The elucidation of aqueous silicate complexes is of particular interest to geochemists as the composition of the Earth is dominated by silicate minerals—the natural abundance of silicon in the crust is almost 29% by weight1. Much of the chemistry that is known for inorganic silicate solutions and associated thermodynamics is based upon hydrothermal solubility experiments2,3,4,5, which are often coupled with vibrational6, optical7,8, X-ray9,10, or nuclear magnetic resonance (NMR)11 spectroscopies. Silicon is present in biological systems, but the role of this silicon is not fully understood12,13.

In dilute, near-neutral solutions aqueous silicon generally exists as silicic acid, or Si(OH)40, which deprotonates at high pH. This monomer, depending on the alkalinity and speciation of the counterions of the system, can polymerize and form a myriad of soluble oligomeric species including the cubic octamer14,15,16. The oligomers maintain a coordination number of four through either terminal hydroxides or oxo-bridges to other silicon atoms. In solids, very high pressures, such as are found deep in the Earth, can cause an increase in coordination number to six. Silicon in aqueous complexes usually coordinates to four oxygen molecules at ambient conditions, except when bonded to certain aromatic and sugar alcohols that cause the Si(IV) coordination to increase from four to five or even six. In particular, sugar alcohols with a threo-configured tetraol, such as xylitol, have been shown to complex the central Si(IV) in up to a hexacoordinated species17,18, and aromatic diols, such as catechol, have also been shown to create tris-diolate-silicon species19,20.

Each of the complexation reactions, either the oligomerization or the complexation of the silicic acid, follows a similar reaction pathway, that of a simple condensation reaction (where R can be either the Si(OH)3 fragment or an organic structure):14

$${\mathrm{Si}}\left( {\mathrm{OH}} \right)_{4({\mathrm{aq}})}^0 + {\mathrm{R}} - {\mathrm{OH}} \rightleftharpoons {\mathrm{Si}}\left( {\mathrm{OH}} \right)_3\left( {{\mathrm{O}} - {\mathrm{R}}} \right)_{({\mathrm{aq}})} + {\mathrm{H}}_2{\mathrm{O}}$$

In the case of the oligomerization of the silicic acid, this step can be repeated until, for example, the octamer species is formed (Q38, using the nomenclature of Engelhardt21), and ultimately solid precipitates.

Because so few silicate complexes are chromophores, NMR has become one of the most useful techniques for following speciation. This method has yielded much insight into the character of the polymeric species in solution22, and also the presence of new organic complexes in solution. 29Si is the NMR active nucleus of silicon, which has a natural abundance of 4.7% and is a nucleus with medium sensitivity23.

While the solution chemistry of silicon has been extensively studied using NMR, particularly at high pH, it has not previously been studied at gigapascal pressures. At these high pressures, the static dielectric constant of water (ε = 78 at ambient conditions), reaches values greater than 100. This increase allows the solvent to support more ionized solutes24.

In recent years, developments of high-pressure solution NMR probes yielded information about the complexation of borate species, up to 1.2 GPa in pressure25,26. Efforts have also been made to increase the pressure threshold, robustness, and sensitivity of these types of probes, to the point where they can now reach 2.8 GPa27,28,29. Furthermore, there are active efforts to increase this pressure threshold but most designs limit the sample volume to sub microliter volumes where induction NMR detection of solutes is difficult (see Supplementary note 1)30,31,32,33.

Herein, we report the use of a narrow-bore high-pressure NMR solution probe coupled with a high-power shim stack that is used to explore the complexation chemistry of aqueous silicate solutions to 1.8 GPa. An image of the probe, coupled with a schematic, is included in Fig. 1. When pressure is applied to different aqueous silicate complexes, the silicate center tends to minimize the volume of the solution by favoring complexation of water into the inner sphere of the silicon. Interestingly, the tris-catecholate silicate complex is such a tightly bound complex that pressure does not induce a change in the coordination environment; we detect no species as pressure is increased other than the tris-catechol silicate complex.

