Feeling Smooth: Psychotribological Probing of Molecular Composition

In order to explore the surface chemical tactile space, we prepared a series of surfaces with similarly low roughness, where the chemical constituents vary systematically from polar to non-polar, and the water contact angles of the surfaces vary from 0° to approximately 120°.

Float glass surface was taken as a smooth substrate. The smooth homogeneous thin films (< 10 nm) were deposited on clean glass. In total we prepared 1500 samples: 150 samples of each of 10 coating types. The 1500 samples were sufficient to conduct all physical measurements and to present a fresh sample for each comparison in the perception studies without the need to develop cleaning procedures. The samples used in the experiments are presented in Table 1. There are three classes of films: #1–4 inorganic, #6–8 hybrid (organic/inorganic), #9–10 organic.

Finger friction or tactile friction was measured with a ForceBoard™ (Industrial Dynamics AB, Sweden), which is equipped with two load cells, one horizontal and one tangential, consisting of strain gauges in a Wheatstone bridge configuration. A mechanical load is converted into voltage signals that are amplified and proportional to the applied load. The tangential force, i.e., friction force, and vertical force, i.e., applied load, were continuously recorded using DAQFactory software at a rate of 100 Hz as a finger was moved over the glass surface. The output data were a text file of columns with the time and respective forces and the analysis was performed using MATLAB.

The glass surfaces were mounted on the top plate with double-sided adhesive tape, and one female experimenter probed the surface with her dominant index finger with a finger sliding angle of approximately 30°. The sliding direction was along the finger axis for all tests. The average stroking distance was about 80 mm and the average speed of all measurements was approximately 20 mm/s. Moreover, the average applied load for all measurements was 1.2 ± 0.15 N. Friction coefficients were calculated as the ratio of friction force and applied load after removing the turning points associated with changing direction of the finger. The friction force in the “backwards” direction (pulling the finger toward the body) is recorded as a negative number: we present all friction coefficients as the absolute values of the measurements in this paper. One measurement consisted of 15 stroking cycles; however, our observation is that in general perceptual discrimination experiments would be better represented by three strokes since most participants are able to make decisions using less exposure. Thus, the friction data shown are for the first three strokes. In the supporting information, the data for 15 strokes are also shown.

In total, six friction measurements were conducted on each glass sample. Measurements were performed in a climate-controlled room with a temperature of 22 °C and relative humidity of 45%. One glass surface was used for three friction measurements, where each friction measurement was obtained from a fresh track on the surface. The ten samples were measured in six series in which the order of the ten samples within each series was randomized. The skin hydration was monitored before each series of measurements with a Corneometer CM 825 (Courage + Khazaka GmbH, Germany) and was in the range of 65–70 arbitrary units which is considered a “moist finger.”

The participants in the tactile perception study were ten women in the ages of 20–34, (24.8 average age), all right handed. Women were preferentially selected in order to avoid possible convolution of gender differences with the perceptual differences between the surfaces. All procedures were in accordance with the ethical standards of the institutional and the national research committee and with the 1964 Helsinki declaration and its later amendments of ethical standards regarding studies with human participants. Written consent was given, and the participants were informed that they could quit the experiment at any time if they so wished. The only risk for physical injury detected in the risk assessment was the sharp edges of the samples. This was addressed by constructing two sample holders that covered the edges and only exposed the safe area of the samples, see Fig. 1. To run a successful data collection, each sample holder needed to be safe, fast and easy to load, and be able to hold two samples for the paired comparisons. The experimenter handling the samples and loaded/unloaded the sample holders wore protective gloves. The first experimenter loaded the sample holders according to the randomized orders (a unique order for each participant) and the second experimenter presented the samples to the participants and recorded the data.

The participants were instructed to scale the perceived haptic similarity between all possible combinations of pairs of stimuli in one-half of a 10 × 10 matrix, i.e., 1–2, 1–3, … ,1–10; 2–3, 2–4 … 2–10; … 9–10. Before the experiment started, the participants washed and dried their hand according to earlier protocols using an unformulated surfactant. After receiving their instructions, they were blindfolded and given a practice session during which they could practice the scaling method and get a feel for the range of materials. The scale for the similarities ranged from 0 to 100%, where 100% meant that the samples were identical and 0% as different as possible within the given range. Halfway through the experiment the participants had a short break, during which they washed and dried their hands again. After the completion of the experiment, the participants were asked to rate the pleasantness of each coating type on a scale from 1 to 10, where 10 was most pleasant and 1 least pleasant.

The data were analyzed with multidimensional scaling (ALSCAL) using the IBM statistics software SPSS. The scaling of similarity is a relatively unbiased measurement. The scaling was interval data and allowed for individual differences weights to compensate for differences in dimensional preference from the participants, see [10] for further theory and application of multidimensional scaling.

