Trap emission by nanocrystalline anatase in visible range studied by conventional and synchronous luminescence spectroscopy: Adsorption and desorption of water vapor

Journal of Luminescence 192 (2017) 388–396

Contents lists available at ScienceDirect

Journal of Luminescence journal homepage: www.elsevier.com/locate/jlumin

Trap emission by nanocrystalline anatase in visible range studied by conventional and synchronous luminescence spectroscopy: Adsorption and desorption of water vapor

MARK

Alexander Samokhvalov Chemistry Department, Rutgers University, 315 Penn St., Camden, NJ 08102, USA

A R T I C L E I N F O

A B S T R A C T

Keywords: Anatase Luminescence Synchronous Water Sorption Midgap

We report the spectroscopic study of adsorption and desorption of water vapor on nanocrystalline anatase in air under ambient conditions. We also report the energy level diagram of absorption and emission through the three electronic midgap states of different origin in anatase as determined by absorption spectroscopy, “conventional” PL emission spectroscopy and, for the first time, by synchronous luminescence spectroscopy at 25 °C. Under the extrabandgap photoexcitation, visible luminescence from nanocrystalline anatase contains peaks of “blue light” at ca. 450 nm and “green light” at ca. 500 nm due to the two different midgap states. Photoluminescence in visible range at 25 °C is quenched after water vapor adsorption with formation of “hydrated” anatase and increased in “dried” anatase after water desorption. For both “dried” anatase with strong emission of visible light and “hydrated” anatase with weak (quenched) emission, synchronous luminescence spectra are superior to “conventional” PL emission spectra in resolving characteristic emission peaks due to midgap states. Photoexcitation of self-trapped exciton (STE) as intrinsic electronic midgap state in anatase at 420 nm results in a strong emission of “green light” at 500 nm, which is preferentially quenched upon water vapor adsorption vs. “blue light” at 450 nm. Visible PL in anatase decreases when the average nanocrystal size increases within 5–30 nm range, suggesting participation of surface midgap states in water adsorption/desorption.

1. Introduction Nanocrystalline titanium dioxide can be present as anatase, rutile or brookite, the least abundant and studied form of TiO2. Rutile is thermodynamically the most stable phase of “bulk” titanium dioxide [1], so rutile is formed at high temperatures and is the most naturally abundant form of TiO2. On the other hand, anatase is more stable than rutile as small nanoparticles ca. < 15 nm in size [1]. Nanocrystalline titania containing anatase has been studied as sorbent in aqueous [2] and nonaqueous [3] solutions, photocatalyst in aqueous [4] and non-aqueous [5] solutions, photocatalyst [6] and sorbent in the gas phase [7], as material for solar cells [8], electronic [9] and biomedical [10] devices, in chemical sensors [11] including humidity sensors [12], etc. For instance, the benchmark TiO2 photocatalyst P25 Degussa aka Evonik consists of ca. 80% anatase and 20% rutile with average nanocrystal size of about 20 nm. Anatase in its “bulk” form has an indirect bandgap of 3.2 eV; anatase is considered photocatalytically more active than rutile, despite the smaller optical bandgap of the latter. Electronic states in the bandgap of semiconductor (often termed “midgap states”) participate in relaxation and transfer of excited charge

E-mail address: [email protected]. http://dx.doi.org/10.1016/j.jlumin.2017.07.009 Received 16 March 2017; Received in revised form 21 June 2017; Accepted 8 July 2017 Available online 10 July 2017 0022-2313/ © 2017 Elsevier B.V. All rights reserved.

which affects functioning of electronic devices [13], rates of photocatalytic reactions [14], etc. The midgap states affect rates of photocatalytic reactions via an improved absorption of visible light and an enhanced rate of electron-hole recombination [15]. “Engineering” of midgap states is promising for development of new functional materials; for example, nanocrystalline anatase was chemically reduced to “black titania” containing surface oxygen vacancies V(O) and Ti3+ sites [14], and the latter has shown an improved absorption of visible light and enhanced photocatalytic hydrogen generation rate. The photoluminescence (PL) spectroscopy is one of the best methods to learn about electronic states in the nanocrystalline semiconductors and about midgap states in particular, due to its non-destructive character and high sensitivity. The PL from midgap states in semiconductors is often named “trap emission”. Water molecule is a common adsorbate on metal oxides, and adsorption/desorption of water may strongly affect the properties of nanocrystalline functional materials. The increase or decrease (quenching) of the PL in nanocrystalline anatase after adsorption and desorption of water vapor was not reported, to our knowledge. In the most frequently used “conventional” PL emission

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

reference. Chemical composition was verified by X-Ray Fluorescence (XRF) using the instrument MESA-50 (Horiba Scientific).

spectroscopy, the spectrum is obtained by scanning the emission wavelength λemiss under the photoexcitation at a chosen constant photoexcitation wavelength λexc. Unfortunately, the “conventional” PL emission spectra of nanocrystalline solids, including titanium dioxide are often poorly resolved at room temperature and feature a significant peak broadening [16] which makes interpretations difficult. For compounds in solution, it is usually possible to achieve a significant narrowing of emission peaks by using synchronous luminescence spectroscopy [17]. The experiment by synchronous luminescence spectroscopy is performed by varying simultaneously (“synchronously”) both the photoexcitation λexc and emission λemiss wavelengths at a constant difference termed “the delta lambda parameter” Δλ = λemiss - λexc. Recently, synchronous fluorescence spectroscopy of compounds in solution has been critically reviewed [18,19]. To our knowledge, synchronous luminescence spectra of anatase were not reported. Recently, we reported a mechanistic study of adsorption of aromatic and heterocyclic compounds in suspension using “conventional” PL emission spectroscopy [20–22]. Herein, we report characterization of nanocrystalline anatase by “conventional” in comparison to synchronous luminescence spectroscopy in visible range, at 25 °C in air under atmospheric pressure. We report the pathways of relaxation and recombination of excited charge in nanocrystalline anatase through the three specific electronic midgap states, some of which are strongly and preferentially modified by periodic adsorption/desorption of water. We also studied how visible photoluminescence in anatase changes when the nanocrystal size changes in the range of 5–30 nm.

