1. Introduction

The recently developed two-photon polymerization (2PP) technology is attracting increasing interest as a viable ‘top-down’ approach for the fabrication of high resolution three-dimensional microstructures, useful in many application areas such as photonics and biomedicine [1–7]. To date, the structuring ability of many materials has been investigated, and lateral resolution below 100 nm has been achieved [8–10]. However, one reason this technology is not yet widespread is that the majority of the materials used suffer from shrinkage during the structuring and development procedures, which results in the deformation, and therefore limits the reproducibility of the fabricated structures.

Although shrinkage is one of the fundamental aspects of photopolymerization and its understanding is important for both the design and the process reproducibility of devices [11], in case of 2PP, volume reduction still has not been fully investigated. Unlike conventional methods, 2PP enables the study of the shrinking processes locally, at the microscale. In this work, we suggest a methodology for the measurement of material shrinkage during 2PP. This method can be applied even to materials exhibiting very low shrinkage. For the zirconium-silicon composites investigated in this paper, the prediction of volume reduction is particularly important when the structures are fabricated near the polymerization threshold. The controllable shrinkage has been exploited to improve the structural resolution further, without compromising the functionality of the fabricated structures.

2. Experimental methods and applied materials

For the investigation of the structural shrinkage dependence on the average laser power, a novel microstructuring approach was applied. It relied on the fabrication of structures on support columns and the direct measurement of the structure deformation. The method was tested by fabricating photonic crystal woodpile structures. These consist of layers of one-dimensional rods with a stacking sequence that repeats itself every four layers. The distance between four adjacent layers is ‘a’ and within each layer, the axes of the rods are parallel to each other with a distance ‘b’ between them. The adjacent layers are rotated by 90∘. Between every other layer, the rods are shifted relative to each other by ‘b/2’. For the case of ‘a/b = √2’, the lattice can be derived from a face-centred-cubic (fcc) unit cell with a basis of two rods. The support columns were attached to the glass substrate at one end, and to a woodpile structure at the other end.

The material employed was a 20:80 zirconium-silicon sol-gel, containing 4,4̀-bis(diethylaminobenzophenone) as a photoinitiator [12]. Samples were prepared by drop casting the solution onto conventional glass coverslips, and prebaking them at 100°C for one hour, to form a hard gel. The woodpile structures were fabricated at the scanning speed of 200 μm/s, the in-layer rod distance is set to a constant value of 1 μm, and the distance between the neighbouring layers is set to 390 nm. In this experiment, a 100× microscope objective (Zeiss, Plan Apochromat, N.A. = 1.4) was used to focus radiation of a Ti: Sapphire femtosecond laser (120 fs, 90 MHz, 780 nm) into the volume of the sol-gel material. Detailed description of the experimental setup can be found elsewhere [13]. To remove the unpolymerized material, the structures were developed in 1-propanol. 1-propanol was subsequently substituted by ethanol prior to insertion into a critical point dryer chamber (Quorum Technologies, UK).

3. Results and discussions

A set of woodpile structures fabricated at different laser powers is shown in Fig. 1. Woodpiles obtained at average laser powers well-above the polymerization threshold do not exhibit any shrinkage. These structures retain their “original” size, and the vertical support columns show no sign of deformation (see Fig. 1 (a)). However, at lower laser powers material shrinkage results in the overall size reduction of the fabricated structures, while column deformation increases gradually with decreasing laser power (see Fig. 1 (b)(c)). The woodpile walls remain vertical, suggesting that the shrinkage has progressed to its full extent throughout the structure. By measuring the structure dimensions and comparing it to those of the structure design, the linear strain values were obtained. The shrinkage strain along one direction was evaluated as a change in the structure width (W ₀ - W) divided by the design width value W ₀. The strain is therefore defined as γ = (W ₀ - W)/W ₀. The dependence of the linear strain on the average laser power is presented in Fig. 2 (a). The maximum shrinkage strain of 18% was observed for the structures fabricated at 4.5mW. As the laser power increases, shrinkage decreases and becomes negligible for laser powers above 6.5 mW. Below this value, 2PP also occurred, however the mechanical stability of the structures was insufficient to survive the development process.

