Crosslinking Behavior of UV-Cured Polyorganosilazane as Polymer-Derived Ceramic Precursor in Ambient and Nitrogen Atmosphere

Polymer-derived ceramics (PDCs) based on silicon precursor represent an outstanding material for ceramic coatings thanks to their extraordinary versatile processibility. A promising example of a silicone precursor, polyorganosilazane (Durazane 1800), was studied concerning its crosslinking behavior by mixing it with three different photoinitiators, and curing it by two different UV-LED sources under both nitrogen and ambient atmosphere. The chemical conversion during polymerization and pyrolysis was monitored by FTIR spectroscopy. Pyrolysis was performed in a nitrogen atmosphere at 950 °C. The results demonstrate that polyorganosilazane can be cured by the energy-efficient UV-LED source at room temperature in nitrogen and ambient atmosphere. In nitrogen atmosphere, already common reactions for polysilazanes, including polyaddition of the vinyl group, dehydrogenation reactions, hydrosilylation, and transamination reaction, are responsible for crosslinking. Meanwhile, in ambient atmosphere, hydrolysis and polycondensation reactions occur next to the aforementioned reactions. In addition, the type of photoinitiator has an influence on the conversion of the reactive bonds and the chemical composition of the resulting ceramic. Furthermore, thermogravimetric analysis (TGA) was conducted in order to measure the ceramic yield of the cured samples as well as to study their decomposition. The ceramic yield was observed in the range of 72 to 78% depending on the composition and the curing atmosphere. The curing atmosphere significantly impacts the chemical composition of the resulting ceramics. Depending on the chosen atmosphere, either silicon carbonitride (SiCN) or a partially oxidized SiCN(O) can be produced.


Introduction
In recent years, polymer-derived ceramics (PDCs) based on silicon precursor such as polyorganosilazane have been used as precursors for ceramics [1], as they offer many advantageous properties, in particular excellent adhesion on numerous surfaces [2,3] as well as high thermal and chemical stability [4]. In addition, polysilazane serves as a component for the preparation of ceramic matrix composites produced by means of additive manufacturing [5,6]. Polyorganosilazane consists of an alternating silicon and nitrogen backbone and is produced on a large scale via ammonolysis of dichlorosilane (R 2 SiCl 2 ) [7,8]. Like polymers, PDCs benefit from the extraordinary versatility of processing when compared with bulk ceramics. After processing, it is necessary to perform crosslinking followed by pyrolysis, leading to the desired ceramic. The crosslinking can be conducted thermally by adding a free radical initiator such as peroxides [9][10][11], catalysts [12], or photochemically [13][14][15], to form an infusible network. Another well-established method for the

Preparation and Characterization
In a first series of experiments, the preceramic polymer Durazane 1800 was mixed with three different photoinitiators to obtain photosensitive mixtures (see Table 1) and to investigate the polymerization behavior.

Preparation and Characterization
In a first series of experiments, the preceramic polymer Durazane 1800 was mixed with three different photoinitiators to obtain photosensitive mixtures (see Table 1) and to investigate the polymerization behavior.
Each photoinitiator (1 wt% or 3 wt%) was dissolved in Durazane 1800 (100 wt%) using a high shear disperser (T-10 basic Ultra-Turrax®, IKA, Germany) for 120 s at 9400 rpm. After the dissolution of the photoinitiator in Durazane 1800, the samples were irradiated for 300 s with LED light sources (LED-Spot-100 lamp, Dr. Hönle UV Technology, München, Germany) with different wavelengths (385 and 405 nm) in ambient (47% RH) or nitrogen atmosphere (1.1% RH). The relative humidity (RH) inside the UV chamber was measured by testo 608-H2 thermo hygrometer (Testo SE & Co. KGaA, Titisee-Neustadt, Germany). The intensity I 0 of the UV lamps (λ = 385 or 405 nm) was 569 mW/cm 2 and 553 mW/cm 2 , Table 1. Overview of the sample description with respect to the used UV lamp and atmosphere as parameters for the polymerization process. All mixtures contain 100 wt% Durazane 1800 and were cured for 300 s.

Sample
Initiator c (Initiator) The Fourier-transform infrared (FT-IR) spectrometer was conducted to investigate the crosslinking behavior of the material. The spectra of all samples were recorded within a wavenumber range of 4000 and 500 cm −1 by averaging 32 scans at a resolution of 4 cm −1 using a Bio-Rad FTS 3000 Excalibur spectrometer (Varian, Palo Alto, CA, USA). The samples were coated with a doctor blade onto a shiny etched silicon wafer (~600 µm) and measured both uncured and UV-cured. The thickness of the samples on the silicon wafer was about 2 µm. A background measurement of a blank silicon wafer was taken first and substracted from each measured spectra.
The pyrolysis was performed in an alumina tube furnace (Carbolite, Neuhausen, Germany) at 950 • C, because the wafer is not resistant to higher temperatures. The heating rate was set at 1 • C/min and the temperature was held for 1 h at 950 • C using a nitrogen flow rate of 120 mL/min. Because of the different coefficient of thermal expansion of silicon compared with the resulting amorphous SiCN(O), the pyrolysis temperature was set to 950 • C to avoid cracks in the layer on the wafer.
