Diverting the use of reduced TiO₂ from a carrier of electron–hole pairs to a reservoir of Brönsted–Lewis acidic pairs for sustainable, non-optical H₂O₂ homolysis

Abstract

Defective TiO₂ bears intra-bandgap states (INTRA) and a valence band (VB) that host electrons and holes, respectively, upon photo-excitation, enabling ˙OH/O₂˙⁻/¹O₂ formation via semi-conducting pathways. However, the energetic window of INTRA/VB is often narrow or mismatches with that of semi-conducting pathways, restricting ˙OH/O₂˙⁻/¹O₂ generation. Here, TiO₂ was calcined at 300 °C (O300) and reduced at 300–600 °C (R300–R600) to tune quantities (N)/strengths (E) of Brönsted acidic (BA⁻–H⁺) and Lewis acidic sites with coordination numbers of 5 (LA_I) or ≤4 (LA_II). The catalysts revealed distinct N/E values with hierarchies of N_LAII < N_LAI ≪ N_BA--H⁺ or E_BA--H⁺ < E_LAI ∼ E_LAII. Non-photonic H₂O₂ homolysis (H₂O₂ → 2˙OH) proceeded via BA⁻–H⁺/LA_II, yet hinged on BA⁻–H⁺ with E_BA--H⁺ and N_BA--H⁺ dictating the energy barrier (E_BARRIER) and lumped collision frequency , respectively, as corroborated by the H₂O₂ consumption rate (−r_H2O2) law. −r_H2O2 values of O300/R300/R600 were challenging to assess, thus being assessed using their initial acetaminophen degradation rates in a per-BA⁻–H⁺ site (−r_{ACETAMINOPHEN, 0, CO2}) or on a per-gram basis (−r_{ACETAMINOPHEN, 0}). R600 showed the highest E_BA--H⁺ and smallest N_BA--H⁺, thus yielding the lowest E_BARRIER and lowest , respectively, where the former overrode the latter to achieve the highest −r_{ACETAMINOPHEN, 0, CO2} for R600. R600 displayed the highest resistance to the deposition of poisonous oligomer/˙OOH on BA⁻–H⁺/LA_II or poisonous O₂ on LA_I due to its smallest N_BA--H⁺/highest E_BA--H⁺/highest E_LAII for oligomer/˙OOH and the smallest N_LAI/highest E_LAI for O₂, as also substantiated by density functional theory calculations. Consequently, −r_{ACETAMINOPHEN, 0} values of R600 were higher than those of O300/R300 throughout acetaminophen decomposition recycle runs, whereas R600 was moderately recovered upon regeneration. −r_{POLLUTANT, 0, CO2} values of R600 were 2–23-fold and 3–10-fold higher than those of O300/R300 and ZrO₂/UiO-66, respectively, in disintegration of diverse aqueous pollutants (analgesic/endocrine disruptor/pesticide/antibiotic).

---

Introduction

Titanium (Ti; [Ar] 3d² 4s²) readily donates four electrons (e⁻) to two oxygens (O; [He] 2s² 2p⁴) under oxidative conditions.^1,2 This results in the crystallization of TiO₂ polymorphs such as tetragonal anatase, tetragonal rutile, or orthorhombic brookite, each of which consists of octahedral [Ti⁴⁺–(O²⁻)₆]⁸⁻ sub-units.³ The Ti⁴⁺ cation of [Ti⁴⁺–(O²⁻)₆]⁸⁻ is hexa-fold coordinated by O²⁻ anions and lacks e⁻ in its valence 3d/4s sub-shells, thereby rendering it inaccessible to e⁻-ample molecules.^1,3–5 These features make Ti⁴⁺ bind rigidly with O²⁻ and reveal high resistance to e⁻ gain/loss or leaching for TiO₂ even upon its exposure to severe redox environments.^6–8 TiO₂ also imparts a characteristic band structure, whose conduction band (CB) and valence band (VB) comprise Ti⁴⁺ cations and O²⁻ anions, respectively. TiO₂ unveils the bandgap energy (E_g) defined by the energetic difference between the CB and VB edges (E_CB–E_VB) of 3.0–3.2 eV.^9–11 TiO₂ thereby needs ultraviolet (UV; ≥3.1 eV) light for its photo-excitation, wherein e⁻ can migrate from the VB to the CB, generating CB electrons (e_CB⁻) and VB holes (h_VB⁺). These e_CB⁻/h_VB⁺ pairs trigger semi-conducting mechanisms that yield H⁺, H₂O₂, ˙OH, O₂˙⁻, or ¹O₂ via e⁻ donation/withdrawal under aqueous conditions (Table S1). Among these, ˙OH outcompetes O₂˙⁻/¹O₂ in destabilizing refractory toxicants, as proved by redox potentials (E_REDOX) of ∼2.7 V for ˙OH and ≤∼2.2 V for O₂˙⁻/¹O₂.^9,10,12–18

Despite the marked oxophilicity of TiO₂, it consists of O–Ti⁴⁺–O–Ti⁴⁺ channels, a fraction of which are surface-terminated and facilitate the release of oxygen (1/2O₂) via O–Ti⁴⁺–O–Ti⁴⁺ → O–•Ti⁴⁺–O_V–•Ti⁴⁺ + 1/2O₂ → O_V–•Ti⁴⁺•–O_V–•Ti⁴⁺ + O₂ or the formation of labile oxygen (˙O_L) dangling to Ti⁴⁺ via O–Ti⁴⁺–O–Ti⁴⁺ → O–•Ti⁴⁺–O_V–Ti⁴⁺–˙O_L (*-***; Fig. 1A/C). These processes also involve the generation of oxygen vacancies (O_V) adjacent to Lewis acidic Ti⁴⁺ defects with coordination numbers of 5 (•Ti⁴⁺; LA_I) or ≤4 (•Ti⁴⁺•; LA_II).^19–22 Notably, LA_I and LA_II sites trap e⁻ (•) in proximity in lieu of filling their empty 3d/4s sub-shells with e⁻ (•), indicating that e⁻ (•) can be delocalized across the CB.^1,19–24 However, when O_V sites are bountiful on the TiO₂ surface, intra-bandgap states can be formed between the CB and VB, while localizing e⁻ on the intra-bandgap state (e_INTRA⁻) with an energy of E_INTRA.^23–26 This reduces the energy required to commence semi-conducting mechanisms from E_g (e_CB⁻/h_VB⁺ in defect-less TiO₂) to E_INTRA–E_VB (e_INTRA⁻/h_VB⁺ in defective TiO₂).^23–26 In some cases, E_INTRA–E_VB is lower than the UV threshold (≥3.1 eV), thereby prompting the production of e_INTRA⁻/h_VB⁺ pairs under visible light irradiation (<3.1 eV) and accelerating semi-conducting pathways.^23–26

Fig. 1 (A) Illustration of tetragonal TiO₂ (anatase) adducted to H₂O molecules (TiO₂·H₂O), most of which undergo dissection to generate H˙ and ˙OH (****) under calcination or reduction conditions. This results in the production of TiO₂ architectures, whose surfaces are terminated to bear Brönsted acidic bonds (–OH; BA⁻–H⁺) and Lewis acidic Ti⁴⁺ defects with coordination numbers of 5 (LA_I) or ≤4 (LA_II). (B) Representation of an octahedral TiO₂ sub-unit, where some of O²⁻ anions hexa-fold coordinated to a Ti⁴⁺ center are released via dehydration (*****) under H₂-mediated reductive conditions, making the Ti⁴⁺ center and O²⁻ anions act as an LA_I (or LA_II) defect and conjugate bases of BA⁻–H⁺ bonds (BA⁻), respectively. Moreover, this can also distort the TiO₂ sub-unit, where the Ti⁴⁺ center and O²⁻ anions are displaced from their original positions (e.g., d_Ti–O reduction) and subtly alter their electronic features, thereby making the Ti⁴⁺ center and O²⁻ anions exhibit elevated electron (e⁻) affinity (E_LAI (or E_LAII)↑) and deprotonation tendency (E_BA--H⁺↑), respectively. (C) Illustration of a surface Ti⁴⁺–O–Ti⁴⁺–O–Ti⁴⁺–O–Ti⁴⁺ channel prone to fragmentation under calcination or reduction conditions, where 1/2O₂ (*/**/***) is released, labile O (˙O_L) is produced and functions as BA that can be covalently bonded to H˙ (****), LA_I/LA_II are generated with LA_II functioning as an anchoring spot for BA⁻–H⁺ (˙OH; ****), or BA⁻–H⁺ is lost yet replaced by LA_I (or LA_II) via H₂-enabled dehydration. (D) Representation of elementary steps proposed to depict homolytic H₂O₂ dissection activated by BA⁻–H⁺/LA_II on the terminated TiO₂ surface by way of the formation of a hydrogen bond between BA⁻ and H of H₂O₂ (i), a coordinate bond between LA_II and O of H₂O₂ (ii), and a covalent bond between LA_II and O of ˙OH (iii), among which hydrogen (i) and covalent bonds (iii) should be weakened by elevating E_BA--H⁺ and E_LAII, respectively, for lowering the energy barrier required to proceed with the rate-determining step (˙OH desorption) of catalytic H₂O₂ homolysis.

Nonetheless, photo-excitation of defective TiO₂ polymorphs using low-energy light sources remains challenging for several reasons. First, O_V formation energies (E_OV) of TiO₂ polymorphs increase in the sequence of anatase (∼4.4 eV) < brookite (∼5.2 eV) < rutile (∼5.6 eV), as assessed via density-functional theory (DFT) calculations.²⁷ In this context, anatase can be most suitable to afford plentiful O_V (or LA_I/LA_II) sites on the surface, yet it can readily undergo reductive transition to rutile at ≥700 °C, unless the rutile is stabilized in amorphous or nano-scale forms.^27,28 Second, E_INTRA and E_VB of defective TiO₂ immersed in aqueous media can be sensitive to pH values or quantities of aqueous species vicinal to TiO₂. These can also alter the E_REDOX values of semi-conducting pathways, which should lie between E_INTRA and E_VB for optical activation.^29–31 Indeed, it is inherently challenging to fine-tune E_INTRA, E_VB, and E_REDOX for selective acceleration of semi-conducting pathways on defective TiO₂ via aqueous photo-excitation.^29–31 Moreover, reduction of E_INTRA–E_VB of defective TiO₂ can elevate productivities of e_INTRA⁻/h_VB⁺ pairs under visible light, yet simultaneously narrows the optical energy window, thereby limiting the range of semi-conducting pathways that can be activated.^{23–26,29–31}

The limitations of semi-conducting mechanisms stated earlier motivate the use of O_V (LA_I/LA_II) to prompt ˙OH generation on defective TiO₂ via a non-optical route, H₂O₂ homolysis in particular (vide infra). Although it is arduous to significantly increase LA_I/LA_II quantities on the defective TiO₂ surface, LA_I and LA_II can serve as anchoring spots for oxygenated species with unpaired e⁻ (˙OH or ˙OOH) via covalent interactions and those with an unpaired e⁻ or a lone e⁻ pair (˙OH/˙OOH or H₂O/H₂O₂) via covalent or coordinate interactions, respectively (ii and iii; Fig. 1D).^19–22 Compared with LA_II sites, their LA_I counterparts are more prevalent on the defective TiO₂ surface due to strong oxophilicity for Ti, indicating N_LAII < N_LAI.^9,10 LA_II sites can homolytically cleave surface-wandering, plentiful H₂O to yield H˙ and ˙OH with the former and the latter binding with ˙O_L and LA_I/LA_II, respectively, to form abundant Brönsted acidic –OH groups (BA⁻–H⁺; ****; Fig. 1A/C).^19–22 Consequently, the amount of BA⁻–H⁺ sites can be far larger than those of their LA_I/LA_II counterparts (N_LAII < N_LAI ≪ N_B--H⁺).^9,10,19–22 Defective TiO₂ can then undergo thermal H₂ reduction, where BA⁻–H⁺ bound to LA_I (or LA_II) can be lost via H˙–assisted dehydration to recover LA_I (or LA_II), whereas O located in the middle of two Ti⁴⁺ for O–Ti⁴⁺–O–Ti⁴⁺ can be consumed via 2H˙-assisted dehydration to evolve additional LA_I/LA_II (*****; Fig. 1C).^19–22 Hypothetically, thermal H₂ reduction can help increase the number of LA_I (or LA_II) species at the expense of a decrease in the number of BA⁻–H⁺ sites. Nevertheless, the hierarchy of N_LAII < N_LAI ≪ N_BA--H⁺ is likely to be preserved, given the comparatively larger population of BA⁻–H⁺ sites relative to those of LA_I/LA_II species, as reported previously (hypothesis I; N_LAI↑, N_LAII↑, N_BA--H⁺↓, and N_LAII < N_LAI ≪ N_BA--H⁺ via H₂ reduction).^19–22

Meanwhile, the [Ti⁴⁺–(O²⁻)₆]⁸⁻ units in TiO₂ can undergo thermal H₂ reduction, where terminal O²⁻ species are removed via H⁺-mediated dehydration and generate O_V sites. This converts Ti⁴⁺ centers into open LA_I/LA_II sites (N_LAI↑ and N_LAII↑), whereas Ti⁴⁺ centers move toward O²⁻ anions positioned at the opposite sides of O_V sites, and shorten the Ti⁴⁺–O²⁻ bond lengths (d_Ti–O↓) to compensate for e⁻ loss triggered by O²⁻ removal (*****; Fig. 1B).²⁵ These structural changes suggest that thermal H₂ reduction can elevate Lewis acidic strengths (e⁻ deficiency) of LA_I and LA_II sites (hypothesis II; E_LAI↑ and E_LAII↑ via H₂ reduction), whose magnitudes can further increase when LA_I and LA_II sites are positioned on kinks rather than on ledges/terraces at higher H₂ reduction temperatures (T_REDUCTION), as reported previously.^{19–22,25,32,33} Concomitantly, the four equatorial O²⁻ anions vicinal to the open Ti⁴⁺ center move toward O_V to sterically minimize e⁻ repulsion (Fig. 1B), leading to distortion of the [Ti⁴⁺–(O²⁻)₆]⁸⁻ unit that facilitates e⁻ migration from O²⁻ to the open Ti⁴⁺ center.^19–22,25 Notably, O²⁻ anions can function as conjugate bases of BA⁻–H⁺ sites (BA⁻) inclined to bind with surface-roaming H⁺ species.^19–22 In this context, it is sound that O²⁻ (BA⁻) can be less nucleophilic and more vulnerable to H⁺ loss (N_BA--H⁺↓) post O_V inclusion in the [Ti⁴⁺–(O²⁻)₆]⁸⁻ unit distorted via thermal H₂ reduction (hypothesis III; Brönsted acidic strength of BA⁻–H⁺ (E_BA--H⁺)↑ via H₂ reduction).^19–22,25 Overall, hypotheses I–III boost the claim that thermal H₂ reduction can modulate the defective TiO₂ surface conducive to activate non-optical H₂O₂ homolysis (H₂O₂ → 2˙OH) on LA_II/BA⁻–H⁺ pairs via the elementary steps proposed below.^19–22

Two BA⁻–H⁺ sites can exothermically adsorb H₂O₂ via the formation of two hydrogen-bonds between BA⁻ species of BA⁻–H⁺ sites and the H atoms of H₂O₂ (i; Fig. 1D).^19–22 This suggests that H₂O₂ can interact preferentially with BA⁻–H⁺ sites rather than with their LA_II counterparts on account of E_BA--H⁺ < E_LAII (hypothesis IV).^19–22 One O atom of the H₂O₂ bound to two BA⁻–H⁺ can then endothermically coordinate to a LA_II site via distortion of the H₂O₂ geometry. This interaction is more pronounced at a higher E_LAII due to greater intimacy between LA_II with the lone e⁻ pair of H₂O₂ (i and ii; Fig. 1D).^19–22 Afterward, the remaining O atom of H₂O₂ bound to two BA⁻–H⁺/LA_II sites undergoes coordination to another adjacent LA_II site, highlighting the significance of LA_II sites adjacent to two BA⁻–H⁺ species.^19–22 The H₂O₂ bound to two BA⁻–H⁺ and two LA_II sites experiences exothermic dissection into two ˙OH radicals, whose O atoms are covalently bonded to LA_II sites, whereas their H atoms are hydrogen-bonded to BA⁻ species of BA⁻–H⁺ sites (i and iii; Fig. 1D).^19–22 Endothermic ˙OH desorption can be expedited at a higher E_BA--H⁺ or a higher E_LAII, where E_BA--H⁺ < E_LAII permits the speculation that detachment of ˙OH from LA_II is energetically more favorable than that from BA⁻–H⁺.^19–22 This leaves ˙OH desorption from BA⁻–H⁺ as the rate-determining step (RDS) of H₂O₂ homolysis (hypothesis V).^19–22 The energy barrier (E_BARRIER) of the RDS can be lower at a higher E_BA--H⁺, whose achievement is of high likelihood on defective TiO₂ subjected to thermal H₂ reduction, as speculated in hypothesis III.^19–22 This makes it compelling that BA⁻–H⁺, rather than LA_II, is the primary active site for H₂O₂ homolysis (hypothesis VI). The lumped collision frequency can be higher at a greater N_BA--H⁺, suggesting that an optimum T_REDUCTION is required to balance E_BARRIER (or E_BA--H⁺) and (or N_BA--H⁺) for maximizing ˙OH productivity on defective TiO₂.^19–22

Notably, ˙OH adsorbed on BA⁻–H⁺/LA_II can react with nearby H₂O₂ to release H₂O and generate ˙OOH bound to BA⁻–H⁺/LA_II (poison I; Fig. 8A). Some of the ˙OOH species can deprotonate at pH values of ≥4.8 to yield O₂˙⁻ analogues bound to LA_II sites (poison II; Fig. 8A).^21,34 In addition, two vicinal •Ti⁴⁺ sites (LA_I; O_V–•Ti⁴⁺–O–•Ti⁴⁺) on the defective TiO₂ surface can also interact with H₂O₂, release two H˙ species, and evolve O₂ with two O atoms sharing e⁻ with two LA_I sites via the formation of covalent bonds (poison III; Fig. 8A).^34–36 Poisons I–III can scarcely be desorbed from ˙OH generators (BA⁻–H⁺/LA_II) or LA_I under certain conditions: a lower E_BA--H⁺ or a lower E_LAII retains poison I, a higher E_LAII stabilizes poison II, and a lower E_LAI persists poison III.²¹ The relative abundance of N_LAII < N_LAI can further exacerbate the poisoning issue because LA_I clusters adjacent to LA_II species can sterically hamper the latter to bind with H₂O₂, thus suppressing the H₂O₂ distortion and cleavage steps needed for H₂O₂ homolysis (ii–iii; Fig. 1D). Again, ˙OH is highly effective in degrading aqueous organics via addition or H˙ abstraction, either of which, however, facilitates the production of oligomeric by-products. Such oligomers often contain e⁻-donating groups (e.g., –OH, –NH₂, or –NH) that bind strongly with N_BA--H⁺ or LA_II sites at a higher E_BA--H⁺ or a higher E_LAII, in particular.^21,37–39 All of these factors are combined to highlight the centrality of T_REDUCTION optimization for defective TiO₂ to minimize the formation and deposition of Ti-peroxy species (poisons I–III) and oligomers on ˙OH producers (BA⁻–H⁺/LA_II) or LA_I.

