Self‑Assembled Gold Nanorod Vertical Arrays: Mechanistic Insights and High‑Sensitivity SERS Applications

Introduction

Noble‑metal nanostructures—gold, silver, copper, and others—generate intense localized EM fields under visible illumination, enabling surface plasmon resonance (SPR) and powerful optical phenomena such as surface‑enhanced fluorescence (SEF) and surface‑enhanced Raman scattering (SERS) [1, 2]. SERS offers exceptional sensitivity, rapid response, and a molecular fingerprint, making it ideal for material analysis, biomedical diagnostics, and sensor development [3‑7]. The enhancement originates mainly from EM field amplification, which can reach 4–11 orders of magnitude, and from subtle chemical interactions [8]. Hot spots—narrow gaps between adjacent nanoparticles—are key to achieving large EM amplification. Therefore, the morphology of the metal substrate, the choice of probe molecules, and the excitation conditions are all critical for maximizing SERS performance [9].

Despite numerous reports on SERS, challenges remain in fabricating low‑cost, large‑area, and uniformly reproducible substrates with ultra‑high sensitivity. Self‑assembled metal nanostructures have emerged as attractive candidates because they combine low fabrication cost, easy handling, and large‑area uniformity while preserving the unique EM properties of individual nanoparticles [13‑18]. Recent work has demonstrated gold nanorod (GNR) self‑assembled substrates for SERS [19‑21], yet the influence of morphological evolution on Raman signals has not been thoroughly investigated. In this study, we synthesize GNR vertical arrays via an evaporation method, regulate their morphology through controlled soaking times, and evaluate their SERS performance for Rhodamine 6G (Rh6G) and crystal violet (CV). Finite‑element modeling (FEM) of the EM field distribution corroborates the experimental observations, and we assess sensitivity, reproducibility, and stability to demonstrate the substrate’s practical potential.

Methods and Experiment

Materials

Rhodamine 6G (laser grade) was purchased from Exciton (USA), crystal violet (CV) from Sigma‑Aldrich, gold chloride tetrahydrate, ethanol, silver nitrate, and hydrochloric acid from Sinopharm Chemical Reagent Co., Ltd. (China). Cetyltrimethylammonium bromide (CTAB), sodium borohydride, and ascorbic acid were obtained from Shanghai Aladdin Bio‑Chem Technology Co., Ltd. (China). Silicon wafers (Si) were supplied by Li Jing Photoelectric Technology Co. Ltd. (Zhejiang, China). All reagents were used as received, and deionized water (18 MΩ cm) was used throughout the experiments.

Preparation of GNR Vertical Arrays

Gold nanorods were synthesized by a seed‑mediated growth protocol [23, 24]. The as‑grown GNRs were centrifuged three times at 10 000 rpm for 5 min to remove excess CTAB. Following a previously reported evaporation procedure [22], a 5 µL drop of the purified GNR suspension was deposited on a 6 × 6 mm² silicon wafer. The wafer was then placed in a controlled environment (21 °C, 85 % RH) to allow slow evaporation. After 72 h, self‑assembled vertical arrays formed on the wafer surface. The substrate was gently withdrawn, rinsed with ethanol, and dried prior to use. The complete workflow is illustrated in Fig. 1.

The scheme of the preparation process of GNR vertical arrays

Characterization

The size and morphology of the GNR arrays were examined with a scanning electron microscope (SEM, Nova Nano 450). Raman spectra were recorded using a confocal Raman microscope (LabRAM HR Evolution, HORIBA Jobin Yvon SAS). A continuous‑wave 532 nm laser was employed as the excitation source, with an incident power of 0.5 mW. The samples were positioned on a × 50 objective, and an integration time of 1 s was used for each spectrum.

Results and Discussion

Mechanism of Gold Nanorod Self‑Assembly

During solvent evaporation, capillary flows drive GNRs toward the droplet edge, producing a coffee‑ring deposition of disordered nanorods [25, 26]. However, in aqueous solution, GNRs can arrange side‑by‑side into a six‑fold lattice under the influence of attractive van der Waals forces, depletion interactions, and electrostatic repulsion [27]. Marangoni convection and the receding contact line promote the aggregation of GNRs around the initial lattice, expanding the vertical‑array region. The interplay between attraction (van der Waals + depletion) and repulsion (electrostatic) establishes a highly ordered, densely packed array. Temperature, humidity, and surfactant concentration critically affect this self‑assembly. A higher temperature or lower humidity intensifies the coffee‑ring effect, whereas sufficient CTAB concentration fosters Marangoni flow, enabling the formation of large‑area vertical arrays [28‑29]. By fine‑tuning these parameters, we achieved uniform, high‑density GNR vertical arrays that serve as robust SERS substrates.