The absorption spectra of the photoinitiators were measured by a UV-VIS spectrometer (Cary 500 Bio, Varian, Palo Alto, CA, USA) within a wavelength range of 450 and 220 nm. The photoinitiators were diluted in 2-propanol, filled into a quarz cuvette, and measured by dual-beam mode.
Thermogravimetric analysis (TGA) was performed using STA-409C (Netzsch Group GmbH &Co, Selb, Germany) to evaluate the polymer-to-ceramic conversion behavior and to examine the residual ceramic yield of all samples. Around 300 µL of each mixture was placed on a specimen and then the cured sample was scraped from the specimen. Around 20 mg of each cured sample was heated up to 1200 • C using a heating rate of 10 • C/min and a nitrogen flow rate of 100 mL/min. The experimental uncertainty for the residual mass measured by TGA is around ±1%.

Degree of Conversion of the Reactive Bonds in Durazane 1800
FTIR spectroscopy was used to determine the relative degree of conversion DC (%) of Durazane 1800 and the photoinitiator after curing.
The degree of conversion was calculated using Equation (1) [32] for the chemical groups, which undergo chemical reactions during the crosslinking process: ≡Si-H, N-H, and the vinyl group.
Thereby, the content of the integrated peak areas of the reactive bonds of the cured samples was defined as A t . The content of the reactive bond of the uncured sample, which was defined as A 0 , was taken as 100%. To minimize the influence of deviations in sample thickness and instrument recording, all integrals were normalized by the integral of the Si-CH 3 bond at 1253 cm −1 , as this bond does not change throughout the whole crosslinking process.

Crosslinking Mechanism of Polysilazane
Curing of preceramic polymers, such as Durazane 1800, usually takes place thermally by the addition of a free radical initiator like peroxides or catalysts. In this work, this approach was adopted, replacing the high temperature with the energy-efficient LED source to start the crosslinking reaction. Hence, as indicated before, three different photoinitiators were used for the polymerization of Durazane 1800.
All used photoinitiators are derivates of benzophenone, which are Norrish type II photoinitiators [33], meaning that, when irradiated with UV light, they are excited to the singlet state, which subsequently changes to the triplet state via intersystem crossing (ISC) [34]. The mechanism is schematically illustrated in Figure 3 for the photoinitiator DMABP.  In inert atmosphere, Durazane 1800 can be primarily crosslinked through three different chemical bonds: ≡Si-H, ≡Si-NH-Si≡, and RCH=CH2 (vinyl group). The crosslinking reactions are shown in Figure 4. In inert atmosphere, Durazane 1800 can be primarily crosslinked through three different chemical bonds: ≡Si-H, ≡Si-NH-Si≡, and RCH=CH 2 (vinyl group). The crosslinking reactions are shown in Figure 4. In inert atmosphere, Durazane 1800 can be primarily crosslinked through three different chemical bonds: ≡Si-H, ≡Si-NH-Si≡, and RCH=CH2 (vinyl group). The crosslinking reactions are shown in Figure 4. The main group that undergoes a crosslinking reaction is the vinyl group. It is involved in the radical vinyl polymerization and hydrosilylation of Si-vinyl and ≡Si-H groups. The possible radical reaction of methyl and vinyl groups, which occurs at temperatures above 200 °C [35], can be neglected in this work. Moreover, the ≡Si-H bond undergoes dehydrogenation reactions between two ≡Si-H bonds and/or between ≡Si-H and ≡Si-NH-Si≡ groups. Finally, the ≡Si-NH-Si≡ group can crosslink via transamination reaction [7,9,36].  The main group that undergoes a crosslinking reaction is the vinyl group. It is involved in the radical vinyl polymerization and hydrosilylation of Si-vinyl and ≡Si-H groups. The possible radical reaction of methyl and vinyl groups, which occurs at temperatures above 200 • C [35], can be neglected in this work. Moreover, the ≡Si-H bond undergoes dehydrogenation reactions between two ≡Si-H bonds and/or between ≡Si-H and ≡Si-NH-Si≡ groups. Finally, the ≡Si-NH-Si≡ group can crosslink via transamination reaction [7,9,36].
In ambient atmosphere, the crosslinking reactions are mostly hydrolysis and polycondensation reactions [37]; these are shown in Figure 5. Thereby, the ≡Si-NH-Si≡ group reacts with a water or an oxygen molecule under formation of silanol groups, which subsequently polymerize to polysiloxane via polycondensation. In ambient atmosphere, the crosslinking reactions are mostly hydrolysis and polycondensation reactions [37]; these are shown in Figure 5. Thereby, the ≡Si-NH-Si≡ group reacts with a water or an oxygen molecule under formation of silanol groups, which subsequently polymerize to polysiloxane via polycondensation.

