André
Knapp
*a,
Lisa Julia
Nebel
b,
Mirko
Nitschke
c,
Oliver
Sander
b and
Andreas
Fery
*ad
aInstitute of Physical Chemistry and Polymer Physics, Leibniz Institute of Polymer Research Dresden e. V., Hohe Str. 6, 01069 Dresden, Germany. E-mail: fery@ipfdd.de; Fax: +49 351 4658 281; Tel: +49 351 4658 225
bInstitute for Numerical Mathematics, Technical University Dresden, Zellescher Weg 12-14, 01069 Dresden, Germany. E-mail: oliver.sander@tu-dresden.de; Tel: +49 351 463 35049
cMax Bergmann Center of Biomaterials, Leibniz Institute of Polymer Research Dresden e. V., Hohe Str. 6, 01069 Dresden, Germany
dChair for Physical Chemistry of Polymeric Materials, Technical University Dresden, Mommsenstr. 4, 01062 Dresden, Germany
First published on 10th May 2021
We demonstrate a novel approach for controlling the line defect formation in microscopic wrinkling structures by patterned plasma treatment of elastomeric surfaces. Wrinkles were formed on polydimethylsiloxane (PDMS) surfaces exposed to low-pressure plasma under uniaxial stretching and subsequent relaxation. The wrinkling wavelength λ can be regulated via the treatment time and choice of plasma process gases (H2, N2). Sequential masking allows for changing these parameters on micron-scale dimensions. Thus, abrupt changes of the wrinkling wavelength become feasible and result in line defects located at the boundary zone between areas of different wavelengths. Wavelengths, morphology, and mechanical properties of the respective areas are investigated by Atomic Force Microscopy and agree quantitatively with predictions of analytical models for wrinkle formation. Notably, the approach allows for the first time the realization of a dramatic wavelength change up to a factor of 7 to control the location of the branching zone. This allows structures with a fixed but also with a strictly alternating branching behavior. The morphology inside the branching zone is compared with finite element methods and shows semi-quantitative agreement. Thus our finding opens new perspectives for “programming” hierarchical wrinkling patterns with potential applications in optics, tribology, and biomimetic structuring of surfaces.
However, so far, the complexity of artificial wrinkling structures is far from examples found in biological systems, where distinctive hierarchical structures can be found.3,5,13,14 In many biological systems, areas of different wavelengths can be encountered instead of a uniform wavelength across the whole surface. Branching of wrinkles is thus a commonly found structural effect. In contrast, wrinkling wavelengths in artificial wrinkling structures are usually constant, and branching often occurs in the form of random line defects at microscopic heterogeneities or due to the wrinkling process getting locked in metastable states, which cannot be controlled.15–17
Controlled branching can be facilitated if a lateral change in properties determining the wrinkle wavelength is achieved. Such properties can be film thickness, film stiffness, or substrate stiffness. Vandeparre et al. showed the evolution of wrinkles and found line defects in wrinkling patterns.18 They used a time-dependent decrease of substrate modulus due to a solvent diffusion into titanium sputtered polystyrol sample and thus generated a lateral gradient in the substrate modulus. This resulted in a continuous change of wrinkle amplitude A and wavelength λ. With this diffusion-controlled wrinkling, they fabricated patterns with defined gradual changes in topography, especially regarding λ. These changes show forced line defects as cascade-like Y-branches, which necessarily form as a physical consequence of a growing λ. Further examples of cascade-like patterns consisting of a hierarchy of successive generations of Y-branches were observed in “one-side” compressed free-standing membranes e.g. curtains or graphene19 or in compressed floating thin polymer sheets, where the fluid meniscus follows the contour of the edge of the sheet and reduce the amplitude and wavelength of the wrinkles at the free edge.20
Branching as a structuring tool, however, needs a more controlled process with static gradients in the wrinkling-related properties independent of the diffusion time.
Such static gradients can be created by changing the thickness of polymeric or metallic films on top of an elastomeric substrate,21–23 by changing the substrate stiffness due to asymmetric curing or continuous component shift of a PDMS substrate,24,25 and by changing both parameters via modification of PDMS surface due to plasma/UVO treatment.26–28 Simulations were carried out by Yin et al.29 for these kinds of gradients. They found Y-branched or multi-branched structures in the case of uniaxial compression perpendicular to a gradient direction, which represent an example of cascade-like branching with randomly located Y-branches.
To use line defects as a structuring tool in a well-defined branching process, such defects need to be controlled and precisely localized. This can be realized by a defined change of the wrinkling parameters perpendicular to the wrinkle direction, which is defined as parallel to the lines of maximum wrinkle amplitude. To quantify the shift in wavelength from one area to the other, we introduce the branching degree as the ratio of wavelengths. Previous investigations used stiffness changes in the substrate10 or different treatment conditions combined with covered sample areas to achieve changes in wavelength.25,30–32 These previous approaches are, however, limited in exact localization and controllability of the branching events. Further, the branching degree was limited to a maximum of 1.5. Also, the quality of wrinkled surfaces in the sub-μm range limited a well-defined branching control, primarily through high λ-deviation and randomly distributed line defects.
