A Low Cost & Open Access System for Rapid Synthesis of Large Volumes of Gold and Silver Nanoparticles

Rapid synthesis of nanomaterials in scalable quantities is critical for accelerating discovery and commercial translation of nanoscale-based technologies. The synthesis of metal nanogold and silver in volumes larger than 100 mL is not automatized and might require of the use of harsh conditions that in most cases is detrimental for the production of nanoparticles with reproducible size distribution. In this work, we present the development and optimization of an open access low-cost NanoParticle Flow Synthesis System (NPFloSS) that allows for the rapid preparation of volumes of up to 1L of gold and silver nanoparticle aqueous solutions.


Introduction
The application of nanoparticles has rapidly emerged in a variety of elds spanning from molecule sensing, 1 drug delivery, 2 nuclear targeting, 3 biosensing, 4 to transfecting agents. 5 Furthermore, the inclusion of nanoparticles in biomimetic matrices can provide a variety of properties such as antimicrobial ability 6,7 and electrical conductivity. 8 However, nanoparticle synthesis has intrinsic batch-tobatch variability in particle diameter polydispersity and surface composition that contributes to discrepancies in nanoparticle performance and activity. 9 Metal nanoparticle synthesis is normally conducted using chemical reduction in batch reactor setups. 10 The photochemical reduction of metal ions to produce nanoparticles is an interesting alternative route for producing nanostructures from metals due to its simplicity and compatibility with a wide variety of capping agents. Both gold and silver nanoparticles can be synthesized from the photoinitiated reduction of HAuCl 4 via the Norrish Type I cleavage reaction of Irgacure 2959 11,12 in an aqueous solution. 13 However, scaling up photochemically prepared nanoparticles remains a challenge particularly when considering limited UVA light penetration in large containers.
Flow-based chemistry is becoming increasingly popular due to its advantages over batch chemistry such as reduced variability 14 Previous work on ow irradiation suggests that particle size and polydispersity can be adjusted simply by altering the ow rate or irradiance intensity. 15 However, reactor fouling remains a serious problem with ow-based nanoparticle synthesis. Tightly controlled conditions containing alkali metals can alleviate fouling, 16 but use of other metals leads to nanoparticle contamination and their implementation in a commercial photoreactor can be prohibitively expensive.
Herein, we report a low-cost open access system called the Nanoparticle Flow Synthesis System (NPFloSS) that utilizes LED-UVA light sources to generate a continuous ow of nanoparticles. NPFloSS produces batches of metal nanoparticles within minutes, which is much faster than the normal photoreactor. Furthermore, the use of individual LEDs considerably lowers the cost, increases the system lifetime, and reduces its physical space requirement when compared to commercial photoreactors.

NPFloSS development
The NPFloSS consists of three main simple components: (1) a ow line driven by a peristaltic pump containing a UV transparent quartz cell (recycled from a decommissioned HPLC), (2) two 365 nm UVA LEDs, and (3) two heat sinks (Fig. 1). Assembly instructions and additional details are available in the tutorial Video S1 and in the ESI. In the NPFloSS, the input tubing leads to a peristaltic pump that ows the reactants into a quartz ow cell, which is housed within a 3D printed compact UVA irradiation system containing two LEDs to provide high intensity irradiation (up to 14 mW/cm 2 ). The LEDs' temperature is maintained using Alpine GT CPU fans. After UVA exposure, the products are generated and ow into the output collector. The 3D printed compact UVA irradiation case allows for benchtop setup and reproducibility while removing the need for additional components or holders. The system can be continuously run to generate + 100 mL of product nanoparticle solution an hour or 1.0 L of product if left overnight.

