference was observed after 4 min of polarization (Table 2). A comparison with previously reported V₂O₅ actuators is presented in Table S4.†

Actuation response and changes in lattice parameters. As an alternative to experimentally estimating Young's modulus and Poisson ratio via nanoindentation, instead the crystallographic strain determined for Na_xV₂O₅·0.5H₂O can be interpolated using Vegard's law.⁶⁸ First the charge passed during electrochemical measurements was used to estimate the equivalents of sodium cations intercalated into V₂O₅·0.5H₂O. This ratio was then used to determine lattice parameters Na_xV₂O₅·0.5H₂O based on a linear fit to reported lattice parameters of known compounds in the series between 0 ≤ x ≤ 1.38.^69,70 Given the crystallites' observed preferential orientation, crystallographic strain could not be experimentally determined by PXRD. Instead, the strain was calculated from the change in area of the unit cells' ab-planes with respect to x = 0. Changes along the c-axis of the unit cell of V₂O₅·0.5H₂O do not significantly contribute to the bulk displacement of the electrode since this direction is largely normal to the electrode plane. This area strain was then used to estimate the effective modulus of the active layer and actuation stress (ESI Table S1†).

On V₂O₅·0.5H₂O|Au|Kapton actuation strains (from Vegards Law) and stresses were determined to be 1.84(10)% and 16(2) MPa via CV at 10 mV s⁻¹ and 2.15(10)% and 18(3) MPa via CA. In contrast, when using a Ni current collector, strains and stresses were determined to be 2.24(10)% and 11.7(10) MPa via CV at 10 mV s⁻¹ and 2.47(10)% and 17(1) MPa via CA (ESI Table S1†). Calculated strains were larger compared to those derived from the film's modulus determined by nanoindentation. This relative overestimation of strain by using the crystallographic values may be expected given imperfect packing of the polycrystalline film. This subsequently resulted in an estimated lower bound for the Young's modulus of the V₂O₅·0.5H₂O layer 0.78(6) GPa averaged across all Au and Ni electrodes. The Young's modulus measured here is approximately one half the value of that observed by nanoindentation (1.46 GPa).

Interatomic interactions at the V₂O₅·0.5H₂O|Au(Ni) interface. Overall, for the case of V₂O₅·0.5H₂O significant deviations from either the multilayer or Stoney models (ESI Table S2†)⁷¹ were not observed as a difference in actuation stress to the degree observed for MoS₂-based electrodes. Heats of adsorption of methanol to V₂O₅ (−70 kJ mol⁻¹) and methanol to Au (−39 kJ mol⁻¹) are reported and may serve as a self-consistent ad hoc estimate of the difference in interaction strength between hydrogen bonding at V₂O₅–NiOH interfaces and the van der Waals interactions at V₂O₅–Au.^72,73 Following this little difference may be expected in the interfacial interaction strengths between both systems.

Loss of weak interfacial interactions during actuation. Noticeably, in MoS₂|Au and MoS₂|Ni electrodes average differences were observed of up to 0.74% and 10.4 MPa in actuation strains and stresses respectively. At the atomic scale, this is attributed to the stronger interaction between the sulfide in MoS₂ and the Au substrate, compared to the considerably weaker Ni–S interaction. By increasing interlayer adhesion, stronger interatomic interactions between the active layer and the metallic allow for greater traction forces to be applied before a potential slip dislocation or detachment. As the MoS₂ layer expanded, the Au–S interactions induce a tensile force across the Au surface. However, in the MoS₂|Ni system where van der Waals interactions are weaker these interatomic tensile stresses are significantly smaller. This is similar to crosslinked polymers and hydrogels where upon mechanical deformation weaker bonding interactions are easily broken and reformed while stronger bonding interactions are retained.^74–77

