Photoablative lithography of cellulose acetate at 172 nm: Subtractive 3D printing of biodegradable optical microstructures and molds for polydimethylsiloxane patterning

Cellulose is a biodegradable polymer consisting of D-glucose monomers linked by glycosidic β 1–4 bonds to form linear macromolecules. In the solid state, these macromolecules assemble into complex supramolecular structures that account for the exceptional mechanical strength and unique optical properties of cellulosic materials, which are integral to all lignocellulosic plants.^1,2 Several of the remarkable chemical, mechanical, and optical properties of cellulose have made this renewable material increasingly attractive for medical applications, sensors, and green photonics, in particular.^3–5 Because the mechanical properties of interleaving microfibrils resemble those of human tissue, cellulose often serves as a scaffold for wound disinfection and repair, and drug delivery, and cartilage regeneration.^6–9 Optical applications of cellulose include fiber matrices for random lasers,¹⁰ optical fibers,¹¹ and ensembles of nanocrystals for personal identification security¹² or cellulose elastomers for stretchable, biomimetic optics.¹³ 

Because cellulose comprises a strongly bound hydrogen-bonded network, its melting point lies far above its degradation temperature. Consequently, processing from the melt is not feasible, and the subtle interplay between hydrophilic and hydrophobic interaction forces in the supramolecular structure renders cellulose insoluble in water and common organic solvents. These factors impede or preclude efforts to process bulk cellulose by dissolution and subsequent molding. For applications requiring homogeneous thin films, however, the challenge of processing cellulose is greater, largely due to the microfibril formation of cellulose dispersions.¹⁴ A potential strategy to overcome such barriers is to convert cellulose into soluble derivatives amenable to further chemical or mechanical processing and, if required, subsequently, converted back to cellulose. With this approach, homogenous cellulose-based thin films can be fabricated reproducibly and with low surface roughness,^15,16 both of which are essential for advanced material applications such as photoresists.

The potential of cellulose-derived polymers as self-developing photoresists was recognized in the 1980s by Deutsch and co-workers^17,18 who reported the patterning of nitrocellulose films with a pulsed, 193 nm laser. These and subsequent experiments at 157 nm¹⁹ observed threshold laser pulse energy densities of 20–25 mJ/cm² for ablation and feature sizes as small as 200 nm. However, the interaction of MW peak-power laser pulses with nitrocellulose releases considerable energy (40 eV/nm³),¹³ which severely degrades both depth and lateral resolution, and precludes contact lithography. Recently, Wolfberger et al.²⁰ patterned cellulose films through the irradiation of a trimethylsilyl cellulose (TMSC)/photoacid generator mixture with near-ultraviolet (UV) photons (λ ∼ 350–400 nm). In contact lithographic experiments, UV exposures were typically $>$5 J cm⁻² and both positive and negative patterns were obtained with post-irradiation solvent processing. Direct laser writing of cellulose features has been realized with a focused electron beam²¹ or through two-photon absorption at 780 nm in cellulose diacetate²² or TMSC/photoacid generator films.²⁰ 

We report here the patterning of organosoluble cellulose acetate (CA) and cellulose acetate butyrate (CAB) polymer films by photoablative contact lithography with flat lamps emitting at 172 nm (hν = 7.2 eV) in the vacuum-ultraviolet (VUV) spectral region. Photoetching rates for these biodegradable polymers in an atmospheric nitrogen environment with VUV intensities of only 10 mW cm⁻² are linear in lamp exposure time (∼0.8 μm/h) for depths up to at least 25 μm, which corresponds to the removal of one anhydroglucose molecule from the surface for ∼100 absorbed photons. If desired, higher material removal rates are available with recently developed 172 nm lamps, which offer spatially uniform intensities $>$100 mW cm⁻² over radiating areas of up to 100 cm² (Ref. 26). Mass spectrometric analysis of the effluent from an irradiated CA surface confirms the primary nascent photoproducts at this wavelength to be H, O, OH, CH₃, and CH₂. The precision in specifying trench depth is consistently 15 nm, which allows for complex nanostructures and optical and biomedical components, such as optical gratings and microfluidic channels, to be fabricated quickly and without any solvents by this subtractive three-dimensional (3D) printing process. Sophisticated optical components, such as sinusoidal gratings, have been realized through multiple exposures and control of mask undercutting by uncollimated VUV radiation. The ability to precisely photoetch cellulose polymers has also opened the door to fabricating molds for micro- and nano-imprinting other materials. As an example of the potential of cellulose VUV photochemistry, a reusable mold for patterning polydimethylsiloxane (PDMS) by a solvent-free process is also described.

