Multiscale in-situ quantification of the role of surface roughness and contact area using a novel Mica-PVS triboelectric nanogenerator

Triboelectric nanogenerators (TENGs) are energy harvesters generating electricity via the triboelectric effect and electrostatic induction. However, the influence of interface mechanics on TENG performance requires attention. Here, we study the effect of random multiscale surface roughness on TENG performance using a novel in-situ optical technique to directly visualise the contact interface. To achieve this, a new type of TENG is developed based on transparent mica in contact with polyvinyl siloxane (PVS). A wide range of surface roughness instances were created on the PVS surface ( Sq from 1.5 to 82.5 µm) by replicating 3D-printed masters developed from numerically generated rough surfaces. TENG output was found to be highly sensitive to surface roughness over a wide range of forces and frequencies. The dependence of real contact area on


Introduction
Sustainable energy generation is a pressing global concern at the present time and critical in our daily lives. As technology becomes increasingly miniaturised and society marches toward a ‗smart world' where sensors and various self-powered devices become more popular, the need for sustainable power sources with decreased carbon emissions is clear [1,2]. Currently, batteries are used; however, these are cumbersome and have negative environmental implications [3]. A great way to make use of the world's abundant energy is to utilise devices that can integrate themselves into the surroundings and harness ambient energy. Triboelectric nanogenerators (TENGs) are an emerging and rapidly growing technology that can sustainably harvest electrical power from mechanical energy. TENGs rely on an oscillating contact scenario. During contact, charges transfer across the interfacea phenomenon known as triboelectrification. After separation of the charged planes, an electric field develops and induces charge on backing electrodes via electrostatic induction [4][5][6]. Connecting with an external circuit causes flow of charge to balance the potential difference. As the gap is closed, current flows in the opposite direction and an AC signal continues as long as the contact cycle repeats. The practical operation of TENGs can be broadly classified into two modes: normal contact-separation mode and sliding mode [6][7][8]. Over the last decade, TENGs have gained a significant amount of interest from researchers particularly in the materials science and electrical engineering communities. In recent years, a very wide range of material combinations have been trialled in the quest to optimise TENG output [4,9,10].
Usually, selecting materials widely spaced on the triboelectric series (ranking of a material's tendency to gain or lose electrons) promotes a stronger triboelectric effect and yields higher electrical output [11,12]. Polymers are commonly used materials in triboelectric nanogenerators [13]: they offer multiple advantages such as light-weight, flexibility, energy absorption, and corrosion resistance, just to name a few [14]. Common polymers used in triboelectric systems include polydimethylsiloxane (PDMS), polyethylene terephthalate (PET), polytetrafluoroethylene J o u r n a l P r e -p r o o f (PTFE), polyurethane rubber, polyethylene (PE) and polyvinyl chloride (PVC), etc. [13,[15][16][17][18][19].
Even though a significant amount of research has been done in the field of TENGs, the interfacial contact mechanics and the role of multi-scale surface texturing on TENG output performance is not well understood yet [4].
In general, the output performance of TENGs is strongly influenced by the physio-chemical characteristics of the contact pair. Parameters such as applied contact pressure, oscillating frequency, surface topography, and real contact area all have a bearing on performance [5,6,9,15,20]. This multi-parameter dependency makes it a complex system to systematically investigate and to independently evaluate the role of individual parameters. For instance, the role of surface roughness has not been systematically explored. However, keeping the same combination of materials and developing multi-scale surface roughness on one of the tribosurfaces could provide a way to systematically investigate the role of multi-scale topography on the electrical performance of TENGs. Over the last decade, a substantial literature has grown up around device fabrication, and physico-chemical modification of the bulk and surface properties of the tribo-materials. Several researchers have explored functionalisation (such as, silane monolayer deposition and coating) of TENG surfaces to achieve high power output [21][22][23][24][25][26][27][28][29].