NPFloSS optimization and nanoparticle synthesis
Initial attempts at synthesis showed that in order to produce quality nanoparticles at an acceptable rate, high light intensity was critical. As such, two powerful LEDs were installed, and the effect of different light intensities evaluated (see Figure S1). The highest intensity (≈ 14 mW/cm 2 ) resulted in the best outcome for producing nanoparticles with the most intense plasmonic absorption at their respective maxima absorption (see Figure S1A). As higher intensity seemed bene cial, the LEDs were run at their maximum recommended operating current at 365 nm to produce a light dosage of 37 mJ/cm 2 for gold and 128 mJ/cm 2 for silver.
At the optimal light dosage, the LED circuit board was found to stabilize at an operating temperature of 33 °C while the irradiated solution inside the quartz cell reached a temperature of 31 °C after 120 seconds ( Figure S2). Note that the time the solution is exposed to 31ºC is less than 3 seconds for gold and 10 seconds for silver, which is enough time for the metal reductions to take place. Furthermore, maintaining the LED at this temperature allows for extending the LED lifespan.
At this stage of the NPFloSS development, nanoparticles could be made but reactor fouling was a major issue, which led to increasing the risk for line clogging and system failure. Attempts to reduce fouling were made by modifying several characteristics such as ow rate, pH, temperature, and capping agent. Ultimately, the use of 10 mM Tween-20 surfactant added to the ow mixture was found to be effective to prevent fouling. Tween-20 is a mild surfactant capping agent 17 that can be displaced by other capping agents and shows good compatibility with biomolecules. 18 Tween-20's anti-fouling action comes from its dual behaviour as both a surfactant and a capping agent that causes forming nanoparticles to better interact with the aqueous environment and remain in solution instead of building-up on the quartz. The protocol used for preventing fouling is shown in Scheme 1 (for further details see ESI).
With the operating conditions optimized (see Scheme 1), the stability of both "uncapped" and citrate capped gold nanoparticles was tested. Citrate capped nanoparticles showed better stability over a month than uncapped gold nanoparticles ( Figure S3). Representative absorption spectra for nanogold and nanosilver colloidal solutions are showed in Fig. 2A and B, respectively. The effect of citrate as a capping agent in the case of gold showed a net decrease in the hydrodynamic size from ≈ 60 nm to ≈ 25 nm for the highest concentration of citrate (Fig. 2C, p < 0.05 t-test, see Table S1 for stats summary). Zeta potential values remained mostly unchanged, while the maximum absorption wavelength in the visible (tau) blue shifted by ≈ 30 nm to shorter wavelengths at the highest citrate concentration used. This was also accompanied by a reduction on the spectra broadness or full width at half max (Fig. 2C). Nanoparticle sizes measured by TEM showed a statistically signi cant reduction of the nanoparticle sizes from 62 to 25 nm (p < 0.05, values were determined by one-way ANOVA using Holm's multiple comparison, see Table S2 for stats summary), at 5 and 50 citrate equivalents (Fig. 3).
For the case of nanosilver, uncapped synthesis was not possible. Fine tuning of the ow rate for synthesizing nanosilver was required (see ESI and Materials and Methods for details). Nanosilver was stable with negative zeta potentials when using citrate as a capping agent ( Fig. 2B and 2D), with particle hydrodynamic sizes that became considerably polydisperse at 5 and 50 equivalents of citrate (Fig. 2D).
Interestingly, a blue shift in the plasmonic absorption of ≈ 8 nm was observed upon increasing citrate concentration which suggests a decrease in the nanoparticle mean size. This was corroborated by TEM experiments showing that nanosilver was statistically smaller (< 5 nm) for the 5 and 50 equivalents of citrate compared to the 0.5 equivalents group ( Figure S4). The apparent change in the nanoparticle hydrodynamic diameters could have been an artifact of the DLS system when measuring smaller particles.
For Tween-20 capped nanogold there were no changes in the hydrodynamic sizes. However, there is a light increase in the zeta potential values of about ≈ 5 mV at 1 and 10 equivalents of the surfactant. In contrast, there is a marked blue shift of ≈ 10 nm in tau at larger Tween-20 equivalent numbers (Fig. 2C). FWHM remained mostly unchanged. TEM measurements, however, of individual nanoparticles prepared with 10 or 100 equivalents of Tween-20 showed signi cantly different values for the mean size (p < 0.05, determined by one-way ANOVA using Holm's multiple comparison, see Table S2). As for nanosilver, the Tween-20 did not change the nanoparticle hydrodynamic sizes. Similar data was observed for zeta potential, maximum absorption and FWHM (Fig. 2D). TEM images of the 10 and 100 equivalent Tween-20 capped silver nanoparticles also indicated that size was similar between groups ( Figure S4).