UV-VIS Spectroscopy
The UV-VIS spectra of the photoinitiators were recorded in order to distinguish at which absorption maximum the PIs absorb UV-light for starting the polymerization reaction of polysilazane. Figure 6 illustrates the absorbance spectra of the photoinitiators and Table 2 summarizes the absorption maxima.

UV-VIS Spectroscopy
The UV-VIS spectra of the photoinitiators were recorded in order to distinguish at which absorption maximum the PIs absorb UV-light for starting the polymerization reaction of polysilazane. Figure 6 illustrates the absorbance spectra of the photoinitiators and Table 2 summarizes the absorption maxima.

FTIR Analysis
FTIR analysis was carried out before and after the UV curing to analyze the crosslinking behavior and to characterize the chemical structure of the samples. The relative degree of conversion was calculated for the reactive bonds (≡Si-H, ≡Si-NH-Si≡, and the vinyl group), which are detected by FTIR spectroscopy. The non-reactive group ≡Si-CH3, with a sharp band at about 1256 cm −1 , is a common characteristic of all FTIR spectra of Durazane 1800, as it was synthesized through an ammonolysis reaction of dichloromethylsilanes (RCH3SiCl2, R=H, CH3, or CH=CH2) [7]. The ≡Si-CH3 group is used as a reference for all calculations of the crosslinking process.
The FTIR spectra of DMABP01-DMABP08 are presented in Figure 7a-d. The uncured samples (black curves) show the characteristic bands of pure Durazane 1800, which are mainly summarized in Table A1. The absorption bands below 1000 cm −1 relate to the stretching and deformation vibrations of Si-C, Si-N, C-C, and C-H bonds, which, however are overlapping and consequently cannot be accurately assigned.
In this work, the focus of attention lies on the reactive chemical bonds in the case of the uncured samples, including the vinyl group as well as the ≡Si-H and ≡Si-NH-Si≡ bond ( Table 3). The absorption band at 1050 cm −1 , corresponding to the ≡Si-O-Si≡, is only visible in Figure 7b,d. A possible reason for the ≡Si-O-Si≡ band in the uncured samples is the higher amount of photoinitiator, which might induce a slight silanol reaction during preparation in ambient atmosphere. After using Durazane 1800, the bottle is flooded with nitrogen to prevent oxygen contamination. Nevertheless, the oxygen cannot be completely avoided, as in the case of inert atmosphere.   The photoinitiator shows different absorption maxima, starting with ITX ( Figure 6, green curve), which has two major absorption maxima at 258 nm and 382 nm in the UV-C and UV-A region, respectively, combined with two small shoulders between 290 and 301 nm in the UV-B region. The photoinitiator ITX is a derivate of benzophenone with a modification of a sulfur atom bridge between the two phenyl groups. Owing to the sulfur atom, the UV absorption spectra are redshifted compared with benzophenone [34].
As ITX, DBS ( Figure 6) absorbs in the UV-region, showing the major maximum at 255 nm, followed by an absorption maximum at 307 nm and a small broad shoulder between 340 and 380 nm. Lastly, DMABP ( Figure 6) is the only PI, which does not absorb in the UV-B region, showing its major absorption maximum in the UV-A region at 351 nm and the smaller absorption maximum at 248 nm. In this work, the UV-A region is of interest, as the employed light source has a light emitting diode for the emission of homogenous irradiation at a wavelength of 385 or 405 nm.

FTIR Analysis
FTIR analysis was carried out before and after the UV curing to analyze the crosslinking behavior and to characterize the chemical structure of the samples. The relative degree of conversion was calculated for the reactive bonds (≡Si-H, ≡Si-NH-Si≡, and the vinyl group), which are detected by FTIR spectroscopy. The non-reactive group ≡Si-CH 3 , with a sharp band at about 1256 cm −1 , is a common characteristic of all FTIR spectra of Durazane 1800, as it was synthesized through an ammonolysis reaction of dichloromethylsilanes (RCH 3 SiCl 2 , R=H, CH 3 , or CH=CH 2 ) [7]. The ≡Si-CH 3 group is used as a reference for all calculations of the crosslinking process. The FTIR spectra of DMABP01-DMABP08 are presented in Figure 7a-d. The uncured samples (black curves) show the characteristic bands of pure Durazane 1800, which are mainly summarized in Table A1. The absorption bands below 1000 cm −1 relate to the stretching and deformation vibrations of Si-C, Si-N, C-C, and C-H bonds, which, however are overlapping and consequently cannot be accurately assigned. The FTIR spectra of sample DMABP03/04 and DMABP07/08 were recorded after UV curing by a LED source (λ = 385 and 405 nm) in nitrogen atmosphere. In these conditions, the spectra of the cured samples reveal similar features as the uncured reference samples, except for the decrease of the deformation vibrations of the vinyl group (Figure 7, blue curves) at the wavenumber of 1596 cm −1 . This indicates that the conversion of the vinyl group via radical polymerization is mainly responsible for the curing of Durazane 1800. The observation is supported by the calculation of the degree of conversion (DC, vinyl group) shown in Figure 8 and Table A1, which is three