Many others considered various modifications of the bulk tribo-materials: for instance, by developing nano-composites to improve TENG output [22,24,[30][31][32][33]. Surprisingly, on the mechanics and tribology side, the role and understanding of interface mechanics (at a local microscopic scale) has only recently begun to be addressed [4]. In the triboelectrification process, the contact mechanics of the interacting surfaces is likely to play a crucial role in determining electrical performance. Some interesting recent research has moved in the direction of exploring the effect of different kinds of interface topography. Surface topographies with well-defined and engineered shapes (for example, square, cylinder, half-spherical, pyramid arrays, etc.) have been deployed to explore their effect on TENG performance [34][35][36][37][38][39][40]. Choi et al. [34] were able to boost output using nano-pillars on either and both tribolayers (Nickel-PDMS TENG), but the mechanisms remained largely unexplored. Work by Kim et al. [37] used well-defined laserpatterned textures (concave hemispherical shaped) on a PDMS tribo-layer (in contact with aluminium) to optimise TENG performance. Their analysis suggested contact area as the critical parameter in governing the response, but they made no measurements of contact area. In another work by Seol et al. [38], the authors considered pyramid textures on PDMS in (contact with flat Ag). They showed that electrical output increased with contact pressure and later saturated at higher pressures. They also gave a simple optical demonstration of the square contact areas at the pyramid contacts increasing simultaneously as a glass slide was pushed into the surface. All these studies focused on engineered topographies with well-defined structured (non-random) J o u r n a l P r e -p r o o f geometries and presented only a qualitative description of the real contact area. However, standard engineering surfaces usually possess multi-scale surface roughness varying from nanometres to a few hundred microns, at different hierarchical levels. In a recent work by Min et al. [15], a systematic study of the role of contact area was carried out for TENGs with random rough surfaces (copper foil and PET). Here, the key finding was that the electrical output (e.g., open circuit voltage and short circuit current) increased with contact pressure in almost the same way as the contact area, with both saturating at approximately the same pressure levelthus, indicating that real contact area governs electrical output (mostly because intimate contact promotes electron transfer). Indeed, Xu et al. [20] has successfully modelled the pressuredependent behaviour of TENGs by combining the mechanics of rough surface deformation with established electrostatic approaches. However, in Min et al. [15], the real contact area measurement was carried out using an indirect method involving insertion of a third body (a combination of two pressure sensitive films) between the tribolayers. In addition, the study by Min et al. [15] looks at only a single surface roughness.
Recent advances in optical visualisation techniques and their growing application in contact mechanics investigations have made it possible to precisely visualize the distribution of real contact junctions down to sub-micron scale and also to investigate the contact dynamics in detail [41][42][43][44][45][46]. In this paper, we develop a direct in-situ optical contact area visualisation technique for TENGs. This allows for a detailed quantitative description of the contact interfaces during the TENG operation and a simultaneous comparison between electrical power output and real contact area. We then use the technique to systematically investigate the role of multi-scale surface roughness and real contact area (while keeping the tribo-material properties and surface chemistry constant). We begin by developing a new high-performance TENG based on the novel combination of hard (and transparent) mica in contact with soft polyvinyl-siloxane (PVS)hereafter referred to as a ‗Mica-PVS TENG'. Polyvinyl siloxane (PVS) is a silicone based viscoelastic polymer. It is a well-known biomaterial and widely used in dentistry to develop tooth imprints [47,48]. Key characteristics of PVS polymer include low-cost, easy commercial availability, simple processing, low modulus, high flexibility, and non-toxicity. Mica, on the other hand, is a hard phyllosilicate mineral characterised physically by a perfect basal cleavage capable of yielding atomically smooth surfaces. It has high dielectric strength and, crucially for the optical visualisation, is transparent.
Surfaces with different scales of surface roughness (ranging Sq from 1 µm to 100 µm) were numerically designed and fabricated on the PVS polymeric material using a combined 3D printing and micro-moulding technique. An electrodynamic mechanical test rig was then modified to perform highly controlled electro-mechanical measurements with a surface self-J o u r n a l P r e -p r o o f aligning facility and a reflection interference microscopy-based optical setup to precisely record the distribution of real contact area. The highly accurate setup was also used to quantitatively investigate the effect of contact pressure, oscillating frequency, and external load resistance on TENG output. The work systematically characterises the effect of surface roughness (and contact pressure and frequency) on TENG power output across a wide surface roughness range and interprets the results using direct in-situ contact area measurements. In summary, the key novelties of this work are: the new high-performance TENG based on PVS in contact with mica, the direct optical technique for interface visualisation & contact area measurement in TENGs and the comprehensive study on the effect of multiscale surface roughness on TENG output.