Morphology of Gold Nanorods and Vertical Arrays

The GNRs synthesized by the seed‑mediated method exhibit a length of 69 ± 5 nm, a width of 24 ± 2 nm, and an aspect ratio of ~3, as confirmed by UV–vis absorption (Fig. 2a) and SEM (Fig. 2b). The longitudinal plasmon peak appears at 690 nm, while the transverse peak is at 520 nm. The GNRs self‑assemble into a monolayer of vertically aligned rods on the silicon surface, with a hexagonal close‑packed arrangement (Fig. 2c). The inter‑rod spacing (~3 nm) matches the CTAB bilayer thickness, creating abundant hot spots. SEM images also reveal the contrast between coffee‑ring (Fig. 2e) and coffee‑stain (Fig. 2f) morphologies, confirming the role of Marangoni flow in achieving ordered arrays.

a Ultraviolet–visible absorption spectrum of GNRs. b–d SEM images of the GNR vertical arrays. e, f SEM images of coffee‑ring and coffee‑stain samples.

Spectral Enhancement with GNR Vertical Arrays

We investigated how soaking time in probe‑molecule solutions affects the SERS performance. Using Rhodamine 6G, the Raman intensity at 1650 cm⁻¹ peaked after a 30‑min soak (Fig. 3a,b). The same trend was observed for crystal violet (CV) at 1619 cm⁻¹ (Fig. 3c,d), indicating that the vertical array collapses after ~60 min, likely due to CTAB dissolution and weakened inter‑rod forces. SEM images (Fig. 4) confirm that the arrays remain intact up to 30 min but become disordered after 60 min. The enhanced Raman signal originates from the dense hot spots within the nanorod gaps; loss of order reduces the number of hot spots, diminishing the signal.

a Raman spectra of 10⁻⁷ M Rh6G on GNR vertical arrays with varying soak times. b Intensity ratio of the 1650 cm⁻¹ peak versus soak time. c Raman spectra of 10⁻⁶ M CV on GNR vertical arrays with varying soak times. d Intensity ratio of the 1619 cm⁻¹ peak versus soak time.

a–d SEM images of GNR arrays soaked for 5, 10, 30, and 60 min, respectively.

Finite‑element simulations (Fig. 5) modeled the local EM field under 532 nm illumination, revealing a markedly stronger field for the ordered GNR array compared to the disordered counterpart. The EM enhancement factor (|M_EM|²) scales with the fourth power of the local field, underscoring the critical role of nanorod proximity in generating hot spots. These computational insights align closely with the experimental SERS trends, validating the 30‑min soak as optimal.

a GNR hexagonal array simulation pattern. b Local EM field distribution of the GNR vertical array. c Local EM field of the disordered GNRs.

With the optimized 30‑min soak, we evaluated the SERS performance using Rhodamine 6G as the probe. The spectra (Fig. 6a) show clear Raman peaks at 613, 774, 1185, 1311, 1360, 1508, and 1650 cm⁻¹, with signal intensity decreasing as concentration lowers. At 10⁻¹¹ M, only the 613, 1360, 1508, and 1650 cm⁻¹ peaks remain detectable, demonstrating a detection limit of 10⁻¹¹ M. The enhancement factor (EF) was calculated using Eq. (2), yielding EF ≈ 9.65 × 10⁵, comparable to state‑of‑the‑art self‑assembled substrates [17, 35, 36].

a Raman spectra of Rh6G on the GNR vertical array from 10⁻⁶ to 10⁻¹¹ M. b Raman spectrum of 10⁻³ M Rh6G on a bare silicon wafer. c Raman spectrum of 10⁻⁷ M Rh6G. d, e Intensity distribution of the 1360 cm⁻¹ and 774 cm⁻¹ peaks across 10 random spots on the GNR vertical array.

Reproducibility was assessed by measuring 10 random points on a single substrate. The peak positions remained unchanged, and the relative standard deviation (RSD) for the 1360 cm⁻¹ and 774 cm⁻¹ peaks were 10.7 % and 9.0 %, respectively (Fig. 6d,e), indicating excellent uniformity.

a Raman spectra of 10⁻⁷ M Rh6G on the GNR vertical array after 30 and 60 days. b Comparison of the peak intensities at 774 and 1360 cm⁻¹ over time.

Stability testing over 60 days showed minimal signal loss (<11 %) for the 1360 cm⁻¹ peak and <9 % for the 774 cm⁻¹ peak, confirming that the GNR vertical arrays retain robust SERS activity over extended storage periods.

Conclusion

We have successfully fabricated self‑assembled gold nanorod vertical arrays using a straightforward evaporation technique. By controlling the soaking time in probe‑molecule solutions, we tuned the array morphology to maximize Raman enhancement. Finite‑element analysis of the local EM field corroborated the experimental SERS results. The optimized substrate detected Rhodamine 6G down to 10⁻¹¹ M, while maintaining high reproducibility (RSD < 11 %) and stability (signal loss < 11 % over 60 days). These attributes make the GNR vertical array an attractive platform for sensitive, reliable biosensing and molecular detection.

Abbreviations

CTAB:

Cetyltrimethylammonium bromide

CV:

Crystal violet

FEM:

Finite element method

GNRs:

Gold nanorods

Rh6G:

Rhodamine 6G

RSD:

Relative standard deviation

SEF:

Surface‑enhanced fluorescence

SEM:

Scanning electron microscope

SERS:

Surface‑enhanced Raman scattering

Si:

Silicon wafers

SPR:

Surface plasmon resonance