Thermal-Induced Morphological Evolution of Highly Ordered Gold Nanoarrays for Ultrasensitive Digital Biodetection

Abstract

Digital biodetection has revolutionized clinical research and diagnostics by utilizing single-molecule counting to overcome the limitations of traditional analog signals. While plasmon-enhanced fluorescence (PEF) offers a rapid, enzyme-free method for signal amplification, the inconsistent distribution of chemically synthesized nanoparticles on sensing interfaces remains a major barrier to detection robustness. In this study, we address this challenge by developing highly ordered gold nanoparticle arrays using nanoimprint lithography and physical vapor deposition. We systematically investigated the effect of thermal annealing temperature on the morphological and optical evolution of these arrays. Our findings demonstrate that high-temperature thermal annealing restructures as-deposited Au nanodisks into compact, quasi-spherical nanoparticles while driving grain growth and enhanced crystalline coherence. This structural refinement blue-shifts the localized surface plasmon resonance (LSPR) by approximately 100 nm, significantly optimizing spectral coupling with Cy5 fluorophores. Arrays annealed at 750 °C yield strong dark-field scattering, excellent spectral overlap with the Cy5 emission band, and superior particle–substrate adhesion, together generating the maximum fluorescence enhancement factor. Consequently, the optimized nanoarrays achieve a one-order-of-magnitude boost in detection sensitivity, enabling an attomolar-level limit of detection (LOD) for model DNA targets in buffer. Furthermore, the platform reliably detects temporally resolved single-molecule photobleaching events even in complex biological matrices such as 1 pM target in fetal bovine serum (FBS). Overall, this work presents a robust and scalable fabrication strategy for high-performance plasmonic arrays tailored for digital bioanalysis.

Introduction

Digital biodetection has transformed clinical diagnostics and biological research by shifting the focus from bulk analog signals to discrete single-molecule counting. By quantifying individual molecular events, this approach effectively overcomes the limitations of signal fluctuations and background noise, providing significantly higher sensitivity than traditional methods. , While digital PCR and digital ELISA have become benchmarks in the field, their reliance on complex enzymatic amplification often increases costs and complicates assay protocols. − Consequently, there is significant interest in developing enzyme-free signal amplification strategies that maintain high sensitivity while improving cost-effectiveness and ease of use. −

Plasmon-enhanced fluorescence (PEF) has emerged as a compelling alternative for biomolecular detection. − This technique utilizes the localized surface plasmon resonance (LSPR) of metal nanostructures to concentrate electromagnetic fields, thereby amplifying the excitation and emission rates of nearby fluorophores. However, the practical application of PEF in digital assays is often hindered by the poor distribution uniformity of chemically synthesized nanoparticles. − When nanoparticles are randomly deposited on a sensing interface, the resulting “hot spots” are inconsistent, leading to poor reproducibility and limited detection robustness.

To address these challenges, researchers have turned to top-down fabrication techniques to create ordered plasmonic microstructures. − These arrays offer the performance of plasmonic nanoparticles with the added benefits of high spatial uniformity and controllability. , Among these methods, the combination of nanoimprint lithography and physical vapor deposition provides a scalable, high-throughput route to creating precise metallic nanopatterns. However, as-deposited metallic nanodisks typically possess small crystalline domains, dense grain boundaries, and high surface roughness. These structural defects induce electron scattering and nonradiative damping, which attenuate plasmon resonance and diminish overall fluorescence enhancement.

In this work, we demonstrate a strategy to optimize the performance of gold nanoparticle arrays by systematically controlling the thermal annealing process. We systematically investigated how thermal annealing governs both the morphological evolution and internal grain growth of prepatterned gold nanodisks. High-temperature annealing converts flat nanodisks into compact, quasi-spherical nanoparticles while enhancing crystalline coherence, inducing a marked blue-shift in the localized surface plasmon resonance (LSPR). Among the tested conditions, arrays annealed at 750 °C achieve an optimal balance of dark-field scattering intensity, spectral overlap with Cy5, fluorescence amplification, and structural array integrity. Consequently, the optimized platform yields a 10-fold improvement in detection sensitivity, reaching an attomolar-level limit of detection (LOD) for model DNA targets in buffer, while maintaining robust temporally resolved single-molecule fluorescence resolution in complex biological serum (1 pM target concentration). This heat-optimized platform provides a robust and efficient solution for high-fidelity digital bioanalysis.

