Crystalline silicon (c-Si) is superior to other semiconductors in its low precursor cost, relatively simple fabrication procedures, and high material stability. These superiorities make it currently the best compromise as the workhouse semiconductor in modern industries. There has been a dramatic increase in demand of c-Si in the expanded electronic and photovoltaic markets. The supply of this treasurable material, however, is largely plagued by the intensive energy consumption and increased environmental impacts such as CO2 emission and wastewater generation in its synthesis.1–3
A large amount of energy is required to release silicon atoms from their precursors, which reconstruct to yield diamond-structured silicon. Figures 1a and 1b summarize the established methods of c-Si production. All these synthetic routes require high-temperature conditions. Deoxidization process is commonly used to produce metallurgical-grade c-Si by reducing silicon dioxide (SiO2) to yield elemental silicon. Due to the large binding energy of Si-O-Si network in SiO2, the reaction to produce the high purity level of c-Si requires substantial amount of reductant and energy (Fig. 1a). Pyrolysis is another established strategy to fabricate c-Si by the cleavage of reactive Si-H or Si-Cl bonds at high temperature to form c-Si with high purity (Fig. 1b). However, this reaction involves highly flammable silane molecules as reactants and proceeds at high temperature (> 500°C); the materials require special care during the transportation and the reaction.
Dehydrogenative coupling reactions are known to create Si-Si linkage in organosilane chemistry.4–11 The process generally involves the oxidative addition of the Si-H bonds with the predesigned molecular catalysis activation and the subsequent elimination of hydrogen (H2).4,6 Dehydrogenative coupling is thus potentially a powerful tool to prepare chemicals with Si-based backbones. Indeed, compounds with Sin units such as organodisilanes and linear polysilanes have been successfully synthesized through dehydrogenative coupling reactions.7,8,10 However, the crosslinking of Si-Si bonds through dehydrogenative reactions is difficult due to the significant steric hindrance of the bulky organic groups. On the other hand, volatile silanes without the bulky groups (e.g., SiH4 and HSiCl3) are too reactive to be used for such solution-based coupling reactions. The potential for dehydrogenative reactions to construct more complex, organized inorganic Si-based network (such as c-Si) have so far been largely ignored.
Two-dimensional (2D) nanomaterials can behave as the building blocks to construct three-dimensional (3D) bulk architectures.12–15 Silicane owns a sp3-hybridized single-layered Si backbone with surface passivation and thus its atomic arrangement is similar to that on the hydride-terminated c-Si(111) surface.16,17 Inspired by the organosilane coupling reactions promoted by Lewis base (LB) reagents, we sought an approach that would chemically “suture” hydride-terminated silicane (termed HSi henceforth) to form c-Si through dehydrogenative coupling reaction in a typical weak LB-type solvent. Our aim was to create covalent Si-Si bonds between HSi flakes through wet chemistry method at mild conditions (Fig. 1c).
In the proof-of-concept experiment, HSi was prepared using the chemical exfoliation of CaSi2 and purified by HF-etching, followed by multiple washing steps to remove unreacted precursors and impurities (See Methods). In a typical process of chemical suture of HSi, purified HSi was added to dimethyl sulfoxide (DMSO), a weak LB reagent, under a nitrogen atmosphere at room temperature. A substantial quantity of gas formed from the reaction, and the original greenish powders changed to shinny and dark-grey crystalloid pellets after the immersion in DMSO (Fig. 2a). The detailed morphology was verified by scanning electron microscopy (Figs. 2b, 2c, Supplementary Figs. 1 and 2). The original layered structure of HSi converted to micrometer-scaled crystalline trunks after the 12-h reaction with DMSO. The crystal evolution can also be monitored by the powdery X-ray diffraction (XRD). The signal at 2θ = ~15° in the XRD pattern of freshly prepared HSi is assigned to (001) lattice plane (see the inset of Fig. 2d). This signal disappeared after the DMSO treatment. It is important to note that c-Si and HSi share similar strong features in the XRD patterns (e.g., 28°, 47°, and 56°), which can be assigned to the common Si(111)-like facets.16
The generation of H2 was detected by gas chromatography (GC), confirming that the cleavage of Si-H occurs during the DMSO mixing process; in contrast, no observable H2 formation was noticed from the HSi powders or the controlled HSi/toluene solution (Fig. 2e), suggesting that the H2 formation is promoted by the presence of DMSO. Absorption and photoluminescence spectroscopies further revealed the bonding evolution during the DMSO treatment. The original absorption edge at 483 nm shifted to 1086 nm after the reaction (Fig. 2f), and the green photoluminescence (PL) at 495 nm is notably diminished (Supplementary Fig. 3). We ascribe this to the formation of c-Si which has a bandgap of ~ 1.1 eV, a significant reduction from the ~ 2.6 eV optical gap of silicane. Raman measurements confirm the presence of the distinguishable 2D Si six-member-ring (~ 380 and ~ 490 cm− 1) and Si-H vibration modes (~ 640 and ~ 730 cm− 1) in silicane, whereas only the bulk Si-Si signal at ~ 515 cm− 1 was observed in the treated sample (Fig. 2g). Similarly, the original Si-H stretching signal (~ 2100 cm− 1) disappeared from the Fourier transfer infrared (FT-IR) spectrum of the DMSO-treated sample (Fig. 2h). Both Raman and FT-IR results indicate that the DMSO treatment effectively cleaves Si-H bonds and converts the layered structure of silicane into bulk c-Si.