The local variation of wavelength, localization of defects on the micron-scale, and pronounced changes of wavelength by more than a factor of 2 remain challenging. The present study aims to introduce patterned plasma treatments of PDMS to overcome these limitations and shows a comparison between experimental results on wavelength, morphology and mechanical properties, and analytical/numerical simulation results. This paves the way for predicting branching geometries for this approach. Thus complexity that can be reached in artificial wrinkling structures approaches further to biological examples.
To get an overview of the different substrate elasticities, the component ratios of Sylgard 184 were changed from 2:
1 to 20
:
1 (base polymer to hardener component), which allowed the variation of the stiffness in the range between 0.5 and 8 MPa.33–36 The plasma treatments were performed with N2 and H2 for 2 and 4 min with all substrate variations. AFM images of all tested variants are summarized in the ESI† (Fig. S2 and S3). The characterization was carried out regarding the homogeneity of the wrinkled surface, e.g., relating to uncontrolled branching or cracks in the surface. The best homogeneity was found for the component ratio of 2
:
1 (CR2) with post-curing for 4 h at 80 °C, which was then used in further experiments.
Gradient samples were used to determine the possible λ range. They were produced with a retractable shield inside the vacuum chamber, which allowed a plasma exposure with a graded treatment time. Single gradient samples with treatment times of 0 to 480 s were produced. Nitrogen and hydrogen were used as process gases. Using N2 generates lower absolute wavelengths compared to using H2, which provides the highest possible absolute wavelengths with the plasma treatment device. Wavelength and the amplitude increased nearly linearly with the given treatment time for both plasma gases. The amplitude was measured as the difference between the minima and maxima of the wrinkles. Fig. 1 summarizes the obtained wavelength and amplitudes for each process gas measured on gradient samples.
![]() | ||
Fig. 1 Treatment time dependence of wavelength (black) and amplitude (green) for plasma process gases N2 (square) and H2 (triangle). |
N2 provided a λ range of 444 to 3285 nm. The amplitude ranged from 47 to 865 nm. The λ range is substantially higher for H2 treatment, and it ranges from 929 to 7364 nm, as well the similar trend was observed regarding the amplitude. The corresponding amplitude was significantly higher than the amplitude obtained with N2 and ranged from 250 to 2100 nm.
When a polymer is exposed to a low-pressure plasma a variety of high energy species including vacuum ultraviolet photons but also accelerated ions and neutral molecules contribute to the overall effect. The respective energy distributions, fluxes, and interaction mechanisms determine type, degree, and depth profile of the actual surface modification. For that reason different low pressure plasmas operated with the same excitation but with different process gases are expected to have different interaction profiles with a given organic material. In case of H2 and N2 plasma exposure of PDMS this leads to rather different properties of the stiff surface layer. In this study we take advantage of this effect to cover a wavelength range as wide as possible with one and the same low-pressure plasma device.
The wavelength calculations with the analytical eqn (1) and (2)37 were used to verify the experimental results and provide a benchmark for further finite element simulations.
![]() | (1) |
![]() | (2) |
Time [s] | h f [nm] | E f [MPa] | λ calc [nm] | λ topo [nm] |
---|---|---|---|---|
30 | 128 ± 23 | 11.3 | 978 ± 218 | 624 ± 69 |
60 | 143 ± 22 | 12.1 | 1115 ± 199 | 831 ± 76 |
120 | 133 ± 5 | 20.4 | 1217 ± 67 | 1289 ± 96 |
240 | 255 ± 33 | 13.9 | 2043 ± 376 | 1983 ± 56 |
The well-known analytical eqn (1) and (2) are valid for a bi-layer. The consequence is, that the exponential stiffness decrease into the substrate material had to be averaged to get such a bi-layer like model, represented by the film thickness hf and a step like constant mean film stiffness Ef. This approximation was done by mean calculation of the stiffness in the range of the measured thickness. This leads to the effect that the film stiffness Ef passes through an ostensible maximum, what is only a consequence of the average process and does not represent an absolute stiffness maximum. The thickness increases continuously with treatment time within the experimental error bars.
