Green AgNPs/TiO2 photocatalyst degrades azo dye using white LED light

In a development that blends kitchen-chemistry simplicity with cutting-edge computation, researchers have crafted a photocatalyst from rosemary extract and titanium dioxide that can completely destroy a stubborn textile dye under nothing more exotic than a white LED lamp. The study, published in Environmental Science and Pollution Research, describes a “green” silver-decorated titanium dioxide nanocomposite—green AgNPs/TiO2—synthesized using phytochemicals derived from Rosmarinus officinalis, the common rosemary plant. The work, led by Ahmed Halfadji of Ibn Khaldoun University of Tiaret in Algeria and Emil Obeid of the American University of the Middle East in Kuwait, combines laboratory experiments, density functional theory (DFT) calculations, and a machine learning-guided optimization scheme to tackle one of the most persistent categories of water pollutants: azo dyes.

The target molecule in the study was Acid Orange 10 (AO10), an anionic azo dye widely used in the textile industry and a representative of the broader class of synthetic colorants that resist conventional wastewater treatment. Azo dyes owe their vivid colors to the azo linkage—a nitrogen–nitrogen double bond—that also makes them chemically stable and, in many cases, toxic or recalcitrant in the environment. Textile effluents carrying such dyes reduce light penetration in receiving waters, disrupt aquatic photosynthesis, and pose health risks to humans and wildlife. Traditional treatment approaches, from adsorption onto activated carbon to biological degradation, often fall short because the dyes are designed from the ground up to resist fading, washing, and microbial attack.

The Algerian–Kuwaiti team’s answer was to build a photocatalyst that attacks these molecules at the molecular level using light. Titanium dioxide has long been the workhorse of photocatalysis: when photons with sufficient energy strike its surface, they excite electrons from the valence band to the conduction band, leaving behind holes. These electron–hole pairs migrate to the surface and trigger redox reactions with water and dissolved oxygen, generating reactive oxygen species—chiefly hydroxyl radicals and superoxide anions—that can oxidize organic pollutants into harmless smaller molecules, ultimately carbon dioxide, water, and inorganic ions. The problem is that pristine TiO2, with its wide bandgap of roughly 3.2 electron volts for the anatase phase, absorbs only ultraviolet light, which is a small fraction of natural sunlight and virtually absent from ordinary indoor lighting. Decorating the TiO2 surface with silver nanoparticles changes that equation dramatically.

Silver nanoparticles act as plasmonic antennas. Their conduction electrons oscillate collectively in response to visible light—a phenomenon known as surface plasmon resonance—which enhances light absorption and promotes the transfer of energetic electrons into the TiO2 conduction band. The silver also serves as an electron sink, trapping photoexcited electrons and thereby suppressing the recombination of electron–hole pairs, the wasteful process that otherwise limits photocatalytic efficiency. By extending the catalyst’s absorption into the visible spectrum and improving charge separation, the silver decoration makes the material responsive to the cool white light of household LEDs—an illumination source that is inexpensive, low-temperature, energy-efficient, and far more practical for real-world water treatment facilities than UV lamps.

What distinguishes this work is how the silver nanoparticles were made. Rather than employing chemical reducing agents such as sodium borohydride or citrate—reagents that are themselves hazardous and generate toxic byproducts—the researchers exploited the phytochemicals naturally present in rosemary extract. Plant-derived polyphenols, flavonoids, and terpenoids act as both reducing agents, converting silver ions into metallic silver nanoparticles, and capping agents, stabilizing the nascent nanoparticles and controlling their growth. The result is a one-pot, eco-friendly synthesis in which the plant chemistry performs the work of industrial reagents. The team confirmed the successful formation of the nanocomposite through a battery of characterization techniques: UV–Vis spectroscopy, which reveals the characteristic plasmon absorption of silver; X-ray diffraction (XRD), which identifies the crystalline phases; scanning electron microscopy (SEM), which images the morphology; and Fourier-transform infrared spectroscopy (FTIR), which detects the organic functional groups from the plant extract bound to the nanoparticle surfaces.

With the catalyst in hand, the researchers systematically varied the operating conditions of the photocatalytic degradation of AO10 under white LED irradiation. Four parameters emerged as critical: the pH of the solution, the catalyst loading, the concentration of hydrogen peroxide added as an oxidizing co-reagent, and the initial concentration of the dye itself. pH is particularly influential in TiO2 photocatalysis because it governs the surface charge of the catalyst particles and, consequently, the electrostatic attraction or repulsion between the catalyst and the dye molecules. At acidic pH values, the TiO2 surface is positively charged and attracts the anionic AO10 molecules; at high pH, the surface becomes negatively charged and repels the dye, while hydroxyl radicals are also scavenged more readily. Hydrogen peroxide, meanwhile, acts as a sacrificial electron acceptor that captures conduction-band electrons and generates additional hydroxyl radicals, accelerating the oxidative degradation—provided its concentration is optimized, since excess peroxide can quench radicals and even poison the catalyst surface. Catalyst loading presents a similar trade-off: more catalyst means more active sites, but excessive loading increases turbidity and scatters light, reducing the photon flux reaching each particle.

