Impact of size effects on photopolymerization and its optical monitoring in-situ

Photopolymerization processes are exploited in light exposure-based 3D printing technologies, where either a focused laser beam or a patterned light sheet allows layers of a UV curable, liquid pre-polymer to be solidified. Here we focus on the crucial, though often neglected, role of the layer thickness on photopolymerization. The temporal evolution of polymerization reactions occurring in droplets of acrylate-based oligomers and in photoresist films with varied thickness is investigated by means of an optical system, which is specifically designed for in-situ and real-time monitoring. The time needed for complete curing is found to increase as the polymerization volume is decreased below a characteristic threshold that depends on the specific reaction pathway. This behavior is rationalized by modelling the process through a size-dependent polymerization rate. Our study highlights that the formation of photopolymerized networks might be affected by the involved volumes regardless of the specific curing mechanisms, which could play a crucial role in optimizing photocuring-based additive manufacturing.


Introduction
Photopolymerization processes, which lead to the formation of solid polymer networks starting from monomers and oligomers, are arousing a continuously increasing interest in many fields of science and technology. This is motivated by the importance of predicting the resulting physical properties of polymerized systems with improved accuracy, and of enhancing various applications in surface coatings, adhesives, dentistry, lab on chips, soft actuators and photonics [1][2][3]. In photopolymerization, light is used for curing pre-polymer solutions which contain suitable photoinitiators. These photosensitive molecules produce reactive species upon photon absorption and initiate reactions proceeding through free-radical or cationic species [4,5]. Such processes show features that make them highly suitable for micro-and nanofabrication for various aspects. First, the intensity profile of a light beam can be structured by using a physical shadow mask, a digital micromirror array or a liquid-crystal phase mask [6][7][8], and such structured light beams can be used to pattern thin layers of photosensitive pre-polymer mixtures. Moreover, light can be focused into diffraction-limited volumes through optical systems, controlling the focal spot size by the numerical aperture (NA) of the focusing optics, thus allowing photopolymerization to be localized at submicrometric scale, whereas complex patterns can be realized by scanning the light beam [9]. The absorption of light can occur also by nonlinear phenomena, such as two-photon absorption [10,11], and single or multicolor activation and deactivation processes [12][13][14], enabling additional options for spatial localization. Technologies impacted by these aspects span from photolithography to laser writing and 3-dimensional (3D) vat photopolymerization [15][16][17][18]. Moreover, photopolymerization can be combined with self-assembling, phase-separation and ink-jet processes, which are used, for instance, to fabricate arrays of lenses and micro-structures [19][20][21]. 4 include oxygen-due inhibition, thickness-dependent heating, and others), the reduction of the polymerization rate upon decreasing thickness appears to be general. These findings are rationalized with a model which introduces a thickness-dependent polymerization rate, well describing the trends measured for the intensity of the backscattered light. Monitoring of polymer photocuring in-situ and in real-time, and accounting for size-dependent polymerization, might be relevant for nextgeneration, intelligent additive manufacturing technologies.

