The 28th Special CU-af Seminar 2020

The 28th Special CU-af Seminar 2020 October 20, 2020 Figure 4 FT-IR spectra of unmodified and modified SiO2. The elemental analysis was also introduced to confirm the amino-organoalkoxyl modification on SiO2 surface as exhibited in Table 2. The C, H, N contents of the unmodified SiO2 particles showed a hydrogen content of only 1.43 wt%, which was from the adsorbed water on the unmodified SiO2 surface. After modification with amine modifiers (1NES-Si, 1NMS-Si, 2NMS-Si or 3NES-Si) the carbon, hydrogen and nitrogen contents tented to be increased. Accordingly, these results supported that the surface of SiO2 particles had been successfully modified by amino-organoalkoxysilane. Moreover, Table 2 summarizes the surface properties of silica particles before and after modification. The BET surface area of unmodified SiO2 was 28.16 m2/g. After modification, the specific surface area was changed due to chemical structure of modifier.The increase of substituted size of silanol group (1NES, 1NMS) and amine content of modifier (1NMS, 2NMS and 3NMS) affected the decrease of surface properties. It may be because the modifier with large amount or steric effect could block the pores of modified silica particles. TGA thermographs of unmodified and modified silica particles are represented in Figure 5. The unmodified SiO2 particles showed only a one step of decomposition (50 - 100 oC) due to the moisture evaporation. While the modified silica particles showed two steps of decomposition. The first step in the range of 50 – 100 oC 93

The 28th Special CU-af Seminar 2020 October 20, 2020 Table 2 CHN analysis of the unmodified and modified SiO2 particles. was the decomposition of moisture evaporation and the second step (200 – 650 oC) indicated the degradation of organic compounds in the modifier. The amount of weight loss was related to the modifier content on the silica surface. Morphological study of unmodified and modified silica particles by SEM is represented in Figure 6. SiO2 showed the spherical particle size in the range of 200-250 nm. After modification, modified silica particles were significantly aggregated or agglomerated due to the increased basicity of each modifier. Accordingly, the modifier was acting as a catalyst to increase the rate of sol-gel reaction, resulting in the aggregates or agglomerates of modified silica particles. Figure 5 TGA thermographs of unmodified and modified silica particles. Figure 6 SEM mocrographs of (a) SiO2 and (b)-(e) modified SiO2 particles. 94

The 28th Special CU-af Seminar 2020 October 20, 2020 Characterization of MNR foam composite The morphology of MNR foam composite filled with unmodified/modified SiO2 particles was investigated by SEM technique as shown in Figure 7. Both NR and MNR foam showed a close cell structure with an oval shape. The cell size of MNR foam was smaller than that of NR foam. After filling with the unmodified SiO2 (SiO2), the morphology of MNR foam composite did not change. The good distribution of SiO2 particles in rubbery matrix was clearly observed at the low silica content (5, 10, 15 phr). By increasing SiO2 content up to 20 phr, the SiO2 particles tended to be aggregated and poor distribution. The similar morphology of MNR foam composite filled with modified silica particles (15 phr) was also observed (Figure 7). However, the modified silica particles tended to be more aggregated than the unmodified SiO2 particles in MNR foam composite material. This result was in agreement with the morphology of modified silica particles in Section 2. Figure 7 SEM micrographs of MNR foam composite 95

The 28th Special CU-af Seminar 2020 October 20, 2020 CO2 adsorption Effect of MNR on CO2 adsorption capacity Figure 8 shows the breakthrough curves (the relative of a ratio of CO2 concentration at the outlet to CO2 concentration at the inlet of reactor versus time) of NR and MNR. The CO2 breakthrough time (the time at which the effluent CO2 concentration reaches 10% allowable breakthrough concentration) of NR and MNR are 67 and 73 s, respectively. This result confirmed that MNR showed the better CO2 adsorption than NR. Based on the area above the breakthrough curves, the corresponding total CO2 adsorption capacity of NR and MNR was calculated to be 0.79 and 1.26 mg/g. Accordingly, the increase of amine groups in MNR structure the increase of CO2 selectivity was obtained, resulting in the increase of CO2 adsorption capacity. Effect of unmodified SiO2 content on CO2 adsorption capacity The effect of unmodified SiO2 content on CO2 adsorption was summarized in Table 3. The breakthrough time and CO2 adsorption capacity of MNR composite increased after filling with SiO2 due to the increased surface area from SiO2 particles to adsorb CO2. Moreover, the CO2 adsorption capacity increased with the increase of SiO2 content and it was reached to the maximal CO2 adsorption capacity at 15 phr SiO2. After filling with SiO2 content more than 20 phr, the decreased breakthrough time and CO2 adsorption capacity tended to be decreased because the poor distribution of SiO2 in the rubbery matrix. Accordingly, the optimum content of modified silica particles was chosen at 15 phr for further preparation of MNR foam composite. Effect of amine type of modified silica particles on CO2 adsorption capacity To improve the adsorption capacity, 15 phr of modified silica particles with 1NES, 1NMS, 2NMS or 3NMS were filled in MNR foam. The results in Table 3 showed that the CO2 adsorption capacity of MNR foam composite increased after filling with modified silica parti- cles when compared to fill with unmodified silica particles due to the increase of selectivity to CO2 from amine groups on modified silica particles surface. Additionally, the effect of amine type of modified silica particles on CO2 adsorption was also investigated. In case of variation of substituted size of silanol group (1NES-Si and 1NMS-Si), MNR-15(1NMS-Si) showed the high CO2 adsorption capacity compared to MNR-15(1NES-Si). It is because the steric effect of 1NES-Si structure prevented the active site for CO2 capture. Moreover, the increase of amine contents in modifier chains (1NMS-Si), 2NMS-Si and 3NMS-Si) resulted the increased CO2 adsorption capacity of MNR foam composite owing to the high CO2 selectivity achieved by the amine groups on modified silica particles surface. Figure 8 Breakthrough curves of NR and MNR (C is the CO2 concentration at the outlet, C0 is the CO2 concentration at the inlet of reactor). 96

The 28th Special CU-af Seminar 2020 October 20, 2020 Table 3 Breakthrough time and CO2 adsorption capacity of MNR foam composite material Effect of temperature on CO2 adsorption capacity The CO2 adsorption capacity of MNR foam composite was studied at various temperature: ambient temperature, 55 oC, 70 oC and 100 oC and the result was shown in Figure 9. The CO2 adsorption capacity of MNR foam composite tended to be increased with the increase of temperature. It was because the structure of MNR foam composite and the modifier on modified silica surface become more flexible at high temperature, so the active site for CO2 capture was not hidden [15]. However, CO2 adsorption capacity of MNR foam decreased when the temperature was increased more than 70 oC because the structure of MNR may shrink compared to MNR composite. Adsorption kinetic of MNR foam composite The behavior of CO2 adsorption process for each adsorbent was studied by using three models of adsorption kinetic: pseudo-first order, pseudo-second order and Avrami’s models as follows [15,16]: 97

The 28th Special CU-af Seminar 2020 October 20, 2020 where k1 (min-1), k2 (g/mmol•min) and kA (min-1) are the rate constants, qe (mmol/g) is the adsorption capacity and qt (mmol/g) is the amount of gas adsorbed in specific time, t (min) is time and n is the order of kinetic equation. Table 4 shows the linear regression parameters (R2) of kinetic data for CO2 adsorption on MNR foam composite. The CO2 adsorption process for all samples was fitted with the Avrami’s model (R2 value > 0.95). The result might indicate that the adsorption mechanism was the combination of physisorption and chemisorption [16]. For NR foam, the physisorption appeared at the pores of NR foam and the chemisorption appeared via the chemical bonded between amine groups in the structure of curing agents (CBS and DNT) and CO2 molecules. In case of MNR, MNR-15SiO2 and MNR-15(3NM-Si), both the physisorption occurred at pores of MNR foam composite or modified silica particles and chemisorption appeared at the amine groups in the MNR structure, curing agents or modified silica particle surface. Figure 9 CO2 adsorption capacity of MNR foam composite at different temperature. Table 4 Linear regression parameters of kinetic data for CO2 adsorption on MNR foam composite. 98

