Graphene-TiO2 hybrids for photocatalytic aided removal of VOCs and nitrogen oxides from outdoor environment.

Outdoor and indoor air pollution has become a global concern in modern society. Although many policies and regulations on air quality have been promulgated worldwide over the past decades, airborne pollution still negatively affects health and therefore the life-style of human beings. One of the strategies to challenge this problem might be reducing the amount of airborne pollutant by mineralising them via photoinduced reactions. Photocatalytic oxidation of gaseous pollutants via titanium dioxide is one of the most promising solar photochemical reactions. In this research work, by means of a green sol-gel procedure, we have coupled titania to graphene (0.5 and 1.0 wt%) aiming to increase the solar photocatalytic activity of the produced hybrid materials. Transient paramagnetic species formed upon UV-A irradiation were detected by means of EPR spectroscopy. The photocatalytic reactions were assessed by monitoring the removal of nitrogen oxides and two different volatile organic compounds (benzene and isopropanol), which has never been assessed before. Our results highlight the exceptional characteristics of the TiO2/graphene hybrid material synthesised with 1.0 wt% graphene, and its excellent suitability for multi-purpose applications in the field of environmental remediation. Compared to unmodified titania, it shows a clear enhancement in the photocatalytic removal of those hazardous pollutants, having a photocatalytic degradation rate twice higher. In addition, the same material is highly stable and shows fully recyclability over repeated tests. Hybrid titania-graphene materials could thus be exploited to grant safer outdoor and indoor environments, having thus a beneficial impact on public health and on the quality of our lives.

Introduction

Air pollution is defined by the World Health Organisation as: “contamination of the indoor or outdoor environment by any chemical, physical or biological agent that modifies the natural characteristics of the atmosphere” [1]. Seinfeld and Pandis proposed an extended definition, stating that: ”air pollution is the presence in the air of one or more substances at a concentration or for duration above their natural levels, with the potential to produce an adverse effect” [2]. This implies that some of the substances that are regarded to be “air pollutant”, are naturally present in the air [3]. Those air pollutants have both natural and anthropogenic origin. Indeed, anthropogenic activity has had, and is still having, a dominant role in the increase of air pollutants to levels that yield adverse impact on human health and on the environment. Exposure to air pollution has been reported to be as the 5th ranking human health risk factor worldwide [4]. Several studies also recognised an association between air pollution and the climate changes that we are idly witnessing [5].

Common (indoor and outdoor) anthropogenic air pollutants are: particulate matter (PM), nitrogen oxides (NO), and volatile organic compounds (VOCs). PM air pollution is given by a mixture of microscopic solid particles and liquid droplets – PM with an aerodynamic diameter equal to, or smaller than, 10 and 2.5 µm is named PM10 and PM2.5, respectively [6]. NO emissions originate themselves from the combustion processes in stationary and mobile units [7], thus they are commonly related to traffic-sources – incidentally, around 80% of the NO emitted by automobiles comes from diesel-powered ones [8]. It is interesting to note that the lockdown for COVID-19 in Italy lead to a significant reduction of pollutants concentration, mainly due to vehicular traffic (PM10, PM2.5, benzene, CO, and NO) [9]. NO are related to several short- and long-term health effects [10], even to carcinogenicity [11]. VOCs are defined, according to the directive from the European Union, as: “any organic compounds having at 293.15 K a vapour pressure of 0.01 kPa or more, or having a corresponding volatility under the particular conditions of use”.[12] Anthropogenic VOCs emission was estimated to be around 142 Tg.C.yr−1 in the atmosphere [13]. Moreover, ground-level ozone (O3) is formed by photochemical reactions between sunlight and NO and VOCs [14], giving rise to the so-called photochemical smog, this again having an adverse impact on human health and climate. Rubber and paint production plants, varnishes, households products are amongst the main sources of isopropanol (2-propanol, IPA) [15]. The majority of IPA exposures are unintentional and are related to the central nervous system and respiratory depression [16]. Benzene can be found primarily outdoor, due to traffic sources [17]. Toxicity of this VOC involves both bone marrow depression and leukemogenesis [18].

Given this context, it is imperative to significantly reduce this global threat. Thus, renewable sources of energy are becoming more relevant. Among those, solar energy is taking an increasingly important role. Photocatalysis is one of the most investigated solar (photo)reactions, as it has different levels of applications, ranging from water and air purification, hygiene and sanitation, environmental remediation, automotive industry [19]. The term photocatalysis defines the acceleration of a photoreaction by a material, the (photo)catalyst, that it is not consumed during said reaction [20]. Semiconductors are the materials used as (heterogeneous) photocatalyst. Amid those photocatalytic semiconductors, titanium dioxide (titania, TiO2), an Earth-abundant material, is one of the most promising candidates, as it possesses: high oxidising power and resistance to photocorrosion in water environment, non-toxicity, cost-effectiveness [21]. However, one of TiO2′s drawback is the relatively high recombination rate of the photogenerated exciton, dissipating the energy as light or heat, thus strongly reducing the resulting photocatalytic activity [22]. Coupling/modifying TiO2 with other semiconductors, metals, noble-metals has been proposed to improve the spatial charge carrier separation, thus increasing the quantum yield of the photocatalytic process [23].