This study targets to reframe H₂ reduction-subjected, defective TiO₂ as a reservoir of BA⁻–H⁺/LA_II pairs, not a carrier of e_INTRA⁻/h_VB⁺ pairs, where the former and the latter act as ˙OH evolvers in the absence and the presence of low-energy light, respectively. To this end, TiO₂ catalysts were synthesized oxidatively at 300 °C or reductively at 300–600 °C. Their morphological, textural, crystallographic, band-structural, electronic, and acidic features were systematically inspected via microscopy, spectroscopy, isotherm, and chemisorption techniques, some of which were complemented by DFT calculations. The catalysts were further examined via mechanistic and kinetic analysis under controlled conditions. Their activities , reusabilities (poison resistance), or adaptabilities in degrading diverse model organics such as analgesics, antibiotics, pesticides, or endocrine disruptors were explored rigorously to identify the optimal T_REDUCTION for defective TiO₂ and formulate a rate law for H₂O₂ homolysis that has been sparsely elaborated to date.

Results and discussion

Generic traits of the catalysts

TiO(SO₄) and urea (CO(NH₂)₂) were dissolved in an aqueous medium for use as precursors of Ti⁴⁺ and NH₃·H₂O, respectively. The latter was generated via urea pyrolysis under reflux and subsequently interacted with Ti⁴⁺ to precipitate TiO₂·H₂O (TiO(OH)₂; Fig. 1A).^40,41 The TiO₂·H₂O surface could contain O–Ti⁴⁺–O–Ti⁴⁺ channels inclined to termination assisted by NH₃ release from NH₃·H₂O, resulting in the formation of LA_I/LA_II/˙O_L species (*-***; Fig. 1C).^40,41 TiO₂·H₂O then transformed into R300–R600 via exposure to reductive conditions at 300–600 °C, where H˙/˙OH produced from H₂O homolysis and H₂ supplied could compete in modulating N_BA--H⁺/N_LAI (or N_LAII) such as N_BA--H⁺↑/N_LAI (or N_LAII)↓ for H˙/˙OH (****; Fig. 1A/C) and N_BA--H⁺↓/N_LAI (or N_LAII)↑ for H₂ (*****; Fig. 1B/C).^19–22 The temperature range of 300–600 °C was chosen because anatase with the lowest E_OV among TiO₂ polymorphs began to crystallize at 300 °C and remained stable without transition to rutile at ≤600 °C (not shown). For comparison, TiO₂·H₂O was also converted to O300 via exposure to oxidative environments at 300 °C (T_OXIDATION), where H₂O could dissociate into H˙ and ˙OH with the former being covalently bonded with ˙O_L and the latter being anchored on LA_I/LA_II, tentatively leading to N_BA--H⁺↑/N_LAI (or N_LAII)↓ (****; Fig. 1A/C).^19–22 O300 thus served as a reference to disclose the merits of R300–R600 in accelerating H₂O₂ homolysis (vide infra).

To ensure that the catalysts should not be activated optically under visible light (<3.1 eV), they were subjected to diffuse reflectance ultraviolet-visible (UV-vis) spectroscopy experiments. Plots of (F(R_∞)·hν)^1/2 versus hν were constructed for the catalysts with F(R_∞), h, and ν corresponding to the reflectance of the catalyst, the Planck constant, and the frequency of the photon, respectively (Fig. S1 and eqn (S1)).^42–44 The plots were thereafter analyzed using Tauc fits, whose X-intercepts corresponded to the E_g values of the catalysts (3.1–3.3 eV). These E_g values were comparable to those reported for TiO₂ polymorphs (Table S2).^9,10,42–45 The results indicated that the catalysts could hardly generate e_CB⁻/h_VB⁺ pairs for activating semi-conducting pathways when visible light was incident. This also proved that the catalysts contained only a limited number of O_V (or LA_I/LA_II) sites that were insufficient to generate abundant intra-bandgap states. Indeed, the catalysts could barely localize e⁻ on the intra-bandgap states under ambient conditions.^23–26

Morphological and surface structural features of the catalysts were explored using their high-resolution transmission electron microscopy (HRTEM) images (Fig. 2A–E). The images showed TiO₂ congregates with sizes of ≤50 nm and lattice fringes with d-spacings (d_LATTICE) of 0.35 nm, corresponding to the surface (101) facet of tetragonal anatase. This was consistent with the bulk crystallographic traits of the catalysts examined using their XRD patterns, where all the bulk diffractions were indexed to those of tetragonal anatase (Fig. S2). Notably, the HRTEM images of the catalysts also bore bunched dark spots near local lattice dislocations (Fig. S3). These were attributed to dark e⁻-rich domains that can readily absorb light, strongly suggesting the presence of O_V sites in proximity to LA_I/LA_II sites or ˙O_L/BA⁻–H⁺ species used to activate H₂O₂ homolysis.^46–48 Overall, despite the challenge of sufficient O_V (or LA_I/LA_II) production on the catalyst surfaces (Tauc fits; Fig. S1), the catalysts (anatase) still provided a higher propensity to bear defective LA_I/LA_II sites in comparison with brookite/rutile under O₂ calcination or H₂ reduction conditions, as conjectured from the E_OV values reported for TiO₂ polymorphs.²⁷

Fig. 2 HRTEM images of the catalysts (O300 for (A); R300 for (B); R400 for (C); R500 for (D); R600 for (E)), where highlighted with yellow lines and arrows are lattice fringes with d-spacings (d_LATTICE) of 0.35 nm, which are indexed to the surface (101) facet of tetragonal TiO₂ (anatase; JCPDF no. of 01-071-1166). Raman spectra of the catalysts (O300 for (F); R300 for (G); R400 for (H); R500 for (I); R600 for (J)), where gray solid lines and black empty circles indicate raw spectra with a resolution of 0.5 cm⁻¹ and those fitted using Lorentzian functions, respectively, whereas orange/green/blue/cyan/magenta empty circles correspond to Raman-active stretching vibrational modes for Ti⁴⁺–O–Ti⁴⁺–O–Ti⁴⁺–O–Ti⁴⁺ channels belonging to tetragonal TiO₂ (along the a-axis of TiO₂ lattices for E_g; along the c-axis of TiO₂ lattices for A_1g/B_1g). Moreover, in (F–J), marked with orange arrows are E_g bands with Raman shifts centered at 146–151 cm⁻¹ and their full width at half maximum intensity (FWHM) values.

Meanwhile, TiO₂ crystallite sizes (d_CRYSTALLITE) of the catalysts on the bulk (101) or (200) facets were evaluated using the Scherrer equation (Table S2 and eqn (S2)). The d_CRYSTALLITE values increased in the sequence of O300 ∼ R300 ∼ R400 (∼6.3 nm) < R500 (∼7.8 nm) < R600 (∼14.7 nm), indicating noticeable aggregation of TiO₂ particulates pertaining to R500 and R600.^49,50 This was in line with the textural features of the catalysts inspected using their type IV N₂ isotherms (Fig. S4 and Table S2). The N₂-accessible Brunauer–Emmett–Teller surface areas (S_BET) increased in the order of R600 (∼35 m² g_CAT⁻¹) < R500 (∼170 m² g_CAT⁻¹) < R400 (∼210 m² g_CAT⁻¹) < R300 ∼ O300 (∼250 m² g_CAT⁻¹), whereas their N₂-accessible Barrett–Joyner–Halenda pore volumes remained similar (0.1–0.2 cm³ g_CAT⁻¹). The d_CRYSTALLITE/S_BET values of the catalysts suggested that the congregation of H₂-exposed TiO₂ poly-crystallites at ≥500 °C could markedly reduce N_BA--H⁺/N_LAI/N_LAII of R500 and R600 and therefore should be deemed as considerable as H₂O homolysis-induced H˙/˙OH and H₂ treatment in varying their N_BA--H⁺/N_LAI/N_LAII.

Interestingly, the catalysts (anatase) are terminated on the primary bulk (004) and (200) facets identified in their XRD patterns, leading to exposure of the surfaces abundant with LA_II sites with coordination numbers of 3–4 that were accessible to H₂O₂/˙OH/˙OOH/O₂˙⁻ (Fig. S5). If this held true, shifts in d_LATTICE values or 2θ values of major bulk diffractions should be seen in HRTEM images and XRD patterns of the catalysts, respectively, yet were verified not to be so. This suggested that LA_I/LA_II formation could incur only short-range distortions in the TiO₂ lattice ([Ti⁴⁺–(O²⁻)₆]⁸⁻ units).¹ In this sense, the catalysts were analyzed via Raman spectroscopy to further inspect their LA_I/LA_II properties or those associated, where N_LAI/N_LAII could be smaller at a lower S_BET, a longer d_Ti–O, or a greater d_CRYSTALLITE.²⁵ The Raman spectra were curve-fitted to unveil five characteristic bands indexed to either E_g or A_1g/B_1g modes (Fig. 2F–J and Table S3), resulting from Raman-excited stretching vibrations of defective O–Ti⁴⁺–O–Ti⁴⁺ channels along the a-axis for E_g and along the c-axis for A_1g/B_1g.^51,52 Notably, it was reported that Raman-unscattered photons could be confined in LA_I/LA_II-proximal O_V sites and enabled to broaden the E_g bands centered at Raman shifts of 146–151 cm⁻¹ to greater extent if the quantity of photons trapped in O_V sites were larger.^53,54 In this regard, the full width at half maximum intensities (FWHM) of the E_g bands served to gauge their broadness and were found to increase in the sequence of R600 ∼ R500 (∼16.6 cm⁻¹) < R400 (∼18.6 cm⁻¹) < R300 ∼ O300 (∼21.3 cm⁻¹). This paralleled the trend in S_BET values of R600 < R500 < R400 < R300 ∼ O300, suggesting an increase in N_LAI/N_LAII in the order of R600 ∼ R500 < R400 < R300 ∼ O300.^53,54

Meanwhile, the E_g bands were red-shifted with the increase in T_REDUCTION from ∼151 cm⁻¹ at 300–400 °C, to ∼149 cm⁻¹ at 500 °C, and to ∼146 cm⁻¹ at 600 °C.^51,52 These shifts indicated that the photon energy required for defective O–Ti⁴⁺–O–Ti⁴⁺ vibration increased in the order of R600 < R500 < R400 ∼ R300 ∼ O300. This could originate from the increase in the d_Ti–O values for the catalysts mediated by a decrease in N_LAI/N_LAII with the elevation of T_REDUCTION. This trend matched that of the d_CRYSTALLITE values: O300 ∼ R300 ∼ R400 < R500 < R600.²⁵ This proved that TiO₂ aggregation (N_BA--H⁺↓; N_LAI/N_LAII↓) could override the competing effects of H₂O homolysis-induced H˙/˙OH (N_BA--H⁺↑; N_LAI/N_LAII↓) and H₂ treatment (N_BA--H⁺↓; N_LAI/N_LAII↑) in varying N_BA--H⁺/N_LAI/N_LAII of the catalysts subjected to H₂ reduction at ≥500 °C. Overall, the Raman spectral results substantiated that N_BA--H⁺/N_LAI/N_LAII of R500 and R600 could be significantly smaller than those of O300, R300, and R400.

Acidic traits of the catalysts: XP and EPR spectroscopy

Surface elemental features of the catalysts were examined using X-ray photoelectron (XP) spectroscopy. Again, the catalyst surfaces comprised defective O–Ti⁴⁺–O–Ti⁴⁺ channels, where e⁻ trapped in LA_I/LA_II sites (defective •Ti⁴⁺/•Ti⁴⁺• with coordination numbers of 5/≤4) cannot occupy the empty 3d/4s sub-shells of LA_I/LA_II or alter their formal oxidation states (+4), unless the electrophilicities of LA_I/LA_II (E_LAI/E_LAII) are sufficiently high.^1,19–24 In this sense, the XP spectra of the surfaces in the Ti 2p domains were curve-fitted to unveil two sub-bands (Fig. 3A and Table S4).^55,56 These could be assigned to Ti⁴⁺ species with binding energies centered at 464.9–465.4 eV and 459.2–459.7 eV in the Ti 2p_1/2 and in the Ti 2p_3/2 regimes, respectively. These binding energies reflected the combined influence of N_LAI/N_LAII and E_LAI/E_LAII for defective •Ti⁴⁺/•Ti⁴⁺• (N_LAI/N_LAII↑; E_LAI/E_LAII↑ → Ti⁴⁺ binding energy↑).^55,56 The binding energies of Ti⁴⁺ species for O300 were 0.5 eV lower than those for R300 (e.g., 459.2 eV for O300; 459.7 eV for R300 in the Ti 2p_3/2 regions), suggesting that N_LAI/N_LAII or E_LAI/E_LAII could be higher in R300 than in O300. However, considering that O300 and R300 were similar with regard to Raman spectral parameters of E_g and FWHM used to gauge their N_LAI/N_LAII, it was reasonable to conclude that E_LAI/E_LAII of R300 were higher than those of O300 (N_LAI/N_LAII of O300 ∼ R300; E_LAI/E_LAII of O300 < R300). This proved hypothesis II (E_LAI↑ and E_LAII↑ via H₂ reduction). Moreover, binding energies of Ti⁴⁺ species were invariant across the catalysts subjected to H₂ reduction at 300–600 °C (e.g., 459.7 eV for R300–R600 in the Ti 2p_3/2 regions). This was coupled with the trends in E_g/FWHM (or N_LAI/N_LAII) for R300–R600 to permit N_LAI/N_LAII hierarchy of R600 < R500 < R400 < R300 and E_LAI/E_LAII hierarchy of R300 < R400 < R500 < R600, where the former could be offset by the latter to induce similar binding energies of Ti⁴⁺ species for R300–R600.

Fig. 3 XP spectra of the catalysts in the Ti 2p (A) and O 1s regimes (B), where gray solid lines and black empty circles indicate raw spectra with a resolution of 0.05 eV and those fitted using Gaussian/Lorentzian functions, respectively. In (A), highlighted with green and blue empty circles/arrows are locations and relative concentrations of surface Ti⁴⁺ species belonging to the Ti 2p_1/2 and Ti 2p_3/2 regions, respectively. Moreover, in (B), highlighted with orange, green, and blue empty circles/arrows are locations and relative concentrations of O species of H₂O molecules chemisorbed on oxygen vacancies adjacent to LA_I/LA_II defects (O_CHEMISORBED), those of Brönsted acidic –OH bonds (O_BRÖNSTED), and those of TiO₂ lattices (O_LATTICE), respectively. (C) EPR spectra of O300, R300, and R600 collected at 10 K, where two signals centered at g-factors of 2.00 and 1.92 stem from LA_I/LA_II defects (Ti⁴⁺) trapping free electrons (e⁻) in proximal oxygen vacancies and those (Ti³⁺) filling their empty 3d/4s orbitals with free e⁻, respectively.

To further explore the trends in N_LAI/N_LAII, E_LAI/E_LAII, and N_BA--H⁺/E_BA--H⁺ for the catalysts, their XP spectra in the O 1s regimes were curve-fitted into three sub-bands (Fig. 3B and Table S4).^57,58 These could be indexed to lattice O (O_LATTICE) and O of H₂O chemisorbed on LA_I/LA_II via the formation of pseudo-covalent/coordinative bonds (O_CHEMISORBED) with binding energies centered at 530.4–530.9 eV and 532.4–532.9 eV, respectively, in conjunction with O of BA⁻–H⁺ (O_BRÖNSTED) with binding energies centered at 530.8–531.8 eV.^57–60 Notably, both O_LATTICE and O_CHEMISORBED could donate e⁻ to nearby LA_I/LA_II sites (•Ti⁴⁺/•Ti⁴⁺•) more vigorously at larger N_LAI/N_LAII or at higher E_LAI/E_LAI (N_LAI/N_LAII↑; E_LAI/E_LAII↑ → O_LATTICE/O_CHEMISORBED binding energies↑).²⁵ Again, the trend in N_LAI/N_LAII of R300–R600 (R600 < R500 < R400 < R300) was opposite to that in their E_LAI/E_LAII (R300 < R400 < R500 < R600), potentially resulting in comparable binding energies for O_LATTICE and O_CHEMISORBED species across R300–R600. Moreover, despite N_LAI/N_LAII of R300 being similar to those of O300, E_LAI/E_LAII of the former were higher than those of O300. These potentially rendered the binding energy (e⁻ deficiency) of O_LATTICE or O_CHEMISORBED to be higher in R300 than in O300. Apparently, binding energies of O_LATTICE and O_CHEMISORBED for O300 were 530. 4 eV and 532.4 eV, respectively, both of which were 0.4–0.5 eV lower than those for R300–R600. These could support the combined trends in N_LAI/N_LAII (R600 < R500 < R400 < R300 ∼ O300) and E_LAI/E_LAII (O300 < R300 < R400 < R500 < R600) for the catalysts, with the former being further refined by the trend in N_LAI/N_LAII assessed using O_CHEMISORBED contents (R600 < R500 < R400 ∼ R300 ∼ O300).