In V₂O₅·0.5H₂O based electrodes, the formal interaction between the oxide in V₂O₅ and the metallic surface are certainly weaker than the MoS₂'s sulfide-based van der Waals interactions with the same metals. Additionally, the difference in interaction strength based on molecular heats of adsorption for Au–thiol and Ni–thiol systems is much larger than that for V₂O₅·0.5H₂O–Au and V₂O₅·0.5H₂O–Ni systems (78 vs. 31 kJ mol⁻¹),^58,59,72,73 and thus a more pronounced difference in actuation stress may be expected. This may account for the lack of a noticeable difference in observed actuation strains and stresses between V₂O₅·0.5H₂O|Au and V₂O₅·0.5H₂O|Ni because both interactions are much weaker overall. Thus, we hypothesize that for only sufficiently strong interatomic forces across the heterointerface changes the surface chemistry can modulate actuation strain in bilayer systems for single component active layer films with thicknesses in the range of 100–1000 nm.

It is also possible the rougher crystal faces of V₂O₅·0.5H₂O and relatively low surface concentration of hydrogen bond donors on the nickel surface afford much weaker interfacial interactions than are estimated from molecular adsorption. In the case of MoS₂, the flat hexagonal close packing of sulfides and the oriented Au (1 1 1) counterparts likely results in a relatively higher interfacial concentration of these particularly strong van der Waals interactions. Finally, uniform 1T-MoS₂ films could be prepared with a thinner average thickness, as the films become thinner interatomic interactions across heterointerfaces may be expected to become more significant.

Conclusions

Through the fabrication of electrodes composed of two distinct electrochemically active materials—1T-MoS₂ and V₂O₅·0.5H₂O—on two different metal surfaces—Ni and Au—the dependance of bonding interactions across heterointerfaces on chemomechanical actuation was examined. In the case of a 1T-MoS₂ actuation on Au was significantly higher than on Ni, consistent with the expectedly stronger Au–S interactions at the heterointerface. In the case of V₂O₅·0.5H₂O electrodes, the lower conductivity resulted in relatively slower electrode kinetics and no significant difference in actuation was found between electrodes composed of either Au or Ni. Future actuating systems, particularly those employing thin films less than 100 nm, could be improved by the inclusion of strong chemical bonds across heterointerfaces.

Data availability

The data supporting this article have been included as part of the ESI.† This includes supplementary electrochemical characterization, nanoindentation results, ESI tables of experimentally determined mechanical properties, adsorption spectroscopy, and ESI videos† of.

Author contributions

M. L. A conceived of and directed the project. J. B. and K. M. wrote the manuscript. J. B. performed V₂O₅·0.5H₂O experiments. K. M performed all experiments using 1T-MoS₂.

Conflicts of interest

The authors declare are no conflicts of interest.

Acknowledgements

This work was supported by the University of Texas at Austin. The authors thank Elizabeth Recker at the University of Texas at Austin for the collection of Nanoindentation measurements.

