Interfacial Electronic Regulation of GO/Na_0.5Bi_2.5Nb₂O₉ Aurivillius-Layered Perovskite Heterointerfaces for Enhanced Photocatalysis

Highlights

What are the main findings?

• Intimate GO/Na_0.5Bi_2.5Nb₂O₉ heterointerfaces were successfully constructed.
• Interfacial electronic regulation was evidenced by XPS and flat-band potential analyses.

What are the implications of the main findings?

• GO induced interfacial electronic interactions in Na_0.5Bi_2.5Nb₂O₉.
• GO modulated the photocatalytic reaction pathway.

Abstract

Constructing heterointerfaces between semiconductor photocatalysts and carbonaceous materials is an effective strategy for enhancing photocatalytic performance. However, the role of graphene oxide (GO) in modulating the interfacial electronic properties of Aurivillius-layered perovskites has remained unexplored, until now. In this work, GO/Na_0.5Bi_2.5Nb₂O₉ (ABNO) composites, denoted as GABNO, with different GO loadings were successfully synthesized using a hydrothermal method. Structural, chemical and surface characterizations confirmed the formation of intimate GO/ABNO heterointerfaces without altering the crystal structure of ABNO. GO incorporation established electronically coupled heterointerfaces and modified the interfacial electronic environment, as evidenced by XPS and flat-band potential analyses. Among all samples, GABNO 0.1 (0.1 mg GO loading) exhibited the highest photocatalytic activity, achieving nearly 100% Rhodamine B removal within 120 min and significantly enhancing methyl orange removal under UV–visible light irradiation. Radical trapping experiments further revealed that GO altered the dominant photocatalytic reaction pathway from hole-dominated oxidation to superoxide-radical-mediated degradation. The superior photocatalytic performance is attributed to GO-induced interfacial electronic interactions, which facilitate efficient interfacial electron utilization and alter the dominant photocatalytic reaction pathway. This work provides new insight into the design of GO/Aurivillius-layered perovskite and highlights interfacial electronic regulation as an effective strategy for developing high-performance photocatalysts for environmental remediation.

1. Introduction

Semiconductor photocatalysis has emerged as a promising technology for environmental remediation and green energy production owing to its ability to utilize solar energy under mild operating conditions [1,2]. In particular, heterogeneous photocatalysis has attracted increasing attention for the degradation of organic pollutants and the development of sustainable energy-conversion processes, owing to its potential to efficiently utilize solar energy and promote environmentally friendly reactions [3]. However, the photocatalytic efficiency of many oxide semiconductors remains limited by sluggish charge transfer, rapid recombination of photogenerated charge carriers and inefficient utilization of reactive oxygen species [1]. Therefore, developing photocatalysts with well-regulated interfacial electronic interactions and charge-transfer behaviour remains an important challenge.

Among various oxide photocatalysts, Aurivillius-type layered perovskites have attracted considerable attention owing to their unique layered crystal structure, excellent chemical stability and intrinsic spontaneous polarization, which can facilitate the separation of photogenerated charge carriers [4,5]. These materials consist of alternating (Bi₂O₂)²⁺ layers and perovskite-like slabs, which generate internal electric fields that facilitate directional charge migration. The unique sandwich-like layered architecture also provides readily accessible active sites for photocatalytic reactions [6,7]. In particular, Na_0.5Bi_2.5Nb₂O₉ (ABNO) has emerged as a promising lead-free Aurivillius photocatalyst because of its ferroelectric characteristics, in which spontaneous polarization induces internal electric fields that promote the separation and migration of photogenerated charge carriers, thereby reducing charge recombination [8]. Combined with its excellent structural stability (i.e., high resistance to photocorrosion and good recyclability) and environmental benignity, ABNO has attracted increasing attention for photocatalytic applications [8]. Nevertheless, its photocatalytic performance remains limited by its intrinsically low electrical conductivity and inefficient interparticle charge transfer, resulting in poor utilization of photogenerated charge carriers during photocatalytic reactions [4,5].

Constructing heterojunctions is an effective strategy for improving photocatalytic performance by promoting the separation and utilization of photogenerated charge carriers [9]. Particularly, constructing heterostructured composites with carbon-based materials has emerged as a promising approach for enhancing interfacial charge transfer and photocatalytic activity. Among them, graphene oxide (GO) has attracted considerable attention owing to its two-dimensional sheet-like morphology, large specific surface area, abundant oxygen-containing functional groups and excellent electron-accepting capability, which collectively promote interfacial electronic interactions and provide abundant adsorption sites for organic pollutants [10]. Consequently, GO has been widely employed as an interfacial modifier to tailor the physicochemical properties of semiconductor photocatalysts. Previous studies have demonstrated that coupling GO with layered bismuth oxide photocatalysts, particularly Bi₂O₂CO₃, can enhance photocatalytic activity by improving light harvesting, increasing adsorption capacity through a larger specific surface area, and facilitating charge transfer and separation [11,12,13]. Nevertheless, these studies have been limited to non-Aurivillius-layered bismuth oxides and have primarily focused on reduced graphene oxide (rGO), including rGO/Bi₂WO₆ and rGO/Bi₂MoO₆ systems [14,15,16,17]. To the best of our knowledge, however, the integration of GO with Aurivillius-type layered perovskites has not yet been reported. In particular, the formation of GO/ABNO heterointerfaces and their influence on the interfacial electronic structure and photocatalytic behaviour remain unexplored.

To bridge this knowledge gap, GO/ABNO composites with different GO loadings were synthesized via a hydrothermal method. The formation of intimate GO/ABNO heterointerfaces was systematically characterized by electron microscopy, elemental mapping and surface chemical analyses. Their interfacial electronic properties were investigated using X-ray photoelectron spectroscopy, flat-band potential analysis and transient photocurrent measurements, while the photocatalytic activities toward Rhodamine B (Rh B) and methyl orange (MO) removal were evaluated under UV–vis light irradiation, with pathways elucidated through radical scavenging experiments. By correlating the structural, electronic and photocatalytic properties, this work demonstrates that GO functions as an active interfacial electronic regulator and modulates the photocatalytic reaction pathway of ABNO. These findings provide new mechanistic insight into interfacial engineering strategies for developing Aurivillius-type layered perovskite photocatalysts.