To further inspect e⁻ affinity (E_LAI/E_LAII) of the catalysts, electron paramagnetic resonance (EPR) spectroscopy experiments were performed at 10 K on the catalysts being selected: O300 with the lowest E_LAI/E_LAII among the catalysts, R300 with the lowest E_LAI/E_LAII among R300–R600, and R600 with the highest E_LAI/E_LAII among the catalysts (Fig. 3C). The EPR spectra displayed signals centered at g-factors of 2.00, which could stem from e⁻ delocalized on the CB or weakly localized at O_V-proximal •Ti⁴⁺/•Ti⁴⁺• sites (LA_I/LA_II) as well as a minor contribution from shallow intra-bandgap states.^23–26,61 Notably, the EPR spectrum of R600 showed an additional signal centered at a g-factor of 1.92. This could originate from e⁻ migration from intra-bandgap states (vide supra) to the empty 3d/4s sub-shells of •Ti⁴⁺/•Ti⁴⁺• with phenomenal electrophilicities, resulting in their reduction to Ti³⁺/•Ti³⁺.^23–26,61 The emergence of this signal was caused by the strong E_LAI/E_LAII of R600, which could further be stabilized under low-temperature analytic conditions (10 K).^23–26,61 This proved the centrality of T_REDUCTION dialed in to yield defective TiO₂ with high E_LAII or E_BA--H⁺ suitable to decline E_BARRIER for elevation of ˙OH productivity via non-optical H₂O₂ homolysis.

Although the catalyst surfaces were potentially featured by the marginality of N_LAI/N_LAII compared to N_BA--H⁺ (N_LAII < N_LAI ≪ N_BA--H⁺; hypothesis I), LA_I/LA_II profoundly affected the locations and contents of O_BRÖNSTED species (BA⁻ of BA⁻–H⁺) across the surfaces, as proved by their binding energy range (530.8–531.8 eV) and relative contents (6.2–24.7%; Fig. 3B and Table S4). Notably, e⁻ of BA⁻–H⁺ can move toward BA⁻ more conspicuously and make BA⁻ less nucleophilic (E_BA--H⁺↑) particularly at larger N_LAI/N_LAII or at higher E_LAI/E_LAI (N_LAI/N_LAII↑; E_LAI/E_LAII↑ → O_BRÖNSTED binding energy↓). This occurred because neighboring LA_I/LA_II could withdraw e⁻ from BA⁻.^19–22,25 This thereby induced a tentative trend in E_BA--H⁺ values of O300 < R300 ∼ R400 ∼ R500 ∼ R600. However, e⁻ abundance of BA⁻ (E_BA--H⁺) inherent to the surfaces increased in the sequence of O300 < R300 < R400 < R500 ∼ R600, as proved by binding energies of their O_BRÖNSTED species. This discrepancy could partially be attributed to the congregation of TiO₂ poly-crystallites, which was more pronounced at a higher T_REDUCTION and thus dynamically altered the traits of BA⁻–H⁺ sites composing the majority of acidic sites on the surfaces. Nonetheless, hypothesis III (E_BA--H⁺↑ via H₂ reduction) was proven to permit the prediction that BA⁻–H⁺ sites of R300 (or R600) could outperform those of O300 in facilitating ˙OH desorption (RDS) via E_BARRIER reduction. Moreover, O_BRÖNSTED contents of the surfaces also relied highly on aggregation of TiO₂ particulates, which was more pronounced at a higher T_REDUCTION. Accordingly, the O_BRÖNSTED contents gradually increased in the order of R600 ∼ R500 < R400 < R300 < O300. Overall, the XP spectra of the surfaces partially validated the merits of R300 over O300 (hypotheses II–III), as well as those of R600 over R300–R500, in achieving higher E_LAII and/or E_BA--H⁺ proper to prompt the RDS of H₂O₂ homolysis. Nevertheless, in the XP spectral results, the binding energies (e⁻/H⁺ affinity) of all surface species hinged on N_LAI/N_LAII that were highly coupled with E_LAI/E_LAII. In contrast, O_CHEMISORBED contents of the surfaces relied primarily on N_LAI/N_LAII.

Acidic traits of the catalysts: isotherm and TPD

H₂O₂/˙OH/˙OOH/O₂˙⁻ possess kinetic diameters (2.4–2.9 Å) similar to those of CO/CO₂ (3.2–3.3 Å).^62,63 C of CO is coordinate-bonded to LA_II (defective •Ti⁴⁺• with a coordination number of ≤4), whereas C and O of CO₂ are pseudo-hydrogen-bonded to BA⁻ and H⁺ of BA⁻–H⁺, respectively, with O of CO₂ also being coordinate-bonded to LA_II.^64–67 Considering hypothesis I (N_LAII ≪ N_BA--H⁺), CO₂ was expected to bind preferentially with BA⁻–H⁺ rather than with N_LAII. Accordingly, CO and CO₂ served as probes for assessing N_LAII/E_LAII and N_BA--H⁺/E_BA--H⁺ of the catalysts, respectively, as also reported elsewhere.^19,21,22 In this regard, the catalysts were subjected to CO or CO₂ isotherm experiments at −10 °C, 0 °C, and 20 °C (near 25 °C), from which the amounts of CO-accessible LA_II sites or CO₂-accessible BA⁻–H⁺ sites present per-gram of the catalysts (N_{CO, ISO} or N_{CO2, ISO}) as a function of relative pressure (P/P₀) were collected. These isotherms were fitted using the Toth model (Fig. S6, S7, Tables S5, S6, and eqn (S3)).^19,22 The fitted CO or CO₂ isotherms of the catalysts were then implemented to relate the isosteric heats of CO adsorption (E_{CO, ISO}) with CO coverages (N_{CO, ISO}) or the isosteric heats of CO₂ adsorption (E_{CO2, ISO}) with CO₂ coverages (N_{CO2, ISO}) using the Clausius–Clapeyron equation (Fig. S6F/S7F and eqn (S4)).^19,22 Although the E_{CO, ISO} or E_{CO2, ISO} values were evaluated near 25 °C, precise quantification of CO binding energies with LA_II sites (E_LAII) and CO₂ binding energies with BA⁻–H⁺ sites (E_BA--H⁺) require consideration of the lowest CO and CO₂ coverages, respectively (i.e., minimum N_{CO, ISO} and N_{CO2, ISO}).^19,22 This is because the increase in N_{CO, ISO} (or N_{CO2, ISO}) brings about either an elevation or reduction of E_{CO, ISO} (or E_{CO2, ISO}). The former arises from multi-layer CO (or CO₂) adsorption on LA_II (or BA⁻–H⁺), whereas the latter results from lateral CO–CO (or CO₂–CO₂) interactions above LA_II (or BA⁻–H⁺; Fig. S6F/S7F). There effects can severely distort the E_{CO, ISO} and E_{CO2, ISO} values of the catalysts and lead frequently to over-estimation of their N_LAII/N_BA--H⁺.^19,22,68

Meanwhile, the trend in E_{CO, ISO} (∼E_LAII) values of the catalysts was O300 < R300 < R400 < R500 < R600 (Fig. 4C) and matched those on their O_LATTICE and O_CHEMISORBED electrophilicities. In addition, the trend in E_{CO2, ISO} (∼E_BA--H⁺) values of the catalysts was O300 < R300 < R400 ∼ R500 < R600 (Fig. 4D), which was in partial agreement with that in their O_BRÖNSTED nucleophilicities (O300 < R300 < R400 < R500 ∼ R600). However, these correlations should be interpreted with caution, as the lowest CO and CO₂ coverages used to assess E_{CO, ISO} and E_{CO2, ISO} values differed markedly (N_{CO, ISO} of 0.1–6.7 µmol_CO g_CAT⁻¹; N_{CO2, ISO} of 14.8–83.9 µmol_CO2 g_CAT⁻¹; Fig. S6F/S7F).^19,22,68

Fig. 4 The numbers of CO-accessible LA_II defects (A) or CO₂-accessible LA_II defects/BA⁻–H⁺ bonds (B) included per gram of the catalysts (N_{CO, TPD} for A; N_{CO2, TPD} for B) near 25 °C. Averaged CO/CO₂ binding energies of the catalyst surfaces (E_{CO, TPD} for C; E_{CO2, TPD} for D) and their isosteric heats of CO/CO₂ adsorption (E_{CO, ISO} for C; E_{CO2, ISO} for D) near 25 °C. In (A–D), N_{CO, TPD}/N_{CO2, TPD}/E_{CO, TPD}/E_{CO2, TPD} values of the catalysts were assessed via CO/CO₂-TPD techniques with CO/CO₂ molecules being chemisorbed on the catalyst surfaces at 50 °C prior to their desorption, whereas E_{CO, ISO}/E_{CO2, ISO} values of the catalysts were evaluated using their CO/CO₂ isotherms recorded near 25 °C.

To promote the accuracy of E_LAII/E_BA--H⁺ and N_LAII/N_BA--H⁺ quantification, temperature-programmed desorption (TPD) profiles were collected with the use of CO and CO₂ as probe molecules for LA_II and BA⁻–H⁺ sites, respectively.^21,69–71 Typically, the purged catalysts chemisorbed CO (or CO₂) molecules at 50 °C and heated to 350 °C for desorbing CO (or CO₂) molecules with diverse ramping rates (β) of 3–7 °C min⁻¹ (or 2–7 °C min⁻¹).^21,69–71 The areas under the resulting CO-TPD (or CO₂-TPD) profiles (thermal conductivity detector (TCD) signal versus temperature) of the catalysts at distinct β values then served to determine their N_{CO, TPD} (∼N_LAII) (or N_{CO2, TPD} (∼N_BA--H⁺)) values, representing the amount of CO-accessible LA_II (or CO₂-accessible BA⁻–H⁺) included per-gram of the catalyst at 50 °C (near 25 °C; Fig. S8, S9/S11 and S12).^21,69–71 In addition, their CO-TPD (or CO₂-TPD) profiles were curve-fitted into three sub-bands (I–III), all of which had peak temperatures (T_MAX; Fig. S8, S9/S11, S12, Tables S7 and S8).^21,69–71 β and T_MAX for the individual sub-bands were thereafter associated by constructing the plots of ln (β/T_MAX²) versus 1/T_MAX, whose slopes are identical to (−E_{CO, TPD}/R) (or (−E_{CO2, TPD}/R)) values. E_{CO, TPD} (or E_{CO2, TPD}) and R denote the CO (or CO₂) binding energy with LA_II (or BA⁻–H⁺) sites innate to the catalysts at 50 °C (near 25 °C) and the ideal gas constant, respectively, according to TPD theory (E_{CO, TPD}–E_LAII; E_{CO2, TPD}–E_BA--H⁺; eqn (S5)), as detailed in Fig. S10/S13, Tables S7 and S8.^21,69–71

Interestingly, N_{CO, TPD} (∼N_LAII) values of the catalysts were similar (∼50 nmol_CO g_CAT⁻¹; Fig. 4A), which indicated that O_CHEMISORBED contents used to probe N_LAI/N_LAII of the catalysts could depend highly on their N_LAI values with the hierarchy of R600 < R500 < R400 ∼ R300 ∼ O300 (Fig. 3B). Furthermore, E_{CO, TPD} values of the catalysts increased in the sequence of O300 ≤ R300 ≤ R400 ≤ R500 ≤ R600, whose hierarchy was in line with those on their E_{CO, ISO} values and on their O_CHEMISORBED electrophilicities (O300 < R300 < R400 < R500 < R600; Fig. 3B/4C). Here, E_{CO, TPD} and E_{CO, ISO} values of the catalysts reflected their E_LAII values, whereas O_CHEMISORBED electrophilicities of the catalysts captured the combined contributions of their E_LAI and E_LAII values. These rendered it to be compelling that E_LAI values of the catalysts increased in the order of O300 < R300 < R400 < R500 < R600, thereby validating hypothesis II (E_LAI↑ and E_LAII↑ via H₂ reduction).

On the other hand, N_{CO2, TPD} (∼N_BA--H⁺) values of the catalysts increased in the order of R600 < R500 < R400 ∼ R300 ∼ O300, which differed from the trend observed for their N_{CO, TPD} values (∼N_LAII; Fig. 4A and B). This substantiated that CO₂ could bind primarily with BA⁻–H⁺ sites of the catalysts. Importantly, N_{CO2, TPD} (∼N_BA--H⁺) values of the catalysts was ∼10³-fold higher than their N_{CO, TPD} values (∼N_LAII; Fig. 4A and B), which corroborated hypothesis I that BA⁻–H⁺ and LA_I/LA_II constituted the majority and minority of acidic sites innate to the catalysts, respectively (N_LAII < N_LAI ≪ N_BA--H⁺ via O₂ calcination/H₂ reduction). Moreover, E_{CO2, TPD} (∼E_B--H⁺) values of the catalysts were elevated in the sequence of O300 ≤ R300 ≤ R400 ≤ R500 < R600 (Fig. 4D), thereby proving hypothesis III (E_BA--H⁺↑ via H₂ reduction). Notably, E_{CO2, TPD} (∼E_BA--H⁺) values of the individual catalysts were 10–15 kJ mol⁻¹ lower than their corresponding E_{CO, TPD} (∼E_LAII) values except for R600 (∼68.8 kJ mol_CO⁻¹ for E_{CO, TPD}; ∼66.4 kJ mol_CO2⁻¹ for E_CO2-TPD; Fig. 4C and D), thereby demonstrating hypothesis IV (E_BA--H⁺ < E_LAII). This indicated the energy needed to detach the H atom of ˙OH from BA⁻ of BA⁻–H⁺ was higher than that required to withdraw the O atom of ˙OH from LA_II throughout the catalyst surfaces. This substantiated hypotheses V–VI that BA⁻–H⁺ could function as the major activator of H₂O₂ homolysis and thereby outweighed LA_II in dictating the RDS (˙OH desorption from BA⁻–H⁺; Fig. 1D) in conjunction with the demonstration of preferential H₂O₂ adsorption on BA⁻–H⁺ over LA_II.

NH₃ can readily coordinate with BA⁻–H⁺ and LA_II to form BA⁻–NH₄⁺ and LA–NH₃, respectively, on the catalyst surface with their NH₃ binding strengths being lower at higher surface temperatures.^21,37 NH₃ was thus utilized as a probe molecule to further evaluate the acidic features of O300/R300/R600 via the in situ NH₃-diffuse reflectance infrared Fourier transform (DRIFT) spectroscopy technique. The choice of O300/R300/R600 for these experiments originated from their distinct N_BA--H⁺/N_LAII and E_BA--H⁺/E_LAII characteristics.^21,37 Typically, the purged catalysts were exposed sequentially to N₂ and NH₃/N₂ for recording their background signals and their in situ NH₃ spectral signals, respectively, at 50 °C or at 150 °C (Fig. S14).^21,37 The resulting background-corrected, in situ NH₃-DRIFT spectra of the catalysts revealed multiple bands, resulting from asymmetric (ASYM)/symmetric (SYM) stretching (STR)/wagging vibrations of O–H bonds inherent to BA⁻–NH₄⁺ species or N–H bonds innate to BA⁻–NH₄⁺/LA_II–NH₃ species.^21,37 Notably, bands centered at wavenumbers of <3000 cm⁻¹ were unsuitable for quantitative comparison of acidic traits for the catalysts due to significant band overlaps and baseline distortion.^21,37 Conversely, bands centered at wavenumbers of ≥3000 cm⁻¹ were well-resolved, thus allowing for comparison of N_BA--H⁺ and N_LAII values for the catalysts using the areas under bands indexed to BA⁻–H⁺_ASYM/STR and LA_{II ASYM/STR}/LA_{II SYM/STR}, respectively.^21,37 The areas under LA_{II ASYM/STR}/LA_{II SYM/STR} bands were nearly identical across all the catalysts at 50 °C or at 150 °C, supporting the comparable N_{CO, TPD} (N_LAII) values for O300 ∼ R300 ∼ R600. In contrast, the area under the BA⁻–H⁺_ASYM band was higher in the sequence of R600 < R300 ∼ O300 at 50 °C or at 150 °C, which was in accordance with their N_{CO2, TPD} (N_BA--H⁺) hierarchy. Again, the elevation of analytic (surface) temperature from 50 °C to 150 °C led to a decrease in band areas with the extent of band area reduction being more pronounced at a lower E_LAII or a lower E_BA--H⁺.^21,37 Apparently, areas under the LA_{II ASYM/STR}/LA_{II SYM/STR} bands of the catalysts were comparable and revealed minimal dependence of analytic temperatures, further verifying E_{CO, TPD} (E_LAII) values of O300 ∼ R300 ∼ R600. Meanwhile, areas under the BA⁻–H⁺_ASYM/STR bands for O300 and R300 decreased substantially with their surfaces being heated. This was opposed to the trend in the areas under the BA⁻–H⁺_ASYM/STR bands for R600 with surface temperature elevation from 50 °C to 150 °C, substantiating the E_{CO2, TPD} (E_BA--H⁺) hierarchy of O300 ≤ R300 < R600.

Catalytic H₂O₂ homolysis: rate law

H₂O₂ homolysis efficiencies (˙OH productivities) of the catalysts on a per-gram basis can rely on their N_BA--H⁺, E_BA--H⁺, N_LAII, and E_LAII (Fig. 1D).^19–22 Therefore, for rigorous comparison, these should be assessed on a per-major active site (BA⁻–H⁺) basis, considering both the energetic burdens (E_BARRIER) and the frequencies of H₂O₂ collisions with BA⁻–H⁺ and/or LA_II sites in a per-unit time . Both E_BARRIER and of the catalysts were found to hinge on their BA⁻–H⁺ properties, as evidenced by the derivation of a rate law for H₂O₂ homolysis (−r_{H2O2, 0, CO2}; vide infra) alongside with the relationships of E_BA--H⁺ < E_LAII (associated with E_BARRIER) and N_LAII ≪ N_BA--H⁺ (associated with ).