Textured PVS tribo-negative layer fabrication
A number of random rough surface realisations covering a range of areal RMS surface roughness (Sq) values were designed and fabricated. The surfaces were first designed numerically based on a selected power spectrum density using the numerical algorithm reported in Perris et al. [49].  (Figure 1a). One could also develop a larger or smaller surface area, if required for a specific application. A controlled thick border of ca. 300 μm in height was created just outside the textured region to achieve a constant thickness for the developed tribo-layers (Figure 1a).
The border edge helps maintain the same active surface area for all tribo-layers. As mentioned, polyvinyl siloxane (or PVS) was used to create the textured surfaces. To the best of the authors' knowledge, PVS has not been used previously in the fabrication of TENGs. PVS is usually used in biomedical (dental) applications. Here, we use it because it has good mould replication capability and is a flexible and cost effective tribo-negative material that can be moulded and cured very easily and rapidly. The dielectric constant of PVS was measured at around 3.15.
Measurement details and dielectric constant vs. frequency curve ( Figure S1) are provided in the Supporting Information. The PVS (President -The Original, Coltène Switzerland) polymer mixture was slowly poured on the 3D printed moulds using an automatic mixing and dispensing gun and then gently pressed in place using a flat glass plate (maintaining a uniform pressure J o u r n a l P r e -p r o o f distribution). While pressing down, the excess polymer mixture comes out of the mould volume.
After polymerization for ca. 10 min at room temperature, the PVS replicas with micro-textures were gently removed from the mould. Produced PVS replicas were then used to form the negative tribo-layer in a contact-separation mode TENG (PVS Tribo-layer in Figure 2b-c). For convenience, we refer to the four PVS replica samples using the following designation: FlatPVS for the nominally flat PVS surface with Sq = 1.5 µm and PVS25, PVS50 and PVS100 for the PVS surfaces having Sq values of 25, 50 and 100 μm, respectively. Negative moulds with numerically generated multi-scale topographies were manufactured using a stereolithography 3D printer. PVS mixture was slowly poured onto the mould (a) and uniformly pressed with a flat glass plate (b). The textured PVS replica was then slowly peeled off from the mould (c).

Mica tribo-positive layer and Mica-PVS TENG fabrication
For the tribo-positive layer, we used ultra-smooth ruby muscovite mica thin-sheets (Agar Scientific Ltd, Stansted, Essex, CM248GF, UK)again, a material little used in TENG research.
There are a number of advantages to using mica in the present study. First, thin mica sheet exhibits very high optical transparency, which is crucial for the in-situ optical visualization of real contact area [42,50,51]-one of the key objectives of our work. Also, mica sheets are easily cleavable and give a highly smooth and atomically flat surface [52]. Combined with the fact that mica is hard (Young's Modulus ≈ 5.4 GPa [53]) in comparison to soft PVS (Youngs modules ≈ 3.4 ± 0.2 MPasee Section 3.4), this allows us to consider the contact as equivalent to a rigid flat in contact with the soft rough surface. Another advantage of mica for TENG construction is its high dielectric strength (κ ≈ 7.0 [54]): higher capacitance of the tribo-layer means greater ability to induce charge on the electrode. The commercial mica sheet comes with a thickness of 150 μm.
It was carefully cut into 25 × 25 mm 2 area, using a sharp scalpel. A thin copper tape (3M Copper foil with conductive adhesive, UK) was attached on the back side of the mica tribo-layer, which acts as an electrode. After this, the back side of the copper tape was insulated with an ultra-thin Kapton tape. For the other tribo-layer, the four different textured PVS sheets described in Section 2.1 were used (active area of 25 × 25 mm 2 ). The PVS sheets were securely pasted on copper tape J o u r n a l P r e -p r o o f and then insulated with Kapton tape. A schematic of the TENG device in the test rig is presented in Figure 2a with photos of the two separate tribo-layers (with connected electrodes) shown in Figure 2c. Since this study focuses on the role of multi-scale surface roughness and real contact area, it was crucial to keep both samples globally flat. Due to the very low thickness of the tribolayers (they are easily bended or distorted), all tribo-layers were attached to smooth glass backing plates using transparent double-sided tape. A long copper lead-wire was attached to the electrodes and this was done outside of the active contact region (Figure 2c) to avoid disturbing the contact formation and pressure distribution. With both tribo-layers having the same active area, the effective dimension of the TENG devices was 25 × 25 mm 2 .