Fig. 1
figure 1

The high-pressure cell and NMR circuit. a The schematic is of the pressure cell alone, where the labels correspond to: (i) the beryllium copper (BeCu) outer shell, (ii) the inner BeCu cylinder, (iii) a tungsten-carbide (WC) spacer, (iv) the BeCu feedthrough with RF coil and Delrin® cap, (v) the BeCu anti-extrusion disc, (vi) the WC plunger, (vii) the BeCu pressure plate, and viii) the BeCu locknut. Pictures of the feedthrough (b), the pressure cell (c), and the NMR circuit (d) are shown, with the feedthrough taken apart to show detail. Scale can be gauged by the diameter of the smaller parts, which are 6.3 mm in diameter. The three-turn solenoid is 28 AWG BeCu wire


Silicate oligomers

A solution containing various oligomers of Si(IV) was prepared using 29Si-labeled SiO2, and spectra of this solution were initially collected in a high-resolution NMR spectrometer before compressing the sample. The solution contained the silicic acid monomer (Q0), dimer (Q12), cyclic trimer (Q23), cyclic tetramer (Q24), and prismatic hexamer (Q36), consistent with assignments in the literature. The monomer, dimer, and cyclic trimer are the dominant species in this solution, which can be seen in Fig. 2a. This solution sample was placed within the high-pressure cell and subjected to a series of experiments at ambient temperature where the pressure was increased stepwise to 1.8 GPa. The sample was then returned to ambient conditions and changes in the peaks were found to be reversible.

Fig. 2
figure 2

29Si NMR of the silicate oligomers. a High-resolution 29Si NMR spectrum of silicate oligomers in solution. Chemical diagrams modified from Kinrade et al.14 b 29Si NMR in the high-pressure NMR probe, with pressures ranging from ambient conditions up to 1.8 GPa. Note a reduction in the peak intensities of both the dimer (Q12) and the cyclic trimer (Q23)

In high-pressure NMR measurements (Fig. 2b), the monomer, dimer, and cyclic trimer are all detectable, while the cyclic tetramer and prismatic hexamer were undetectable. As pressure is increased, the signal for the cyclic trimer steadily diminishes to the point that it disappears at a nominal pressure of 0.9 GPa. The signal assigned to the dimer also diminishes relative to the monomer peak. The signals assigned to the dimer and cyclic trimer both return to their initial intensities once pressure is relieved. Uncertainties in the pressure estimates are ±0.25 GPa, unless otherwise indicated, and were determined as the 95% prediction interval of five separate calibrations (see Supplementary Figure 1).

Sugar alcohol complexes

A solution of 29Si labeled-SiO2 was created where the five-carbon sugar alcohol, xylitol ((2R, 3R, 4S)-Pentane-1,2,3,4,5-pentol), was doped into the system, similar to the composition used by Kinrade et al.17 The NMR spectrum of this solution was first collected using a high-resolution NMR spectrometer at ambient pressures. This spectrum confirmed the presence of the four-coordinate silicic-acid species (includes both Si(OH)4o+SiO(OH)3-) as well as both the five-coordinate and six-coordinate Si-sugar alcohol complexes (Fig. 3a). This solution was then placed within the high-pressure cell and subjected to a series of pressures, up to 1.2 GPa. Higher pressures caused freezing of the solution, but this freezing was reversible upon release of the pressure and all peaks recovered.