Again, H₂O₂ homolysis proceeds through five elementary steps (Fig. 1D): H₂O₂ adsorption via its linkages (hydrogen bonds) with two BA⁻–H⁺ (BA⁻: S₁) to generate 2S₁·H₂O₂ (eqn (1)), 2S₁·H₂O₂ distortion via its coordination with a LA_II (S₂) to yield 2S₁·H₂O₂·S₂ (eqn (2)), 2S₁·H₂O₂·S₂ dissection via its coordination with an additional S₂ to form two S₁·˙OH·S₂ species (eqn (3)), disintegration of two S₁·˙OH·S₂ species via S₂ detachment to produce two S₁·˙OH species (eqn (4)), and disintegration of two S₁·˙OH species via S₁ detachment to release two ˙OH into the aqueous phase (eqn (5)), as summarized in Table S9.^19–22

Eqn (5) is more endoergic than eqn (4) (E_BA--H⁺ < E_LAII) and thereby could be considered as the RDS of H₂O₂ homolysis, in which K₁–K₄, k₅, C_i, and [j] indicate the quasi-equilibrium constants for eqn (1)–(4) at or near standard states (25 °C and 1 atm), the forward reaction rate constant for eqn (5), the concentration of aqueous species i, and the concentration of surface species j, respectively. Site balances were thereafter formulated for S₁ and S₂ (eqn (6) and Table S9), where the content of S₁ could be substantially higher than that of S₂ ([S₂] ≪ [S₁]), as boosted by their initial quantities of N_LAII ≪ N_BA--H⁺ in 100 mL of aqueous reaction medium ([S₂]₀ ≪ [S₁]₀). Accordingly, the terms [S₁]₀–[S₂]₀ and [S₁]–[S₂] could be simplified to [S₁]₀ and [S₁], respectively, as depicted in eqn (7).

[S₁]₀ − [S₂]₀ = [S₁] − [S₂] + [S₁](K₁C_H2O2[S₁] + K₁^0.5K₂^0.5K₃^0.5K₄^0.5C_H2O2^0.5) (6)

[S₁]₀ = [S₁](1 + K₁C_H2O2[S₁] + K₁^0.5K₂^0.5K₃^0.5K₄^0.5C_H2O2^0.5) ∼ [S₁] (7)

Moreover, the equilibrium constant of H₂O₂ adsorption on S₂ sites of Au/titanium silicate or Fe/silicon carbide (S₂ + H₂O₂ ↔ S₂·H₂O₂) was reported to be ≤∼10¹ L mol⁻¹ under near standard conditions.^72,73 Hence, K₁ could be reasonably estimated as ≤∼10¹ L² mol⁻², considering that E_LAII values (E_{CO, TPD} of 45–70 kJ mol_CO⁻¹; Fig. 4C) of the catalysts were of a similar magnitude to their E_BA--H⁺ counterparts (E_{CO2, TPD} of 35–70 kJ mol_CO2⁻¹; Fig. 4D), albeit with the hierarchies of E_BA--H⁺ < E_LAII for the individual catalysts. In addition, S₂ ≪ S₁ of the catalysts suggested that S₂ sites vicinal to two S₁ sites could be rare, thus making the endothermic quasi-equilibria for H₂O₂ distortion (eqn (2)) and ˙OH desorption from S₂ (eqn (4)) shift to the left, while K₂ and K₄ values could be lower than K₁. A similar constraint applied to the exoergic quasi-equilibrium of H₂O₂ dissection (eqn (3)), whose K₃ could thus be only slightly higher than 10⁰ and below 10¹. Overall, it was logically sound to approximate K₁, K₂, K₃, and K₄ as ≤∼10¹ L² mol⁻², <∼10⁰ L mol⁻¹, ∼10⁰, and <∼10⁰ mol² L⁻², respectively.

Meanwhile, the H₂O₂ concentrations (C_H2O2) evolved upon exposure of the catalysts to reaction (aqueous electric) conditions were evaluated to be ≤∼5.4 × 10⁻³ mol_H2O2 L⁻¹ (Fig. S17; vide infra). Furthermore, [S₁] values of the catalysts should be similar to or smaller than their [S₁]₀ counterparts. Here, [S₁]₀ values corresponded to the moles of BA⁻ (or BA⁻–H⁺) present in 0.2 g of the catalysts soaked in 100 mL of aqueous reaction media (C_{BA--H⁺, 0.2 g} ∼ C_{CO2-TPD, 0.2 g}) and were assessed to be ≤∼3.6 × 10⁻⁴ mol_B--H⁺ L⁻¹, as detailed in Table S9. All of these were gathered to render the terms K₁C_H2O2[S₁] and K₁^0.5K₂^0.5K₃^0.5K₄^0.5C_H2O2^0.5, located on the left-hand side of eqn (7), ≪1, permitting the simplification of eqn (7) to [S₁]₀ ∼ [S₁].

The H₂O₂ dissection rate (−r_H2O2) is defined by the moles of H₂O₂ dissected per unit volume and per unit time (eqn (8)). −r_H2O2 served to represent the initial H₂O₂ dissection rate (−r_{H2O2, 0, CO2}) defined by the moles of H₂O₂ initially dissected per CO₂-accessible BA⁻–H⁺ and per unit time upon the reformulation of eqn (8) using C_{CO2-TPD, 0.2 g} (∼[S₁]₀) along with the substitution of the initial aqueous H₂O₂ concentration (C_{H2O2, 0}) and [S₁]₀ for C_H2O2 and [S₁], respectively (eqn (9) and Table S9).^19–22

−r_H2O2 = k₅K₁K₂K₃K₄C_H2O2[S₁]² (8)

Notably, −r_{H2O2, 0, CO2} values of the catalysts remained formidable to directly evaluate owing to the self-dissection of H₂O₂ to ˙OH/˙OOH/O₂˙⁻ even in the absence of the catalysts under aqueous electric conditions, in addition to the difficulty of precisely tracing ˙OH/˙OOH/O₂˙⁻ contents with short lifetimes of ≤∼10⁷ µs.^{19–22,74,75} Therefore, −r_{H2O2, 0, CO2} values of the catalysts were assessed indirectly using acetaminophen as a model of aqueous wastes. Acetaminophen is particularly suitable because it preferentially undergoes fragmentation via ˙OH-enabled addition or H˙ abstraction compared with the other organics being examined herein (Fig. 11A).^76–78 This was primarily because the energy needed to take e⁻ away from the e⁻-donating group (–OH/–NH/–NH₂) is lower in acetaminophen than in the other organics (Table S19), whereas the other organics were more prone to disintegration via (H⁺-couple) e⁻ transfer on SO₄˙⁻, , , or Fe^IV=O, as we investigated earlier.^{21,37,62,76–79} Moreover, acetaminophen subjected to ˙OH addition or H˙ abstraction readily yields oligomers containing –OH/–NH/–NH₂ groups, all of which can strongly interact with and poison ˙OH generators (BA⁻–H⁺/LA_II).^80,81 Consequently, acetaminophen degradation efficiencies on the catalysts could be deemed a useful probe to compare their activities, resistance to oligomer formation, and reusability in ˙OH production via H₂O₂ homolysis, if ˙OH acted as the dominant degrader of acetaminophen and if diffusional constraints between H₂O₂ and BA⁻–H⁺/LA_II or between ˙OH and acetaminophen were negligible. In this context, the initial acetaminophen degradation rate (−r_{ACETAMINOPHEN, 0, CO2}) defined by the moles of acetaminophen initially decomposed per CO₂-accessible BA⁻–H⁺ and per unit time could be represented using a pseudo-1st-order kinetic model (eqn (10)/S7).^19–22 Moreover, −r_{ACETAMINOPHEN, 0, CO2} could be identical to −r_{H2O2, 0, CO2}, where k_APP and C_{ACETAMINOPHEN, 0} denote the lumped apparent reaction rate constant and the initial aqueous acetaminophen concentration, respectively (Table S9).^19–22

Notably, the RDS of H₂O₂ homolysis (eqn (5)) was expected to obey Arrhenius behavior, from which k₅ could be represented using the Arrhenius equation (eqn (11)). , , and T_REACTION are referred to as the pre-factor of the RDS, the lumped pre-factor, and the reaction temperature, respectively (Table S9).^19–22

Obviously, (eqn (12)) and E_BARRIER (eqn (11)) were found to rely highly on [S₁]₀ (∼C_{CO2, TPD, 0.2 g} ∼ C_{BA--H⁺, 0.2 g}) and E_BA--H⁺, respectively, indicating that BA⁻–H⁺ sites could play a more dominant role than their LA_II counterparts in directing −r_{ACETAMINOPHEN, 0, CO2} (∼−r_{H2O2, 0, CO2}).

In addition, to make the claims on −r_{ACETAMINOPHEN, 0, CO2} ∼ −r_{H2O2, 0, CO2} and on Arrhenius relationships (eqn (10)–(12)) valid, E_BARRIER values associated with all ˙OH-mediated acetaminophen degradation routes had to be insignificant. This has previously been demonstrated by a spectrum of computational studies on ˙OH-induced decomposition of aqueous organics.^82,83 Furthermore, the equivalence of −r_{ACETAMINOPHEN, 0, CO2} ∼ −r_{H2O2, 0, CO2} could be compelling if the catalyst scarcely leached one of H₂O₂ evolvers (LA_II) to the aqueous phase under electric conditions, which could be verified by filtration runs (vide infra). Again, −r_{ACETAMINOPHEN, 0, CO2} ∼ −r_{H2O2, 0, CO2} could also be contingent upon negligible mass-transfer limitations for both H₂O₂ transport to BA⁻–H⁺/LA_II sites and ˙OH transport to acetaminophen, as demonstrated by diffusion runs (vide infra). Finally, ˙OOH/O₂˙⁻ should be marginal relative to ˙OH in acetaminophen disintegration, as substantiated by scavenging runs (vide infra).

Catalytic H₂O₂ homolysis: kinetic assessment

For kinetic inspection of H₂O₂ homolysis upon validation of −r_{H2O2, 0, CO2} ∼ −r_{ACETAMINOPHEN, 0, CO2}, O300, R300, and R600 were selected owing to their distinct Brönsted acidities including O300 ≤ R300 < R600 for E_BA--H⁺ and R600 < R300 ∼ O300 for N_BA--H⁺. These trends suggested that R600 should unveil the lowest E_BARRIER (−r_{H2O2, 0, CO2} (or −r_{ACETAMINOPHEN, 0, CO2})↑; eqn (11)) and the lowest (−r_{H2O2, 0, CO2} (or −r_{ACETAMINOPHEN, 0, CO2})↓; eqn (12)), where these opposing effects were anticipated to compete in determining −r_H2O2 (or −r_{ACETAMINOPHEN, 0, CO2}) for R600.^19–22

Acetaminophen was degraded under aqueous electric conditions. A bare graphite plate (anode) and a catalyst (0.2 g)-coated graphite plate (cathode) were immersed in 100 mL of an aqueous acetaminophen solution (1.0 × 10⁻³ mol L⁻¹) containing Na₂SO₄ as the supporting electrolyte with an electric input being applied.^19–21,79 At the anode, H₂O can be oxidized via H₂O → 2H⁺ + 1/2O₂ + 2e⁻. The generated H⁺, O₂, and e⁻ then migrate to the cathode, where H₂O₂ is continuously generated via 2H⁺ + O₂ + 2e⁻ → H₂O₂ as long as an electric input is maintained across the electrodes.^19–21,79 An electric potential of 3 V was selected to maximize C_H2O2 values at or near the catalyst surfaces, while minimizing the release of the catalysts from the cathode to the aqueous phase.^19–21,79 Moreover, H₂O oxidation at the anode can also proceed via H₂O → H⁺ + ˙OH + e⁻, where ˙OH can react with the anode surface to form surface O highly prone to degrade acetaminophen (anodic oxidation).^19–21,79 Moreover, H₂O₂ produced at or near the cathode tends to undergo self-cleavage to generate ˙OH, ˙OOH, or O₂˙⁻, all of which can also participate in acetaminophen degradation (H₂O₂ self-dissection).^19–21,79 Anodic oxidation and H₂O₂ self-dissection were regarded as backgrounds and thereby served to correct kinetic datasets of acetaminophen conversion (X_{ACETAMINOPHEN}; eqn (S6)) as a function of time. X_{ACETAMINOPHEN} values of <25% were selected to shift the equilibrium of acetaminophen degradation (or H₂O₂ homolysis) to the right at T_REACTION values of 25–55 °C.^19–21,79 The kinetic datasets subjected to background correction were then fitted using a pseudo-1^st-order kinetic model to evaluate k_APP values (Fig. S16/S18/S20, S21/S24/S27 and Table S10), with which background-corrected −r_{ACETAMINOPHEN, 0, CO2} and initial acetaminophen degradation rate defined by the moles of acetaminophen initially decomposed per gram and per unit time (−r_{ACETAMINOPHEN, 0}; eqn (S8)) were assessed.^19–21,79 Importantly, −r_{ACETAMINOPHEN, 0, CO2} and −r_{ACETAMINOPHEN, 0} of the catalysts served to compare their activities and recyclabilities in decomposing acetaminophen, respectively, whereas −r_{POLLUTANT, 0, CO2} values were also used to compare the adaptabilities of the catalysts in disintegration of other aqueous organic pollutants.^19–21,79

Notably, acetaminophen degradation should proceed under LA_II leaching-free aqueous electric conditions, under which the presumption concerning the rarity of S₂ (LA_II) sites vicinal to two S₁ (BA⁻) sites used in deriving −r_{H2O2, 0, CO2} (eqn (9)) and −r_{ACETAMINOPHEN, 0, CO2} (eqn (11)) could remain valid. To verify this, acetaminophen degradation runs were performed at 3 V and 25 °C up to 360 minutes according to the methods depicted earlier and in the caption of Fig. S15.^19–21,79 Exceptions were the replacement of the catalyst-coated cathodes with bare graphite electrodes at 60 minutes, where the aqueous reaction solutions were filtered to collect the corresponding reaction filtrates (Fig. S15B–D).^19–21,79 Acetaminophen degradation runs were then resumed using two bare graphite plates soaked in the aqueous reaction filtrates for monitoring X_{ACETAMINOPHEN} values at 60–360 minutes (filtration runs). X_{ACETAMINOPHEN} values monitored in the filtration runs could originate from anodic oxidation, ˙OH/˙OOH/O₂˙⁻ generated via H₂O₂ self-dissection, and those produced via homogeneous H₂O₂ homolysis on leached LA_II and aqueous OH⁻ (BA⁻–H⁺).^19–21,79 For comparison, an acetaminophen degradation control run was conducted using two bare graphite electrodes at 3 V and 25 °C up to 360 minutes, where X_{ACETAMINOPHEN} values could arise only from anodic oxidation and ˙OH/˙OOH/O₂˙⁻ generated via H₂O₂ self-dissection (blank run; Fig. S15A).^19–21,79 The extent of acetaminophen degradation between 60 and 360 minutes (ΔC_{ACETAMINOPHEN} or the change in X_{ACETAMINOPHEN} values) was similar for the filtration and blank runs (∼0.4 × 10⁻³ mol_{ACETAMINOPHEN} mol⁻¹). This indicated that H₂O₂ homolysis contributed negligibly to ˙OH/˙OOH/O₂˙⁻ production, confirming that LA_II quantities leached from the catalyst surfaces were minimal under the aqueous electric conditions.

Acetaminophen decomposition should proceed under reaction-controlled conditions without significant internal or external mass-transfer limitations. In particular, diffusion of H₂O₂ to BA⁻–H⁺/LA_II sites and that of ˙OH to acetaminophen should not dominate the observed kinetics under aqueous electric conditions. In this context, acetaminophen degradation runs were performed at 3 V and 25 °C for 6 minutes, while varying the catalyst particulate size (<90 µm, 90–200 µm, and <200 µm) or stirring speed (300 rpm and 400 rpm) with the former and the latter being utilized to inspect the repercussions of internal and external diffusions on −r_{ACETAMINOPHEN, 0, CO2} values of the catalysts, respectively (diffusion runs; Fig. S16 and Table S10).^19–21,79 The resulting −r_{ACETAMINOPHEN, 0, CO2} values of the catalysts were essentially independent of both particulate size or stirring speed (Fig. 5A). This indicated that −r_{ACETAMINOPHEN, 0, CO2} values were evaluated in reaction-limited domains and were barely affected by mass-transfer limitations of H₂O₂ transport to BA⁻–H⁺/LA_II sites or ˙OH transport to acetaminophen. Accordingly, acetaminophen (or aqueous contaminant) degradation runs were conducted using 0.2 g of the catalyst with sizes of <200 µm at an electric input of 3 V, a T_REACTION of 25 °C, and a stirring speed of 300 rpm, unless otherwise specified. Inductively coupled plasma-atomic emission spectroscopy (ICP-AES) served to quantify Ti leaching during the diffusion runs. All the catalysts released minute Ti contents (<10⁻⁴ mol%) to aqueous phases, which was in line with the results on the filtration runs (Fig. S15). These could be gathered with the results on the diffusion runs (Fig. 5A) to support the validity of −r_{H2O2, 0, CO2} ∼ −r_{ACETAMINOPHEN, 0, CO2} for the catalysts. Importantly, −r_{ACETAMINOPHEN, 0, CO2} values of O300 and R300 were comparable (∼0.02 min⁻¹), yet were only around one-fifth of that for R600. This suggested that E_BARRIER could outstrip in determining −r_{ACETAMINOPHEN, 0, CO2} of R600, highlighting the merit of H₂ reduction-subjected, defective TiO₂ (R600) for ˙OH generation via non-optical H₂O₂ activation.