Surface characterization and mechanical testing of PVS samples
Surface morphology visualization and characterization of all PVS samples was carried out using a 3D optical profiler (InfiniteFocus, Alicona-Bruker, Austria). All measurements were carried out at 5x magnification. For the 3D visualization of captured images, MountainsMap (Digital Surf, France) surface metrology software was used. In addition, 3D printed moulds were quickly visualized under the optical light microscope to check any surface imperfection. A universal compression testing machine (Instron 3367, USA) was used for the mechanical characterization of PVS material (Supporting Information, Figure S2). Cylindrical test samples were prepared using the moulding technique. A stress-strain curve was obtained by applying the normal load at a slow loading rate of 1 mm/min. Young's modulus (modulus of elasticity) was then estimated as the slope of stress-strain curve within the proportional limit (straight-line portion). To calculate the Poisson's ratio of PVS, a micro-mechanical testing setup (Dual leadscrew, Deben, UK) along with a high-definition camera system was used (see Figure S3 in Supporting Information). An insitu synchronised video of the polymer sample was recorded during the deformation stage. A digital-image-correlation (DIC) based method (Imetrum Video Gauge, Imtetum, UK) was used to estimate the Poisson's ratio (see Perris et al. [55]).

TENG test rig and electrical measurement
The TENG characterisation work was then carried out in normal contact-separation mode. A detailed schematic of the test rig is presented in Figure 2a. A high-precision linear electrodynamic fatigue testing machine (Electropuls E3000, Instron UK) was adapted for the tests. A 50 N capacity load cell (tension and compression) was fitted at the actuation drive. A special tribo-surface holder with a flat base was designed and machined to maintain the flatness of the top tribo-layer. Since the work is exploring a flat-on-flat contact configuration, it was J o u r n a l P r e -p r o o f crucial to achieve a perfectly parallel contact. To achieve this, a 360° alignment bed was incorporated to attach the lower tribo-layer onto it (Figure 2d). The alignment bed offers a large degree of freedom in rotation, yawing and pitching along the centre axis. During the development of this test rig, we noticed that a slight alternation in contact alignment could drastically alter the electrical output of the TENGs. The same issue has been reported in a previous work by Hong et al. [56]. Before starting each test, a pre-load (about 10 N) was slowly applied to establish the contact and alignment between the two tribo-layers. Once an aligned contact is established, the lower tribo-layer position was locked by tightening the knob on the adjustable table (Figure 2d).
The test rig is well capable of high frequency oscillations with synchronised data acquisition at high-resolution. Before starting the tests for each type of TENG device, the stiffness of the contacting system was estimated. This is an important step to obtain the correct dynamic impact behaviour via the test machine's feed-back loop. To investigate the effect of contact pressure, Figure 2: (a) Schematic of the test rig for high-precision contact-separation TENG measurements, (b) Schematic of the Mica-PVS TENG construction, (c) Photograph of developed TENG tribo-layers and (d) Close-up photo of the test rig around the TENG including the 360° alignment stage. The mica tribo-layer (upper surface) was attached on base plate and then fixed to an in-house fabricated holder connected to a load cell. The PVS tribo-layer (lower surface) was fixed on the 360° alignment stage. A sample video of the operating test rig is provided in Supplementary Information (Video S1). N = normal load, d = separation distance, and f = oscillating frequency.

