Investigation of the photoreaction
For the investigation of the photoinduced desilylation reaction, thin films of TMSC containing 2 wt% of the non-ionic photoacid generator N-hydroxynaphthalimide triflate (NHNA) were prepared by spin coating onto CaF2 plates, leading to films with a thickness of approximately 190 nm. In order to prevent a photodegradation of the material, which preferentially occurs at wavelengths of λ ≤254 nm, UV-irradiation was carried out under nitrogen atmosphere at wavelengths higher than 300 nm. UV-exposure of a TMSC/NHNA blend yields triflic acid as the main photoproduct, which subsequently causes a cleavage of the trimethylsilyl (TMS) groups, resulting in a conversion of TMSC to cellulose as depicted in Scheme 1.
The photoinduced desilylation was followed by means of FTIR spectroscopy. In Fig. 1, the obtained FTIR spectra of thin TMSC/NHNA films prior to and after exposure to UV-light (E = 5.2 J cm−2) are displayed. The FTIR spectrum before UV-illumination shows a weak signal at 3,490 cm−1, which can be assigned to residual hydroxyl moieties stemming from an incomplete silylation of the hydroxyl moieties as expected for TMSC with a degree of substitution (DSSi) of 2.8, which was used for this study. UV-illumination leads to a significant increase of the O–H stretching vibration, while the intensities of the Si–C rocking vibrations at 1,250 cm−1 and C–H stretching vibrations at 2,957 cm−1 decreased, which is consistent with the proposed desilylation reaction in Scheme 1. The observed changes in the FTIR spectra reveal a nearly complete conversion of TMSC to cellulose. These findings were also supported by X-ray photoelectron spectroscopy (XPS). In accordance with the FTIR study, C1s detail spectra (shown in the supporting information) exhibit a significantly lower Si–C signal at 284.6 eV in relation to the C–O signal at 286.9 eV after illumination. Additionally a decrease in the total Si content from 11.7 to 2.7 at.% at the sample surface can be observed, which additionally confirms the UV-induced desilylation reaction. The photo-generated hydroxyl groups influence the surface energy of the TMSC/NHNA layers towards a higher polarity, which has already been observed for desilylation of TMSC using vapors of hydrochloric acid (Mohan et al. 2011, 2012). A study describing this behavior in detail can be found in the supporting information.
Furthermore, it was found that the amount of NHNA present in the TMSC films significantly influences the conversion of TMSC to cellulose. The kinetic behavior of the desilylation reaction for different PAG concentrations was determined by evaluating the decrease of the Si–C band during UV-illumination as shown in Fig. 2a. While a UV exposure of TMSC, containing 5 and 10 wt% NHNA leads to an almost complete desilylation (approx. 11 % remaining silyl ether groups after illumination with E ≥1.8 J cm−2), lower NHNA concentrations of 1 and 2 wt%, lead to an insufficient conversion of 7 and 27 %, respectively, after a prolonged illumination with an irradiation dose of E = 7.3 J cm-2. TMSC films without photoacid generator show no decrease of the Si–C band, leading to the conclusion that an UV induced cleavage of the silyl ether bond does not occur without an additional PAG component.
It has to be mentioned that the desilylation reaction also proceeds in the TMSC films after UV-irradiation. This phenomenon is already known from cationic photo-polymerization reactions and is referred to as “dark reaction”. In this case, the photo-generated protons catalyze the desilylation reaction in absence of UV-light. This reaction causes a further decrease of the silyl ether content from 93 to 82 % and from 73 to 58 % in TMSC films containing 1 and 2 wt% NHNA, respectively, after storage under exclusion of light for 24 h as depicted in Fig. 2a. FTIR spectra, recorded after 48 h, did not exhibit any further decrease of the Si–C band.