Figure 1(a) illustrates the chemical structure of anhydrocellobiose, which is the repeating unit forming the planar cellulose backbone. Typically, the polymerization order n for cellulose lies in the 5000–15 000 range. The primary impact of the acetylation process transforming cellulose into CA is to attach hydroxymethanolate groups (⁠$CH3O2−$⁠) to the backbone, which has the effect of improving the mechanical properties of cellulose while retaining its biocompatibility and insolubility in water.⁷ A comparison of the photoetching rates for CA and polymethylmethacrylate (PMMA)²³ films at 172 nm and 300 K is presented in Fig. 1(b) for the VUV intensity at the film surface fixed at 10 mW cm⁻². These data were acquired by photoablating the films in a π²-Cygni exposure system (Cygnus Photonics, Inc.) having a 43 × 43 mm² open aperture. The flat, 172 nm source was situated 3 cm from, and parallel to, the film. The system was purged with dry N₂ at a flow rate of 4 l/min because of the transparency of nitrogen at 172 nm. All of the custom contact masks were of Cr/fused silica construction, and surface topographies of the patterned surfaces were measured by laser confocal microscopy. Further details concerning the experimental arrangement and data acquisition processes can be found in Sec. IV, and the PMMA etching profiles of Fig. 1 were derived from the data of Ref. 23. Although the PMMA photoetching rate is initially ∼4 nm/s for exposure times t between ∼10 and 100 s, it quickly saturates, and as shown in Fig. 1(c), the maximum depth attained to date is ∼1–1.5 μm. In contrast, photoablation of CA at 172 nm for t $<$ 2 min proceeds at a rate of ∼1.7 nm s⁻¹ but subsequently slows to 0.8 μm h⁻¹, as represented by the linear least-squares fit to the data of Fig. 1(c) (indicated by the black solid line). Following this transition, the CA photoetching rate shows no signs of saturation, even after more than 50 h of VUV irradiation. Similar removal rates and exposure behavior were observed for CAB. It is difficult to overstate the importance of this result because previous experiments with other polymers, including PMMA, acrylonitrile butadiene styrene (ABS), and polyvinyl acetate (PVA), observed a trench depth limit of ∼1.5 μm.²³ Although this barrier could be bypassed with a multistep exposure process, rinsing with solvents was required after the removal of 200–500 nm of material, which also necessitated a realignment procedure.

The self-limiting nature of PMMA photoablation at 172 nm, for example, arises from the formation of non-volatile products in an ablated region such as a trench. The thickness of this material layer grows with exposure time and eventually must be removed with solvents. Otherwise, the etching process is terminated. Organic material is also observed within features etched in CA, but the thickness of this non-volatile photochemical precipitate saturates because it is eventually volatilized by the VUV radiation in a secondary photolytic process. That is, a steady-state is reached at which the rates for volatilization and photoproduction of the precipitate are equalized. The optical processes driving the photolysis of the bulk CA and the precipitate are assumed to entail the absorption of a single photon because the peak power of the pulses generated by the lamp is $<$100 W. This presumption is supported by mass analysis of the effluent generated at the film surface by the lamp. For these measurements, the chamber was evacuated to a base pressure of $<10−5$ Torr, and the mass spectrum, observed up to 47 amu, is given in Fig. 1(d). The dominant peaks, associated with H, OH, H₂O, and CH₃, are consistent with the expectation that the primary nascent species produced by the photolysis of the CA surface are H, O, OH, and CH₃. However, peaks due to HCO and the CH₂ radical are also noticeable, which confirms the importance of collisional kinetics occurring in the effluent and the absorption of a second 172 nm photon by a fraction of the volatile by-products of VUV/cellulose photochemistry. Photodecomposition of the hydroxymethanolate group (⁠$CH3O2−$⁠) in CA, substituted during the acetylation process for OH in cellulose, is responsible for the strength of the methyl, hydrogen, and oxygen features in Fig. 1(d). Another source of oxygen in the mass spectrum is cleaving of the covalent bonds linking the anhydroglucose chain. Because the 7.2 eV energy of each photon is insufficient to photoionize most of the atomic and molecular photofragments (such as H and O, both with ionization potentials of ∼13.6 eV), the number density of ions in the effluent is expected to be negligible.

