Silicon-Based Composite Photocatalyst for Solar-Driven Photocatalytic Water Purification

Abstract

In this work, we report the fabrication of a relatively efficient, cheap, and stable silicon-based photocatalyst for water purification. Silicon-nickel composites were fabricated by mixing and grinding metallurgical silicon with commercial nickel nanoparticles. Experiments demonstrate a significant improvement in methylene blue degradation under UV—and especially under solar—conditions. About half of the contaminants were degraded after 100 min. Theoretical modeling demonstrates the effects of impurities in subsurface layers on absorbance and on the release of hydroxyl radicals from silicon.

1. Introduction

Fabrication of low-cost, simple-to-produce, stable, and abundant-material photocatalysts for water purification is needed for the well-being of humanity. The most efficient photocatalysts are usually composed of costly and/or rare elements [1] or require state-of-the-art fabrication techniques [2]. This has gravely limited the large-scale manufacturing of widely available photocatalysts. Thus, turning to cheaper, more abundant sources is a reasonable choice. Silicon is the first choice.

Silicon-metal composites have emerged as a promising yet still developing class of photocatalytic materials because they combine the abundance, low cost, and tunable electronic structure of silicon with the charge separation and catalytic benefits of metallic or metal-containing components. In the current literature, most progress is concentrated on silicon-based heterostructures, where metals, metal oxides, or metal phosphides are integrated with silicon to improve light harvesting, suppress electron–hole recombination, and create active interfaces for redox reactions. This is especially important because bare silicon, despite its favorable band structure and broad availability, suffers from photocorrosion and limited long-term stability in aqueous environments [3,4].

Recent studies show that the most effective designs usually rely on nanoscale morphology control, such as silicon nanowires, porous silicon, or silicon nanocrystals, combined with catalytic metallic decorations or protective layers. These architectures can enhance photocatalytic performance in water treatment, hydrogen evolution, and carbon dioxide conversion by increasing surface area, improving band alignment, and facilitating interfacial charge transfer. At the same time, the field is moving beyond simple activity enhancement toward balancing activity with durability, since oxidation, surface reconstruction, and unstable interfaces remain major bottlenecks. For that reason, modern work increasingly emphasizes protective coatings, Schottky junction engineering, plasmonic effects, and co-catalyst optimization [3,4,5].

From a broader perspective, the current state of silicon-metal composites can be described as a transition from proof-of-concept systems to more rationally designed photocatalysts. Machine learning is also beginning to enter the field as a tool for predicting redox behavior and accelerating materials optimization, which may help shorten the development cycle for next-generation composites. Overall, silicon-metal composites are not yet the most mature family of photocatalysts, but they are a strong platform for sustainable solar-driven chemistry because they combine scalability with high tunability [3,4,5].

In this work, we propose an alternative solution to this problem: the migration of atoms from metallic nanoparticles to silicon, which is usually considered challenging [6,7]. In our work, we used this phenomenon to increase the concentration of metal dopants in the subsurface of silicon nanoparticles (SiNPs). Based on theoretical predictions, nickel was chosen as the simplest and most efficient dopant for silicon. Methylene blue has been chosen as the molecule of choice for testing photocatalytic efficiency.

2. Experimental

Metallurgical silicon of solar grade [8] was subjected to multi-stage mechanical grinding: first in a ball mill with balls with a diameter of 5 mm for 10 min to obtain a fraction with a particle size of less than 200 microns, then in a planetary ball mill sequentially at 600 rpm (d = 3 mm, t = 1 h) and 800 rpm (d = 0.1 mm, t = 1 h). To remove external impurities on the surface of the particles, the resulting powder was subjected to wet chemical etching in a mixture of HF:HCl:HNO₃ (0.5 M:5 M:5 M) acids at 60 °C for 2 h, followed by particle selection by two-stage centrifugation (3000 rpm, 15 min) and (4000 rpm, 15 min). The size of nanoparticles ranges from 150 nm to 1 µm. The silicon particles at the particle outlet were dried in an electric drying oven at 105 °C for 12 h to remove residual moisture. Next, the resulting silicon powder was mixed with commercial Ni nanoparticles at a 3:1 (by mass) ratio. Ni nanoparticles with a diameter of 100 nm and 99.5% purity were obtained from ACS MATERIAL. The two-component powder was additionally ground using a mechanical ball mill at 800 rpm (d = 0.1 mm, t =1 h). The general scheme of the Si/Ni sample preparation process is shown in Figure 1.