Photolithographic patterning of TMSC
Negative type development
Due to the fact that the photo-generated cellulose is insoluble in common organic solvents such as chloroform or toluene, a direct photolithographic patterning of TMSC, containing small amounts of PAG is possible, yielding negative type cellulose structures after development. The changes in solubility of TMSC films, caused by the photoinduced desilylation reaction were assessed by means of sol–gel analysis. The insoluble fraction (gel fraction) was determined by FTIR spectroscopy by evaluating the intensity of the C–O–C stretching vibration of the glycosidic bond at 1,150–1,170 cm−1 and comparing the peak height before and after development in chloroform for 10 min. Figure 2b represents the gel fraction of TMSC-films containing 2 wt% NHNA and 5 wt% NHNA as a function of the irradiation dose, revealing maximum gel fractions of 73 and 98 %, respectively, after UV-illumination with an irradiation dose of E ≥0.97 J cm−2. Prolonged exposure does not further influence the gel fraction, which is in good accordance with the kinetic behavior shown in Fig. 2a. A decrease in solubility after an additional exposure (irradiation doses up to 70 J cm−2) could not be observed, which excludes a photodegradation of the cellulose backbone. Compared to commercially available photoresists, the required illumination doses in the range of 1 J cm−2 are rather high for photolithographic patterning. This can be attributed to an insufficient spectral overlap of the UV absorption spectrum of NHNA and the used polychromatic irradiation source (as displayed in the supporting information). It can be assumed that a better matching of the spectral overlap leads to a reasonable photoresist performance. With respect to a possible application of these films as a dielectric material in organic thin film transistors, further photo-patterning experiments were performed with TMSC films, containing 2 wt% NHNA. Although, this concentration leads to an incomplete conversion of the TMSC to cellulose, the changes in solubility are sufficient for a successful photopatterning. Higher PAG contents (and their ionic photocleavage products) may also negatively influence the device stability, therefore we aimed for a compromise between low PAG concentration and a high obtainable gel fraction. A NHNA concentration of 2 wt% was found to be ideal in this respect.
Positive type development
Going a step beyond conventional resist development using organic solvents, enzymatic digestion has been evaluated for the realization of positive type cellulose structures. It is well known that cellulose can be digested by appropriate enzymes, e.g. cellulase from T. viride (Kargl et al. 2013; Mohan et al. 2013a; Ahola et al. 2008). Utilizing this fact, the selective enzymatic digestion of photo-regenerated cellulose has been investigated. Consequently, the illuminated films were immersed in an acetate buffer solution containing a mixture of cellulose digesting enzymes. For a complete enzymatic development, the kinetic behavior of digestion has been determined by measuring the decrease in the film thickness of UV-illuminated TMSC films (E = 4.6 J cm−2), prepared on silicon wafers, for different periods of time of enzymatic digestion. The determination of the remaining film thickness by means of FTIR spectroscopy was not applicable in this case, because the glycosidic bonds which were used for the previous sol–gel analysis are expected to be hydrolyzed due to enzyme activity. Since other FTIR signals were not found to be suitable for a sol–gel analysis, AFM was employed to characterize the films, exhibiting a film thickness of approximately 180 nm before UV-illumination. Figure 3a shows the remaining film thickness (related to the film thickness after illumination) as a function of the immersion time. It turned out that an almost complete digestion (4 % of the original film thickness) is achieved after 8 h of immersion. Moreover, the rms roughness Rq of the TMSC films significantly increases during enzymatic digestion and decreases again after 8 h of immersion in enzyme solution, indicating that only some residue of original film is remaining on the surface. In a control experiment, a degradation of non-illuminated TMSC films, containing 2 wt% NHNA could not be observed, demonstrating the selectivity of the used enzymes. In order to determine the sensitivity of this positive type photoresist in combination with an enzymatic development, a sol–gel analysis has been performed. Consequently, TMSC films, containing 2 wt% NHNA were illuminated for different periods of time and developed by an enzymatic treatment (16 h of immersion at room temperature). In Fig. 3b, the remaining film thickness (determined by means of AFM) of the illuminated TMSC film is plotted as a function of the illumination dose. The graph shows a pronounced threshold at an irradiation dose of 0.49 J cm−2 (69 % remaining thickness), leading to a complete digestion after an illumination dose of 1.9 J cm−2. These findings reveal that also partially regenerated cellulose (Si–C band intensity reveals a DSSi of approximately 0.5 after irradiation) is digested by the used enzymes, while pristine TMSC (DSSi = 2.8) is not affected at all. Although it appears surprising on first glance, the digestion of lowly substituted celluloses (e.g. cellulose acetate) has been already described in literature some time ago in the context of biodegradability of cellulosic materials (Reese 1957). In order to determine if the photogenerated acid in the illuminated films can also lead to a formation of water-soluble degradation products, TMSC films, containing 2 wt% NHNA were illuminated (E = 4.6 J cm−2) and immersed in buffer solution for an extended period of time without adding any enzyme. It was observed that the regenerated cellulose in the illuminated areas remained on the substrates, even after 72 h of immersion in acetate buffer.