References

1. M. Shi and E. M. Yeatman, Microsyst. Nanoeng., 2021, 7, 95.
2. J. D. W. Madden, N. A. Vandesteeg, P. A. Anquetil, P. G. A. Madden, A. Takshi, R. Z. Pytel, S. R. Lafontaine, P. A. Wieringa and I. W. Hunter, IEEE J. Oceanic Eng., 2004, 29, 706–728.
3. Q. Deng, H. Jia, C. An, S. Wu, S. Zhao and N. Hu, Composites, Part A, 2023, 165, 107336.
4. D. Li, J. Li, P. Wu, G. Zhao, Q. Qu and X. Yu, Adv. Intell. Syst., 2024, 6, 2300070.
5. P. Brochu and Q. Pei, Macromol. Rapid Commun., 2010, 31, 10–36.
6. R. Pelrine, R. Kornbluh, Q. Pei and J. Joseph, Science, 2000, 287, 836–839.
7. E. Hajiesmaili and D. R. Clarke, J. Appl. Phys., 2021, 129, 151102.
8. K. J. Kim, J. Caligiuri, K. Choe and M. Shahinpoor, Contraction/elongation behavior of cation-modified polyacrylonitrile fibers, ed. Y. Bar-Cohen, San Diego, CA, 2003, p. 207.
9. C. Lexcellent, Shape-memory Alloys Handbook, Wiley, Hoboken, New Jersey, U.S., 1st edn, 2013.
10. Q.-M. Wang, Q. Zhang, B. Xu, R. Liu and L. E. Cross, J. Appl. Phys., 1999, 86, 3352–3360.
11. J. D. Madden, R. A. Cush, T. S. Kanigan and I. W. Hunter, Synth. Met., 2000, 113, 185–192.
12. G. Gu, M. Schmid, P.-W. Chiu, A. Minett, J. Fraysse, G.-T. Kim, S. Roth, M. Kozlov, E. Muñoz and R. H. Baughman, Nat. Mater., 2003, 2, 316–319.
13. M. Hakamada, S. Matsumura and M. Mabuchi, Mater. Lett., 2012, 70, 132–134.
14. K.-W. Kwan, N.-Y. Hau, S.-P. Feng and A. H.-W. Ngan, Sens. Actuators, B, 2017, 248, 657–664.
15. C. Cheng and A. H. W. Ngan, ACS Nano, 2015, 9, 3984–3995.
16. C. Lu, Y. Yang, J. Wang, R. Fu, X. Zhao, L. Zhao, Y. Ming, Y. Hu, H. Lin, X. Tao, Y. Li and W. Chen, Nat. Commun., 2018, 9, 752.
17. S.-Q. Wang, B. Zhang, Y.-W. Luo, X. Meng, Z.-X. Wang, X.-M. Luo and G.-P. Zhang, ACS Appl. Mater. Interfaces, 2022, 14, 9281–9291.
18. N. T. Hung, A. R. T. Nugraha and R. Saito, J. Phys. D: Appl. Phys., 2018, 51, 075306.
19. W. Ma, K. W. Kwan, R. Wu and A. H. W. Ngan, J. Mater. Chem. A, 2021, 9, 6261–6273.
20. L. Liu, L. Su, Y. Lu, Q. Zhang, L. Zhang, S. Lei, S. Shi, M. D. Levi and X. Yan, Adv. Funct. Mater., 2019, 29, 1806778.
21. M. Acerce, E. K. Akdogan and M. Chhowalla, Nature, 2017, 549, 370–373.
22. I. W. Hunter and S. Lafontaine, in Technical Digest IEEE Solid-State Sensor and Actuator Workshop, IEEE, Hilton Head Island, SC, USA, 1992, pp. 178–185.
23. D. Pang, M. Alhabeb, X. Mu, Y. Dall'Agnese, Y. Gogotsi and Y. Gao, Nano Lett., 2019, 19, 7443–7448.
24. Y. Ling, L. Li, J. Liu, K. Li, C. Hou, Q. Zhang, Y. Li and H. Wang, Adv. Mater., 2024, 36, 2305914.
25. Y. Li, P. Yu, W. Ma and L. Mao, ACS Appl. Mater. Interfaces, 2023, 15, 56158–56166.
26. K. K. C. Lee, Y. Dobashi, P. R. Herman, J. D. W. Madden and V. X. D. Yang, Smart Mater. Struct., 2019, 28, 015028.