When Triton was used as a capping agent, at 10 equivalents there was a decrease in the hydrodynamic sizes to values closer to those observed for the 1 equivalent Triton (Fig. 2C). For tau, the most pronounced change was found for Triton, which blue shifted by ≈ 15 nm when increasing the concentration of the surfractant. FWHM values narrowed at 10 and 100 equivalents of the surfactant. The changes in the nanogold hydrodynamic sizes align well with the TEM ndings that showed a size decrease from ≈ 53 nm to 28 nm at 1 and 10 Triton equivalents (see Fig. 3, p < 0.05 Table S2, determined by one-way ANOVA using Holm's multiple comparison). Further increasing Triton concentration beyond 10 equivalents did not have a signi cant effect on the nanoparticle size.
Using smaller molecules such as HEPES, Lipoic acid, and the CLKRS peptide led to the production of stable nanoparticles with hydrodynamic and TEM sizes in the ≈ 10-60 nm range (Figs. 2C and 3). Lipoic acid coated nanoparticles showed a broader absorption spectrum, FWHM, with zeta potential values in the order of -12 mV, which is considerably lower than the other thiol containing capping agent (CLKRS) that displayed + 15 mV. This aligns well with the reported ndings of this short peptide as a superior capping agent for metal nanostructures. 19 Similar trends were observed for nanosilver and CLKRS (Fig. 2D). When using bovine serum albumin (BSA) as a capping agent for nanogold, nanoparticles with sizes in the 50 nm range were produced with strong positively charged surfaces (Fig. 2C), which resembles ndings by our group and others in the production of protein capped metal nanostructures. 6,7,20,21 TEM measurements for the BSA coated nanogold indicated considerably smaller nanoparticle diameters of ≈ 13 nm than those measured by dynamic light scattering. These differences might be related to the formation of a protein corona around the nanoparticle when in solution, a phenomenon that does not exist under the experimental conditions of TEM. 22 Discussion Flow systems have recently received a lot of attention and as a result systems ranging from simpli ed to highly automated have been produced to facilitate reagent mixing along with thermal, photochemical, and electrochemical stimulus. 23 The advent of ow photochemical systems has helped solve some of the problems leading to the under-utilization of photosynthetic routes such as speci c glassware requirements, scalability, safety, and technical knowledge. 24 Thus, to encourage use of promising photosynthetic techniques, their application should be made as simple and cost-accessible as possible.
Due to their wide variety of uses, there is an increasing demand for producing cost-effective methodologies for large scale synthesis of metal nanoparticles. In designing such systems, considerations for having reliable methodologies that encompass rapid capping agent replacement and batch-to-batch reproducibility are key. Flow reactors, such as NPFloSS, provide a cost-effective solution for use in academic or industrial settings.
Reactor fouling is the most common and serious issue when designing ow systems, as the metals tend to deposit on surfaces. In NPFloSS, we used low concentrations of Tween-20 that decreases the contact angle between quartz and water and thus increases the wettability of quartz while reducing interfacial tension. 25 Furthermore, Tween-20 acts as a stabilizing nanoparticle capping agent 26 while interacting with water through its hydrophilic domains, thus increasing nanoparticle-solution interaction. Through these dual functions, Tween-20 prevented reactor fouling and allowed the system to operate continuously without line blockage. As a biocompatible additive used in food and cosmetics that increases nanoparticle stability in biological media, 18 Tween-20 is a safe addition to biomedical nanoparticle formulations.
NPFloSS' photochemical reduction methodology may be extended to allow for ow synthesis of other nanoparticle types. Further improvements of NPFloSS could include operation inside a glovebox for synthesis of oxygen sensitive nanoparticles like copper 27 or its use as an in line system for dual photochemical synthesis of nanoparticles and photocatalysis. Also, the use of a bidirectional inlet system for in situ mixing in the "reaction cell" could be used for the preparation of core-shell nanoparticles.