In order to demonstrate the applicability of this material as a dual tone photoresist, thin TMSC films, containing 2 wt% NHNA were photopatterned on a mask aligner system. After patterned UV-irradiation (E = 5.4 J cm−2), the development of the TMSC films was either performed in chloroform or, alternatively, in a cellulase solution as depicted in Fig. 4a–c. For the fabrication of positive type cellulose structures, patterned TMSC films were subsequently treated with vapors of hydrochloric acid for 90 s (Fig. 4d), leading to a complete regeneration (Petritz et al. 2013). In addition, the intensity of the signal of the glycosidic bond at 1,150–1,170 cm−1 did not change after HCl treatment, leading to the conclusion that no significant degradation occurred. The observed decrease in the film thickness after the acid induced conversion of TMSC to cellulose can be related to a change in the density of the film due to the formation of hydrogen bonds of the resulting cellulose. Although both approaches yield well-defined cellulose structures with lateral resolutions in the range of 1–2 µm, the enzymatic development results in less well-defined edges as shown in the cross section in Fig. 4c. This can be explained by the comparably poor resist behavior, which is also reflected by the sol–gel analysis in Fig. 3b. Additionally, the extended development time of 24 h as well as the subsequent treatment with hydrochloric acid can negatively influence the quality of the obtained structures. Accordingly, the surface roughness Rq of these films increases during the patterning procedure from 0.70 ± 0.04 nm (non-illuminated sample) to 1.82 ± 0.02 nm (after patterned illumination, enzymatic development and subsequent hydrochloric acid treatment). Detailed AFM micrographs, revealing the surface morphology prior to and after the patterning procedure can be found in the supporting information. A negative type development of the patterned films in chloroform does not influence the surface roughness significantly (Rq = 0.75 ± 0.05 nm after development).
Two-photon absorption (TPA) lithography
Advancing from micro- to submicrometer resolutions, TPA lithography was evaluated for the fabrication of cellulose patterns. The two-photon excitation technique has attracted interest due to its intrinsic 3D processing capability and is considered as a promising technology with unique advantages regarding nanofabrication, enabling lateral resolutions of less than 100 nm (Lee et al. 2008; Juodkazis et al. 2005). Because of the aforementioned negative effects of enzymatic digestion on the achievable resolution, a negative type development using organic solvents was chosen in these experiments. For sub-µm patterning, TMSC films containing 10 wt% NHNA were patterned on a commercial lithographic setup with a laser power of 15 mW and lateral feed rate of 50 µm s−1. In this patterning setup, the laser beam (λ = 780 nm, repetition rate 100 MHz, pulse width 150 fs) causes multi-photon absorption in its focus, leading to an energy transfer from the laser to the PAG which initiates the desilylation reaction. Figure 5a shows the obtained pattern directly after TPA lithography visualized by means of AFM. A subsequent development of the film in toluene for 15 min yields freestanding cellulose structures with a height of approx. 180 nm and a lateral resolution of approx. 550 nm (full width at half maximum, FWHM) as shown in Fig. 5b, c. One important prerequisite for a two-photon induced patterning process is a sufficient TPA efficiency, i.e. a high TPA cross section of the used photoinitator, which mainly depends on the delocalized p-electron system of the photoinitiator (Pucher et al. 2009). Although the used photoacid generator offers an appropriate UV absorption behavior that corresponds with the applied laser wavelength under two-photon conditions, the TPA activity of this commercially available PAG has not been determined in detail. It can be assumed that tailored photoinitiators which provide better TPA coefficients enable even higher resolutions, comparable to those reported for cationic photoresists (e.g. SU-8).