In 2004, Novoselov and Geim extracted from graphite a single layer of carbon atoms tightly packed into a two-dimensional honeycomb crystal structure [24]. This discovery led to the spark of the graphene-driven “gold rush”, as the discoverers themselves dubbed graphene to be the “rising star on the horizon of materials science and condensed matter physics” [25]. Indeed, as of 2004, there has been remarkable progress in graphene research, quicker than that of any other material [26]. Graphene and titania can be combined [27], and their hybridisation results in a built-in electric field that increases the conductivity of that system by promoting interfacial charge transfer and separation [28]. That is a decrease in the recombination rate of the photogenerated exciton (a critical step for a photocatalytic reaction), which in turn enhances the photocatalytic activity of the material. This led to massive research focussed on graphene-TiO2 composite materials for environmental applications [29], [30], [31], [32]. However, the vast majority of the literature deals with photocatalytic water treatment [33], [34], [35], [36], as only few papers, in recent literature based on graphene-TiO2 composites, are about the photocatalytic degradation of air pollutants. Those latter works deal or with the abatement of nitrogen oxide(s) [37], or with the removal of organic compounds – like acetone [38], or benzene [39]. Still, a single comprehensive study dealing with the gas–solid phase photocatalytic removal of different kinds of hazardous organic compounds (i.e. alcohol and aromatic compounds), and inorganic nitrogen compounds (i.e. NO and NO2) over graphene/TiO2 hybrid materials has never been assessed before. To fill that literature gap, in this paper we have used a green sol–gel synthesis method to produce TiO2 and hybrid TiO2/graphene nanostructures [40]. Their photocatalytic activity was assessed in the gas phase, by monitoring the degradation of NO and two different types of VOCs, namely: benzene and IPA. The light source used in the photocatalytic experiments imitated the solar spectrum, so as to reproduce an outdoor environment. However, an artificial visible light emitting diode (LED) lamp – i.e. a lamp with no UV components, thus imitating and indoor environment, was also used in the photocatalytic tests. Results showed that the addition of graphene boosted titania’s photocatalytic activity. With 1.0 wt% graphene, and using a light imitating the solar spectrum, the photocatalytic activity of TiO2 was (more than) doubled in NO abatement, benzene removal, and IPA oxidation, never being reported before. Materials produced in this research work possess high throughput against those hazardous pollutants, therefore, being excellent candidates for offering a safer outdoor environment.

Experimental section

Sample preparation

We followed an aqueous sol–gel procedure for the synthesis of TiO2-based photocatalytic materials; details of the synthesis can be found in a paper previously described by the authors [40]. Briefly, aqueous sols were made by means of the controlled hydrolysis and peptisation of titanium(IV)isopropoxide [Ti-i-pr, Ti(OCH(CH3)2)4] with Milli-Q water (18.2 MΩ.cm, H2O:Ti-i-pr = 4:1) diluted in isopropyl alcohol (2-propanol, IPA). To make the graphene-modified TiO2 sols, 0.5 and 1.0 wt% of graphene nanoplatelets dispersed in ethanol (supplied by Graphenest) were added to the TiO2-based sol when this had 1 M concentration. The sols were dried in an oven at 80 °C, and then the dried gels were thermally treated under static air atmosphere using an electric muffle furnace, following thermal cycles reaching two maximum temperatures: (i) 250 °C with a heating rate of 15 °C.min−1, pre-heating the furnace at 200 °C, with a 8 h dwell time; (ii) 450 °C, with a heating rate of 5 °C.min−1, with 2 h dwell time. Specimens were referred to as: G were × stands for the graphene wt% (0.5 or 1.0 wt%), and Y is a number indicating the maximum temperature reached (i.e. 250 or 450 °C). For instance, the TiO2 sol in which was added 0.5 wt% graphene, and thermally treated at 450 °C / 2 h, will be referred to as: G0.5/450. Unmodified titania has been labelled: Ti250 or Ti450, according to the maximum temperature reached.

Sample characterisation

X-ray powder diffraction (XRPD) was used to reveal the mineralogical composition of the prepared specimens. The relative fractions of crystalline phases were determined via Rietveld refinements of the XRPD data. The XRPD patterns were collected at room temperature on a θ/θ diffractometer (PANalytical X’Pert Pro, NL), equipped with a fast RTMS detector (PIXcel 1D, PANalytical), with Cu Kα radiation (45 kV and 40 mA, 20–80 °2θ range, with a virtual step scan of 0.02 °2θ, and virtual time per step of 200 s). The Rietveld data analysis for obtaining semi-quantitative phase analysis (QPA) information was assessed using the GSAS-EXPGUI software packages [41], [42]. The instrumental broadening was obtained from the refinement of LaB6 standard (NIST SRM 660b) and included in all of the Rietveld refinements. The refinement strategy that we followed has been reported in a previous work [40]. XRPD was also used to gain microstructural information. At this purpose, the same instrument and setup as per the QPA was employed, but XRPD patterns were recorded in the 20–145 °2θ range, with a virtual step-scan of 0.1 °2θ, and virtual time per step of 500 s, so to work with data having a higher signal-to-noise ratio. The whole powder pattern modelling (WPPM) [43], as implemented in the PM2K software package [44], was used for the microstructural analysis of the diffraction data. In this work we adopted the same modelling approach as that described in previous works of us, assuming the crystalline domains to be spherical, and their diameter distributed according to a log-normal curve [40].

High-resolution scanning transmission electron microscopy (HR-TEM) imaging and electron energy-loss spectroscopy analysis (SR-EELS) were performed using a FEI Titan Cubed Themis microscope operated at 80 kV. This microscope is equipped with a double Cs aberration-corrector, a monochromator, a X-FEG gun, an ultrahigh resolution energy filter (Gatan Quantum ERS), which allows working in dual-EELS mode, and a super X EDS detector. The monochromator was excited during imaging and spectroscopic experiments to minimize chromatic aberrations. Convergence and collection angles during EELS experiments were 5 and 21 mrad, respectively. The probe diameter for EELS acquisition in TEM mode was around 30 nm. The energy resolution, measured as the full width at half maximum of the zero loss peak, was 0.75 eV with a dispersion of 0.25 eV.pixel−1. Background subtraction for the core-loss spectra was performed by using the usual inverse power law function. Interpretations of the fast Fourier transform (FFT) patterns in HR-TEM images were performed by using the JEMS software [45].

Optical spectra of the prepared graphene-TiO2 hybrid materials were recorded on a Shimadzu UV-3100 spectrometer (JP), equipped with an integrating sphere, and a white reference material made of Spectralon; UV − vis spectral range was explored (250–850 nm), using 0.2 nm in resolution. Raman spectra were acquired on a RFS 100/S (Bruker, DE) equipped with a 1,064 nm Nd:YAG laser as the excitation source, in the 50–1,800 cm−1 wavenumber range, with 2 cm−1 in resolution. Specific surface area (SSA) of the specimens was measured by means of the Brunauer–Emmett–Teller (BET) method (Micromeritics Gemini 2380, US) using N2 as the adsorbate gas, on samples degassed at 120 °C.