times as high as the DC of the Si-H and N-H bonds. The vinyl groups of samples DMABP03/04 convert almost completely (DC = 89%) using the LED source (λ = 385 nm) compared with the samples DMABP07/08. A reason for this occurence might be the absorption maximum of DMABP, which is at 351 nm (Table 2), being closer to 385 nm than to 405 nm. The DC (Si-H and N-H, Figure 8) of DMABP03/04 is half as high as that of DMABP01/02, owing to the absence of oxygen during the crosslinking process and, accordingly, little to no poly-condensation reactions of the Si-H and N-H groups occurred in inert atmosphere. The conversion of these bonds in inert atmosphere is based on transamination (N-H bond) and dehydrogenation (Si-H) reactions (see Figure 4). By comparing the spectra of DMABP03/04 (blue curves) with DMABP01/02 (red curves), the samples cured in an inert atmosphere exhibit In this work, the focus of attention lies on the reactive chemical bonds in the case of the uncured samples, including the vinyl group as well as the ≡Si-H and ≡Si-NH-Si≡ bond ( Table 3). The absorption band at 1050 cm −1 , corresponding to the ≡Si-O-Si≡, is only visible in Figure 7b,d. A possible reason for the ≡Si-O-Si≡ band in the uncured samples is the higher amount of photoinitiator, which might induce a slight silanol reaction during preparation in ambient atmosphere. After using Durazane 1800, the bottle is flooded with nitrogen to prevent oxygen contamination. Nevertheless, the oxygen cannot be completely avoided, as in the case of inert atmosphere.
The FTIR spectra of sample DMABP03/04 and DMABP07/08 were recorded after UV curing by a LED source (λ = 385 and 405 nm) in nitrogen atmosphere. In these conditions, the spectra of the cured samples reveal similar features as the uncured reference samples, except for the decrease of the deformation vibrations of the vinyl group (Figure 7, blue curves) at the wavenumber of 1596 cm −1 . This indicates that the conversion of the vinyl group via radical polymerization is mainly responsible for the curing of Durazane 1800. The observation is supported by the calculation of the degree of conversion (DC, vinyl group) shown in Figure 8 and Table A1, which is three times as high as the DC of the Si-H and N-H bonds. The vinyl groups of samples DMABP03/04 convert almost completely (DC = 89%) using the LED source (λ = 385 nm) compared with the samples DMABP07/08. A reason for this occurence might be the absorption maximum of DMABP, which is at 351 nm (Table 2), being closer to 385 nm than to 405 nm. The DC (Si-H and N-H, Figure 8) of DMABP03/04 is half as high as that of DMABP01/02, owing to the absence of oxygen during the crosslinking process and, accordingly, little to no poly-condensation reactions of the Si-H and N-H groups occurred in inert atmosphere. The conversion of these bonds in inert atmosphere is based on transamination (N-H bond) and dehydrogenation (Si-H) reactions (see Figure 4). By comparing the spectra of DMABP03/04 (blue curves) with DMABP01/02 (red curves), the samples cured in an inert atmosphere exhibit a minor absorption band of ≡Si-O-Si≡ between 1080 and 1040 cm −1 , which could be completely avoided when operating in a glovebox. Moreover, the photoinitiatior DMABP is resistant against oxygen inhibition, as the aminoalkyl radical can form a peroxide radical by reacting with oxygen. In turn the peroxide radical generates another aminoalkyl radical by hydrogen abstraction [34]. This property of DMABP can be very useful when operating with material suffering from oxygen inhibition. Hence, it can be concluded that successful curing of Durazane 1800 took place using a low energetic LED source within minutes. ating with material suffering from oxygen inhibition. Hence, it can be concluded that successful curing of Durazane 1800 took place using a low energetic LED source within minutes.  The FTIR spectra of DMABP01/02 and DMABP05/06 were measured after UV curing by an LED source (λ = 385 and 405 nm), respectively, in ambient atmosphere. For all samples cured in ambient atmosphere, the DC of the ≡Si-NH-Si≡ group was neglected, because the peak could not be distinguished properly owing to the peak broadening caused by water. The intensities of the bands assigned to the stretching of the C-H bonds of the vinyl group (at 3050 cm −1 ) as well as of the C=C double bonds at 1596 cm −1 disappeared for the samples DMABP01/02 and decreased for the samples DMABP05/06. Besides, the absorption band of the ≡Si-H bond decreased for all these samples. This observation indicates a hydrosilylation reaction (Figure 4, (4)) of the ≡Si-H group with the vinyl group and radical polymerization of the vinyl group (Figure 4, (5)). The disappearance of the vinyl group of the samples DMABP01/02 may be due to the absorption maximum of the photoinitiator around 355 nm ( Figure 6 and Table 2), thus corresponding more closely to the wavelength at 385 nm than to the 405 nm UV-lamp. This result is confirmed by the DC of the vinyl group, which is higher for DMABP01/02 than for DMABP05. However, it is in the same order of magnitude when comparing the samples DMABP01/02 and DMABP06, considering the measurement inaccuracy.