Synchrotron-based X-ray characterizations gave further insight into the impact of the oxidation state of silicon atoms and the crystal evolution induced by the DMSO treatment. The surface-sensitive total electron yield (TEY) X-ray absorption near-edge structure (XANES) spectra at the Si K-edge reflects the formation of surface oxide layer on both pristine and DMSO-treated samples (Supplementary Fig. 4), which is not unexpected because the surfaces of both HSi and c-Si are prone to oxidation.18,19 The total fluorescence yield (FLY) XANES spectra, on the other hand, reveals the bulk crystal information. The presence of the strong signal at ~ 1847 eV in both pristine and DMSO-treated samples is consistent with the feature of the standard SiO2 sample (Fig. 2i), indicating the presence of oxidation in both samples. The absorption onset energy of the Si 1s feature of HSi (1840.0 eV) is higher than that of c-Si (1839.1 eV). The high absorption onset energy in HSi may originate from the slightly positively charged Si atoms in HSi (χSi = 1.90 vs. χH = 2.20, and natural atomic charges calculated to be + 0.1 and − 0.1 e for Si and H (see Supplementary Note 1 and Note 2). The shoulder signal shifts by 0.9 eV following the DMSO treatment with a similar feature of the control c-Si sample and can be assigned to the resonance excitation of Si atoms in c-Si.20 These observations are in excellent agreement with XRD and Raman results, and together these findings suggest that the DMSO treatment leads to the complete conversion from the layered HSi structure to bulk c-Si crystalline.
Interestingly, it has been reported that DMSO, serving as a weak oxidizing reagent, can oxidize hydride-terminated porous silicon and form dimethyl sulfide (DMS) through a radical-driven mechanism.21 However, no DMS was detected from either the gas phase or the solution, suggesting that DMSO does not serve as the oxidant in this study. All these experimental results hint on a face-to-face suture of silicanes catalyzed by DMSO at room temperature, yielding c-Si and H2 gas.
To assure this is the case, we resourced to quantum chemistry calculations using density functional theory (DFT) method. A HSi(SiH3)3 molecule was used to model a HSi flake to study the suture mechanism. The reaction occurs locally at one Si site, which is surrounded by three Si atoms, and all the Si atoms are in sp3 hybridization. This model preserves the silicane chemical environment for the central Si and H atoms (Fig. 3a). With this model, we investigated the 2HSi(SiH3)3 = (SiH3)3Si-Si(SiH3)3 + H2 reaction, and the conclusions are transferrable to the suture of two silicanes flakes. The reaction was calculated to have \(\varDelta G\) = -15.7 kcal/mol when the translational and rotational contributions to the Gibbs free energies of the HSi(SiH3)3 monomer and the (SiH3)3Si-Si(SiH3)3 dimer are ignored. This approximation is reasonable since the HSi flakes, before and after they are connected by a Si-Si bond, cannot freely rotate and translate in the solution. All reported energies below were obtained under this approximation, unless further specified. Such a favorability is reduced to -8.5 kcal/mol, if the translational and rotational Gibbs energies of the H2 molecule are excluded. These computational results are in consistence with the observed mild exothermicity of the reaction, and the effusion of H2 gas itself contributes half of the thermodynamics favorability.