EPR spin trapping experiments were performed using the spin trapping agent 5,5-dimethyl-1-pyrroline N-oxide (DMPO; Sigma-Aldrich, distilled prior to the application). Benzene (anhydrous, 99.8%) and 2-propanol (IPA, puriss. p.a., ACS reagent, ≥99.8%) were purchased from Sigma-Aldrich. The DMPO solutions in benzene and IPA (0.2 M) were used to impregnate the powdered photocatalysts placed in a thin-walled quartz Q-band EPR tube (Bruker). As-prepared photocatalyst was mixed by a gentle stream of argon in order to reduce the concentration of dissolved oxygen, and to eliminate the EPR line broadening in benzene and IPA caused by oxygen dissolved in these solvents [46]. The sample was directly irradiated in the cavity of EPR spectrometer EMX (Bruker) operating at 100 kHz field modulation in the standard TE102 rectangular cavity (ER 4102 ST, Bruker) at room temperature. An UV radiator [λmax = 365 nm; Bluepoint LED, Hönle UV Technology; irradiance of 15 mW.cm−2 within the EPR cavity determined by a UVX radiometer (UVP, USA)] was used as irradiation source. The standard EPR spectrometer settings were as follows: microwave frequency: ~9.447 GHz; microwave power: 11.5 mW; center field: 363 mT; gain: 1 × 105; modulation amplitude: 0.05 mT; sweep time: 25 s; time constant: 10.24 ms; number of scans: 10. The experimental EPR spectra were analysed by the WinEPR software (Bruker), and the simulated spectra were calculated with the EasySpin toolbox working under Matlab (MathWorks) software [47].

Functional properties: Photocatalytic experiments

De-NO and benzene removal gas–solid phase photocatalytic experiments were carried out in the same reactor, that operated in continuous mode [48]. It comprised a cylindrical chamber (3.8 L internal volume) built from a high grade stainless steel with a stainless steel top with a sealed glass window to allow the light to reach the sample that was placed inside – a scheme of the reactor employed is displayed in Figure S1a of the Supplementary information file. A light imitating the solar spectrum (Osram Ultra-Vitalux, 300 W) was used as light source. This was positioned 33 cm above the photocatalyst; the radiant flux per unit area (irradiance) reaching the surface of the photocatalyst was then equal to 3.5 mW.cm−2 in the UV-A range, and 35 mW.cm−2 in the visible spectral range – values measured with a radiometer (Delta OHM, HD2302.0, IT). A LED lamp (Philips warm white, 2700 K LED bulb, 9.5 W) was employed to mimic an indoor environment. The LED was placed 12 cm above the photocatalyst. The irradiance reaching the photocatalyst was estimated to be approximately 5.8 mW.cm−2 in the visible-light range and nil in the UV-A range. Emission spectra of the solar and visible-light LED lamps are displayed in Figure S2.

A Petri dish, 6 cm in diameter, was covered with a layer of photocatalyst (0.10 g). The tests were performed at 24 ± 1 °C (temperature inside the reactor) with a relative humidity of 36%. These parameters were controlled by a thermocouple that was placed inside the chamber, and a humidity sensor placed in the inlet pipe, and they were stable throughout the tests.

As per the de-NO photocatalytic experiments, the initial concentration of NO (prepared using synthetic air and NO gas) was set at 200 ppb. That concentration is higher than that commonly reported outdoor, e.g. outdoor urban levels have an annual mean range of 11–48 ppb (around 20–90 µg.m−3) [49]. However, the safety limit of 200 µg.m−3 (approximately 110 ppb) on a daily basis is often breached in many European Countries [50].

The outlet concentration of NO gases was measured using a chemiluminescence analyser (AC-31 M, Environment SA, FR), according to a procedure that we described in very detail previously [51], [52].

When benzene was used as model gas, its concentration exiting through the outlet was measured using a VOC-72 M gas analyser (Environment SA, FR), that is based on gas chromatography and photoionisation detector. The inlet gas mixture (prepared mixing gas cylinders containing synthetic air and benzene) was allowed to flow into the chamber until it stabilised at a concentration of ~ 260 μg.m−3 (~80 ppb). This concentration was chosen because the World Health Organisation recommends an exposure to total VOCs < 300 μg.m−3 during 8 h/day [53]. The mixture of air with that concentration of benzene was guaranteed using two mass flow controllers with a flow rate of 150 mL.min−1. The photocatalytic experiments were assessed by placing the photocatalyst inside the reactor and covering the glass window. Once the desired NO/benzene concentration was reached and it attained a stable level into the reactor, the window glass was uncovered, the lamp turned on, and the photocatalytic reaction started. Photocatalytic activity for de-NO abatement and benzene removal was reported as formal quantum efficiency (FQE) [54], defined as the number of molecules degraded per incident photon. In this work, based on the results obtained in Section 3.2 in which the optical band-gap was determined, we presumed that prepared photocatalysts exploited the UV–A part of the electromagnetic spectrum (i.e. 315–400 nm) for band-to-band excitation, thus excluding wavelengths > 400 nm. Furthermore, aiming to have a better comparison between the tested specimens, data relative to the first 2 min (de-NO) or 60 min (benzene removal) of reaction time, were interpreted according to a first-order kinetic law.