Fig. 3
figure 3

29Si NMR of the silicon-xylitol complex. a A high-resolution 29Si spectrum of the xylitol-Si(IV) solution, where the four-coordinate, five-coordinate, and six-coordinate silicon species are seen in solution. b 29Si NMR spectra of the same solution using the high-pressure NMR probe, measured up to 1.2 GPa (the solution froze immediately above this pressure). Note that the peak assignable to silicic acid appears with increased pressure and then disappears upon return to ambient conditions

In the high-pressure series (Fig. 3b), the spectra show virtually no change in speciation of the silicon complexes at the lowest pressures (up to 0.2 GPa). Only peaks assignable to the five- and six-coordinate Si(IV) complexes were apparent. The peak assigned to silicic acid, which is visible in the high-resolution NMR spectrum at ambient pressure, has an intensity below the detection limit of the high-pressure probe. However, as pressure is increased, there is a change in the relative peak intensities. The intensities of peaks assigned to the five- and six-coordinate complexes decrease with pressure, and the peak assigned to silicic acid increases up to the highest pressure of 1.2 GPa. The appearance of the peaks is reversible and the peak assigned to silicic acid once again reduces below the signal-to-noise threshold when pressure is relieved. Peak separations are constant with pressure.

Catechol complex

A third solution consisting of dissolved 29Si labeled SiO2 was formulated with catechol (1,2-dihydroxybenzene) to create the tris-catecholate complex. An excess of catechol was used (Table 1). In this spectrum, a single peak was seen at −145 ppm (externally referenced to the silicic acid at −71 ppm), which provides good agreement with the listed chemical shift of −146 ppm found in Evans et al.34 The peak due to the tris-catecholate also happens to reside within the same chemical shift region as the six-coordinate sugar alcohols seen previously, consistent with the assignment of hexacoordinated Si(IV). This spectrum is shown as Fig. 4a.

Table 1 Solution compositions of the analyzed samples
Fig. 4
figure 4

29Si NMR of the tris-catecholate silicon complex. a A high-resolution 29Si NMR spectrum of the tris-catecholate-Si(IV) complex. b 29Si NMR spectra of the same solution using the high-pressure NMR probe, measured up to 1.5 GPa, where the solution was near freezing. Note that changes in the spectrum with pressure are reversible

This solution was then placed within the high-pressure cell and compressed, where the maximum pressure reached was 1.5 GPa. The 29Si NMR pressure series can be seen in Fig. 4b. At this pressure, the peak due to the tris-catecholate broadened, which experience has shown commonly indicates that the solution is close to freezing—as the viscosity of the solution increases, the correlation time τc decreases, which causes broadening of the peak35. When the solution was returned to ambient pressures, the initial peak shape returned.

Chemical interpretation

The pressure-induced speciation changes presented in Figs. 2 and 3 can be understood by following the movement of water between the bulk solution and the silicon species in the reaction. Generally, increases in the coordination numbers of the Si(IV)-sugar complexes are associated with release of water to the bulk solvent. Conversely, these larger silicate oligomers are then reduced to smaller species as bulk waters are packed back into the inner-coordination-sphere of Si(IV) as pressure is increased. Consider as an example the formation of a bis-diolato-hydroxo-structure with Si(IV) resulting in a Si(IV) coordinated by four oxo atoms from the two diolate anion and two hydroxides. This hypothetical six-coordinated Si(IV) is portrayed as: (L)2(SiOH)2, where L represents the diolate anion (-O–R–R–O-):

$${\mathrm{Si}}\left( {\mathrm{OH}} \right)_{4\left( {\mathrm{aq}} \right)}^0 + 2{\mathrm{H}}_2{\mathrm{L}}_{({\mathrm{aq}})} \rightleftharpoons \left( L \right)_2\left( {{\mathrm{SiOH}}} \right)_{2\left( {{\mathrm{aq}}} \right)} + 2{\mathrm{H}}_2{\mathrm{O}}$$

The trend toward smaller complexes with increased pressure is the reverse of this condensation reaction, and is consistent with results presented in Figs. 2 and 3. The volume change due to the elimination of a bulk water molecule during hydrolysis is negative and drives this trend towards a reduction in Si(IV) coordination number. Water molecules are packed more efficiently into coordination sphere of Si(IV) than in bulk solution. Thus the change in volume to eliminate a bulk water (V° ~ −18 cm3mol−1) by packing it around Si(IV) overwhelms any increase in volume from elimination and solvation of the ligand.