Results and Discussion

To evaluate the impact of thermal treatment on sensing performance, we systematically investigated the effects of varying annealing temperatures on the fluorescence enhancement of the resulting gold structures. The initial gold nanodisks were fabricated with a diameter of 200 nm and a thickness of 40 nm, arranged in a highly ordered square lattice with a 1 μm period (Figure a). Following the annealing process, the nanoparticle surfaces were functionalized with DNA capture probes using an efficient one-step “freezing-directed” method. The arrays were then incubated with Cy5-labeled complementary DNA probes, enabling the immobilization of fluorescence tags at the single-molecule level.

1.

Characterization of the plasmonic gold nanoarray and the TIRF imaging system. (a) SEM micrograph of the gold nanoarray (1 μm pitch) and a schematic of the objective-type TIRF microscopy setup. The system features a 100× oil-immersion objective (NA 1.47), a 635 nm excitation laser, and an EMCCD camera. (b–d) Representative TIRF images of the arrays conjugated with equivalent fluorophore concentrations under varying conditions: (b) as-deposited (unannealed), (c) low-temperature annealing (350 °C), and (d) high-temperature annealing (750 °C). (e–g) Representative fluorescence intensity-time traces for individual particles from groups (b–d), respectively. Imaging conditions: 20 mM Tris–HCl (pH 8.0), 0.1 M NaCl, 0.005 M MgCl₂, and 50 pM DNA probes; 10 min incubation. Acquisition: 100 ms exposure, 10 fps, 30 s duration, and 500 EM gain.

The PEF performance was characterized using Total internal reflection fluorescence (TIRF) microscopy to ensure high signal-to-noise ratios. As the annealing temperature increases, the gold nanostructures evolve from flattened disks into compact, near-spherical nanoparticles. This structural evolution markedly improves the fluorescence signal and the detectability of discrete photobleaching transitions. The optimized arrays therefore support more reliable counting of temporally resolved single-molecule events and lower the LOD for model nucleic acid biomarkers by approximately 1 order of magnitude.

The fabrication process for the Au nanoarrays, illustrated in Figure a, utilized a streamlined combination of nanoimprint lithography (NIL) and physical vapor deposition (PVD). Scanning electron microscopy (SEM) characterization confirmed that the resulting arrays comprised gold nanodisks with a uniform diameter of 200 nm and a center-to-center pitch of 1 μm (Figure b,c). This 1 μm periodicity was strategically selected to allow individual nanoparticles to be optically resolved during fluorescence imaging, while simultaneously maximizing the particle density within a single field of view (FOV) to ensure statistical rigor.

2.

Fabrication workflow and morphological characterization of the nanoarrays. (a) Schematic representation of the fabrication process combining NIL and PVD. (b) SEM image of the vertical nanopore array etched into the resist via RIE. (c) SEM image of the precursor gold nanodisk array (diameter: 200 nm, thickness: 40 nm) following the lift-off process. (d) SEM image of the finalized plasmonic gold nanoparticle array after thermal annealing at 950 °C. Scale bar: 1 μm.

Furthermore, experimental observations indicated that 200 nm diameter disks require a specific thickness threshold to transform into single, cohesive particles during thermal annealing rather than fragmenting into multiple isolated islands (Figures d and S1, deposition-thickness-dependent morphologies after annealing). We also observed that the particle morphology became increasingly rounded as the annealing temperature rose. Based on these findings, we used gold nanodisk arrays with a deposited thickness of 40 nm to evaluate four annealing temperatures: 350, 550, 750, and 950 °C. The heating rate and dwell time were kept constant across all thermal treatments.