A different reactor was employed for the gas-phase oxidation of IPA and, unlike the reactor described previously, this was a batch system (Figure S1b). It was a cylindrical reactor (1.4 L in volume) covered by a quartz glass connected by Teflon tubes to a FT-IR spectrometer; the whole batch system was hermetically sealed. IPA photocatalytic oxidation is as follows: IPA acetone further products → CO2 + H2O. Following the works of Munuera and co-authors [55], [56], the IPA to acetone reaction is described as a zero order reaction, whilst the second reaction (acetone to further products) is a first order reaction. Also, Larson et al. [57], demonstrated that k 1 ≫ k 2, therefore, we can genuinely assume the initial reaction rate of acetone formation to be reasonably approximated as linear, and obtain a reliable kinetic for the photocatalytic activity. The working distance between the Petri dish and the lamp was 6 cm. The light source employed for the photocatalytic tests was a 300 W Xenon lamp (Newport Oriel Instruments, US). The lamp imitates the solar light spectrum, both in the UV and visible range, with an irradiance of approximately 0.8 mW.cm−2 in the UV range (250–400 nm), and 8.8 mW.cm−2 in the wavelength range of 400–800 nm. A 400 nm cut-off filter was used to exclude the UV radiation, and thus measurements were also taken only under visible-light irradiation. The filter decreases the light intensity, which reaches a value of 6.5 mW.cm−2 in the 400–800 nm range. The relative humidity in the reacting system was kept constant in the range 25%–30% by means of a flow of air passing through molecular sieves until a pre-defined humidity was attained. Each experiment was performed by injecting 8 µL of IPA (~800 ppm in gas phase) into the reacting system through a septum; the total reaction time was set at 24 h, and the lamp was turned on after a certain period of time after IPA injection, to allow for an adsorption/desorption equilibrium. IPA degradation, as well as the acetone formation-degradation process, was followed by monitoring the calculated area of their characteristic peaks located at 1,104 and 1,207 cm−1, respectively, by using a FT-IR spectrometer (Perkin Elmer Spectrum BX, US). The acetone formation kinetics was taken as a measurement of photocatalytic activity, because the photocatalytic oxidation of IPA is the only source of acetone. Thus, although an indirect evidence, it unequivocally proves IPA oxidation. On the other hand, the decreasing concentration of IPA may be due to photocatalytic activity and some other undesired processes, like adsorption and possible minor leaking. [It has to be stressed here that the radiant fluxes (in the UV-A and visible regions) entering the reactors are significantly lower than those of xenon arc lamps that are normally used by the catalytic community – i.e. in the order of around 8 mW.cm−2 in the UV-A region, and 80 mW.cm−2 in the visible range [58], [59], [60]. All of the photocatalytic tests were repeated in triplicate to check recyclability and photostability of the photocatalysts.

Results and discussion

XRPD and HR-TEM analyses

XRPD patterns of the synthesised materials are shown in Figure S3a,b whilst Rietveld QPA results are listed in Table 1 . A graphical output of a Rietveld refinement is in Figure S4. TiO2 thermally treated at 250 °C / 8 h (Ti250) is composed of anatase (79.6 wt%), rutile (8.1 wt%), and brookite (12.3 wt%). Addition of graphene to the system strongly delayed the anatase-to-rutile phase transition (ART), being the anatase wt% amount in G0.5/250 (G1.0/250): 94.2 wt% (92.8 wt%); that of rutile was: 0.6 wt% (0.4 wt%); brookite: 5.2 wt% (6.7 wt%). Similarly, graphene addition retarded the ART in the specimens thermally treated at 450 °C / 2 h. Ti450 is composed of 70.8 wt% anatase, 19.5 wt% rutile, and 9.7 wt% brookite. Specimens G0.5/450 (G1.0/450) are composed of: 89.1 wt% (88.4 wt%) anatase; rutile: 8.1 wt% (9.4 wt%); brookite: 2.8 wt% (2.1 wt%). As per the microstructure, microstructural information as extracted from the WPPM formalism is listed in Table 1 and S1, and Figure S5 (an output of a WPPM modelling is shown in Figure S6). Consistent with QPA data, graphene addition to the system delayed the nucleation-and-growth of anatase crystalline domains – ART being a nucleation-and-growth phenomenon [61]. Indeed, the diameter of anatase crystalline domains in Ti250 is 4.4 nm. That in G0.5/250 (G1.0/250) is equal to: 3.8 nm (3.9 nm). A similar behaviour happens when the specimens are thermally treated at 450 °C / 2 h: the average diameter of anatase crystalline domain in Ti450 is 10.4 nm. That in G0.5/450 (G1.0/450) is: 7.0 nm (6.2 nm). Rutile nanocrystals in Ti450 have a diameter equal to 14.4 nm; those in G0.5/450 (G1.0/450): 8.7 nm (8.3 nm) – cf Table 1.

Table 1

Phase fraction (wt%) of TiO2 polymorphs, average diameter of the crystalline domains, and BET specific surface area in the prepared specimens.

Sample	Phase composition (wt%)			Mean crystalline domain diameter (nm)
	anatase	rutile	brookite	anatase	rutile	brookite	SSABET (m2.g−1)
Ti250/8h	79.6 ± 0.2	8.1 ± 0.3	12.3 ± 0.5	4.4 ± 0.1	8.7 ± 1.4	–	134.4 ± 2.1
G0.5/250	94.2 ± 0.1	0.6 ± 0.1	5.2 ± 0.3	3.8 ± 0.1	–	5.6 ± 0.6	139.1 ± 1.3
G1.0/250	95.3 ± 0.1	0.4 ± 0.1	4.3 ± 0.3	3.9 ± 0.1	–	4.8 ± 0.5	140.6 ± 1.3
Ti450	70.8 ± 0.1	19.5 ± 0.1	9.7 ± 0.2	10.4 ± 0.7	14.4 ± 0.6	7.0 ± 0.1	45.9 ± 0.7
G0.5/450	89.1 ± 0.1	8.1 ± 0.2	2.8 ± 0.3	7.0 ± 0.1	8.7 ± 0.1	–	50.1 ± 0.6
G1.0/450	88.4 ± 0.1	9.4 ± 0.2	2.1 ± 0.2	6.2 ± 0.2	8.3 ± 0.6	–	57.4 ± 0.6

Agreement factors of Rietveld refinements were: R(F2) ≤ 4.73%; Rwp ≤ 4.39%; χ2 ≤ 2.21. There were 2285 observations for every Rietveld refinement; the number of anatase, rutile and brookite reflections was 32, 31 and 158, respectively.

TEM micrographs of pure graphene are displayed in Figure S7. Graphene flakes randomly superposed are visible in Figure S7a. Each flake is around 0.9–1.0 μm in length, and 200–300 nm in width. The thickness of each flake is around 5 nm, cf Figure S7b.