2.4. Measurements by “conventional” PL spectroscopy in the solid phase at 25 °C The photoluminescence (PL) spectra were collected using spectrometer Fluorolog FL3-22 (Horiba Scientific). This instrument is equipped with dual monochromator gratings on the excitation and emission light paths. The instrument was calibrated daily using the Raman signal of liquid water [16] in quartz cuvette at 25 °C. To minimize optical artifacts due to primary and secondary absorption of light in solid specimen [16], all spectra were obtained in the Front Face (FF) geometry using the FL-1001 accessory (Horiba Scientific). To eliminate the consequences of fluctuations of intensity of the photoexcitation source, the emission signal S from the sample has been divided by the reference signal R generated by the excitation beam before reaching the sample, and the (S/R) ratio has always been used as an Y axis of the spectra. All PL spectra were collected at 25 °C with specimen packed into a cleaned and dried cuvette, closed with polytetrafluoroethylene (PTFE) stopper, and sealed with Parafilm tape to protect from ambient moisture. The photoexcitation wavelength λexc was varied from 340 nm to 500 nm at the 10 nm increments. The PL emission spectra were recorded with excitation and emission slits at 5 nm. 2.5. Measurements by synchronous luminescence spectroscopy in the solid phase at 25 °C

2. Experimental Synchronous luminescence spectroscopy measurements were conducted using at same Fluorolog FL3-22 spectrometer in the FF geometry described above. The (S/R) ratio was used to construct the Y axis of all synchronous luminescence spectra, and the specimen was in the same container and physical form as in “conventional” PL measurements. The Δλ parameter, Δλ = λemiss - λexc has been varied at the 10 nm increments. Synchronous luminescence spectra were recorded with excitation and emission slits set at 5 nm. Numeric curve fitting was conducted with Microcal Origin 2015 program.

2.1. Materials Nanopowders of anatase with average nanocrystal size of 30 nm (99.98% purity), 15 nm (99.5% purity) and 5 nm (99.5% purity) were obtained from the U.S. Nanomaterials. Sulfuric acid and hydrogen peroxide were obtained from Sigma. 2.2. Drying and hydration of nanocrystalline anatase The as-obtained material was denoted asisAnatase. A quartz cuvette of 1.5 cc was cleaned with Piranha solution (sulfuric acid and hydrogen peroxide), rinsed well with distilled water, and dried. The asisAnatase with the given nanocrystal size was placed into a cleaned cuvette and dried at 110 °C in the oven in air overnight, resulting in a dried material denoted drAnatase. After drying, a cuvette with the specimen was removed from the oven, promptly closed with a polytetrafluoroethylene (PTFE) stopper, weighed, and sealed with Parafilm to protect the specimen from ambient moisture for the following spectroscopic measurements. Next, the obtained drAnatase was allowed to adsorb water vapor (“hydration”) using a procedure recently reported by us [21]. Namely, the specimen in this cuvette has been placed above the water/ air meniscus inside a tightly closed dessicator with liquid water, and kept in contact with water vapor at relative humidity RH ~ 100% at 25 °C overnight, yielding a “hydrated” material denoted hyd-drAnatase. When testing the stability of nanocrystalline anatase at the higher temperature (Table S1), the specimen was dried at 150 °C and hydrated as described above, and then dried again at 150 °C resulting in dr-hyddrAnatase.

3. Results and discussion 3.1. Characterization of nanocrystalline anatase In order to determine the bandgap in the nanocrystalline semiconductor, the Tauc plots are very useful. The Tauc formula is (α × E)1/ n = E – Eg where E is the photon energy, Eg is the bandgap, α is an absorption coefficient, and n is the Tauc constant [23]. For the nanocrystalline semiconductors, the Kubelka-Munk (KM) function applies F (R) = (1−R)2/2R = k/s, where R is optical reflectance R(λ), k is an absorption coefficient, and s is the scattering constant. The Tauc plots can be conveniently obtained using the KM function (F(R) × E)1/n = E – Eg, where n = 2 is for an indirect optical transition with energy Egindirect. Anatase is commonly believed to be an indirect bandgap semiconductor, e.g. in recent reports by experiment [24] and density functional theory (DFT) calculations [25]. In contrast to direct transition, indirect transition requires lattice vibrations (phonons), so indirect bandgap is often found in optical experiments conducted at room temperature. The magnitude of quantum size effect in nanocrystalline anatase has been the topic of research. Some authors believed that there was no quantum size effect in anatase [26,27], while others found more recently weak quantum size effect based on measurements by optical spectroscopy [24,28]. Fig. 1 shows the Tauc plots to determine Egindirect in anatase of different nanoparticle size. From Fig. 1, the following bandgaps were obtained: Egindirect (5 nm) = 3.30 eV or 375 nm; Egindirect (15 nm) = 3.28 eV or 378 nm, and Egindirect (30 nm) = 3.22 eV or 385 nm. The obtained values are consistent with reported weak

2.3. Sample characterization The XRD patterns were obtained by using Rigaku SmartLab diffractometer with Cu K-alpha line at 0.15418 nm. The UV–Visible diffuse reflectance spectra (UV–Vis DRS) were collected in the reflectance mode at room temperature using Cary 5000 spectrophotometer which is equipped with Praying Mantis attachment (Harrick Scientific). A finely ground BaSO4 of 99.998% purity (Alfa Aesar) was used as white 389

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

Fig. 1. The Tauc plots of asisAnatase with different nanocrystal size. a) 5 nm. b) 15 nm. c) 30 nm.