Representative SEM images (Figure a) and expansive full-field views (Figure S2) reveal a systematic lateral contraction and surface smoothing of the patterned Au nanostructures as the annealing temperature increases. The unannealed disks feature rough surfaces and irregular boundaries. At 350 °C, the nanostructure contours begin contracting inward, and by 550 °C, surface asperities are substantially passivated. Further annealing at 750 and 950 °C transforms the nanodisks into smaller, highly compact, and faceted quasi-spherical nanoparticles, a structural transition driven by mass transport via thermally activated Au surface diffusion and surface energy minimization. ,

3.

Thermally induced structural evolution of Au nanoparticle arrays and its influence on their optical properties. (a) SEM images showing the morphological evolution of as-deposited Au nanodisks (unannealed) into compact nanoparticles after annealing at 350, 550, 750, and 950 °C. Annealing was performed at a heating rate of 10 °C min^–1 with a dwell time of 10 min. (b) Representative AFM height maps of individual particles corresponding to the five thermal conditions in (a). (c) XRD patterns recorded from the patterned Au arrays before annealing and after annealing at 950 °C, showing annealing-induced grain growth and enhanced crystalline coherence. (d) Dark-field scattering spectra of individual Au nanoparticles under the five thermal conditions. (e) Spectral overlap between the scattering spectrum of the Au nanoparticles annealed at 750 °C and the absorption and emission spectra of Cy5. (f) Schematic illustration of TIRF-based PEF characterization, in which Cy5-labeled target DNA is immobilized near the Au surface through hybridization with Au-bound capture DNA. (g) Violin distributions of the background-subtracted initial mean fluorescence intensities measured at individual nanoparticle sites under the five thermal conditions. The connecting lines indicate the group mean and median. Measurements were performed in 20 mM Tris–HCl (pH 8.0), 0.1 M NaCl, and 5 mM MgCl₂ using 50 pM Cy5-labeled DNA probes and a 10 min incubation. Images were acquired with a 100 ms exposure at 10 frames s^–1 for 30 s using an EM gain of 500. Scale bars: 200 nm in (a) and 100 nm in (b).

Corrected AFM height topographies (Figures b and S3) show a progressive increase in mean particle height from 40 nm (unannealed) to 50, 60, 80, and 90 nm for the 350, 550, 750, and 950 °C arrays, respectively. Complementary top-view SEM statistical analysis demonstrates a concurrent decrease in mean lateral diameter from 236.5 ± 4.5, down to 207.6 ± 7.1, 196.6 ± 4.4, 184.3 ± 11.0, and 169.4 ± 12.6 nm. Consequently, the corresponding height-to-diameter aspect ratio increases monotonically from 0.17 to 0.24, 0.31, 0.43, and 0.53. Residual map analysis following baseline spherical-curvature subtraction reveals lower surface roughness parameters (Sa and Sq) for all annealed samples relative to unannealed disks, though without a monotonic trend among the annealed cohorts (Figure S4 and Table S2). This concomitant reduction in lateral footprint, height increase, and aspect ratio enhancement clearly reflects a reduction in the particle–substrate contact area to achieve thermodynamic equilibrium.

X-ray diffraction (XRD) measurements performed directly on the patterned nanoarrays independently verify internal crystalline-domain growth (Figure c). Both unannealed and 950 °C arrays display a prominent Au(111) reflection centered at 2θ = 38.2°. Quantitative fitting of the Cu Kα doublet yields a substantial narrowing of the Au(111) full width at half-maximum (fwhm) from 0.296° down to 0.098°, corresponding to an expansion of the calculated Scherrer coherent domain size from 28.4 to 85.9 nm. This pronounced peak sharpening provides ensemble-averaged evidence of significant grain growth, strain relaxation, and enhanced crystalline coherence upon thermal annealing.

This structural evolution strongly affects the optical response. Representative dark-field micrographs show the scattering color changing from dark red to yellow and then pale green (Figure S5, structural-evolution schematic and temperature-dependent dark-field array images). The normalized single-particle spectra in Figure d confirm that the broad LSPR shifts from approximately 750 nm for the unannealed disks to approximately 650 nm after annealing at 950 °C. The dark-field scattering intensity reaches its maximum at 750 °C and then decreases at 950 °C, consistent with the continued lateral contraction, reduced scattering cross-section, and size-dependent optical response of Au nanoparticles.