On the basis of the photoablation rates of CA in Fig. 1(c), the removal of one anhydroglucose molecule from the surface is estimated to require 80–100 photons. The photoetching of CA and CAB with lamps of modest intensity (⁠$<$50 mW cm⁻²) is remarkable when one considers that the photoablation of organic films and tissue with pulsed 193 nm (ArF) lasers, first reported in the 1980s,^24,25 required peak intensities of $>$1 MW cm⁻². In contrast, the lamps employed for the present study emit trains of 172 nm pulses, each having a peak power of $<$100 W.

Trenches and other features having depths beyond 25 μm can now be fabricated in cellulose acetate and cellulose acetate butyrate with precision. Figure 2 shows false color images of a variety of square and linear trenches photomachined into CA by exposing the surface of commercial-grade material through a Cr/fused silica mask. Acquired by laser confocal microscopy, these images have a depth (axial) and lateral resolution of $<$15 and 600 nm, respectively. Figure 2(a) shows an image of a portion of a periodic arrangement of lines and squares, all having depths of 21 μm. The width of the lines is 15 μm, and it should be noted that all dimensions in the xy plane (surface of the CA sheet), and along the z coordinate oriented orthogonal to the plane, are to scale. Figure 2(b) shows an image of a 50 μm square mesa rising 20 μm above the CA floor, which is noticeably rough. A segment of a pattern comprising 15 μm wide lines and 2.3 μm deep trenches is shown in Fig. 2(c), and it is evident that the rms floor roughness is considerably smaller than that observed for more deeply etched structures such as the mesa of Fig. 2(b). Indeed, extensive measurements of the characteristics of trenches of differing depths found the floor roughness to be negligible for depths below ∼2 μm but to scale linearly with larger etched depths, as illustrated in Fig. 2(d). For deeper features, the rms floor roughness is ∼9% of the trench depth if the process is solvent-free [“dry” process; red curve, Fig. 2(d)]. However, smoother floors are obtained if the CA substrate is rinsed with isopropyl alcohol (IPA) after the etching process, as shown by the black curve of Fig. 2(d). Before leaving this subject, it should be reiterated that all substrates for these experiments are commercial-grade CA sheets and the particles observed on the otherwise smooth floor of Fig. 2(c) are CA conglomerates not removed by filtration prior to the plasticizing process.