Fig. 1 a shows a TEM image acquired on G1.0/450. The presence of clustered TiO2 nanoparticles and graphene flakes in the same area can clearly be highlighted. The flakes have a typical size of about 500 × 1000 nm. Fig. 1b depicts an image of higher magnification taken in the same area. The high crystalline quality of both the nanoparticles and the flake can be clearly highlighted. In addition, it should be noted that most of the flakes are constituted by a multi-layer graphene – the number of layers varying between 3 and ~ 20. Automatic indexation of the FFT pattern acquired on the flake show that this corresponds to a graphitic basal plane described in the [0 0 1] zone axis. TiO2 nanoparticles are clustered, and have a typical size of about 5–7 nm, as displayed in Fig. 1c. Clustering of TiO2 nanoparticles and their distribution over graphene’s surface can be appreciated in Fig. 1d. In it, it is also shown that the thermal treatment did not virtually modify the flake structure.

Fig. 1

a) TEM image acquired on the sample G1.0/450; the orange arrow points TiO2 nanoparticles, the dark grey arrow graphene flakes. b) HR-TEM image acquired on the sample G1.0/450. The inset shows the FFT calculated on the area highlighted by the red square. c) HR-TEM image of G1.0/450 to show clustered TiO2 nanoparticles. d) TEM micrograph showing the distribution of clustered TiO2 nanoparticles on graphene surface. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

To get more insight on the chemical nature of the nanostructures, EELS spectra were acquired on the flakes and nanoparticles of G1.0/450 (Fig. 2 ). The EELS spectrum acquired on the nanoparticles shows the presence of Ti–L 2,3 and O–K edges located around 457 and 530 eV, respectively. On the other hand, the spectrum acquired on the flake shows only the presence of the C–K edge around 285 eV. It should be noted that no other edges, related in particular to oxygen or nitrogen, can be highlighted on the probed energy range. In addition, the C–K edge clearly shows fine-structure characteristics of a highly graphitic carbon [62]. These results show that the hybrid nanostructure is constituted of few-layers of graphene mixed with small TiO2 nanoparticles [62]. Similar results concerning imaging and spectroscopic analyses were obtained on G1.0/250.

Fig. 2

EELS spectra acquired on nanoparticles (orange line) and one flake (dark blue line) in G1.0/450. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

UV–Vis and Raman spectroscopy

Diffuse reflectance (DR) spectra are shown in Fig. 3 a,b. They all display one single absorption edge, located at around 380 nm, and assigned to the band-to-band transition in titania [63]. Addition of graphene increased the absorption in the visible range, consistent with previous literature reports [64]. Values of the apparent optical bandgap energy (E g) of the specimens are listed in Table 2 . These were extracted from the DR spectra by means of the Tauc-plot (considering direct and indirect interband transitions) [65], and of the differential reflectance method (Figure S8) [66]. Specimens are mostly composed of anatase (cf Table 1), and this TiO2 polymorph is commonly reported to have an indirect band-to-band transition [67]. However, the E g values determined via the Tauc procedure, and assuming an indirect model, are in contrast with the expected E g value of anatase (i.e. ~ 3.2 eV; ~388 nm) [68]. Therefore, a direct model for interband transition would seem more appropriate in this system [see also references [39], [52]] – indeed, although unusual, anatase having direct gap transition has already been reported in the literature [69], [70].

Fig. 3

Optical spectra (pseudo-absorption versus energy), as derived by Kubelka − Munk analysis of the DR spectra. a) Specimens thermally treated at 250 °C / 8 h; b) specimens thermally treated at 450 °C / 2 h.

Table 2

Apparent optical Eg, estimated according to the Tauc procedure and the differential reflectance method (dR/dλ).

Sample	Apparent optical Eg
	Tauc plot†				dR/dλ
	direct		indirect		anatase		rutile		brookite
	eV	nm	eV	nm	eV	nm	eV	nm	eV	nm
Ti250/8h	3.25	382	3.00	414	3.28	378	3.16	392	3.41	363
G0.5/250	3.22	385	2.96	418	3.23	384	–	–	–	–
G1.0/250	3.24	382	2.95	421	3.22	384	–	–	–	–
Ti450	3.12	398	2.93	423	3.30	376	3.07	404	3.41	364
G0.5/450	3.15	394	2.92	424	3.32	373	3.08	403	–	–
G1.0/450	3.15	393	2.90	428	3.34	372	3.13	397	–	–

The absolute error in the Tauc-plot measurements was < 1%.

However, the differential reflectance method has been reported to be a more reliable technique for extracting the E g values in the TiO2 polymorphs system [40], [66], [71]. Indeed, values listed in Table 1 (6th to 11th columns), are consistent with anatase, rutile and brookite E gs reported in the literature [68], [72].

Raman spectra are displayed in Fig. 4 a,b. In this figure, the main Raman active mode of anatase are clearly detectable [i.e.: E g(1) (144 cm−1); B 1g(1) (399 cm−1); A 1g,B 1g(2) (515 cm−1); E g(2) (638 cm−1)] [73]. The major Raman bands assigned to graphene (i.e.: D, G, and a very weak D’) [74], are identified in the specimens modified with graphene (see the inset in Fig. 4a,b).

Fig. 4

Raman spectra: a) specimens thermally treated at 250 °C / 8 h; b) specimens thermally treated at 450 °C / 2 h. Insets in a) and b) show the magnified wavenumber region 1200–1850 cm−1, in specimens modified with 1.0 wt% graphene, to highlight the major graphene bands, D (~1,300 cm−1), D′ (around 1,610 cm−1), and G (approximately at 1,580 cm−1).