PEF was evaluated by immobilizing thiolated capture DNA on Au and hybridizing Cy5-labeled target DNA (Figure f). The 26-base-pair duplex and molecular linkers place Cy5 approximately 10 nm from the Au surface, within the localized plasmonic near field while effectively suppressing nonradiative metal quenching. All annealing groups were functionalized using the same capture-DNA concentration, volume, freezing, washing, passivation, and hybridization procedures. The 5 μM capture-DNA solution provided a substantial stoichiometric excess, promoting a dense and saturated probe monolayer across all arrays. Figure g summarizes the mean fluorescence intensity measured at each individual nanoparticle site. Each value is the mean initial intensity measured at one particle-associated site after subtraction of its corresponding baseline. The unannealed disks show negligible fluorescence, whereas annealing at 350 °C produces a pronounced increase. The response rises further from 550 to 750 °C and reaches its maximum at 750 °C; the 950 °C group does not provide an additional improvement.

The optimal performance observed at 750 °C emerges from a synergistic interplay among spectral matching, scattering cross-section, crystalline coherence, and structural integrity. At 750 °C, the plasmonic scattering spectrum achieves favorable overlap with the Cy5 emission band. Although further annealing to 950 °C blue-shifts the resonance closer to the Cy5 excitation maximum (Figure e), the 950 °C particles exhibit reduced absolute scattering intensity due to severe lateral contraction, which diminishes their effective volume and scattering cross-section. This reduced scattering power and attenuated optical-antenna efficiency at 950 °C negate any benefit from improved excitation-band spectral overlap. Furthermore, the drastically minimized particle–substrate contact area at 950 °C compromises interfacial adhesion, predisposing nanoparticles to detachment during routine microfluidic washing and bioassay handling. In contrast, 750 °C arrays maintain excellent substrate anchorage and structural integrity. Thus, 750 °C represents the optimal plasmonic fluorescence enhancement alongside superior mechanical and analytical robustness.

To evaluate annealing-dependent fluorescence kinetics, we compared the unannealed, 350 °C, and 750 °C arrays under identical TIRF conditions. Representative images show sparse and weak fluorescence signals on the unannealed array, a higher spot density after annealing at 350 °C, and bright, well-defined particle-associated signals after annealing at 750 °C (Figure a). For each addressable nanoparticle, the bright-field-derived coordinate was mapped to the corresponding TIRF movie, and the fluorescence intensity within a 3 × 3 pixel ROI was extracted frame by frame. After background correction and smoothing, discrete downward transitions in the fluorescence traces were identified as temporally resolved photobleaching events using consistent criteria across all experimental conditions. Representative fluorescence intensity–time traces further illustrate the progressive improvement in single-molecule signal detectability with increasing annealing temperature (Figure b). The unannealed array exhibited weak fluorescence close to the background level, with no clearly resolvable photobleaching transitions. After annealing at 350 °C, the increased fluorescence intensity allowed multiple discrete intensity levels and stepwise downward transitions to be distinguished. The 750 °C array produced a higher initial fluorescence intensity and more clearly separated fluorescence plateaus, enabling a greater number of photobleaching events to be temporally resolved. The initial fluorescence peak represents the superposition of concurrently emitting fluorophores, whereas the successive downward transitions arise from the sequential photobleaching of detectable fluorophores. The corresponding schematics illustrate how annealing-induced particle reshaping improves the plasmonic optical-antenna response and thereby increases the detectability of particle-associated fluorescence signals. These discrete, stepwise photobleaching trajectories form the quantitative foundation for single-molecule event extraction (Figures b and S6). We emphasize that the measured event counts represent a conservative lower-bound estimate of fluorophore occupancy (N _resolved ≤ N _fluorophore) rather than an absolute molecular quantification. This operational limit arises because subthreshold transitions are merged into background noise, while coincident photobleaching of multiple fluorophores within the ∼0.2 s temporal window registers as a single step. To avoid positional intensity artifacts inherent to plasmonic near-fields, step amplitudes are not subdivided into multiple molecular counts. Accordingly, reported counts reflect temporally resolved single-molecule events under standardized detection parameters, providing an unbiased benchmark for comparative assay performance.