Without the presence of DMSO, the reaction barrier for the suture was calculated to be 62.7 kcal/mol, which is formidable and prevents the reaction from occurring (Fig. 3b). The reaction complex maintains a closed-shell electronic structure at the transition state (TS). We used symmetry-broken unrestricted open-shell DFT method to calculate the barrier and found it to be even higher (> 80 kcal/mol), excluding the possibility of a biradical TS. The closed-shell structure at TS corresponds to heterolytic partial cleavages of the two relevant Si-H bonds (Step i in Fig. 3b). The resultant hydridic and protonic H atoms readily form a H2 molecule, leaving the two Si centers to form a Si-Si bond and stitch up the silicanes (Step ii in Fig. 3b). The high barrier arises from the difficulty in the heterolytic cleavages of the Si-H bonds in the absence of catalyst, especially in the formation of the protonic H atom and the silicide center (the red arrow in Step i), in which the bond pair migrates in a counter-electronegativity manner.
The presence of DMSO lowers the barrier to 28.7 kcal/mol, a surmountable barrier at room temperature (Fig. 3b). DMSO attacks from the opposite direction of a Si-H bond using its O basic center, converting the Si to a siliconium center with a trigonal bipyramidal configuration (Step iii in Fig. 3b). The Si-H bond has been partially heterolytically cleaved. The hydridic center is ready to abstract a proton and thus induce the counter-electronegativity heterolytic cleavage of the Si-H bond of the other HSi(SiH3)3. Clearly, the inductions of the heterolytic cleavages of the two Si-H bonds, first by DMSO, then by the generated hydridic center, cause the 34 kcal/mol barrier reduction, which enables the reaction to occur slowly. The TS can be viewed as being stabilized by a 5-center-6-electron interaction across the O-Si+-H−-H+-Si− bridge (more details available in Supplementary Note 2 and Note 3). Crossing the TS, the H2 formation and Si-Si bond suture follow the same pathway as the uncatalyzed reaction, and the DMSO falls off (Step iv in Fig. 3b). The large availability of the DMSO solvent molecules guarantees that every Si site is accessible by these catalytic bases. The catalytic cycles hence continue and eventually results in the 2D-to-3D suture, as demonstrated in Fig. 3b. It is logical to expect that the basicity of a solvent and the steric accessibility of its basic site are the key to catalyze the suture process. It is noteworthy that the anisotropy of the DMSO in exhibiting its Lewis basicity facilitates the catalytic process. If the solution consists of a large amount of isotropic basic species, such as F−, it is likely that the two HSi bonds in Fig. 3 are heterolytically cleaved in the same Siδ−Hδ+ manner. The two hydridic H atoms will thus not form an effusive H2 molecule, and the reaction does not proceed. In the contrary, the DMSO solvent molecules can approach the backsides of the two Si-H bonds, which are to be sutured, through their oxygen end and methyl end, respectively, so that the Siδ−Hδ+ partial cleavage only occurs in one of the Si-H bonds, which is activated by the oxygen end.
We also performed calculations to investigate the possibility of the DMSO inserting its O atom to the Si-H bond and resulting in the Si-O-H moiety. As expected, this reaction is highly exothermic, with \(\varDelta G\) = -43.1 kcal/mol. However, Gibbs reaction barrier was estimated to be 39.1 kcal/mol high (See Supplementary Fig. 17). This reaction is thus kinetically non-competitive with the suture of HSi flakes, despite the thermodynamics sink of forming Si-O bonds. More details of our computational studies are given in Supplementary Notes 1 to 3.