Fig. 2. The photoexcitation wavelength dependent PL emission spectra of drAnatase at 25 °C. a) Extrabandgap excitation. b) Subbandgap excitation. c) Low-energy subbandgap excitation.

quantum size effect in anatase. The Egindirect (30 nm) = 3.22 eV (Fig. 1c) is close to reported 3.239 eV for anatase with large (29 nm) nanocrystals having a “bulk” bandgap [24]. The Egindirect (5 nm) = 3.30 eV (Fig. 1a) is by 0.08 eV larger than a “bulk” value (Fig. 1c), consistently [24] with reported Egindirect in small anatase nanoparticles (3.8 nm) which exceeds “bulk” Egindirect by ca. 0.05 eV. The Egindirect (15 nm) = 3.28 eV (Fig. 1b) is closer to “bulk” value than Egindirect (5 nm) as expected. In Fig. 1a–c, there is a weak “tail” at photon energy E < Egindirect which is the Urbach tail [29] indicative of midgap states in nanocrystalline anatase. The XRD pattern of asisAnatase (data not shown) corresponds to anatase lattice, PDF 01-084-1285.

amount of water on anatase (101) surface, where the multilayer adsorption occurs and the interface formation energy has a minimum at 3 ML coverage [30]. For spectroscopic studies of nanocrystalline anatase before and after adsorption and desorption of water vapor, we selected anatase with an intermediate nanocrystal size of 15 nm. This choice is convenient, since it a) shows the quantum size effect in comparison to the larger (30 nm) anatase (Fig. 1b), and b) does not suffer from the mass instability in contrast to the smaller (5 nm) anatase (Table S1). In order to investigate radiative charge relaxation, a set of 2-dimensional PL emission spectra is useful with variable photoexcitation wavelength λexc aka the excitation wavelength dependent PL emission spectra [21]. Fig. 2a–c show the photoexcitation wavelength dependent PL emission spectra of anatase dried at 110 °C, drAnatase with 15 nm nanocrystals. An empty quartz cuvette (specimen container) has shown a negligibly small emission as expected (spectra not shown). In Fig. 2, the Rayleigh lines at λexc are partially shown to aid in spectral assignments (labeled “Rayleigh”). At the extrabandgap photoexcitation (Fig. 2a), the PL emission is weaker when the λexc are shorter (at the higher excitation energy). When the energy of the photoexcitation decreases and approaches an indirect bandgap at 3.28 eV (378 nm), the PL intensity steadily increases. However, the highest intensity of photoluminescence in anatase is achieved at the subbandgap photoexcitation with λexc = 420 nm or 2.95 eV (Fig. 2b, a thick blue line). When the λexc further increases (Fig. 2c) the PL intensity decreases. The wavelength of excitation λexc = 420 nm (2.95 eV) observed by us at 25 °C is very close to the wavelength of photoluminescence from self-trapped exciton

3.2. Adsorption and desorption of water vapor by nanocrystalline anatase First, we have tested the stability of nanocrystalline anatase in the “drying/hydration” cycle at drying temperature of 150 °C, using the specimens with nanocrystal size 5 nm, 15 nm and 30 nm, Table S1. Upon hydration and the 2-nd drying, the changes in mass have been nearly constant for anatase with nanocrystals of 15 nm and 30 nm, while the mass change has fluctuated for the smallest (5 nm) nanocrystals. For anatase with an intermediate nanocrystal size (15 nm), the specimen of 0.482 g (6 mmol) as in Table S1 adsorbs and desorbs 0.027 g water (1.5 mmol), and has surface area of ca. 60 m2/g (29 m2). The diameter of water molecule at 2.75 Å yields that one monolayer (1 ML) of adsorbed water has a nominal total volume of 8 × 10−9 m3/ specimen. The adsorbed amount of 0.027 g water/specimen has a nominal volume of 27 × 10−9 m3 assuming the density of water at 1 g/ cm3, thus the total amount of adsorbed/desorbed water corresponds to ca. 3 ML coverage. This finding is consistent with computed adsorbed 390

