Brookite : Nothing New under the Sun ?

Advances in the synthesis of pure brookite and brookite-based TiO2 materials have opened the way to fundamental and applicative studies of the once least known TiO2 polymorph. Brookite is now recognized as an active phase, in some cases showing enhanced performance with respect to anatase, rutile or their mixture. The peculiar structure of brookite determines its distinct electronic properties, such as band gap, charge–carrier lifetime and mobility, trapping sites, surface energetics, surface atom arrangements and adsorption sites. Understanding the relationship between these properties and the photocatalytic performances of brookite compared to other TiO2 polymorphs is still a formidable challenge, because of the interplay of many factors contributing to the observed efficiency of a given photocatalyst. Here, the most recent advances in brookite TiO2 material synthesis and applications are summarized, focusing on structure/activity relation studies of phase and morphology-controlled materials. Many questions remain unanswered regarding brookite, but one answer is clear: Is it still worth studying such a hard-to-synthesize, elusive TiO2 polymorph? Yes.


Introduction
Polymorphism-the ability of a solid material to exist in more than one crystal structure-is a common property of metal oxides [1].Different polymorphs have distinct chemico-physical properties, such as electronic and optical properties, magnetism [2], ion conductivity [3], photo/electro-chromism [4], and surface energy and atom arrangement, which in turn affect the material performance in various applications, such as sensing [5] and photocatalysis [6,7].Rationalizing the effect of crystal structure on the performance of a photocatalytic material is a pivotal research goal, since both Fe 2 O 3 and TiO 2 , the most investigated and widely used photocatalysts, may exist in different crystal structures [7,8].Nonetheless, studying such a relation is not a trivial task, because of the many interconnected factors which influence the rate of photochemical reactions (e.g., particle size, shape, crystallinity, number/type of defects, electron-hole recombination, surface area, reactant and intermediate species adsorption on the catalyst surface) and the difficulty of synthesizing materials with a single crystal structure.The latest advances in nanotechnology are shedding new light on the field of photocatalysis, helping to understand the role of crystal structure and morphology in photocatalytic processes.
There are four commonly known polymorphs of TiO 2 : anatase (tetragonal), rutile (tetragonal, the most thermodynamically stable phase), brookite (orthorhombic) and TiO 2 (B) (monoclinic) [9].The structural parameters of these polymorphs are listed in Table 1.Such different lattice structures cause different mass densities and electronic band structures in TiO 2 polymorphs.The exact position of the conduction and valance band edges of TiO 2 polymorphs is still debated, mostly because it is dependent on the purity of the material, its crystallinity and its particle dimensions.While a consensus has been reached for an approximate band gap value, E g , of anatase (3.2 eV) and rutile (3.02 eV), the reported experimental E g values of brookite range from 3.1 to 3.4 eV: both smaller and larger than that of anatase [10].Even more uncertain is the value of E g of monoclinic TiO 2 (B), reported in the range of 3.1-4.1 [11][12][13].Theoretical approaches (e.g., hybrid functionals, GW methods, the Bethe-Salpeter equation (BSE) method) are also not able to accurately estimate both the fundamental and optical band gap of these materials [10,13,14].Nonetheless, determining the band gap, the flatband potential and the energy level of trap states, defects and impurities is crucial for the design of materials with improved photocatalytic properties, so future theoretical and experimental detailed studies are needed in order to solve these issues [15].Table 1.Structural parameters of the main TiO 2 polymorphs.Date reproduce from reference [9], with permission of the American Chemical Society, 2017.Once considered inactive, brookite is now gaining more and more attention in the field of photocatalysis due to its peculiar performance.A very comprehensive review of brookite was published some years ago by Di Paola et al. [10].while some detailed reviews on TiO 2 for photocatalytic applications can be found in the literature, focusing on electronic properties [16], exposure of tailored facets [17] and nanostructural control [18].The present review is focused on the advancement in brookite synthesis for photocatalytic applications, a research field in rapid expansion.First, the most advanced synthetic methodologies to produce pure brookite and well-defined brookite-containing composites are presented, together with some guidelines for thorough characterization of the materials.Then, some applications of brookite as a photocatalyst are outlined in comparison with the performance of other TiO 2 polymorphs.Finally, structure/activity relations are summarized and a perspective on the future development of brookite nanostructured materials is given.