Moreover, the spectra of DMABP01, DMABP02, and DMAPB06 (Figure 7a, Figure 7b, and Figure 7d, respectively, red curves) show the absorption band of ≡Si-O-Si≡. The reason is the hydrolysis of the ≡Si-NH-Si≡ group resulting in a silanol group (≡Si-OH) and ammonia, followed by a polycondensation reaction ( Figure 5). This crosslinking reaction is confirmed by nearly complete disappearance of both the stretching and deformation vibrations of the Si-NH band. The absorption band of the Si-O-Si of sample DMABP05 (Figure 7c) is not as strong in the case of DMABP01/02 and DMABP06, despite being polymerized similarly. In addition, neither the stretching nor the deformation vibrations of the ≡Si-NH-Si≡ group fully disappeared in this sample, despite curing in ambient atmosphere. One reason for this observation could be the low concentration of the photoinitiator compared with sample DMABP06, and another reason could be the lower energetic irradiation of the LED source (λ = 405 nm). Moreover, the DC of the Si-H bond of samples DMABP01/02 is higher than that of samples DMABP05/06, implying more silanol and polycondensation reactions of the bonds ≡Si-H and N-H occurred at the wavelength 385 nm than at 405 nm. The hydrolysis of Si-H bond can be catalyzed by the ammonia formed from hydrolysis reactions of the ≡Si-NH-Si≡ group [39]. Especially in sample DMABP05, the DC (≡Si-H) is much lower than in sample DMABP06, indicating less hydrolysis and polycondensation reactions, as proven in the spectra (Figure 7c) by the unreacted absorption peaks of the ≡Si-NH-Si≡ bond. This result shows that Durazane 1800 does not completely hydrolyse during curing, even in these conditions. Thermally, this could not be observed, as more energy is given into the system. However, this absorption band occurs for all samples cured in ambient atmosphere owing to the sensitivity of Durazane 1800 to moisture [37,40,41].
Compared with DMABP and ITX, the samples mixed with DBS could not be cured by near visible light LED (λ = 405 nm), owing to its absorbance maxima ( Figure 6) and the structure of DBS ( Figure 2). As it only consists of phenyl-groups without heteroatom like sulfur or the amino alkyl group, the absorbance spectrum is not red-shifted. The photoinitiator concentration does not affect the DC (vinyl group, ≡Si-H, and N-H) for all samples. For the coating application, it is advantageous to use a higher amount of photoinitiator if thinner layers are desired [34].
The samples DBS01-04 were cured by the LED source (λ = 385 nm) in ambient and nitrogen atmosphere followed by the measurement of the FTIR spectra, which are illustrated in Figure 9a,b.
Samples DBS03/04 were cured in nitrogen atmosphere and, as shown in Figure 9, an Si-O-Si absorption band (~1060 cm −1 ) was formed, implying that the samples were partially hydrolyzed. As shown in Figure 9b, the spectra of the uncured mixture already show an Si-O-Si band, indicating the sensitivity of the Durazane 1800/DBS system to oxygen. Nevertheless, the crosslinking behavior is supported by radical polymerization of the vinyl group (DC up to 85%), as the absorbance band of the vinyl group has decreased. Compared with DMABP03/04, the DC (N-H) is higher for DBS03/04 because the N-H group works as a co-initiator in the initiation reaction of DBS. Additionally, transamination and dehydrogenation reactions with the Si-H bond (DC~40%) may have occurred.
itiator concentration does not affect the DC (vinyl group, ≡Si-H, and N-H) for all samples. For the coating application, it is advantageous to use a higher amount of photoinitiator if thinner layers are desired [34].
The samples DBS01-04 were cured by the LED source (λ = 385 nm) in ambient and nitrogen atmosphere followed by the measurement of the FTIR spectra, which are illustrated in Figure 9a,b. Samples DBS03/04 were cured in nitrogen atmosphere and, as shown in Figure 9, an Si-O-Si absorption band (~1060 cm −1 ) was formed, implying that the samples were partially hydrolyzed. As shown in Figure 9b, the spectra of the uncured mixture already show an Si-O-Si band, indicating the sensitivity of the Durazane 1800/DBS system to oxygen. Nevertheless, the crosslinking behavior is supported by radical polymerization of the vinyl group (DC up to 85%), as the absorbance band of the vinyl group has decreased. Compared with DMABP03/04, the DC (N-H) is higher for DBS03/04 because the N-H group works as a co-initiator in the initiation reaction of DBS. Additionally, transamination and dehydrogenation reactions with the Si-H bond (DC ~ 40%) may have occurred.