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

described as an intrinsic localized midgap state [33], in which the density of an excited electron is “spread” over one TiO6 octahedron (at very low temperatures) or over few neighboring TiO6 octahedra. Therefore, a strong emission of visible light by drAnatase at 25 °C occurs after photoexcitation of an electron at the valence band maximum (VBM) to the STE (a high energy intrinsic midgap state). The subbandgap photoexcitation of the STE is consistent with the Urbach's tail in the Tauc plot of nanocrystalline anatase (Fig. 1b). The PL spectra in Fig. 2 contain emission shoulders (shown by arrows) at ca. 450 nm and 500 nm; Fig. 3a shows a multi-Gaussian peak fitting of the PL emission spectrum of drAnatase at the photoexcitation energy close to an indirect bandgap. An adjusted fitting parameter R2 = 0.999 indicates a good quality of fitting; using the Lorentian model resulted in poor fit quality. There are peak #1 at 458 nm with the full width at the half maximum FWHM = 40 nm (“blue light”) and peak #2 at 495 nm with the FWHM = 44 nm (“green light”). In general, the PL emission spectra of anatase may contain peaks due to self-trapped exciton [31], “bulk” oxygen vacancies [34] denoted V(O)bulk and surface states including surface oxygen vacancies V(O)surf e.g. [35]. The recent computational work [36] predicts the three kinds of V(O) for anatase (101) and anatase(001) surfaces: “bulk” oxygen vacancies V(O)bulk, surface oxygen vacancies V(O)surf and subsurface oxygen vacancies V (O)sub. Based upon “conventional” PL emission spectra, we assign the peak #1 (“blue light”) at 458 nm to emission by “shallow trap” formed by the F center (neutral oxygen vacancy) in vicinity of Ti3+ site following Refs. [32,37], and emission peak #2 at ca. 500 nm (“green light”) to emission by “deep trap” formed by the F+ center (oxygen vacancy with one electron). These assignments based on data by photoluminescence spectroscopy [32,37] are consistent with studies of F center and F+ center in anatase by UV–Vis absorption spectroscopy [34]. The emission peak similar to peak #2 (“green light”) was reported [35] in the PL emission spectrum of P25 TiO2 (containing mainly anatase) at 529.5 nm when measured at 77 K, and assigned to surface oxygen vacancy V(O)surf. Localization of these midgap states was determined by synchronous luminescence spectroscopy (see below). Scheme 1 shows the energy level diagram for transitions in dried nanocrystalline anatase under the extrabandgap and subbandgap photoexcitation as determined from”conventional” PL emission spectra. Electrode potential of the

Fig. 3. The PL emission spectra of drAnatase at 25 °C. a) At λexc= 390 nm; b) Excitation of the STE at λexc= 420 nm.

(STE) state in anatase nanoparticles at 425 nm at room temperature [31], as well as the wavelength of luminescence from the STE in TiO2 nanotubes at 415 nm at room temperature [32]. The STE in anatase is

Scheme 1. The energy level diagram of transitions in dried nanocrystalline anatase. a) Extrabandgap photoexcitation. b) Subbandgap photoexcitation.

391

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

Fig. 5. Emission spectra of drAnatase at 25 °C. a) “Conventional” PL emission spectra. b) Synchronous luminescence spectra. Fig. 4. The PL emission spectra of a 15 nm anatase at 25 °C. a) The asisAnatase. b) The drAnatase after drying. c) The hyd-drAnatase after the following hydration.

artifacts of adsorbed water (labeled “R”). We decided to systematically compare the capabilities of synchronous luminescence vs. “conventional” PL emission spectroscopy (Section 3.3) in resolving characteristic emission peaks from midgap states in dried anatase.

conduction band minimum (CBM) in the nanocrystalline anatase at pH = 7 was taken at ca. − 0.5 V vs. normal hydrogen electrode (NHE) as in Refs. [38,39]. It would be desired to use the photoexcitation of the STE in anatase to learn about midgap states yielding “blue light” and “green light” at room temperature. However, the PL emission spectrum in Fig. 3b at λexc= 420 nm is less resolved (numeric fitting with biGaussian function was not successful) than in Fig. 3a; only emission peak #2 (“green light”) with a maximum at ca. 510 nm is observed. Attempts to resolve both peak #1 and peak #2 in the PL emission spectrum at λexc= 420 nm by collecting the spectra with decreased optical slit widths (at 3 nm) were not successful. The two possibilities arise: a) emission of “blue light” does not proceed upon photoexcitation of the STE or b) “conventional” PL emission spectroscopy is unable to detect emission peak #1 upon photoexcitation of the STE. To learn further about visible emission peaks #1 and #2 in anatase, we have used the three approaches: a) water adsorbate as probe molecule plus “conventional” PL emission spectra (Fig. 4); b) synchronous luminescence spectroscopy without water molecule (Figs. 5 and 6 in Section 3.3) and c) synchronous luminescence spectroscopy plus water as probe molecule (Figs. 7 and 8 in Section 3.3). Fig. 4 shows the “conventional” PL emission spectra in visible range for a 15 nm anatase under the subbandgap excitation at variable λexc upon drying at 110 °C and subsequent hydration (excitation and emission slits at 5 nm). Upon drying, there was a factor × 2 increase in the PL emission. Upon the following adsorption of water vapor in air (“hydration”), there was a factor × 2 decrease (quenching) of the PL signal. Weak “conventional” PL emission of hyd-drAnatase does not allow learning about emission from the STE or about emission peaks #1 and #2 due to strong Raman

3.3. Adsorption/desorption of water vapor by nanocrystalline anatase by “conventional” PL emission vs. synchronous luminescence spectroscopy Fig. 5 shows emission spectra of drAnatase with 15 nm nanocrystals (dried at 110 °C) as “conventional” PL spectra at variable excitation wavelengths λexc (Fig. 5a) and synchronous luminescence spectra at variable Δλ parameter (Fig. 5b). The photoexcitation and emission slits were set at 5 nm in both cases for comparison. In Fig. 5a, the highest PL emission is achieved at λexc = 420 nm (thick blue line) upon photoexcitation of the STE in anatase. In Fig. 5b, the highest synchronous luminescence signal is achieved at Δλ = 80 nm (thick blue line); for each Δλ, an empty quartz cuvette (specimen container) has shown a negligibly small signal (data not shown). In “conventional” PL emission spectra (Fig. 5a), peaks #1 and #2 appear as poorly resolved shoulders (labeled with arrows). In addition, in Fig. 5a the high-energy sides of the PL spectra are obscured with strong Rayleigh excitation lines (labeled Rayleigh). Synchronous luminescence spectra (Fig. 5b) have an advantage that the Rayleigh lines are absent which allows more accurate analysis, particularly fitting the high-energy spectral part. In addition, emission peaks are better resolved in synchronous luminescence spectra. Fig. 6a shows “conventional” PL emission spectrum of the highest intensity (excitation of the STE at λexc = 420 nm) and Fig. 6b shows the synchronous luminescence spectrum of the highest intensity (with Δλ 392

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

Fig. 6. The highest intensity luminescence spectra of drAnatase at 25 °C. a) “Conventional” PL emission spectrum. b) Synchronous luminescence spectrum.

