*2.2. Characterizations*

UV-visible spectra were recorded with a 1 cm path length quartz cell using an Agilent Cary 8454 spectroscopy system (Agilent Technologies Inc., Santa Clara, CA, USA). Polydimethylsiloxane (PDMS) surfaces were treated with a BD-20AC laboratory corona treater (Electro-Technic Products Inc., Chicago, IL, USA). Scanning transmission electron microscope (STEM) measurements were conducted on Tecnai G2 F20 (FEI, Hillsboro, OR, USA). The Raman spectra were obtained using a Renishaw inVia Reflex Raman Microscope and Spectrometer (inVia Reflex, Gloucestershire, UK) equipped with a 633 nm laser and 50× objective. The integration time of all spectra acquisition for each measurement was set to be 10 s. The laser power was 1.7 mW. Four spots on the same SERS substrate were examined, and the spectra were averaged for final analysis.

### *2.3. Preparation of Ag NPs*

The Ag NPs were prepared according to our previous reports [30]. Briefly, approximately 1 mL of ammonium hydroxide was added drop by drop into a fast stirring silver nitrate solution (85 mg in 10 mL water). A dark-brownish precipitate was formed and then dissolved as the amount of ammonium hydroxide is increased. Subsequently, a small amount of as-prepared Tollens solution (40 μL) was added into a fresh solution (40 mL) of D-glucose with a concentration of 0.1 M to 0.5 M. After four hours at room temperature, Ag+ ions were reduced by glucose in the presence of ammonia, and the color of the glucose solution turned to yellow, indicating the formation of Ag NPs [31].

### *2.4. Fabrication of SERS Substrates*

Firstly, the thin PDMS films with a thickness of 0.5 mm were prepared by mixing a Slygard 184 elastomer with a curing agen<sup>t</sup> in a 10:1 ratio at 70 ◦C for two hours. Then the PDMS films were punched to form an array of holes with a diameter of 5 mm. Subsequently, the punched PDMS films were treated by O2 plasma using a BD-20AC laboratory corona treater (Electro-Technic Products Inc., Chicago, IL, USA) for 30 s. Then, the treated PDMS surface was bonded to a glass slide under 70 ◦C for 2 h. Finally, 100 μL of prepared colloidal Ag NPs were dropped into each hole of the PDMS films. After the droplets were evaporated, the pie-shaped substrates of Ag NPs were formed, and then stored at room temperature for SERS detecting. The fabricated SERS film array and the SERS detection process is shown in the Figure 1.

**Figure 1.** Schematics of the SERS detection process. The Ag NPs are closely packed and surrounded by massive solid-state glucose. Moreover, analytes can be distributed uniformly in the SERS substrates by infiltration and capillarity.

### **3. Results and Discussion**

### *3.1. Characterization of Ag NPs*

UV-vis absorption spectra of the colloidal Ag NPs synthesized in 0.5 M glucose is given in Figure 2a. The absorption peak was around 415 nm, exhibiting a sharp plasmon absorption maximum. It indicated that the Ag NPs were monodispersed and relatively uniform in a stable colloidal solution [30,31]. A TEM image of Ag NPs is given in Figure 2b. The size of the Ag NPs was 35 ± 3 nm (the UV-vis absorption spectra and TEM images of the Ag NPs synthesized in 0.1–0.5 M glucose are given in Figures S1 and S2). Since the amount of glucose is excessive, the Ag+ ions have been completely reduced. Moreover, it was a green and highly-efficient synthesis method that does not need a heating condition or any relatively toxic organic surfactants.

**Figure 2.** (**a**) UV-vis absorption spectra of the Ag NPs, and the inset shows a photograph of the Ag NP colloid, and (**b**) TEM image of the Ag NPs, and the size of Ag NPs were 35 ± 3 nm.

### *3.2. Suppression of the Coffee-Ring Effect/Pie-Shaped SERS Substrates*

When a drop of liquid dries on a solid surface, the liquid evaporates first from the edge and is replenished by liquid from the interior, which results in capillary flow in the drop during its drying process. The suspended particles are driven to the edge by the capillary flow, and then left highly concentrated along the original drop edge, finally depositing in a ring-like pattern after evaporation [28]. Many attempts to suppress and ameliorate the coffee ring effect have thus far focused on manipulating the capillary flow [32–35]. In our strategy, the solution became viscous owing to the existence of massive glucose during the evaporation of the as-prepared colloidal Ag NPs, and the capillary flow is therefore burdened by viscidity resistance, which prevented the suspended particles from reaching the drop edge and ensured their uniform deposition. Figure 3a shows the evaporation process of a drop of as-prepared colloidal Ag NPs dropped onto a clear glass directly. As the water evaporates, the concentration of the glucose is increased, which resulted in an increasing viscosity of the solution [36]. Finally, the droplet had turned into a rigid film and the coffee-ring effect was almost eliminated after evaporation. As the droplet turned to a rigid film, the volume shrunk dramatically, therefore the gap between the Ag NPs would decrease and even turn into an aggregate, which exhibits higher SERS activity [37]. In order to make the rigid film well-shaped and standardize the fabricating process, the colloidal Ag NPs were dropped into an array of holes, which were defined by PDMS structures bonded on the glass substrates (Figure 3b). Figure S3 shows the extinction spectrum of the SERS substrate with 0.2 M glucose. Thanks to the protection of glucose, oxidation of the Ag NPs was also prevented, which resulted in superior stability of the SERS substrates. From the test, the stability and sensitivity of SERS signals could remain for at least 6 months (Figures S4 and S5). The characteristic peaks of R6G were still very distinct after 6 months, and the intensity of the peaks was comparable to the intensity when the substrate was deposited.