27. D. Pang, X. Wang, C. Liu, H. Xu, G. Chen, F. Du, Y. Dall'Agnese and Y. Gao, Appl. Surf. Sci., 2022, 583, 152403.
28. S.-Q. Wang, B. Zhang, X.-M. Luo, F. Liang and G.-P. Zhang, J. Mater. Chem. A, 2020, 8, 5679–5687.
29. R. H. Baughman, C. Cui, A. A. Zakhidov, Z. Iqbal, J. N. Barisci, G. M. Spinks, G. G. Wallace, A. Mazzoldi, D. De Rossi, A. G. Rinzler, O. Jaschinski, S. Roth and M. Kertesz, Science, 1999, 284, 1340–1344.
30. J. N. Barisci, G. M. Spinks, G. G. Wallace, J. D. Madden and R. H. Baughman, Smart Mater. Struct., 2003, 12, 549–555.
31. K. Mukai, K. Asaka, T. Sugino, K. Kiyohara, I. Takeuchi, N. Terasawa, D. N. Futaba, K. Hata, T. Fukushima and T. Aida, Adv. Mater., 2009, 21, 1582–1585.
32. J. Wilmers and S. Bargmann, Extreme Mech. Lett., 2018, 21, 57–64.
33. J. Weissmüller, R. N. Viswanath, D. Kramer, P. Zimmer, R. Würschum and H. Gleiter, Science, 2003, 300, 312–315.
34. G. Wu, G. H. Li, T. Lan, Y. Hu, Q. W. Li, T. Zhang and W. Chen, J. Mater. Chem. A, 2014, 2, 16836–16841.
35. K. Li, Y. Shao, H. Yan, Z. Lu, K. J. Griffith, J. Yan, G. Wang, H. Fan, J. Lu, W. Huang, B. Bao, X. Liu, C. Hou, Q. Zhang, Y. Li, J. Yu and H. Wang, Nat. Commun., 2018, 9, 4798.
36. A. S. Etman, A. K. Inge, X. Jiaru, R. Younesi, K. Edström and J. Sun, Electrochim. Acta, 2017, 252, 254–260.
37. B. K. Juluri, A. S. Kumar, Y. Liu, T. Ye, Y.-W. Yang, A. H. Flood, L. Fang, J. F. Stoddart, P. S. Weiss and T. J. Huang, ACS Nano, 2009, 3, 291–300.
38. R. Berger, E. Delamarche, H. P. Lang, C. Gerber, J. K. Gimzewski, E. Meyer and H.-J. Güntherodt, Science, 1997, 276, 2021–2024.
39. I. Lee and R. P. Wool, Macromolecules, 2000, 33, 2680–2687.
40. L. Ouyang, B. Wei, C. Kuo, S. Pathak, B. Farrell and D. C. Martin, Sci. Adv., 2017, 3, e1600448.
41. L. B. Freund, J. Mech. Phys. Solids, 1996, 44, 723–736.
42. Y. Wang, S. Li, L. Liu, S. Feng, K. Guan, Y. Shi, F. Sun, X. Wang, Y. Shen, C. Zhang, Q. Liu, T. Li, T. Zhang and S. Qin, J. Mater. Chem. A, 2024, 12, 27549–27557.
43. P. Du, X. Lin and X. Zhang, Sens. Actuators, A, 2010, 163, 240–246.
44. A. Ulman, Chem. Rev., 1996, 96, 1533–1554.
45. L. H. Dubois and R. G. Nuzzo, Annu. Rev. Phys. Chem., 1992, 43, 437–463.
46. E. S. Lambers, C. N. Dykstal, J. M. Seo, J. E. Rowe and P. H. Holloway, Oxid. Met., 1996, 45, 301–321.
47. A. Y. Faid and S. Sunde, Energy Technol., 2022, 10, 2200506.
48. P. Joensen, R. F. Frindt and S. R. Morrison, Mater. Res. Bull., 1986, 21, 457–461.
49. J. Mooney and P. Kambhampati, J. Phys. Chem. Lett., 2013, 4, 3316–3318.
50. F. Wypych and R. Schöllhorn, J. Chem. Soc. Chem. Commun., 1992, 1386–1388.
51. P. Joensen, E. D. Crozier, N. Alberding and R. F. Frindt, J. Phys. C: Solid State Phys., 1987, 20, 4043–4053.
52. B. Vedhanarayanan, J. Shi, J.-Y. Lin, S. Yun and T.-W. Lin, Chem. Eng. J., 2021, 403, 126318.