The line width dependence on the average laser power is shown in Fig. 2 (b). Assuming that the laser beam profile has a Gaussian intensity distribution in the focal plane, the data can be fitted with the following equation [8, 14]:

where d is the width of the photopolymerized line, w ₀ is the focal spot size, P is the applied laser power, and P _th is the laser power threshold for polymerization. At average laser powers below 5.5 mW, the experimental data deviate from the curve fit (dashed curve in Fig.2 (b)), with the actual line widths being smaller. This correlates with the observed material shrinkage dependence on the laser power (Fig. 2(a)). This theory-experiment disagreement can be solved by introducing a correction d × (1 - A × γ) to Eq. (1). When this equation was fitted to the data, (solid curve shown in the Fig. 2(b)) using the strain values from Fig. 2(a), the correction coefficient A = 2 was derived. In this case, the maximum lateral strain of ^{A × γ ≈}35% was obtained, indicating that due to shrinkage the line width was reduced to about one third of its initial value. This value is also suggesting anisotropic shrinkage behavior, where the line width reduces more than its length and the A coefficient describes the degree of this asymmetry.

The SEM image in Fig. 1(c) reveals that the lines are thicker at the intersections when compared to a free-hanging line sections. This is most likely due to a partial arrest of shrinkage at the contact points. The disappearance of these features in the woodpiles fabricated using higher average laser powers (Fig. 1(a)(b)) which do not shrink, confirms this conclusion.

 

Fig. 1. SEM images of representative woodpile structures, applied for investigations of shrinkage. From bottom to the top: structure side view, top view, and a close-up view on the upper layers are presented. The applied average laser powers from left to right are accordingly (a) 8.0 mW; (b) 5.5 mW and (c) 4.5mW. It is observed that the decrease of laser power results in increased material/structure shrinkage, and deformation of the supporting columns.

Download Full Size | PDF

At high average laser powers, the negligible structure shrinkage of the zirconium-silicon composite studied in this paper is due to the double polymerisation ability of sol-gels. During the baking of the sample, condensation of the alkoxy groups takes place to form an inorganic matrix. Alcohol, water, and any other solvents present in the film are released from the system at this stage and the unstructured material shrinks. Subsequently, the structuring by 2PP results in the formation of an organic network without the release of any molecules and thus, to no significant distortion of the structures due to shrinkage.

At illumination parameters slightly above the 2PP threshold, the photopolymerization yield is not 100%. Thus, during developing, the non-polymerized material is removed, leaving a sponge-like material behind. The collapse of this material at a molecular level results in the structural shrinkage observed at low average laser powers (Fig. 3(a) A). On the other hand, for average laser powers above 6 mW the polymerization yield is complete, therefore, the resulting material is mechanically stable and no structural shrinkage is evident (Fig. 3 (a) B).

 

Fig. 2. Characteristics of the polymerized structures: a) linear shrinkage strain in [%] as a function of the applied laser power; b) line width (lateral resolution) dependence on the applied laser power.

Download Full Size | PDF

Non-uniform shrinkage might destroy the structural periodicity of a photonic crystal, resulting in the degradation of its optical quality. To investigate the integrity and functionality of our structures fabricated near the polymerization threshold, their optical properties were analysed by means of microscope-coupled FTIR spectroscopy (Bruker Equinox). Reflectance and transmittance spectra were measured with a probe beam perpendicular to the upper surface of the woodpile. Figure 3(b) shows the spectra obtained for the structure of Fig. 1(c). The reduction in transmission and the respective peak in the reflectance spectra indicate a clear bandstop at around 1330 nm. This suggests that uniformity and structure functionality is preserved even for the structures produced at low average laser powers exhibiting the maximum volume reduction. This is achieved by: a) a building design which allows the structures to shrink freely and b) using a critical point drying process, to eliminate capillary forces during solvent removal.

 

Fig. 3. (a) schematic illustration of the polymerization process at low (A) and high (B) average laser powers; (b) the reflectance and transmittance spectra obtained by a microscope coupled to FTIR spectrometer, indicate a bandstop at 1330 nm

Download Full Size | PDF

It is clear from this study that shrinkage can be used in order to indirectly “improve” the structural resolution. i.e. to decrease feature sizes. In contrast to other proposed post-processing resolution improvement methods [15], in this case not only the line width but also the whole size of the structure is reduced. The minimum lateral resolution obtained for described material is around 100nm. Further studies are in progress in order to optimise the control and benefit of such shrinkage process.

4. Summary

In conclusion, we have introduced a new, simple methodology which allows the direct analysis of shrinkage of photosensitive materials on the microscale. Furthermore, illumination dose dependent shrinkage was investigated and its influence on the structural resolution was studied. In case of woodpile photonic crystal structures, it has been demonstrated that at sufficiently high laser powers the structures produced do not exhibit any measurable shrinkage. At lower average laser powers shrinkage results in a structural reduction of up to 18%, and line width reduction of up to 35% of the original values. At the same time, the shrinkage uniformity was confirmed by the investigation of the optical properties of the woodpile structures. A stopband at around 1330 nm was measured in the reflectance and transmittance.