2. Experimental

2.1. Materials

The starting materials used in this work include Bi(NO₃)₃·5H₂O (≥99%, Sigma-Aldrich, St. Louis, MO, USA), C₄H₄NNbO₉ (99.99%, Sigma-Aldrich), NaOH (≥98%, Sigma-Aldrich) and HNO₃ (68–70%, Carl Roth, Karlsruhe, Germany). These chemicals were used in the experiments as received, without any further purification.

2.1.1. Sample Preparation

GO was synthesized using a modified Hummer’s method, as described elsewhere, and biocarbon was used as the carbon precursor [18]. ABNO was synthesized using a procedure similar to that described in the literature [8]. GABNO samples were synthesized using a similar procedure to that employed for ABNO. After the precursor solution had been prepared, GO was added under continuous stirring for 30 min. The resulting mixture was transferred to a Teflon-lined stainless-steel autoclave and heated at 180 °C for 36 h under hydrothermal conditions. After the reaction was completed, the resulting solid was collected by centrifugation, washed three times with deionized water, rinsed with ethanol and dried at 100 °C overnight. The GO content in the precursor solution was varied at 0.1, 0.5, 1 and 5 mg. The synthesized samples were denoted as GABNO 0.1, GABNO 0.5, GABNO 1 and GABNO 5, respectively.

2.1.2. Sample Characterizations

The XRD patterns were obtained using a Bruker D2 PHASER diffractometer (Bruker AXS GmbH, Karlsruhe, Germany) with Cu K_α radiation (λ = 1.5406 Å), operated at 30 kV and 10 mA. Data were acquired in the 2θ range of 5–80° with a step increment of 0.02° and a dwell time of 0.2 s per step. Raman measurements were carried out with a Horiba XploRA PLUS Raman microscope (HORIBA Scientific (HORIBA Jobin Yvon), Lille, France) using a 532 nm laser as the excitation source. The laser output was adjusted to 1% power, corresponding to approximately 0.86 mW at the sample surface. Spectra were obtained within the range of 52–1200 cm⁻¹, with a resolution of 3.22 cm⁻¹. Each spectrum was recorded using a 20-s acquisition time and averaged over 10 accumulations, giving a total measurement duration of approximately 3.3 min per sample. Field-emission scanning electron microscopy (FE-SEM) was performed using a Zeiss Merlin Compact microscope (Carl Zeiss Microscopy GmbH, Oberkochen, Germany) equipped with an Oxford Instruments Aztec energy-dispersive X-ray spectroscopy (Oxford Instruments plc, Abingdon, Oxfordshire, UK) (EDS) at an accelerating voltage of 10 kV. The samples were dispersed in ethanol, deposited onto aluminium tape mounted on a specimen stub via drop-casting and subsequently coated with gold before analyses. A Talos F200X G2 field-emission transmission electron microscope (FETEM, Thermo Fisher Scientific, Waltham, MA, USA), coupled with energy-dispersive X-ray spectroscopy (EDS), was used to obtain high-resolution TEM images, selected area electron diffraction (SAED) patterns and EDS mappings. Prior to FETEM-EDS measurement, the samples were prepared by dispersing them in ethanol and depositing the suspension onto a copper grid via drop-casting. A Micromeritics 3Flex analyser (Micromeritics Instrument Corporation, Norcross, GA, USA) was used to perform adsorption–desorption measurements using N₂ as the adsorptive gas and the specific surface areas were evaluated based on the Brunauer–Emmett–Teller (BET) approach. Degassing was carried out on each sample (~200 mg) at 80 °C for 12 h to eliminate weakly adsorbed molecules prior to analysis. X-ray photoelectron spectroscopy (XPS) was performed using a PHI5000 VersaProbe III (ULVAC-PHI, Chigasaki, Kanagawa, Japan) at the SUT-NANOTEC-SLRI Joint Research Facility, Synchrotron Light Research Institute (SLRI), Nakhon Ratchasima, Thailand. A monochromated Al K_α X-ray source (hν = 1486.6 eV) was used for analysis. Survey and high-resolution spectra were collected with pass energies of 117.4 and 46.95 eV, respectively. Peak fitting was carried out using CasaXPS software Version 2.3.26 (Casa Software Ltd., Teignmouth, Devon, UK) with Gaussian–Lorentzian functions and a Shirley-type background was applied. All peaks were adjusted to adventitious graphitic carbon at 284.8 eV. The UV–visible diffuse reflectance (DRS) spectra were acquired in the 190–1200 nm range utilizing a Shimadzu UV-2600i UV–visible spectrophotometer (Shimadzu Corporation, Kyoto, Japan), with BaSO₄ as the reference standard. A reflectance mode with a 0.5 nm data interval and moderate scan speed was employed during the measurement. The resultant spectra yielded significant insights into the electronic transitions and band gap energy of the materials. The zeta potentials were measured using a Zetasizer ZS nanoparticle size and zeta potential analyzer (Malvern Panalytical, Malvern, Worcestershire, UK) based on the electrophoretic light scattering (ELS) technique. Prior to measurement, the photocatalyst powders were dispersed in Rh B and MO aqueous solutions at the same concentration employed in the photocatalytic adsorption experiments and ultrasonicated for 10 min to obtain homogeneous suspensions.

2.1.3. Photoelectrochemical Measurements

Stock solutions of each sample were made by dispersing 10 mg of powder in 1 mL of water, 1 mL of ethanol, and 1 mg of Nafion. Each stock solution was sonicated for 10 min to increase dispersion (VWR ultrasonic cleaner, 30 W, 45 kHz). Fluorine-doped tin oxide (FTO; TEC 15) substrates (1 × 2 cm in size) were heated to 100 °C on a hotplate (VWR VMS-C7). For each sample, 80 μL of stock solution was cast onto the heated FTO substrate using a pipette. This was done in 20 μL aliquots, waiting ~5 min for the solvent to evaporate before casting the next aliquot.