Next, consider the last set of experiments involving the catecholate-Si(IV) complex. The hypothesis motivating these experiments is that pressure would favor bis-catecholate or mono-catecholate species as these would pack water more effectively than the bulk solvent and be favored at pressure:

$${\mathrm{Si}}\left( {\mathrm{cat}} \right)_{3\left( {\mathrm{aq}} \right)}^{2 - } + 2\,{\mathrm{H}}_2{\mathrm{O}} + 2\,{\mathrm{H}}_{\left( {{\mathrm{aq}}} \right)}^ + \rightleftharpoons {\mathrm{Si}}\left( {{\mathrm{cat}}} \right)_2\left( {{\mathrm{OH}}_2} \right)_{2({\mathrm{aq}})} + {\mathrm{H}}_2{\mathrm{cat}}_{({\mathrm{aq}})}$$

Note that the coordination number of six around Si(IV) is preserved by the ligand-exchange reaction and such a bis-catecholate complex has been crystallized by Hahn et al.36 However, our experiments show this hypothesis to be refuted, at least at slightly excessive catechol concentrations—there are no aqueous hexacoordinated catecholate complexes of 29Si evident save for the familiar tris-catecholate complex. If some of the oxo ligands bridging to Si(IV) from the catecholate were converted to bridging hydroxo ligands, the number of protons and waters in reaction (3) would change, but not the general trend. Similarly, a competing reaction that reduces the coordination number of Si(IV), while still heavily favored by increases in pressure, would cause silica precipitation:

$${\mathrm{Si}}\left( {{\mathrm{cat}}} \right)_{3\left( {{\mathrm{aq}}} \right)}^{2 - } + 2\,{\mathrm{H}}_{\left( {{\mathrm{aq}}} \right)}^ + + 4\,{\mathrm{H}}_2{\mathrm{O}} \rightleftharpoons {\mathrm{Si}}\left( {{\mathrm{OH}}} \right)_{4({\mathrm{aq}})} + 3{\mathrm{H}}_2{\mathrm{cat}}_{({\mathrm{aq}})}$$

If this were to occur, the solubility of SiO2(s) is less than ~10−3 molal at ambient conditions and near-neutral pH, so destruction of the tris-catecholate-Si(IV) complex by pressure would lead to diminished 29Si signals, which we do not observe in the high-pressure NMR spectra. Instead we observe a clear constant signal that is assignable to the tris-catecholate-Si(IV) complex alone.

Of course, the reaction volume for these Si(IV)-complexation reactions reflects all changes in solvation, including solvation of the released ligands. However, the data described above consistently indicate that dominant term is the uptake or release of bulk waters by the condensation reaction (1), above. (Calculations to identify reasonable stoichiometries based upon energetics are included in the Supplementary Table 1 and visualized in Supplementary Figure 2.)


High-pressure 29Si NMR data indicate that the condensation reactions that lead to the complexation of aqueous silicate are reversed at elevated pressure. At pressures greater than a GPa, it seems that silicate speciation is determined by forming aqueous complexes that pack water more efficiently. These pressures correspond to a column of water reaching to the deepest part of the Earth’s crust; the NMR data were not, however, collected at the elevated temperatures of this region but such design changes are underway. In these regions, reactions between mobile fluids and silicate minerals redistribute mass and alter mineralogical composition. While the reactions of silicon deep in the Earth’s crust are done at hydrothermal temperatures, the ambient data presented herein suggest that much of the pressure-induced speciation change is attributable to simple transfer of waters into, and out of, the inner-coordination spheres of Si(IV).


High-pressure NMR

The high-pressure NMR probe uses a beryllium copper body (BeCu) with a tungsten carbide (WC) plunger and spacer combination to allow for pressures up above 2.0 GPa. The diameter of the probe is 39 mm and is designed for a narrow-bore magnet system. The circuit is an LC-circuit coupled with a three-turn microcoil made of 28-AWG BeCu wire, where the tuning and match capacitors can be exchanged depending on the nuclei studied. The quality factor of the circuit at 79.48 MHz is Q = 23.7.