The synthesis of silicon particles took place in two stages:

• Mechanical stage: The initial powder of metallurgical silicon with large fractions was subjected to multi-stage grinding in a planetary ball mill (PULVERISETTE 7, FRITSCH, Idar-Obersten, Germany, 2022) using grinding balls of different diameters. This stage leads to the formation of particles with a wide size distribution from nanometers to tens of microns, which is typical for this method.
• Chemical step: To remove unwanted impurities and narrow the range of particle distribution, the method of intensive wet chemical etching in an acid solution HF:HCl:HNO₃ (0.5 M:5 M:5 M) at 60 °C for 2 h.

The chemical step is key to understanding the final dimensional distribution. Wet chemical etching of silicon in HF and HNO₃ acids is a well-proven two-stage oxidation process followed by dissolution of oxidized particles [9]. HCl acid removes impurities, enabling effective purification of the initial metallurgical silicon. Thus, all the components of this acid solution perform the following functions: HNO₃ oxidizes, HF «opens» the surface, and HCl «washes out» impurities.

Silicon nanoparticles (less than 100 nm) have a higher reactivity due to their large specific surface area. Consequently, they have a higher rate of dissolution in HF and HNO₃ acids. According to research by Choi et al., the etching rate is inversely proportional to nanoparticle size, meaning smaller particles dissolve faster [10].

For the selection of mechanically ground particles, a two-stage centrifugation protocol was used, which provided quantitative solid-phase deposition of silicon particles. The centrifugation protocol used to separate silicon particles is directly confirmed in the work of Biswas S. and co-authors, where the “cyclic process of ultrasonic treatment and centrifugation” of a colloidal suspension of silicon powders in ethanol makes it possible to accurately select silicon particles by size and obtain fractions with a predominance of nanoparticles of a certain diameter [11]. In addition, a mechanochemical method similar to the laboratory protocol described in our manuscript was used to synthesize the initial particles.

Scanning electron microscopy (SEM, JeoL Ltd., Akishima, Japan) was used to visualize the size of the obtained particles. The SEM images (Figure 2) show that the obtained photocatalyst samples based on metallurgical silicon are particles ranging from 150 nm to 1 µm in size. The absence of metallurgical silicon particles smaller than 100 nm in the images presented is direct evidence of the effectiveness of acid treatment. X-ray spectroscopy (XPS) measurements were performed using an AXIS UltraDLD (Shimadzu Corporation, Kyoto, Japan) with Al Kα radiation. The spectra were fitted using the SpectraFit XPS online app [12]. The peak positions were taken from the NIST XPS database [13].

As for nickel particles and, in particular, the Si/Ni photocatalyst, we used a certified commercial product, thereby eliminating uncertainty about their characteristics. According to the manufacturer’s certificate (ACS Material), nickel nanoparticles have an average size of 100 nm [14].

Methylene blue (MB) was chosen as the standard testing contaminant. An initial 30 mL MB solution at 10 ppm was prepared, followed by the separate addition of two photocatalysts: metallurgical Si particles (20 mg) and composite Si/Ni particles (20 mg). Before light irradiation, the solutions were mixed in the dark for 30 min to establish an adsorption–desorption equilibrium between the catalyst particles and MB molecules (

Table 1. Results of control experiments: dark adsorption and photolysis without catalyst.

Experiment	Time (min)	Decrease in MB Concentration (%)
Dark adsorption Si	30	2
Dark adsorption Si/Ni	30	6
Photolysis without catalyst SM	120	5
Photolysis without catalyst UV-lamp	120	8

).

The samples were irradiated using a Class AAA LED solar simulator (Ossila, Sheffield, UK). The device simulated the 1.5 G AM spectrum over the wavelength range 300–1000 nm, with a radiation intensity of 1000 W/m² at an operating distance. UV lamp: an OSRAM ULTRA-VITALUX mercury lamp (OSRAM, Munich, Germany) was used as a source of ultraviolet radiation; the power is 1.1 W/m² for the UV-B range, 7.3 W/m² for the UV-A range, and 29.7 W/m² for U-Vis. To assess the contribution of direct photolytic decomposition of MB, control experiments were conducted: the dye solution (10 ppm) was irradiated with a solar simulator (300–1000 nm) and a UV lamp without catalysts for 120 min. The obtained data are shown in Table 1.

The distance to the source and the flow intensity were kept constant throughout the series of experiments. 1.5 mL samples were taken from the reactors at regular intervals. All photocatalytic experiments were conducted in at least three repetitions, and the data are presented as average values. The solid catalyst particles were separated from the solution by centrifugation. The concentration of MB remaining in the solution was determined by spectrophotometry using transmission measurements. The transmission and absorption spectra are related by Equation (1):

$$A=-{\mathit{log}}_{10}\left({\displaystyle \frac{T}{100}}\right),$$

(1)

where A is the absorption; T is the transmission coefficient of the solution.