Materials
Various UV-curable feedstock compounds are used as summarized in Table 1. These materials are selected for investigating different polymerization reactions and timescales, the latter depending on the coefficient of light absorption at 405 nm (the wavelength of the light used for curing). The bisphenol-A-ethoxylate dimethacrylate (BisEMA) oligomer (Mn=1700, Sigma Aldrich) is mixed with 2% weight:weight (w:w) of 2,2-dimethoxy-2-phenylacetophenone photoinitiator (Sigma Aldrich, Table 1). The mixture is vortexed until the complete dissolution of the photo-initiator [24] and, afterwards, it is degassed in a vacuum chamber. The absorption spectrum of this pre-polymer mixture is shown in the Supplementary Figure 1. The photoinitiator has a strong absorption band at wavelengths <300 nm and a weaker absorption band in the 300-350 nm range. The latter further extends in the visible range, which allows radiation with a wavelength of 405 nm to be used for photopolymerization. The mixture is deposited on quartz substrates by spin-coating (8000 rpm for 60 s). So-formed films are uniform, but they undergo a dewetting process within a few minutes. This leads to the formation of an ensemble of microdroplets with size ranging from 1 mm down to 1 µm (Supplementary Figure 2). E-shell ® 600 (Envisiontec) is a photocurable resin containing a phosphine oxide photoinitiator [33].
It is used as received by the manufacturer ( Table 1). Droplets of E-shell ® 600 (Envisiontec) are deposited by material jetting (the set-up is illustrated in the Supplementary Figure 3) on a 2×2 cm 2 Published in Additive Manufacturing, doi: 10.1016/j.addma.2022.103020 (2022). 5 quartz substrate, by using a F5200N.2 robotic system (Fisnar). Droplet jetting of E-shell ® 600 is carried out using a gauge 32 needle and by setting a pressure of 8 psi and a deposition time of 0.1 seconds. After printing, the droplets are photopolymerized by using the Micro Plus HD (ENVISIONTEC ® ) 3D printer (exposure time 1-9 s, intensity of UV light of 3 mW cm -2 ). SU-8 (Microchem) is an epoxy resin containing triarylsulfonium salts as photoacid generator [34,35]. It is used as received by the manufacturer without further addition or purification (Table 1). Thin films of SU-8 are deposited on cleaned quartz substrates (1×1 cm 2 ) by spin-coating. The thickness of the films is varied in the range 0.5-5 µm by changing the spin-coating speed (500-8000 rpm) and the used pre-polymer (SU-8 2000.5 for thickness < 1 µm and SU-8 2002 for thickness > 1 µm, respectively, both from Microchem). The deposition of the SU-8 films is followed by a soft-bake step in an oven (~95 °C for 1 hour).  6 The thickness of the samples is measured by a stylus profilometer (Dektak). The morphology of the printed structures is characterized by scanning electron microscopy (SEM), using a Merlin system (Zeiss).

Photopolymerization and in-situ monitoring
The system used for photopolymerization and in-situ monitoring of the process is based on a confocal microscope set-up (mod. FV1000, Olympus). The main components of this set-up are schematically illustrated in Figure 1a. It includes a diode laser source with 405 nm emission wavelength and maximum power of 1 mW, as measured at the sample position by using a power meter (mod. 843-R, Newport). The laser is coupled to an inverted microscope (IX71, Olympus) and focused onto the sample by a 10× objective (Numerical Aperture, NA=0.4). The spot size (2w0, where w0 is the beam waist) is about 1.2 µm. This system can be used also for 3D printing experiments through vat photopolymerization, performed by using an approach similar to the one reported in Ref. [36]. While in standard laser 3D printing, low-NA focusing systems with high focal depth (50-100 µm) are used, in the system sketched in Figure 1a the focusing conditions allow the photopolymerization to occur only in a volume close to the focal spot, providing polymerized layers with thickness of a few micrometers. The focal spot can be moved in 3D by shifting the objective position with respect to the sample along the optical axis, and by 2D-scanning of the laser spot in the sample plane (i.e. the plane perpendicular to the optical axis) by a galvo-mirror ( Figure 1a). 3D objects can be therefore manufactured by these two independent actuation systems. The laser on and off switching can be controlled with microseconds time resolution through an electro-optic shutter. The 3D structures here realized are printed with a layer thickness of 2 µm, much lower than values (10-100 µm) used in standard UV stereolithography systems [16]. After printing, the uncured pre-polymer is rinsed by isopropanol. The maximum area that can be printed is 1.2×1.2 mm 2 .
The monitoring of the photopolymerization process is performed in-situ by collecting the photons of the 405 nm laser (the same used for photopolymerization), that are backscattered by the sample. More specifically, the change of the refractive index of the pre-polymer during the photopolymerization Published in Additive Manufacturing, doi: 10.1016/j.addma.2022.103020 (2022). 7 process can lead to a variation of the intensity of both the back-reflected light and the light that is backscattered within the collection angle of the focusing objective [37,38]. Therefore, the intensity (IBS) of the UV laser beam, focused in the pre-polymer and back-scattered by the polymerized area, is indicative of the local evolution of polymerization. This approach is similar to refractometry experiments [31], with the main advantage of using the same light source for both curing and process sensing.
exploited in order to decrease the depth and diameter of the sample that is probed, and by which backscattered light is collected. This system allows the photopolymerization kinetics occurring in a volume nearby the focal spot of the 405 nm laser to be monitored. The intensity of the 405 nm laser backscattered by the sample is measured with a temporal resolution of a few tens of ms.