More complex patterns fabricated by a double exposure photoablation process are presented by the false color images of Fig. 3. Both structures were obtained with two 1 h exposures of the lamp to the CA surface, and the mask was rotated between each exposure. In Fig. 3(a), the mask comprised equally spaced lines and corners and, after the first exposure, was rotated by 30°. The depths of each layer are indexed to a color with blue representing the deepest portions of the pattern. Each green and red section denotes a layer thickness of 0.7 and 1 μm, respectively, and the precision in defining and reproducing the thickness of a photoablated layer is 15 nm. Chirped gratings consisting of either lines or posts arranged in a sequence having monotonically decreasing spacings are shown in Fig. 3(b). In this instance, the mask comprised a pattern of equally spaced lines and was rotated by ∼30° following the first exposure. The widths of the smallest posts in this image (at right) are ∼5 μm, and the heights of all features are ∼1.7 μm. Because the radiation generated by the 172 nm lamp in the exposure system is not fully collimated and the lamp-mask separation is ∼3 cm, slight undercutting of the Cr mask occurs and was exploited to obtain features having a sinusoidal surface profile. The false color images of Fig. 4(a) provide one such example for an array of lines and cones formed by a double exposure of the surface to the 172 nm lamp. Although a single standard resolution mask was used, the cellulose polymer sample was exposed through two separate regions of the mask for 1 h each. The white dashed line near the top of the image denotes the path for an intensity lineout, which is represented in Fig. 4(b) by the black curve. This topographical profile is almost precisely sinusoidal, as illustrated by the red trace, which is the best-fit of a sinusoid to the data. Another example of the potential of multiexposure subtractive patterning of CA is that of Fig. 4(c). A double-exposure process has again produced sinusoidal ripples, having a period and peak height of ∼5 and 1 μm, respectively, embedded within, and lying in a plane below, an array of lines 0.6 μm in height and also having a sinusoidal transverse spatial profile.

One characteristic of 172 nm photoablation of cellulose polymers is the preservation of the surface topography (prior to etching) in the final surface, regardless of depth. Two false color images of a sinusoidal grating structure, also fabricated by a double-exposure process, are shown in Fig. 5. The first exposure of a CA surface produced a sinusoidal grating with a period of ∼5 μm, and the second 1 h exposure involved a simple mask having an aperture outlined by the blue region of Fig. 5(a). As shown by the magnified portion of Fig. 5(a) given in Fig. 5(b), the periodicity of the original (upper) grating is replicated within the quasi-rectangular trench. These and other experiments conducted over the past few years demonstrate that producing optical structures and components, such as gratings, in the floor of microfluidic components or waveguides, for example, is now feasible. We also note that the performance of gratings similar to those of Figs. 3–5 is virtually identical to that for PMMA gratings reported previously.²³ 

Another application of the precise photoablation of cellulose polymers is fabricating negative molds for patterning polydimethylsiloxane (PDMS) and doing so in a non-cleanroom environment. Microfluidic and other biomedical components are often constructed from PDMS and fabricated by replica molding from a Si master mold. As demonstrated here, the cellulose polymers can be photomachined with submicron precision in 1 atm of nitrogen, thereby dispensing with the requirement for a cleanroom environment and the associated infrastructure. Figure 6 shows two false color images of PDMS line and corner patterns fabricated with a CA negative mold. Figure 6(a) shows a portion of a periodic array of 35 μm wide lines separated by trenches having a depth of 5.5 μm. Panel (b) of Fig. 6 is an image of a segment of another pattern (line and corner) that shows the low value of the rms roughness (⁠$<$35 nm) of the trench floor and the steep sidewalls that can be realized, despite the trench depth of ∼2 μm. An image of a negative mold fabricated in CA is presented in Fig. 7(a). The trench depth throughout the pattern is 18 μm, and the squares are 40 × 40 μm² in surface area. Figures 7(b)–7(d) show the first, second, and third positive replicas from this mold, respectively. A thorough examination of the replicas finds that all of the features of the master mold are reproduced by the replicas with no noticeable loss in resolution.

Advances in solid state photochemistry have been impeded in the past by a lack of efficient VUV sources, particularly those capable of uniformly irradiating substantial surface areas. Few lamps and lasers have existed in the 100–200 nm spectral region, and the latter (157, 193 nm) typically operate at low duty cycles. Recently developed flat lamps offering time-averaged intensities at 172 nm beyond 300 mW cm⁻² and average powers above 30 W represent a turning point for VUV photochemistry in both organics and inorganics.²³ With respect to the cellulose polymers, a 172 nm photon has more than sufficient energy (7.2 eV) to shear any of the chemical bonds of interest to anhydroglucose or other organics, including the C=C bond. Because the photon flux at the surface is ∼10¹⁴ photons cm⁻² s⁻¹, this non-thermal photoablation process does not require two-photon absorption and the concomitant optical field intensities necessary for direct writing with visible or near-infrared radiation.