To navigate this multidimensional parameter space efficiently, the researchers turned to machine learning. Rather than relying on laborious one-factor-at-a-time experiments, they trained a polynomial regression model to capture the relationships among pH, catalyst loading, peroxide concentration, initial dye concentration, reaction time, and the resulting concentration ratio C/C0—the fraction of dye remaining relative to its starting value. Polynomial regression, a workhorse of statistical learning, fits a mathematical surface through the experimental data, allowing the model to interpolate and predict outcomes for conditions never tested in the laboratory. The team first screened several regression algorithms from the scikit-learn library before settling on the polynomial approach, and then coupled the trained model with Pareto optimization, a multi-objective technique that identifies conditions balancing competing goals—in this case, minimizing the residual dye concentration while minimizing reaction time. With the weighting parameter λ set to 1, the optimization predicted an ideal recipe: pH 1.50, a hydrogen peroxide concentration of 3.35 × 10⁻³ M, a catalyst loading of 80.24 mg L⁻¹, and an initial AO10 concentration of 26.40 mg L⁻¹. Under these conditions, the model forecast complete degradation of the dye in just 15.81 minutes—a remarkably short time for a process powered by an ordinary LED.

The machine learning results were complemented by quantum chemical calculations at the DFT level, which probed how the AO10 molecule actually attaches to the AgNPs/TiO2 surface. The team computed relative adsorption energies (Eads) for three possible binding orientations, in which the dye anchors to a silver site through its sulfonate group (SO3–Ag), through a nitrogen atom of the azo linkage (N–Ag), or through a hydroxyl group (OH–Ag). The calculations revealed that the SO3–Ag orientation is energetically the most favorable, meaning the dye preferentially binds through its sulfonate moiety—the acidic functional group characteristic of anionic dyes. This orientation positions the dye in a geometry that presumably facilitates electron transfer during photocatalysis. Perhaps more intriguingly, the DFT analysis showed that adsorption is strongly pH-dependent: at very high pH values, the adsorption process becomes endergonic—that is, thermodynamically unfavorable. This provides a molecular-level explanation for the experimentally observed collapse of photocatalytic activity under strongly alkaline conditions. The dye simply refuses to stick to the catalyst, and a molecule that never adsorbs can never be degraded. It is a satisfying example of theory and experiment converging on the same physical picture.

Kinetic analyses of the degradation experiments added further mechanistic depth. Photocatalytic degradation of dyes on TiO2 typically follows pseudo-first-order kinetics at low dye concentrations, where the rate is proportional to the dye concentration because active sites are not saturated. The kinetic data from this study helped the researchers characterize how the rate responds to each of the parameters they varied, feeding into the machine learning framework and reinforcing the understanding that adsorption of the pollutant onto the catalyst surface is the rate-determining first step in the overall process.

The significance of this work extends beyond a single dye. Acid Orange 10 serves as a model compound for an entire family of azo dyes, and the general strategy—an easily synthesized, plant-derived, silver-plasmonic TiO2 photocatalyst activated by low-energy LED light—could be adapted to other organic micropollutants, from pharmaceuticals to pesticides, that increasingly contaminate water supplies worldwide. The integration of machine learning for process optimization also points toward a broader trend in environmental catalysis: instead of exhaustive empirical screening, researchers can now train predictive models on relatively modest experimental datasets and use them to pinpoint optimal operating windows with minimal laboratory effort. And because the synthesis avoids toxic reagents and the illumination avoids energy-hungry UV lamps, the entire treatment chain aligns with the principles of green chemistry.

There remain, of course, the usual challenges between laboratory success and industrial deployment: scaling up nanomaterial synthesis, recovering and reusing catalyst particles from treated water, and validating performance in real, complex wastewater matrices rather than synthetic dye solutions. But the study offers a coherent proof of concept that a catalyst born from rosemary, powered by a reading lamp, and guided by an algorithm can dismantle one of the textile industry’s most stubborn pollutants in under sixteen minutes. For a field searching for sustainable answers to a global water contamination crisis, that is a compelling demonstration of what happens when green chemistry, surface science, quantum mechanics, and artificial intelligence are made to work together under the same light.