Characterization
TiO 2 polymorphs are usually identified by X-ray diffraction (XRD) or Raman spectroscopy.The presence of brookite in a TiO 2 sample can be evidenced by the presence of the characteristic (121) peak at 2θ = 30.8• in the XRD pattern (Figure 1a).While rutile is also straightforward to identify from XRD patterns, thanks to the (110) peak at 27.43 • , the detection of anatase in an apparently pure brookite sample is complicated by the overlapping of the (101) main diffraction peak of anatase at 25.28 • with the (120) and (111) peaks of brookite, at 25.34 • and 25.28 • , respectively (Figure 1a).In order to quantitatively assess the phase composition of a TiO 2 sample, the whole XRD pattern should be fitted by a structure refinement method, such as the Rietveld method [19].The method is very accurate as it allows us to take into account the broadening of the XRD peaks at higher Bragg diffraction angles and the possible preferred crystal orientations in plate-or rod-like crystallites, for which the reflex intensities vary from that predicted for a completely random distribution.In a study by some of the present authors on the synthesis of TiO 2 nanocrystals with precise and tunable exposed facets, the Rietveld method was used in conjunction with simulations and electron microscopy results to determine the phase and the average nanocrystal dimensions [20].Raman spectroscopy can be used in combination with XRD in order to gain more insight into the phase composition of TiO2 samples.Raman is very sensitive to minor phase impurities which might go undetected by XRD, and allows spatial mapping in the 10 × 10 micrometer resolution in confocal microscope setups.Since the vibrational spectrum of brookite presents more bands than the other two polymorphs, small amounts of brookite can be detected (Figure 1b) [22].The more symmetric anatase and rutile structure gives rise to much simpler Raman spectra, with characteristic features evidenced in Figure 1b [23].Insights into the surface/bulk distribution of phases in a TiO2 sample can be gained by a combination of common visible/IR Raman with UV Raman spectroscopy, which is more surface-sensitive due to the adsorption of UV light by TiO2 [24,25].This approach was used to demonstrate that, in the anatase → rutile transition, rutile forms in the bulk, while anatase can persist on the surface of the sample to higher temperatures than previously thought.Applying this method to the study of brookite-based samples could lead to a deeper understanding of brookite formation and phase transitions.
Transmission electron microscopy (TEM) can provide insights in the phase composition of small areas of the material (hundreds of nanometers in size) by electron diffraction over a selected area electron diffraction pattern (SAED) [26].In this regard, it can be used in combination with XRD, which gives bulk-sensitive results.Moreover, high-resolution TEM (HRTEM) can reveal the presence of amorphous material, which can be also quantified quite precisely by electron energy loss spectroscopy (EELS), provided that a crystalline and amorphous standard are available [27].HRTEM can be used to study preferential exposure of facets, crystal morphology, grain boundaries and presence of defects, which can all influence the photocatalytic activity of a material and are therefore very important in structure/activity relation studies.Nonetheless, in some cases it is hard to distinguish between TiO2 phases, making a definitive assignment difficult if nanocrystals are not suitably aligned with respect to the electron beam [28].Moreover, the presence of defects and lattice deformation can lead to slight changes of diffractograms, so intensive characterization is needed in this kind of HRTEM study [26].Raman spectroscopy can be used in combination with XRD in order to gain more insight into the phase composition of TiO 2 samples.Raman is very sensitive to minor phase impurities which might go undetected by XRD, and allows spatial mapping in the 10 × 10 micrometer resolution in confocal microscope setups.Since the vibrational spectrum of brookite presents more bands than the other two polymorphs, small amounts of brookite can be detected (Figure 1b) [22].The more symmetric anatase and rutile structure gives rise to much simpler Raman spectra, with characteristic features evidenced in Figure 1b [23].Insights into the surface/bulk distribution of phases in a TiO 2 sample can be gained by a combination of common visible/IR Raman with UV Raman spectroscopy, which is more surface-sensitive due to the adsorption of UV light by TiO 2 [24,25].This approach was used to demonstrate that, in the anatase → rutile transition, rutile forms in the bulk, while anatase can persist on the surface of the sample to higher temperatures than previously thought.Applying this method to the study of brookite-based samples could lead to a deeper understanding of brookite formation and phase transitions.
Transmission electron microscopy (TEM) can provide insights in the phase composition of small areas of the material (hundreds of nanometers in size) by electron diffraction over a selected area electron diffraction pattern (SAED) [26].In this regard, it can be used in combination with XRD, which gives bulk-sensitive results.Moreover, high-resolution TEM (HRTEM) can reveal the presence of amorphous material, which can be also quantified quite precisely by electron energy loss spectroscopy (EELS), provided that a crystalline and amorphous standard are available [27].HRTEM can be used to study preferential exposure of facets, crystal morphology, grain boundaries and presence of defects, which can all influence the photocatalytic activity of a material and are therefore very important in structure/activity relation studies.Nonetheless, in some cases it is hard to distinguish between TiO 2 phases, making a definitive assignment difficult if nanocrystals are not suitably aligned with respect to the electron beam [28].Moreover, the presence of defects and lattice deformation can lead to slight changes of diffractograms, so intensive characterization is needed in this kind of HRTEM study [26].
Quite interestingly, X-ray absorption spectroscopy (XAS) techniques have seldom been used to study TiO 2 polymorphism [29][30][31], despite the crystal structure sensitivity of such methods.A reason for this is that XAS techniques are not for routine characterization, since they require access to synchrotron facilities.However, Ti L 3 -edge and O K-edge XAS can be used to study the local Ti and O symmetry and to investigate the crystal structure in great detail.For instance, the pre-edge region of the Ti L 3 -edge allows accurate quantification of the degree of crystalline nature of a sample, in a more sensitive way than is performed by XRD [30].Finally, in-situ studies would give new insights into phase transition processes and in structure-activity relations.
Other characterization techniques, not sensitive to phase composition, are nonetheless essential in order to fully characterize and study TiO 2 materials, especially for their use in photocatalysis and to study structure/activity relations.A comprehensive discussion of all such techniques is out of the scope of this paper, but recent reviews on the topic can be found in the literature [32,33].