**Figure 3.** (**a**) The evaporation process of a drop of the solution, and (**b**) a photograph of colloidal Ag NPs dropped into an array of holes. The holes were 5 mm in diameter, 1 mm in depth, and 10 mm apart for standardized fabrication and convenient measurement of the SERS films in array. (**c**) The fabricated pie-shaped SERS substrates after evaporation. The scale bar is 1 cm.

### *3.3. Optimal Concentration of Glucose*

According to the enhancement mechanism of SERS signals, more sensitive SERS signal could be realized when analytes are close enough to the Ag NPs surface. In our strategy, the glucose protected the Ag NPs from oxidation while preventing the touch between the analyte and the Ag NPs to some extent. Therefore, the optimal concentration of glucose is of grea<sup>t</sup> necessity. In experiments, colloidal Ag NPs with different concentrations of glucose, ranging from 0.1 M to 0.5 M, were used to fabricate the SERS substrates. R6G (10−<sup>8</sup> M) was employed as a SERS probe to evaluate the performance of these SERS substrates. As shown in Figure 4, the SERS signal intensity of these five substrates was in the order: 0.2 M > 0.1 M > 0.3 M > 0.4 M > 0.5 M, suggesting the substrate fabricated with a 0.2 M concentration of glucose showed a stronger signal than the others. When 1 μL of R6G solution was dropped onto the SERS substrates, the water would dissolve the glucose that covered the Ag NPs, and then the Ag NPs were exposed to analytes. Hence, a lower concentration of glucose was more likely to supply bare Ag NPs, resulting in a higher Raman enhancement. As shown in Figure S6, there was a risk of oxidation of Ag NPs with a much lower concentration of glucose (such as 0.1 M). Accordingly, we choose a 0.2 M Ag NPs/glucose solution to fabricate the SERS substrates for later detection.

**Figure 4.** SERS spectra of 10−<sup>8</sup> M R6G as different concentrations of glucose are used to fabricate the SERS substrates.

### *3.4. Sensitivity and Reproducibility of SERS Substrates*

Sensitivity and reproducibility are major concerns for any SERS substrates. Hence, we employed different concentrations of R6G to investigate the performance of the SERS substrates. As shown in Figure 5a, the strong peaks at 612 cm<sup>−</sup>1, 774 cm<sup>−</sup>1, 1127 cm<sup>−</sup>1, 1185 cm<sup>−</sup>1, 1310 cm<sup>−</sup>1, 1360 cm<sup>−</sup>1, 1509 cm<sup>−</sup>1, 1573 cm<sup>−</sup>1, and 1650 cm<sup>−</sup><sup>1</sup> were in good agreemen<sup>t</sup> with previous reports on pure R6G [38–40]. These SERS spectral feature peaks can still be clearly identified even when the concentration was down to 10−<sup>14</sup> M, corresponding to 10 zeptomoles of R6G molecules in a 1 μL sample volume, which is far below previous works [38–41]. The reproducibility of the SERS substrate was further investigated by taking SERS spectra of R6G at the concentration of 10−<sup>10</sup> M from 20 random locations on a single pie-shaped substrate. The average relative standard deviation (RSD) of the intensities (at 1509 cm<sup>−</sup>1) was 6.8% (Figure 5b and Figure S7) which is lower than previously reported [40,42], indicating that the substrates possess good signal uniformity.

**Figure 5.** (**a**) SERS spectra of R6G at a concentration of 10−<sup>8</sup> M to10−<sup>14</sup> M. 1 μL of the R6G solution was dropped to the SERS substrates for detection. The blank is a control group which used deionized water. (**b**) Raman spectra of 10−<sup>10</sup> M R6G collected from 20 random spots. The RSD of the intensity maximum at the peak of 1509 cm<sup>−</sup><sup>1</sup> was 6.8%.

### *3.5. Application for Thiram Detection*

Thiram is a typical sulfur-containing pesticide molecule, which is widely used in agriculture. In this contribution, the SERS substrates were employed to detect thiram for future practical applications. The SERS spectra of thiram with a concentration varying from 10−<sup>6</sup> M to 10−<sup>10</sup> M are shown in Figure 6. The SERS signal intensity gradually increased with the increase of thiram concentration, pointing to the possibility of quantitative analyte determination. The main Raman bands include 563 cm<sup>−</sup><sup>1</sup> attributed to υ(S-S), 1147 cm<sup>−</sup><sup>1</sup> corresponding to ρ(CH3) and υ(C-N), and 1383 cm<sup>−</sup><sup>1</sup> corresponding to *δs*(CH3) and υ(C-N), and 1511 cm<sup>−</sup><sup>1</sup> corresponding to υ(C-N), δ(CH3), and ρ(CH3) [42,43]. We could identify the spectrum of the thiram even at the concentration of 10−<sup>10</sup> M. The result suggested that the ability of SERS substrates to have a detection sensitivity of 10−<sup>10</sup> M of thiram, which is far below the maximum residue limits of thiram for vegetables (5 mg/kg, equal to ≈2 × 10−<sup>5</sup> M). It is evident that such SERS substrates would have grea<sup>t</sup> potential for real world applications.

**Figure 6.** SERS spectra of thiram at a concentration range from 10−<sup>6</sup> M to 10−<sup>10</sup> M. The blank is a control group with DI water.