The concentration of the resulting solutions after photocatalytic purification can be estimated using Equation (2):

$$C_t={\displaystyle \frac{C_o{A}_t}{A_o}},$$

(2)

where C_o is the initial concentration of methylene blue; C_t is the concentration at time t; A_o is the initial absorption value; A_t is the absorption value at time t.

Theoretical modeling was carried out using the SIESTA pseudopotential code [15], employing the generalized gradient approximation (GGA-PBE) [16] for the exchange-correlation potential in spin-polarized mode, and including van der Waals corrections [17]. The latter is essential for simulating hydroxyl groups on the surface. A full optimization of the atomic positions was carried out, during which the electronic ground state was consistently obtained using norm-conserving pseudopotentials [18] for the cores, with a double-ξ-plus polarization basis for non-hydrogen atoms and a double-ξ basis for hydrogen atoms. The forces and total energies were optimized with an accuracy of 0.04 eV Å⁻¹ and 1.0 meV/cell, respectively. The simulations were conducted using a 6 × 6 × 1 Monkhorst–Pack k-point grid for Brillouin-zone sampling [19].

3. Results and Discussion

3.1. Synthesis Yield

When evaluating the yield of the target product, the following data were obtained: mechanical grinding is not associated with significant material losses, which amount to less than 5% of the initial mass. The main losses (about 25% of the initial mass) occur during wet chemical etching in an acid solution, due to the removal of impurities and the partial dissolution of the silicon surface layers in an aggressive environment. The use of two-stage centrifugation and particle drying results in an additional 5% loss. The final yield of synthesized silicon particles was about 65% (6.5 g).

3.2. Discussion of Silicon Particle Sizes

The size of the obtained silicon particles is especially important in the context of the research by Lv and co-authors, in which the dependence of the photocatalytic activity of metallurgical silicon on particle size in the hydrogen evolution reaction was revealed [20]. Large particles of metallurgical silicon (with sizes greater than 800 nm) exhibit extremely low photocatalytic activity due to impurities that, with their large size, remain “trapped” within the crystal volume and effectively quench photogenerated carriers. When the particles are reduced to 400 nm, the charge carriers travel a path that is shorter than the diffusion length [20]. As a result, the probability of their capture on impurity centers in the crystal volume decreases.

The introduction of Cu, Ag, and Pt nanoparticles onto the surface of silicon particles makes it possible to increase the photocatalytic activity due to the formation of a Schottky barrier, which promotes the separation of electron-hole pairs [20]. This also echoes the research presented in this article, which incorporates commercial nickel particles into synthesized silicon particle samples.

According to the results of the aforementioned scientific work, the photocatalytic activity of metallurgical silicon primarily depends on particle size. Based on this, the dominant contribution to the relatively high photocatalytic activity of the silicon particles synthesized by us is due to smaller particles, 150–400 nm in size. Silicon particles included in this size range provide an active surface area and favor the separation of photogenerated charge carriers. An additional factor is the wet chemical cleaning of silicon particles in an acid solution to remove unwanted impurities.

3.3. Theoretical Modeling

Silicon is usually considered a photoactive catalyst in UV light, and doping can make it photoactive under solar light [3]. We examined the effects of two types of impurities (iron, as a natural impurity in metallurgical silicon [6], and nickel) on the surface and subsurface (Figure 3a,b). Our previous density functional theory-based calculations demonstrate that the formation of substitutional (M_Si) and interstitial (M_i) iron impurities in silicon is nearly as favorable for iron as for nickel [7]. The calculations show that substitutional iron impurities have a negligible effect on light absorption in silicon. On the contrary, incorporating iron atoms into interstitial voids leads to the appearance of the tail corresponding to the visible part of the spectra (Figure 3c,d). In the case of nickel, both types of impurities increase absorbance in the visible part of the spectrum, especially the substitutional ones. Incorporation of iron and nickel impurities does not change the positions of CB and VB in the model slab (–3.3 eV and –4.0 eV, respectively).

In addition to generating OH radicals by splitting water molecules [3], hydroxyl groups from the silicon surface can also be released into solution. Our calculations demonstrate that the presence of Fe_Si and Fe_i decreases the energy cost of hydroxyl radical release by 0.62 eV/OH and 1.12 eV/OH, respectively. Incorporation of Ni_Si and Ni_i impurities also decreases the energy costs of hydroxyl radical release by 0.38 eV/OH and 0.45 eV/OH, respectively. However, because the concentration of metallic impurities in solar-grade silicon is very low, their contribution to photocatalytic performance is negligible. Thus, simple methods for incorporating metallic impurities into the silicon surface are needed.

3.4. Experimental Verification

Table 1 presents the results of control experiments and the effects of dark adsorption and direct photolysis on the reduction in MB concentration. The reduction in MB concentration during the dark period (30 min) amounted to only 2% for Si and 6% for Si/Ni. Additionally, direct photolysis of MB without samples was evaluated both under a solar simulator (300–1000 nm) and under a UV lamp for 120 min. The MB concentration decreased by only 5% under solar simulator exposure and by 8% under UV radiation.