Functional properties: Gas-phase photocatalytic experiments

de-NOx properties

Solar de-NO photocatalytic activity results are shown in Fig. 5 a-d. As it is seen in Fig. 5a,b all of the specimens are able to eliminate the NO gases when stimulated by a light simulating the solar radiation (that is with a UV-A component). Specimens thermally treated with the two thermal cycles showed quite similar de-NO properties during the total duration of the experiments. This was confirmed by FQEs values too, as reported in Fig. 5c. Unmodified specimens thermally treated at 250 °C / 8 h and 450 °C / 2 h, after 10 min irradiation time, had de-NO removal rates equal to 6.7 × 10–3 and 9.7 × 10–1 molecules.photon–1, respectively. Upon modification with 0.5 and 1.0 wt% graphene, in specimens thermally treated at 250 °C / 2 h, the FQE rates increased to 11.7 × 10–3 and 13.3 × 10–3 molecules.photon–1, respectively. Better NO abatements FQE rates were reached when specimens were subjected to a higher thermal treatment, G0.5/450 and G1.0/450, these respectively being: 12.8 × 10–3 and 14.0 × 10–3 molecules.photon–1. To have a better comparison amongst the specimens, these results were compared with the initial (first 2 min of reaction) pseudo-first order kinetic constants, in which are reported the values averaged over three consecutive photocatalytic de-NO tests, see Fig. 5d. Indeed, specimens thermally treated at 450 °C/ 2 h exhibited a slightly better photocatalytic activity than those treated at 250 °C / 8 h. Specimens thermally treated at 450 °C / 2 h have in general a larger size of anatase and rutile crystalline domains (cf Table S2). This might (partly) justify the slight increase in the de-NO properties of those specimens: the larger the size, the lower the defect density in the material (that is: lower density of trapping centres), therefore, lower probability of the exciton recombination [75]. In any case, specimens hybridised with graphene showed a higher photocatalytic activity. Addition of 1.0 wt% graphene enhanced the photocatalytic activity: G1.0/250 and G1.0/450 almost doubled their photocatalytic activity compared to Ti250 and Ti450. Values of the initial k’app for Ti250 and Ti450 were 6.9 × 10–2 and 7.1 × 10–2 min−1, respectively. Those of G1.0/250 and G1.0/450: 11.4 × 10–2 and 12.7 × 10–2 min−1, respectively. For practical purposes, a photocatalyst should be active over several runs. We have tested the photocatalytic activity of G1.0/450 over five consecutive tests, and this proved to be fully re-usable, see Fig. 6 .

Fig. 5

Photocatalytic de-NO plots of the samples irradiated with solar lamp. a) Specimens thermally treated at 250 °C / 8 h; b) specimens thermally treated at 450 °C /2h. In a) and b) the initial negative time zone indicates a period of stabilisation of the chamber; positive time indicates that the light has been turned on starting the experiment. c) Histograms depicting FQEs of prepared photocatalysts for NO removal, for a total reaction time of 10 min. d) Histograms showing the first 2 min of reaction pseudo-first order kinetic constants, k’app. Reported values are averaged over three consecutive photocatalytic de-NO tests. The error bars represent the standard error over three photocatalytic experiments.

Fig. 6

Solar photocatalytic de-NO tests (G1.0/450) repeated for five consecutive times. The grey shaded areas represent the time in which the solar lamp was turned off; the white ones the time in which the solar lamp was turned on.

Photocatalytic de-NO results using the visible white LED lamp are shown in Figure S9. Specimens showed themselves to be photocatalytically active even using a lamp with no UV component – although less active than with the solar lamp. Activity under visible light illumination might be explained by the mineralogical composition of the specimens: they are indeed composed of anatase and rutile TiO2 polymorphs (see Table S1). Anatase and rutile nanoparticles when in form of a heterojunction are capable of absorbing visible light, and thus enable photocatalytic activity under visible light illumination [40].

VOCs removal

Benzene removal

Solar photocatalytic benzene removal results are reported in Fig. 7 a,b. These are shown as histograms showing FQEs over a total of 5 h irradiation time, and the first 60 min of reaction pseudo-first order kinetic constants. These data are pretty much comparable with those obtained with the abatement of nitrogen oxides. That is: all of the samples are photocatalytically active; besides, hybridisation of titania with graphene boosted TiO2 photocatalytic activity. This was approximately two-fold higher for specimens thermally treated at 250 °C / 8 h, the overall FQEs values being: 5.4 × 10–8, 10.3 × 10–8, and 10.0 × 10–8 molecules.photon–1 in Ti250, G0.5/250, and G1.0/250, respectively, cf Fig. 7a. Increased thermal treatment, led to higher overall FQEs rates: 7.9 × 10–8, 10.1 × 10–8, and 11.0 × 10–8 molecules.photon–1 in Ti450, G0.5/450, and G1.0/450, over 5 h irradiation time. However, taking into consideration the initial 60 min of benzene removal, Fig. 7b, specimens with 1.0 wt% graphene had a photocatalytic activity that was 3 times higher than that of unmodified titania in the set of specimens thermally treated at 250 °C / 8 h – k’app values of Ti250 (G1.0/250) were: 16 × 10–2h−1 (49 × 10–2h−1). In case of specimens thermally treated at 450 °C / 2 h, the k’app values of Ti450 and G1.0/450 were 20 × 10–2h−1 and 53 × 10–2h−1, respectively – more than two times higher. Like the de-NO tests, G1.0/450 proved itself to be extremely repeatable also in consecutive photocatalytic benzene removal experiments, as depicted in Fig. 8 .

Fig. 7

a) Histograms depicting the FQE rates of prepared photocatalysts for benzene removal using the solar lamp, for a total reaction time of 5 h. b) Histograms showing the pseudo-first order kinetic constants (first 1 h reaction), k’app (solar lamp). Reported values are averaged over three consecutive photocatalytic benzene removal tests. The error bars represent the standard error over three photocatalytic experiments.

Fig. 8

Photocatalytic benzene removal tests (G1.0/450) repeated for five consecutive times. The grey shaded areas represent the time in which the solar lamp was tuned off; the white ones the time in which the solar lamp was turned on.

Visible-light photocatalytic benzene removal results are displayed in Figure S10. Unlike de-NO abatement, graphene addition did not increase the visible-light photocatalytic activity of titania in the removal of benzene.

Isopropanol oxidation

UV–Vis photocatalytic activity results of the samples against the removal of IPA, shown as FQE and formation rates of acetone (ppm.h−1), are summarised in Fig. 9 a,b. Consistent with the data obtained with the NO abatement and benzene removal (cf Section 4.1, and Section 4.2.1), addition of graphene to TiO2 aided the photocatalytic assisted IPA oxidation. This is shown by FQEs values (Fig. 9a): pure TiO2 at 250 °C / 8 h (Ti250) showed an FQE acetone formation rate of 1.8 × 10–2 molecules.photon–1. Upon hybridisation with 0.5 and 1.0 wt% graphene, the FQE acetone formation rate increased to 2.9 and 2.6 × 10–2 molecules.photon–1 (G0.5/250 and G1.0/250, Fig. 9a). This is also displayed by acetone formation rates of 210, 330, and 295 ppm.h−1 for Ti250, G0.5/250 and G1.0/250, respectively, Fig. 9b).