Similar to DMABP01/02, samples DBS01/02 cured in ambient atmosphere (red curves) also undergo hydrolysis and polycondensation reactions illustrated in Figure 9 by the formation of an Si-O-Si absorption band. Moreover, the vinyl group is converted as Similar to DMABP01/02, samples DBS01/02 cured in ambient atmosphere (red curves) also undergo hydrolysis and polycondensation reactions illustrated in Figure 9 by the formation of an Si-O-Si absorption band. Moreover, the vinyl group is converted as the absorption band decreases and the resulting DC (vinyl group) accounts for approximately 75% (Figure 10). The vinyl group possibly reacts with the Si-H bond (DC~60%) by hydrosilylation reaction, forming an Si-C linkage, or via radical polymerization.  By mixing Durazane 1800 with the photoinitiator ITX, both LED sources could be used for curing the mixtures owing to the red shifting of ITX generated by the sulfur atom. The FTIR spectra and the resulting degree of conversion of the reactive bonds are shown in Figure 11 and Figure 12, respectively.
The initial aspect that is noticeable when analyzing the spectra of the samples (ITX03/04 and ITX07/08) polymerized in nitrogen atmosphere is the formation of the ≡Si-O-Si≡ bond in spite of the inert atmosphere. The ≡Si-O-Si≡ bond is more evident for samples ITX03/07 than for ITX04/08. This result is represented by the corresponding DC (N-H), which is larger for ITX03/07 than for ITX04/08. One reason for this occurence might be the lower photoinitiator concentration, leading to a slower, nevertheless more homogenous, initiation of the photoinitiator, and thus to a slower gelation. However, as the N-H By mixing Durazane 1800 with the photoinitiator ITX, both LED sources could be used for curing the mixtures owing to the red shifting of ITX generated by the sulfur atom. The FTIR spectra and the resulting degree of conversion of the reactive bonds are shown in Figures 11 and 12, respectively.  The FTIR spectra of the samples pyrolyzed at 950 °C are shown in Figure 13 of the   ITX01  ITX02  ITX03  ITX04  ITX05  ITX06  ITX07   The FTIR spectra of the samples pyrolyzed at 950 °C are shown in Figure 13 of the samples cured in nitrogen and ambient atmosphere. The ceramization process was com- ITX01  ITX02  ITX03  ITX04  ITX05  ITX06  ITX07  The initial aspect that is noticeable when analyzing the spectra of the samples (ITX03/04 and ITX07/08) polymerized in nitrogen atmosphere is the formation of the ≡Si-O-Si≡ bond in spite of the inert atmosphere. The ≡Si-O-Si≡ bond is more evident for samples ITX03/07 than for ITX04/08. This result is represented by the corresponding DC (N-H), which is larger for ITX03/07 than for ITX04/08. One reason for this occurence might be the lower photoinitiator concentration, leading to a slower, nevertheless more homogenous, initiation of the photoinitiator, and thus to a slower gelation. However, as the N-H group co-initiates the photoinitiator, the conversion is higher with the lower concentrations of the initiator than at higher concentrations, because more radical centers could be activated owing to the slower reaction. The DC (vinyl group) of ITX04 is higher than ITX03, which is desirable, as the Si-C linkage observed by hydrosilylation reaction is stable at elevated temperatures.
As can be seen in the spectra (Figure 11), the samples (ITX01/02, ITX05/06) cured in ambient atmosphere were hydrolyzed similarly to the aforementioned mixtures of Durazane 1800 with DBS and DMABP. The calculated DC of the vinyl and the Si-H bonds are in the same order of magnitude for ITX01/02 and ITX05/06 and are to be attributed to hydrosilylation reactions and radical polymerization of the vinyl group.
The FTIR spectra of the samples pyrolyzed at 950 • C are shown in Figure 13 of the samples cured in nitrogen and ambient atmosphere. The ceramization process was completed as no absorption of Si-H or C-H bonds is present. A broad absorption band is visible between 1100 and 680 cm −1 attributed to Si-C, Si-N-Si, and Si-O-Si bonds. As previously mentioned, the presence of oxygen refers to the sensitivity of Durazane 1800 to moisture. Even if the crosslinking and pyrolysis processes took place in nitrogen atmosphere, the mixtures were prepared in ambient atmosphere. As expected, the Si-O-Si bond is more evident for DBS04 and ITX04 compared with DMABP04, as the spectra of the cured samples already showed the Si-O-Si absorption band.