Photoelectrochemical analyses were carried out in 0.1 M phosphate buffer (0.05 M K₂HPO₄, 0.05 M KH₂PO₄; pH = 7) in Milli-Q-water (Millipore Corp., Bedford, MA, USA, 18.2 MΩ cm at 25 °C). For all measurements, a three-electrode configuration was used, with a Pt mesh counter electrode and a Ag/AgCl reference electrode in sat. KCl(aq.) (0.197 V_NHE at 25 °C; Metrohm AG, Herisau, Switzerland). All applied voltages are reported vs. the reversible hydrogen electrode (V_RHE), converted using the Nernst equation:

V_RHE = V_Ag/AgCl + (0.0591 × pH) + V^∅_Ag/AgCl

(1)

where V_Ag/AgCl is the applied potential vs. the reference electrode and V^∅_Ag/AgCl is the standard potential of the reference electrode. Current-voltage (jV) and chronoamperometry (CA) experiments were conducted in a custom-built PEEK cell with quartz windows with an Autolab potentiostat (PGSTAT 101, Metrohm Autolab B.V., Utrecht, The Netherlands) used for these measurements. Light of 350 nm (3.5 mW.cm⁻²) was provided by a 75 W Xe lamp coupled to a monochromator (OBB-2001, Photon Technology International, Birmingham, NJ, USA). Light power was measured using a Thorlabs optical power meter with photodiode sensor (PM100D with S120VC). Flat-band potentials were determined from chopped-light jV curves, with experiments conducted in a Redox.me Teflon Photo-electrochemical H-Cell with an Autolab PGStat302N potentiostat (Metrohm). The primary light source used for photoelectrochemical measurements was a Class A Abet Technologies Sun 2000 Solar Simulator (ABET Technologies, Inc., Milford, CT, USA) with a 550 W Xe lamp and AM 1.5 G filter. A variable working distance was used to provide an equivalent incident light intensity to the reactor of one sun (100 mW cm⁻²). The light incident on the electrochemical cells was characterised using StellarNet Black Comet UV/Vis (280–900 nm) and Dwarf Star (900–1700 nm) spectrophotometers, interfaced with SpectraWiz software, Version 6.32 (StellarNet, Inc., Tampa, FL, USA).

The flat band potential was determined by chopped linear sweep voltammetry (chopping frequency = 1 Hz, scan rate = 5 mV.s⁻¹) at the point of photocurrent reversal [19]. For flat band potential measurements, samples were irradiated by a Sun2000 solar simulator, equipped with an AM 1.5G filter (100 mW.cm⁻²). The electrode potential was controlled using an Autolab PGStat302N potentiostat (Metrohm), interfaced with Nova 2.1.8.

2.1.4. Photocatalytic Test

Rh B and MO were employed as model organic pollutants to evaluate the photocatalytic activity of the synthesized materials. Photocatalytic experiments were conducted in a temperature-controlled double-walled quartz reactor (7 × 9.5 cm) connected to a circulating water system. In each run, 50 mg of the catalyst was introduced into 50 mL of an aqueous solution containing 10 ppm of the target pollutant. Prior to irradiation, the suspension was first stirred in the dark for 30 min for Rh B and 60 min for MO to achieve adsorption–desorption equilibrium. UV–visible-light irradiation was supplied by a 300 W xenon lamp (PLS-SXE300D/300DUV, PerfectLight, Beijing, China), positioned at a fixed distance of 7.5 cm from the reactor and 11.5 cm above the solution surface. The reaction temperature was maintained at 25 °C. The same irradiation configuration was maintained for all photocatalytic experiments. During the reaction, 5 mL of the suspension was periodically withdrawn and centrifuged to separate the solid catalyst. The concentrations of Rh B and MO were subsequently determined using a UV–Vis spectrophotometer (Mettler Toledo UV5Bio, Greifensee, Switzerland) by monitoring their characteristic absorption maxima at 554 and 464 nm, respectively.

To probe the active species responsible for photocatalysis, radical scavenger tests were performed under identical experimental conditions. Different quenching agents were introduced, including EDTA-2Na (1 mM) for holes (h⁺), isopropyl alcohol (1 mL) for hydroxyl radicals (^•OH), and benzoquinone (1 mM) for superoxide radicals (O₂^•−) [20,21].

3. Results and Discussion

3.1. Crystal Structure and Composition

Figure 1a presents the XRD patterns of ABNO, GO and GABNO samples. The diffraction peaks of ABNO can be well indexed to the Aurivillius-type layered perovskite structure (PDF No. 42-0397), confirming the successful formation of phase-pure Na_0.5Bi_2.5Nb₂O₉, as similarly reported in the literature [8]. The GO sample exhibits a broad diffraction peak at around 12°, which is attributed to the (001) plane of GO, corresponding to the enlarged interlayer spacing induced by oxygen-containing functional groups [22]. In addition, a weak and broad diffraction feature observed at around 43–44° can be assigned to the (100) and (101) planes of graphitic carbon, which is associated with the in-plane structure of the sp² carbon network [22]. The broad nature and low intensity of this peak reflect the partially disordered structure of GO. After the incorporation of GO into ABNO, all GABNO composite samples exhibit diffraction patterns similar to those of ABNO. This suggests that the crystal structure of ABNO is well preserved during composite formation.

Figure 1b shows the Raman spectra of ABNO and GABNO composites. The ABNO exhibits several characteristic vibrational modes in the range of 50–950 cm⁻¹ [23], which are associated with the lattice vibrations of Nb–O bonds within the perovskite slabs and Bi–O layers of the Aurivillius structure. These modes confirm the layered structural nature of ABNO. Figure S1 shows the Raman spectrum of GO, in which the typical D band (~1350 cm⁻¹) and G band (~1580 cm⁻¹) are observed, which are characteristic of graphene oxide [18]. Nevertheless, the incorporation of GO induces subtle changes in the Raman features of ABNO. Slight variations in peak intensity and band broadening are observed in the composite samples compared to ABNO. Such changes suggest perturbations in the local bonding environments and lattice vibrations, possibly arising from the interaction between ABNO and GO. Notably, no new Raman modes corresponding to secondary phases are observed, further confirming the structural integrity of ABNO in the composite system with GO.