Synthesis
The first synthesis of pure brookite dates back to the 1950s [34].Since then, a lot of progress has been made in controlling the phase composition of TiO 2 materials, and now a number of synthetic methods exist to produce pure and mixed-phase brookite, rutile and anatase.The effect of the preparation parameters of TiO 2 materials on the final phase composition has also been widely investigated and was recently reviewed by Kumar and Rao [35].However, the development of new phase-specific synthetic routes remains a trial-and-error process, due to the many thermodynamic and kinetic factors influencing the final phase composition.Indeed, the thermodynamic stability of polymorphs in nanocrystal systems is strongly affected by surface energy contributions, which in turn depend on the nanoparticles size, shape, exposed facets and surface adsorbates, as recently reviewed [36].Kinetic processes such as reactant diffusion and ripening can be even more influential in particle formation and growth regimes, which usually occur far from equilibrium.The fact that brookite synthesis is more challenging than that of anatase and rutile is due to the narrower energy stability window of brookite compared to other polymorphs.
Moreover, brookite, anatase and monoclinic TiO 2 (B) are all metastable phases, and are transformed to the most thermodynamically stable polymorph-rutile-by annealing at high temperatures.Therefore, post-synthetic temperature treatments, usually employed to improve the degree of crystallization, to reduce the sample or to remove organic precursors, should be carefully controlled in order to avoid the phase transformation of metastable polymorphs.The process of phase transformation has been widely investigated, especially on anatase, but its mechanism is not yet consolidated, to the point that all these transitions were observed: anatase to brookite to rutile [37], brookite to anatase to rutile [38], anatase to rutile [39,40], and brookite to rutile [41].This is due to the fact that the phase transformation process depends on many factors, such as starting material, temperature, particle size, surface area [37], surface defect sites and presence of adsorbates [42].The transformation sequence was rationalized in terms of surface enthalpy crossover upon coarsening [37], but further studies are needed in order to shed some light on such a complex process.
Hydrothermal methods are the most commonly employed to synthesize brookite.Briefly, a TiO 2 precursor is mixed with water or organic solvents, an acid or base and additives such as capping/chelating agents or salts, and heated in an autoclave to moderately high temperatures and pressures.Typical Ti precursors are chlorides (TiCl 3 , TiCl 4 ), Ti(SO 4 ) 2 , alkoxides (e.g., isopropoxide, butoxide), alkaline titanates and amorphous titania.In general, there is no one-fit-all strategy in order to obtain brookite; acidic or basic conditions can be employed to obtain pure brookite, depending on the set of conditions, and additives which are necessary in one case can be avoided by changing other synthetic parameters.Nonetheless, varying the reaction parameters in a systematic way, different polymorphs can be obtained [43,44].Adapting the procedure reported by Zhao et al. [45], some of us showed that pure brookite, anatase and rutile with a well-defined shape and dimension can be obtained using titanium (IV) bis(ammonium lactate) dihydroxide, Ti(NH 4 C 3 H 4 O 3 ) 2 (OH) 2 , and varying the concentration of urea in aqueous solution (0 to 7.0 M) at 160 • C for 24 h (Figure 2).Pure brookite nanorod formation was favored by higher urea concentrations, corresponding to higher pH values due to urea decomposition, in accordance with the literature [45,46].Interestingly, the use of comparable urea concentrations to those reported by Zhao et al. resulted in slightly different phase compositions of the final material, probably because of the influence of setup parameters such as autoclave free volume and heating/cooling method [21].The drawbacks of hydrothermal synthesis are high energy consumption, low throughput and of difficult scalability due to the high pressure and temperature conditions of the sealed synthesis vessel, which also makes in-situ monitoring of the synthesis a challenging task.However, to the best of our knowledge, pure brookite has not been synthesized so far by other wet-chemistry methods, such as sol-gel, ionic liquids assisted synthesis, ultrasound and microwave-assisted synthesis and microemulsion [35,47].The reason for this is most probably related to the aforementioned narrower stability window of brookite compared to other polymorphs, but this aspect surely deserves more investigation in the future.
Mechanochemical activation methods, such as ball milling (BM), are emerging as more sustainable synthetic methods because of the absence of bulk solvents and the possibility to use metal oxides or hydroxides as starting materials [35,48].While pure brookite synthesis by BM has not been reported so far, future studies could clarify the role of the reaction parameters on the phase composition of the final material.Moreover, brookite could be used as a starting material in BM synthesis in order to study its stability (and therefore the feasibility of the synthetic method) and the phases formed under different conditions.
Modification of TiO2 by doping (H, N, C, metal ions), co-doping and self-doping (Ti 3+ , black titania [49]) are general strategies to enhance TiO2 light harvesting and photocatalytic properties.Doped brookite synthesis by the introduction of dopants or their precursors in brookite hydrothermal synthesis was recently reported [10].As suggested by a computational study of TiO2 polymorphs (H,N)-co-doping, the effect of doping on the electric properties of anatase, rutile and brookite can be different.For instance, (H,N)-co-doping reduced the band gap of anatase and brookite, but had little effect on rutile [50].Combined experimental and theoretical results on this topic could greatly extend our knowledge of TiO2, and some efforts have been recently spent in this sense [51].
Self-doping of TiO2 by Ti 3+ can be accomplished by various reduction methods, using plasma or thermal, chemical or electrochemical activation [49].With the intent of avoiding harsh, harmful and The drawbacks of hydrothermal synthesis are high energy consumption, low throughput and of difficult scalability due to the high pressure and temperature conditions of the sealed synthesis vessel, which also makes in-situ monitoring of the synthesis a challenging task.However, to the best of our knowledge, pure brookite has not been synthesized so far by other wet-chemistry methods, such as sol-gel, ionic liquids assisted synthesis, ultrasound and microwave-assisted synthesis and microemulsion [35,47].The reason for this is most probably related to the aforementioned narrower stability window of brookite compared to other polymorphs, but this aspect surely deserves more investigation in the future.
Mechanochemical activation methods, such as ball milling (BM), are emerging as more sustainable synthetic methods because of the absence of bulk solvents and the possibility to use metal oxides or hydroxides as starting materials [35,48].While pure brookite synthesis by BM has not been reported so far, future studies could clarify the role of the reaction parameters on the phase composition of the final material.Moreover, brookite could be used as a starting material in BM synthesis in order to study its stability (and therefore the feasibility of the synthetic method) and the phases formed under different conditions.
Modification of TiO 2 by doping (H, N, C, metal ions), co-doping and self-doping (Ti 3+ , black titania [49]) are general strategies to enhance TiO 2 light harvesting and photocatalytic properties.Doped brookite synthesis by the introduction of dopants or their precursors in brookite hydrothermal synthesis was recently reported [10].As suggested by a computational study of TiO 2 polymorphs (H,N)-co-doping, the effect of doping on the electric properties of anatase, rutile and brookite can be different.For instance, (H,N)-co-doping reduced the band gap of anatase and brookite, but had little effect on rutile [50].Combined experimental and theoretical results on this topic could greatly extend our knowledge of TiO 2 , and some efforts have been recently spent in this sense [51].
Self-doping of TiO 2 by Ti 3+ can be accomplished by various reduction methods, using plasma or thermal, chemical or electrochemical activation [49].With the intent of avoiding harsh, harmful and expensive methods, recent studies focused on the development of yet other synthetic routes [52,53].For instance, pure black brookite synthesis was recently reported, in which melted Al was used as a reducing agent in a two-zone vacuum furnace, instead of employing more dangerous H 2 [52].In this case, the Al reduction reportedly starts from the brookite surface and causes lattice disorders and oxygen vacancies in the surface layer.Xi et al. reported a hydrothermal synthesis of black brookite using post-annealing treatment in order to introduce large amount of Ti 3+ defects in the bulk of the nanoparticles [53].Interestingly, no rutile or anatase were detected after N 2 annealing treatment at elevated temperatures (up to 700 • C), suggesting that brookite can be stabilized by reduction.Systematic studies comparing different synthetic routes are greatly desired, since the synthetic method can affect the properties and performances of black TiO 2 materials, leading to the diverse results rationalization found in the literature [49].Controlling the amount and spatial distribution of surface lattice disorder, oxygen vacancies, Ti 3+ ions, Ti-OH and Ti-H groups present in black TiO 2 would also open the way to band-gap engineering in order to enhance photocatalytic performances in targeted reactions.
Finally, in order to scrupulously study structure/activity relations in TiO 2 materials, the synthesized samples should not only have controlled phase composition, but also well-defined crystallite dimensions and morphology.Indeed, different facets of TiO 2 crystals have distinct photocatalytic properties as a result of different atomic and electronic structure [17,54].The nanostructure also plays a pivotal role in photocatalysis.For instance, the rate of electron-hole recombination in brookite nanorods can be controlled by engineering their length, as recently reported by some of us [28].The synthesis of nanostructured TiO 2 has been widely investigated in the literature [18,20,55], so that a wide library of morphology-controlled materials can now be produced.Structure/activity studies on TiO 2 are thriving, as they can provide great insights into reaction mechanisms and give an indication of how to enhance the performance of state-of-the-art photocatalysts.