Thermogravimetric Analysis and Ceramic Yields
Thermogravimetric analysis was conducted under nitrogen flow with 10 °C/min to study the decomposition of the preceramic polymer and to measure the ceramic yield of the sample. The ceramic yield of all samples was taken of the residual mass at 1200 °C and is plotted for each mixture in Figure 8, Figure 10, and Figure 12. The thermograms measured in nitrogen atmosphere are shown in Figure A1 and Figure A2 for the samples cured by the LED source (λ = 385 and 405 nm), respectively. Thermal behavior in nitrogen atmosphere will be explained using the sample DMABP04 as an exemplary sample in comparison with uncured Durazana 1800, as the cured samples exhibit very similar thermal

Thermogravimetric Analysis and Ceramic Yields
Thermogravimetric analysis was conducted under nitrogen flow with 10 • C/min to study the decomposition of the preceramic polymer and to measure the ceramic yield of the sample. The ceramic yield of all samples was taken of the residual mass at 1200 • C and is plotted for each mixture in Figures 8, 10 and 12. The thermograms measured in nitrogen atmosphere are shown in Figures A1 and A2 for the samples cured by the LED source (λ = 385 and 405 nm), respectively. Thermal behavior in nitrogen atmosphere will be explained using the sample DMABP04 as an exemplary sample in comparison with uncured Durazana 1800, as the cured samples exhibit very similar thermal behavior. Sample DMABP04 was cured by LED source (λ = 385 nm). The thermograms of uncured Durazane 1800 measured in ambient and nitrogen atmosphere and of DMABP04 are presented in Figure 14. Pure and uncured Durazane 1800, which was measured both in nitrogen and ambient atmosphere, decomposes in a three-step process. The degradation starts below 100 °C with a mass loss of 16 and 10%, respectively, because of the non-crosslinked volatile oligomers. It is worth mentioning that the first step is interrupted by a small plateau between 200 and 300 °C, which is attributable to the free-radical polymerization of the vinyl groups. This plateau is missing in the UV-cured samples because the vinyl group was already radically converted. The second step begins between 300 and 530 °C and results in a mass loss of about 10 and 7%, respectively, owing to dehydrogenation and transamination reactions. During the third step, which starts between 530 and 850 °C, the polymerto-ceramic transformation takes place as described before and results in a mass loss of 10 and 4%, respectively. The ceramic yield of pure Durazane 1800, measured in nitrogen and ambient atmosphere, is 63 and 80%, respectively. The main reason for the increase of ceramic yield is the incorporation of oxygen through hydrolysis and polycondensation reactions of Durazane 1800 [40,42].
The degradation of the cured samples proceeds in a two-step process. The first step starts at temperatures between 120 and 400 °C. During this step, further dehydrogenation and transamination reactions occur, leading to mass loss due to volatile groups like ammonia and hydrogen. Furthermore, at temperatures higher than 100 °C, the non-crosslinked volatile oligomers degrade, resulting in mass loss. On the other hand, at temperatures higher than 200 °C, the residual vinyl groups undergo further radical polymerization and form a thermoset. Therefore, the curve between 200 and 400 °C is more flat than sharp. The second step starts between 500 and 750 °C, resulting in a mass loss due to the organic-inorganic transformation of the thermoset into amorphous SiCN and SiCO, depending on the curing atmosphere. The organic substituents degrade into methane and other volatile hydrocarbons, hydrogen, ammonia, and volatile silicon derivatives [12], mainly owing to rearrangements and radical reactions leading to bond breaking and new bond formation [43]. No further mass loss is observed at temperatures higher than 800 °C. Pure and uncured Durazane 1800, which was measured both in nitrogen and ambient atmosphere, decomposes in a three-step process. The degradation starts below 100 • C with a mass loss of 16 and 10%, respectively, because of the non-crosslinked volatile oligomers. It is worth mentioning that the first step is interrupted by a small plateau between 200 and 300 • C, which is attributable to the free-radical polymerization of the vinyl groups. This plateau is missing in the UV-cured samples because the vinyl group was already radically converted. The second step begins between 300 and 530 • C and results in a mass loss of about 10 and 7%, respectively, owing to dehydrogenation and transamination reactions. During the third step, which starts between 530 and 850 • C, the polymer-toceramic transformation takes place as described before and results in a mass loss of 10 and 4%, respectively. The ceramic yield of pure Durazane 1800, measured in nitrogen and ambient atmosphere, is 63 and 80%, respectively. The main reason for the increase of ceramic yield is the incorporation of oxygen through hydrolysis and polycondensation reactions of Durazane 1800 [40,42].
The degradation of the cured samples proceeds in a two-step process. The first step starts at temperatures between 120 and 400 • C. During this step, further dehydrogenation and transamination reactions occur, leading to mass loss due to volatile groups like ammonia and hydrogen. Furthermore, at temperatures higher than 100 • C, the non-crosslinked volatile oligomers degrade, resulting in mass loss. On the other hand, at temperatures higher than 200 • C, the residual vinyl groups undergo further radical polymerization and form a thermoset. Therefore, the curve between 200 and 400 • C is more flat than sharp. The second step starts between 500 and 750 • C, resulting in a mass loss due to the organic-inorganic transformation of the thermoset into amorphous SiCN and SiCO, depending on the curing atmosphere. The organic substituents degrade into methane and other volatile hydrocarbons, hydrogen, ammonia, and volatile silicon derivatives [12], mainly owing to rearrangements and radical reactions leading to bond breaking and new bond formation [43]. No further mass loss is observed at temperatures higher than 800 • C.