The TEM, HRTEM and SAED patterns of GO and ABNO were investigated. Figure 2a illustrates that the GO sample exhibits typical thin and wrinkled sheet-like morphology, indicating its two-dimensional structure [11,12,24]. The corresponding SAED pattern shows diffuse ring features, which are characteristic of the partially ordered graphitic structure of GO. For ABNO (Figure 2b), aggregated nanoparticles with well-defined crystalline domains are observed. The HRTEM image reveals clear lattice fringes with an interplanar spacing of approximately 0.35 nm, which can be indexed to the (113) plane of Na_0.5Bi_2.5Nb₂O₉, confirming its high crystallinity [8]. The corresponding SAED pattern exhibits distinct diffraction rings that can be indexed to the crystallographic planes of ABNO, indicating its polycrystalline nature [8].

To investigate the formation of the composites, the TEM image of the GABNO 0.1 sample (Figure 2c) exhibits a heterogeneous structure consisting of aggregated particles anchored on thin and wrinkled sheet-like features. This observation confirms the successful integration of ABNO with GO, forming a composite structure. Notably, the intimate contact between the ABNO particles and the GO sheets can be clearly observed, indicating strong interfacial coupling at the nanoscale. A high-resolution TEM image (Figure 2c, inset) shows that the aggregated particles exhibit well-resolved lattice fringes with a d₁₁₃ spacing of 0.35 nm, corresponding to ABNO. The corresponding selected area electron diffraction (SAED) pattern (Figure 2d) exhibits distinct diffraction spots of ABNO, indicating its single-crystalline nature. In addition, the observed diffuse ring-like features correspond to graphitic GO. The coexistence of spot and ring patterns further confirms the formation of a composite system composed of crystalline ABNO and graphitic GO. This observation is further supported by the EDS spectrum (Figure 2e), in which the characteristic peaks of Na, Bi, Nb, O and C are clearly detected, confirming the coexistence of ABNO and GO in the composite system. The corresponding elemental mapping images (Figure 2f) reveal the homogeneous distribution of these elements throughout the composite particles, further confirming the presence of the ABNO phase with the uniformly distributed GO. This observation suggests the successful integration of ABNO with GO, even at a very low loading.

To further elucidate the formation of the composite with higher GO loading, Figure S2 shows that the particle morphology gradually evolves with increasing GO content. As discussed earlier, in the GABNO 0.1 sample, ABNO nanoparticles are well dispersed on the GO sheets, forming an interfacial composite structure while preserving the crystallinity of ABNO. The corresponding SAED pattern still shows distinct diffraction spots, indicating that the single-crystalline nature of ABNO is largely retained at low GO loading. With increasing GO content (GABNO 0.5 and GABNO 1), ABNO particles remain attached to the GO sheets; however, the dispersion becomes less uniform, and partial aggregation of ABNO nanoparticles can be observed. In these samples, the SAED patterns exhibit a combination of diffraction spots and faint ring-like features, suggesting the coexistence of crystalline ABNO domains and disordered graphitic structures from GO. Notably, at the highest GO loading (GABNO 5), the morphology shows more pronounced aggregation of ABNO nanoparticles on the GO surface. The corresponding SAED pattern is dominated by ring-like features rather than discrete spots, indicating a transition toward a polycrystalline-like structure. This change suggests that excessive GO content may disrupt the long-range ordering or preferred orientation of ABNO domains, leading to a reduction in apparent crystallinity.

To further verify the elemental composition and spatial distribution within the composite system, TEM-EDS mapping and spectral analyses were performed, as shown in Figure S3 and Figure S4, respectively. In Figure S3a, the GO sample exhibits a dominant C signal with a uniform distribution across the sheet-like structure, accompanied by a relatively weaker O signal, confirming the presence of oxygen-containing functional groups. For ABNO (Figure S3b), elemental mapping reveals homogeneous distributions of Na, Bi, Nb and O throughout the particles, indicating the formation of a uniform oxide phase without elemental segregation. In the composite samples (Figure S3c–f), the coexistence of C, O, Na, Bi and Nb signals is clearly observed, confirming the successful integration of GO with ABNO. The C signal is mainly distributed along the sheet-like regions, while Na, Bi and Nb are predominantly localized within the particle domains, suggesting that ABNO nanoparticles are anchored onto the GO sheets. Notably, the overlapping distribution of O across both regions further supports the intimate contact between GO and ABNO. With increasing GO content, the intensity and spatial coverage of the C signal become more pronounced, indicating the gradual increase in GO within the composite. Despite this, the distributions of Na, Bi and Nb remain relatively uniform, suggesting that the incorporation of GO does not disrupt the compositional homogeneity of ABNO. The corresponding EDS spectra (Figure S4) further confirm the presence of all constituent elements in the composite samples, with characteristic peaks of Bi, Nb, Na, O and C clearly detected. These results are consistent with the TEM observations and demonstrate the successful formation of GO/ABNO composites with well-dispersed components.

To determine elemental composition at the bulk scale, SEM-EDS mapping and spectral analyses were performed for all samples, as shown in Figure S5. Consistent with the TEM-EDS mapping results, a uniform distribution of constituent elements is observed throughout the particles of each sample, indicating a homogeneous bulk composition. The GO sample shows a uniform distribution of C and O across the sheet-like structure, while the elemental maps of the ABNO sample confirm the presence of Na, Bi, Nb and O with a homogeneous distribution throughout the particles. All constituent elements (C, O, Na, Bi and Nb) are consistently detected, confirming the successful incorporation of GO within the ABNO matrix in the composite samples, with no indication of phase segregation or compositional inhomogeneity. Importantly, the increasing intensity of the C signal with higher GO loading further supports the systematic incorporation of GO into the composite. The corresponding EDS spectra corroborate these observations, showing characteristic peaks of all expected elements in each sample. Although minor variations in atomic percentages are observed, they fall within the expected range for semi-quantitative EDS analysis.