Fig. 5 Background-corrected, initial acetaminophen degradation rates normalized on a per-CO₂-accessible site basis (−r_{ACETAMINOPHEN, 0, CO2}) for the catalysts assessed upon the variation of catalyst particulate sizes/stirring speeds (A) or in the absence/presence of a scavenger used to quench ˙OH/˙OOH/O₂˙⁻ (B). Reaction environments: 0.2 g of the catalyst with sizes of <90 µm (A), 90–200 µm (A), or <200 µm (A and B); 100 mL of de-ionized H₂O; 0.1 mmol of acetaminophen; 0 mmol (A and B) or 0.82 mmol (B) of a scavenger; 0.14 mmol of Na₂SO₄; 3 V; 25 °C; 300 rpm (A and B) or 400 rpm (A); initial pH of 7.0; final pH of 4.5 (±0.3).

Notably, acetaminophen degradation should primarily be driven by ˙OH rather than ˙OOH/O₂˙⁻ under aqueous electric conditions. At circum-neutral pH values (4.5–7.0), ˙OOH readily deprotonates to form O₂˙⁻, thus making it challenging to conclude which reactive oxygen species (ROS; ˙OH, ˙OOH, or O₂˙⁻) could dominate acetaminophen degradation.^21,34 H₂O₂ evolution runs were thus carried out identically to the diffusion runs except for the exclusion of acetaminophen from aqueous reaction solutions with C_H2O2 values being monitored for up to 30 minutes (Fig. S17).^19–21,79 The C_H2O2 values rapidly increased and plateaued within 5–30 minutes with averaged C_H2O2 (C_{H2O2, AVG}) values in the range of 2.6 × 10⁻³–5.4 × 10⁻³ mol_H2O2 L⁻¹. The observed trend in C_{H2O2, AVG} values did not correlate with the acidic traits of the catalysts, yet fell outside the scope of the current study and therefore was not further elaborated.

Instead, the C_H2O2 values monitored at ≥5 minutes were averaged and doubled to determine the scavenger content added to the aqueous reaction solution (8.2 × 10⁻³ mol_SCAVENGER L⁻¹). The quenching rate of ˙OH, ˙OOH, or O₂˙⁻ can depend highly on the type of the scavenger utilized, as corroborated by its 2^nd-order rate constants for ROS removal (k_ROS↑ → −r_{ACETAMINOPHEN, 0, CO2}↓; Table S11).^84–86 To this end, acetaminophen degradation runs were performed identically to the diffusion runs except for the inclusion of a scavenger (catechol or hydroquinone) in aqueous reaction solutions (scavenging runs; Fig. S18 and Table S10). The pH range was comparable to that of the diffusion runs (4.5–7.0),^19–21,79 suggesting that ˙OH/˙OOH/O₂˙⁻ generation was of likelihood across the scavenging runs. The −r_{ACETAMINOPHEN, 0, CO2} values of the individual catalysts increased in the sequence of ‘with catechol’ ∼ ‘with hydroquinone’ < ‘without scavenger’ (Fig. 5B). This trend was inconsistent with those anticipated for −r_{ACETAMINOPHEN, 0, CO2} values under the speculation that either ˙OOH or O₂˙⁻ could function as the main degrader of acetaminophen (‘with hydroquinone’ < ‘with catechol’ < ‘without scavenger’). The hierarchy of −r_{ACETAMINOPHEN, 0, CO2} values for the individual catalysts, however, was in agreement with that expected with the assumption that ˙OH could act as the major decomposer of acetaminophen, thereby boosting the claim that ˙OH outcompeted ˙OOH/O₂˙⁻ in acetaminophen disintegration in addition to demonstration of −r_{H2O2, 0, CO2} ∼ −r_{ACETAMINOPHEN, 0, CO2}.

Catalytic H₂O₂ homolysis: EPR spectroscopy and kinetic parameter

˙OH bound to LA_II/BA⁻–H⁺ sites can react with aqueous H₂O₂ to form ˙OOH species bound to these sites (poison I). At pH values of ≥4.8, some of these ˙OOH species undergo deprotonation to yield O₂˙⁻ species bound to LA_II sites (poison II; Fig. 8A).^21,34 The ˙OOH/O₂˙⁻ species can poison H₂O₂ activators (LA_II/BA⁻–H⁺), unless they are released into the aqueous media via desorption from LA_II/BA⁻–H⁺ sites.^21,34 Aqueous ˙OH productivities increased in the sequence of O300 ∼ R300 < R600, as evidenced by the hierarchy of their −r_{ACETAMINOPHEN, 0, CO2} (∼−r_{H2O2, 0, CO2}) values assessed from the diffusion runs. In contrast, aqueous ˙OOH productivities were expected to be lower for R600 than for R300 (or O300). This was ascribed to higher E_LAII and E_BA--H⁺ values of R600, whose surface could desorb ˙OH species from LA_II/BA⁻–H⁺ sites more promptly than the others. As a result, ˙OH species bound to the R600 surface has a lower probability of reacting with H₂O₂ to yield ˙OOH species (poison I). Similarly, aqueous O₂˙⁻ productivities were also predicted to be lower for R600 than for R300 (or O300), considering their hypothetic coverages of ˙OOH species functioning as O₂˙⁻ precursors (R600 < R300 ∼ O300) coupled with the comparable pH range (4.5–7.0) across all catalysts immersed under aqueous electric conditions (i.e., similar tendencies of ˙OOH → O₂˙⁻).

Overall, aqueous ˙OH and ˙OOH/O₂˙⁻ productivities were expected to increase in the sequence of O300 ∼ R300 < R600 and R600 < R300 ∼ O300, respectively (vide supra). These hypotheses were investigated via EPR spectroscopy experiments.^19–22 The catalysts were mixed with aqueous H₂O₂ to form aqueous reaction mixtures, where the quantity of H₂O₂ was regulated to be in excess relative to that of BA⁻–H⁺ sites on the catalyst, given that C_{BA--H⁺, 0.2 g} ≪ C_H2O2 in the diffusion runs (Fig. S17).^19–22 Notably, 5,5-dimethyl-1-pyrroline N-oxide (DMPO) can be adducted to aqueous ˙OH and aqueous ˙OOH/O₂˙⁻ for trapping their spins via production of aqueous DMPO–OH and aqueous DMPO–OOH, respectively (* and **; Fig. 6A).^19–22 Moreover, a fraction of aqueous DMPO–OH adducts can transform into aqueous HDMPO analogues via adsorption on LA_II sites (defective Ti⁴⁺; ****; Fig. 6B) followed by deprotonation (***; Fig. 6B) and desorption from •LA_II sites (defective •Ti⁴⁺; *****; Fig. 6B).^19–22 Notably, the adsorption stage marked with **** could be decelerated on the catalyst surface with a higher E_LAII, whereas the deprotonation stage marked with *** could proceed similarly across all the catalyst surfaces due to similar pH range of the corresponding aqueous reaction solutions subjected to EPR spectroscopy experiments (vide infra). All of these were coupled with the trend in N_LAII values of O300 ∼ R300 ∼ R600 to induce the speculation that aqueous HDMPO productivities could increase in the order of R600 < O300 ∼ R300. Furthermore, aqueous HDMPO can also be adducted to aqueous ˙OH to yield aqueous HDMPO–OH, which made it convincing that aqueous HDMPO–OH contents should be included in the quantification of aqueous ˙OH productivities on the catalyst surfaces.^19–22

Fig. 6 (A) Depiction of formation routes for 5,5-dimethyl-1-pyrroline N-oxide (DMPO, spin trapper) species adducted to aqueous ˙OH (aqueous DMPO–OH) or aqueous ˙OOH (*)/O₂˙⁻ (**; aqueous DMPO–OOH), for which ˙OH bound to BA⁻–H⁺/LA_II innate to the catalyst surface should interact with aqueous H₂O₂ to generate ˙OOH prior to its deprotonation and transition into O₂˙⁻ on the surface. (B) Illustration of the formation route for an aqueous HDMPO–OH adduct by way of aqueous DMPO–OH adsorption on a LA_II defect (Ti⁴⁺; ****), deprotonation (***), and HDMPO desorption from a •LA_II defect (•Ti⁴⁺; *****) prior to ˙OH addition to aqueous HDMPO. In (A and B), ˙OOH formation or aqueous DMPO–OH adsorption on the surface (****) hinges on its E_LAII value, whereas deprotonation of ˙OOH or Ti⁴⁺-coordinated DMPO–OH on the surface highly relies on the pH value of the aqueous phase. EPR spectra of the reaction solutions collected via syringe filtration of the corresponding reaction mixtures containing H₂O, H₂O₂, DMPO, and the catalyst (O300 for (C); R300 for (D); R600 for (E)). In (C–E), gray solid lines and purple empty circles correspond to raw spectra and those simulated using Gaussian/Lorentzian functions, whereas orange solid circles, green solid squares, and blue solid diamonds refer to aqueous DMPO–OH, aqueous DMPO–OOH, and aqueous HDMPO–OH adducts, respectively. Analytic conditions: 0.1 g of the catalyst with sizes of <200 µm; 3.0 mmol of H₂O₂; 0.2 mL of de-ionized H₂O; 0.7 mmol of DMPO; 25 °C; stirred using a vortex for 2 minutes; 0.45 µm PTFE syringe filtration; final pH of 4.2 (±0.3).

To this end, the aqueous reaction mixtures stated above were then mixed with DMPO, stirred vigorously, and filtered to isolate aqueous reaction solutions with pH values of 4.2, under which a fraction of ˙OOH poisons could transform into their O₂˙⁻ counterparts on the catalyst surfaces exposed to aqueous H₂O₂ conditions.^19–22 The aqueous reaction solutions were subjected to EPR spectroscopy experiments at 25 °C. The resulting EPR spectra were curve-fitted according to the methods outlined in Table S12 for identifying DMPO adducts and quantifying their concentrations (Fig. 6C–E and Table S13).^19–22 The EPR spectra disclosed three characteristic signal patterns: a quartet-signal set with an intensity ratio of 1 : 2 : 2 : 1 assigned to aqueous DMPO–OH, a sextet-signal set with an intensity ratio of 1 : 1 : 1 : 1 : 1 : 1 indexed to aqueous DMPO–OOH, and a trio-signal set with an intensity ratio of 1 : 1 : 1 assigned to aqueous HDMPO–OH.^{19,22,87–90} The aqueous HDMPO–OH contents increased in the order of R600 (5.6%) < R300 ∼ O300 (16.7–20.8%), which was in accordance with the trend anticipated (R600 < R300 ∼ O300). Moreover, the combined contents of the aqueous DMPO–OH and aqueous HDMPO–OH increased in the sequence of O300 (66.7%) < R300 ∼ R600 (83.3–88.9%), which was in partial line with aqueous ˙OH productivities inferred from −r_{ACETAMINOPHEN, 0, CO2} values (O300 ∼ R300 < R600) and supported the conclusion that defective TiO₂ subjected to H₂ reduction at 600 °C could maximize acceleration of H₂O₂ homolysis under aqueous conditions. Conversely, the aqueous DMPO–OOH contents increased in the order of R600 ∼ R300 (11.1–16.7%) < O300 (33.3%), which partially matched our prediction depicted earlier (R600 < R300 ∼ O300). This suggested that H₂ reduction-subjected, defective TiO₂ could reveal greater tolerance to ˙OOH/O₂˙⁻ poisons than H₂ reduction-unsubjected TiO₂ (O300), thereby tentatively sustaining H₂O₂ homolysis more effectively under aqueous conditions. Importantly, aqueous ˙OH-induced DMPO adducts of aqueous DMPO–OH and HDMPO–OH accounted for the majority of aqueous DMPO adducts monitored earlier (66.7–88.9%). This could further substantiate that aqueous ˙OH was the primary reactive species responsible for acetaminophen degradation.

Notably, aqueous DMPO–OOH might be decomposed to form aqueous DMPO–OH, which should be considered for the rigor of our quantitative elaboration of all aqueous DMPO adducts.^91,92 To this end, ethanol (CH₃CH₂OH) served as a ˙OH quencher due to its high k_˙OH value of 1.9 × 10⁹ L mol⁻¹ s⁻¹. Aqueous ˙OH can readily abstract H˙ from CH₃CH₂OH to generate aqueous ˙CHCH₃OH and H₂O. The resulting aqueous ˙CHCH₃OH can then be adducted to DMPO to yield aqueous DMPO–CHCH₃OH.⁹³ In this context, the aforementioned DMPO-containing aqueous reaction mixtures were further mixed with excess CH₃CH₂OH, stirred vigorously, and filtered to isolate the corresponding aqueous reaction solutions prior to their EPR spectroscopy experiments. The resulting EPR spectra were curve-fitted into bunched signal sets indexed to aqueous DMPO–OH, aqueous DMPO–OOH, and aqueous HDMPO–OH adducts alongside with additional sextet-signal sets with an intensity ratio of 1 : 1 : 1 : 1 : 1 : 1 assigned to the aqueous DMPO–CHCH₃OH adducts (Fig. S19, Tables S14 and S15).⁹⁴ As noted above, aqueous ˙OH can react with both DMPO and HDMPO to produce aqueous DMPO–OH and HDMPO–OH (Fig. 6A and B), respectively, whose contents should thus be combined to compare aqueous ˙OH productivities of the catalysts. If aqueous DMPO–OH arose primarily from aqueous ˙OH yielded via catalytic H₂O₂ homolysis, the sum of aqueous DMPO–OH and HDMPO–OH contents should decrease substantially upon the addition of CH₃CH₂OH (a potent ˙OH scavenger) in the reaction mixture subjected to filtration. This claim held true across all quantitative results for the CH₃CH₂OH-bearing EPR spectra of the simulated reaction solutions (Tables S13 and S15; e.g., 88.9% without CH₃CH₂OH; 40.7% with CH₃CH₂OH for R600). Moreover, if aqueous DMPO–OH originated predominantly from aqueous ˙OH produced via H₂O₂ homolysis on the catalyst surface, the sum of aqueous DMPO–OH and HDMPO–OH contents with CH₃CH₂OH being absent in the reaction solution should be comparable to that of aqueous DMPO–OH, HDMPO–OH, and DMPO–CHCH₃OH contents with CH₃CH₂OH being present in the reaction solution. This relationship was consistently observed across all quantitative results for the CH₃CH₂OH-bearing EPR spectra of the simulated reaction solutions (Tables S13 and S15; e.g., 88.9% without CH₃CH₂OH; 87.3% with CH₃CH₂OH for R600). These results substantiated that aqueous DMPO–OH primarily stemmed from aqueous ˙OH generated via catalytic H₂O₂ homolysis rather than DMPO–OOH decomposition. This was in line with previous literature arguing for minimal conversion of aqueous DMPO–OOH to aqueous DMPO–OH (≤3%) under circum-neutral aqueous pH conditions.^91,92 Overall, all the results from the filtration/diffusion/scavenging runs and the EPR spectra of the aqueous reaction solutions validated −r_{H2O2, 0, CO2} ∼ −r_{ACETAMINOPHEN, 0, CO2} as well as the rate law dictated mainly by BA⁻–H⁺ features of the catalysts (eqn (10)–(12)).

To ensure the obedience of the catalysts to Arrhenius behavior, the catalysts were subjected to acetaminophen degradation runs identical to the diffusion runs except for the alteration of T_REACTION values from 25 to 55 °C.^19–22 The resulting −r_{ACETAMINOPHEN, 0, CO2} values of the catalysts were higher at greater T_REACTION values (∼0.02 min⁻¹ → ∼0.07 min⁻¹ for O300; ∼0.02 min⁻¹ → ∼0.06 min⁻¹ for R300; ∼0.09 min⁻¹ → ∼0.13 min⁻¹ for R600 at T_REACTION of 25 °C → 55 °C), which proved their conformity to Arrhenius law (Fig. S20 and Table S10).^19–22 Arrhenius plots of logarithmic −r_{ACETAMINOPHEN, 0, CO2} (ln (−r_{ACETAMINOPHEN, 0, CO2})) versus reciprocal of T_REACTION (1/T_REACTION) were constructed for the catalysts (eqn (13)/S9). The resulting slopes and y-intercepts served to assess E_BARRIER and values required to activate H₂O₂ homolysis on the catalyst surfaces. Again, E_BARRIER and values were governed by E_BA--H⁺ (O300 ≤ R300 < R600; E_BA--H⁺↑ → E_BARRIER↓) and N_BA--H⁺ values (R600 < R300 ∼ O300; ), respectively, for the catalysts (eqn (12)).^19–22

The catalysts exhibited E_BARRIER values in the order of R600 (10.8 kJ mol⁻¹) < R300 ∼ O300 (33.3–34.5 kJ mol⁻¹) and values in the order of R600 (5.9 × 10⁻¹ min⁻¹) < R300 ∼ O300 (3.4 × 10³–6.1 × 10³ min⁻¹; Fig. 7A). This proved that E_BARRIER could override in determining −r_{H2O2, 0, CO2} values, enabling R600 to display the highest activity among the catalysts (vide supra). This finding highlighted the centrality of H₂ reduction conditions in tuning the acidic strength (E_BA--H⁺) of the major H₂O₂ activator (BA⁻–H⁺) suitable for desorbing ˙OH with lower energy input under optical energy-free, aqueous conditions.

Fig. 7 (A) Arrhenius plots of logarithmic −r_{ACETAMINOPHEN, 0, CO2} values (ln (−r_{ACETAMINOPHEN, 0, CO2})) versus the reciprocal of reaction temperatures (1/T_REACTION) for the catalysts, in which slopes and y-intercepts served to evaluate energy barriers (E_BARRIER) and lumped pre-factors needed to activate ˙OH-mediated acetaminophen degradation. (B) Background-corrected initial acetaminophen degradation rates normalized on a per-gram basis (−r_{ACETAMINOPHEN, 0}) for the (used) catalysts subjected to acetaminophen degradation recycle runs. Reaction environments: 0.2 g of the (used) catalyst with sizes of <200 µm; 100 mL of de-ionized H₂O; 0.1 mmol of acetaminophen; 0.14 mmol of Na₂SO₄; 3 V; 25 °C (A and B) or 35–55 °C (A); 300 rpm; initial pH of 7.0; final pH of 4.5 (±0.3).