The 28th Special CU-af Seminar 2020 October 20, 2020 Conclusion In this study, MNR was successfully prepared by modified NR with diallylamine. The CO2 adsorption capacity of MNR increased compared to NR because the amine group in MNR structure increased the CO2 selectivity. The effect of substituted size of silanol group (1NES-Si and 1NMS-Si) and amine content (1NMS-Si, 2NMS-Si and 3NMS-Si) of modified silica particles on CO2 adsorption capacity showed that the long chain of substituted of silanol group (1NES-Si) showed the decreased CO2 adsorption capacity owing to the stearic effect. The increase of amine content enhanced the CO2 adsorption capacity of MNR foam composite. Finally, the kinetic studied of MNR foam composite was fitted with Avrami’s model. Accordingly, these MNR foam composites can be a candidate for using as the adsorbent material to decrease CO2 gashouse gas in the atmosphere. References 1) Loganathan, S., Tikmani, M., Ghoshal, A.K., Langmuir, 2013(29): 3491-3499. 2) Mello, M.R., Phanon, D., Silveira, G.Q., Llewellyn, P.L., Ronconi,C.M., Microporous and Mesoporous Materials, 2011 (143): 174-179. 3) Sarver, J.A., Sumey, J.L., Williams, M.L., Bishop, J.P., Dean, D.M.,Kiran, E., Journal of Applied Polymer Science, 2018 (135): 45841 (1 of 24). 4) Abady, A.R.N., Movahed, S.O., Journal of Applied Polymer Science, 2017 (134): 45357 (1 of 8). 5) Han, D.H., Choi, M.C., Jeong, J.H., Cho, K.M., Kim, H.S., Composite Interfaces, 2016 (26) : 771-780. 6) Almeida, M.L.B., Ayres, E., Moura, F.C.C., Oréfice, R.L., Journal of Hazardous Materials, 2018 (346): 285–295. 7) Rose, K., Steinbüchel, A., Applied and environmental microbiology,2005 (71): 2803-2812. 8) RUHIDA, A.R., AMIR-HASHIM, M.Y., Journal of Rubber Research,2012 (15): 81-95. 9) Johns, J., Rao, V., International Journal of Polymeric Materials and Polymeric Biomaterials, 2011 (60): 766-775. 10) Sawatdiwutthipong, N., Khumngern, T., Poompradub, S., Chalermsinsuwan, B., Senior project: Development of CO2 of Chemical Technology, Faculty of Science, Chulalongkorn University, 2016. 11) Tunlert, A., Prasassarakich, P., Poompradub, S., Materials Chemistryand Physics, 2016 (173): 78-88. 12) Rolere, S., Liengprayoon, S., Vaysse, L., Sainte-Beuve, J., Bonfils,F., Polymer Testing, 2015 (43): 83-93. 13) Jong, L., Industrial Crops and Products, 2017 (105): 53-62. 14) Lu, H.-T., Colloid Journal, 2013 (75): 311-318. 15) Wang, X., Chen, L., Guo, Q., Chemical Engineering Journal, 2015(260): 573-581. 16) Keramati, M., Ghoreyshi, A.A., Physica E, 2014 (57): 161-168. 99



Efficiency and photostability of visible-light driven metal-doped NaTaO3 photocatalysts for environmentalpurification and clean energy production Assoc. Prof. Dr. Pornapa Sujaridworakun



The 28th Special CU-af Seminar 2020 October 20, 2020 Efficiency and photostability of visible-light driven metal-doped NaTaO3 photocatalysts for environmentalpurification and clean energy production Assoc. Prof. Dr. Pornapa Sujaridworakun Abstract This research aimed to synthesis of visible-light responsive Cr-doped NaTiO3 photocatalyst for degradation of organic pollutants and H2 production from water splitting reaction. It was obtained that doping NaTaO3 with Cr3+ could enhance the photocatalytic activity of NaTaO3 for degradation of MB dye under visible light irradiation, but the photocatalytic water split- ting to produce H2 could not be observed due to its improper electronic band structure. In this report, the effects of Cr ions doping amount, morphology, size of NaTaO3 photocatalyst on its photocatalytic activity was performed. In addition, the development of synthesis process to improve photodegradation efficiency and H2 evolution activity of NaTaO3 under UV irradiation was achieved. Thus a study on doping of metal ions into the high efficiency NaTiO3 in order to enhance visible-light responsive photocatalytic activity is an ongoing research. Department of Materials Science Faculty of Sciencee, Chulalongkorn University Bangkok, Thailand 103

The 28th Special CU-af Seminar 2020 October 20, 2020 Introduction and Objectives In recent years, growing concern and awareness for the exhaustion of fossil fuel coupled with their environmental impact leads researchers to search and development for efficient route to generate clean energy. Researchers has played much attention to hydrogen which is considered to be a promising energy carrier. Semiconductor photocatalysis using sunlight to break down harmful organic pollutants and produce hydrogen is one of the effective approaches to solve the energy and environmental problems. Hydrogen production via photocatalytic overall spitting of water has attracted research interest since the pioneering work conducted by Honda and Fujishima on a photoelectrochemical [1]. Until now, a wide range of photocatalysts has been developed, most of which are oxide-based semiconductors and recently various perovskite-type materials have also attracted immense interest due to their potential catalytic property to split water. Among the vast majority of active photocatalysts, perovskite-type sodium tantalate compound (NaTaO3) has been found to exhibit the most active for H2 production as well as the degradation of toxic organic pollutants [2,3]. Unfortunately, owing to their large band gap (4.1 eV), their practical applications have been restricted to the ultraviolet region which occupy only about 4% of the whole solar spectrum. Thus, research on the development of visible-light active NaTaO3 photocatalysts for water splitting has attracted much attention recently. It has been reported that doping with foreign elements is one of the effective strategies to improve visible light photocatalytic activity of semiconductor materials [4]. Doping NaTaO3 compounds with the dopant like Bi, Co, Cr, Cu, Nb, N and Fe have been studied to induce visible light absorption in NaTaO3. These studies show that doping NaTaO3 with metal, lanthanide ions offer potential to reduce band gap by introducing defect states [4-7]. The photocatalytic activities of semiconductors nanoparticles are known to be critically influenced by their crystallinity and surface area in that the former reduces the number of electron-hole recombination center, and the latter provides more surface active sites for photocatalytic reactions. To achieve high efficiency, high crystallinity and high surface area of photocatalyst are required. Thus various synthesis methods have been established to enhance the photocatalytic activity. Generally, tantalate photocatalysts are mostly synthesized by a conventional solid-state reaction at high temperature which is difficult to control crystallinity, particles localized segregation of components and stoichiometry of products [8-10]. For low temperature process such as sol-gel or precipitation routes are required calcination for the formation of final product [10-12]. Among different methods, it is well known that hydrothermal synthesis has been widely used to prepare of advanced nanostructure due to it can generate highly crystalline product with high purity, narrow size distribution and low aggregation. In addition, the morphology and crystal form can also be easily controlled by adjusting hydrothermal conditions [7,10,13-14]. For these reasons, our attention has been focused on the preparation of high performance visible light responsive NaTaO3 by doping with Cr3+ using hydrothermal process. The concentrations of dopants and optimum hydrothermal synthesis conditions will be investigated to achieve the high efficiency photocatalyst for degradation of organic pollutants and H2 production through overall splitting of water. Moreover, the chemical state of metal dopants introduced in the photocatalyst will be evaluated using the feasible Extended X-ray Absorption Fine Structure (EXAFS). 104

The 28th Special CU-af Seminar 2020 October 20, 2020 Materials and Methods Synthesis of photocatalysts Part I: NaTaO3samples were synthesized by hydrothermal process using Ta2O5 and NaOH as starting materials. Ta2O5 powder was added to NaOH solution to form a mixed suspension by magnetically stirring at room temperature. After that the mixed dispersion was transferred into Teflon-lined stainless-steel autoclave (40 mL capacity). Then, the autoclave was kept in a furnace at a temperature between 200 oC for a period between 24 h. After that the autoclave was cooled to room temperature naturally, and the resultant precipitates was separated by centrifuge, washed with distilled water and ethanol several times, then it was completely dried at 60 oC in an oven. Metal-doped NaTaO3 samples was synthesized by the same procedure as NaTaO3 excepted that Cr2O3 was added to the above mixture in which the amount of metal ion dopants is at 2-8% in a weight ratio to that of the tantalum source. Part II: Ta2O5 0.441 g was mixed with NaOH solid (8.4 –25.2 g) and maintained in the oven at 180°C for 20 h. After air-cooled to room temperature, 30 ml aqueous solution of glycerol and deionized water (glycerol: DI water ratio = 1:2 ) were added into the mixture. The suspension was continuously stirred for 2 h. before transferred into an autoclave with 50 ml volume Teflon line. The autoclave was heated in an oven at 180 °C for 2-10 h. After that, the autoclave was cooled down to room temperature. The samples were centrifuged and washed with distilled water until pH neutral and finally washed with ethanol. Subsequently, the white product was dried at 60 °C. For this process, it was mentioned later as 2-steps hydrothermal process. And the samples prepared without heat treatment before hydrothermal of starting materials in an oven was called as 1-step hydrothermal. Characterizations of synthesized photocatalyst samples The properties of as-synthesized photocatalyst samples were characterized using various techniques. The phase structure and crystallinity will be evaluated by X-ray diffractometry. The specific surface area was determined using the Brunauer-Emmett-Teller (BET) method. The morphologies were observed by scanning electron microscopy. The optical properties were investigated using UV-vis diffuse reflectance spectroscopy. The chemical states of dopants introduced in the photocatalyst was investigated by Extended X-ray Absorption Fine Structure (EXAFS) at the Synchrotron Light Research Institute (SLRI), Thailand. Photocatalytic activity evaluation Photocatalytic activities of the samples were evaluated by the decomposition of methylene blue (MB) solution. Asolar simulator (ASAHI spectra HAL-302, 300W Xenon arc lamp) was used as a light source, and the light intensity was set at 360 W/m2. The cutoff filter was placed above reactor to completely remove any radiation below the wavelength of 420 nm to provide visible light irradiation. Water was circulated through the annulus to avoid heating during the reaction. The initial concentration of MB is about 0.02 mM and the weighed amount of the catalyst used was 0.1 g. All the experiments were performed in the natural pH of the MB solution. The suspension was stirred continuously for 120 min in the dark before reaction. When the MB concentration become constant, light irradiation was irradiated for 2 hrs. The absorbency of the sample solution was detected volumetrically at certain times with a UV-visible spectrometer and was used to calculate the photo-degradation rate of the MB. In case of pure NaTaO3 prepared by method in part II, the photocatalytic activity of all samples was investigated under UV-light irradiation (Phillip Actinic BL TL-D 18W / 10 1SL/25). 105