The wavelength of the wrinkle pattern for different plasma treatments was compared by two independent approaches: (a) directly measured with topographical AFM images (λtopo) and (b) indirectly calculated from QNM cross-section analysis via Ef and hf (λcalc). The calculation using eqn (1) and (2) approximates the complex depth profile of Young's modulus (see Fig. S5 and S6, ESI†) with the simple bi-layer model. This requires an averaging procedure to determine Ef which is less or more appropriate depending on the actual shape of the modulus profile. In the particular case of the presented data the procedure provides more reliable results for longer treatment times. The data on effective modulus and thickness of the modified layers show that both parameters do not exhibit a simple proportionality to the treatment time, which is not surprising due to the complex processes involved. To the best of our knowledge, there is no physical model which explains the linear trend of wavelength with treatment time. This finding will be subject to further studies which go beyond the current report. But nevertheless, the comparison showed reasonable accordance between the directly measured and indirectly calculated wavelengths, but also the limitation of the analytical equations to approximate the experimentally observed dependency of the wavelength shown in Fig. 1, especially for low treatment times.
Masks with a thickness as low as possible are required for this purpose. The reason for this is a penumbra effect at the mask edge that occurs for any directed component of the complex plasma impact due to the given exposure geometry (source diameter 160 mm).
To quantify the mechanical changes induced by the plasma treatment as well as to quantify the sharpness of the transition zone, unstrained masked samples were treated for 4 min with N2 plasma. To evaluate the transition zone, the surface stiffness in the transition area was measured with QNM. Fig. 3 shows a measured stiffness map of the transition zone between the treated and untreated areas. The determination provided a stiffness transition zone of (630 ± 188) nm for N2 plasma. The measured transition zone is smaller than the minimum wavelength for both process gases, which is about 800 nm for N2 plasma treatment.
With this masking technique, location-independent combinations of plasma treatment parameters like time variation and process gas are possible. This allows the combination of process conditions leading to different wavelengths and enabling the possibility of high wavelength-differences, high branching degrees, and consequently a controlled branching process of surface wrinkles. Fig. 4 shows an example of controlled localized branching. The photograph in Fig. 4a illustrates the typical angle-dependent reflection colors of wrinkled surfaces caused by interference. Differences in structural colors correspond to regions of different wavelengths. The homogeneous color of the strips and the surrounding area gives a first impression of the homogeneity of the produced wrinkles. The microscopy image in Fig. 4b with a magnification factor of 50× allows resolving the individual wrinkles and the transition zone. The wrinkles of the bigger wavelength split into wrinkles with smaller λ exactly along the line representing the mask edge. The different λ areas are homogeneous over a large area (90 μm × 76 μm). Fig. 4c shows an AFM image, which allows the quantification of wavelength, amplitude, and branching. The branching degree as a consequence of the wavelength change and the amplitude transition from one area to a neighboring one can be investigated. It is visible that the branching zone is not affected by the transition zone we described before. The branching events were localized along a line with dislocations smaller than 1 μm.
In practice, the reached branching degree for single plasma gases shows that the useable λ range enables BDs of up to 4, but not higher. The limiting factors for N2 are the randomly located line defects with low λ and of course the maximal reachable λ. The used two-step masking process allows the combination of different plasma gases to combine the specific reachable λ ranges of the single plasma gases. To produce BDs higher than 4 a combination of H2 and N2 was used. First, the substrate was treated with H2 to reach a λ above the maximum λ for N2 plasma, followed by a second treatment-step with N2 plasma to get a λ below the minimum λ for H2. This process used different process gases for the masked and unmasked treatment steps and led to an area treated twice (H2 + N2) with a resulted λ of 5650 nm, whereas the area treated with only one process gas (N2) led to a λ of 850 nm. In this way, the BD could be increased to up to 7. The described sharp localized stiffness steps in combination with transition zones in the range of the used wavelengths, suppressed a cascade-like branching like by the treatment with single-process gases. The branching started immediately and without clear visible cascade branching (see ESI,† Fig. S7).
Fig. 5 gives an overview of branching scenarios depending on the process gases or the combinations thereof.
![]() | ||
Fig. 5 Branching scenarios sorted by branching degree for the process gases N2, H2, and their combinations, produced with the masking technique and measured with AFM. |
High-quality N2 branching without randomly localized line defects was successful in the wavelength range from 1000–3000 nm and enabled, BDs up to 3. H2 branching was possible in the λ range from 2000–7000 nm and enabled BDs up to 3.5. In principle, every process gas enables BDs of around 1 up to 3–4 in their specific absolute λ range. Higher BDs up to 7 and more could be obtained with a combination of the process gases.
We model the substrate as a rectangular block of dimensions 60 μm in x-direction, 20 μm in y-direction and 20 μm in z-direction. We stretch it uniaxially in x-direction by 30% using displacement boundary conditions on one side while fixing the rectangular block on the opposite side using Dirichlet boundary conditions. We use a Mooney–Rivlin material33 with parameters derived from uniaxial tensile tests (Table S1, ESI†). On the deformed upper surface, we then attach a stiffer material layer modeled by a geometrically exact Cosserat shell as described in ref. 39. The material parameters are summarized in Table S2 (ESI†). While the shell model is two-dimensional, the thickness h appears as a material parameter. This thickness parameter and the stiffness parameters μstiff and λstiff, which determine the elastic behavior of the Cosserat shell, take position-dependent values. This allows us to simulate the masking process described above. These changes in parameters trigger the wrinkle branching behavior. We release the uniaxial component of the displacement boundary conditions, thus releasing the 30% stretch, and observe how wrinkling patterns form due to the stress mismatch of substrate and stiff layer.