4.

Annealing-dependent fluorescence kinetics and single-molecule digital counting performance. (a) Representative TIRF images of Au nanoparticle arrays in the (a1) unannealed, (a2) 350 °C-annealed, and (a3) 750 °C-annealed states. Red boxes indicate the ROIs selected for digital mapping. (b) Representative fluorescence intensity–time traces acquired from individual nanoparticle sites, together with schematics illustrating morphology-dependent PEF. Colored shaded regions indicate the intensity intervals used to resolve successive photobleaching transitions. (c) Spatial distributions of the resolved photobleaching-event counts within the corresponding 10 × 10 subarrays selected in (a). The mean ± SD of the event count per nanoparticle site is displayed above each map. Fluorescence videos were acquired with a 100 ms exposure at 10 frames s^–1 for 30 s. Scale bar: 60 μm in (a).

The resolved event counts were spatially reconstructed within the 10 × 10 ROIs marked in Figure a. The mean ± SD values were 0.00 ± 0.00, 2.21 ± 1.09, and 4.43 ± 0.59 resolved events per nanoparticle for the unannealed, 350 °C, and 750 °C arrays, respectively (Figure c). This increase reflects the improved detectability of single-molecule fluorescence signals under identical acquisition and analysis criteria. Spatial uniformity across the nanoarray surface was evaluated using a five-region sampling strategy, analyzing 10 × 10 particle ROIs situated at the upper-left, upper-right, lower-left, lower-right, and geometric center of each chip. Across arrays annealed at 350 °C, 550 °C, and 750 °C, the maximum variation between the highest and lowest regional means was less than 0.5 resolved events per nanoparticle, demonstrating exceptional chip-scale homogeneity without center-to-edge spatial bias (Figure S7). Furthermore, batch-to-batch fabrication reproducibility was confirmed across three independently prepared array batches (40 nm nominal Au thickness, annealed at 750 °C). Following a 10 min incubation with 10 pM Cy5-DNA, single-molecule fluorescence measurements acquired under standardized imaging and analytical conditions yielded closely overlapping distributions of background-subtracted initial intensities without systematic batch-to-batch shifts (Figure S8), ensuring the operational reliability and reproducibility of the platform.

The main advantage of the fluorescence enhancement produced by optimized annealing is the reduction in LOD for low-concentration biomarkers. Concentration–response assays were performed in Tris–HCl/MgCl₂/NaCl buffer using Cy5-labeled DNA probes on arrays annealed at 350, 550, and 750 °C. Across concentrations from 1 fM to 100 pM, the number of temporally resolved photobleaching events increased with concentration for all arrays (Figure a). At every concentration tested, the 750 °C arrays yielded more resolved events than the lower-temperature arrays. At 1 pM, the 750 °C array produced 55.5% more events than the 550 °C group and 79% more than the 350 °C group. This improved detectability is particularly important in the ultralow-concentration regime, where target occupancy is sparse. Below 10 fM, the 350 °C array approached its detection threshold, whereas the 750 °C array retained a measurable concentration-dependent response. The LOD was estimated from the buffer-phase calibration curves using the mean blank response plus three times the standard deviation of the blank and back-calculation from the fitted log–log relationship. Estimated LODs of 1.8 fM and 540 aM were obtained for the 350 and 750 °C arrays, respectively.

5.

Quantitative detection sensitivity and robustness in complex matrices. (a) Buffer-phase concentration–response curves for Cy5-labeled DNA probes (1 fM to 100 pM) on arrays annealed at 350, 550, and 750 °C. Responses are reported as temporally resolved photobleaching-event counts normalized to the analyzed nanoparticle population. (b) Fixed-concentration serum robustness test comparing event counts in buffer and processed FBS for the 350 and 750 °C arrays (1 pM target; 30 min incubation). Error bars indicate the standard error of the mean.

