Ruthenium nitrosyl nitrate solution (Ru(NO)(NO₃)₃, 1.5 w/v% in dilute nitric acid), alkaline sodium borohydride solution (NaBH₄, ∼12 wt% in 14 M sodium hydroxide), powdered sodium borohydride (NaBH₄, ≥98.0%), sodium hydroxide solution (NaOH, 0.1 M, 95–100%), powdered sodium hydroxide (NaOH, 97%), polyvinylpyrrolidone powder (PVP, average M_r 40 000), and bovine serum albumin powder (BSA, lyophilized, ≥96%) were purchased from Sigma Aldrich. Nitric acid solution (HNO₃, 1 M) was purchased from Fisher Scientific. All chemicals were used without further purifications within 3 months of purchase to ensure fresh solutions free of degradation. Ultra-high-purified 18.2 MΩ cm MilliQ water was used in preparing all solutions and washing.

Ruthenium nanoparticles (Ru NPs) were synthesised via wet reduction of ruthenium precursors (Ru(NO)(NO₃)₃ and RuCl₃) using NaBH₄ as a reducing agent. In a typical flow synthesis, 2.5 mM Ru(NO)(NO₃)₃ and 6.25 mM of alkaline NaBH₄ aqueous solutions were each fed at 115 mL h⁻¹ by a Chemyx Fusion 6000 syringe pump through a continuous coiled flow inverter reactor (CFIR) (Fig. 1).

The precursors entered a 0.02-inch T-mixer in a crossflow configuration, where the Ru-containing stream entered the system perpendicular to the resulting direction of flow. The resulting solution passed through the CFIR, consisting of 3.16 m of perfluoro alkoxy (PFA) tubing (0.02-inch inner diameter and 1 cm outer diameter) coiled around a 3D printed support (printed using a Formlabs Form 2 3D stereolithographic printer with a high temperature resin and 0.1 mm resolution). Both the mixer and the reactor were immersed in a water bath maintained at 25 °C. Steady-state conditions were assumed after 2 min of operation, after which samples were collected. For post-synthesis characterisation, 5 mL of Ru NPs were then collected in a glass vial containing 0.56 mL of 1 M HNO₃ solution (final HNO₃ concentration of 0.1 M) to instantaneously decompose any excess of NaBH₄ to prevent any possible further reaction, and to provide electrostatic stabilisation.

For comparison, a batch synthesis method was used whereby the same precursor solutions were simultaneously added into a round bottle flask maintained at 25 °C with a 500 rpm magnetic stirrer. The solutions reacted during 10 s (emulating the residence time in the standard flow synthesis) before the addition of nitric acid to quench the reaction (final HNO₃ concentration of 0.1 M). In an initial batch synthesis screening of different molar ratios of NaBH₄ : Ru, 2 mL of each precursor solution was ejected simultaneously from pipettes into a glass vial at room temperature. The pH values of the precursors were measured, and the pH values of mixed solutions were measured 10 s after mixing, after which UV-vis spectrums were obtained for the solutions that did not display immediate large agglomeration.

pH measurements were carried out with a Fisherbrand™ accumet™ XL150 pH Benchtop Meter with a Tris-compatible double junction refillable glass-body pH electrode. Synthesised Ru NP suspensions were analysed by Dynamic Light Scattering (DLS) to estimate average size and zeta potential (ZP) to assess colloidal stability using a Malvern Zetasizer Nano S90 instrument equipped with a 633 nm laser. Ultraviolet-visible (UV-vis) spectrometry was conducted with an Agilent Cary-60 UV-vis spectrometer. Transmission Electron Microscopy (TEM) was conducted with a Talos, FEI microscope (200 kV), with samples prepared by drop-casting 3 μl of colloidal Ru NP suspension onto a carbon-coated copper TEM grid (400 mesh, Agar Scientific) and drying naturally for 30 min. Particle size distribution histograms were constructed using ImageJ open-source software using a minimum of 200 nanoparticles from at least 3 TEM images taken from different mesh sections of the TEM grid. Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) was used to measure the concentration of Ru in solution using a Thermo-Fisher Nexion 2000-S machine.