3.2. Surface Chemical Composition and Oxidation States

The surface chemical composition and oxidation states of GO, ABNO and GABNO composites were investigated by XPS analysis, as shown in Figure 3. The survey spectra (Figure S6a) confirm the presence of Na, Bi, Nb, O and C elements in the composite samples, indicating the successful incorporation of GO into the ABNO matrix. No additional impurity peaks are detected, suggesting the high purity of the synthesized materials. The high-resolution C 1s spectrum of GO (Figure S6b) can be deconvoluted into three main components assigned to C–C/C=C (284.8 eV), C–O (286.81 eV), and C=O (288.67 eV), corresponding to characteristic carbon-bonding environments in graphene oxide (Table S1) [11,12,13].

The high-resolution O 1s spectra provide further insight into the evolution of surface oxygen species from GO to ABNO and the corresponding GABNO composites, as shown in Figure 3a. For GO, the O 1s spectrum can be deconvoluted into two components associated with oxygen-containing functional groups, including O–C and O=C species, reflecting the oxidized nature of graphene oxide [18]. In contrast, ABNO exhibits an O 1s spectrum that can be deconvoluted into three components associated with lattice oxygen, O–C and oxygen in surface hydroxyl or adsorbed oxygen species (O-ads) [25]. In the composite sample, noticeable shifts of these oxygen species toward higher binding energies are observed in GABNO 0.1 as well as in the other GABNO samples (Table S1). This provides further insight into the modification of the surface oxygen environment upon GO incorporation. Bi 4f spectra (Figure 3b) display two characteristic peaks corresponding to Bi 4f_7/2 and Bi 4f_5/2, confirming the presence of Bi³⁺ in all samples [26]. Similarly, the Nb 3d spectra (Figure 3c) show two peaks assigned to Nb 3d_5/2 and Nb 3d_3/2, indicating the Nb⁵⁺ oxidation state [8]. The Na 1s spectra (Figure 3d) further confirm the presence of Na⁺ in the ABNO structure of all samples [8]. Notably, slight positive shifts in binding energies are observed for Bi 4f, Nb 3d and Na 1s in the composite samples, particularly in GABNO 0.1, compared to ABNO. These positive shifts suggest a collective redistribution of electron density across the constituent elements in the composite system, which is attributed to interfacial electronic interactions between GO and ABNO, consistent with previous reports on GO- or rGO or graphene-based oxide heterostructures [27,28].

3.3. Morphology and Surface Properties

The surface morphology of GO, ABNO and GABNO composite samples was investigated by SEM, as illustrated in Figure 4. GO (Figure 4a) exhibits a typical wrinkled and sheet-like morphology with thin and stacked layers, which is a characteristic of graphene oxide [11,18]. In contrast, ABNO (Figure 4b) displays a plate-like structure composed of irregularly stacked nanoflakes with relatively smooth surfaces, consistent with its layered Aurivillius-type structure, which agrees well with observations in other studies that have synthesized Na_0.5Bi_2.5Nb₂O₉ utilizing a hydrothermal method [8].

Upon incorporation of GO, significant changes in morphology are observed in the GABNO samples. The composite particles form a loosely packed plate-like structure, with small irregular nanoparticles homogeneously distributed on the surfaces (Figure 4c–f). However, the overall particle morphology of the GABNO samples does not show significant variation with different GO loading contents, suggesting that the fundamental structure of ABNO is largely preserved after composite formation. As shown in Figure S5, the homogeneous distribution of constituent elements further confirms the uniform dispersion of GO and ABNO across the probed areas of all GABNO samples. These observations indicate the formation of a well-interconnected composite structure, in which ABNO particles are uniformly distributed and closely attached to the GO sheets, providing intimate contact between the two components.

Nitrogen adsorption–desorption analysis is shown in Figure 5. As shown in Figure 5a, the GO sample exhibits a typical type IV isotherm with an H3-type hysteresis loop, along with a relatively high specific surface area of 47.91 m²/g. The relatively low adsorption uptake over most of the relative pressure range, followed by a sharp increase near P/P₀ ≈ 1, suggests that the mesoporosity is weakly developed and primarily originates from interlayer voids formed by the stacking 2D structure of GO sheets rather than well-defined intrinsic pores. The pore size distribution of the GO sample (Figure 5a, inset) shows a very low contribution in the mesoporous range (below 50 nm), while a significant distribution is observed at larger pore sizes (>100 nm). This indicates that the pore structure is dominated by macroporous voids rather than well-defined mesopores. These large pores are attributed to interlayer and interparticle spaces formed by the restacking and wrinkling of GO sheets, consistent with its slit-like structural nature. For ABNO (Figure 5b), a similar type IV isotherm is observed; however, the hysteresis loop is weak and not well-defined, indicating poorly developed mesoporosity. The corresponding pore size distribution (Figure 5b, inset) shows a relatively low contribution in the mesoporous range (<50 nm), while a broader distribution extending toward larger pore sizes is observed. This suggests that the pore structure of ABNO is predominantly governed by interparticle voids rather than well-defined intrinsic mesopores, which is consistent with the aggregation of plate-like particles observed in SEM and TEM analyses. Compared to GO, ABNO exhibits a lower specific BET surface area (31.35 m²/g), reflecting its more compact structure.

For the GABNO composite samples (Figure 5c–f), the isotherms retain a type IV-like profile with similarly weak and poorly defined hysteresis loops, indicating that the incorporation of GO does not introduce well-developed intrinsic mesoporosity. The pore size distributions show only minor contributions in the mesoporous range (<50 nm), while a significant portion of the pore volume is distributed at larger pore sizes (>100 nm). This behavior suggests that the textural properties of the composites are still dominated by interparticle and interlayer voids, arising from the assembly of plate-like ABNO particles and stacked GO sheets. With increasing GO content, slight variations in the pore size distribution can be observed, which may be associated with changes in particle packing and the degree of sheet restacking. Overall, the introduction of GO primarily modifies the structural arrangement rather than creating new intrinsic pore structures. Notably, the BET surface areas of the composites (18.65–29.36 m²/g) decrease with increasing GO content, which can be attributed to the partial coverage of ABNO surfaces by GO sheets and increased aggregation, leading to reduced accessible surface area.