The materials characterizations and control experiments lead us to an overall model of the LB-driven c-Si formation (Fig. 4a), beyond the understanding of the formation of each Si-Si bond that was gained by DFT calculation. The details of this model are presented in Supplementary Note 5. Overall, the suture starts at one pair of Si-H bonds on two HSi flakes, with the catalysis of DMSO (the first arrow in Fig. 4a). This first Si-Si bond, which we call anchored bond, brings nearby Si-H bonds of the two flakes close to each other and more suturing processes ensue. The suture propagates in two dimensions to completely stitch up the two flakes (the propagation arrows in Fig. 4a). Ideally, the other HSi flakes spectate the suture of the two flakes (see Supplementary Note 5 for more detailed discussion). After the suture of the first two flakes into the double-layered HSi2 intermediate structure, the third and fourth flakes are sutured with the first and second, respectively, on the two sides of the HSi2. Since the backsides of all Si-H bonds of the HSi2 are enclosed by Si-Si bonds and are not accessible to DMSO, the DMSO activation of Si-H bonds can only occur at the third and fourth flakes. On and on, the thickness of the HSin increases as n HSi flakes are sutured in this layer after layer manner, and eventually, the perfect c-Si is formed. Two HSi2 double-layered flakes cannot be sutured due to their DMSO-inaccessible backsides of Si-H bonds.
However, such an adiabatic (in thermodynamics sense) layering process hardly happens. This is because multiple anchored Si-Si bonds can be formed between several HSi flakes, which then evolve to some HSi2 layer fragments (see the “Reality” part of Supplementary Note 5). These layer fragments cannot be stitched up through the DMSO catalysis and therefore, the space between the HSi2 layer fragments eventually become enclosed cavities in the final product of c-Si. Some of them become pores at the surface of c-Si, if they are not enclosed in three directions.
The proposed mechanism suggests that the reaction rate to form the c-Si product and the properties of the product depend on LB concentration. In order to confirm this proposal, we implemented the dehydrogenative coupling reaction in solvents with various DMSO/toluene mixtures. Notable differences in the H2 production rate were observed (Fig. 4b), and the reactions with only 10% or 1% DMSO were far from completion even after 500 min. We attributed this to the reduced probability of the Si-H bond heterolytic cleavage by the lower concentration of LB. Interestingly, there is no obvious correlation between surface area, pore size, and DMSO concentration (Fig. 4d and Supplementary Fig. 5), while we expect enhanced porosity with a higher DMSO concentration. Clearly, the high DMSO concentration facilitates the formation of the enclosed cavities within the bulk of c-Si, instead of pores. The inner surfaces of these “bubbles” are inaccessible to the N2 gas in the isotherm experiment, and thus do not contribute to the measured surface areas. We observed that the DMSO serves as mild oxidizing agent in generating surface Si-O-Si bonds. This is verified by FT-IR results in Supplementary Fig. 7, which show clear signal of OSiSi-H bond stretching.
We applied electron paramagnetic resonance (EPR) analysis to further investigate the detailed surface states of c-Si. Two sets of strong signals can be found from all samples regardless of the concentration of DMSO used in the complete treatment (i.e., reaction for over 12 h, Fig. 4c). The strong absorption in the g = 2.0039 region is characteristic of the Pb defects attributed to the surface silyl radical groups.22–24 Another notable feature at g = 2.0111 can be assigned to the oxygen-related trap-hole centres (OHC) associated with the surface oxide species.25–27 Both absorptions exhibit obviously higher intensities for the sample prepared using 100% DMSO concentration. This is reasonable since the higher concentration of the catalyst results in a more rapid (i.e., less adiabatic) suture of HSi flakes into c-Si. Consequently, more enclosed cavities, and correspondingly a larger inner surface area and more radical sites in the inner surface result. Again, the inner surface of the enclosed cavities is inaccessible to the N2 adsorbate in measuring surface area.
The function of surface oxides and their derivatives (e.g., silyl and oxide radicals) on the photophysical properties of c-Si has been a subject of ongoing discussion.28,29 We therefore performed solid-state density functional theory (DFT) calculations on seven classes of surface state structures (Fig. 4e and Supplementary Fig. 8), including oxide/suboxides and silyl radicals, to investigate their impact on the electronic structures of c-Si. The projected density of state (pDOS, Fig. 4f and Supplementary Fig. 7) results clearly show that both states of valence band maximum (VBM) and conduction band minimum (CBM) are dominated by silicon contribution. The midgap state of the structure with a surface Si• is dominated by the pDOS of this Si atom, as expected. The midgap state of the structure with a surface O-Si• radical has not much contribution from the O atom. It mainly arises from the Si atom and its Si neighbors. Also, the midgap state is not far away from the original VBM (Fig. 4f). The surface O atom formally adopts a -2 valence state, while the unpaired electron is distributed around the nearby Si atoms. This is consistent with the large difference in electronegativities of O and Si and the easiness of electron migration in silicon (a semiconductor). These two midgap states are responsible for the two g-factor values shown in Fig. 4c. They are believed to effectively enhance the charge carrier migration under photoexcitation.