The calculations were done using the DUNE libraries for C++ for solving partial differential equations (PDEs) with grid-based methods.40 We discretize the model using second-order (27-nodes) Lagrange finite elements for the substrate, and second-order (9-nodes) geodesic finite elements for the shell.41 As shown experimentally in ref. 42, no locking occurs for this type of shell discretization. We start with a cubical grid for the substrate which is then manually graded resulting in a hierarchical grid with a high resolution in the vicinity of the shell. In total, the substrate grid has 15642 vertices. The attached shell model is discretized on the two-dimensional restriction of the substrate grid to the upper boundary in y-direction, which has 3993 vertices. This setup results in a nonlinear minimization problem which we solve using a trust-region method, an iterative method considering a linearized version of the problem in each step. As an experimental reference sample and base for the material characterization, we used a branching case that was realized by treatment with N2 plasma for 60 s and 240 s. To trigger branching of the wrinkling pattern, we divided the shell horizontally into two halves in strain direction. We assigned thickness parameters h1 = 238.5 μm in the upper half and h2 = 142.5 μm in the lower half, and elastic moduli of E1 = 13.87 MPa in the lower half and E2 = 12.14 MPa in the upper half. These parameters were measured with QNM cross-section measurement (Fig. S5, ESI†) for known wrinkle patterns and their corresponding thickness and stiffness values (Fig. S6, ESI†). Simulation results and the corresponding experiments are shown in Fig. 6.
The wavelengths resulting from the numerical simulations are (994 ± 136) nm for the upper half and (1932 ± 184) nm for the lower half calculated with the same Python script as we used for the experiments. The corresponding experimental wavelengths are (831 ± 77) nm and (1983 ± 56) nm, respectively. These results are closer to the experiments than the results of the analytical equations (Table 1), especially for low wavelengths. The ratio of the number of wrinkles resulting from the simulation is 1.94, which is close to the wrinkle ratio of 2.37 of the corresponding experimental results. The resulting amplitudes are (120 ± 25) nm for the upper half and (354 ± 46) nm for the lower half, compared with experimental results of (101 ± 10) nm and (416 ± 40) nm, respectively.
The localization of line defects for the branching occurs in the transition zone between the adjacent wrinkling areas. The correspondent adjustment of the amplitude, induced by the change of wrinkle shape in the branching zone, depends on the branching degree. This behavior can be observed both in the experimental and in the simulation results (Fig. 7). The wrinkle shape transition starts with a high wavelength and amplitude; then branching occurs when large wrinkles split into smaller ones, i.e., each wrinkle maximum splits into local maxima with a new small minimum developing between them. The new small minimum increases while the initial wrinkle minimum decreases. This transition process ends as soon as the small wavelength and the corresponding amplitude are fully reached.
![]() | ||
Fig. 7 Comparison of experimental branched wrinkle (right) shape with simulation outcome (left) (arrows represent the new minima, where the wrinkle splits into smaller ones). |
The experimental wrinkle shapes differ from the simulated ones for several reasons. The used grid size in the simulation resulted in a more triangular wrinkle shape compared to the more sinusoidal shape in the experiment. Further, the symmetry break in the wrinkle shape was more pronounced in the experiment than in the simulation results. This asymmetric wrinkle shape depends on the ratio of film and substrate stiffness as well as the used substrate pre-stretch.43 In the experiment we can see V-shaped minima in contrast to more sinusoidal maxima.
We were able to semi-quantitatively reproduce the wrinkling behavior in a finite element simulation using a geometrically exact Cosserat shell on a hyperelastic Mooney–Rivlin-substrate, which shows perspectives for the future rational design of intricate wrinkling patterns.
Hierarchical wrinkling offers a range of perspectives. First, wrinkles are well known to influence optical, wetting, and tribological properties as well as the interaction with biological systems.44–47 Surfaces featuring different wrinkle wavelength and amplitude allow combinatorial approaches for investigating these effects. As well, the combination of different wrinkled surfaces (e.g., the effect of wrinkles on friction, capillary forces, and adhesion48,49) can thus be efficiently screened. Second, the branching points themselves are of interest as symmetry-breaking elements, which could, for example, serve as optical vertices or allow for selective deposition of nanoparticles at branching points, to mention just a few perspectives.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0sm02231d |
This journal is © The Royal Society of Chemistry 2021 |