Conclusions
We developed, optimized, and validated a low-cost and open access ow reactor for rapid synthesis of gold and silver nanoparticles: the NPFloSS. The practicality of this device allows for rapid synthesis of stable nanoparticles whose surfaces can be readily protected with capping agent molecules such as citrate, Triton X-100, Tween, HEPES, lipoic acid, peptides such as CLKRS, and bovine serum albumin. The open access and low-cost features of NPFloSS make this instrument attractive for use by anyone to expand its nanoparticle synthesis potential to other metals and capping agents.

CAD modeling and 3D printing
All 3D printed components were designed using CAD software (Fusion 360, Autodesk Inc) and printed using an Ultimaker S5 with ABS lament. The instructions and tutorial video for assembling the system are made available in the ESI of this article.
Other components 2.0 mm Te on tubing was used to connect all components. A Manostat cassette pump was used to generate ow at a rate of 0.5-1.8 mL/min. Arctic Alpine 64 GT CPU fans were used as a heat sink for two 365 nm LEDs (LZ4-V4UV0R, Mouser) running at 2.8A 4.0V powered by two adjustable power sources (Yihua PSN-305D). A quartz cell was used as the irradiation chamber.

Light irradiance calculations
Radiance was directly measured at the ow cell's position using a Luzchem L-0487 power meter. The obtained reading in lux was converted to W/m 2 power for 365 nm light using the conversion factor 0.0027. The total light dosage was obtained by multiplying the power by residence time.

Temperature measurement
A thermal probe was placed in contact with the LED starboard and the temperature was continuously measured until thermal equilibria had been reached and for 10 minutes thereafter. Solution temperature was measured by owing water into the quartz cell, turning on the LEDs, and then quickly extruding the water onto a thermal probe after a given time.

Absorption spectra
To characterize the nanoparticles, a UV absorbance spectrum from 350 nm to 750 nm was taken using a SpectraMax M2e with 2 nm steps. Three independent samples were prepared for each type of nanoparticle/capping agent. The plotted data in this article corresponds to the mean of those measurements.
Dynamic light scattering and zeta potential measurements Hydrodynamic sizes and zeta potential measurements were carried out in a Malvern Zetasizer Nano ZS at 20 o C in 1.0 cm pathlength disposable plastic cuvettes. Reported values correspond to the average of three independent batches, each measured in triplicate.

Nanoparticle stability assays
Stability tests for nanogold particles was carried out by measuring the absorbance spectra from 350 nm to 750 nm i a SpectraMax M2e with 2 nm steps immediately after synthesis (day 0), a day after synthesis (day 1), 7 days after synthesis (day 7), and 28 days after synthesis (day 28).

Transmission electron microscopy
Samples for electron microscopy were prepared by delivering ~ 10 µL of solution to carbon-coated copper grids (400 mesh) and dried in a vacuum system for three days. Electron microscopy images were taken in a FEI Tecnai F20 G2 FE-TEM, operating in the transmission mode (TEM) at 200 kV. Nanoparticle mean size was calculated from TEM imaging by using ImageJ software 28 to manually measure 100 individual particles from different areas in the grid.

Statistical analysis
Unless otherwise indicated, p-values were calculated using t-test or one-way ANOVA in KaleidaGraph 4.5 software. For t-test a p < 0.05 was considered as statistically signi cant. For the one-way ANOVA, a Holm's multiple comparison analysis was performed. Declarations