In-situ optical real contact area measurement:
To achieve real contact area visualization for TENG contacts, a novel setup was developed and incorporated into the TENG testing rig. Owing to one surface being transparent (the mica), the reflection interference microscopy approach could be used to record images of the actual contact interfaces [51,58,59] clearly distinguishing the areas of contact and non-contact. A simplified schematic showing the working principle and optical setup is presented in Figure 3a. The telecentric lens used provides a unique in-line illumination lighting option. A fibre optic light source was connected to the illumination port to direct the light onto the lens. The optical system is capable of horizontal and vertical resolution of ca. 5.5 µm. It offers a field of view of 11.26 mm × 5.9 mm and this was used as the region for contact area analysis (an area large enough to be representative of the full nominal contact area). The small working distance (40 mm) of the optical lens allowed us to insert the whole optical visualization system within the compact volume available. A high-definition CMOS sensor camera (Pixelink C-Mount USB 3.0 Camera), with progressive scan and high frame rate, was attached to the telecentric lens and fixed on a focusing stage. Getting a perfect focus is very important to achieve a sharp contrast between the contact and non-contact area. Therefore, after application of normal load, the focus was finely adjusted (Figure 3a). A representative real contact image captured using this setup is shown J o u r n a l P r e -p r o o f  Recorded real contact images were analysed with Fiji, an open-source image processing package based on ImageJ2, (National Institutes of Health, USA). The Fiji tool allows for uneven illumination corrections, homogenous filtering, and thresholding procedures [60]. Thresholding was performed based on the pixel grayscale intensity difference between contact and out-ofcontact regions (Figure 3c). All the data analysis and graphing was performed in Python. Owing to the non-transparent nature of the electrodes, the optical images were obtained from separate quasi-static load-up of the TENG samples (without an electrode on the mica) to the contact pressure used for analysis (64 kPa for the TENG tests comparing surface roughness instances in Section 3.3).

Morphological & mechanical characterization of tribo-layers
The results of surface morphological characterization of the PVS tribo-layers are presented in  Table 1 and compared to the design roughness values (see SI for more details on Sq). We note that the as-produced roughness values are somewhat less than the design values: this is likely due to the resolution limitation of the 3D printing technique hampering the realisation of the more steeply sloped features (i.e. those contributing to higher surface roughness) [49]. However, the technique is perfectly adequate for the present purpose which is to achieve rough PVS surfaces covering a wide roughness range. The surfaces comprise an ideal sample-set for our examination of the role of surface roughness in effecting TENG performance.
The method of fabricating the PVS tribo-layers is reliable and robust in producing high-precision controlled textured samples. Indeed, the developed 3D printed mould can be used for rapid production of multiple PVS samples. In addition, the procedure easily allows scaling of the TENG nominal area to produce larger or smaller TENG devices. PVS is a low-cost silicone based soft viscoelastic polymer. The Young's modulus and Poisson's ratio of the PVS material was experimentally measured at 3.4 ± 0.2 MPa and 0.45 respectively. The complete stress-strain curve obtained from the mechanical testing of PVS is provided in the Supporting Information ( Figure S4). A key advantage of using PVS lies in its ultra-fasting curing time (ca.10 min), simple handling and easy processing [61]. PVS tribo-samples could be developed without using any expensive and sophisticated machinery. Before discussing the results of the in-situ optical characterisation of the TENG interface, we first explore the contact pressure and frequency dependence of the Mica-PVS TENG and how these are effected by roughness.    [15]; therefore, calculating the nominal contact pressure allows for a more meaningful comparison. The working mechanism of the PVS-Mica TENG is schematically demonstrated in Figure S5 in the Supporting Information. Figure 5 shows peak-to-peak output voltage (V out ) vs time for the Mica-FlatPVS TENG at different nominal contact pressures. The output voltage was recorded after a few hundred cycles (at the stable functioning stage). After running the TENG device at a given condition, the tribo-layers were kept relaxed for few minutes to allow time for the material to relax. This is especially important when one of the tribo-layers (PVS) is a viscoelastic material [42]. A notable feature of Figure 5 is the symmetric (about the zero voltage level) and uniform voltage peak output signal (at all pressures). Given the authors experience with other testing setups (e.g. using  contact pressure for four different frequencies in Figure 6. It is clear from the results that V out increases with increasing contact pressure, regardless of tribo-layer surface roughness or oscillating frequency. This contact pressure-dependent behaviour of V out has been noted previously [15,38,63,64]. It is understood to be due to the real contact area increasing with pressurea phenomenon well accepted in the tribology literature [41,42,65,66]. Indeed the results in Min et al. [15] clearly indicate that electrical output tracks real contact area as contact pressure is increased. It is considered that the total surface tribo-charge increases when the real contact area increasesthus leading to the increase in output voltage. Interestingly, if we closely examine the pressure dependent results in Figure 6, the sensitivity J o u r n a l P r e -p r o o f of output voltage to contact pressure (slope of the graph) appears to be frequency dependent with the lowest frequency (3 Hz) being least sensitive to pressure. This is likely to be due to the amount of real contact area being more sensitive to pressure for smoother surfaces (where the initial contact area at low pressures is sparse, but can increase quickly as load is applied). This phenomenon could be further explained using the Greenwood-Williamson rough contact model or Persson's theory [67]. According to this J o u r n a l P r e -p r o o f model, at low pressure loading situations, the real contact area evolution can be estimated by following equation: where | | is the surface gradient, is a constant of √ and √ and is plane strain modulus. Assuming the (mean) charge density is constant, then the slope of output voltage vs.
