A frugal implementation of Surface Enhanced Raman Scattering for sensing Zn2+ in freshwaters – In depth investigation of the analytical performances

Surface Enhanced Raman Scattering (SERS) has been widely praised for its extreme sensitivity but has not so far been put to use in routine analytical applications, with the accessible scale of measurements a limiting factor. We report here on a frugal implementation of SERS dedicated to the quantitative detection of Zn2+ in water, Zn being an element that can serve as an indicator of contamination by heavy metals in aquatic bodies. The method consists in randomly aggregating simple silver colloids in the analyte solution in the presence of a complexometric indicator of Zn2+, recording the SERS spectrum with a portable Raman spectrometer and analysing the data using multivariate calibration models. The frugality of the sensing procedure enables us to acquire a dataset much larger than conventionally done in the field of SERS, which in turn allows for an in-depth statistical analysis of the analytical performances that matter to end-users. In pure water, the proposed sensor is sensitive and accurate in the 160–2230 nM range, with a trueness of 96% and a precision of 4%. Although its limit of detection is one order of magnitude higher than those of golden standard techniques for quantifying metals, its sensitivity range matches Zn levels that are relevant to the health of aquatic bodies. Moreover, its frugality positions it as an interesting alternative to monitor water quality. Critically, the combination of the simple procedure for sample preparation, abundant SERS material and affordable portable instrument paves the way for a realistic deployment to the water site, with each Zn reading three to five times cheaper than through conventional techniques. It could therefore complement current monitoring methods in a bid to solve the pressing needs for large scale water quality data.

1 Transduction and recognition properties of the sensor Figure S1: Absorption spectra of Lee-Meisel Ag NPs upon interaction with spermine as a function of time. The spectra were obtained by mixing in a quartz cuvette 100 µL of as synthesised NPs with 10 µL of a 0.15 mM spermine solution and subsequently adjusting the total volume to 2 mL. The stock solutions were adjusted to a pH of 7 prior to mixing. The dashed line indicate the wavelength of the laser used for Raman acquistions.   (left). Note that at a pH of 7 and within the investigated Zn concentration range, the free ligand exists in two main forms (three times and four times deprotonated respectively) and the bound ligand exists in two forms with one or two chelated Zn 2+ ions respectively. Here the free receptor and bound receptor molar fractions represent the sum of the free forms and bound forms respectively. The speciation is largely unaffected by pH fluctuations about 7 (right).
± .  2 Chemometric analysis Figure S4: Schematics of information that can be derived from validation plots, also called response plots.

Estimation of cost per analysis
The cost of performing one spectral acquisition has two main contributors: the consumption of stock solutions of NPs, spermine, XO and Zn standard, and the labour time needed to prepare the sample in the cuvette and record the corresponding spectrum. We estimate below the costs of these contributors.

Consumption of stock solutions of NPs, spermine, XO and Zn standard:
Consumption of chemicals from commercial supplier:

Labour for the preparation of the stock solutions:
The synthesis of the nanoparticles at a scale of 500 mL is completed in one hour, within which the settingup of the reaction mixture only takes the first 10 min. During the remaining 50 min of NP growth, the stock solutions of spermine, xylenol orange and zinc nitrate can be prepared and adjusted in pH. Reasoning in terms of PhD labour cost, these operations amount to 18€/hour (charged cost). The overall cost is dominated by the labour time needed to prepare the NPs and the XO, spermine and Zn solutions. Hence the consumption of stock solution is estimated to cost 18€ for 5000 acquisitions, or 4 m€ per acquisition. This cost is about 3 order of magnitude lower than the cost of spectral acquisition. It will therefore be hereafter neglected.

SERS acquisition:
For calibration: First, a Zn-free standard sample is prepared by mixing the NP, XO and spermine solutions in the cuvette and then addding pure water and mixing again. The spectrum is then recorded. An aliquot of Zn concentrated standard is then spiked into the cuvette, which is further mixed and a new spectrum is recorded. The procedure is repeated to span the desired Zn sensitivity range (Fig. S15). The cuvette is then cleaned in an ultrasonic bath and another titration series is conducted. To build a training set equivalent to that used for building the M50 calibration model, 8 concentration steps are explored in each titration series and 10 replicate titrations must be conducted. From the performed measurements, we estimated that each titration series takes on average 9' to complete. Acquisition of the training set therefore takes 90 min which amounts to 27€ worth of labour.

Figure S15
: graphical depiction of the timeline of sample preparation, for calibration or for acquisition in an unknown solution. Note that 10 titrations series are needed to achieve the analytical performances of model M50 and 6 acquisitions on the unknown solution are needed to give readouts with 5 % precision.

For Zn determinations on unknown samples:
The NP, XO and spermine solutions are mixed in the cuvette and then the water sample is added and the cuvette is again homogenised. The spectrum is then recorded and the cuvette is cleaned. The whole acquisition procedure can be comfortably performed in 4 min (Fig. S15), which amounts to 1.2€ worth of labour. To achieve a standard error of the mean predicted Zn concentrations of 5% at most, 6 replicate measurements are needed and therefore each mean determination of Zn concentration costs 7.2 € worth of labour.

Cost per Zn determination, including calibration
Assuming that the spectrometer is deployed in a neighbourhood for one day (8 hours), 6 hours 30 min can be dedicated to actual sample measurements after the 1h30 calibration, corresponding to 98 spectral acquisitions or 16 determinations of Zn concentration. The overall cost of each of these Zn readings, including calibration is therefore 8.9€.
4 Ageing and stability of Lee-Meisel NPs  The NP batch used is the c one in Figure S16. The dashed lines correspond to the normal Raman signals of the PMMA spectrophotometric cuvettes.