The 28th Special CU-af Seminar 2020 October 20, 2020 The initial concentration of 200 ml Rhodamine B solution was around 0.01 mM and the weighed amount of the catalyst used was 0.1 g. All the experiments were performed in the natural pH of the RhB solution. The suspension was stirred continuously for 5 hrs in the dark before reaction. When the RhB concentration became constant, light irradiation was irradiated for two hours. Photocatalytic test for overall water splitting The reactions were carried out in a gas-closed circulation system. 0.1 - 1 g of the photocatalyst was dispersed in 100 ml water by a magnetic stirrer in an inner irradiation reactor cell. Then, the reaction system was evacuated several times to remove air prior to irradiation. Next, sample solution was irradiated under UV light using a 200 W Hg lamp. Finally, the amount of H2 produced was measured using a gas chromatography system. Results and Discussion Part I: Fig.1 shows XRD patterns of pure NaTaO3 and NaTaO3 doped with different amount of Cr ions (2-8 wt%). As shown, all synthesized samples were well crystallized NaTaO3 which indexed as orthorhombic structure (JCPDS card: 25-0863). There were no additional phases except the very low intensity of Cr2O3 characteristic peaks were observed in samples doped with Cr ion from 4-8 wt%. It appeared that doping with Cr ion did not result in significant structural changes for NaTaO3. When enlarging the peak at 2 theta = 22.9 degree of 2-8wt% Cr-doped samples (Fig.1, right), a systematic shift of diffraction peak toward larger angles was detected. This could cause by lattice shrinkage due to the substitution of Ta5+ by smaller size Cr3+ ions. In addition, the oxygen vacancies in lattice structure might be occurred [15]. Fig.1 XRD patterns of pure NaTaO3 and Cr3+- doped NaTaO3 with various amount dopants. Figure 2 presents microstructure observed by SEM of (a) pure NaTaO3 and (b) 8% Cr-doped NaTaO3 (b). The pure NaTaO3 particles exhibit the well-defined cubic shaped morphology with edge size around 300 nm. The 8% Cr-doped NaTaO3 sample composes of cubic NaTaO3 106

The 28th Special CU-af Seminar 2020 October 20, 2020 particles (~200 nm) which were attached with large and irregular shape of Cr2O3 particles. It was obtained that the size of NaTaO3 particles was decreased in Cr-doped samples, which could cause by the substitution of smaller Cr3+ for Ta5+ site inducing a decrease of the unit cell volume. In addition, the elemental composition was verified by using the energy dispersive spectroscopy (EDS) technique as shown in Figure 2(c). It was confirmed the existence of Na, Ta, O and Cr without other impurity in 8% Cr-doped NaTaO3 sample. For 2% Cr-doped NaTaO3 sample, only cubic NaTaO3 particles with similar shape and size to pure NaTaO3 were observed, and confirming the XRD results. The specific surface area value of pure NaTaO3 and 2,4,6,8 wt% Cr-doped NaTaO3 measured by BET method were 3.07, 3.29, 3.54, 3.38 and 2.98, respectively. It revealed that Cr-doped samples had larger surface area than that of pure NaTaO3 due to the decreasing in particle size. However, the decrease in surface area when doping amount was higher than 4wt% might be caused by the existence of Cr2O3 particles as shown in SEM results. Fig. 2 SEM images of pure (a) NaTaO3 (a), (b) 8wt% Cr-doped NaTaO3. and (c) EDS result of sample (b). EXAFS measurements at the Cr K-edge were also acquired for 2-8%Cr-doped NaTaO3 samples, which all sample exhibited very similar results as shown in Fig. 3. The adsorption edge of all Cr-doped NaTaO3 samples show the same position of Cr3+ pre-edge from Cr2O3 foil [16]. It confirmed that the existent specie of Cr ion is only in form of Cr3+ in all samples. Fig.3 EXAFS results measured at the Cr K-edge for 2-8wt%Cr-doped NaTaO3 samples. 107

The 28th Special CU-af Seminar 2020 October 20, 2020 The UV-Vis absorption spectra in Fig. 4(a) show that Cr-doped NaTaO3 samples could be responded for visible-light. Compared with the spectrum of pure NaTaO3 that has the absorption edge at around 310 nm, the red shift was clearly observed in the 2% Cr-doped NaTaO3 and the absorption edge of the Cr-doped NaTaO3 sample has shifted towards the longer wavelength when the amount of Cr2O3 was increased. Persistently, the band gap energy was decreased from 4 to 3.8 eV as shown in the Fig. 4 (b). This indicated that the doping of Cr3+ ions led the red shift of the absorption edge into the visible regions. Fig. 4 UV-Vis absorption spectra (a) and band gap energy of as-synthesized samples. Figure 5 shows the photoluminescence spectra of NaTaO3 and 2-8 wt% Cr-doped NaTaO3. The PL emission spectra have been used in the field of photocatalysis over solid semiconductors as a useful probe for understanding the surface processes that involve the photogenerated electrons and holes. Obviously, Cr-doped NaTaO3 samples exhibited the decreasing of PL intensities as increasing Cr doping contents indicating that the PL intensities are quite affected by the doping level. The relative intensity of the Cr-doped NaTaO3 is lower than pure NaTaO3 and decreased when amount of Cr doping increased implied a decrease of charge recombination [10]. The critical reasons for promoting the effective separation of electron–hole pairs might be the decreased particle size and the shrinkage lattice volume of the catalysts [17]. To evaluate the photoactivity of the as-prepared samples, the methylene blue dyes (MB) photodegradation over all as-synthesized doped and un-doped NaTaO3 samples were performed as shown in Figure. 6. It was demonstrated that, the undoped NaTaO3 irradiated under visible light (λ > 400 nm) did not generate the decrease of MB concentration by comparing with MB solution without sample, since NaTaO3 was not responsible for visible light. On the other hand, the degradation of MB was observed from all Cr-doped NaTaO3 samples. As the results, the highest efficiency for photodecomposition was presented in 2% Cr-doped NaTaO3 by removal 99% of MB after 2 h irradiation. Increasing Cr doping amount led to decreasing in photodegradation efficiency. 108

The 28th Special CU-af Seminar 2020 October 20, 2020 Fig.5 The PL emission spectra of NaTaO3 and all Cr-doped NaTaO3 samples. Fig.6 The photocatalytic activity of the as-prepared NaTaO3 and Cr-doped NaTaO3 samples under visible light irradiation. From the previous reports, they suggested that doping with metal elements such as chromium (Cr) into NaTaO3 structure can promote the photocatalytic efficiency under visible light of NaTaO3. The doping elements creates a new energy level in the band gap energy of NaTaO3 results in reducing its band gap energy. Therefore, NaTaO3 can be activated under UV and visible light irradiation. Dopants also promote the separation of charge carriers used in the reactions. Therefore, the recombination of electrons and hole is reduced which provides more stability for long-term photocatalytic reaction [14,18]. For photocatalytic water splitting evaluation, it could not detect the H2 evolution from the reaction of produced pure and Cr-doped NaTiO3 samples. Generally, for hydrogen production, the CB level should be more negative than hydrogen production level (EH2/H2O) while the VB should be more positive than water oxidation level (EH2/H2O) for efficient oxygen production from water by photocatalysis. Therefore, the reason why these samples did not provide H2 production was attributed to the undesired band structure (VB and CB positions) for photocat- alytic water splitting reaction which was confirmed by Cyclic voltammetry (CV) measurement. 109