Photocatalytic Studies
The phase composition of TiO 2 nanomaterials is one of the factors strongly influencing photocatalytic reaction activity and selectivity [56].This is mainly due to the phase-dependency of electronic structures (band gap, charge-carrier lifetime and mobility, exciton diffusion length, trapping sites) and of preferentially exposed surfaces with different energy, atom arrangements and adsorption sites.While anatase-and rutile-based nanomaterials have been widely investigated and their photocatalytic performances are well established [9,17], brookite has slowly gained the status of an efficient photocatalyst only in the last couple of decades, as recently reviewed [10].Here, we report the latest advances in photocatalysis over brookite-based materials, focusing on comparative and in-depth studies, to give a critical view of the present challenges in structure/activity relation studies.
One of the reasons for the different photocatalytic performances of TiO 2 polymorphs is the trapping of photogenerated charge carriers (electrons and holes) in defect sites, resulting in slower electron-hole recombination kinetics and phase-dependent charge carrier stabilization.The density of such occupied mid-gap trap states (DOTS) can be studied by time-resolved vis-IR absorption spectroscopy (TRAS) [57,58].In a recent experimental study, the depth of electron-traps in brookite was estimated to be 0.4 eV, which is deeper than that of anatase (<0.1 eV), but shallower than that of rutile (∼0.9 eV) [57].These results are in good accordance with the calculated stabilization energies for rutile (0.8-1 eV) and anatase (0-0.2 eV), in which oxygen vacancies or interstitial Ti atoms were considered as trap site defects [59,60].Notably, to the best of our knowledge, no theoretical study has been reported on electron trapping at the defects of brookite, even if similar results to anatase and rutile are expected.Some of the authors recently reported a study on trapped carriers in different TiO 2 polymorphs in the ms-s timescales, typical of photocatalytic reactions [58].Also in these slow timescales, anatase and brookite exhibit dispersive power law recombination dynamics, in accordance with shallow charge trapping, while rutile exhibits logarithmic decay kinetics, indicative of deeper charge trapping.
The stabilization of electrons in trap states has two opposite effects: on the one side, it reduces the reactivity of electrons, and on the other, it causes an increase in the lifetime of electrons and holes, because the probability of recombination is decreased.Therefore, an appropriate electron trap depth can help in maximizing the yield of long-lived, but still reactive, charge carriers.In the presence of traps of moderate depth, such as in brookite, electrons can take part in reduction reactions, while long-lived holes can participate in photocatalytic oxidation reactions, such as water oxidation (Figure 3).In anatase, the trap-depth is too shallow to extend the lifetime of holes, while in rutile it is too deep for electron-consuming reactions (Figure 3).Notably, deep electron trapping in rutile extends the lifetime of holes, which could promote water oxidation.However, the low reactivity of deeply trapped electrons causes the overall activity to be low [61].
Catalysts 2017, 7, 304 7 of 19 because the probability of recombination is decreased.Therefore, an appropriate electron trap depth can help in maximizing the yield of long-lived, but still reactive, charge carriers.In the presence of traps of moderate depth, such as in brookite, electrons can take part in reduction reactions, while long-lived holes can participate in photocatalytic oxidation reactions, such as water oxidation (Figure 3).In anatase, the trap-depth is too shallow to extend the lifetime of holes, while in rutile it is too deep for electron-consuming reactions (Figure 3).Notably, deep electron trapping in rutile extends the lifetime of holes, which could promote water oxidation.However, the low reactivity of deeply trapped electrons causes the overall activity to be low [61].Particle size can also have a great influence on photocatalytic activity due to geometric and electronic effects.While a smaller particle size results in larger surface areas (which means more active sites per volume) and shorter mean distances for carriers to migrate to the surface, it also brings about some drawbacks such as less volume to generate and separate charge carriers, an increase of band gap due to quantum confinement and incomplete band relaxation to the bulk level (which means a smaller potential drop and electric field in the space charge region, and hindered charge separation) [62].Therefore, a compromise in particle size between the positive influence of having more reactive sites and the negative influence of increased recombination is wanted.Nonetheless, electron-hole separation can be improved by controlling the morphology of nanocrystals, while maintaining high specific surface areas.
As recently demonstrated by some of the authors by a combination of electron paramagnetic resonance (EPR) spectroscopy and in-situ TRAS [28], H2 production rates in the photoreforming of ethanol, glucose and glycerol increases with the increasing length of Pt-decorated brookite nanorods due to the longer lifetime of charge-separated states (Figure 4).A broad photoinduced absorption feature in the visible range is obtained in TRAS when pumping above the brookite band gap, which was assigned to trapped holes lying energetically within the band gap and physically on or near the surface of the TiO2 nanoparticles [63].The TRAS normalized kinetics reported in Figure 4 show an ultrafast rise due to the trapping of photoexcited holes of the order of a few hundred fs, followed by a decay due to the recombination of holes with electrons or reaction with adsorbates.The increase in Particle size can also have a great influence on photocatalytic activity due to geometric and electronic effects.While a smaller particle size results in larger surface areas (which means more active sites per volume) and shorter mean distances for carriers to migrate to the surface, it also brings about some drawbacks such as less volume to generate and separate charge carriers, an increase of band gap due to quantum confinement and incomplete band relaxation to the bulk level (which means a smaller potential drop and electric field in the space charge region, and hindered charge separation) [62].Therefore, a compromise in particle size between the positive influence of having more reactive sites and the negative influence of increased recombination is wanted.Nonetheless, electron-hole separation can be improved by controlling the morphology of nanocrystals, while maintaining high specific surface areas.
As recently demonstrated by some of the authors by a combination of electron paramagnetic resonance (EPR) spectroscopy and in-situ TRAS [28], H 2 production rates in the photoreforming of ethanol, glucose and glycerol increases with the increasing length of Pt-decorated brookite nanorods Catalysts 2017, 7, 304 8 of 19 due to the longer lifetime of charge-separated states (Figure 4).A broad photoinduced absorption feature in the visible range is obtained in TRAS when pumping above the brookite band gap, which was assigned to trapped holes lying energetically within the band gap and physically on or near the surface of the TiO 2 nanoparticles [63].The TRAS normalized kinetics reported in Figure 4 show an ultrafast rise due to the trapping of photoexcited holes of the order of a few hundred fs, followed by a decay due to the recombination of holes with electrons or reaction with adsorbates.The increase in the lifetime of the trapped holes with nanorod length was attributed to the different surface energy of crystal faces exposed on the tips and the sides of the nanorods, which drives the electrons and holes to different crystal faces, leading to enhanced charge carrier separation in longer nanorods [64].Studies of photocatalytic coupling of methanol to methyl formate and the photo-oxidation of acetaldehyde to acetate on brookite nanorods revealed similar length-dependent rate enhancement, corroborating the proposed model [65].