The pleasantness rating is not formally a robust scale and should therefore not be taken too literally. Its purpose was to anchor the dimensionality, i.e., in which direction does a general pleasantness lie, not as an absolute scale. However, the similarity scale can then be consulted for interpreting the relative pleasantness between the samples.

From the considerable spread in similarity scores across surface pairs (range 19–79), it was clear that our participants were able to detect perceptual differences between the surfaces, see Table 2.

The multidimensional scaling analysis treats the perception similarities as distances and identifies and reveals how many important factors the participants relied on to differentiate between the surfaces. For this data set, the analysis suggested a 2D solution. The stress (0.39) and RSQ (0.16) values for the 2D solution suggest that although participants were able to successfully perform the task, it was relatively difficult. From the derived subject weights in the scaling analysis, it was also noted that there was variability between participants as to which dimension was primarily responsible for discrimination. The resultant tactile map based on the dissimilarity scores (100—similarity score) is shown in Fig. 1. From the distribution of the samples within the map, it is apparent that the participants perceived the smooth glass surfaces as different. Although this analysis revealed that the participants relied on two perceptual dimensions to discriminate the stimuli, it cannot tell us what these dimensions are. To do this, we attempted to relate the perceptual dimensions to instrumental data.

The average, dynamic finger friction coefficients and standard error of the first three stroking cycles based on six measurements per sample are presented in decreasing order in Fig. 2a. The water contact angle is plotted in Fig. 2b. The uncoated glass has a low contact angle and the fluorocarbon surfaces, as expected, have highest contact angle. Contact angle hysteresis data are plotted in supporting info and is positively related to contact angle. Typically, there is no clear static friction peak in such measurements, so the friction coefficient is obtained from the sliding, or dynamic region of the force traces. Derler et al. [4] have earlier investigated the finger tribology on glass of various roughness and identified adhesive friction as the dominant underlying mechanism. For smooth glass the friction coefficient was around 2.2. Overall, the glass surfaces in this study are also regarded as high friction surfaces with friction coefficients from 2 and above. In comparison, tactile friction coefficients for printing paper lie in the range of 0.2–0.5 [11]. As can be seen, the fluoropolymer-coated glass surfaces (#9 and #10) display the lowest friction coefficient and two of the SiOx–CHy (#6 and #7) display highest friction together with two of the SiOx and SixNy (#2 and #3).

Smooth and high-friction surfaces are often associated with a high degree of stick–slip. A detailed look at the load and friction for all measurements show that stick–slip is present on these glass coatings, especially strongly for coatings 2, 3, 4, 6, 7, and 8. A typical plot of the two forces can be seen in Fig. S1, where obvious stick–slip is “visually” observed during the first few strokes. The disappearance of the stick–slip over time may have several explanations. It could be the result of transfer of lipid material from the fingers that lubricates the surface or, if sufficient moisture is transferred from the finger, this moisture may also act as a lubricant [4, 12]. Although the finger pad does not have sebaceous glands, there is nonetheless lipid material transfer from the finger to the surface during touch. Similarly, increased moisture in general increases friction due to the softening of the stratum corneum [13] and occlusion [5]. However, because glass is non-absorbent, the accumulation of moisture in the contact also provides a possibility of reducing friction via a lubrication mechanism.

Previous experience [1, 12] from tactile friction measurements has shown a friction dependence on the surface roughness where there is an optimum surface structure range or roughness for low friction. This is explained by the real contact area at the finger-surface interface. A surface structure may decrease the real contact area, i.e., smaller number of contact points, compared to a smooth surface, resulting in a lower friction. The glass surfaces within this project are all very smooth, enabling a large contact area between a finger and the surface, resulting in high friction and little expected frictional variation due to roughness differences.

The root mean square (RMS) roughness and average height for the coatings were obtained from AFM images and are listed in Table 3. As can be seen in Fig. S2, the tactile friction coefficient of the coatings does not seem to depend on the RMS roughness (nor average height). It is observed that the SiOx–CHy coating (#8) displays higher roughness compared to #6 and #7 which may explain why #8 shows lower friction. However, film #8 contains the lowest amount of carbon of the samples in the SiOx–CHy film family, and within this family a clear trend suggesting that higher C-content is associated with lower friction is evident. AFM images of these three SiOx–CHy surfaces are shown in Fig. S3. These samples appear to act as outliers in the tribology behavior and this is discussed further below.

The evaluation of the friction response to differences in surface chemistry is addressed by plotting the friction coefficient versus the water contact angle in Fig. 3. Figure 4 shows that the surface energy and the contact angle are inversely related to one another and can therefore regarded as degenerate parameters. As can be seen, the friction is not a systematic function of solely water contact angle (or surface free energy). However, within each sample family, differing behaviors emerged. The two most hydrophobic surfaces (high water contact angle and low surface free energy), the fluoropolymer coatings, produced the lowest friction coefficients. Within the SiOx–CHy family (#6–8), the friction coefficient increases with increasing water contact angle and decreases with the surface free energy.