The 28th Special CU-af Seminar 2020 October 20, 2020 It is known that, the particle size, structure and morphology of the photocatalyst, which can be affected by the synthesis methods, play an important role in tuning their photocatalytic activity. For example, it was reported that the NaTaO3 microspheres obtained by hydrothermal showed much better photocatalytic performance than the NaTaO3 microcubes prepared by the solid-state method for overall water splitting activity under UV light [19]. For these reasons, it shifted our attention to focus on the development of synthesis process for enhancing properties and photocatalytic activity of pure NaTaO3 which was performed in part II. Part II: XRD results in Fig. 7 shows the diffraction patterns of high crystallinity pure NaTaO3 with orthorhombic phase of samples prepared from both 1 and 2-steps hydrothermal synthesis. The crystallite size calculated using Scherrer equation of samples synthesized by 1 and 2-steps hydrothermal were 39.76 and 37.98. It revealed that 2-steps synthesis led to the decrease in crystallite size of NaTaO3 crystals. Fig.7 XRD patterns of pure NaTaO3 prepared by 1-step hydrothermal at 180 oC for 24 hrs. and 2-steps hydrothermal synthesis for 6 hrs. SEM images in Fig. 8 revealed that the NaTaO3 obtained from 2-step hydrothermal synthesis for 6 hrs. composed of spherical particles with size varies in range of 100-300 nm which was smaller and lower agglomeration than samples prepared for longer time. While particles obtained from hydrothermal synthesis for 24 hrs composed of sharp edge cubic particles with size varies in range of 100-400 nm. It demonstrated that synthesis conditions and processes had significantly affected on size and morphology of the obtained products. In 2-steps synthesis, the heat treatment of starting materials before hydrothermal enhanced the crystallization of NaTaO3 under hydrothermal at shorter reaction time comparing with 1-step process, and the addition of glycerol played a role on the morphology controlled of spherical particle. In addition, hydrothermal reaction time influenced the crystal size. 110

The 28th Special CU-af Seminar 2020 October 20, 2020 Fig.8 SEM images of pure NaTaO3 prepared by (a)1-step hydrothermal at 180 oC for 24 hrs. and (b) 2-steps hydrothermal synthesis for 6 hrs. As shown in Fig. 9 both NaTaO3 samples prepared from 1 and 2-steps showed a similar absorption behavior in that the edge position is at around 310-320 nm. The band gap energy estimated from the onsets of the diffuse reflection spectra is about 4 - 3.88 eV. Fig.9 UV-Vis absorption spectra of pure NaTaO3 prepared by 1-step hydrothermal at 180 oC for 24 hrs and 2-steps hydrothermal synthesis for 6 hrs. 111

The 28th Special CU-af Seminar 2020 October 20, 2020 Fig.10 The photocatalytic activity of pure NaTaO3 prepared by 1- step hydrothermal at 180 oC for 24 hrs and 2-steps synthesis for 6 hrs under UV irradiation. As the photocatalytic activity results shown in Fig.10, the NaTaO3 sample prepared by 2-step hydrothermal for 6 hrs and it exhibited the highest efficiency for degradation of RhB under UV irradiation comparing with other samples, which was much higher than sample prepared by 1- step process. This was attributed to its smaller particle with lower agglomeration. Moreover, it was explained in previous work that the superior photocatalytic performance of NaTaO3 microspheres with rough surface is easily understandable in terms of the increases in the available amounts of the photogenerated charges and adsorbed water by rough surface [20]. For water splitting activity of the NaTaO3 sample prepared by 2-step hydrothermal for 6 hrs., 1 wt% RuO2 co-catalyst was loaded onto the surface of the NaTaO3 spheres in order to improve the photocatalytic activity. It was successfully observed the H2 evolution in overall water splitting reaction which was 0.6 µmol/h/gcat. Conclusion In this work, the synthesis of visible light responsive Cr-doped NaTaO3 photocatalysts was achieved using hydrothermal process. It was obtained that doping amount played an important role on the photocatalytic activity. The highest photocatalytic efficiency for MB degradation under visible light irradiation was obtained from sample doped with 2wt% Cr3+. However, H2 production from Cr-doped samples could not be detected under visible light. Thus, the improve- ment of photocatalytic activity of pure NaTaO3 particles was proposed using 2-steps hydrothermal process. It demonstrated that spherical NaTaO3 with the size ranging from 100 to 300 nm obtained by 2-steps synthesis performed higher photocatalytic activity for RhB degradation than cubic particles prepared by 1-step hydrothermal, in that the sample prepared at optimum conditions successfully produced H2 via photocatalytic water splitting under UV irradiation. As the results, the development of visible-light responsive NaTaO3 for overall water splitting will be further studied. 112

The 28th Special CU-af Seminar 2020 October 20, 2020 References 1) Xu, D., et al., Langmuir, 2015. 31(35): 9694-9699. 2) Yan, S.C., et al., Solid State Ionics, 2009. 180 (32-35): 1539-1542. 3) Jana, P., et al., International Journal of Hydrogen Energy, 2014. 39(10): 5283-5290. 4) Modak, B., Srinivasu, K. and Ghosh, S.K., The Journal of Physical Chemistry C. 2014. 118(20): 10711-10719. 5) Su, Y., et al., RSC Advances. 2012. 2(33): 12932. 6) Kang, H.W., et al., International Journal of Hydrogen Energy, 2013. 38(15): 6323-6334 7) Wang, X., et al., Journal of Nanoscience and Nanotechnology, 2010.10(3): 1788-1793. 8) Zhou, X., Shi, J. and Li, C., The Journal of Physical Chemistry C,2011. 115(16): 8305-8311. 9) Iwase, A., Kato, H. and Kudo, A., Applied Catalysis B: Environmental,2013. 136-137: 89-93. 10) Hu, C.-C., Tsai, C.-C. and Teng. K., Journal of the American CeramicSociety, 2009. 92(2) : 460-466 11) Husin, H., et al., Applied Catalysis B: Environmental, 2011. 102(1-2): 343-351. 12) Lin, W.-H., et al., Applied Physics Letters, 2006 (89): 211904. 13) Grewe, T. and Tuysuz,H., ACS Appl Mater Interfaces, 2015. 7(41): 23153-62. 14) Jiang, W., Jiao, X. and Chen, D., International Journal of HydrogenEnergy, 2013. 38(29) : 12739-12746. 15) Lan, N. T., et al., Materials Transactions, 2016. 57(1): 1-4. 16) Martis, V., et al., Physical Chemistry Chemical Physics, 2013. 15(1): 168-175. 17) Wang, X., et al., Journal of Nanoscience and Nanotechnology, 2010. 10(3): 1788-1793. 18) Grewe, T. and H. Tuysuz., ACS Appl Mater Interfaces, 2015. 7(41): 23153-62. 19) Li, Y., Gou, H., Lu, J and Wang, X., International Journal of Hydrogen Energy, 2014. 39(23) : 13481-13485 Acknowledgements We would like to thank the Synchrotron Light Research Institute, Nakhon Ratchasima, Thailand, for the XANES analysis. References and Prof. Nobuo Saito, Nagaoka University of Technology for supporting water splitting measurement. 113



Intelligent Monitoring and Estimation of Surface Roughness and Straightness in CNC Turning Prof. Dr. Somkiat Tangjitsitcharoen



The 28th Special CU-af Seminar 2020 October 20, 2020 Intelligent Monitoring and Estimation of Surface Roughness and Straightness in CNC Turning Prof. Dr. Somkiat Tangjitsitcharoen Abstract The relations of the surface roughness, the straightness and the cutting conditions are investigated to realize an intelligent CNC machine by monitoring the in-process cutting forces during CNC turning of aluminium 6063 with the use of coated carbide tools. The cutting force is proposed to predict the straightness and surface roughness. The Fast Fourier Transform (FFT) is used to prove the relations of them by checking the frequencies of them. The cutting force ratio is proposed and normalized to predict the in-process surface roughness and straightness regardless of the cutting conditions. The surface roughness and the straightness are calculated simultaneously by employing the two-layer feed-forward neural networks with sigmoid hidden and linear output neurons. The neural networks is trained by using the Levenberg-Marquardt back propagation algorithm. It is understood that the surface roughness and the straightness can be estimated well by utilizing the proposed method under various cutting conditions. Department of Industrial Engineering Faculty of Engineering, Chulalongkorn University Bangkok, Thailand. 117