53. M. Acerce, D. Voiry and M. Chhowalla, Nat. Nanotechnol., 2015, 10, 313–318.
54. R. C. Cooper, C. Lee, C. A. Marianetti, X. Wei, J. Hone and J. W. Kysar, Phys. Rev. B: Condens. Matter Mater. Phys., 2013, 87, 035423.
55. L. Agmon, R. Almog, D. Gaspar, G. Voscoboynik, M. Choudhary, J. Jopp, Z. Klausner, A. Ya’akobovitz and R. Berkovich, Tribol. Int., 2022, 174, 107734.
56. S. B. Desai, S. R. Madhvapathy, M. Amani, D. Kiriya, M. Hettick, M. Tosun, Y. Zhou, M. Dubey, J. W. Ager, D. Chrzan and A. Javey, Adv. Mater., 2016, 28, 4053–4058.
57. G. Z. Magda, J. Pető, G. Dobrik, C. Hwang, L. P. Biró and L. Tapasztó, Sci. Rep., 2015, 5, 14714.
58. T. Miyahara, K. Stolt, D. A. Reed and H. K. Birnbaum, Scr. Metall., 1985, 19, 117–121.
59. P. Carro, G. Andreasen, C. Vericat, M. E. Vela and R. C. Salvarezza, Appl. Surf. Sci., 2019, 487, 848–856.
60. C. Sanchez, J. Livage and G. Lucazeau, J. Raman Spectrosc., 1982, 12, 68–72.
61. Y. Wang, H. Shang, T. Chou and G. Cao, J. Phys. Chem. B, 2005, 109, 11361–11366.
62. H. An, X. Li, C. Chalker, M. Stracke, R. Verduzco and J. L. Lutkenhaus, ACS Appl. Mater. Interfaces, 2016, 8, 28585–28591.
63. K. W. Kwan, R. Wu and A. H. W. Ngan, IEEE Trans. Compon. Packag. Manuf. Technol., 2023, 13, 928–934.
64. L. Ji, Y. Yu, Q. Deng and S. Shen, Nanoscale, 2020, 12, 15643–15651.
65. T. E. Herod and J. B. Schlenoff, Chem. Mater., 1993, 5, 951–955.
66. R. Wu, K. W. Kwan, Y. Wang and A. H. W. Ngan, Adv. Mater. Technol., 2023, 8, 2202062.
67. Y. Zhu, Y. Zhang, L. Dai, F.-C. Cheong, V. Tan, C.-H. Sow and C.-T. Lim, Acta Mater., 2010, 58, 415–420.
68. L. Vegard, Z. Phys., 1921, 5, 17–26.
69. V. Petkov, P. N. Trikalitis, E. S. Bozin, S. J. L. Billinge, T. Vogt and M. G. Kanatzidis, J. Am. Chem. Soc., 2002, 124, 10157–10162.
70. J. Orive, R. Fernández De Luis, E. S. Larrea, A. Martínez-Amesti, A. Altomare, R. Rizzi, L. Lezama, M. I. Arriortua, J. L. Gómez-Cámer, M. Jauregui, M. Casas-Cabanas and J. Lisoni, Dalton Trans., 2020, 49, 3856–3868.
71. C. A. Klein, J. Appl. Phys., 2000, 88, 5487–5489.
72. P. Boulet, A. Baiker, H. Chermette, F. Gilardoni, J.-C. Volta and J. Weber, J. Phys. Chem. B, 2002, 106, 9659–9667.
73. L. Mastroianni, T. Weckman, K. Eränen, V. Russo, D. Yu. Murzin, K. Honkala and T. Salmi, J. Catal., 2023, 425, 233–244.
74. C. Creton, Macromolecules, 2017, 50, 8297–8316.
75. K. Mayumi, C. Liu, Y. Yasuda and K. Ito, Gels, 2021, 7, 91.
76. T. L. Sun, T. Kurokawa, S. Kuroda, A. B. Ihsan, T. Akasaki, K. Sato, Md. A. Haque, T. Nakajima and J. P. Gong, Nat. Mater., 2013, 12, 932–937.
77. P. Cordier, F. Tournilhac, C. Soulié-Ziakovic and L. Leibler, Nature, 2008, 451, 977–980.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta02659h

‡ These authors contributed equally to this work.

---