Fig. 9

Photocatalytic isopropanol degradation under UV–Vis light exposure, expressed as: a) FQEs of initial acetone formation; b) rate constant of the initial acetone formation. In a) and b), the total irradiation time was set at 1 h. The error bars represent the standard error over three photocatalytic experiments.

At the higher thermal treatment, addition of 0.5 and 1.0 wt% graphene resulted in acetone formation rate (in terms of FQE) more than the double than that of TiO2: 2.2 × 10–2, 4.4 × 10–2, and 4.8 × 10–2 molecules.photon–1 acetone FQE rate for Ti450, G0.5/450, and G1.0/450, respectively, cf Fig. 9a. This correlated well with acetone formation rates values of: 250, 510, and 550 ppm.h−1, respectively (Fig. 9b).

Contrasting results occurred when visible-light radiation was used, cf Figure S11. At 250 °C, acetone formation rates decreased with the increase of graphene in the system (i.e., the photocatalytic activity was: G1.0/250 < G0.5/250 < Ti250). On the contrary, when the specimens were thermally treated at 450 °C / 2 h, IPA oxidation had this trend: G1.0/450 > G0.5/450 > Ti450, even though the rise in photocatalytic activity was little (18, 15, and 14 ppm.h−1, respectively).

Therefore, results reported here emphasise once again the extreme complexity of photocatalytic air purification processes (using different molecules as pollutant to be removed and different light sources).

EPR spin trapping technique

The EPR spin trapping technique was applied for the detection of transient paramagnetic species formed upon UV-A exposure of powdered photocatalysts impregnated with benzene or IPA containing DMPO spin trap. The analogous DMPO spin-adducts were detected for the photocatalytic systems prepared with Ti450 and G0.5/450 in the presence of IPA or benzene. Fig. 10 represents the experimental EPR spectra obtained upon continuous UV-A exposure of the powdered photocatalyst G1.0/450 impregnated with IPA/DMPO or benzene/DMPO, along with their simulations. The UV-A photoactivation of G1.0/450/IPA/DMPO resulted in the generation of superimposed signals corresponding to the simultaneous production of various DMPO-adducts. The asymmetric six-line signal represents the spin-adducts of two carbon-centred radicals with slightly different spin-Hamiltonian parameters, i.e. •DMPO-CR1 (a N = 1.453 mT, a H β = 2.178 mT, g = 2.0057) and •DMPO-CR2 (a N = 1.527 mT, a H β = 2.185 mT, g = 2.0057) that dominates the EPR spectrum (Fig. 10a). Additionally, the spin-adduct of oxygen-centred radical (a N = 1.287 mT, a H β = 0.999 mT, a H γ = 0.098 mT, g = 2.0057) was detected as minor spectral component, and was attributed to the trapped O2 •–/HO2 • radical generated via interaction of photoelectrons with O2 [76]. The formation of oxygen- and carbon-centred radicals detected as the corresponding DMPO-adducts is well compatible with the oxidation of IPA adsorbed on TiO2 surface. The photocatalytic oxidative transformation of IPA is a complex process and acetone as intermediate may be formed by several reaction pathways [77], [78]. Under those experimental conditions, IPA molecules adsorbed on the photocatalysts’ surface reacts with the photogenerated holes forming the primary alkoxy radicals (CH3)2CHO•. These species rapidly react with other IPA molecules and/or with hydroxyl radicals producing hydroxyalkyl radical [(CH3)2 •C(OH)], transformed via consecutive reactions to acetone [77], [78].

Fig. 10

Experimental (orange solid line) and simulated (blue dashed line) EPR spectra (magnetic field sweep width 8.0 mT) of: (a) IPA, and (b) benzene G1.0/450 impregnated powders in the presence of DMPO spin trapping agent upon LED@365 nm exposure (dose 4.5 J.cm−2). The concentration of oxygen was reduced by the slight stream of argon before exposure. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Upon continuous LED@365 nm exposure of G1.0/450/benzene/DMPO, the signals of oxygen-centred spin-adducts dominate the EPR spectrum (Fig. 10b). The crucial twelve-line signal was attributed to the superoxide radical anion added on the DMPO molecule (•DMPO-O2 –/O2H, spin-Hamiltonian parameters a N = 1.279 mT, a H β = 1.012 mT, a H γ = 0.139 mT, g = 2.0061). The simulation analysis of the complex experimental EPR spectrum revealed the presence of DMPO-adducts with carbon-centred and alkoxy radicals, originating from benzene molecule [79]. The photocatalytic oxidation of benzene on titania surface is a very complex process including photogenerated holes, electrons and reactive oxygen species [80], [81]. The photogenerated holes may interact with benzene molecule forming benzene radical cation. However, in the presence of HO–/H2O adsorbed on photocatalysts’ surface, the holes effectively react with HO–/H2O producing hydroxyl radicals. The addition of HO• to benzene molecule leads to the formation of hydroxy cyclohexadienyl radical, which can enter in consecutive reactions with oxygen, benzene or other species. The photogenerated electrons react with molecular oxygen forming superoxide radical anions. The benzene radical cation may interact also with both O2 •– and O2 and via transient benzene peroxy radicals, and phenol is formed. The DMPO spin-adducts of peroxy radicals are not stable and may rapidly disproportionate to the detected alkoxy-radical adducts. Alternatively, benzene-based alkoxy radicals formed by the transformation of benzene peroxy radicals, can be directly trapped by DMPO.

Mechanistic insight of the photocatalytic activity

Typically, in a photocatalytic process, the following reactions happen at the surface of a semiconductor material:

It has been shown that graphene addition improves the electron transport in TiO2/graphene hybrid materials [31], therefore being a support with an added-value for TiO2 based photocatalytic reactions. The photogenerated electrons, from the TiO2 component, are swiftly transferred to the surrounding graphene sheets. Graphene sheets behave as a highway for the mobility of such electrons, allowing for an increase in the separation of the photogenerated exciton [37]. This is schematically drawn in Fig. 11 .