average contact pressure is proportional to the partial derivative of real contact area with respect to average contact pressure, i.e., | | . Since work by Perris et al. [49] showed that the surface gradient increases with increasing the , it is expected that the slope of output voltage decreases with . The surface roughness-dependent pressure sensitivity of the output voltage in Figure 7 will be an important consideration for those designing TENG based pressure sensorsas sensitivity can clearly be adjusted by surface roughness. The frequency of contact-separation is a crucial parameter influencing TENG performance.
For efficient and versatile operation, TENGs should be able to produce power in multifrequency situations. To evaluate the frequency dependence of output voltage and the effect of surface roughness, the voltage vs. frequency is plotted for each surface roughness case in

Role of surface roughness and real contact area
The short circuit current (I sc ) and open circuit voltage (V oc ) are the maximum current and maximum voltage produced from a TENG device [69]. In real engineering applications, an electrical load is connected to the power generating device (TENG) and it's power output can change depending on the external load resistance [69][70][71][72].  Figure 9 for each (PVS) surface roughness. As expected (Figure 9a-d) Figure 9). Clearly, the V oc reduces with increasing roughness on the PVS tribo-layerwe will probe this further later. Figure 9a-d) show the peak-to-peak current decreasing with increasing external resistance and reaching the lowest value (nearly zero) at the highest R (~5 GΩ). This is simply attributed to very high ohmic losses at high resistance [7,71]. The current value recorded at the very low (or negligible) R can be considered as the I sc [69]. The maximum I sc of 33.2 μA was measured for the Mica-FlatPVS TENG at a low value of R (~100 kΩ). I sc then decreased with increasing surface roughness to 22.8 μA, 13.8 μA and 9.36 μA for PVS25, PVS50 and PVS100, respectively. Instantaneous power was calculated using the recorded voltage and current (P = V × I) at each external load resistance. Power output against load resistance exhibits the characteristic n-shaped curve reported in previous work [7,30,70,73,74]. For all the four TENG devices, the P max was achieved in the range 30 MΩ -70 MΩ external load resistance.

Results on current-load (I -R) investigation (red plots in
J o u r n a l P r e -p r o o f  Figure 10b-e). Achieving direct and accurate real contact visualization on multi-scale rough contact interfaces is challenging. Min et al. [15] investigated the contact area in a TENG having nominally flat Cu and PET tribo-layers. However, the approach was indirect using a third body (pressure sensitive film) between the contact. Lee et al. [75] utilised an ink-transfer based approach, however the resolution remained unclear. A few other researchers qualitatively observed the contact area in TENGs, but on regular and defined structured surfaces [36,39]. It is clear from the visuals in Figure 10 that contact area reduces dramatically as the roughness is increased from the smooth Mica-FlatPVS case (Figure 10b) to the rough Mica-PVS100 case (Figure 10e). However, to quantify the amount of contact, we define the real contact area percentage as: The real contact area (A r ) is the sum of all the areas forming real solid-solid contact (black spots in Figure 10b-e) and the nominal contact area (A n ) is simply the apparent area of the imaged area for analysis. Real contact area and peak power are plotted together against PVS tribo-layer surface roughness in Figure 11. The Mica-FlatPVS TENG exhibited the highest real contact area percentage (47.7 %). Real contact area percentage then reduced with increasing surface roughness to 33.8, 21.6 and 12.6 % for PVS25, PVS50 and PVS100, respectively. The reason for this is clear from the images in Figure 10. For the Mica-FlatPVS TENG, the roughness is very low (Sq = 1.5 μm) and this permits a large contact area (Figure   10b), but for the Mica-PVS100 TENG, roughness is high (Sq = 82.5 μm) and far less contact is formed due to the steep slopes and deep valleys on the surface. The crucial result is that peak power output closely tracks real contact area in the plots against surface roughness in Figure 11. Thus, the underlying explanation for the roughness-dependent power output is the following. As roughness increases, contact area decreases and this reduces the amount of Mica-FlatPVS, PVS25, PVS50 and PVS100 tribolayers, respectively. The reason for reduction in the surface potential with high surface roughness on the PVS tribo-layers is likely attributable to the reduced real contact area contributing towards tribo-electrification (which should result in lower surface charge). Interestingly, the surface potential reduction trend follows the same trend observed for power output and real contact area (in Figure 11a).