In comparison with the uncured Durazane 1800 (63% in N 2 atmosphere), the ceramic yield of the samples crosslinked via LED source (λ = 385 and 405 nm) increased significantly up to the range of between 72 and 78% depending on the mixture and crosslinking atmosphere (see Appendix A, Table A1). Crosslinking of the preceramic polymers is necessary for the polymer-to-ceramic transformation, as it increases the ceramic yield by reducing the volatilization of the oligomers [44]. The high ceramic yield confirmed the high degree of conversion. In addition, it demonstrates the response of Durazane 1800 to the chosen crosslinking approach. For increasing the ceramic yield and reducing the shrinkage of the final ceramic, crosslinking is an indispensable process, as the uncured sample is at least 10% lower than the crosslinked samples in inert atmosphere.
In this work, different parameters were changed to examine the behavior of Durazane 1800. First, the influence of the polymerization atmosphere on the ceramic yield was studied. As summarized in Table A1, the ceramic yield of DBS01/02 is slightly higher than that of DBS03/04, owing to the already mentioned oxygen incorporation, which is visible in the FTIR spectra ( Figure 9). A higher ceramic yield was also observed for DMABP02 in comparison with DMABP03/04. For all the other samples, the polymerization atmosphere did not affect the ceramic yield. Furthermore, the photoinitiator concentration was varied for the purpose of examining the influence on the ceramic yield. The higher photoinitiator concentration does not affect the ceramic yield for most samples, except for sample ITX08, which has a higher ceramic yield than ITX07. Finally, two UV-lamps with different wavelengths were used for the photoinitiators ITX and DMABP. The change of the LED source did not influence the ceramic yield, because the ceramic yield depends on the conversion of the reactive bonds. As previously described, the mixtures presented in Table 1 were cured properly by the chosen conditions and parameter.

Conclusions
In conclusion, the present work focuses on the crosslinking behavior of polyorganosilazane (Durazane 1800), which was mixed with three different photoinitiatorsdibenzosuberone (DBS), 4-(dimethylamino)benzophenone (DMABP), and isopropylthioxanthone (ITX)-and UV irradiated by two LED sources (λ = 385 and 405 nm) in ambient and nitrogen atmosphere. It can be concluded that successful curing of Durazane 1800 was established within minutes using energy-efficient LED sources. The crosslinking behavior of the mixtures was investigated in detail by FTIR spectroscopy and the degree of conversion was calculated for the reactive bonds. The curing atmosphere has a significant impact on the crosslinking behavior of the reactive bonds and the chemical composition of the resulting ceramic. In nitrogen atmosphere, it turned out that the DMABP mixtures observed the least oxygen incorporation in comparison with the other two photoinitiators. The main reason for this occurence is the aminoalkyl radical, which can form a peroxide radical by reacting with oxygen. The peroxide radical in turn generates another aminoalkyl radical by hydrogen abstraction [34]. The resulting ceramic of the samples cured in nitrogen atmosphere is SiCN(O). The amount of oxygen depends on the curing kinetics of the photoinitiator. If the reactivity of the photoinitiator is slow, the N-H bond will hydrolyse beforehand. Therefore, DMABP is recommended for usage as it shows the best results in terms of insensitivity to hydrolysis reactions.
The approach presented in this work can be applied to various applications, including coatings' or microelectromechanical systems' (MEMS) fabrication. Especially in the field of MEMS, the preparation of precise structures is required, so it is of great interest if a targeted curing can be carried out by means of a mask, which would not be possible with thermal curing. With the help of this approach, UV-curable inks for inkjet printing can be prepared or even 3D structures can be fabricated by stereolithography.
In ambient atmosphere, curing of all samples is given by partial hydrolysis and polycondensation reactions owing to the sensistivity of poly(organo)silazane to water. Therefore, the resulting ceramic will be SiCO(N).
As only one spectroscopic method (FTIR) was applied in this work, it is of great importance that other methods are used to better understand and compare the presented system. FTIR spectroscopy is a fast and easy to use method. However, one disadvantage of the technique is the overlapping of the bands, which makes the method imprecise. Therefore, the use of another method such as NMR spectroscopy is indispensable, as it would identify individual bonds quite accurately.
By comparing the two used LED sources (λ = 385 and 405 nm) of the DMABP samples, it was found that the samples cured by the LED source (λ = 385 nm) reached the higher degree of conversion of the reactive bonds, because this wavelength corresponds more closely to the absorption maximum of the photoinitiator. When increasing the photoinitiator concentration, a high radical concentration is available near the surface, resulting in sufficient surface cure. Accordingly, the thickness of the sample needs to be reduced to achieve complete curing as most of the light is absorbed on the surface, leading to a top to bottom amount of initiated species [45]. A lower photoinitiator concentration causes a homogeneous radical distribution, leading to good through-curing, but poor surface cure [34]. The photoinitiator concentration should be adjusted for each application depending on the film thickness desired. Moreover, the higher amount of photoinitiator results in a faster gelation owing to the higher radical concentration.
The ceramic yield was obtained by using the residual mass of the TGA and was observed in the range of 72 to 78% depending on the composition and crosslinking behavior. Hence, the curing atmosphere only influenced the ceramic yield of the DBS mixtures. In addition, the two different LED sources and the photoinitiator concentration do not affect the ceramic yield. Thus, the ceramic yield observed by the approach presented is high enough to form dense amorphous ceramics.