**3. Results**

### *3.1. Spectral Features of Tartrazine Dye*

Figure 1 shows the Raman spectrum of tartrazine solid. The peaks at 485, 618, 638, 1129, 1358, 1477, 1503 and 1599 cm<sup>−</sup><sup>1</sup> were easily recognizable. Among these peaks, the most prominent tartrazine peak at 1599 cm<sup>−</sup><sup>1</sup> was assigned to the in-plane bending of OH, the asymmetric stretching of COO<sup>−</sup>, the overlapping effect of symmetric stretching and the out-of-plane C–H deformation of two phenyl rings. The other major peak at 1358 cm<sup>−</sup><sup>1</sup> was attributed to the rocking of the phenyl ring, the symmetric stretching of COO− and the stretching vibrations of –C–N=N–C–. Table 1 shows the assignments of their vibrational bands [25].

**Figure 1.** Raman spectrum of tartrazine.

**Table 1.** Major band assignment of Raman spectra for tartrazine.


<sup>υ</sup>s—symmetric stretching; <sup>υ</sup>as—asymmetric stretching; δ—in-plane bending; γ—out-of-plane bending; ip—in-plane; op—out-of-plane; ρ—rocking; ph—phenyl ring; pyr—pyrazole ring; Def—deformation.

### *3.2. Ag Nanowire Characterization*

The UV-Vis absorbance spectrum of the as-prepared Ag nanowires exhibits two characteristic absorption peaks at 349 nm and 380 nm, as shown in Figure 2a. The peak located at 349 nm is ascribed to the out-of-plane quadrupole resonance of Ag nanowires while another peak with high intensity at 380 nm is attributed to the out-of-plane dipole resonance of Ag nanowires [26].

**Figure 2.** (**a**) Ultraviolet-visible (UV-Vis) absorbance spectrum of Ag nanowires; (**b**) transmission electron microscopy (TEM) images of Ag nanowires.

To determine the normal distribution and a certain amount of homogeneity, we carried out TEM tests. The TEM images of Ag nanowires are shown in Figure 2b. The figure shows that the diameter of the as-prepared Ag nanowires was relatively uniform. An average diameter of 49.4 ± 3.9 nm and length of 7–10 μm were observed by calculating 100 nanowires.

### *3.3. SERS Analysis of Tartrazine*

Figure 3 shows the SERS spectral features of the standard tartrazine solutions. The characteristic peaks of standard tartrazine solution can be clearly distinguished at a concentration as low as 0.01 μg/mL. The limit of detection, calculated based on HPLC, was 5.2 ng/mL [27], which was close to the minimum concentration visually observed by our SERS method.

The strongest peak of tartrazine was still visible at 1599 cm<sup>−</sup>1. However, some peaks may red shift or blue shift compared with the normal Raman spectrum shown in Figure 1. For instance, the characteristic peak at 1503 cm<sup>−</sup><sup>1</sup> in the Raman spectrum of tartrazine red shifted to 1504 cm<sup>−</sup><sup>1</sup> in the SERS spectra. The characteristic peak at 1477 cm<sup>−</sup><sup>1</sup> in the Raman spectrum of tartrazine red shifted to 1474 cm<sup>−</sup><sup>1</sup> in the SERS spectra. The shifts in the characteristic peaks were due to changes in site and adsorption orientations or to changes in polarizability after interactions between the analyte and substrate [28,29]. The as-prepared Ag nanowires as SERS substrate could greatly enhance Raman scattering signals of tartrazine. Based on the method described by Ru et al. [30], the enhancement factors for tartrazine were calculated as 3.9 × 10<sup>6</sup> (based upon the strongest peak at 1599 cm<sup>−</sup>1), the reproducibility of the as-prepared Ag nanowires was reported in our previous study [23].

**Figure 3.** Surface-enhanced Raman scattering (SERS) optical spectra of the standard tartrazine solution (*n* = 40).

As shown in Figure 3, the intensity of the characteristic peaks increased remarkably with increasing concentration. Therefore, the quantitative analysis model of the standard tartrazine solutions may be established based on the SERS spectra. The SERS spectra of tartrazine solutions at eight different concentration levels (10 ng/mL–10 μg/mL) were collected to establish the PLS model (*n* = 32, four spectra for each concentration). As shown in Figure 4, a high linear correlation (R<sup>2</sup> = 0.970) was observed between the actual and the predicted concentrations. High RPD (6.10) and low RMSE (0.26) values also demonstrate that this model has relatively good predictability [16].

**Figure 4.** Partial least square (PLS) model of the standard tartrazine solution.

### *3.4. SERS Analysis of Tartrazine in Large Yellow Croaker*

To simulate the process of the illegal addition of tartrazine to the large yellow croaker, we experimentally spiked large yellow croaker peel extracts with tartrazine (Figure 5a). We aimed to analyze the large yellow croaker matrix from the tartrazine extracts in the large yellow croaker peel experiment. Figure 5a,b show more impure peaks, such as the peak at 1428 cm<sup>−</sup>1, than in Figure 4 due to the interference of non-targeted components. For the first preparation of the W sample, an approximate 12% decrease was observed at the intensity peak of 1599 cm<sup>−</sup><sup>1</sup> in Figure 5a compared with that in Figure 5b, thereby indicating that the recovery rate of the large yellow croaker was approximately 88% of the sample preparation. Similarly, the recovery rates of the preparation of the W+M and W+M+C samples were 84% and 81%, respectively. From W to W+M+C, the decrease in recovery rate may have been due to the increase in purification steps. As shown in Figure 5b, W+M+C had a significant increase in SERS signal, although it had more purification steps than the afore mentioned methods. The W+M+C method took less than 20 mins to prepare the sample, the decrease in recovery rate was minimal and the method showed positive SERS performance. Therefore, this sample preparation method was chosen as the optimal method for further tartrazine detection in large yellow croaker.

**Figure 5.** *Cont*.

**Figure 5.** (**a**) Surface-enhanced Raman scattering (SERS) spectra extract from large yellow croaker peels spiked with 509.55 ng/cm<sup>2</sup> of tartrazine; (**b**) extract added with tartrazine equivalent (a) but assumed 100% recovery rate; (**c**) Raman intensity of tartrazine at 1599 cm<sup>−</sup><sup>1</sup> via different sample preparation methods (*n* = 20).

Tartrazine is used illegally to color fish to make low-priced fish, such as albiflora croaker, appear similar to high-priced yellow croaker or putrescent yellow croaker. Dripping standard tartrazine solutions on a white paper showed that 2.5 ppm was the lowest critical concentration identified by the naked eye, converting 25.48 ng/cm<sup>2</sup> to 200 μL in an area of 19.62 cm2. Hence, the color was completely invisibility when this concentration was added to yellow croaker peels.

We detected tartrazine on the basis of the optimal sample preparation. Figure 6 illustrates the SERS spectra obtained from seven different tartrazine concentrations adsorbed in the large yellow croaker. The spectra revealed that the lowest detectable concentration was 20.38 ng/cm2, which was lower than 25.48 ng/cm2. Thus, the as-prepared Ag nanowires coupled with SERS technology can meet the requirement of zero-tolerance residue limit. Compared to the HPLC method, this method has the advantage of a short detection time. However, there is still a long way to go before the SERS technique overcomes the matrix interference absolutely.

**Figure 6.** Representative surface-enhanced Raman scattering (SERS) spectra of tartrazine extracts from large yellow croaker skin (*n* = 30).