The photocatalytic characteristics of the purified Si powder were evaluated by measuring the decomposition of 10 ppm MB under ultraviolet and full-spectrum solar simulator irradiation at room temperature. The results are shown in Figure 4a. Changes in the intensity of the MB absorption peak at 664 nm after irradiation with a solar simulator and a UV lamp indicate effective MB decomposition. Figure 4b shows the MB concentration ratio over time under UV lamp and solar simulator irradiation. It is clear that the photocatalyst composition and irradiation wavelength significantly affect photocatalytic activity, thereby promoting MB dye decomposition. In all cases, there is a gradual decrease in the value of C_t/C_o, which indicates the decomposition of MB molecules under the influence of photocatalytic processes. In the presence of silicon, photodegradation occurs most slowly under a solar simulator. Throughout the experiment, the dye concentration decreases gradually, and by 120 min, the C_t/C_o ratio remains relatively high (0.3), indicating limited photocatalytic activity of pure silicon under irradiation conditions similar to those of solar radiation. When irradiated with UV light, the degradation rate increases markedly, and C_t/C_o reaches 0.18. The observed photoactivity of undoped silicon can be attributed to iron impurities in the samples used [5]. X-ray photoelectron spectroscopy of Ni/Si composites demonstrates the oxidation of the surface layers of Ni nanoparticles (see Figure 5). These spectra also show valuable contributions from NiSiO₃ and NiSi-related peaks, which correspond to Ni impurities in the surface (Figure 3a) and subsurface layers (Figure 3b).

The details of the process are reported in the methods section. The experiment demonstrates that the formation of Si/Ni composites leads to a visible improvement in photoactivity in UV and under a solar simulator. The degradation rate of MB increases significantly compared to undoped samples, especially at short times, which is essential for the purification of running water (see Figure 4b). After one hour, the degradation rate increases by more than twice, from about 0.2 to about 0.5, for both types of light. Note that after one hour, the switch in the kinetics of MB decomposition occurred in both samples. The addition of commercial Ni particles leads to a significant change not only in the photocatalyst’s efficiency but also in the kinetics of the photocatalytic decomposition of methylene blue (MB). If for the studied sample of metallurgical silicon, a convex profile of the kinetic curve is observed (where the velocity is maximum at the initial moment and gradually slows down), then for the Si/Ni composite, the curve acquires an S-shape: after the initial section, a sharp acceleration of the photocatalytic process follows, forming a concave profile at the middle stages of the reaction. These changes can be attributed to the switch between contributions from different reaction mechanisms [3]. The first is the release of OH radicals from the surface of silicon nanoparticles. This process is facilitated by the presence of Ni impurities. The second is the standard mechanism for inducing radicals by the migration of photoelectrons from the surface to molecules that generate radicals in liquid media, as shown in Figure 6.

After two hours, only 14% of MB remained under a solar simulator, and 14% under UV light. The time required for the total degradation of 10 ppm of MB, estimated from Figure 4b, is about 2.5 h for doped samples and 3.2 h for undoped SiNPs. Since the oxidation of metallurgical silicon is self-limiting [21] and the extraction of the impurities discussed requires temperatures above 800 °C [6], the studied Ni-doped silicon nanoparticles can be considered stable over time. Our recent experiments also demonstrate the performance of Ni-doped porous silicon after two years [22].

Summarizing the above experimental data, we can identify key patterns: these data clearly demonstrate the synergistic effect of the introduction of nickel into the initial composition of the photocatalyst; the composite sample is not only significantly superior to metallurgical silicon in efficiency, but also demonstrates lower sensitivity to the spectral composition of radiation, which is an extremely valuable property for practical use in sunlight. However, to fully understand the role of Ni in photocatalytic mechanisms, further experimental studies of the band structure (e.g., C_B, V_B, and Fermi level) of the obtained Si/Ni composite are required [23].

4. Conclusions

The combination of theoretical predictions and experimental verification via methylene blue degradation measurements demonstrates that even undoped metallurgical SiNPs exhibit photocatalytic activity under UV and visible light, with half of the contaminant degraded after 100 min. Theoretical simulations suggest that residual metallic impurities may be a source of this effect. A silicon-nickel composite fabricated by mixing and grinding nanoparticles demonstrates improved photocatalytic performance, achieving half of the contaminant’s degradation within an hour. Theoretical modeling attributes this phenomenon to nickel incorporation into the subsurface, which enhances absorbance in the 300–1000 nm range (simulated sunlight) and releases hydroxyl radicals at the surface.