The 28th Special CU-af Seminar 2020 October 20, 2020 Introduction The surface roughness and the straightness of the machined cylindrical parts is considered as the critical dimensions in CNC turning. However, a control of those dimensions is difficult since it cannot be measured directly during turning. It is required to know the in-process surface roughness and straightness but they depend on the cutting conditions such as the cutting speed, the feed rate, the depth of cut, the tool nose radius and the rake angle. Hence, the relations of the surface roughness, the straightness and the cutting conditions will be examined in order to develop an intelligent CNC machine. Many researches have been developed so far to monitor the surface finish and the straightness. Kim et al. [1] proposed a multi-step straightness control system (MSSC) through the fuzzy self-learning method to minimize the straightness error. Jialiang et al. [2] considered the cutting conditions to predict the error of slenderbar using the statistical model. The results had been proved that the error mainly depends on the depth of cut and the feed rate. Bugra et al. [3] utilized the cutting forces to estimate the bar deflection, which is accurate enough to estimate the deformed surfaces. Xuefeng et al. [4] proposed the model to calculate the dimensional error which uses the flank wear land. It is shown that the dimension error increases with an increase in cutting speed, cutting time, tool nose radius and clearance angle. Shawky et al. [5] proposed an on-line evaluation system of the workpiece size in bar turning process. Thamizhmanii et al. [6] optimized the cutting conditions to find the minimum surface roughness in turning. The results showed that the surface roughness is affected by depth of cut and feed rate. Rao et al. [7] studied the cutting conditions on cutting force and surface roughness in turning process. It is concluded that the feed rate affected on both the cutting force and the surface roughness. The depth of cut influenced on the cutting force. Lalwani et al. [8] monitored the effect of cutting force and surface roughness in hard turning of steel. It is understood that the feed rate is the most significant factor on the surface roughness. The models of surface roughness have been proposed by many researchers [9-12]. Tangjitsitcharoen, S. [12] modeled the in-process surface roughness with the cutting force ratio by utilizing the exponential function. It is proved that the cutting force can aid to predict the in-process surface roughness under various cutting conditions during cutting. Hence, the relations among the surface roughness, the straightness, the cutting conditions and the cutting forces are monitored while turning process using the coated carbide tool and the work material of aluminium 6063. In order to realize an intelligent CNC machine to estimate the in-process surface roughness and straightness concurrently, the two-layer feed-forward neural networks with Levenberg-Marquardt back propagation learning algorithm is introduced [13-14]. The number of nodes in the hidden and output layers are 10 and 3 respectively while the input layer has 7 nodes, which consist of five cutting parameters and two proposed cutting force ratios. It is expected that the proposed system can predict the in-process surface roughness and straightness well in CNC turning. Monitoring of surface roughness and straightness The cutting forces in turning process are detected by dynamometer which is installed on the turret of CNC turning machine [14]. The obtained cutting forces consist of three components, which are the radial force (Fx), the feed force (Fy) and the main force (Fz) as shown in Fig. 1. The feed force is most sensitive to the surface roughness while the main force is affected by the cutting conditions [10,15]. The main force and the feed force are hence normally adopted to predict the surface roughness during the in-process cutting. 118

The 28th Special CU-af Seminar 2020 October 20, 2020 The cutting conditions such as the feed rate, the cutting speed, the depth of cut, and the tool nose radius affect the surface roughness. The feed rate and the tool nose radius are significantly monitored, as they are the function of the theoretical surface roughness. It is known that the larger tool nose radius improves the surface roughness, vice versa an increase in feed rate causes the poor surface finish. The better surface finish is obtained at small depth of cut. Since the cutting force depends on the depth of cut which leads to the vibration of cutting tool and causes the higher surface roughness. It is necessary to normalize the cutting forces to eliminate those combinations of the cutting conditions. The ratio of the cutting forces (Fy/Fz), which can cut-off the combinations of the cutting conditions in its components, is proposed to predict the in-process surface roughness although the cutting conditions are changed [12]. The relation between the dynamic cutting force and the straightness profile can be checked by comparing the frequencies of them in frequency domain using the Fast Fourier Transform (FFT), which is similar to the relation between the dynamic cutting force and the surface roughness profile [15]. The feed force had been utilized to monitor the in-process surface roughness during cutting in the previous research [10]. In the same way, the feed force is also applied to monitor the straightness in this research.The feed force is most sensitive to the straightness and surface roughness profiles referring to the measuring direction which is the same as the feed rate direc- tion. The roughness left on the machined surface is normally created by the feed rate and the tool nose radius of the cutting tool which also generate the straightness profile. Hence, the height and depth of the peaks and valleys on the workpiece are considered to predict the straightness. It is expected that the straightness profiles are correspond with the amplitudes of dynamic feed forces as shown in Fig. 1. Figure 1. Illustration of relations of surface roughness, straightness and cutting forces. 119

The 28th Special CU-af Seminar 2020 October 20, 2020 However, the dynamic feed force is proposed to be generalized and dimensionless by taking the ratio of dynamic feed force to its static feed force as shown in Eq. (1) [16] and Fig. 1. The highest peak-to-valley amplitude of the dynamic feed force Fy(dynamic) is obtained from the difference between the maximum peak Fy(max) and the minimum valley Fy(min). The cutting force ratio is calculated as the ratio of the highest peak-to-valley amplitude of dynamic feed force to its static feed force Fy(static). The highest peak-to-valley amplitude of dynamic feed force is expected to be correspondent with the straightness profile as shown in Fig. 1. Neural network approach Figure 2 shows the two-layer feed-forward neural networks with a sigmoid transfer func- tion in the hidden layer and a linear transfer function in the output layer. The network is trained with the Levenberg-Marquardt back propagation algorithm and evaluated its performance using mean square error and regression analysis. The input layer consists of 7 input which are cutting speed, feed rate, depth of cut, tool nose radius, rake angle, cutting force ratios of (Fy/Fz) and [(Fy(max)- Fy(min)]/Fy(s). The hidden layer is designed to use the 10 neurons to obtain the weight and the bias with the smallest mean square error. While the outputs compose of surface roughness (Ra, Rz) and straightness (St), respectively. Figure 2. Illustration of two-layer feed-forward neural networks with the back propagation Experimental setup and cutting conditions The measuring direction of the surface roughness and straightness profiles on the cylindrical workpiece is measured parallel to the axis of it [17-18]. The experiments are conducted on the CNC turning machine. The aluminium alloy (Al 6063) is adopt-ed for the cutting tests with the coated carbide tools, which have the different tool nose radiuses of 0.4 and 0.8 mm. The rake angle of cutting tool is 11°. The dynamometer (Kistler, model: 9121) is installed on the turret to obtain the cutting force signals while turning. The cutting forces are amplified through the charge amplifier (Kistler, model: 5083) before digitization and calculation within the personal computer as shown in Fig. 3. The straightness (St), the average surface roughness (Ra) and the surface roughness (Rz) are measured by the surface roughness tester (Mitutoyo, SJ-400). The cutting conditions are summarized in Table 1. 120

The 28th Special CU-af Seminar 2020 October 20, 2020 Figure 3. Illustration of experimental setup. Table 1. Cutting conditions. The following procedures are proposed to obtain the relation of the surface rough-ness, the straightness, the cutting force ratio and the cutting conditions. 1. Start cutting and monitor the cutting forces referring to the cutting conditions. 2. Check the relations of the surface roughness and straightness profiles and the cutting forces in the time domain and the frequency domain. 3. Measure the surface roughness and straightness profiles for each cutting condition. 4. Calculate the cutting force ratios of (Fy/Fz) and [(Fy(max)-Fy(min)]/Fy(s), which are correspondent with the time records of the surface roughness and straightness, respectively. 5. Analyze the relation of surface roughness, straightness, cutting conditions and cutting force ratios. 6. Repeat the procedures (1) to (5) for other cutting conditions. 7. Train the two-layer feed-forward neural networks for the relation of average surface roughness (Ra), surface roughness (Rz), and straightness (St) with the cutting force ratio under various cutting conditions using back propagation algorithm as shown in Fig. 2. 121

The 28th Special CU-af Seminar 2020 October 20, 2020 8. Test the in-process prediction of surface roughness (Ra, Rz) and straightness (St) via obtained neural networks system which has the 7 input with 10 hidden neurons and 3 outputs. 9. Verify the obtained neural networks weights and bias by using the new cutting conditions and the new cutting tool with the different rake angle. Experimental results and discussions The preliminary experiments have been conducted with the aluminium alloy (Al 6063) and the coated carbide tools to monitor the dynamic cutting forces as shown in Figs. 4 to 7, which are the redial force (Fx), the feed force (Fy) and the main force (Fz). The surface roughness and straightness profiles can be measured off-line firstly. Figure 4 illustrates the dynamic cutting force and the corresponding time records of the surface roughness in time domain which are obtained from the cutting speed of 250 m/min, the feed rate of 0.125 mm/rev, the depth of cut of 0.2 mm, the tool nose radius of 0.4 mm and the rake angle of 11°. Figure 5 shows their power spectrum density (PSD) in the frequency domain by using the Fast Fourier Transform (FFT). It is understood that the frequency of the surface roughness is correspondent with the frequency of the dynamic cutting forces in frequency domain which is 28 Hz. Figure 4. Example of dynamic cutting forces and surface roughness signal in time domain. Figure 5. Example of PSD of dynamic cutting forces and surface roughness in frequency domain. 122

The 28th Special CU-af Seminar 2020 October 20, 2020 Figure 6. Example of dynamic cutting forces and straightness signal in time domain. Figure 7. Example of PSD of dynamic cutting forces and straightness in frequency domain. The examples of the experimentally obtained dynamic cutting forces and the corresponding time record of the straightness profile in time domain which are obtained from the cutting speed of 150 m/min, the feed rate of 0.15 mm/rev, the depth of cut of 0.1 mm, the tool nose radius of 0.8 mm and the rake angle of 11°. It is shown that the dynamic cutting force signals are similar to the straightness profile as shown in Fig. 6 due to the numbers of peak-to-valley cycles. The FFT is applied to check the frequencies of the dynamic cutting forces and the straightness profile which are the same frequency at 17 Hz as shown in Fig. 7. It has been proved that the dynamic cutting forces correspond to the straightness pro-file. Hence, it is understood that the dynamic cutting force can be used to predict the in-process straightness of the workpiece while cutting. In order to monitor the in-process surface roughness and straightness, the cutting force ratios of (Fy/Fz) and [(Fy(max) -Fy(min)]/Fy(s) as well as the cutting conditions, which are the cutting speed, the feed rate, the depth of cut, tool nose radius and the rake angle are employed as the input of the back-propagation neural networks in order to predict the surface roughness and straightness. 123