Fig. 7. Emission spectra of hyd-drAnatase at 25 °C. a) “Conventional” PL emission spectra. b) Synchronous luminescence spectra.

= 80 nm) for drAnatase. While “conventional” PL emission spectrum at λexc = 420 nm seems to show weakly resolved shoulders, our attempts to numerically fit them were not satisfactory. The synchronous luminescence spectrum in Fig. 6b is of lower intensity than the spectrum in Fig. 6a, but the former is better resolved into shoulders which were fitted with a multi-Gaussian function. In Fig. 6b, a shoulder at λemiss = 450 nm (“blue light”) originates from photoexcitation at λexc = λemiss – Δλ = 450 − 80 nm = 370 nm, i.e. at the indirect bandgap in anatase. Therefore, this synchronous luminescence peak is labeled “BG to #1″, i.e. excitation across bandgap leading to emission from midgap state #1 (“blue light”). Similarly, the maximum in synchronous luminescence spectrum at λemiss = 500 nm (“green light”) originates from photoexcitation at λexc = λemiss – Δλ = 500 − 80 nm = 420 nm, i.e. self-trapped exciton (STE) in anatase. This synchronous luminescence peak is labeled “STE to #2”, i.e. excitation of the STE leading to emission from the midgap state #2 (“green light”). In Fig. 6b, there is also a weakly resolved shoulder at ca. λemiss = 550 nm which gives λexc = λemiss – Δλ = 550 − 80 nm = 470 nm. The λexc = 470 nm is close to the wavelength of excitation of “blue light”, while λemiss = 550 nm apparently reflects emission from some shallow midgap states. This emission peak is of low intensity in both “conventional” and synchronous luminescence spectra; its studies are in progress. Numeric multi-Gaussian fitting of synchronous luminescence spectrum at Δλ = 100 nm (data not shown) also reveals emission peaks for transitions “BG to #1” and “STE to #2”. For hyd-drAnatase after adsorption of water vapor, Fig. 7a shows “conventional” PL emission spectra at the same λexc as in Fig. 5a, and Fig. 7b shows synchronous luminescence spectra at the same Δλ parameter as in Fig. 5b. After hydration of drAnatase in moist air towards hyd-drAnatase,

the “conventional” PL emission spectra have been strongly quenched (compare Fig. 7a and Fig. 5a). In Fig. 7a the Rayleigh lines and strong Raman peaks of adsorbed water (labeled “R”) dominate and the characteristic peaks #1 (“blue light”) and #2 (“green light”) of anatase are not observed, so numeric peak fitting could not have been performed. As expected, synchronous luminescence spectra (Fig. 7b) do not contain Rayleigh or Raman lines, and emission peaks are observed. It would be of interest to examine how synchronous luminescence spectra of nanocrystalline anatase change after adsorption of water vapor. Fig. 8a shows synchronous luminescence spectrum at Δλ = 80 nm for drAnatase (solid blue line) in comparison to that of hyd-drAnatase (dashed blue line), and Fig. 8b shows the ratio of these spectra. For λemiss < 450 nm, the ratio is close to unity i.e. no quenching of emission peak #1 (“blue light”) by adsorbed water occurs. This finding is consistent with assignment of emission peak #1 to “shallow trap” formed by the F center (neutral oxygen vacancy) in vicinity of Ti3+ site [32,37] and located in the “bulk” of anatase nanocrystal. Indeed, it is unlikely to expect the Ti3+ site on surface of anatase in our studies, which were conducted in air at 25 °C or at the elevated temperature. In contrast, synchronous luminescence spectrum at λemiss > 450 nm including the peak #2 at ca. 500 nm (“green light”) is quenched after adsorption of water vapor, as the ratio of spectra is close to 2. The preferential quenching of visible emission peak #2 by water vapor adsorbed at 25 °C indicates that electronic states due to emission peak #2 (“deep trap”) are on surface of anatase. This finding is consistent with assignment of peak #2 in dried anatase to surface oxygen vacancy V(O)surf as F+ center using “conventional” PL emission spectra (Fig. 3) and Refs. [32,37]. Thus, synchronous luminescence spectra are much better suited than “conventional” PL emission spectra to observe characteristic visible emission from the particular midgap states of 393

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

Fig. 8. a) Synchronous luminescence spectra of drAnatase and hyd-drAnatase with Δλ = 80 nm at 25 °C. b) The ratio of these spectra.

Fig. 9. Synchronous luminescence spectra of drAnatase at 25 °C. a) Spectra at variable Δλ. b) The first derivative of spectrum at Δλ = 110 nm.

anatase, even when a strong quenching of photoluminescence by water adsorbate occurs.

Table 1 The wavelengths λexc to excite the STE in drAnatase determined from synchronous luminescence spectra.