Other parameters that may affect the friction response that have not been addressed are the surface charge, and how the moisture in the sliding contact is affected by the surface chemistry of the coating. The differences in friction response between the coatings may be an effect of this, as water contact angle alone was not enough to describe them. The surface free energy and the water contact angle is very similar for the uncoated and the coated glasses #2–5 as can be seen in Figs. 2b and S4.

The tactile friction measurements in this study are performed in reciprocating strokes, where a forward stroke (away from the body) is recorded as a positive tangential force and a backward stroke as a negative tangential force (towards the body). Analysis of variations in the two stroking directions shows that it is mainly the friction in the backward stroke that varies between the coatings, and that this value tends to be larger (see Fig. 5).

This friction hysteresis has not previously been observed on any of the other materials studied using our apparatus, which include paper [12], artificial skin [14], tissue and polymer. It was observed that on average the applied load in the backward stroke (all measurements) was 1.0 N while for the forward stroke this value was significantly larger at 1.5 N. When the finger encountered higher friction, a lower load was applied so that the finger could slide on the surface, most probably an unconscious action which has been addressed previously [12, 15] and results from the fact that the tendency is to probe with a constant friction force, rather than constant load. A similar hysteresis was observed by Derler [4] and attributed to mechanical asymmetry. There may also be an additional mechanical contribution to this effect, which is specifically associated with stick slip. Zhang et al. [16] have recently addressed this subject, and conveniently they have done so using glass as a substrate. In this elegant study, they demonstrated that there is a strong hysteretic effect in friction coefficient with sliding direction with a mechanical origin. In fact, the hysteresis depends strongly on the angle of the finger to the surface, and reverses as the angle becomes steeper. At lower sliding angles, analogous to those used here, the friction is higher as the finger is slid towards the body. The authors also measure the spacing of the fingerprint ridges within the nominal area of contact and show that this changes with the sliding direction and thus ascribed the hysteresis to this biomechanical variation. In that study, it was also shown that more reproducible friction data were obtained for an immobilised finger, though this configuration is less representative of the fingers mechanical behavior during actual tactile probing. They also observed changes in load with sliding direction as seen here, so the picture is not fully clear. Because the finger pad is deformable, the friction-load relationship is pseudolinear only over moderate ranges due to the non-linear relationship between nominal contact area and load. Thus, there are likely to be additional hysteretic effects resulting from the variation in load with direction.

Close examination of the extended (15 stroke) friction data indicates that for surface #6 (SiOx–CHy), the first three measurements (all performed on the same piece of glass in different tracks) show an average friction coefficient of 3.2, whereas the last three measurements performed on another piece of glass returned a friction coefficient of 4.9. No studies of the coating robustness were undertaken. The water contact angle hysteresis (Fig. S7) may indicate that there are differences in topography in the coating, or that the surface displays patches of hydrophilic and hydrophobic areas. Coating #6 displays the highest contact angle hysteresis and this may be the reason for the observed variations in friction coefficient—i.e., that friction hysteresis may have a similar advancing/receding mechanistic dependence as that of the contact angle.

To attempt to interpret the dominant perception dimension, the coordinates from dimension one were plotted against the instrumental measurements. In Fig. 6, the dimension 1 scale values are plotted against the WCA. Although there is not a perfect systematic relationship between the perception and WCA, the relationship is statistically significant, see Table S1. Thus, the analysis suggests that the water contact angle describes dimension 1 well. This implies that the surface chemistry was the main cue used for discrimination in this task. Despite this significant correlation, inspection of Fig. 6 suggests that samples #1 through to #5 do not seem to be dependent on the WCA to the same extent as the other samples. We also correlated the coordinates in dimension 1 with the friction coefficient (shown in Fig. 7). Whilst not statistically significant, Fig. 7 suggests that friction contributes to how the participants discriminated the surfaces.

We investigated the possible relative contributions of WCA and friction further by splitting the surfaces into high (#1–5) and low (#6–10) hydrophilicity groups. This roughly mirrors the grouping of the coatings into organic and non-organic. The same correlation analysis was then performed on the two sub-groups. The results are shown in Tables S2 and S3. Interestingly, this analysis implies that for surfaces #6–10, WCA remains the main explanatory parameter for dimension 1, while for surfaces #1–5, the important factor seems to be the friction coefficient of the surfaces. We interpret this as meaning that for highly hydrophobic surfaces, WCA is the primary cue used for distinguishing surfaces. In contrast, for surfaces with low hydrophobicity, WCA is still a factor however the friction coefficient is more important. This is consistent with the conclusion that surface chemistry is an important cue for tactile perception of surfaces.