The 28th Special CU-af Seminar 2020 October 20, 2020 Figure 8. Illustration of experimentally obtained relation between straightness and cutting speed at feed rate of 0.15 mm/rev, depth of cut of 0.3 mm, rake angle of -6° and tool node radius of 0.4 and 0.8 mm. Figure 9. Illustration of experimentally obtained relation between straightness and feed rate at cutting speed of 200 mm/min, depth of cut of 0.3 mm, rake angle of 11° and tool node radius of 0.4 and 0.8 mm. Figure 8 shows the example of relation between the straightness and the cutting speed. While the cutting speed increases, the straightness error decreases because the work material becomes softer which leads to the lower cutting forces, Hence, the surface roughness and the straightness would be improved at the higher cutting speed. Figure 9 shows the example of straightness obtained from the different feed rates comparing with the same tool nose radius at the depth of cut of 0.3 mm and the cut-ting speed of 200 mm/ min. It is understood that an increase in feed rate causes the poor straightness. The higher feed rate increases the cutting forces, consequently the straightness error increases. Figure 10. Illustration of experimentally obtained relation between straightness and depth of cut at cutting speed of 200 mm/min, feed rate of 0.1 mm/rev, rake angle of -6° and tool node radius of 0.4 and 0.8 mm. 124

The 28th Special CU-af Seminar 2020 October 20, 2020 Figure 11. Illustration of experimentally obtained relation between straightness and tool node radius at cutting speed of 200 mm/min, feed rate of 0.125 mm/rev, rake angle of 11° and depth of cut of 0.3 mm. Figure 10 represents the example of relation between the straightness and the depth of cut. An increase in depth of cut will cause the straightness error because of the larger cutting area resulting in higher cutting forces and straightness error. From the Fig. 11, it can be concluded that the larger tool nose radius helps to improve the straightness because it can reduce the feed marks on the machined surface, which is the same reason as the surface roughness. Figure 12. Illustration of the training and validation of the neural networks system to monitor the surface roughness and straightness. The experimental results of training and validation of the neural networks system are shown in Fig. 12 with the high accuracy of 96.189%, which is correlation coefficient. The algorithm to calculate the cutting forces and predict the in-process surface roughness and straightness are proposed in Fig. 13. 125

The 28th Special CU-af Seminar 2020 October 20, 2020 Figure 13. Illustration of algorithm to calculate the cutting forces and predict the in-process surface roughness and straightness. Figures 14. and 15 shows the experimentally measured surface roughness and straightness of 32 cutting tests versus with the in-process predicted surface roughness and straightness obtained from the neural networks. The new cutting conditions and the new cutting tool are shown in Table 2. Table 2. New cutting conditions. 126

The 28th Special CU-af Seminar 2020 October 20, 2020 Figure 14. Illustration of the measured average surface roughness (Ra), the measured surface roughness (Rz) and the in-process predicted surface roughness from neural networks. Figure 15. Illustration of the measured straightness (St) and the in-process predicted straightness from neural networks. The prediction accuracy of the Ra, Rz and St are about 66.40%, 73.72% and 57.28%, respectively. The reason is that the training sets and the new cutting conditions are different. Since the training set is selected only the cutting forces from the continuous chips. However, the new cutting conditions to verify the neural network are from the cutting forces of continuous and broken chips. It indicates that the in-process predicted surface roughness and straightness from neural network can be effectively used to monitor the surface roughness and straightness during cutting. 127

The 28th Special CU-af Seminar 2020 October 20, 2020 Conclusions As the intelligent CNC machines and manufacturing systems are expected to be realized in the near future, the in-process prediction of surface roughness and straightness are hence proposed and developed by utilizing neural networks with the use of Leven-berg-Marquardt back propagation algorithm. The relations of the surface roughness, the straightness, the cutting speed, the feed rate, the depth of cut, the tool nose radius, and the cutting force ratio are investigated.The experimental results showed that the effects of cutting conditions affect the sur-face roughness and the straightness. However, the effects of those cutting conditions can be eliminated by utilizing the cutting force ratio in order to predict the in-process roughness and straightness. It is proved by the cutting experiments that the in-process surface roughness and straightness can be predicted and obtained with the high accuracy via the two-layer feed-forward neural networks with sigmoid hidden and linear output neurons. Acknowledgments This work was performed and supported by The Thailand Research Fund (TRF) and The Asahi Glass Foundation, Japan, 2019. References 1) Kim, S.C. and Chung, S.C., Synthesis of the multistep straightness control system for shaft straightening processes. Mechatronics. 2002(12): 139-156. 2) Jialiang, G. and Rongdi, H., A united model of diametral error in slender bar turning with a follower rest. International Journal of Machine Tools & Manufacture. 2006(46): 1002-1012. 3) Kilic, B., Aguirre-Cruz, J.A. and Raman, S., Inspection of the cylindrical surface feature after turning using coordinate metrology. International Journal of Machine Tool & Manufacture. 2007(47): 1893-1903. 4) Bi, X.F., Liu, Y.X. and Liu, Y., Analysis and control of dimensional precision in turning process. Chinese Control and Decision Confer ence. IEEE, 2009: 3456-3459. 5) Shawky, A.M. and Elbestawi, M.A., In-process evaluation of workpiece geometrical tolerances in bar turning. International Journal of Machine Tool & Manufacture. 1996(36): 33-46. 6) Thamizhmanii, S., Saparudin, S. and Hasan, S., Analyses of surface roughness by turning process using Taguchi method. Journal of Achievements in Materials and Manufacturing Engineering. 2007(20): 503-506. 7) Rao, C.J., Rao, D.N. and Srihari, P., Influence of Cutting Parameters on Cutting Force and Surface Finish in Turning Operation. Procedia Engineering. 2013(64): 1405-1415. 8) Lalwani, D.I., Mehta, N.K. and Jain, P.K., Experimental investigations of cutting parameters influence on cutting forces and surface roughness in finish hard turning of MDN250 steel. Journal of Materials Processing Technology. 2008(206): 167-179. 9) Thammasing, V. and Tangjitsitcharoen, S., In-Process Prediction of Surface Roughness in Grinding Process by Monitoring of Cutting Force Ratio. Applied Mechanics and Materials. 2014(627): 29-34. 10) Tangjitsitcharoen, S., Samanmit, K. and Ratanakuakangwan, S., Development of surface roughness prediction by utilizing dynamiccutting force ratio. Applied Mechanics and Materials. 2014(490-491) : 207-212. 11) Tangjitsitcharoen, S. and Damrongthaveesak, S., Advance in monitoring and process control of surface roughness. World Academy of Science, Engineering and Technology. 2013(7) : 07-29. 128

The 28th Special CU-af Seminar 2020 October 20, 2020 12) Tangjitsitcharoen, S., In-process prediction of surface roughness by utilizing the cutting force ratio. Transactions of NAMRI/SME.2013(38): 307-315. 13) Tangjitsitcharoen, S. and Senjuntichai, A., Comparison of In-ProcessCutting State Detection in CNC Turning using Different Neural Network Systems. Applied Mechanics and Materials. 2012(121): 1942-1946. 14) Tangjitsitcharoen, S., Rungruang, C. and Pongsathornwiwat, N., Advanced Monitoring of Tool Wear and Cutting States on CNC Turning Process by Utilizing Sensor Fusion. Advanced Materials Research, Trans Tech Publications. 2011(189): 377-384. 15) Tangjitsitcharoen, S., Advanced prediction of surface roughness by monitoring of dynamic cutting force in CNC process. Applied Mechanics and Materials. 2013(239-240): 661-669. 16) Shansungnoen, T. and Tangjitsitcharoen, S., Investigation of relation between straightness and cutting force in CNC turning process. Applied Mechanics and Materials. 2015(789-790) : 812-820. 17) Liotto, G. and Wang, C., Straightness measurement of a long guide way A comparison of dual-beam laser technique and optical collimator. International Symposium on Precision Mechanical Measurements (ISPMM’2004). 2014: 24-28. 18) Ali, S.H.R, Mohamed, H.H. and Bedewy, M.K., Identifying Cylinder Liner Wear using Precise Coordinate Measurements. International Journal of Precision Engineering and Manufacturing. 2019(10-5), 19-25. 129



Investigation of Defect States from Radiative Emissions in Culn1-xGaxSe2 / Cu(In1-xGax)3Se5 Bi-Layer Systems by Photoluminescence Technique Asst. Prof. Dr. Sojipong Chatraphorn