The EPR results further indicated that electrons in longer nanorods transfer to Pt faster than electrons in shorter nanorods.The capture of electrons by Pt increases the lifetime of the chargeseparated state, as indicated by the much faster decays observed for nanorods without Pt in ethanol [28].A similar effect was observed for Au surface decoration of brookite nanorods [66], which enabled electrons a lifetime of four orders of magnitude longer due to efficient hopping on brookite lateral facets.Conversely, when Au nanoparticles are introduced in the bulk of nanorods, they act as recombination centers for plasmonic carriers in the fs timescale.From the above discussion, it is clear that the control over exposed crystal faces is an essential aspect in fundamental photocatalytic studies, as the surface energy, chemical surface state, number/type of defects and reactivity strongly depend on the atomic arrangement of the exposed surfaces [17,54,67,68].In the last decade, the synthesis of TiO2 anatase, rutile and brookite nanocrystals with controlled exposed faces has been reported by several group, employing specific precursors/conditions or suitable structure-directing agents.All polymorphs show pronounced structure/activity relations, with distinct crystal faces showing dramatically different reactivity and specifically promoting reduction or oxidation [64].Ohno et al. reported clear evidence of such facedependent reactivity by studying the photodeposition of Pt (by reduction of PtCl6 − ) and of PbO2 (by oxidation of Pb 2+ nitrate) on rutile, anatase [69], and brookite [68,70].On brookite TiO2 nanorods, reduction and oxidation reactions were observed to proceed predominantly on {210} and {212} The EPR results further indicated that electrons in longer nanorods transfer to Pt faster than electrons in shorter nanorods.The capture of electrons by Pt increases the lifetime of the charge-separated state, as indicated by the much faster decays observed for nanorods without Pt in ethanol [28].A similar effect was observed for Au surface decoration of brookite nanorods [66], which enabled electrons a lifetime of four orders of magnitude longer due to efficient hopping on brookite lateral facets.Conversely, when Au nanoparticles are introduced in the bulk of nanorods, they act as recombination centers for plasmonic carriers in the fs timescale.
From the above discussion, it is clear that the control over exposed crystal faces is an essential aspect in fundamental photocatalytic studies, as the surface energy, chemical surface state, number/type of defects and reactivity strongly depend on the atomic arrangement of the exposed surfaces [17,54,67,68].In the last decade, the synthesis of TiO 2 anatase, rutile and brookite nanocrystals with controlled exposed faces has been reported by several group, employing specific precursors/conditions or suitable structure-directing agents.(by oxidation of Pb 2+ nitrate) on rutile, anatase [69], and brookite [68,70].On brookite TiO 2 nanorods, reduction and oxidation reactions were observed to proceed predominantly on {210} and {212} exposed crystal faces, respectively [68,70].In brookite nanosheets, {201} facets acted as oxidation sites, while {210} and {101} facets acted as reduction sites [64].Face-specific reactivity was also employed to selectively modify the tips of brookite nanorods with Fe 3+ ions, to enhance the activity in acetaldehyde photocatalytic oxidation [68].
Theoretical studies also evidenced the importance of the energetics of crystal faces.The sequence of surface energies of ten stoichiometric 1 × 1 low-index surfaces of brookite was calculated by first-principle density functional theory (DFT) simulations, showing that the electronic and chemical properties of brookite and the other TiO 2 phases can be significantly different [67].Recently, combined theoretical and experimental studies of brookite nanomaterials demonstrated that preferential exposure of {121} faces, with under-coordinated atoms and lower VB potential, led to higher performance in photodegradation, while preferential exposure of {211} surface, with higher CB potential, resulted in enhanced H 2 productivity [71,72].Such calculations also suggested that electrons struggle to migrate from bulk to {121} faces, in accordance for their poor H 2 production efficiency.
While pure phase TiO 2 materials are extensively investigated for fundamental studies, mixed phase TiO 2 materials have been shown to provide enhanced photocatalytic performances, due to charge transfer across the interface of different phases [73].For instance, the widely investigated TiO 2 Degussa P25, used as a standard in the majority of TiO 2 -related studies, is a mixture of anatase and rutile particles (anatase/rutile ratio ~70:30) in which a synergistic effect of the two phases is ascribed to the spatial separation of photogenerated charge carriers [74].Despite the general consensus in such mechanisms, there is disagreement about the direction of charge transfer, which depends on the relative positions of the conduction bands and trap states in the two polymorphs.Electron transfer from anatase to rutile was proposed by Kawahara et al., based on results of the photodeposition of Ag on a patterned bilayer-type TiO 2 photocatalyst consisting of anatase and rutile phases [75].SEM images clearly showed silver particles mostly deposited on the anatase surface, except at the interfacial region, in which silver deposition occurred preferentially at the rutile layer boundary, suggesting electron transfer from anatase to rutile.On the other hand, the opposite mechanism was proposed by other studies basing on EPR spectroscopy results, in which photogenerated electrons migrated from rutile to anatase trapping sites [74,76].
Recently, a similar enhancement of photocatalytic activity in the photodegradation of organic molecules and CO 2 photoreduction has been also observed for brookite/anatase [45,77,78] and brookite/rutile mixtures [79,80].Moreover, anatase-rich/brookite mixture (75:25) was shown to be much more active for CO 2 photoreduction than Degussa P25, an anatase-rich/rutile mixture with similar anatase fraction [45].These results suggest that electron transfer may take pace also in brookite-based TiO 2 composites, which may be a new direction for the development of efficient photocatalysts.However, to the best of our knowledge, no direct evidence for electron transfer, e.g., by EPR, has been reported so far.DFT calculations on pairs of rutile, anatase and brookite TiO 2 slabs showed that in most cases highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) states were separated in the two phases, indicating that charge transfer may take place upon photoexcitation [81].Such separation was observed to be dependent on the native HOMO-LUMO states of the two components used to build the composites and on the lattice match between the surfaces of the two phases.Future experimental and theoretical investigations focused on the structural and electronic properties of brookite TiO 2 composite interfaces are needed to gain further insights into efficient photocatalyst preparation.On the other hand, since the presence of interfaces can drastically improve the activity of TiO 2 materials, careful characterization is essential to rule out contamination in fundamentals studies aimed to assess the differences in the activity of polymorphs.