The aqueous sample is held within a small portion of PEEK tubing, which is glued in place within the microcoil. The coil is immersed in fluorocarbon oil (Halocarbon 6.3) as the pressure transmission fluid and the probe is assembled as in Fig. 1. A hydraulic press is used to apply pressure to the sample while a lock-nut is screwed in place to lock in pressure. The gauge pressure of the press has been calibrated to the internal pressure of the sample via the technique of ruby fluorescence, measured from a spherical ruby within the sample cell (See Supplementary Methods). This ruby is coupled to a fiber-optic cable and the fluorescence is measured via an external optical spectrometer37.

The high-powered shim stack was designed by Resonance Research, Inc., and is designed for use inside a wide-bore superconducting magnet. The water-cooled shim coils have been specifically designed to provide higher power (up to 5 Amps per channel) for shimming this probe. The channels available are Z0, X, Y, Z, Z2, Z3, X2Y2, and 2XY. The probe is tuned down to 2H (ωRF = 61.40 MHz at 9.4 T) and shimmed using the D2O peak in all samples. All high-pressure measurements were carried out using a wide-bore Oxford superconducting magnet (9.4 T) coupled to a Bruker DRX console. 29Si (ωRF = 79.47 MHz at 9.4 T) measurements for all three samples used a single-pulse “zg” experiment with a smaller RF tip-angle38. The peaks in the high-pressure spectra were referenced using the high-resolution spectra, with the Q0 peak set at −71 ppm. High-resolution spectra were measured using a Bruker DRX500 system (11.75 T) using a 5 mm PABBO broadband probe. Typical quality factors for the resonance circuit are 20, but can of course fluctuate as each experiment is run with a different coil assembly.

Details for each experiment are: (i) 29Si oligomer solution: A 30° Ernst angle was used to create a pulse length of 3.6 µs at 34 W in power (3.5 dB using a Bruker BLAX300 amplifier). The smaller tip angle was used to reduce the acquisition time per scan as the longest measured T1 was 6.6 s. Total acquisition time per scan (AQ + D1) was 1.0 s, and 50,000 transients were measured, resulting in ~14 h experiments. (ii) 29Si sugar alcohol (xylitol) solution: A 30° Ernst angle was used to create a pulse length of 4.8 µs at 34 W in power (3.5 dB using a Bruker BLAX300 amplifier). The smaller tip angle was used to reduce the acquisition time per scan as the longest measured T1 was 7.0 s. Total acquisition time per scan (AQ + D1) was 1.0 s, and 30,000 transients were measured, resulting in ~8 h experiments. (iii) 29Si catecholate solution: A 15° Ernst angle was used to create a pulse length of 2.0 µs at 34 W in power (3.5 dB using a Bruker BLAX300 amplifier). The smaller tip angle was used to reduce the acquisition time per scan as the T1 of the catecholate complex was 60.6 s. Total acquisition time per scan (AQ + D1) was 2.1 s, and 26,000 transients were measured, resulting in ~16 h experiments.

Solution Preparation

The three silicate solutions were made using 98.7% isotopically enriched SiO2 (Isoflex, USA) digested in NaOH and D2O using a Teflon-lined stainless-steel Parr reactor vessel at 80 °C for 12 h. The resulting solutions were clear and free of colloids. For the solution containing only silicate oligomers (Solution 1), nothing more was added. For solutions 2 and 3, xylitol and catechol were added after the solution was allowed to cool to room temperature. No other adulteration was done to the solutions. The exact solution composition is included in Table 1. The resulting solutions were stable for weeks with no sign of gelling. The solution containing catechol was made under an inert argon atmosphere, where the initial solution had a slight rose tint. However, the solution began to darken slowly, even after sealing the sample in the argon environment, covering with foil, and placing in the refrigerator. Initial high-resolution 29Si NMR was run on the fresh solution and eventually compared to the NMR of the dark brown solution—no difference between the two spectra was seen.