Optical properties
Micro-Raman measurements are carried out with a Renishaw InVia spectrometer equipped with a confocal optical microscope, by using a 532 nm laser as excitation source and a 50× long working , where n is the refractive index of the sample. These measurements are performed with the BisEMA samples, for which, n=1.49 [42].

Experimental results
Figure 1b shows a 3D structure printed by BisEMA, and imaged by SEM. No footprint of the printed layers can be appreciated in the realized object. Such layered footprint, possibly leading to a step-like surface structure [43], could constitute a major flaw, especially when 3D printing is used to realize optical components. Various approaches have been developed to overcome such drawback, such as post-processing by mechanical polishing [44], depositing and curing additional layers [44][45][46], using micro-continuous liquid interface production [47], or optimized polymer formulations [48]. Here we use a highly focused laser, and very thin printing layers (2 µm), leading to surfaces of the printed structures with submicrometric roughness. This roughness is estimated through the height of the features protruding from the surfaces of the printed structures (<0.5 µm, Figure 1d). Flat windows (area  1 mm and thickness in the range 400-450 µm) printed in this way display optical transmittance higher than 90% in the whole visible range (Figure 1c), corresponding to an attenuation loss coefficient of the order of 1 cm -1 . While a low layer thickness is beneficial for the surface quality and the optical properties of the printed structures, it might raise issues related to the photopolymerization rate. Both the presence of many interfaces between adjacent layers and the favored oxygen diffusion are expected to be detrimental for the free-radical photopolymerization process of acrylate-based prepolymers. To investigate such aspects, we exploit the system schematized in Figure 1a for in-situ, 11 of E-shell ® 600 occurs on timescales of the order of seconds and with much lower irradiation intensity with respect to BisEMA, because of the more effective free-radical generation of phosphine oxide photoinitiators at the 405 nm curing light [51]. Raman spectra show peaks at about 1608 cm -1 and 1638 cm -1 , which are due to C=C aromatic ring stretching and to aliphatic C=C stretching, respectively [39] (Figure 2c and Supplementary Figure 4). Upon increasing the UV exposure time, the intensity of the peak at 1638 cm -1 (I1638) decreases while that of the peak at 1608 cm -1 (I1608) remains almost constant, as expected due to the photopolymerization [39,40]. The conversion factor, CR, reaches a plateau after 7 s of UV exposure (Supplementary Figure 4).
Interestingly, by measuring the spatially-resolved confocal Raman spectra at various distances, h, from the substrate (scheme in the inset of Figure 2d) and after 9 s of UV exposure, a monotonic dependence of CR on h is found as shown in Figure 2c Such findings can be explained by considering diffusion of oxygen, which is an efficient inhibitor of free-radical polymerization of acrylate-based pre-polymers [23]. This is in agreement with experiments that exploit such so-called dead layers, where polymerization is completely inhibited by oxygen, to print 3D objects in continuous runs [52]. Such effects have been also used for printing 3D micro-objects by continuous flow lithography [53]. In such works, the typical thickness of dead layers is from  1 m to a few tens of m, similar to the one measured here. Our methods provide in-situ, quantitative measurements of the layers where free-radical polymerization is inhibited, which can be highly useful for optimizing printing processes relying on a precise reaction control.
The second geometry investigated is constituted by thin films, which we made by SU-8, an epoxy resin with high mechanical stability, largely used in additive manufacturing [54][55][56][57]. In SU-8, the absorption of UV light generates reactive species, and cationic polymerization typically occurs by thermal treatment at T>Tg, where Tg is the glass transition temperature (Tg~50 °C). The intensity of the backscattered signals upon curing and the corresponding tpol measured for SU-8 films with various thicknesses are shown in Figure 3. respectively. Therefore, for thicker films the local temperature can be > 80 °C (i.e. >Tg), and polymerization occurs in less than a minute as observed in Figure 3a, whereas for thinner films the local temperature is lower, thus requiring the exposure times to be increased by about a factor 2 for polymerization to occur.