The speciation of Ru and NaBH₄ are both dependent on pH. Ru³⁺ is thermodynamically stable in very acidic pH values below 1.8, above which it can hydrolyse forming Ru(OH)₃.⁴⁹ Hence, Ru(iii) salt solutions are often stored under acidic conditions to maintain stability. Although there exists pH speciation studies for Ru³⁺ from RuCl₃,⁴⁹ the pH-speciation of Ru(NO)(NO₃)₃ has not been documented to the best of our knowledge. Ru(NO)(NO₃)₃ contains a nitrosyl (NO) group that is π back bonded to Ru.⁵⁰ This is such a stable bond that the Ru–NO(iii) group often behave like a single element.⁵¹ Indeed, we believe that the nitrosyl group (NO) is responsible of the small Ru NP observed as its bond to the Ru atom stabilises it against agglomeration during the nanoparticle synthesis. To prove this theory, palladium nanoparticles were synthesised using a counterpart K₂[Pd(NO)(NO₂)₄(NO₃)] leading to 4.9 ± 0.9 nm particles in a batch synthesis (Fig. S2†), comparable sizes to the ones achieved with Ru(NO)(NO₃)₃ precursor in batch (Fig. 2c).

To investigate the strength of the nitrosyl ligand in the Ru(NO)(NO₃)₃, its hydrolysis was investigated under different pH values from 3.0 (after dilution to 2.5 mM) to 12.6 using NaOH. No colour change was observed across all pH values, nor the presence of particles by DLS. UV-vis measurements in Fig. S3a† reveal the appearance of an absorbance shoulder at ∼380–400 nm on the precursor spectrum, which increases in magnitude as pH increases, indicating changes in the structure of the Ru(NO)(NO₃)₃ complex. The nitrate (NO₃⁻) groups can be easily replaced by OH⁻ ligands upon addition of NaOH, however, there is not a complete Ru hydrolysis to form oxide particles or precipitates. This likely suggests the formation of a new complex as pH increases, whereby the stable nitrosyl (NO) group is hydrolysed into a nitrite group, as reported for other alkyl, amino, and cyano ruthenium nitrosyl complexes.^50–52 Such partial hydrolysis of Ru(NO)(NO₃)₃ in basic conditions is partially reversible as the new UV-vis absorption shoulder at ∼380–400 nm partially relaxes over 24 hours (Fig. S3b†) although it does not fully return to its initial spectrum. Visually, a change in the colour of the solution is not apparent to the naked eye even for the most alkaline Ru(NO)(NO₃)₃ solution (Fig. S3c†). These observations are opposite to those for RuCl₃ (ref. 53) or other metal nitrates⁵⁴ which hydrolyse and form metal oxides particles or precipitates in basic pH conditions.

To investigate the effect of the reducibility of Ru(NO)(NO₃)₃ speciation at different pH values, the pH of 2.5 mM Ru(NO)(NO₃)₃ precursor solution was first varied between 1.8 to 12 and subsequently 6.25 mM NaBH₄ solution (aqueous solutions of powdered NaBH₄ in the absence of NaOH) was added (note that these conditions are similar to those in the flow synthesis above). Observations are summarised in Table 1. The most alkaline Ru(NO)(NO₃)₃ solution (pH 12) experienced no colour change within 20 min of NaBH₄ addition with DLS results showing poor repeatability, substantially low count rates, and unstable raw correlogram plots due to low Ru conversion. However, agglomerates were observed after 20 h, suggesting that under alkaline conditions, the reduction reaction is very slow. Similar observations took place on the other side of the pH spectrum. The most acidic Ru(NO)(NO₃)₃ solution (pH 1.8) experienced no subsequent colour change after 20 min of NaBH₄ addition and showed unstable DLS results, also indicating low Ru conversion. No agglomerates or colour change were observed even the following day, likely because the very acidic conditions instantly hydrolytically decomposed the NaBH₄ solution, preventing any further reduction, as well as the high stability of Ru(NO)(NO₃)₃ under these conditions as previously discussed.

The Ru(NO)(NO₃)₃ solutions at pH values between 6.3–9.7, experienced a delay of ∼20 min for the colour change after NaBH₄ addition with formation of relatively large particles (<150 nm) according to DLS. These suspensions showed repeatable DLS measurements across 3 scans (Fig. S4†) with stable raw correlogram plots, indicating the formation of stable NPs. On the other hand, the Ru(NO)(NO₃)₃ solutions with acidic pH values (3.3 and 5.4) showed immediate colour changes from light yellow to brown-grey upon NaBH₄ addition, indicating faster reduction of the Ru precursor under these acidic conditions.¹¹ However, agglomerates observable with a naked eye formed rapidly from the Ru solution of pH 3.3 (Fig. S4†).

These results suggest that the partial hydrolysis of Ru(NO)(NO₃)₃ with NaOH under basic conditions form Ru complexes with a slow reduction by NaBH₄. However, NaOH is still needed to prevent the immediate hydrolytic decomposition of NaBH₄ in acidic conditions. Therefore, a compromise in the NaOH concentration is required to achieve fast reduction for the formation of small NPs.