The 28th Special CU-af Seminar 2020 October 20, 2020 Investigation of Defect States from Radiative Emissions in Culn1-xGaxSe2 / Cu(In1-xGax)3Se5 Bi-Layer Systems by Photoluminescence Technique Asst. Prof. Dr. Sojipong Chatraphorn Abstract CuIn1-xGaxSe2 (112-CIGS) and Cu(In1-xGax)3Se5 (135-CIGS) are the photon absorber materials used in CIGS thin film solar cells. It has been reported that a 135-CIGS phase accidentally exists on a Cu-poor 112-CIGS surface of the solar cells. The 135-CIGS has larger band gap energy (Eg) than the 112-CIGS and only cause the valence band offset in the band alignment between 112-CIGS and 135-CIGS which depends on the defect states in both layers. In order to investigate the defect states in these materials of such a structure, an ultra-thin layer of 135-CIGS with various thicknesses will be deposited on the 112-CIGS. A temperature-dependent and an p-dependent photoluminescence (PL) techniques will be used to probe the radiative emissions from the samples. The obtained PL spectra can be used to model the possible defect states in the CIGS materials. The results show that the general defect states detected in 112-CIGS and 135-CIGS/112-CIGS thin films are Cu vacancies (VCu) and Se vacancies (VSe) that are the acceptor defects and donor defects, respectively, whereas, the deep donor defects, i.e., group-III elements on the Cu site (IIICu) is found only in 135-CIGS. Physics of Energy Materials Research Unit, Department of Physics Faculty of Science, Chulalongkorn University Bangkok, Thailand 133

The 28th Special CU-af Seminar 2020 October 20, 2020 Introduction and Objectives CuIn1-xGaxSe2 (CIGS or 112-CIGS) is a chalcopyrite semiconducting material and one of the promising materials for high efficiency thin film solar cells. Its band gap energy can be tuned from 1.02 to 1.66 eV by varying the ratio of group-III elements (x = [Ga]/([Ga]+[In])) that lies within the structure of CuInSe2 (Eg = 1.02 eV) and CuGaSe2 (Eg = 1.66 eV). The performance of high efficiency CIGS solar cells is confronted with the limitation of increasing open-circuit voltage (Voc) and short-circuit current density (Jsc). In other words, these are two competing parameters which are difficult to increase simultaneously. The alternation increase in the Voc with larger band gap energy of CIGS absorber can cause photocurrent loss. On the other hand, the increase in Jscalso causes the reduction in Voc. There are several attempts to increase both Voc and Jsc by tuning of conduction band energy, e.g., front band gap grading, back band gap grading and double grading. However, it is difficult to control the precise fluxes of Ga and In to achieve the conduction band grading. The minimum front grading should be contained within the space charge region (SCR) without losing fill factor (FF) and Jsc [1]. However, there is one technique that can also enhance both Voc and Jsc and can be achieved by the modification of the surface of 112-CIGS absorber by introducing Cu(In,Ga)3Se5 (135-CIGS or also known as ordered defect compound (ODC) or phase for CIGS structure) layer. This phase is unexpectedly found on 112-CIGS absorber, especially Cu-poor film [2]. Theoretically, 135-CIGS is believed to induce the valence band offset at the interface region [3]. There have been several attempts trying to understand and to confirm the existence and effects of this ODC. The doping of both 112-CIGS and 135CIGS are due to defects whose states lie in the band gap. The optical properties measurement is one of various methods to probe the existence and effects of 135-CIGS to the 112-CIGS. In this work, optical transmission and reflection spectroscopy are used to obtain the optical band gap energy (Eg) of the materials while photoluminescence (PL) technique is used to probe the radiative recombination transitions for the defect states or impurity levels of the 112-CIGS, 135-CIGS and 135-CIGS/112- CIGS heterostructures. Experimental 112-CIGS thin films were deposited by molecular beam deposition method under ultra-high vacuum environment using EIKO model EW-100 MBE system and employed the thin film growth technique known as the 3-stage deposition process. Briefly, In, Ga and Se were first deposited onto (600 m) Mo-coated (3cm x 3cm) soda-lime glass (SLG) substrate at 300oC (as measured by a thermocouple behind the substrate) to form (In,Ga)2Se3 precursor, followed by evaporation of Cu and Se at higher substrate temperature of 450oC in the second stage until the overall composition was Cu-rich, e.g., y = [Cu]/([Ga]+[In]) ~ 1.3 in this work. Again, In, Ga and Se were co-evaporated in the third stage at the same substrate temperature as performed in the second stage until the required composition was Cu deficient with y ~ 0.9. Ga composition ratio, i.e., x = [Ga]/([Ga]+[In]) was set at ~ 0.37. The Cu, In and Ga fluxes were calibrated prior to the deposition. We note that Se flux was always over supplied throughout the growth process. The thickness of 112-CIGS films was set at 1.8 m. For the samples with Cu-depleted layer, the 135-CIGS thin films with various thicknesses were deposited onto the 112-CIGS surfaces. In order to limit the diffusion of the constituent elements, the 135- CIGS layers were grown at 400oC within 90–180 seconds after the completion of 112-CIGS thin films. The y and x composition ratios of 135-CIGS capping layers were aimed to be 0.33 and 0.37, respectively.The deposition fluxes of 112-CIGS and 135-CIGS thin film were verified by the energy dispersive x-ray spectroscopy (EDS, Oxford model PentaFET X3). 134

The 28th Special CU-af Seminar 2020 October 20, 2020 The accelerating voltage used for the EDS measurements of all samples was 15 kV which was sufficient for all constituent elements of the samples. UV-VIS-NIR spectrophotometer (Perkin-Elmer model Lambda 900) was used to measure the optical transmittance and reflectance of 112-CIGS, 135-CIGS and 135-CIGS/112-CIGS films in the range of 700 – 1400 nm. For the PL measurements, the films were excited with 35 mW He-Ne laser (632.8 nm) and were performed from 10 – 300 K in a closed cycle liquid - He cryostat. The emission was detected liquid – nitrogen cooled Ge detector together with high resolution monochromator. The power of excitation laser incident on the sample surface was varied by using the continuous neutral density filter wheel. A schematic diagram of PL measurement is shown in Fig. 1. Results and Discussion The optical transmittance of the as-grown 112-CIGS, 135-CIGS and 135-CIGS/112-CIGS with various thicknesses of 135-CIGS was measured through an intentionally unfilled region of Mo layer on the SLG substrate. The reflectance spectra over the same area of the transmittance measurements were obtained using the reflection Fig. 1. Schematic diagram of photoluminescence spectroscopy setup. Fig. 2. Plots of (a) optical transmittance and (b) reflectance spectra of 112-CIGS, 135-CIGS and 135-CIGS/112-CIGS with various thicknesses of 135-CIGS; 10, 20, 40, 80, 200 and 300 nm. Inset in (a) shows the results of the transmittance of physi- cally stacked layers of 112-CIGS and 135-CIGS by laying two samples together. 135

The 28th Special CU-af Seminar 2020 October 20, 2020 measurement assembly of the spectrophotometer. The transmission and reflection spectra are shown in Fig. 2 in the range of 700–1400 nm. The absorption edge significantly shifts toward shorter wavelengths with increasing thickness of 135-CIGS layer as seen in the transmission spectra of Fig. 2(a). It is relatively unclear to indicate the shifts from the reflection spectra in Fig. 2(b). The thickness of all 112-CIGS and 112-CIGS with 135-CIGS capping layer films is approximately 1.8 µm while the thickness of 135-CIGS film is only 1 µm. The transmittance and reflectance were used to determine the absorption coefficient (α) of thin films of thickness d according to where Tmeas and Rmeas are the transmittance and reflectance measured by the spectrophotometer, respectively [4]. We note that expression (1) is an approximation based on the assumption that αd is large near the absorption edge. The energy gap Eg of each sample was obtained using Tauc plot governed by the relation where A is a constant, h v is the incident photon energy [5, 6]. The Eg of direct bandgap semiconductors can be obtained from the extrapolation of the linear section to the intercept on h v axis in the plot of (αhv)2 vs. hv as shown in Fig. 3. The Eg of 112-CIGS and 135-CIGS were found to be ~1.15 and ~1.46 eV, respectively. The Eg of 135-CIGS in this work agrees with the Eg of single crystal and polycrystalline 135-CIGS of 1.40 and 1.49 eV, respectively, reported elsewhere [7, 8]. The Eg of 135-CIGS thin film is about 0.3 eV larger than that of 112-CIGS and agrees with others [9, 10]. The absorption edge of thin 135-CIGS capping layer on 112-CIGS shifts toward higher energy, i.e., from ~1.15 eV (112-CIGS) to ~1.23 eV (300 nm thick of 135- CIGS capping layer). The inset in Fig. 2(a) shows the results of the transmittance of physically stacked layers of 112-CIGS and 135-CIGS by laying two samples together, each of about 1.5 µm thick, e.g., 135-CIGS/SLG on 112-CIGS/SLG (red dash-dotted line), and vice versa (green dashed line). The transmission spectra are almost identical to the transmission of only 112-CIGS (blue solid line) as expected, i.e., 112-CIGS, with sufficient thickness whose Eg 136