These surface potential results support our argument on the critical role of surface roughness and real contact area in governing the the electrical performance of TENGs. Analysis of the contact visuals during the dynamic contact-separation cycle, indicated that the attachment and detachment phases were very smooth and stable. No explicit disturbances or sudden fluctuations were noticed and we did not observe any bending deformation. We believe that the smooth attachment and detachment contributed to the symmetric, uniform and noise-free electrical output signals shown in Figure 5. In addition, the distribution of contact areas in  Although electrical output decreased with increasing surface roughness in the present study, this is not a universal conclusion and the response to surface roughness can be rather subtle.
In fact, when surface roughness is increased, the real contact area between two contacting bodies may decrease or increase as compared to the idealised nominal smooth flat-on-flat contact depicted in Figure 12a. This depends on a number of factors such as the relative mechanical properties of the contact bodies, surface energy, interfacial loading conditions and on which surface the roughness was introduced (i.e. on the softer or harder surface) [65,79,80]. In this work, surface roughness was introduced on the softer tribo-layer (the PVS) and the counter contact surface (mica) was atomically flat. PVS is significantly softer than mica (Young's modulus of PVS and mica and are 3.4 ± 0.2 MPa and 5.4 GPa, respectively) [53]. This is somewhat equivalent to flattening a roughened soft surface by a rigid flat: at low and moderate loads, contact area will only occur around the higher roughness features.
Therefore, the case in the present work corresponds to Figure 12b. On the other hand, if we press a hard (rigid) rough surface into a soft flat surface to form a conformal contact, a contact area greater than even the nominal contact area is achievable (as indicated in Figure 12c).
This type of contact configuration has been reported in the literature [81][82][83][84]. Such real contact area enhancement could be used to advantage in future TENG design if the right combination of materials and topography are selected for the tribo-layers. A hard rough surface pressed into a soft surface will likely produce a 3D contact areait is not possible to J o u r n a l P r e -p r o o f visualise this using the optical technique in the present paper but a study on this contact scenario in TENGs would make an interesting future study. These results on power output and real contact area suggest the real contact area is a crucial parameter to consider in the optimisation TENG performance. Interestingly, Qiao et al. [85] showed that, if a high flexoelectric effect is present (i.e. high electrical polarisation in response to strain gradients in the material), then rougher surfaces (pyramid arrays in this case) can produce higher or lower TENG output (compared to flat-flat) depending on whether the flexoelectric and triboelectric effects add or subtract from each other. However, the appearance of a notable flexoelectric contribution may be limited to materials exhibiting a high flexoelectric coefficients.

PVS-Mica: A regenerative TENG, device stability and application
To examine the stability and durability of the Mica-PVS TENG, continuous contact-   Usually, mica sheet is easily cleavable using a scalpel and offers a highly flat (atomically smooth) and clean surface area [50]. After running several thousands of contact-separation cycles on the Mica-FlatPVS TENG, an ultra-thin layer was carefully cleaved from the bulk mica. Interestingly, after cleaving off a layer of mica (at about 26,000 cycles), the Mica-PVS TENG showed a surprising jump in output (V out from 240 to 278 V) as compared to the earlier un-cleaved situation (See Figure 13a). A possible explanation for the increment in power output after cleaving may be the fact that, the freshly cleaved mica is extremely clean and contamination free and facilitates a more intimate contact against the contacting PVS surface (the mica prior to cleaving will have been exposed to the ambient atmosphere for a long time and may have accumulated adsorbed molecules and dust particles). Moreover, pristine mica (i.e. obtainable after cleaving) exhibits a very high surface energy and strong wettability which could also serve to increase real contact area via increased adhesion [52,88]. This  Figure   13c shows the capacitor charging characteristics and confirms the smooth output behaviour during charging. Charging (oscillation on) and discharging (oscillation off) cycles with the 1 μF capacitor is presented in Figure 13d. Next, we offer a brief demonstration of the Mica-FlatPVS TENG operating various low-power electrical devices. A pattern of 42 large-sized (5 mm diameter) commercial LEDs (as the external load resistance) connected in series were simultaneously illuminated (Figure 13e). A full video of the LED illumination is provided in the Supporting Information (Video S2). To convert the AC signals generated from the TENG to usable DC output, a bridge rectifier was attached between the TENG device output and the LED circuit, as shown in Figure 13e. Furthermore, a commercial calculator (Figure 13f), a digital alarm clock (Figure 13g) and a digital vernier caliper (Figure 13h) were successfully powered using the Mica-FlatPVS TENG through a 100 μF capacitor (corresponding circuit diagram is presented in Figure S7). Full videos of each application are provided in the Supporting Information (Video S3-S5).