3.4. Determination of energy of self-trapped exciton in nanocrystalline anatase by synchronous luminescence spectroscopy We have found from “conventional” PL emission spectra (Fig. 2) that the strongest emission of visible light by nanocrystalline anatase occurs upon the subbandgap photoexcitation of self-trapped exciton (STE) state. To our knowledge, synchronous luminescence spectroscopy was not reported for determination of energy of the STE in anatase or other semiconductors. Fig. 9a shows synchronous luminescence spectra of drAnatase with variable Δλ at 25 °C; large values of Δλ > 100 nm are suitable for detection of “green light” (emission peak #2). The two peaks in each spectrum in Fig. 9a correspond to photoexcitation across an indirect bandgap (a shoulder at the shorter wavelengths) and photoexcitation of the STE (major peak at the longer wavelengths). For instance, the shoulder at λemiss = 470 nm for Δλ = 100 nm in Fig. 9a corresponds to λexc = λemiss - Δλ = 470 nm − 100 nm = 370 nm, i.e. photoexcitation across an indirect bandgap (“BG to #1”). When the Δλ increases in Fig. 9a, spectral maximum shifts to the longer wavelengths; the maxima in synchronous luminescence spectra were determined through their first derivatives. Fig. 9b shows the 1-st derivative of the spectrum at Δλ = 110 nm, where the maximum of luminescence is at λemiss = 524 nm. We find λexc = λemiss - Δλ = 524 nm −110 nm = 414 nm, which is close to λexc = 420 nm to photoexcite the STE as determined from the set of “conventional” photoexcitation wavelength dependent PL emission spectra (Fig. 2). Numeric values of λexc determined by the first derivatives of synchronous luminescence spectra at different Δλ are shown in Table 1.

Δλ, nm

The λemiss at a maximum, nm

λexc = λemiss - Δλ, nm

90 100 110 120 130 140 150

508 518 524 536 544 557 564

418 418 414 416 414 417 414

The advantage of synchronous luminescence spectroscopy is that a desired value of λexc is obtained upon analysis of only one synchronous luminescence spectrum vs. multiple PL emission spectra at the variable photoexcitation wavelength (Fig. 2). 3.5. Visible emission from midgap states of anatase with different nanocrystal size We have studied how visible emission from anatase depends on the nanocrystal size. While photoexcitation of self-trapped exciton (STE) in nanocrystalline anatase results in the strongest emission of visible light (Fig. 2), the STE is a localized excited state [33] with electron density spread over just few adjacent TiO6 octahedra. Therefore, photoexcitation of the STE and the following electron transfer to the respective emissive midgap states proceeds locally, mostly within the “bulk” of the nanocrystal. On the other hand, upon the photoexcitation across an indirect bandgap, the photoexcited electron can travel at the conduction band minimum (CBM) to reach emissive midgap states on the 394

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

had resulted in the apparent instability of mass of anatase with the smallest nanocrystal size (5 nm). To avoid instability, the drying temperature (for water desorption) was set at 110 °C. Fig. 10a shows the PL emission spectra of drAnatase with λexc = 370 nm (extrabandgap photoexcitation). In Fig. 10a, blue and green arrows show emission shoulders at ca. 450 nm (“blue light”) and 500 nm (“green light”) present in all three samples of anatase, and the smallest nanocrystals (5 nm) with the largest surface area showed the strongest PL emission. Fig. 10b shows integrated PL spectra (within λemiss = 400–565 nm) divided by mass of drAnatase in each case. Again, the smallest nanocrystals of 5 nm showed the strongest PL emission signal. When the integrated PL emission (without normalization by mass of drAnatase) was plotted against the nanocrystal size, the decreasing function was also observed (data not shown). These results are consistent with reported stronger photoluminescence in nanocrystalline anatase with larger nanoparticles in the size range 12.3–25.4 nm at room temperature [40]. This suggests that initial states for visible emission are due to electronic midgap states on surface of anatase which are quenched by adsorbed water, Scheme 2. The molecule of water adsorbate would interact with surface midgap state, oxygen vacancy V(O)surf by donating electron density, and the increase in the Fermi level causes population of previously vacant midgap states with electrons. Hence, emission of “green light” by the midgap state upon photoexcitation of the STE would be quenched by adsorbed water. Upon desorption of water, the Fermi level would be lowered, resulting in the recovery of “green” photoluminescence. The reported quenching and increase of visible luminescence in nanocrystalline anatase after adsorption/desorption of water vapor at room temperature in air can be used in development of new humidity sensors, sorbents in the gas and vapor phase, heterogeneous photocatalysts, etc. More importantly, the superior sensitivity and resolution of synchronous luminescence spectroscopy of midgap states in anatase under ambient conditions in detecting weak emission peaks has a significant potential for advanced spectroscopic characterization of other metal oxides, other semiconductors, and possibly nanocrystalline functional materials beyond semiconductors. Further, the combination of synchronous luminescence spectroscopy under ambient conditions with suitable probe molecule as adsorbate offers additional capabilities in determination of localization of electronic midgap states in semiconductors beyond titanium dioxide.

Fig. 10. Luminescence of drAnatase with 5 nm, 15 nm and 30 nm nanocrystals at 25 °C under extrabandgap photoexcitation. a) The PL emission spectra. b) Integrated PL emission spectrum divided by mass of drAnatase.

surface. From the Tauc plots (Fig. 1a–c), the following λindirect wavelengths corresponding to bandgaps Egindirect were obtained for different nanocrystal size: λindirect (5 nm) = 375 nm; λindirect (15 nm) = 378 nm, and λindirect (30 nm) = 385 nm. The low-intensity PL emission spectra of hydrated anatase were not suitable for analysis due to strong water Raman artifacts. The higher drying temperature of 150 °C (Table S1)

Scheme 2. Changes in emission transitions in nanocrystalline anatase under the subbandgap photoexcitation. a) Dried anatase. b) Hydrated anatase.