For comparison purposes, RuCl₃, a more common precursor in the literature,^11,13–15 was used for the synthesis of Ru NP colloids via NaBH₄ reduction. The reduction of 2 mM RuCl₃via dropwise addition of NaBH₄ solution, in the absence of ligands, led to significant precipitation as soon as the pH rose above 5 (Fig. S5†), in contrast to previous observations.¹³ Attempts to maintain the pH of 2 mM RuCl₃ under acidic conditions (through HCl addition) to prevent the aforementioned precipitation were also unsuccessful, resulting in initially large DLS sizes and significant agglomeration after 24 h (Table S2†).

The precursor concentrations used in the flow synthesis were selected after conducting an initial 6 point-screening in batch mixing conditions to investigate the effect of NaBH₄ : Ru molar ratios. The Ru(NO)(NO₃)₃ concentration was kept constant at 1.25 mM and the concentration of NaBH₄ (prepared from NaBH₄ in 14 M NaOH solution, with an NaBH₄ : NaOH ratio of 4.4) was varied to study NaBH₄ : Ru molar ratios in range of 1.5–3 – a sensitive region identified by preliminary tests. The reduction and colour change of the Ru precursor was followed with UV-vis spectroscopy. The formation of any agglomerates visible by the naked eye was recorded. Initial and final pH's upon mixing the precursors were measured. Results are summarised in Fig. 3 and Table 2.

At low NaBH₄ : Ru ratios of 1.5, 1.8 and 2, minor colour changes are observed which indicate incomplete reduction of the Ru precursor (Fig. 3). This is likely due to the instant hydrolytic decomposition of the NaBH₄ reducing agent, as indicated by the acidic final pHs for these molar ratios of 2.4, 3.5 and 5.3, respectively (Table 2). At medium NaBH₄ : Ru ratios of 2.3 and 2.5, a colour change from light yellow to brown-grey (Fig. 3) is observed within seconds of mixing, indicating reduction of Ru(iii) to Ru(0).¹¹ No agglomerates are observed with the naked eye. The corresponding UV-vis spectra for NaBH₄ : Ru ratios of 2.3 and 2.5 show a shoulder between 280–300 nm, characteristic of OH– anions stabilising Ru NPs^18,21 – which agrees with the final pH becoming basic at exactly these ratios (Table 2).

At the highest NaBH₄ : Ru ratios of 2.8 and 3, immediate large agglomerates are observed for the ratio of 2.8 but not for the ratio of 3 (Fig. 3). For the highest NaBH₄ : Ru ratio of 3, the reduction of the Ru precursor is likely inhibited by the formation of the stable Ru nitrite complex under the basic conditions (final pH of 11.0, Table 2), as discussed in the previous section on the pH speciation of the Ru precursor. This reveals that the Ru complex formation occurs faster than reduction at these conditions.

Based on these initial screenings, a NaBH₄ : Ru molar ratio of 2.5 is selected for flow synthesis to achieve sufficient conversion of the Ru precursor while avoiding the formation of agglomerates or the stable Ru nitrite complex. Both NaBH₄ : Ru molar ratios of 2.3 and 2.5 are deemed suitable, however, 2.5 is selected based on ensuring that there is an excess of NaBH₄ to effectively reduce the stable Ru nitrite complex which forms in the presence of NaOH. This ratio of 2.5 is substantially larger than the stoichiometric ratio of NaBH₄ : Ru of 0.375 for the reduction of Ru³⁺.

The relatively fast reduction of Ru(NO)(NO₃)₃ by NaBH₄ at a pH of 9.5 and the differences in Ru particle size in batch and flow reactors indicate that mixing time (i.e. the time required to achieve homogeneous mixing between both reactant streams) has a similar scale to the reaction time, an aspect usually overlooked in material synthesis.⁵⁵ In these cases, a fast-mixing rate is critical to trigger a homogeneous nucleation of Ru NP in the whole reactor volume leading to narrow size distributions.⁵⁶ Flow reactors are particularly suited to evaluate transport phenomena effects as their fluid dynamics are well-defined and can be manipulated easily by simply varying their design and operating conditions. To understand these effects, Ru NPs were synthesised in flow systems under the same conditions (1.25 mM Ru(NO)(NO₃)₃ and 3.13 mM NaBH₄, pH upon mixing: 9.5, 25 °C), but with varying reactor design parameters and operating conditions as summarised in Table 3, alongside resulting Ru particle size and distribution (TEM). Further DLS and zeta potential characterisation is provided in Table S3.† The disparity between the DLS and TEM results in Table S3† are because the DLS measurements have a low limit of particles size of ∼10 nm.⁵⁷ DLS measures the hydrodynamic sizes of particles, being particularly sensitive to the presence of large particles (a sixth-power relationship).^44,58 Hence, DLS for these sub 4 nm Ru NPs should only be used qualitatively, as an initial indication of the presence of a NP colloidal suspension before conducting TEM to obtain more accurate quantitative results.