The serum experiment was conducted separately as a fixed-concentration robustness test. Processed FBS was spiked with 1 pM Cy5-labeled target DNA and incubated with the functionalized arrays for 30 min. Relative to buffer at the same concentration, the resolved event count decreased by 36.2% for the 350 °C array but by only 16.5% for the 750 °C array (Figure b), demonstrating improved signal retention in a high-background biological matrix.

Conclusions

In summary, we have established a scalable, deterministic approach for fabricating highly uniform gold nanoparticle arrays by combining nanoimprint lithography with thermal annealing. Thermal processing governs the structural, crystalline, and plasmonic evolution of the nanoarrays, driving the transition from shallow, rough nanodisks into smooth, high-aspect-ratio quasi-spherical nanoparticles with enhanced crystalline coherence. This structural reorganization yields a tunable LSPR response and enhanced scattering efficiency, with 750 °C identified as the optimal annealing condition that balances spectral matching with Cy5, scattering cross-section, fluorescence enhancement, and interfacial mechanical stability. When applied to single-molecule digital fluorescence counting, the optimized 750 °C arrays achieved an estimated limit of detection of 540 aM for a model DNA target, while maintaining robust single-molecule event resolution in complex biofluids (1 pM target in fetal bovine serum). This heat-controlled nanofabrication strategy offers a practical, high-throughput pattern for manufacturing standardized plasmonic chips for digital molecular diagnostics, demonstrating the potential for future translation toward simplified optical instrumentation and clinical biomarker validation.

Experimental Section

Materials and Reagents

All DNA oligonucleotides (Table S1, DNA sequences used in this work, Supporting Information), Tris–HCl buffer, DNase kits, and DNase inhibitors were purchased from Takara Biomedical Technology (Beijing) Co., Ltd. Inorganic reagents were obtained from Shanghai Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). Serum samples were obtained from Sangon Biotech (Shanghai, China) Co., Ltd. Tris­(2-carboxyethyl)­phosphine hydrochloride (TCEP, 98%), 6-mercapto-1-hexanol (MCH), and bovine serum albumin (BSA) were sourced from Shanghai Macklin Biochemical Co., Ltd. (Shanghai, China). Quartz glass slides and glass cleaning solutions were acquired from Donghai Jingqi Optical Technology Co., Ltd. Polydimethylsiloxane (PDMS) was purchased from The Dow Chemical Company (Midland, MI, USA). Poly­(methyl methacrylate) (PMMA) was purchased from Sigma-Aldrich (St. Louis, MO, USA). The nanoimprint resist was prepared in our laboratory.

Fabrication of Au Nanoarrays

The Au nanoparticle arrays were fabricated using a combination of nanoimprint lithography (NIL) and physical vapor deposition. Initially, a silicon mold with a periodic pit pattern (diameter: 200 nm, depth: 100 nm, period: 1 μm) was treated with an antiadhesive layer. Then the polydimethylsiloxane (PDMS) was dropped onto the silicon template and cured for 4 h in a vacuum at 80 °C. Subsequently, a 200 nm thick composite photoresist layer, consisting of custom-made nanoimprint resist and PMMA, was spin-coated onto the quartz substrates. After drying for 10 min, a PDMS mold with the target array patterns was laminated onto the resist. Following the removal of the residual resist layer through reactive ion etching (RIE), a 40 nm gold film was deposited using an electron-beam evaporator. The final nanodisk arrays were obtained after a standard lift-off process in acetone.

Thermal Annealing and Characterization

To investigate the structural evolution of the nanoarrays, samples were annealed in a muffle furnace at 350–950 °C using a heating rate of 10 °C min^–1 and a dwell time of 10 min. External morphology was characterized by SEM and AFM. XRD patterns were acquired from the actual patterned arrays before annealing and after annealing at 950 °C. X-ray diffraction (XRD) patterns were collected using Cu Kα radiation at an operating voltage of 40 kV and a current of 40 mA. The measurements were performed in a coupled θ–2θ geometry over a 2θ range of 30–80°, with a step size of approximately 0.0205°. The wavelength used for data acquisition was 1.5418 Å (Cu Kα₁ = 1.5406 Å and Cu Kα₂ = 1.54439 Å). Single-particle dark-field scattering spectra were measured using an objective-coupled microspectrometer.