395

Journal of Luminescence 192 (2017) 388–396

A. Samokhvalov

[12] M. Ghadiry, M. Gholami, L.C. Kong, C.W. Yi, H. Ahmad, Y. Alias, Nano-anatase TiO2 for high performance optical humidity sensing on chip, Sensors 16 (2016) 39. [13] P. Nagpal, V.I. Klimov, Role of mid-gap states in charge transport and photoconductivity in semiconductor nanocrystal films, Nat. Commun. 2 (2011) 486. [14] X. Chen, L. Liu, Z. Liu, M.A. Marcus, W.-C. Wang, N.A. Oyler, M.E. Grass, B. Mao, P.A. Glans, P.Y. Yu, J. Guo, S.S. Mao, Properties of disorder-engineered black titanium dioxide nanoparticles through hydrogenation, Sci. Rep. 3 (2013) Article number 1510. [15] A. Samokhvalov, Hydrogen by photocatalysis with nitrogen codoped titanium dioxide, Renew. Sustain. Energy Rev. 72 (2017) 981–1000. [16] J.R. Lakowicz, Principles of Fluorescence Spectroscopy, third ed, Springer, New York, 2006. [17] T. Vo-Dinh, Multicomponent analysis by synchronous luminescence spectrometry, Anal. Chem. 50 (1978) 396–401. [18] A. Andrade-Eiroa, G. de-Armas, J.-M. Estela, V. Cerda, Critical approach to synchronous spectrofluorimetry. I, Trends Anal. Chem. 29 (2010) 885–901. [19] A. Andrade-Eiroa, G. de-Armas, J.-M. Estela, V. Cerda, Critical approach to synchronous spectrofluorimetry. Ⅱ, Trends Anal. Chem. 29 (2010) 902–927. [20] M. Demir, M.L. McKee, A. Samokhvalov, Interactions of thiophenes with C300 Basolite MOF in solution by the temperature-programmed adsorption and desorption, spectroscopy and simulations, Adsorption 20 (2014) 829–842. [21] J. Dai, M.L. McKee, A. Samokhvalov, Fluorescence of A100 MOF and adsorption of water, indole, and naphthalene on A100 by the spectroscopic, kinetic, and DFT studies, J. Phys. Chem. C 119 (2015) 2491–2502. [22] J. Dai, M.L. McKee, A. Samokhvalov, Adsorption of naphthalene and indole on F300 MOF in liquid phase by the complementary spectroscopic, kinetic and DFT studies, J. Porous Mater. 21 (2014) 709–727. [23] D.L. Wood, J. Tauc, Weak absorption tails in amorphous semiconductors, Phys. Rev. B: Condens. Matter 5 (1972) 3144–3151. [24] H. Lin, C.P. Huang, W. Li, C. Ni, S.I. Shah, Y.H. Tseng, Size dependency of nanocrystalline TiO2 on its optical property and photocatalytic reactivity exemplified by 2-chlorophenol, Appl. Catal. B: Environ. 68 (2006) 1–11. [25] J.F. Zhang, P. Zhou, J.J. Liu, J.G. Yu, New understanding of the difference of photocatalytic activity among anatase, rutile and brookite TiO2, Phys. Chem. Chem. Phys. 16 (2014) 20382–20386. [26] N. Serpone, D. Lawless, R. Khairutdinov, Size effects on the photophysical properties of colloidal anatase TiO2 particles: size quantization versus direct transitions in this indirect semiconductor? J. Phys. Chem. 99 (1995) 16646–16654. [27] S. Monticone, R. Tufeu, A.V. Kanaev, E. Scolan, C. Sanchez, Quantum size effect in TiO2 nanoparticles: does it exist? Appl. Surf. Sci. 162–163 (2000) 565–570. [28] K. Madhusudan Reddy, S.V. Manorama, A. Ramachandra Reddy, Bandgap studies on anatase titanium dioxide nanoparticles, Mater. Chem. Phys. 78 (2003) 239–245. [29] S. John, C. Soukoulis, M.H. Cohen, E.N. Economou, Theory of electron band tails and the Urbach optical-absorption edge, Phys. Rev. Lett. 57 (1986) 1777–1780. [30] Z. Zhao, Z. Li, Z. Zou, Structure and properties of water on the anatase TiO2(101) surface: from single-molecule adsorption to interface formation, J. Phys. Chem. C 116 (2012) 11054–11061. [31] L. Kernazhitsky, V. Shymanovska, T. Gavrilko, V. Naumov, L. Fedorenko, V. Kshnyakin, J. Baran, Room temperature photoluminescence of anatase, J. Lumin. 146 (2014) 199–204. [32] Z. Hongchao, Z. Min, F. Qun, L. Bo, L. Wei, G. Heshuai, W. Minghong, L. Yong, Observation of defect state in highly ordered titanium dioxide nanotube arrays, Nanotechnology 25 (2014) Article No: 275603. [33] Y. Toyozawa, Elementary processes in luminescence, J. Lumin. 12 (1976) 13–21. [34] A.V. Emeline, V.N. Kuznetsov, V.K. Rybchuk, N. Serpone, Visible-light-active titania photocatalysts: the case of N-doped TiO2—properties and some fundamental issues, Int. J. Photoenergy 2008 (2008) Article ID 258394. [35] A. Stevanovic, M. Buttner, Z. Zhang, J.T. Yates, Photoluminescence of TiO2: effect of UV light and adsorbed molecules on surface band structure, J. Am. Chem. Soc. 134 (2012) 324–332. [36] H.F. Li, Y.Z. Guo, J. Robertson, Calculation of TiO2 surface and subsurface oxygen vacancy by the screened exchange functional, J. Phys. Chem. C 119 (2015) 18160–18166. [37] Y. Lei, L.D. Zhang, G.W. Meng, G.H. Li, X.Y. Zhang, C.H. Liang, W. Chen, S.X. Wang, Preparation and photoluminescence of highly ordered TiO2 nanowire arrays, Appl. Phys. Lett. 78 (2001) 1125–1127. [38] A. Di Paola, M. Bellardita, R. Ceccato, L. Palmisano, F. Parrino, Highly active photocatalytic TiO2 powders obtained by thermohydrolysis of TiCl4 in water, J. Phys. Chem. C 113 (2009) 15166–15174. [39] J.A. Rengifo-Herrera, K. Pierzchała, A. Sienkiewicz, L. Forró, J. Kiwi, C. Pulgarin, Abatement of organics and Escherichia coli by N, S co-doped TiO2 under UV and visible light. Implications of the formation of singlet oxygen (1O2) under visible light, Appl. Catal. B: Environ. 88 (2009) 398–406. [40] J. Yan, G. Wu, N. Guan, L. Li, Z. Li, X. Cao, Understanding the effect of surface/bulk defects on the photocatalytic activity of TiO2: anatase versus rutile, Phys. Chem. Chem. Phys. 15 (2013) 10978–10988.