3.4. Optical Properties

The optical properties of GO, ABNO and GABNO composites were investigated using UV–vis diffuse reflectance spectroscopy (DRS) and photoluminescence (PL) analyses (Figure 6). As shown in Figure 6a, all GABNO samples exhibit a similar absorption edge in the UV region, indicating that the incorporation of GO does not significantly alter the fundamental light absorption behavior of ABNO. The estimated band gap energies, derived from Tauc plots, are approximately 3.00 eV for ABNO and remain nearly unchanged for GABNO 0.1 and GABNO 0.5, while slightly decreasing to ~2.90 eV for GABNO 1 and GABNO 5. This marginal variation suggests that GO incorporation does not introduce substantial electronic band structure modification, and the composites retain the intrinsic wide band gap nature of ABNO. Despite the minimal change in band gap energy, a noticeable difference in absorption intensity can be observed, particularly at higher photon energies, which may be attributed to the presence of GO and its light-harvesting contribution.

The PL spectra (Figure 6b) exhibit the emission peak centered around ~500 nm, which is associated with band-to-band recombination and defect-related transitions [29]. All GABNO samples exhibit emission profiles similar to that of ABNO, with no significant quenching observed in most samples. However, GABNO 5, with the highest GO loading, exhibits a slight decrease in PL intensity.

3.5. Photocatalytic Properties

To comprehensively evaluate the photocatalytic performance of the samples, the removal of two representative organic dyes, Rh B and MO, was investigated under UV–visible-light irradiation. Rh B and MO were selected because they possess different physicochemical properties, particularly their molecular charge. Rh B is a cationic dye, whereas MO is an anionic dye. These differences strongly influence their adsorption behaviour on the photocatalyst surface and the predominant degradation pathway, including surface-mediated and solution-phase photocatalytic processes.

3.5.1. Photocatalytic Rh B Removal

The photocatalytic activities of GO, ABNO and GABNO composites were evaluated through the degradation of Rh B under light irradiation (Figure 7). Changes in Rh B concentration are represented by the ratio C/C₀, where C₀ and C correspond to the initial and instantaneous concentrations, respectively. As shown in Figure 7a, most samples exhibited negligible Rh B removal during the dark adsorption stage, indicating that adsorption plays a minor role. In contrast, the GO sample showed approximately 40% dye adsorption, which is well known for GO materials.

Upon light irradiation for 2 h, ABNO achieved moderate removal efficiency (81%), while GO showed limited activity (25%), confirming that GO alone is not an effective photocatalyst. In the absence of a photocatalyst, Rh B removal was negligible, indicating that photolysis does not contribute to changes in Rh B concentration. In contrast, the GABNO composite exhibited significantly enhanced photocatalytic performance, with the GABNO 0.1 sample showing the highest activity, and achieving nearly complete removal (~100%) within 120 min. The evolution of the UV–vis absorption spectra of the Rh B solution during the photocatalytic process was monitored to further evaluate the degradation behaviour. As shown in Figure S7a, the characteristic absorption peak of Rh B at ~554 nm gradually decreases in intensity with increasing irradiation time, accompanied by a slight blue shift, indicating the stepwise degradation of the chromophoric structure. This behaviour suggests that the photocatalytic process involves not only decolorization but also the breakdown of the conjugated structure of Rh B molecules [8]. Further increasing the GO content leads to a gradual decline in performance (GABNO 0.5 > GABNO 1 > GABNO 5), suggesting that an optimal GO loading is critical for maximizing photocatalytic efficiency. Excessive GO may shield active sites or hinder light absorption, resulting in reduced activity. This effect is particularly evident in GABNO 5, where the performance decreases to below that of ABNO. The reaction kinetics were further analyzed using a pseudo-first-order model (Figure S7b), and the apparent rate constant (k_app) was summarized in Figure 7b. The k_app value of ABNO is higher than that of GO and is significantly enhanced upon GO incorporation, with GABNO 0.1 exhibiting the highest value, approximately 2.5 times greater than that of ABNO.

The reusability and stability of the best photocatalyst, GABNO 0.1, were evaluated over three consecutive cycles (Figure 7c). The photocatalytic activity of the composite improved, reaching 100% removal within 60 min, and was further enhanced in the third cycle, achieving complete removal within 30 min. The evolution of the UV–vis spectra of the Rh B solution during the second and third cycles is illustrated in Figure S7c and Figure S7d, respectively. This gradual enhancement in photocatalytic activity may be associated with changes at the catalyst surface during repeated irradiation, such as surface cleaning or activation that exposes additional active sites [30]. However, the exact origin of this enhancement remains unclear, and possible irradiation-induced modification of the GO component cannot be conclusively excluded based on the present data. Therefore, the increase in activity is not attributed to improved bulk charge separation. Instead, it may reflect changes in surface accessibility and interfacial electronic interactions during repeated use. This overall structural stability is consistent with the preserved structural, chemical and physical integrity, as confirmed by post-reaction characterizations (Figure S8), including XRD analysis (Figure S8a), morphological observations (Figure S8b), and EDS elemental mapping, all of which demonstrate the stability of the GABNO 0.1 composite (Figure S8c). These results indicate that the GABNO 0.1 composite retained its overall structural integrity after repeated photocatalytic use.

To further elucidate the photocatalytic mechanism of Rh B degradation, radical scavenging experiments were conducted (Figure 7d). The addition of benzoquinone (BQ), a superoxide radical (O₂^•−) scavenger, resulted in the most significant inhibition (~87%), indicating that O₂^•− is the dominant reactive species. Moreover, the addition of EDTA-2Na (hole scavenger) led to lower inhibition (~63%), suggesting that holes play a secondary role. Meanwhile, the presence of isopropanol (IPA), a hydroxyl radical (^•OH) scavenger, showed a negligible effect, indicating that ^•OH is not a primary species in the Rh B degradation process. Jiamprasertboon et al. [8] previously reported that holes are the dominant active species responsible for the photocatalytic activity of ABNO. In contrast, the present results demonstrate that GO incorporation induces a clear shift in the photocatalytic pathway, from a hole-dominated oxidation mechanism to a superoxide radical-mediated process in the GABNO composites. This transition highlights the role of GO as an electron mediator, facilitating the transfer and utilization of photogenerated electrons for oxygen reduction, thereby enhancing the overall photocatalytic efficiency.