Surface Functionalization and DNA Capture

Prior to the bioassays, the Au nanoarrays were cleaned and functionalized. A semienclosed reaction chamber approximately 5 mm in diameter was constructed over the array using custom-punched PTFE tape. Thiolated capture DNA was reduced with 20 mM TCEP for 30 min and diluted to 5 μM in 20 mM Tris–HCl (pH 8.0). Twenty microliters of capture-DNA solution was added to each chamber, and freezing-directed functionalization was performed by freezing the chamber completely and then thawing it. The arrays were rinsed with salt-free buffer. The supplied DNA amount was several orders of magnitude greater than the estimated surface capacity, favoring high and approximately comparable capture-DNA densities across accessible Au surfaces.

To reduce nonspecific binding, the surfaces were sequentially passivated with 1 mM MCH and 1% BSA for 1 h each and washed with 20 mM Tris–HCl, 5 mM MgCl₂, and 100 mM NaCl (pH 8.0). Cy5-labeled target DNA containing a 26-base sequence complementary to the capture strand was then added in the dark. Annealing-dependent fluorescence comparisons used a 10 min hybridization, whereas concentration–response and serum robustness experiments used the incubation times specified for those assays. After incubation, the supernatant was removed and replaced with 15 μL of imaging buffer.

Single-Molecule Fluorescence Assay and Data Analysis

The Cy5-labeled target DNA was hybridized with the immobilized capture probes. For digital detection, the arrays were imaged using a custom wide-field TIRF microscope equipped with a 635 nm excitation laser and a high-sensitivity EMCCD camera. Single-molecule identification was performed by analyzing the fluorescence intensity profiles; the characteristic “staircase” photobleaching behavior was used to confirm the presence of individual molecules.

Single-Molecule Fluorescence Performance in Serum

To assess robustness in a complex matrix, serum was treated with a DNase kit for 5 min at 37 °C, quenched with a DNase inhibitor, and filtered through a 200 nm membrane. The processed serum was spiked with 1 pM Cy5-labeled target DNA using a 1/10 volume of 10× detection buffer and incubated with the functionalized arrays for 30 min. This experiment was a fixed-concentration robustness test. The sensing interface was imaged without further purification after incubation.

Optical Setup and Characterization

The dark-field scattering properties of the nanoarrays were characterized using an upright metallurgical microscope (Zeiss AX10, Zeiss, Germany) equipped with a dedicated dark-field objective (50×, NA 0.60). Scattering images were captured with a digital single-lens reflex (DSLR) camera (EOS R6II, Canon, Japan) using a fixed exposure time of 1 s and an ISO of 500 to ensure consistency. The dark-field scattering spectra of individual nanoparticles were acquired using a microspectrometer (Shamrock, Andor, UK) coupled to the microscope port.

Single-molecule fluorescence signals on the nanoarrays were collected using an inverted fluorescence microscope (DMI 8, Leica, Germany) featuring an objective-type total internal reflection fluorescence illumination system. The system was equipped with a high-numerical-aperture oil-immersion objective (100×, NA 1.47, Leica, Germany). Fluorescence emission was detected by an electron-multiplying charge-coupled device (EMCCD) camera (iXon Ultra 897u, Andor, UK) known for its single-photon sensitivity. A 635 nm laser was employed as the excitation source. To isolate the fluorescence signal and minimize background noise, the optical path included a 650 nm dichroic mirror combined with a 700/60 nm bandpass emission filter (Semrock, US).

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnanoscienceau.6c00110.

• DNA sequences; deposition-thickness-dependent and annealing-dependent Au-array morphology; corrected AFM height maps, particle-core roughness statistics, and dimensional analysis; structural-evolution schematics and dark-field array images; fluorescence colocalization, trace extraction, event-detection criteria and error interpretation; region-to-region uniformity; batch reproducibility (PDF)

†.

L.S. and X.H. contributed equally to this work.

The authors declare no competing financial interest.