The present experiments demonstrate the ability to pattern cellulose polymers in a solvent-free process that takes place at atmospheric pressure and does not require conventional lithographic and micro- or nano-processing tools. Photoablation of CA and CAB at 172 nm with flat lamps is a nonthermal etching process that dispenses with the photoresist and solvents and yet produces intricate patterns and optical structures by contact photolithography with a depth precision of 15 nm. Because of the drawbacks of contact photolithography with respect to large scale production, the development of a lamp-based projection lithography system is in progress. Although the lateral resolution limit of this photoetching process has not yet been determined, the structures reported here demonstrate wall profiles and depth precision that provide access to microparticle and cell sorting applications, for example. Capable of producing trenches and other features $>$25 μm in depth at a photoetching rate of ∼0.8 μm/h (for 10 mW cm⁻² intensity at the mask), this dry process is well-suited for the direct fabrication of biomedical, photonic, and electronic devices and systems. Furthermore, the application of VUV-machined micro- and nano-structures in cellulose as a mold or scaffold for PDMS microfluidic components by replica molding has also been demonstrated. Finally, we note that higher photoablation rates are available with lamps offering greater intensities. Consequently, by overcoating photomachined cellulose or PDMS structures with dielectric and conducting films, a wide range of photonic, biochemical, and biomedical devices and components will be forthcoming.

The cellulose acetate and cellulose acetate butyrate samples in this study were irradiated in a Cygnus Photonics exposure system containing a 43 × 43 mm² emitting aperture, flat 172 nm lamp, and a Cr/fused silica mask. Samples were prepared from commercially available sheets of 0.5–2 mm thickness. No detectable variations in the photoablation patterns generated were observed for samples of differing thicknesses. The presence of particles up to several μm in size in the CA samples appears to be the result of limitations in filtering when CA was plasticized by the manufacturer. The CA and CAB samples were situated in contact with a Cr/fused silica mask and ∼3 cm from the lamp surface. The fused silica substrate for the mask is necessary for acceptable transmission at 172 nm. The effective pulse repetition frequency (PRF) for the 172 nm lamp is 40 kHz because the voltage waveform driving the lamp has a PRF of 10 kHz and four VUV pulses are emitted for every waveform cycle. The photolithographic masks were fabricated with a Heidelberg MLA150 maskless aligner. Prior to each experiment, the chamber was purged with nitrogen at 1 atm pressure for 15 s. The VUV lamp intensity at the surface of the mask was measured with a calibrated Hamamatsu detector designed specifically for the 172 nm spectral region.

Patterned CA and CAB surfaces were examined by laser confocal microscopy with a Keyence VHX-1000 laser profilometer operating at 405 nm and having a lateral and axial (depth) resolution of $<$600 and $<$15 nm, respectively. Shallow optical gratings (⁠$<$2 μm deep features) photoablated into CA were characterized by visible laser diffraction, but grating order diffraction intensities were not measured for deeper gratings because of the rough floor topography. To measure the photofragment product distribution generated by 172 nm irradiation of CA, a sample was placed into a vacuum system having a residual gas analyzer (Inficon Transpector MRH) and evacuated by a turbomolecular pump to a base pressure of 5 × 10⁻⁶ Torr. A MgF₂ window on the vacuum system permitted access for the lamp VUV radiation to the substrate. Figure 1(c) shows a representative mass spectrum (with background subtracted), recorded 20 s after the onset of 172 nm radiation to one sample.

Access to surface profilometry and mask manufacturing tools of the Materials Research Laboratory at the University of Illinois and the support of this work by the U.S. Air Force Office of Scientific Research (AFOSR) under Grant No. FA9550-18-1-0380 (G. Pomrenke and J. Luginsland) are gratefully acknowledged.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.