It should be noted that the relative activity of polymorphs is dependent on the considered photocatalytic reaction, so that a general activity trend among brookite, anatase and rutile (and their composites) does not exist [82,83].For instance, two distinct activity trends were observed for anatase, brookite and their composites in MeOH photoreforming and in dichloroacetic acid (DCA) degradation [82].This was explained by two different rate-determining factors: H 2 production was favored by the cathodically-shifted flatband potentials of brookite materials with respect to anatase, while DCA degradation rates were highest for anatase, and observed to increase with surface area (Figure 5a).Organic degradation studies can be further complicated by a non-trivial effect of calcination temperature and the sorption capacity of ions and of dissolved gases (e.g., O 2 ) in water on the relative activity of brookite and anatase [83].Normalizing the photocatalytic activity by a certain value may help in rationalizing the results, as observed by normalization to initial absorbed Ag(I) or Cr(VI) amount [83] and by normalization to the surface area in anatase/brookite materials used in photocatalytic reforming of ethanol and glycerol [21].In the latter case, the activity normalized to the surface area continuously increased with the brookite content, indicating that the exposed facets of brookite nanorods possess an intrinsic higher activity in H 2 production than that of the other polymorphs (Figure 5b,c) [21].
Some insights in the factors determining the high intrinsic activity of brookite surfaces come from DFT calculations on the structure and adsorption properties of brookite (210) and anatase (101) surfaces, the most commonly exposed facets in nanocrystals [84,85].The two surfaces are structurally similar, but the brookite (210) surface presents shorter interatomic distances and a different block arrangement.These features result in enhanced reactivity toward strong dissociative adsorption of H 2 O and HCOOH, and generate highly active junction sites, all factors entailing higher specific activity of brookite (210) surfaces [84].On the other hand, charge analysis calculations revealed that CO 2 •− radical anion (the first intermediate in CO 2 reduction) formed on the negatively charged anatase (101) surface, but not on the negatively charged brookite (210) surface, indicating that brookite is not a suitable catalyst for CO 2 reduction [85].Nonetheless, the study confirmed the previous results on the favorable adsorption energetics of brookite (210), and revealed that the presence of oxygen vacancies on the brookite (210) surface enhanced the charge transfer to the CO 2 molecule, promoting CO 2 •− radical formation and CO 2 reduction.
These results were supported by in-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) and CO 2 photoreduction photocatalytic experiments on defect-free and oxygen-deficient brookite, anatase and rutile TiO 2 nanocrystals [86].Oxygen vacancies (V O ) and Ti 3+ sites were created by helium pretreatment of the as-prepared TiO 2 at a moderate temperature on anatase and brookite, but not on rutile.The production of CO and CH 4 from CO 2 photoreduction was remarkably enhanced on defective anatase and brookite TiO 2 (up to 10-fold enhancement) as compared to the defect-free surfaces, thanks to enhanced light harvesting and CO 2 adsorption on defect sites.Defective brookite was the most active photocatalyst among the investigated TiO 2 polymorphs, which was tentatively explained by the favored formation of oxygen vacancies, faster reaction rate of CO 2 − with adsorbed H 2 O or surface OH groups, and an additional reaction route involving an HCOOH intermediate.High CO 2 photoreduction rates were also observed for defective brookite nanosheets with Ti 3+ self-doping, synthesized by an innovative hydrothermal method combined with post-annealing treatment, avoiding commonly employed harsh and costly physical methods [53].Defective brookite nanomaterials are surely worthy of more extensive investigation, in terms of the development of new synthetic routes and of in-depth characterization (e.g., by EPR [87,88]) and photocatalytic studies.
(Figure 5a).Organic degradation studies can be further complicated by a non-trivial effect of calcination temperature and the sorption capacity of ions and of dissolved gases (e.g., O2) in water on the relative activity of brookite and anatase [83].Normalizing the photocatalytic activity by a certain value may help in rationalizing the results, as observed by normalization to initial absorbed Ag(I) or Cr(VI) amount [83] and by normalization to the surface area in anatase/brookite materials used in photocatalytic reforming of ethanol and glycerol [21].In the latter case, the activity normalized to the surface area continuously increased with the brookite content, indicating that the exposed facets of brookite nanorods possess an intrinsic higher activity in H2 production than that of the other polymorphs (Figure 5b,c) [21].with respect the surface area.Markers correspond to the urea/Ti molar ratios that allow the preparation of pure phase materials: rutile (R), anatase (A) and brookite (B).Reproduced from reference [21] with permission of the American Chemical Society, 2017.
Some insights in the factors determining the high intrinsic activity of brookite surfaces come from DFT calculations on the structure and adsorption properties of brookite (210) and anatase (101) Figure 5. (a) Effect of brookite ratio on the photonic efficiency of TiO 2 composites in dicholoracetic acid (DCA) degradation (green triangles) and H 2 production (blue circles).BET surface area (red squares) is reported for comparison with the efficiency trends.Reprinted with permission of reference [82]; (b) H 2 production rate by photocatalytic ethanol dehydrogenation over Pt(0.2 wt.%)/TiO 2 prepared using different urea/Ti molar ratios, normalized with respect to the mass of catalysts and (c) with respect the surface area.Markers correspond to the urea/Ti molar ratios that allow the preparation of pure phase materials: rutile (R), anatase (A) and brookite (B).Reproduced from reference [21] with permission of the American Chemical Society, 2017.
Selectivity studies in photocatalysis have long been lacking, but are now gaining momentum due to the significant advantages presented by photocatalytic reactions as alternative, green routes in organic and inorganic chemistry [89].The selectivity of photocatalytic systems can be enhanced by two main strategies: modification of the photocatalyst and optimization of reaction conditions (e.g., aeration, type of solvent, concentration and type of anions [90], use of membranes) [91].Not surprisingly, the TiO 2 phase composition was demonstrated to have an effect on the selectivity of many photocatalytic reactions, such as ammonia oxidation [92], photoinduced decomposition of acetone, oxygenate photoreforming [6], and selective oxidation of alcohols to aldehydes [91,93].For instance, the product distribution of ammonia oxidation was different in the case of rutile, yielding nitrates as a major product, and of anatase and brookite, yielding mild-oxidation products such as nitrites and N 2 [92].In a study by some of the authors, brookite was observed to be less active in the complete mineralization of glycerol during photocatalytic reforming, leading to higher H 2 /CO 2 and opening interesting opportunities in selective oxidation of sacrificial agents.On rutile, on the other hand, a selective dehydrogenation of glycerol through the secondary OH group was observed (although with very low activity) [6].Furthermore, brookite nanorods were observed to selectively produce H 2 O 2 , a high value-added and green oxidant, during photoelectrochemical water splitting, also showing a very low onset potential for water oxidation (E onset ~−0.2 V vs. reversible hydrogen electrode (RHE)) [66].These findings suggest that the effect of TiO 2 phase composition should be investigated more in detail in the highly novel field of selective photocatalysis.
Future theoretical and experimental studies on well-defined nanostructured TiO 2 materials will be essential to unravel the conundrum regarding the relative photocatalytic performances of different polymorphs and their mixtures, considering the discrepancies observed in the literature.We strongly advise the pursuit of rigorous studies of the electronic structure of TiO 2 nanomaterials and of their facet-specific properties, which requires the development of precise and tunable synthesis methods, yielding size and morphology-controlled materials.The task is arduous, but can be now undertaken using the tools of simulations, high-throughput synthesis and nanotechnology.