Hydrogen peroxide (H2O2) has been widely used in industries. However, its current synthetic approach (i.e., the anthraquinone process) causes large energy consumption and inevitably generates a significant amount of organic wastes. Motivated by the enhanced charge carrier migration of c-Si by the introduction of abundant surface radical states, we sought to investigate the possibility of the chemically sutured c-Si samples being used as the photocatalysis for the green production of H2O2. In a typical reaction, 5 mg of c-Si powders were dispersed in deionized water with continuous supply of oxygen (Supplementary Fig. 9). Triethylamine (TEA) and water served as the sacrificial agent and the hydrogen source, respectively. The oxygen reduction product, H2O2, was immediately formed when the solution was irradiated by the white light source (λex > 420 nm). The highest H2O2 yield was found on the c-Si formed by the treatment with 100% DMSO (Fig. 4g), which is expected to possess more enclosed cavities and larger inner surface area. The average yield for all measured 100% DMSO treated c-Si samples was 219.0 ± 4.9 µmol L− 1 under irradiation for 1 h at room temperature, over 35 × and 2.7 × higher than those of the reactions using HSi (6.2 ± 3.4 µmol L− 1) and commercial c-Si nanoparticles (81 ± 6.2 µmol L− 1), respectively. Considering the DMSO concentration does not significantly influence the band positions as verified by X-ray photoelectron spectroscopy (XPS, Supplementary Fig. 10), we attribute the yield improvement to the increased concentration of inner surface radical sites which facilitate the charge carrier migration and separation, and chemical adsorption of reactant and intermediate radical species (see Supplementary Note 6 for a detailed discussion based on solid state DFT calculations). The c-Si catalysts retained over 95% H2O2 production rate after 5 cycles (1 h per cycle) of photocatalytic reactions (Supplementary Fig. 11), indicating that the catalytic active sites and the structure of c-Si are sufficiently stable for efficient aqueous oxygen reduction reaction (ORR).
In the context of our observations, we provide two possible mechanisms of the photocatalytic H2O2 production (Supplementary Fig. 12). Two types of surface radical sites, siloxyl (≡ Si-O•) and silyl (≡ Si•) groups, of c-Si firstly react with H2O and/or O2, producing highly-active surface peroxide intermediates (•O-O-H or ≡ Si-O-O-H). Under the light irradiation, holes and electrons were separated and traveled to the surface of c-Si. The holes are abstracted by TEA to form TEA+, whereas the electrons further interact with the peroxide intermediates, leading to the formation of H2O2 and the re-generation of surface radical species.
In conclusion, we have developed a new wet chemistry route to prepare crystalline Si at room temperature. We achieved this via a Lewis-based-promoted chemical suture reaction on 2D silicanes. The basic catalyst heterolytically cleaves a Si-H bond in a Siδ+Hδ- manner, the hydridic H atom induces a counter-electronegativity Siδ-Hδ+ cleavage of a Si-H bond from another silicane, the hydridic and protonic H atoms form an effusive H2 molecule, and then the leftover silide and siliconium centers form a Si-Si bond between two silicanes. This whole process is thermodynamically favorable and features a surmountable energy barrier. Such a Si-Si bond formation propagates to suture silicanes into Si crystals. A higher concentration of the catalytic Lewis base facilitates the suture and the formation of crystalline Si with more abundant nanometer-scaled enclosed cavities. Although the inner surfaces of the enclosed cavities are not directly accessible to reactants, they provide inner surface radical sites that facilitate photoexcited charge separation and migration, which enhance the capability of the Si crystal as photocatalyst for oxygen reduction reactions. This is proved by the present work. Conceptually, this work demonstrates that the organosilane coupling strategy can be extended to inorganic synthesis of Si-based materials and provides an approach to further tailor the electronic and photophysical properties of silicon in optoelectronic and catalytic applications.