Application of photopatterned cellulose films as OTFT gate dielectrics
In order to highlight a potential application for this cellulose based photoresist in the emerging area of organic electronics, photopatterned films were implemented as dielectric layers in low-voltage OTFTs with a device setup as shown in Fig. 6a. In general, patternable dielectric materials have to offer reasonable dielectric properties in addition to an appropriate resist behavior (in terms of sensitivity and resolution). For that purpose, the dielectric properties of 32 nm thin cellulose films, obtained by negative type photolithographic patterning of spin coated TMSC films, containing 2 wt% NHNA were investigated in capacitor structures. A current/voltage (I/V) measurement is plotted in Fig. 6b, revealing leakage currents in the order of 10−6 A cm−2 at an electric field of 1.3 MV cm−1, which is quite low, considering that a 32 nm thin patterned polymeric dielectric is used. A capacitance of approx. 130 nF cm−2 was measured, corresponding to a dielectric constant of εR = 5.4 ± 0.7 at 1 kHz with a sufficiently high frequency stability of εR up to 100 kHz (illustrated in the inset in Fig. 6b). The observed permittivity is significantly lower in comparison to the permittivity which we observed for cellulose films, fabricated by vapor phase acid hydrolysis in our previous work (εR = 8.4 ± 0.5) (Petritz et al. 2013). As already mentioned above, XPS as well as FTIR measurements reveal a remaining Si–C signal after the photochemical regeneration of cellulose which clearly indicates that the TMSC is not completely regenerated. In contrast, the cellulose films fabricated by vapor phase acid hydrolysis previously showed no remaining Si–C-signals and proved to be completely regenerated (Petritz et al. 2013). Therefore, a permittivity of 5.4 of the photochemically regenerated cellulose films, lying between fully regenerated cellulose (εR = 8.4) and pristine TMSC with a DSSi of 2.8 (εR = 2.3) can be explained by the residuals of TMSC in the films.
The electrical characteristics of a pentacene based OTFT with a negative type photolithographically patterned cellulose film as gate dielectric is plotted in Fig. 7. The output characteristics in Fig. 7a display a clear saturation of the drain current level with the gate bias. Furthermore, no hysteresis between forward and reverse drain voltage sweep is observed. In Fig. 7b the transfer characteristic is plotted, showing ID(VG) and IG(VG) in a semi-logarithmic representation as well as the square root of the drain current as a function of the gate bias. From the transfer characteristics the important performance parameters of the OTFT can be extracted, revealing an onset voltage Von = −0.8 V, a threshold voltage of Vthr = −1.25 V and a subthreshold swing S as low as 110 mV dec.−1. The subthreshold swing S is the inverse of the maximum slope of the (quasi)linear part of the subthreshold current (dashed line in the semi-logarithmic plot of the transfer curve, as shown in Fig. 7b. Furthermore, the fabricated low voltage OTFTs show no hysteresis, gate leakage currents in the range of 80 pA, OFF-currents around 60 pA and a linear field effect mobility µlin of 0.08 cm2 Vs−1.
As a final remark we want to comment on the interface properties of our photopatternable dielectric material with the organic semiconductor in the OTFT, which is directly affecting the device performance. A low interface charge trap density is particularly essential for the fabrication of fast and stable organic electronic circuits. An upper limit of the density of interfacial trap states Nss,max can be calculated from the obtained subthreshold swing according to the method reported by Rolland (1993). For the determined subthreshold swing of 110 mV dec.−1 an upper trap density limit of Nss,max = 6.9 × 1011 cm−2 eV−1 is calculated. The extracted NSS values for cellulose based OTFTs are exceptionally low and are in fact much smaller than the average interface trap density of states observed in amorphous silicon TFTs being in the range of 1012 cm−2 eV−1.