In each size control experiment, the Ru NPs were electronically stabilised in the presence of HNO₃ in the receiving outlet vial (0.1 M final concentration). The acidic conditions also quenched the reaction by rapidly decomposing any NaBH₄ remaining in the solution. Keeping constant the flow system configuration and decreasing the total precursors flowrate from 230 to 115 ml h⁻¹ leads to an increase in particle size from 2.9 ± 0.4 to 3.9 ± 0.5 (experiments 1 and 2, Table 3, Fig. 4). Further decrease of the total flowrate to 20 ml h⁻¹ does not lead to further increases of sizes but instead a considerably broader size distribution, 4.0 ± 1.4 nm (experiment 3, Table 3, Fig. 4). This relative upper limit of ∼4 nm agrees with other observations in the literature using NaBH₄ as a strong reducing agent.^11,13–15 Increasing total flowrate has two convoluted effects on mixing. On one hand, it increases the early mixing in the T-mixer promoted by a higher level of engulfment of the streams.⁵⁹ On the other hand, it also promotes a higher level of mixing in the reactor itself by increasing the magnitude of the Dean vortexes formed in the coiled flow inverter reactor. Dean vortexes consist of the rotation of the fluid enhancing the radial mixing. They are created by the centripetal force experienced by fluids flowing inside curved reactor. The dimensional Dean number provides a quantification of the difference forces acting in the fluid and represent the magnitude of the vortices.⁴¹ As the Dean number is directly proportional to the Reynolds number, an increase of the flowrate increases the magnitude of the vortices and thus, the mixing in the reactor (within laminar flow).

To decouple the mixing effects in the T-mixer and the reactor, the total flowrate (230 mL h⁻¹) and reactor configuration (CFIR) was fixed, while the internal diameter of the T-mixer was varied between 0.01 and 0.05′′ (experiments 1, 4 and 5, Table 3, Fig. 5). The Ru NPs sizes increased from 2.1 to 2.9 and 3.1 nm as the T-mixer internal diameter increased from 0.01 to 0.02 to 0.05′′ respectively. In all cases, a similar size distribution of ±0.3–0.4 nm was measured (Fig. 5). As the fluid dynamics in the reactor is the same in the three experiments, with constant Reynolds and Dean numbers (179 and 40 respectively), any differences in size can only be attributed to the early mixing in the T-mixer, within the first 50 ms. As a result, it can be concluded the faster the mixing rate (i.e. rate at which homogeneous mixing is achieved), the faster the average nucleation rate which leads to a higher number of nuclei and consequently a smaller resulting size of Ru NPs. Similar conclusions have been confirmed using computational fluid dynamic simulations for gold nanoparticles³⁷ and perovskite nanocrystals.⁶⁰

On the other hand, keeping constant the flowrate (230 mL h⁻¹) and T-mixer (0.02′′) but varying the curvature of the coiled reactor from 1 cm to 10 cm (experiments 1 and 6, Table 3) led to almost negligible differences in the resulting particle size and distribution (2.9 ± 0.4 nm and 3.0 ± 0.5 nm respectively, Fig. S6†) despite the considerable differences in Dean number (40 and 13 respectively) and thus, radial mixing in the reactor. It is important to note that the flow inversions were omitted in the latter experiment to further minimise any radial mixing. The high early mixing rate promoted at high Reynolds number in the T-mixer likely leads to a homogeneous distribution of nuclei across the reaction volume promoting homogeneous growth in the reactor, independent on any further passive mixing during this stage (>50 ms). Further comparison of experiments 2 and 3 (Table 3) where the synthesis of Ru NPs were carried out in the same flow set-up but with decreasing flowrates (58 and 10 mL h⁻¹ leading to 3.9 ± 0.5 nm and 4.0 ± 1.4 nm respectively) demonstrates that in order to achieve a homogeneous growth in the reactor and thus, narrow size distribution, it is critical to have previously achieved a high level of mixing during the nucleation stage. It must be noted that due to the absence of stabilisers, there is also the possibility that the inner walls of the reactor function as sites for heterogeneous nucleation of the NPs. Further research is needed to investigate this synthetic avenue.