Fig. 11

Proposed mechanism for improved spatial charge carrier separation in a graphene-TiO2 hybrid material (following to UV–Vis excitation). In the band diagram, relative to the absolute vacuum scale, the dashed yellow arrow shows the electron transferring in titania (from the valence band to the conduction band) due to photocatalysis; the dotted yellow arrow depicts the successive electron transferring from the conduction band of TiO2 to graphene. Energy levels of TiO2 are from specimen G1.0/450 – estimated assuming a direct Eg model, and considering the absolute electronegativity values of the constituent atoms of the material [82], [83]. The value for the working function (WF) of graphene is from Liang and Ang [84]. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Those electrons are then involved in reductive reactions, i.e. the formation of active superoxide ion (O2 •−) species with oxygen from the environment (equation (3)).

According to what is reported above, Section 4.3, and looking at Fig. 12 , it is reasonable assuming NO abatement to mainly follow the NO oxidation path by O2 •− species [85]:

Fig. 12

Concentration of NO and NO2versus reaction time in G1.0/450.

It has been proposed that the de-NO cycle shall be ideally closed by HNO3 removal via water treatment from the photocatalyst surface [Equation (7)] [86]:

In the plot displaying the concentration of NO and NO2 versus reaction time (Fig. 12), we observe the (photocatalytic) abatement of NO, whilst the concentration of NO2 is virtually stable over the time. The final product of NO abatement is HNO3, which can be easily eluted from the catalyst into water [Equation (7)] [87].

A similar hypothesis can be made for IPA degradation. Indeed, Palmisano and colleagues proposed the photocatalytic oxidation of IPA to acetone to follow this path [88]:

The EPR spin trapping technique indeed showed the effective generation of carbon-centred radicals upon exposure, supporting the proposed mechanism.

However, photocatalytic removal of hydrocarbons can be a much more complicated issue. It has been found by Rickard and colleagues that a complete isoprene (C5H8) atmospheric oxidation involved 1,926 different degradation steps, together with the formation of 602 unique intermediate organic species before CO2 and water were eventually formed [89]. Therefore, recognising (qualitatively and quantitatively) all of the intermediated of benzene photocatalytic oxidation is beyond the scope of this work. However, with the available experimental data together with the aid of previous literature, we can speculate about (some of) the degradation products of benzene. The generation of transient radical intermediates upon exposure responsible for the degradation processes of the benzene molecule was confirmed by EPR spin trapping technique. In Fig. 13 are reported the gas-chromatograms as extracted (at selected irradiation times) from the gas-analyser during the photocatalytic experiments – specimen G1.0/450. At time zero, just before switching the solar lamp on, benzene is virtually the only compound found (Fig. 13a). Lighter and heavier degradation compounds than benzene are detected just after 0.5 h of solar irradiation, together with a decrease in benzene concentration, Fig. 13b. Phenol (molar mass, M = 94.11 g.mol−1), 1,4-benzoquinone (M = 108.095 g.mol−1), hydroquinone (M = 110.11 g.mol−1), and benzoic acid (M = 122.12 g.mol−1) were detected by means of GC/MS and DRIFT techniques as photocatalytic benzene removal by-products [90], [91], [80], [92]. Those compounds all have a higher molar mass than benzene (M = 78.11 g.mol−1), therefore their presence is reasonable. Ollis and co-authors also found carboxylic acids – acetic and formic acid (M = 60.052, and 46.03 g.mol−1, respectively) – as intermediates of benzene oxidation [91]. The amounts of benzene and its degradation compound gradually decrease as the exposure time increases (Fig. 13c,d). Eventually, after 3 h irradiation time and ongoing (Fig. 13e-h), chromatograms virtually report only benzene and lighter degradation compounds of it. Therefore, from these results, VOCs should not be treated as a single pollutant [93]. Besides, we can postulate that, as proposed by Lichtin and Sadeghi [94], the formation of those intermediates might delay/inhibit the reaction(s) responsible for benzene’s mineralisation, therefore competing with benzene, and themselves, in the photocatalytic reaction, hampering the overall kinetic of benzene removal. TiO2 gas-phase photocatalytic oxidation of benzene confirms to be a complex matter, having different reaction mechanisms and pathways under different reaction conditions [81]. Those pollutants (i.e. nitrogen oxides and VOCs) might be found mixed together in a real outdoor environment. At this aim, a study by Chan and co-authors showed that the presence of NO promoted the photodegradation of mixed VOCs (BTEX) over Evonik Aeroxide TiO2 P25 under low humidity level; vice versa, the presence of BTEX reduced the conversion of NO, and the generation of NO2 [95].

Fig. 13

G1.0/450 gas-chromatograms as extracted from the VOC-72 M gas analyser at: a) t = 0; b) t = 0.5 h; c) t = 1 h; d) t = 2 h; e) t = 3 h; f) t = 4 h; g) t = 5 h; h) t = 6 h of solar lamp irradiation time. Benzene (symbol 1 in the gas-chromatograms) is at 202 s retention time.

Conclusions

Air pollution is a cause of adverse health effects worldwide, therefore has to be treated as a global problem. To (partly) tackle this concern, we have synthesised, via a green, clean, and simple sol–gel method, titania/graphene (0.5 and 1.0 wt% graphene) hybrid nanomaterials. Photocatalytic activity of as-prepared materials was tested against three major pollutants found outdoor: nitrogen oxides (i.e. NO + NO2), and two different kinds of volatile organic compounds, namely: benzene and isopropanol. EPR spin trapping technique proved that, upon UV-A irradiation, distinct radicals were formed in presence of isopropanol and benzene, thus leading to different yet complex reaction pathways. Results showed that the addition of 1.0 wt% of graphene to TiO2 enabled (at least) a two-fold increase in the photocatalytic removal of those gaseous pollutants, being fully recyclable over repeated tests – thus showing the stability of that system during the oxidising working conditions of photocatalytic reactions. Indeed, the addition of graphene to that semiconducting material acted as a highway for the electron mobility, enhancing the separation of the photogenerated exciton, decreasing their recombination rate. This makes our material an ideal candidate for multi-purpose environmental applications.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.