H₂O₂ homolysis on the catalysts: reusability, poison tolerance, and regenerability

To inspect recyclabilities of the catalysts in acetaminophen degradation, recycle runs were performed identically to the diffusion runs except for exposure of the (used) catalysts to acetaminophen disintegration conditions multiple times. After each reaction run, the used catalysts were rinsed with de-ionized H₂O and dried overnight at 110 °C prior to the commencement of the subsequent runs (Fig. S21 and Table S10).^19–22 As anticipated, −r_{ACETAMINOPHEN, 0} values of the catalysts were gradually reduced over successive cycles (Fig. 7B). This could be ascribed to oligomeric by-products yielded via aqueous ˙OH-enabled addition to or H˙ abstraction from acetaminophen. These oligomers rich in e⁻-donating groups can interact with H₂O₂ activators of BA⁻–H⁺ and LA_II sites via pseudo-hydrogen bonding and coordination, respectively, thereby hindering their access to aqueous H₂O₂.^21,37–39

For this investigation, the reaction solutions subjected to the 1st recycle runs were examined using ultra-performance liquid chromatography equipped with mass spectrometry (UPLC-MS) in positive mode. The UPLC-MS spectra of the reaction solutions were then used to identify intermediates based on their mass (m)/charge (z) values (Fig. S22).^{19–22,95–97} Aside from revealing acetaminophen (m/z of 152), the UPLC-MS spectra showed several intermediates formed via ˙OH-mediated addition/H˙ abstraction or H⁺/e⁻-enabled addition. The latter could proceed with the use of bountiful H⁺ and e⁻ produced at the anode via two H₂O oxidation pathways (H₂O → 2H⁺ + 1/2O₂ + 2e⁻ and H₂O → H⁺ + ˙OH + e⁻).^19–22 Most of the intermediates monitored were oligomeric species except for those with m/z values of 91, 100, and 110. The oligomers possessed functional groups of –OH/–NH–/–NH₂ with m/z values of 227, 259, 301, 307, 323, 408, and 450.^19–22

To further prove catalyst deactivation, Ti and C/N contents of the used catalysts collected post the 1^st, 3^rd, and 5^th recycle runs were evaluated using ICP-AES and elemental analysis (EA), respectively, with which bulk atomic ratios of C to Ti (C/Ti) and N to Ti (N/Ti) could be compared for the used catalysts (Fig. 8C, D and Table S2). The C/Ti values were similar across the used catalysts at each of the recycle runs, yet increased steadily throughout the recycle runs (∼2.6 × 10⁻¹ mol_C mol_Ti⁻¹ → ∼3.5 × 10⁻¹ mol_C mol_Ti⁻¹). This trend indicated progressive accumulation of oligomers on BA⁻–H⁺/LA_II sites of the used catalysts and could justify the gradual decline in their −r_{ACETAMINOPHEN, 0} values throughout the recycle runs. Moreover, the N/Ti values of used O300 continuously increased throughout the recycle runs (∼10.0 × 10⁻³ mol_N mol_Ti⁻¹ → ∼17.8 × 10⁻³ mol_N mol_Ti⁻¹). This was partially the case with the N/Ti values of used R300, whose magnitude was ∼8.5 × 10⁻³ mol_N mol_Ti⁻¹ prior to the 4^th run and reached ∼13.7 × 10⁻³ mol_N mol_Ti⁻¹ post the 5^th run. In contrast, the N/Ti values of used R600 were negligible prior to the 4^th run and only ∼3.8 × 10⁻³ mol_N mol_Ti⁻¹ post the 5^th run. The catalysts showed the trend in E_BA--H⁺ or E_LAII values of O300 ≤ R300 < R600, suggesting that the extent of oligomer deposition (N_OLIGOMER) could increase in the identical order.^21,37–39 However, this was opposite to the trend in N/Ti (N_OLIGOMER) values for the used catalysts (used R600 ≪ used R300 < used O300), demonstrating that oligomer accumulation on BA⁻–H⁺/LA_II sites of the used catalysts was not governed solely by their E_BA--H⁺/E_LAII values. Instead, considering the N_LAII/N_BA--H⁺ values of the catalysts (O300 ∼ R300 ∼ R600 for N_LAII; R600 < R300 ∼ O300 for N_BA--H⁺), (used) R600 with the smallest N_BA--H⁺ could detour oligomer deposition more effectively than (used) O300 or (used) R300. Consequently, (used) R600 revealed the highest −r_{ACETAMINOPHEN, 0} values among the (used) catalysts over successive recycle runs (Fig. 7B).

Fig. 8 (A) Depiction of a BA⁻–H⁺-modified Ti⁴⁺–O–Ti⁴⁺–O–Ti⁴⁺–O–Ti⁴⁺ channel exposed to calcination or reduction environments, under which 1/2O₂ species (*/** or */***) are released to afford LA_II (highlighted with gray arrows) or two LA_I defects (highlighted with sky blue arrows). In (A), BA⁻–H⁺ and LA_II can serve to activate H₂O₂ homolysis up to the H₂O₂ cleavage step, upon which ˙OH bound to BA⁻–H⁺/LA_II can potentially undergo transition to surface ˙OOH (poison I) and O₂˙⁻ (poison II) via interaction with aqueous H₂O₂ and deprotonation, respectively (highlighted with gray arrows), in lieu of being released into an aqueous phase via desorption. Moreover, in (A), two LA_I defects can tentatively interact with aqueous H₂O₂, liberate two H˙ species, and covalently bind O₂ (poison III), as highlighted with sky blue arrows. (B) XP spectra of the homolytic H₂O₂ dissection-subjected catalysts in the O 1s regimes, where gray solid lines and black empty circles indicate raw spectra and those fitted using Gaussian/Lorentzian functions with a resolution of 0.05 eV. In (B), purple, orange, green, and blue empty circles indicate ˙OOH/O₂˙⁻/O₂ (poison I/II/III; O_PEROXY), O_CHEMISORBED, O_BRÖNSTED, and O_LATTICE, respectively, among which locations and relative concentrations of O_PEROXY and O_BRÖNSTED are highlighted with purple and green arrows, respectively. Atomic ratios of C to Ti (C/Ti; C) or N to Ti (N/Ti; D) included in the (used) catalysts subjected to acetaminophen degradation recycle runs. Reaction environments: 0.2 g of the (used) catalyst with sizes of <200 µm; 100 mL of de-ionized H₂O; 0.1 mmol of acetaminophen; 0.14 mmol of Na₂SO₄; 3 V; 25 °C; 300 rpm; initial pH of 7.0; final pH of 4.5 (±0.3).

Again, ˙OOH bound to BA⁻–H⁺/LA_II sites (poison I) and O₂˙⁻ bound to LA_II sites (poison II) can severely limit H₂O₂ access to these sites (Fig. 8A).^21,34 Moreover, H₂O₂ dehydrogenation can generate O₂ covalently bonded to two vicinal LA_I sites (poison III). These LA_I sites could be located near H₂O₂ activators (BA⁻–H⁺/LA_II sites) and therefore impede H₂O₂ distortion and dissection, as conjectured based on N_LAII < N_LAI ≪ N_BA--H⁺ for the catalysts (Fig. 8A).^34–36 Notably, the formation of poison II requires cleavage of the pseudo–hydrogen bond between BA⁻ of BA⁻–H⁺ and the H atom of ˙OOH (poison I). However, this pathway was expected to be limited across the catalyst surfaces because of E_BA--H⁺ < E_LAII (i.e., cleavage of the covalent bond between LA_II and the O atom of ˙OOH in advance of cleavage of the pseudo–hydrogen bond between BA⁻ of BA⁻–H⁺ and the H atom of ˙OOH).^21,34 Moreover, poison I can bind more strongly with BA⁻–H⁺ or LA_II at a lower E_BA--H⁺ or a lower E_LAII, whereas its surface coverage (N_{POISON I}) can increase with a larger N_BA--H⁺ or a larger N_LAII.²¹ Considering the trends in E_BA--H⁺/E_LAII values of O300 ≤ R300 < R600, N_BA--H⁺ values of R600 < R300 ∼ O300, and N_LAII values of O300 ∼ R300 ∼ R600, resistance to poison I (N_{POISON I}↓) was speculated to increase in the order of O300 ≤ R300 < R600. Similarly, poison III binds more firmly to LA_I at a lower E_LAI, while its accumulation (N_{POISON III}) can increase at a larger N_LAI.²¹ Considering the trends on E_LAI values of O300 < R300 < R600 and N_LAI values of R600 < R300 ∼ O300, tolerance to poison III (N_{POISON III}↓) was expected to increase in the order of O300 ≤ R300 < R600.

For inspecting the aforestated hypotheses, the catalysts were mixed with aqueous H₂O₂ and ground with a mortar and pestle to yield H₂O₂ activation-subjected, wet catalysts for Raman spectroscopy experiments. The resulting Raman spectra were curve-fitted to five Raman-active bands indexed to E_g or A_1g/B_1g.^51,52 The locations of the E_g bands (O300 ∼ R300 > R600) and their FWHM values (O300 ∼ R300 > R600) were consistent with those of the H₂O₂ activation-unsubjected, dry catalysts (Fig. 2F–J/S23 and Tables S3/S16). This indicated that the N_LAI/N_LAII hierarchies of O300 ∼ R300 > R600 were retained post their H₂O₂ activation.^51–54 Interestingly, additional Raman-excited bands appeared in the H₂O₂ activation-subjected, wet catalysts with Raman shifts centered at 856–871 cm⁻¹. These bands could be ascribed to photon-excited stretching vibrations of O–O bonds for O_PEROXY species (O_PEROXY). Although their intensities were lower relative to those of the E_g bands (*; Fig. S23), their presence was seemingly indispensable, as postulated based on literature reports to explore O_PEROXY bands via Raman spectroscopy.^98,99 Notably, the O_PEROXY band for wet R600 exhibited a red-shift of ∼15 cm⁻¹ relative to those for wet O300 and R300. This could partially result from elongation of O–O bonds in O_PEROXY species on the wet R600 surface, which might arise from its higher E_LAII/E_BA--H⁺ values (O300 ∼ R300 < R600) for poison I or its higher E_LAI values (O300 < R300 < R600) for poison III.^51–54 Moreover, the area ratio of the O_PEROXY band to the E_g band (marked with *; Fig. S23) was higher for wet O300/R300 (0.07–0.08) than for R600 (0.05). This proved that R600 could exhibit greater resistance to poison I/III accumulation than O300/R300.

To further examine the hypotheses set earlier, the catalysts underwent H₂O₂ dissection runs identical to the 1^st recycle runs except for the exclusion of acetaminophen from aqueous reaction solutions and the absence of an applied electric input. The latter was compensated by the inclusion of H₂O₂ in aqueous reaction solutions. Notably, C_H2O2 was set to 4.1 × 10⁻³ mol_H2O2 L⁻¹, corresponding to averaged C_H2O2 at ≥5 minutes in H₂O₂ evolution runs (Fig. S17) with a target to simulate the aqueous electric conditions used to carry out the 1st recycle runs (Fig. 7B). The catalysts subjected to H₂O₂ dissection runs were collected via filtration, dried at 110 °C for an hour, and analyzed via XP spectroscopy to explore the type or contents of surface O species (Fig. 8B and Table S17). The XP spectra of the H₂O₂ dissection-subjected catalysts in the O 1s regimes were curve-fitted into four sub-bands, three of which were assigned to O_LATTICE, O_CHEMISORBED, and O_BRÖNSTED species with their peak locations (binding energies) being identical to those of the H₂O₂ dissection-unsubjected catalysts.^57–60 This highly suggested that the trends in E_BA--H⁺/E_LAII/E_LAI and N_BA--H⁺/N_LAII/N_LAI values of the catalysts were maintained after undergoing H₂O₂ dissection runs. However, O300 showed a decrease in its O_BRÖNSTED content (24.7% → 18.6%), whereas only minor changes were observed for R300 (12.7% → 10.7%) and R600 (6.2% → 5.2%). This substantiated that poison I deposition on BA⁻–H⁺/LA_II sites was unavoidable on the O300 surface and was unlikely on R300 and R600 surfaces. Similarly, O_CHEMISORBED content decreased substantially for O300 (24.5% → 11.2%) and R300 (24.1% → 12.4%), yet remained almost unaltered for R600 (6.0% → 5.5%). This demonstrated that poison III deposition on LA_I sites was imperative on O300 and R300 surfaces and was minimal on the R600 surface.

The remaining sub-bands of the XP spectra in the O 1s regions could be indexed to O of Ti-peroxy species (poison I or III; O_PEROXY) with binding energies centered at 533.0–533.4 eV.²¹ Notably, the binding energies of O_PEROXY species inherent to H₂O₂ dissection-subjected R300 and R600 were 0.4 eV higher than that inherent to H₂O₂ dissection-subjected O300 (533.0 eV). This shift could be justified by differences in E_LAI (O300 < R300 < R600) and N_LAI values (R600 < R300 ∼ O300), both of which were lumped, as discussed earlier.^25,55,56 Hence, e⁻ donation from the O atom of poison III to proximal LA_I sites could be more pronounced for R300/R600 than for O300. Nevertheless, the O_PEROXY contents were substantially higher for H₂O₂ dissection-subjected O300 and R300 (17.8–18.6%) than for H₂O₂ dissection-subjected R600 (3.3%), which was in agreement with the trend in their O_CHEMISORBED contents directed by poison III accumulation on LA_I sites of the catalysts subjected to H₂O₂ dissection runs. Overall, R600 with the smallest N_BA--H⁺, highest E_BA--H⁺, and highest E_LAI showed the strongest reluctance to accumulation of oligomeric species, poison I, and poison III, respectively, upon exposure to aqueous organic degradation conditions.

Despite validation concerning the superiority of R600 over O300 and R300 in tolerating oligomeric species and poison I/III, their regenerabilities to sustain H₂O₂ homolysis remained unexamined. (Used) O300 and (used) R600 were thus chosen because of their similar resuabilities and poisoning patterns (Fig. 7B/8/S21 and Table S17). The −r_{ACETAMINOPHEN, 0, CO2} values of (used) O300 and R600 were then monitored over extended recycle runs up to the 10^th cycle (Fig. 9/S24 and Table S10). −r_{ACETAMINOPHEN, 0, CO2} values of O300 and R600 were continuously declined, whereas −r_{ACETAMINOPHEN, 0, CO2} values of the latter were 1.2–2.3-fold higher than those of the former at each of the recycle runs, clearly demonstrating that R600 outstripped O300 in sustaining ˙OH production via H₂O₂ homolysis.

Fig. 9 Background-corrected, initial acetaminophen degradation rates normalized on a per-gram basis (−r_{ACETAMINOPHEN, 0}) for (used) O300 and R600 subjected to acetaminophen degradation recycle runs and −r_{ACETAMINOPHEN, 0} values for O300 and R600 subjected to the 10^th cycle of acetaminophen degradation recycle runs followed by regeneration (marked with *). Reaction environments: 0.2 g of the (used or regenerated) catalyst with sizes of <200 µm; 100 mL of de-ionized H₂O; 0.1 mmol of acetaminophen; 0.14 mmol of Na₂SO₄; 3 V; 25 °C; 300 rpm; initial pH of 7.0; final pH of 4.5 (±0.3). Regeneration conditions: under N₂; ramping rate of 5 °C min⁻¹; 300 °C; 4 hours.

The R600 and O300 collected after being subjected to the 10^th cycle were further examined using the thermo-gravimetric analysis (TGA) technique, with which profiles of weight percent (wt%) versus temperature under N₂ were recorded (Fig. S25). The decrease in wt% monitored for used O300 and R600 corresponded to 11.4 wt% and 4.5 wt%, respectively, at 110–800 °C. This verified greater resistance to oligomeric poison or poison I/III deposition was achievable for R600 rather than for O300, while supporting greater recyclability of R600 in activating H₂O₂ homolysis compared to that of O300.

Moreover, the used catalysts exhibited continuous weight loss at 110–800 °C, where H₂O molecules chemisorbed on LA_I/LA_II sites could be desorbed at 110–400 °C, whereas acetaminophen (or its oligomeric derivative) and poison I/III (O_PEROXY) could be pyrolyzed or desorbed at 200–350 °C and at 110–170 °C, respectively. This was postulated based on a spectrum of literature reports on pyrolysis or desorption of H₂O, acetaminophen, and O_PEROXY via TGA under inert conditions.^100–103 Considering thermal behaviors of the aforementioned species and considering the synthetic conditions of O300 (air; 300 °C) and R600 (H₂; 600 °C), the used catalysts were regenerated under N₂ at 300 °C for 4 hours to restore their surface acidic features. The catalysts subjected to regeneration were then reevaluated under acetaminophen degradation conditions. The resulting −r_{ACETAMINOPHEN, 0, CO2} values were recovered to around 70% of those in the 1st cycle of the recycle runs (Fig. 9/S24 and Table S10). This highly suggested a fraction of oligomeric poisons or poison I/III could be pyrolyzed under inert conditions.

The catalysts subjected to the 10^th cycle or regeneration were further characterized by XP spectroscopy to explore their surface properties. The XP spectra in the O 1s regimes revealed such O_PEROXY contents that were around 3-fold higher in used O300 (32.6%) than in used R600 (Fig. S26 and Table S18). This was in accordance with the trends in −r_{ACETAMINOPHEN, 0, CO2} values (Fig. 9), wt% loss (via TGA; Fig. S25), surface C/Ti ratios (2.3 for used O300; 1.4 for used R600), and surface N/Ti ratios (0.26 for used O300; 0.11 for used R600) for the used catalysts (Table S18). Nonetheless, O_PEROXY contents of the used catalysts were markedly reduced to nearly 0% post their regeneration. This indicated that residual oligomeric poisons remained on the used catalyst surfaces even after regeneration. This was also proved by only partial recovery of O_BRÖNSTED/O_CHEMISORBED contents (∼9.8% for recovered O300), surface C/Ti ratios (1.5 for recovered O300), and surface N/Ti ratios (0.15 for recovered O300) for the used catalysts in comparison with those of the corresponding catalysts as-synthesized (O_BRÖNSTED/O_CHEMISORBED of ∼24.6%, C/Ti of ∼0, and N/Ti of ∼0 for as-synthesized O300; Tables S17 and S18). Nevertheless, surface C/Ti and N/Ti values were higher in recovered O300 than in regenerated R600 (C/Ti of 1.1 and N/Ti of 0.07 for recovered R600; Table S18). This corroborated that R600 outstripped O300 and R300 in sustaining repeated H₂O₂ homolysis and in restoring acidic features desired to activate H₂O₂ homolysis.