3.5.2. Photocatalytic MO Removal

The photocatalytic performance of the prepared photocatalysts was further evaluated using MO as a model pollutant (Figure 8a). Prior to light irradiation, all samples exhibited negligible changes in concentration during the dark adsorption stage, indicating limited adsorption capacity toward MO. Upon light irradiation, the GABNO composites showed enhanced photocatalytic activity compared to GO or ABNO, demonstrating the beneficial effect of GO incorporation. Among all samples, GABNO 0.5 exhibited the highest removal efficiency, while higher GO loading (GABNO 1 and GABNO 5) led to a noticeable decline in activity.

Figure 8b shows the k_app values derived from pseudo-first-order kinetic plots (Figure S9a). The results confirm that GABNO 0.5 possesses the highest removal rate, followed by GABNO 0.1 and GABNO 1, whereas GO shows negligible activity. The highest GO loading (GABNO 5) exhibited the worst catalytic activity compared to ABNO, indicating that excessive GO loading adversely affects performance.

The reusability of the optimal sample (GABNO 0.5) was evaluated over three consecutive cycles (Figure 8c). Interestingly, the photocatalytic activity increased in the subsequent cycles, showing faster removal rates compared to the first cycle, consistent with the trend observed in Rh B removal. This behavior might be associated with a surface activation effect, where repeated irradiation removes surface contaminants and exposes additional active sites, thereby improving photocatalytic performance [30,31,32,33]. The UV–vis absorption spectra of MO during the removal process over GABNO 0.5 in the third cycle (Figure 8d) show a gradual decrease in the characteristic absorption peak at ~464 nm, accompanied by a noticeable blue shift. This behaviour indicates not only a reduction in dye concentration but also the progressive breakdown of the conjugated structure. The observed shift suggests cleavage of the azo (−N=N−) bond and the formation of intermediate species with shorter conjugation lengths, confirming effective photocatalytic removal of MO molecules [34,35]. We note that the changes in the characteristic absorption peak are minimal in the first cycle (Figure S9b), more apparent in the second cycle (Figure S9c) and clearly pronounced in the third cycle. This progressive behaviour correlates with the increasing reaction rates and k_app values, confirming the enhanced photocatalytic performance upon cycling.

To further understand the role of surface electrostatic interactions in dye adsorption, the zeta potentials of GO, ABNO, and GABNO samples dispersed in Rh B and MO solutions were measured. The zeta potential values are summarized in Table S2. All samples exhibited negative zeta potentials in both dye solutions, indicating negatively charged catalyst surfaces under the photocatalytic conditions. In Rh B solution, the zeta potentials varied slightly (−25.2 to −31.8 mV), suggesting comparable electrostatic attraction toward the cationic Rh B molecules. In contrast, more negative zeta potentials were observed for all GABNO composites in MO solution (−26.3 to −28.9 mV) compared with GO (−13.5 mV) and ABNO (−17.8 mV). The increased negative surface charge is expected to strengthen electrostatic repulsion toward anionic MO molecules, partially limiting their adsorption on the catalyst surface [36]. This result is consistent with the lower photocatalytic removal efficiency of MO than Rh B, indicating that electrostatic interactions influence dye adsorption.

3.6. Charge Carrier Properties

To further elucidate the origin of the enhanced photocatalytic performance, the charge separation efficiency and interfacial electronic charge transfer were systematically investigated using photoelectrochemical (PEC) analyses.

The photoelectrochemical properties of the samples were evaluated by linear sweep voltammetry (j-V) and chronoamperometry (CA) measurements under back-side irradiation (350 nm, ~3.5 mW.cm⁻²) in a phosphate buffer solution (pH 7), as shown in Figure 9. The j-V curves (Figure 9a) reveal that all samples exhibit anodic photocurrent under illumination, indicating typical n-type photoanode behavior capable of driving photoelectrochemical water oxidation. The incorporation of GO leads to an increase in the dark current background, particularly at higher GO loadings, while the photocurrent response shows only moderate variation among the samples. Notably, the GABNO series does not exhibit a proportional increase in photocurrent with increasing GO content.

The transient photocurrent responses obtained from CA measurements at 1.23 V vs. RHE (Figure 9b) provide insight into charge separation efficiency. All samples display stable and reproducible photocurrent under chopped illumination, confirming their photo-responsiveness. The photocurrent density of GO is negligible (~0.02 μA.cm⁻²), while ABNO shows a higher value of 0.29 μA.cm⁻². The GABNO composites exhibit slightly enhanced photocurrent densities (GABNO 0.1: 0.31 μA.cm⁻²; GABNO 0.5: 0.34 μA.cm⁻²; GABNO 1: 0.32 μA.cm⁻²; GABNO 5: 0.31 μA.cm⁻²). Despite the relatively low photocurrent densities (corresponding to an IPCE on the order of ~0.03%, typical for powder-based photoelectrodes), the results demonstrate that controlled incorporation of GO can improve charge carrier separation and photoelectrochemical performance, with an optimal composition observed for GABNO 0.5.

To gain further insight into the interfacial electronic structure of the GO/ABNO heterointerface, the flat-band potentials (E_fb) of all samples were measured. Owing to the highly insulating nature and large dielectric constant of the ABNO-based materials, the flat-band potentials were determined using the photocurrent reversal method under chopped illumination, following the procedure proposed by Hankin et al. [19]. The photocurrent reversal curves used for determining the flat-band potentials are shown in Figure S10. The obtained flat-band potentials were determined to be 0.745 ± 0.050 V vs. RHE for GO, 0.557 ± 0.050 V for ABNO, 0.562 ± 0.050 V for GABNO 0.1, 0.638 ± 0.050 V for GABNO 0.5, 0.537 ± 0.050 V for GABNO 1, and 0.596 ± 0.050 V for GABNO 5. GO exhibits the highest flat-band potential, whereas ABNO and the GABNO composites show values within a relatively narrow range. These results imply that the incorporation of GO into ABNO does not cause the flat-band potential to change significantly.