J o u r n a l P r e -p r o o f
Overall, the Mica-PVS TENG, demonstrates a competitive output performance (peak power density of 4256 mW/m 2 for contact pressure of 64 kPa), but with the benefit of low-cost construction, good flexibility and ease of manufacture (i.e. no chemical modifications are required and the PVS can be rapidly moulded and cured). Figure S8 in Supporting Information illustrates the flexibility of the device. A roughly indicative performance comparison with other TENG devices in the literature is given in  Table S1). The approximate cost for developing a working model of the Mica-PVS TENG comes out considerably lower than several other approaches giving comparable output (see Table S2 in Supporting Information for cost details).

Conclusions
TENGs are promising energy harvesters that generate electrical power from the rapid and repeated contact of a suitable tribo-electric material pair. However, the effect of the tribology and mechanics of the interface on TENG performance has not been explored comprehensively enough yet. This paper introduces a high-definition in-situ optical technique to image the real contact area developed at TENG interfaces. In addition, it develops a new type of TENG based on muscovite mica in contact with polyvinyl siloxane (or PVS). The new TENG and visualisation technique were then used to study how TENG electrical output responds to a wide variation in random multiscale surface roughness. Four instances of surface roughness were generated on the PVS surface (these were denoted according to their design roughness values as FlatPVS, PVS25, PVS50 and PVS100, but had measured roughness's of Sq = 1.5, 22.5, 40.5 and 82.5 µm). These were realised by 3D printing numerically designed multiscale rough surfaces and then moulding PVS tribo-layers from the 3D-printed master. The advantages of using a mica counter-surface include high dielectric strength, but crucially, high transparency to enable optical visualisation of the interface. Mica is also hard (compared to PVS) and near atomically smooth so that the contact scenario is equivalent to a rigid flat contacting a rough soft surface (the PVS).
The results show that electrical output is highly sensitive to surface roughness. The highest peak output power was found for the smoothest surface (FlatPVS with Sq = 1.5 µm) and this J o u r n a l P r e -p r o o f dropped by almost 70% when the roughness was increased as high as 82.5 µm (i.e. for PVS100). The in-situ interface imaging technique proved effective in accurately calculating the real contact area for each roughness. However, it is limited to the two-dimensional visualization of contact area. Crucially, real contact area reduced similarly (to power) as surface roughness was increased, thus indicating that the roughness effect on contact area is the likely reason for the roughness sensitivity of TENG performance. Clearly solid-solid contact is more likely to promote triboelectric charge transfer across the interface. The smooth surface can facilitate a high contact area and this reduces as roughness increases due to deep valleys on the surface which are not likely to make contact at low and moderate pressures. As we mention, this result is not universal and can depend on relative material properties. For example, if a hard rough surface is pushed into a soft countersurface, TENG output might be expected to rise with increasing roughness. Likewise, materials with a high flexoelectric effect have been shown to be capable of increasing TENG output with increasing roughness.
These results highlight the importance of understanding the contact area effect in TENG design because surface topography and material properties can clearly be chosen to optimise performance. The response of the Mica-PVS TENG to contact pressure and frequency was also explored: results confirm the characteristic increase in TENG output in response to increasing pressure or frequency, and in addition, show that TENG frequency and pressure response is more sensitive for smoother surfaces.
There were also some interesting findings on the testing of triboelectric nanogenerators: a very accurate mechanical testing approach (using an electrodynamic test machine) with surface self-alignment produced highly symmetric and uniform electrical output signals. The Mica-PVS TENG also showed excellent stability over large numbers of cycles and even