H₂O₂ homolysis on the catalysts: DFT calculations

To further justify our findings on poison resistance for the catalysts, density functional theory (DFT) calculations were carried out according to methodologies specified in the SI. A TiO₂ (anatase) slab was initially terminated on its major (101) facet (evidenced by XRD patterns; Fig. S2) and subjected to geometric optimization, upon which a fraction of hexa-fold coordinated, defect-free Ti atoms transformed into penta-fold coordinated analogues (LA_I) alongside the impartation of BA⁻–H⁺ sites (Fig. 10A; surface I).

Fig. 10 Representation of a tetragonal TiO₂ (anatase) slab subjected to termination on its (101) facet followed by geometric optimization (A; denoted as surface I), upon which 5-fold coordinated Ti atoms (LA_I) and BA⁻–H⁺ were imparted, whereas the O atom marked with * was eliminated for transition of vicinal LA_I sites to 4-fold coordinated Ti analogues (LA_II; D; denoted as surface II). Bond lengths marked with i, ii, and iii were computed to be 2.06 Å, 2.00 Å, and 1.86 Å, respectively, for surface I (A), whereas those were calculated to be 2.04 Å, 1.96 Å, and 1.88 Å, respectively, for surface II (D). (C) Bader charges of 4/5/6-fold coordinated Ti atoms pertaining to surface II. Depiction of surface I adsorbing O₂ on LA_I (B) in addition to the change in thermodynamic energy involved in O₂ adsorption on surface I (ΔE_ADSORPTION (O₂)). Illustration of surface II adsorbing ˙OH on BA⁻–H⁺/LA_II (E), ˙OOH on BA⁻–H⁺/LA_II (F), or O₂˙⁻ on two LA_II (G) alongside the change in thermodynamic energy involved in ˙OH, ˙OOH, or O₂˙⁻ adsorption on surface II (ΔE_ADSORPTION (˙OH), ΔE_ADSORPTION (˙OOH), or ΔE_ADSORPTION (O₂˙⁻)). Bond lengths marked with iv and v were computed to be 2.05 Å and 1.61 Å, respectively, for surface II (E), whereas those marked with vi and vii were calculated to be 2.10 Å and 1.01 Å, respectively, for surface II (F).

Poison III (O₂) was first attempted to undergo chemisorption on two adjacent LA_I sites of surface I, yet it did not converge to a stable local energy minimum. This could stem from octahedral [Ti⁴⁺–(O²⁻)₆]⁸⁻ sub-units of the TiO₂ lattice (Fig. 1B), whose geometric energy minimum is contingent upon stable coordination of each Ti⁴⁺ center with six O²⁻ anions (i.e., one for O₂ (poison III); five for O²⁻ anions; Fig. 10B). Nonetheless, it should not be disregarded that poison III could settle stably down to dynamic TiO₂ surface facets via the formation of covalent bonds with two adjacent LA_I sites (Fig. 8A).^3,34–36 In contrast, poison III adsorption on a single LA_I site on surface I was computed to be feasible, as boosted by the change in thermodynamic energy involved (ΔE_ADSORPTION (O₂) of −6.8 kJ mol⁻¹; Fig. 10B), proving the significance of poison III as a potential surface contaminator.

Moreover, an O atom positioned between two LA_I sites on surface I was removed to yield an O_V and two tetra-fold coordinated Ti centers (LA_II) in resulting surface II (Fig. 10D). Upon relaxation, the LA_II sites moved away from the O_V and revealed LA_II–O bond lengths of 2.04 Å (i), 1.96 Å (ii), and 1.88 Å (iii) (Fig. 10D). In particular, one of the LA_II–O bond lengths (ii) was 0.04 Å shorter than the corresponding LA_II–O bond on surface I (2.00 Å; Fig. 10A), validating that geometric distortion of [Ti⁴⁺–(O²⁻)₆]⁸⁻ sub-units was highly likely, as discussed earlier (Fig. 1B).³ Meanwhile, Bader charges of LA_I (n of 5) and LA_II (n of 4) sites on surface II were computed for comparison with that of the defect-less Ti center (n of 6). Bader charges were elevated in the sequence of LA_II < LA_I < defect-free Ti (Fig. 10C), substantiating that the quantity of e⁻ trapped near the sub-shells of the LA_II site was greater than that of either the LA_I site or defect-less Ti center, as stated earlier (Fig. 1C).^1,19–24

Surface II experienced ˙OH and poison I (˙OOH) adsorption on BA⁻–H⁺/LA_II sites (Fig. 10E and F). The likelihood of ˙OH adsorption was energetically spontaneous, as proved by the change in thermodynamic energy involved (ΔE_ADSORPTION (˙OH) of −209.4 kJ mol⁻¹), whereas the likelihood of poison I adsorption was energetically unspontaneous, as evidenced by the change in thermodynamic energy involved (ΔE_ADSORPTION (˙OOH) of 114.8 kJ mol⁻¹). Again, it should be stressed that surface facets of poly-crystalline TiO₂ were highly entangled, facilitating H₂O₂-mediated ˙OH transition to poison I on BA⁻–H⁺/LA_II sites (Fig. 8A).^21,34 Notably, in ˙OH-bound surface II (Fig. 10E), the bond between BA⁻ of BA⁻–H⁺ and the H atom of ˙OH(v) was 0.44 Å shorter than that between LA_II and the O atom of ˙OH(IV), proving E_BA--H⁺ < E_LAII for the catalyst surfaces in conjunction with the centrality of BA⁻–H⁺ functioning as the primary activator of H₂O₂ homolysis. In addition, despite energetic instability of ˙OOH-bound surface II (Fig. 10F), the bond between BA⁻ of BA⁻–H⁺ and the H atom of ˙OOH (vii) was 1.09 Å shorter than that between LA_II and the O atom of ˙OOH (vi), indicating that bond strength was higher in vii than in vi. This could support the unlikelihood of poison I (˙OOH)-induced formation of poison II (O₂˙⁻) on LA_II sites (Fig. 8A) because bond vii should be cleaved to yield poison II.^21,34 Indeed, the change in thermodynamic energy involved in poison II adsorption on two LA_II sites (not a single LA_II; due to marked stability of [Ti⁴⁺–(O²⁻)₆]⁸⁻ sub-units; vide supra) was computed to be 93.6 kJ mol⁻¹ (ΔE_ADSORPTION (O₂˙⁻); Fig. 10G), further demonstrating that poison II could scarcely be deposited on the catalyst surfaces.

H₂O₂ homolysis on the catalysts: adaptability

To examine the adaptabilities of the catalysts in disintegrating aqueous organics, acetaminophen, bisphenol A, aniline/phenol, and sulfanilamide/sulfamethoxazole were selected as model compounds representing an analgesic, endocrine disruptor, pesticide, and antibiotic, respectively (Fig. 11A).^14,21,22,37 These compounds contain distinct combinations of e⁻-donating groups (–OH/–NH–/–NH₂) and e⁻-withdrawing groups (–SO₂–) and thereby reveal dissimilar energies needed to deprive –OH/–NH–/–NH₂ of e⁻ (ionization potential; Table S19).^76–78 Accordingly, the −r_{POLLUTANT, 0, CO2} values of the catalysts should scale inversely with the ionization potentials of the pollutants, if their degradation primarily proceeded via (H⁺-coupled) e⁻ transfer on ROS (e.g., , Cl˙, or Fe^IV=O).^14,21,22,37 However, the catalysts elaborated herein could produce ˙OH, which is implemented to fragment contaminants primarily via addition or ˙H abstraction. Hence, the −r_{POLLUTANT, 0, CO2} values were expected to show little correlation with the ionization potentials of the pollutants.^14,21,22,37

Fig. 11 (A) Model compounds of hard-to-fragment aqueous pollutants with distinct electronic features with regard to e⁻ donation/acceptance or ionization tendency. (B) Background-corrected initial pollutant degradation rates normalized on a per-CO₂-accessible site basis (−r_{POLLUTANT, 0, CO2}) for the catalysts and (C) their plots of −r_{POLLUTANT, 0, CO2} values versus ionization potentials of pollutants. In (C), shown with dashed lines are the fits of the plots, across which their regression factors (R²) were evaluated. Reaction conditions: 0.2 g of the catalyst with sizes of <200 µm; 100 mL of de-ionized H₂O; 0.1 mmol of acetaminophen, 0.1 mmol of aniline, 0.1 mmol of sulfanilamide, 0.1 mmol of phenol, 0.1 mmol of sulfamethoxazole, or 0.004 mmol of bisphenol A; 0.14 mmol of Na₂SO₄; 3 V; 25 °C; 300 rpm; initial pH of 7.0; final pH of 4.5 (±0.3).

The catalysts underwent contaminant degradation runs identical to the diffusion runs, except for the replacement of acetaminophen with those listed earlier and the adjustment of initial concentration of bisphenol A to 0.4 × 10⁻⁴ mol_{BISPHENOL A} L⁻¹, with consideration of its aqueous solubility (Fig. S27 and Table S10).^{14,19–22,37,104,105} Moreover, kinetic datasets with X_POLLUTANT values of <30% were considered to precisely assess k_APP values of the catalysts, while ensuring that the equilibrium of pollutant degradation (or H₂O₂ homolysis) was effectively driven forward at T_REACTION of 25 °C.^{14,19–22,37} The −r_{POLLUTANT, 0, CO2} values of the catalysts revealed little correlation with the ionization potentials of the corresponding pollutants. This was demonstrated by the poor linear fits (R² < 0.1) obtained from plots of −r_{POLLUTANT, 0, CO2} versus ionization potential (Fig. 11C).^14,21,22,37 This indicated that the catalysts deployed ˙OH as the major decomposer of aqueous pollutants under the electric conditions, as also demonstrated by the scavenging runs and EPR spectroscopy experiments (Fig. 5B and 6). Importantly, the −r_{POLLUTANT, 0, CO2} values of R600 were around 2–17-fold and 3–23-fold higher than those of O300 and R300, respectively (Fig. 11B). The −r_{POLLUTANT, 0} values and extent of active metal cation (Ti⁴⁺) leaching for R600 were further compared with those of ZrO₂ (or the Zr-based metal–organic framework; UiO-66) and Fe₂O₃ (or Fe₃O₄), which we studied elsewhere^20,22,106 ZrO₂/UiO-66 and Fe₂O₃/Fe₃O₄ can produce ˙OH via H₂O₂ homolysis and H₂O₂ heterolysis (H₂O₂ + e⁻ → ˙OH + OH⁻), respectively, under the electric conditions (Table S20).

Despite negligible Ti⁴⁺/Zr⁴⁺ leaching (∼10⁻⁴ mol%), the −r_{POLLUTANT, 0} values of R600 were around 3–10-fold higher than those of ZrO₂ (or UIO-66).^20,22 In addition, the number of Ti⁴⁺ cations released from R600 was 3 orders of magnitude smaller than that of Fe^2+/3+ cations liberated from Fe₂O₃ (or Fe₃O₄; ∼10⁻¹ mol%) in bisphenol A degradation runs, although the −r_{BISPHENOL A, 0} values were comparable.¹⁰⁶ Overall, these results corroborated the superiority of R600 to O300/R300 or ZrO₂/UiO-66/Fe₂O₃/Fe₃O₄ in disintegrating refractory organic wastes in optical energy-free, aqueous media.

Conclusions

TiO₂ subjected to thermal H₂ reduction generates BA⁻–H⁺ and LA_I/LA_II sites. The latter create intra-bandgap states (INTRA) that localize e⁻ near the valence band (VB) to bear h⁺ with h⁺ and e⁻ being photo-excited to produce ˙OH, O₂˙⁻, or ¹O₂ (ROS) via semi-conducting routes. However, control over the type of ROS generated through selective acceleration of these pathways remains challenging. This is because E_INTRA and E_VB rely highly on the aqueous pH range and because the narrow optical energy gap of E_INTRA–E_VB can limit the range of accessible optical transitions, thereby restricting the number of ROS production pathways that can be activated.

Alternatively, non-optical H₂O₂ homolysis (H₂O₂ → 2˙OH) was envisioned as a route to deploy BA⁻–H⁺ and LA_II sites on H₂-reduced TiO₂. These sites proceed with H₂O₂ adsorption, H₂O₂ distortion, H₂O₂ dissection, and ˙OH desorption (RDS) with a greater N_BA--H⁺ (or N_LAII) and a higher E_BA--H⁺ (or E_LAII) favoring elevation and E_BARRIER reduction, respectively, for promoted ˙OH productivity. In this regard, TiO₂ experienced O₂ calcination at 300 °C and H₂ reduction at 300–600 °C to produce O300 and R300–R600, respectively, all of which revealed acidic trends of N_LAII < N_LAI ≪ N_BA--H⁺ and E_BA--H⁺ < E_LAI ∼ E_LAII. This made it compelling that the H₂O₂ consumption rate (−r_H2O2) law could hinge on N_BA--H⁺/E_BA--H⁺ rather than on N_LAII/E_LAII. Meanwhile, −r_H2O2 is challenging to directly monitor owing to the short life-span of ˙OH (∼10⁻³ µs) and thereby was assessed using background-corrected −r_{ACETAMINOPHEN, 0, CO2} and/or −r_{ACETAMINOPHEN, 0}. This approach was supported by the minute LA_II leaching, the absence of internal/external mass-transfer limitations for H₂O₂ ↔ BA⁻–H⁺/LA_II and ˙OH ↔ acetaminophen interactions, and the identification of ˙OH (rather than ˙OOH/O₂˙⁻) as the primary ROS responsible for acetaminophen decomposition.

In particular, O300, R300, and R600 were distinct and tunable in terms of acidic traits such as R600 < R300 ∼ O300 for N_BA--H⁺, O300 ∼ R300 ∼ R600 for N_LAII, and O300 ≤ R300 < R600 for E_BA--H⁺/E_LAII, in conjunction with H₂O₂ dissector (BA⁻–H⁺/LA_II)-proximal LA_I features of R600 < R300 ∼ O300 for N_LAI and O300 < R300 < R600 for E_LAI. R600 outstripped O300 and R300 in multiple aspects. R600 revealed the lowest E_BARRIER and lowest , yet displayed the highest −r_{ACETAMINOPHEN, 0, CO2}. This indicated that E_BARRIER outweighed in directing −r_{ACETAMINOPHEN, 0, CO2} for TiO₂ subjected to H₂ reduction. R600 also unveiled the greatest tolerance to deposit poisonous oligomers on BA⁻–H⁺/LA_II sites, arising from its smallest N_BA--H⁺. In addition, R600 showed the highest resistance to accumulate poisonous ˙OOH on BA⁻–H⁺/LA_II sites or poisonous O₂ on LA_I sites, arising from its smallest N_BA--H⁺, highest E_BA--H⁺, and highest E_LAII for ˙OOH and its smallest N_LAI coupled with highest E_LAI for O₂. All of these could be gathered with the DFT calculations to justify the highest −r_{ACETAMINOPHEN, 0} values for R600 throughout acetaminophen disintegration recycle runs in conjunction with its moderate recovery upon exposure to regeneration environments. Furthermore, R600 disclosed the greatest adaptability to fragment diverse aqueous pollutants including an analgesic (acetaminophen), endocrine disruptor (bisphenol A), pesticide (aniline/phenol), and antibiotic (sulfanilamide/sulfamethoxazole). This was corroborated by −r_{POLLUTANT, 0, CO2} values of R600, which were 2–23-fold higher than those of O300/R300 and 3–10 fold higher than those of ZrO₂/UiO-66.

Author contributions

Minsung Kim: methodology, validation, data curation, writing – original draft, formal analysis, investigation, visualization. Gyuchan Kim: methodology, validation, data curation, writing – editing, formal analysis, investigation, visualization, software. Junseo Kim: methodology, validation, data curation, writing – editing, formal analysis, investigation, visualization. Sang Hoon Kim: writing – editing, resources, funding acquisition. Byung-Hyun Kim: methodology, validation, data curation, writing – original draft, investigation, visualization, software, supervision, resources. Jongsik Kim: conceptualization, methodology, validation, data curation, writing – original and revised draft, investigation, visualization, supervision, resources, project administration, funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information: experimental and computational sections; semi-conducting mechanisms; secondary quenching rate constants for the scavengers; ionization potentials of the aqueous toxicants; properties, surfaces, Tauc fits, XRD patterns, Raman/XP/EPR spectra, N₂/CO/CO₂ isotherms, and CO-TPD/CO₂-TPD profiles for the (used) catalysts; reaction profiles, rates, k_APP values, Arrhenius plots, rate laws, and toxicant degradation pathways of the (used) catalysts. See DOI: https://doi.org/10.1039/d6ta01339b.

Acknowledgements

This work was supported by the Korea Institute for Advancement of Technology (KIAT) grant funded by the Korea Government (MOTIR) (RS-2026-25509542, HRD Program for Industrial Innovation). This study was conducted with the support of the equipment and facilities provided by the ACE Center at Kyung Hee University. We are thankful to the Korea Institute of Science and Technology for supporting this project through Future R&D.