While the friction coefficient seems to be important for discrimination in at least some of the samples and explains some of the variance in dimension 1, sample #6 appears to be something of an exception. As discussed above, there appear to be questions concerning its robustness, and it displays both high contact angle hysteresis and frictional hysteresis. Bearing these reservations in mind a possible explanation for its placement in Fig. 7 is that despite the high friction coefficient, the high WCA was a more dominant feature for the perception of the surface. It thus appears that a finger friction model is suitable for characterizing the perception of glass coatings lower than 100° WCA, and that for WCAs over 100° it becomes less meaningful. This is consistent with the analysis for high and low hydrophilic surfaces conducted above. However, there was only one high friction, high WCA sample in this stimuli range so this explanation should be regarded with caution until further research addresses this issue.

The relationship is between dimension 1 and friction is thus not as clear cut as in earlier studies where there was an obvious relationship between friction coefficient and the sticky-slippery dimension for micro textures [1, 18]. Surface chemistry appears to dominate at high contact angles while friction coefficient becomes more important at low, or very similar, contact angles. Dimension 1 is thus NOT unequivocally a “sticky–slippery” perceptual dimension.

No systematic relationship between dimension 2 and any physical parameters was found, though the samples that displayed most stick slip (Fig. S1) behavior are well spread in this dimension. The sampling rate of the device is too small to allow the frequencies to be reliably extracted but these frequencies appear to be high enough to be detected by Pacinian mechanoreceptors under the skin [17, 18]. This type of dynamic, “pre-constant slip” has been implicated in perceptual discrimination earlier [7, 19]. It thus seems likely that dimension 2 is related to such an effect, though we have not identified a physical parameter that we can extract from the data to correlate with this dimension. Neither can we rule out that effects involving water occlusion in the sliding contact are responsible for this dimension since the surfaces are both smooth and non-porous/adsorbent [5]. Gueorguiev et al. have studied glass and polymer samples to study the effect of varying surface chemistry on perception. In this case, both the topography and the elastic modulus are varied simultaneously with the surface chemistry, and both these parameters can be used as perceptual discriminators [1, 20]. Nonetheless, the authors were able to successfully couple related perceptual discrimination to differences in an arbitrary spatial frequency parameter extracted from the tangential force. They thus show very clearly that dynamic effects are implicated, particularly in the so-called partial slip region. This lends further weight to the argument that dynamic events related to slip are responsible for dimension 2.

To our knowledge, there is no study where the surface chemistry has been varied as systematically as done here. While it is relatively easy to obtain hydrophobic monolayers by self-assembly approaches which can be compared to hydrophilic untreated glass on a friction device, it is untenable to use that approach to produce a large number of surface chemistries or to produce the hundreds of samples for each surface chemistry which are necessary for pairwise comparisons using fresh surfaces for every comparison. It should also be noted that while the wetting properties are varied, different chemical approaches are used, which may well have different binding affinities to the glass, or local variations over the large areas of float glass employed. There is no evidence for this, but it cannot be completely ruled out. The use of a single operator in performing friction measurements is not completely without risk. As has earlier been discussed, the variation between individuals can be much larger than the variation between substrates, leading to difficulties in analyzing group data, though the trends are preserved at the individual level [15]. Thus, it should be established that the operator has representative tribological properties, rather than being an outlier, in terms of absolute friction coefficients. Neither is it practical to record friction data on the individual participants—their sessions take several hours for the psychophysical experiments alone.

Complementary to the similarity data, the ten surfaces were ranked in order from least pleasant (1) to most pleasant (10) by each participant. Note that this is a ranking and not a score, so the abscissa labels of Fig. 8 should not be seen as values, or as representing absolute differences in pleasantness. To assess if assessments of pleasantness were associated with any of the physical parameters, we ran correlations between the mean pleasantness ranking and selected instrumental measurements (Table S4). This analysis revealed a significant correlation between pleasantness and the friction coefficient (COF 3). From Fig. 8, it can be seen that participants ranked #10 (high contact angle, low friction coefficient) as the most pleasant and #2, #3, & #7 (high friction coefficients) as the least pleasant. Again, #6 was a something of an exception. Similar findings relating pleasantness and friction coefficient are discussed by Adams et al. [5]. Interestingly, pleasantness was not significantly correlated with WCA or surface energy. This suggests that although WCA was a primary contributor to participants’ ability to discriminate between the surfaces, it alone did not affect their pleasantness ratings. Instead, pleasantness is strongly related to the friction coefficient. Together, these findings tend to cement the importance of tribology in affective touch, and suggest that psychotribology is likely to emerge as a distinct field.