4. Discussion

The present results collectively demonstrate that the enhanced photocatalytic performance of the GABNO composites cannot be attributed to a single factor but rather originates from the synergistic effects of structural characteristics, physicochemical properties, heterointerface formation, interfacial electronic interactions, and optimized charge-transfer processes.

Structural characterizations by XRD and Raman spectroscopy confirm that the crystal structure of ABNO is preserved after GO incorporation, indicating that GO is integrated without altering the intrinsic Aurivillius framework. TEM, SAED, elemental mapping, and XPS further verify the formation of intimate GO/ABNO heterointerfaces, where ABNO nanoparticles are uniformly anchored onto GO sheets, providing close interfacial contact for electronic communication.

Nitrogen adsorption–desorption measurements further reveal that GO incorporation primarily modifies the structural arrangement and interfacial contact rather than generating additional intrinsic porosity. Collectively, these results indicate that an appropriate amount of GO maximizes the interfacial contact area without significantly disturbing the crystal framework, whereas excessive GO promotes particle aggregation and partially reduces surface accessibility. This interpretation is further supported by the normalized photocatalytic rate constants (k_norm), which account for the influence of specific surface area [37]. For Rh B removal, the k_norm increases from 1.95 × 10⁻⁴ g.m⁻².min⁻¹ for ABNO to 6.75 × 10⁻⁴ g.m⁻².min⁻¹ for GABNO 0.1, representing an approximately 3.5-fold enhancement for the composite. A similar trend is observed for MO removal, where GABNO 0.5 exhibits the highest normalized activity (1.96 × 10⁻⁴ g.m⁻².min⁻¹), approximately 3.4 times higher than that of ABNO (5.84 × 10⁻⁵ g.m⁻².min⁻¹). These results further confirm that the enhanced photocatalytic activity cannot be simply explained by surface area effects but is predominantly governed by the nature of the GO/ABNO heterointerface.

The physicochemical characterizations further suggest that the improvement in photocatalytic performance originates primarily from interfacial electronic regulation rather than changes in the intrinsic optical properties. Although GO incorporation results in only negligible changes in the optical absorption edge, band-gap energy and photoluminescence intensity, a contribution from GO-mediated light harvesting or photosensitization cannot be completely excluded. However, the negligible changes in the absorption edge and band-gap energy, together with the absence of pronounced PL quenching, do not support optical sensitization or enhanced bulk charge separation as the dominant factors governing the improved photocatalytic performance. The XPS spectra exhibit systematic positive shifts of the O 1s, Bi 4f, Nb 3d and Na 1s core levels, indicating electron redistribution at the GO/ABNO interface. Such electronic redistribution is further supported by the measurable shifts in flat-band potential after GO incorporation, suggesting Fermi-level equilibration between GO and ABNO upon interfacial contact. Together, these results provide stronger evidence for modification of the interfacial electronic environment following GO/ABNO contact. These observations demonstrate that GO effectively perturbs the interfacial electronic environment. Instead, the available evidence suggests that GO primarily facilitates interfacial charge transfer and more efficient utilization of photogenerated electrons during surface redox reactions, highlighting the significance of heterointerface construction in enhancing photocatalytic performance [9].

These structure–property relationships are well reflected in the photocatalytic performance. The optimum GO loading significantly enhances the removal of both Rh B and MO, whereas excessive GO leads to a gradual decline in activity because of increased aggregation and partial shielding of active sites. The zeta-potential measurements reveal that all samples possess negatively charged surfaces, favoring adsorption of the cationic Rh B while partially suppressing adsorption of anionic MO. Nevertheless, the superior activity of the GABNO composites toward both pollutants indicates that adsorption is not the sole determining factor. More importantly, radical trapping experiments demonstrate that GO incorporation changes the dominant active species from hole-dominated oxidation in ABNO to a superoxide-radical-mediated pathway in the composites, confirming more efficient electron utilization after interfacial contact with GO. Therefore, enhanced photocatalytic performance is attributed to the synergistic effects of well-defined GO/ABNO heterointerfaces, interfacial electronic regulation, and optimized interfacial charge-transfer behavior, rather than to modifications of the intrinsic crystal structure, optical properties, or surface area of ABNO. The proposed photocatalytic mechanism is illustrated in Figure 10, for which AI-assisted image generation (Google Gemini, Gemini 1.5 Pro/Imagen 3, Google) was used solely to prepare the schematic illustration, whereas all scientific concepts, mechanistic interpretation, verification and final editing were performed by the authors.

5. Conclusions

GABNO heterostructured composites with different GO loadings were successfully synthesized via a hydrothermal method. Structural characterizations confirmed that GO was intimately integrated with ABNO while preserving the intrinsic Aurivillius-type crystal structure, leading to the formation of well-defined heterointerfaces. Although GO incorporation caused negligible changes in the crystal structure, band-gap energy and photoluminescence behavior, XPS and flat-band potential analyses revealed measurable electronic perturbations at the GO/ABNO interface, indicating electron redistribution and Fermi-level alignment at the interface. Photoelectrochemical measurements further demonstrated that GO primarily facilitates interfacial charge transfer rather than substantially enhancing bulk charge separation.

Among all compositions, GABNO 0.1 (0.1 mg GO loading) exhibited the highest photocatalytic activity toward Rh B removal, while GABNO 0.5 (0.5 mg GO loading) showed the best performance for MO removal. The normalized reaction rate constants confirmed that enhanced photocatalytic activity cannot be explained solely by differences in specific surface area, highlighting the dominant contribution of interfacial electronic interactions. Radical trapping experiments further revealed that GO incorporation modulates the photocatalytic reaction pathway by promoting superoxide-radical-mediated oxidation through more efficient utilization of photogenerated electrons.

Overall, this work demonstrates that GO functions as an interfacial electronic regulator rather than merely a conductive additive, where the intimate GO/ABNO heterointerface governs electron redistribution, interfacial charge transfer and photocatalytic reaction pathways. These findings provide new insight into the rational design of carbon-modified Aurivillius-layered perovskite photocatalysts and establish interfacial electronic regulation as an effective strategy for developing high-performance photocatalysts for environmental remediation.