References

1. V. K. Paidi, B.-H. Lee, D. Ahn, K.-J. Kim, Y. Kim, T. Hyeon and K.-S. Lee, Nano Lett., 2021, 21, 7953–7959.
2. J. Chapman, K. E. Kweon, Y. Zhu, K. Bushick, L. B. Bayu Aji, C. A. Colla, H. Mason, N. Goldman, N. Keilbart, S. R. Qiu, T. W. Heo, J. Rodriguez and B. C. Wood, J. Mater. Chem. A, 2023, 11, 8670–8683.
3. D. Van den Eynden, R. Pokratath and J. De Roo, Chem. Rev., 2022, 122, 10538–10572.
4. Ľ. Benco, J.-L. Barras, C. A. Daul and E. Deiss, Inorg. Chem., 1999, 38, 20–28.
5. Z.-W. Qu and G.-J. Kroes, J. Phys. Chem. B, 2006, 110, 8998–9007.
6. J. Shim, K. Lee, Y. Yu, H. S. Lee, H. Shin, K.-S. Lee, M. S. Bootharaju, S. Han, G. S. Yi, H. Ko, S. Lee, J. Ryu, M. Kim, B.-H. Lee, T. Hyeon and Y.-E. Sung, J. Am. Chem. Soc., 2025, 147, 16179–16188.
7. B. Ji, J. Gou, Y. Zheng, X. Pu, Y. Wang, P. Kidkhunthod and Y. Tang, Adv. Mater., 2023, 35, 2300381.
8. H. Song, W. Xie, Y. Tian, M. Guo, T. Wang, D. Kang, M. Jia and X. Zhang, J. Mater. Chem. A, 2024, 12, 15154–15162.
9. L. Jing, B. Xin, F. Yuan, L. Xue, B. Wang and H. Fu, J. Phys. Chem. B, 2006, 110, 17860–17865.
10. M. S. Hamdy, R. Amrollahi and G. Mul, ACS Catal., 2012, 2, 2641–2647.
11. S. Oh, J.-H. Kim, H. M. Hwang, D. Kim, J. Kim, G. H. Park, J. S. Kim, Y. H. Lee and H. Lee, J. Mater. Chem. A, 2021, 9, 4822–4830.
12. A. J. Hoffman, E. R. Carraway and M. R. Hoffmann, Environ. Sci. Technol., 1994, 28, 776–785.
13. X. Li, C. Chen and J. Zhao, Langmuir, 2001, 17, 4118–4122.
14. M. Kim, H. Lee, J. Kim, H. Yu, T. Yu, K. Jeong and J. Kim, J. Alloys Compd., 2024, 1008, 176782.
15. W. He, H.-K. Kim, W. G. Wamer, D. Melka, J. H. Callahan and J.-J. Yin, J. Am. Chem. Soc., 2014, 136, 750–757.
16. L. Wang, J. Li, X. Liu, J. Zhang, P. Zeng and Y. Song, Chemosphere, 2023, 324, 138264.
17. P. Chen, Y. Zhao, C. Li, X. Chen and J. Wang, J. Mater. Chem. A, 2025, 13, 25829–25841.
18. L. Wu, H. Li, W. Wang, S. Zhang, S. Zeng and X. Zhou, J. Mater. Chem. A, 2025, 13, 37008–37017.
19. M. Kim, J. Park, S. H. Kim, J.-H. Lee, K. Jeong and J. Kim, Carbon, 2023, 203, 630–649.
20. Y. J. Choe, S. Lee, M. Kim, S. H. Kim, I.-S. Choi, K. Jeong and J. Kim, Sep. Purif. Technol., 2023, 310, 123146.
21. J. Kim, M. Kim, B. Y. Woo, S. Byun, G. Bang, K. M. Cho, H. Jung, K. Jeong and J. Kim, Chem. Eng. J., 2025, 519, 165063.
22. M. Kim, M. Al Mamunur Rashid, Y. J. Choe, S. H. Kim, J.-H. Lee, K. Jeong and J. Kim, J. Mater. Chem. A, 2023, 11, 9436–9454.
23. N. M. Dimitrijevic, Z. V. Saponjic, B. M. Rabatic, O. G. Poluektov and T. Rajh, J. Phys. Chem. C, 2007, 111, 14597–14601.
24. L.-B. Xiong, J.-L. Li, B. Yang and Y. Yu, J. Nanomater., 2012, 2012, 831524.
25. T. S. Rajaraman, S. P. Parikh and V. G. Gandhi, Chem. Eng. J., 2020, 389, 123918.
26. B. Bharti, S. Kumar, H.-N. Lee and R. Kumar, Sci. Rep., 2016, 6, 32355.
27. D. Nanda Gopala Krishna, R. P. George and J. Philip, J. Phys. Chem. C, 2021, 125, 16136–16146.
28. E. Albanese, A. Ruiz Puigdollers and G. Pacchioni, ACS Omega, 2018, 3, 5301–5307.
29. Y. Li, W. Zhang, J. Niu and Y. Chen, ACS Nano, 2012, 6, 5164–5173.
30. Q. Xiang, J. Yu and P. K. Wong, J. Colloid Interface Sci., 2011, 357, 163–167.
31. Z. Feng, Q. Tian, Q. Yang, Y. Zhou, H. Zhao and G. Zhao, Appl. Catal., B, 2021, 286, 119908.
32. T. D. Gould, A. M. Lubers, B. T. Neltner, J. V. Carrier, A. W. Weimer, J. L. Falconer and J. Will Medlin, J. Catal., 2013, 303, 9–15.
33. R. Barth, R. Pitchai, R. L. Anderson and X. E. Verykios, J. Catal., 1989, 116, 61–70.
34. K. Sahel, L. Elsellami, I. Mirali, F. Dappozze, M. Bouhent and C. Guillard, Appl. Catal., B, 2016, 188, 106–112.
35. C. M. Lousada, A. J. Johansson, T. Brinck and M. Jonsson, J. Phys. Chem. C, 2012, 116, 9533–9543.
36. W.-F. Huang, P. Raghunath and M. C. Lin, J. Comput. Chem., 2011, 32, 1065–1081.
37. M. Kim, S. H. Kim and J. Kim, Sep. Purif. Technol., 2025, 378, 134563.
38. S. Yu, S. Tong, M. Chen, H. Zhang, Y. Xu, Y. Guo and M. Ge, Environ. Sci. Technol., 2025, 59, 11666–11676.
39. T.-Y. Liu, C. Wang, Y.-Z. Han, C. Bai, H.-T. Ren, Y. Liu and X. Han, Chem. Eng. J., 2022, 435, 134816.
40. J. Kim, S. Lee and H. P. Ha, ACS Catal., 2021, 11, 767–786.
41. S. Bakardjieva, J. Šubrt, V. Štengl, M. J. Dianez and M. J. Sayagues, Appl. Catal., B, 2005, 58, 193–202.
42. W. Zhang, X. Liu, L. Li, Z. Sun, S. Han, Z. Wu and J. Luo, Chem. Mater., 2018, 30, 4081–4088.
43. S. Kumar, S. Khanchandani, M. Thirumal and A. K. Ganguli, ACS Appl. Mater. Interfaces, 2014, 6, 13221–13233.
44. N. Riaz, F. K. Chong, B. K. Dutta, Z. B. Man, M. S. Khan and E. Nurlaela, Chem. Eng. J., 2012, 185–186, 108–119.
45. S. Lettieri, V. Gargiulo, M. Alfè, M. Amati, P. Zeller, V.-A. Maraloiu, F. Borbone, M. Pavone, A. B. Muñoz-García and P. Maddalena, J. Phys. Chem. C, 2020, 124, 3564–3576.
46. Y. Zhang, Z. Su, A. K. Azad, W. Zhou and J. T. S. Irvine, Adv. Energy Mater., 2012, 2, 316–321.
47. S. Jiang, G. Su, J. Wu, C. Song, Z. Lu, C. Wu, Y. Wang, P. Wang, M. He, Y. Zhao, Y. Jiang, X. Zhao, H. Rao and M. Sun, ACS Appl. Mater. Interfaces, 2023, 15, 11787–11801.
48. F. Nekouei, C. J. Pollock, T. Wang, Z. Zheng, Y. Zhang, Z. Fusco, H. Jin, T. R. Ramireddy, A. A. Wibowo, T. Lu, S. Nekouei, F. Keshtpour, J. Langley, E. H. Abdelkader, N. Cox, Z. Yin, H. Nguyen, A. Glushenkov, S. Karuturi, Z. Liu, L. Wei, H. Li and Y. Liu, ACS Catal., 2025, 15, 1431–1443.
49. N. E. Kochkina, A. V. Agafonov, A. V. Vinogradov, N. S. Karasev, N. L. Ovchinnikov and M. F. Butman, ACS Sustainable Chem. Eng., 2017, 5, 5148–5155.
50. R. Scotti, I. R. Bellobono, C. Canevali, C. Cannas, M. Catti, M. D'Arienzo, A. Musinu, S. Polizzi, M. Sommariva, A. Testino and F. Morazzoni, Chem. Mater., 2008, 20, 4051–4061.
51. S. K. Mukherjee and D. Mergel, J. Appl. Phys., 2013, 114, 013501.
52. H. Li, L. Shen, K. Zhang, B. Sun, L. Ren, P. Qiao, K. Pan, L. Wang and W. Zhou, Appl. Catal., B, 2018, 220, 111–117.
53. X. Yu, B. Kim and Y. K. Kim, ACS Catal., 2013, 3, 2479–2486.
54. H. C. Choi, Y. M. Jung and S. B. Kim, Vib. Spectrosc., 2005, 37, 33–38.
55. G. An, W. Ma, Z. Sun, Z. Liu, B. Han, S. Miao, Z. Miao and K. Ding, Carbon, 2007, 45, 1795–1801.
56. R. Katal, M. Salehi, M. H. Davood Abadi Farahani, S. Masudy-Panah, S. L. Ong and J. Hu, ACS Appl. Mater. Interfaces, 2018, 10, 35316–35326.
57. W. Shen, D. Xiang, J. Yang, Y. Tang, C. Xin, Q. Guo and X. Yu, Colloids Surf., A, 2022, 653, 129965.
58. T. Su, Y. Yang, Y. Na, R. Fan, L. Li, L. Wei, B. Yang and W. Cao, ACS Appl. Mater. Interfaces, 2015, 7, 3754–3763.
59. U. Diebold, Surf. Sci. Rep., 2003, 48, 53–229.
60. C. Sun, L.-M. Liu, A. Selloni, G. Q. Lu and S. C. Smith, J. Mater. Chem. A, 2010, 20, 10319–10334.
61. A. Naldoni, M. Altomare, G. Zoppellaro, N. Liu, Š. Kment, R. Zbořil and P. Schmuki, ACS Catal., 2019, 9, 345–364.
62. J. Kim, Y. J. Choe, S. H. Kim, I.-S. Choi and K. Jeong, JACS Au, 2021, 1, 1158–1177.
63. J. Sui, M.-L. Gao, C. Liu, Y. Pan, Z. Meng, D. Yuan and H.-L. Jiang, J. Mater. Chem. A, 2023, 11, 18733–18739.
64. A. Vimont, A. Travert, P. Bazin, J.-C. Lavalley, M. Daturi, C. Serre, G. Férey, S. Bourrelly and P. L. Llewellyn, Chem. Commun., 2007, 3291–3293,  10.1039/B703468G.
65. X. Ma, L. Li, R. Chen, C. Wang, H. Li and S. Wang, Appl. Surf. Sci., 2018, 435, 494–502.
66. V. Bolis, C. Morterra, B. Fubini, P. Ugliengo and E. Garrone, Langmuir, 1993, 9, 1521–1528.
67. X. Wang, L. Lu, B. Wang, Z. Xu, Z. Xin, S. Yan, Z. Geng and Z. Zou, Adv. Funct. Mater., 2018, 28, 1804191.
68. Y. Guo, C. Tan, J. Sun, W. Li, J. Zhang and C. Zhao, Chem. Eng. J., 2020, 381, 122736.
69. S. Lee, S. H. Park and J. Kim, J. Mater. Chem. A, 2024, 12, 24574–24592.
70. S. Lee, J. Choi, H. P. Ha, J.-H. Lee, J. Park and J. Kim, ACS Catal., 2024, 14, 3349–3368.
71. S. H. Park, S. Lee and J. Kim, J. Hazard. Mater., 2025, 484, 136653.
72. G. Zhan, Y. Hong, F. Lu, A.-R. Ibrahim, M. Du, D. Sun, J. Huang, Q. Li and J. Li, J. Mol. Catal. A:Chem., 2013, 366, 215–221.
73. G. Vega, A. Quintanilla, M. Belmonte and J. A. Casas, Chem. Eng. J., 2022, 428, 131128.
74. M. P. Murphy, H. Bayir, V. Belousov, C. J. Chang, K. J. A. Davies, M. J. Davies, T. P. Dick, T. Finkel, H. J. Forman, Y. Janssen-Heininger, D. Gems, V. E. Kagan, B. Kalyanaraman, N.-G. Larsson, G. L. Milne, T. Nyström, H. E. Poulsen, R. Radi, H. Van Remmen, P. T. Schumacker, P. J. Thornalley, S. Toyokuni, C. C. Winterbourn, H. Yin and B. Halliwell, Nat. Metab., 2022, 4, 651–662.
75. J. S. Barroso-Martínez, A. I. B. Romo, S. Pudar, S. T. Putnam, E. Bustos and J. Rodríguez-López, J. Am. Chem. Soc., 2022, 144, 18896–18907.
76. J. Wang, K.-P. Hou, Y. Wen, H. Liu, H. Wang, K. Chakarawet, M. Gong and X. Yang, J. Am. Chem. Soc., 2022, 144, 4294–4299.
77. P. Hu, H. Su, Z. Chen, C. Yu, Q. Li, B. Zhou, P. J. J. Alvarez and M. Long, Environ. Sci. Technol., 2017, 51, 11288–11296.
78. H. Lei and S. A. Snyder, Water Res., 2007, 41, 4051–4060.
79. J. Kim, Y. J. Choe and S. H. Kim, Chem. Eng. J., 2021, 413, 127550.
80. S. O. Ganiyu, N. Oturan, S. Raffy, M. Cretin, C. Causserand and M. A. Oturan, Sep. Purif. Technol., 2019, 208, 142–152.
81. V. K. Parida, S. K. Srivastava, S. Chowdhury and A. K. Gupta, Chem. Eng. J., 2023, 472, 144969.
82. Q. Mei, J. Sun, D. Han, B. Wei, Z. An, X. Wang, J. Xie, J. Zhan and M. He, Chem. Eng. J., 2019, 373, 668–676.
83. J. R. Alvarez-Idaboy, N. Mora-Diez, R. J. Boyd and A. Vivier-Bunge, J. Am. Chem. Soc., 2001, 123, 2018–2024.
84. T. Yasuhisa, H. Hideki and Y. Muneyoshi, Int. J. Biochem., 1993, 25, 491–494.
85. G. V. Buxton, C. L. Greenstock, W. P. Helman and A. B. Ross, J. Phys. Chem. Ref. Data, 1988, 17, 513–886.
86. B. H. J. Bielski, D. E. Cabelli, R. L. Arudi and A. B. Ross, J. Phys. Chem. Ref. Data, 1985, 14, 1041–1100.
87. Y. Feng, D. Wu, H. Li, J. Bai, Y. Hu, C. Liao, X.-y. Li and K. Shih, ACS Sustainable Chem. Eng., 2018, 6, 3624–3631.
88. S. S. Shinde, M. P. Hay, A. V. Patterson, W. A. Denny and R. F. Anderson, J. Am. Chem. Soc., 2009, 131, 14220–14221.
89. S. Dikalov, J. Jiang and R. P. Mason, Free Radical Res., 2005, 39, 825–836.
90. J. M. Fontmorin, R. C. Burgos Castillo, W. Z. Tang and M. Sillanpää, Water Res., 2016, 99, 24–32.
91. L. Khachatryan and B. Dellinger, Environ. Sci. Technol., 2011, 45, 9232–9239.
92. I. Yamazaki, L. H. Piette and T. A. Grover, J. Biol. Chem., 1990, 265, 652–659.
93. H.-Y. Gao, C.-H. Huang, L. Mao, B. Shao, J. Shao, Z.-Y. Yan, M. Tang and B.-Z. Zhu, Environ. Sci. Technol., 2020, 54, 14046–14056.
94. A. Staško, M. Liptáková, F. Malík and V. Mišík, Appl. Magn. Reson., 2002, 22, 101–113.
95. Y. Li, Y. Pan, L. Lian, S. Yan, W. Song and X. Yang, Water Res., 2017, 109, 266–273.
96. W. Li, X. Zhang and J. Han, Environ. Sci. Technol., 2022, 56, 16929–16939.
97. J. Lu, Q. Huang and L. Mao, Environ. Sci. Technol., 2009, 43, 7062–7067.
98. S.-H. Chung, G. H. Park, N. Schukkink, H. Lee and N. R. Shiju, Chem. Commun., 2023, 59, 756–759.
99. H. Park, T. Goto, S. Cho, H. Nishida and T. Sekino, ACS Omega, 2020, 5, 21753–21761.
100. G. Li, L. Li, J. Boerio-Goates and B. F. Woodfield, J. Am. Chem. Soc., 2005, 127, 8659–8666.
101. H.-H. Kim, H. Lee, D. Lee, Y.-J. Ko, H. Woo, J. Lee, C. Lee and A. L.-T. Pham, Environ. Sci. Technol., 2020, 54, 15424–15432.
102. J. de Freitas, A. P. G. Ferreira, C. Gaglieri, G. Bannach and É. T. G. Cavalheiro, J. Therm. Anal. Calorim., 2025, 150, 4165–4176.
103. M. Mgharbel, L. Halawy, A. Milane, J. Zeaiter and W. Saad, J. Anal. Appl. Pyrolysis, 2023, 172, 106014.
104. Y. Zhang, C. Causserand, P. Aimar and J. P. Cravedi, Water Res., 2006, 40, 3793–3799.
105. J. Xu, L. Wang and Y. Zhu, Langmuir, 2012, 28, 8418–8425.
106. Y. J. Choe, S. H. Kim, K. Jeong and J. Kim, Chem. Eng. J., 2023, 455, 140537.

---

Footnote

† Co-1st authors.

---