Modeling of the photopolymerization process
In order to describe the size effects, we develop a phenomenological model of the photopolymerization process. First, the dependence of the probability of the activation of reactive species on the optical path length determining the overall light-polymer interaction time should be taken into account. Indeed, in thin pre-polymer layers this interaction time is short, therefore the probability of activation of reactive species is relatively low, whereas in thick films this probability is much higher. An analogous behavior has been described for polymerization activated by photothermal effects, the increase of temperature is higher for thicker samples, as discussed above.
Taking these considerations into account and following previous studies of the polymerization reactions in bulk materials similar to the one here investigated [24] the photopolymerization process is described on the basis of a simple kinetic equation: where c is the polymerization degree, t is time, k is an effective reaction constant, for which a dependence on c and on the thickness, F d , of the film has been introduced phenomenologically, in order to account for the various effects here observed that might affect the photopolymerization kinetics: In Equation (2) namely it shows an effective reaction constant not depending on c, and the process, which corresponds to the first order reaction, is described by a standard exponential kinetics: The solution of Equation (3) gets the form: Overall, the above expressions are capable of accounting for a dependence of the rate of polymerization on the thickness of the reacting film at the initial stage of the process, while such dependence is smoothed as the thickness increases. With the polymerization reaction flowing, the thickness dependence of the rate of polymerization becomes weaker, and at the final stage of the process it will be almost negligible. The cross-over from the initial (size dependent) regime to the final one occurs at the time: , after which the polymerization reaction will occur according to standard kinetics described by an exponential decay.

Discussion
The phenomenological model introduced above, allows the overall kinetics of photopolymerization to be calculated by using Equation (3), varying the characteristic size parameter of the cured material.
The results for different film thicknesses are shown in the Figure 4a.
The comparison of the curves of Figure 4a with the experimental data (Figure 3a), evidence some differences at the initial stage of the process. More specifically, while the curves of Figure 4a increase monotonically with time, the experimental curves show a non-monotonical behaviour.
The temporal evolution of I(t) for various thickness values is shown in Figure 4b. Notably, the calculated curves display a trend of non-monotonic kinetics that well reproduces the measured one Published in Additive Manufacturing, doi: 10.1016/j.addma.2022.103020 (2022). 18 (IBS in Figure 2a and 3a), evidencing the effectiveness of the developed model for describing photopolymerization processes taking into account potential size effects, independently on the specific nature of the underlying curing mechanism.

Conclusions
The kinetics of photopolymerization reactions are affected by the restricted volume of pre-polymer layers. An in-situ method for process monitoring has been developed based on real-time measurements of the light backscattered by a photo-exposed region. An increase of the time needed for the polymerization by UV light is found as the thickness of the film is decreased below about 1.5 or 10 µm, for materials relying on cationic/photothermal and on free-radical polymerization, respectively. A phenomenological model for photopolymerization has been developed, which is capable of reproducing the trend measured by in-situ monitoring. These results show that while a decrease of the layer thickness is effective for printing 3D objects with improved surface quality, the associated increase of the curing time demands a proper choice of compensating polymerization parameters. Strategies might include the increase of photoinitiator concentration in used materials as well as the increase of light intensity during printing [23,52,53]. In this framework, our study provides an effective experimental tool and theoretical framework for engineering 3D printing based on UV photopolymerization, for improving the achievable spatial resolution, and for enhancing additive manufacturing with composite and heterogeneous materials. inlet To 3D robotic movement system Syringe Quartz substrate