The 28th Special CU-af Seminar 2020 October 20, 2020 Fig. 3. Tauc plot of (αhv)2 vs. incident photon energy (h v) for 112-CIGS, 135-CIGS and 135-CIGS/112-CIGS with various thicknesses of 135-CIGS. A straight line is extrapolated to determine band gap energy (Eg).is also less than that of 135-CIGS, absorbs all photons. In other words, 135-CIGS does not contribute to the absorption in these physical stacking configurations. However, when the thin 135-CIGS capping layers were deposited on top of 112- CIGS, the absorption edges shift towards shorter wavelengths. This indicates that Eg increased when the thin 135-CIGS was deposited on top of 112-CIGS. It is in contrast to the physically stacked layers of 135-CIGS and 112-CIGS. The results of Eg are used as a guideline for the range of the wavelength for photoluminescence measurements. Photoluminescence or PL is a technique to detect and identify probable defect states in semiconductor materials. PL describes the photon emission processes after the materials are irradiated with photons. It relies on the creation of electron-hole pairs by incident radiation and subsequent radiative recombination photon emission. Photons of a particular energy that are absorbed or emitted by a sample provide evidence of electronic states differing by that energy within the band gap. The radiative emission intensity is proportional to impurity density [11]. Depending on the defect or impurity, the state forms as a donor or acceptor of excess electrons or holes in the crystal. Electrons or holes are attracted to the excess or deficiency of local charge due to the impurity nucleus or defect, and thus Coulomb binding occurs. Fig. 4 shows a simplified sketch of the possible PL recombination paths for intrinsic and impurity transitions. When the temperature is sufficiently low, carriers will be trapped at these states. If these carriers recombine radiatively, the energy of the emitted light can be analyzed to determine the energy of the defect or impurity level shown in Fig. 4. Shallow levels that lie near the conduction or valence band edge are more likely to participate in radiative recombination. However, the sample temperature must be low enough to discourage thermal activation of carriers out of the traps. Deep levels tend to facilitate non-radiative recombination by providing a stop-over for electrons making their way between the conduction and valence bands by emitting phonons. 137

The 28th Special CU-af Seminar 2020 October 20, 2020 Fig. 4. A diagram showing possible radiative recombination paths in semiconductor materials; (from left to right) interband transition, donor to valence band transition, conduction band to acceptor transition, donor to acceptor transition and conduction band to intermediate state or intermediate state to valence band transition. A black circle and a white circle represent an electron and a hole, respectively. When the sample temperature is decreased from room to low temperature or conversely in PL measurement, the thermal activation energy of the defect state for the emission peak corresponds to where IT is the intensity of photoluminescence peak at temperature T, I0 is the radiative intensity at T = 0 K, kB is the Boltzmann’s constant (1.38 x 10-23 J/K), C is a constant and ∆E is the thermal activation energy of the donor or acceptor. In the case of donor-acceptor pair (DAP) transition in the semiconductor materials, the activation energy levels of the donor and acceptor can be written by where Eg is the band gap energy, ED and EA are the donor and acceptor ionization energies, respectively. The last term corresponds to the Coulomb’s interaction between the pair, r is the distance between the donor and acceptor that involves in the transition, ε0 and εr are the dielectric constant of the vacuum and material, respectively . When r is small, individual pair line may be observed. When r becomes larger, the emission usually merges into a broad band. 138

The 28th Special CU-af Seminar 2020 October 20, 2020 The peak energy of the broad band corresponds to the suitable separation of the pair. The transition from donor levels to valence band or acceptor levels to conduction band can be related to where ED,A represents the activation energy of either donor (D) or acceptor (A). This is known as the free to bound transition (FB). It can be referred to a free electron recombines with a hole bound to an acceptor or a free hole recombine with a donor bound to an electron. In this work, temperature-dependent and power-dependent PL spectra of 112-CIGS, 135-CIGS, 135-CIGS/112-CIGS thin films were investigated as followed. The excitation power dependence of PL spectra of 112-CIGS, 135-CIGS and 135-CIGS/112-CIGS with 10 thick 135-CIGS from 0.1 mW to 35 mW at 10 K are shown in Fig. 5. The PL spectra show broad donor-to-acceptor pairs (DAPs) transition for 112-CIGS and 135-CIGS/112-CIGS films. It can be seen in Fig. 5 that the broad emission in the PL spectra of 112-CIGS and 135-CIGS at high excitation power can be resolved into two peaks when the excitation intensity is reduced. The two resolved peaks at low excitation are identified as donor-to-acceptor pairs (DAPs) and free (conduction band)-to-bound (acceptor) (FB) transitions. Normally, the nature of peak position of the FB transition is temperature and excitation power independent, whereas the DAPs showed blue-shift for a p-type and red-shift for an n-type semiconductor [12]. It is worth to point out that the broad peak of 135-CIGS cannot be resolved to observe FB and DAP transitions when the excitation power is decreased. There is no significant shift of the peak in the excitation power for the 135-CIGS. In addition, the emission from the 135-CIGS cannot be noticed in the 135-CIGS/112-CIGS heterostructure. Fig. 6 show the temperature-dependent PL spectra of 112-CIGS, 135-CIGS/112-CIGS with 10 nm thick 135-CIGS from 10–300 K and 135-CIGS from 10–200 K. Inset is the fitting result of PL spectrum at 10 K. The broad peaks are observed in range of 1.07–1.12 eV for 112-CIGS and 1.09–1.14 eV for 135-CIGS/112-CIGS which the energy range for 10 nm thick 135-CIGS capping on 112-CIGS is slightly shifted to higher energy about 20 meV. The peak energy increases by reducing of Cu elements for 135-CIGS thin film. With increase in temperature, the transition to higher energy is apparent. The peak positions of the PL measurements from 10–300 K at maximum excitation power are plotted as a function of temperature and shown in Fig. 7. The PL emissions from 112-CIGS and 135-CIGS/112-CIGS show blue-shift as the temperature increases that are the nature of the DAP transition of a p-type semiconductor. It is noted that the peak energy at 300 K coincides with the FB transition at 10 K. On the contrary, a 135-CIGS shows red-shift as the temperature increases that is a characteristic of an n-type semiconductor. The activation energy can be calculated from the two-channel model of Arrhenius function [13]; 139

The 28th Special CU-af Seminar 2020 October 20, 2020 where I(0) is the PL emission intensity at 0 K, Ea1 and Ea2 are the activation energies in the low and high temperature regime, respectively, A and B are fitting parameters and kB is the Boltzmann constant. The results of Arrhenius fitting to obtain Ea1 and Ea2 are shown in Fig. 8. They are used to model the defect states as shown in Fig. 9. Fig. 5. The PL spectra of 135-CIGS, 112-CIGS and 135-CIGS/112-CIGS thin films at 10 K. The black line is for the high power (35 mW) and the red, green and blue lines are for lower excitation power. Fig. 6. Temperature dependent PL spectra of 112-CIGS, 135-CIGS/112-CIGS and 135-CIGSthin films at 10 K. Inset is the fitting result of PL spectrum at 10 K using Gaussian distribution. 140

The 28th Special CU-af Seminar 2020 October 20, 2020 Fig. 7. Peak energy at high excitation power plotted as a function of temperature from 10–300 K for 112-CIGS, 135-CIGS and 135-CIGS/112-CIGS. Fig. 8. Arrhenius plot of temperature-dependent PL intensity vs. 1/T for fitting peak 1 (middle) and peak 2 (right) of 112-CIGS and 135-CIGS/112-CIGS. The PL peaks at 10 K (left) are only shown for examples of fitting. Ea1 and Ea2 are the activation energies that correspond to the VSe below CB and VCu above VB, respectively. 141

The 28th Special CU-af Seminar 2020 October 20, 2020 Fig. 8. (Continued) Arrhenius plot of temperature-dependent PL intensity vs. 1/T for fitting peak 1 (middle) and peak 2 (right) of 112-CIGS and 135-CIGS/112-CIGS. The PL peaks at 10 K (left) are only shown for examples of fitting. Ea1 and Ea2 are the activation energies that correspond to the VSe below CB and VCu above VB, respectively. Fig. 9. The energy level diagram of (a) 112-CIGS (b) 135-CIGS/112-CIGS and (c) 135-CIGS thin films. The types of donor and acceptor in the 112-CIGS and 135-CIGS/112-CIGS could be regarded as Se vacancies (VSe) and Cu vacancies (VCu), respectively, and the estimated activation energies are (Ea1) ~10–30 meV for VSe and (Ea2) ~50–90 meV for VCu, as estimated from the fitting in Fig. 8, compared with the values of 5–10 meV for VSe and 80–90 meV for VCu reported elsewhere [14]. The slight discrepancies are due to the difference of the Ga concentration in this 142