4. Conclusions The photoluminescence emission spectra of nanocrystalline anatase at 25 °C in air have the two distinct peaks in visible range, a “blue light” at ca. 450 nm and “green light” at ca. 500 nm due to charge relaxation through the three distinct midgap states. We have determined the energy level diagram for the photoexcitation and relaxation of photoexcited charge in the nanocrystalline anatase through the three midgap states of different origin and localization, including an intrinsic midgap state self-trapped exciton (STE). Visible photoluminescence through midgap states is quenched upon adsorption of water vapor in air under ambient conditions, and reversibly increased upon water desorption. Synchronous luminescence spectra at 25 °C are superior to “conventional” PL emission spectra in visible range in resolving characteristic emission peaks of anatase. For the first time, the energy of the STE is determined by synchronous luminescence spectroscopy in the solid phase. Photoexcitation of self-trapped exciton in anatase at 420 nm results in strong emission of “green light” at 500 nm. Visible photoluminescence from anatase decreases when the nanocrystal size increases in the range 5–30 nm, indicating participation of surface midgap states in adsorption/desorption of water. In experiments by synchronous luminescence spectroscopy, emission of “green light” at 500 nm is shown to be preferentially quenched upon water vapor adsorption (surface midgap state) vs. “blue light” at 450 nm emitted upon the extrabandgap photoexcitation (bulk midgap state). Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.jlumin.2017.07.009. References [1] M.R. Ranade, A. Navrotsky, H.Z. Zhang, J.F. Banfield, S.H. Elder, A. Zaban, P.H. Borse, S.K. Kulkarni, G.S. Doran, H.J. Whitfield, Energetics of nanocrystalline TiO2, Proc. Natl. Acad. Sci. USA 99 (2002) 6476–6481. [2] P.J. Holliman, B. Vaca Velasco, I. Butler, M. Wijdekop, D.A. Worsley, Studies of dye sensitisation kinetics and sorption isotherms of Direct Red 23 on titania, Int. J. Photoenergy 2008 (2008) Article ID 827605. [3] A. Samokhvalov, S. Nair, E.C. Duin, B.J. Tatarchuk, Surface characterization of Ag/ titania adsorbents, Appl. Surf. Sci. 256 (2010) 3647–3652. [4] R. Ramakrishnan, S. Kalaivani, J.A.I. Joice, T. Sivakumar, Photocatalytic activity of multielement doped TiO2 in the degradation of Congo Red, Appl. Surf. Sci. 258 (2012) 2515–2521. [5] M. Zielinski, L.A. Burke, A. Samokhvalov, Selective activation of C=C bond in sustainable phenolic compounds from lignin via photooxidation: experiment and density functional theory calculations, Photochem. Photobiol. 91 (2015) 1332–1339. [6] C. Ampelli, R. Passalacqua, C. Genovese, S. Perathoner, G. Centi, T. Montini, V. Gombac, J.J. Delgado Jaen, P. Fornasiero, H2 production by selective photodehydrogenation of ethanol in gas and liquid phase on CuOx/TiO2 nanocomposites, RSC Adv. 3 (2013) 21776–21788. [7] K. Demeestere, J. Dewulf, H. Van Langenhove, B. Sercu, Gas–solid adsorption of selected volatile organic compounds on titanium dioxide Degussa P25, Chem. Eng. Sci. 58 (2003) 2255–2267. [8] B. Tan, Y. Wu, Dye-sensitized solar cells based on anatase TiO2 nanoparticle/nanowire composites, J. Phys. Chem. B 110 (2006) 15932–15938. [9] A. Tamilselvan, S. Balakumar, Anatase TiO2 nanotube by electrochemical anodization method: effect of tubes dimension on the supercapacitor application, Ionics 22 (2016) 99–105. [10] L. Visai, L. De Nardo, C. Punta, L. Melone, A. Cigada, M. Imbriani, C.R. Arciola, Titanium oxide antibacterial surfaces in biomedical devices, Int. J. Artif. Organs 34 (2011) 929–946. [11] N. Savage, B. Chwieroth, A. Ginwalla, B.R. Patton, S.A. Akbar, P.K. Dutta, Composite n-p semiconducting titanium oxides as gas sensors, Sens. Actuator BChem. 79 (2001) 17–27.

396