Microplasma Controlled Nanogold Sensor for SERS of Aliphatic and Aromatic Explosives with PCA-KNN Recognition

Abstract

Nanogold is an emerging material for enhancing surface-enhanced Raman scattering (SERS), which enables the detection of hazardous analytes at trace levels. This study presents a simple, single-step plasma synthesis method to control the size and yield of Au nanoparticles by using plasma-liquid redox chemistry. The pin-based argon plasma reduces the Au³⁺ precursor in under 5 min, synthesizing Au spherical particles ranging from ∼20 nm at 0.025 mM to ∼90 nm at 1.0 mM, in addition to plate-like particles occurring at concentrations of 0.25-1.0 mM. The enhanced SERS responses correlated with the UV-vis absorption and reflectance profiles, which can be attributed to synergistic plasmonic hotspots created by the sphere-sphere, plate-sphere, and plate-plate nanogold interactions. This nanogold mixture, combined with gold-plated CPU grid pin arrays, facilitated the detection of trace explosives, including aromatic (TNT, TNB, and TNP) and aliphatic (RDX, PETN, and HMX) compounds. We demonstrate that stabler aliphatic analytes, associated with lower vapor pressure (10^-8-10^-11 atm), exhibit smaller signal fluctuations (RSD ∼ 6-10%) compared to their more volatile (10^-5 atm) aromatic (RSD ∼ 12-17%) counterparts at similar analyte concentrations. The calculated limit of detection (LoD) was found to be ∼2-6 nM and ∼600-900 pM for aromatic and aliphatic explosives, respectively. Finally, we show that the poorer performance of aromatic explosives under the same sensing conditions affects SERS-PCA separation, which can then be improved either by a machine learning approach (PCA with k-NN classification) or by consideration of a specific NO₂ symmetric stretching fingerprint range.

Keywords: SERS; explosives; gold nanoparticles; machine learning; plasma synthesis.

Figure 1

(a) Schematic showing the cold atmospheric pressure plasma (CAP) setup with a diagram highlighting the plasma-liquid interface and proposed particle formation process; (b) voltage and current waveforms during plasma generation; (c) emission spectrum of CAP at the liquid interface; (d) simulated (T_rot ≈ 590 K) and experimental spectra of excited OH; and (e) measured H_β emission line and Voigt fit for n_e calculation.

Figure 2

(a) Photographs showing the progression of nanoparticle synthesis from aqueous precursor solution to colloidal nanogold over 5 min of treatment time for the 0.1 mM concentration case; (b) UV–vis absorbance spectra of the stock solution indicating Au³⁺ peak intensity with concentration; (c) extinction spectra of colloidal nanogold showing an increase in the surface plasmon resonance peak intensity with concentration; (d) total diffuse reflectance of the colloidal nanogold; (e) colloidal nanogold after plasma treatment with increasing precursor concentration from left (0.025 mM) to right (2.5 mM).

Figure 3

(a) Series of SEM and TEM images showing the size and shape of Au nanoparticles with increasing concentration of starting material; (b) Au⁰ peak positions in UV–vis extinction spectra. (c) DLS/MALDS measurements of 0.025 to 1.0 mM nanocolloids and their (d) spherical size and count distribution as a function of precursor concentration; (e) selected area electron diffraction (SAED) ring patterns for nanocolloids created using a 1.0 mM precursor concentration; and (f) deconvoluted XPS spectra showing the Au 4f and Cl 2p core levels for the 1.0 mM sample before and after CAP treatment.

Figure 4

(a) Process diagram showing three independent samples of nanogold created using 1.0 mM precursor, centrifuged and the analyte added for SERS measurements; (b) corresponding spectra of CV (10^–6 M) recorded from dried spots on Si wafer each accompanied by relative standard deviation measurements; (c) detection limit estimation taking intensity profile of a mode located at 1618 cm^–1 from the SERS concentration study of crystal violet Figure S5; (d) deposition of a mixture of water diluted Raman molecule (CV:10^–6 M) with nanoparticles on CPU pins; (e) SERS comparison of CV (10^–6 M) for CPU pins with Au nanoparticles “black” and CV (10^–6 M) without Au nanoparticles “red” curve; (f) SERS test of nanocolloids obtained from different Au³⁺ precursor concentrations; (g) SERS fingerprints (10^–6 M) of crystal violet, alcian blue, and rhodamine using 1.0 mM AuNPs.

Figure 5

(a) SERS of the aromatic TNT, TNP, TNB and (b) aliphatic RDX, HMX, PETN explosives after baseline correction with corresponding (a’, b’) LoD fitting; (c) PC1 vs PC2 and its evaluation with (d) k-NN confusion matrix (up to PC3) within a wide-range Raman spectrum (750–1700 cm^–1) of aromatic and aliphatic explosives; (e) PC1 vs PC2 and (f) corresponding loading plots of ν_s(NO₂) band (1200–1400 cm^–1).