Conclusions and Perspectives
Brookite-based materials are receiving increasing attention in photocatalytic and related applications, due to their novelty and peculiar properties.This "brookite rush" is rapidly producing new evidence of the profound influence of phase composition on photocatalytic activity and selectivity, unraveling the mechanistic aspects and structure/activity relations in photocatalytic processes.Nonetheless, we are still fumbling around in the dark regarding the paramount electronic properties of brookite (e.g., its energy gap), so further theoretical and experimental studies are needed in order to reach a new level of understanding and a wide consensus in the scientific community on this topic.However, the solid research work previously published on anatase and rutile seems to have discouraged similar fundamental investigation of brookite.A prime example of this is the case of the postulated electron transfer in the brookite/anatase and brookite/rutile interfaces, which was not supported by direct evidence, since similar work was done for anatase/rutile composites [45,[77][78][79][80].While such electron transfer is reasonably expected, more caution should be taken in future studies of brookite.
An intrinsic burden to the development of a wide range of morphology-controlled brookite nanomaterials is posed by the limited understanding of solvothermal synthetic mechanisms.This is a common issue in the synthesis of nanomaterials, usually carried out by a trial-and-error approach.Nonetheless, nanostructured brookite (e.g., nanorods [28], nanosheets [64], and nanocubes [94]) has been successfully synthesized, and alternative strategies are being proposed to produce highly-oriented brookite and TiO 2 nanoparticles with the desirable specific external crystal facet (e.g., templating with graphene [95] or 2D oxides [96]).The playground is now open to new areas of investigation, such as the development of scalable and sustainable brookite synthesis, modification by doping [10,[50][51][52][53], controlled introduction of defects, and synthesis of brookite nanocomposites [94,97] and hybrid organic-inorganic systems [98].In particular, the effect of TiO 2 phase composition in nanocomposites with metal organic frameworks (MOFs), quantum dots and nanocarbons is still to be assessed.The formation of phase-dependent charged interfaces in such hierarchically-structured materials could be a promising strategy for the design of improved photocatalysts [62].
Investigation of new composites could lead to a great enhancement of photocatalytic performance, stimulating further advances in synthetic methods, characterization techniques and elucidation of catalytic mechanisms.For instance, a non-trivial influence of co-catalyst loading on CO 2 photoreduction activity and selectivity was recently reported for brookite nanocubes loaded with Ag nanoparticles [94].A maximum yield of CO was observed for 0.5 wt.% Ag loading, which corresponded to the case of the preferential deposition of Ag nanoparticles on {210} faces, while for intermediate loadings (1 wt.%) of Ag, the CH 4 selectivity increased, due to an agglomeration of Ag on {210} faces and the deposition of small nanoparticles on {001} faces (Figure 6).These fascinating results were tentatively rationalized in terms of an interplay of charge transfers, adsorption properties and light harvesting effects, revealing how the synergistic effects in semiconductors/co-catalyst systems are much more complex than expected, and how they can be used to finely tune the performances of photocatalytic systems.In-depth studies concerning different band gap and trap states, face reactivity, the effects of crystal morphology and dimension, the presence of vacancies, and the generation of adsorbates and radicals are of great interest, and recently great progress has been made in the study of these properties.Photo-induced heterogeneous electron transfer (ET) across the semiconductor and adsorbed molecules, leading to the formation of radicals (e.g., •OH) is of particular relevance for any photocatalytic reaction, but studying such highly reactive species is challenging [91,99].An emerging approach is based on the use of organic dye probes in microscopic fluorescence imaging, for the sensitive detection of reactive oxygen species, their diffusion in solution or air, and the identification of photocatalytic active facets on semiconductor surfaces [100,101].Recently, a mass spectrometrybased approach was used to investigate the ultrafast ET of photoelectrons generated by ultraviolet irradiation on the surfaces of semiconductor nanoparticles or crystalline facets, providing a new technique for studying the photo-electric properties of various materials [102].Similar studies of brookite would lead to a deeper understanding of the reaction mechanism and provide valuable insights into strategies to enhance activity and selectivity in many photocatalytic reactions.
The phase-dependent activity and selectivity observed in photocatalytic reactions over brookite, rutile and anatase is probably the most exciting and yet still not well investigated aspect concerning TiO2 polymorphs.Brookite nanomaterials were found to perform better than other polymorphs in various photocatalytic reactions (e.g., CO2 photoreduction [86,103], photoreforming of oxygenates [21,82], ammonia oxidation [92]), and showed promising properties for future applications also in other fields, such as photoelectrochemical devices [66], dye-sensitized solar cells (DSSC) [104,105], bio-applications and self-cleaning materials [106,107].The days in which brookite was considered just an undesirable byproduct are long gone, it is now time to dig deeper and find out even more about this and other less familiar TiO2 polymorphs.

Figure 1 .
Figure 1.(a) XRD patterns (colored dots) and Rietveld analysis (black lines) of pure rutile, anatase and brookite (reproduced from [21] with the permission of the Elsevier, 2017) and (b) corresponding Raman spectra.

Figure 1 .
Figure 1.(a) XRD patterns (colored dots) and Rietveld analysis (black lines) of pure rutile, anatase and brookite (reproduced from [21] with the permission of the Elsevier, 2017) and (b) corresponding Raman spectra.

Figure 2 .
Figure 2. (a) Evolution of phase composition and (b) mean crystallite sizes as a function of the initial urea concentration.Reproduced from reference [21] with the permission of the Elsevier, 2017.

Figure 2 .
Figure 2. (a) Evolution of phase composition and (b) mean crystallite sizes as a function of the initial urea concentration.Reproduced from reference [21] with the permission of the Elsevier, 2017.

Figure 3 .
Figure 3. Schematic representation of electron trapping sites in anatase, rutile and brookite.CB: conduction band.Reproduced from reference[57] with permission of the American Chemical Society, 2017.

Figure 3 .
Figure 3. Schematic representation of electron trapping sites in anatase, rutile and brookite.CB: conduction band.Reproduced from reference [57] with permission of the American Chemical Society, 2017.

Figure 4 .
Figure 4. (a) TEM images and size analysis for the brookite nanorods with different lengths; (b) Normalized TRAS (time-resolved vis-IR absorption spectroscopy) kinetic traces of Pt-decorated nanorods in ethanol.Insets show fitted time constants (τ1 (Top) and τ2 (Bottom)); (c) H2 production rates from ethanol photoreforming on 1 wt.% Pt−brookite nanorods of different length after 10 h under illumination, normalized by the surface area of the photocatalysts.Reproduced from reference [28] with the permission of the PNAS, 2017.

Figure 4 .
Figure 4. (a) TEM images and size analysis for the brookite nanorods with different lengths; (b) Normalized TRAS (time-resolved vis-IR absorption spectroscopy) kinetic traces of Pt-decorated nanorods in ethanol.Insets show fitted time constants (τ 1 (Top) and τ 2 (Bottom)); (c) H 2 production rates from ethanol photoreforming on 1 wt.% Pt−brookite nanorods of different length after 10 h under illumination, normalized by the surface area of the photocatalysts.Reproduced from reference [28] with the permission of the PNAS, 2017.
All polymorphs show pronounced structure/activity relations, with distinct crystal faces showing dramatically different reactivity and specifically promoting reduction or oxidation [64].Ohno et al. reported clear evidence of such face-dependent reactivity by studying the photodeposition of Pt (by reduction of PtCl 6 − ) and of PbO 2

Figure 5 .
Figure5.(a) Effect of brookite ratio on the photonic efficiency of TiO2 composites in dicholoracetic acid (DCA) degradation (green triangles) and H2 production (blue circles).BET surface area (red squares) is reported for comparison with the efficiency trends.Reprinted with permission of reference[82]; (b) H2 production rate by photocatalytic ethanol dehydrogenation over Pt(0.2 wt.%)/TiO2 prepared using different urea/Ti molar ratios, normalized with respect to the mass of catalysts and (c) with respect the surface area.Markers correspond to the urea/Ti molar ratios that allow the preparation of pure phase materials: rutile (R), anatase (A) and brookite (B).Reproduced from reference[21] with permission of the American Chemical Society, 2017.