Flexible Surface-enhanced Raman Scattering Strips Using Colloidal Ink of Gold-silver Alloyed Nanoparticles

Surface-enhanced Raman scattering (SERS) has been used for trace detection at the single-molecule level. The low-cost preparation of high-performance test strips has enabled the development of SERS techniques. In this study, oil-dispersible metal or alloy nanoparticles prepared by the Brust-Schiffrin method were used as "inks" in a ballpoint pen to handwrite SERS test strips on polytetrafluoroethylene (PTFE) membranes. Because of the good PTFE lipophilicity, the flexible substrates had good uniformity. The large laser damage threshold of the PTFE membrane also enabled increased laser powers for SERS testing. The Au and Ag alloy nanoparticle inks exhibited increased performance with larger proportions of Ag. The Au1Ag8 nanoparticles had the best properties, and those strips could detect 10-11-M Rhodamine 6G dyes in a 10-µL volume with an enhancement factor of 5.4×108. The SERS strips were used to demonstrate detection of malachite green, the use of which is prohibited in aquaculture and fish tanks.


Introduction
Surface-enhanced Raman scattering (SERS) spectroscopy has been used for sensitive trace detection [1][2][3] for chemical analysis, archaeological identification, and biological research. SERS substrates are usually rigid, such as glass or silicon. Electron-beam lithography [4], reactive-ion beam etching [5], and focused-ion beam etching [6] have been used to fabricate metallic nanostructures on substrates to obtain high-performance SERS. However, wipeable sampling with flexible SERS substrates is more useful [7]. Recently, Tay et al. [8] prepared flexible SERS substrates via inkjet printing of gold nanoparticles on cellulose-based filter paper to detect fentanyl aerosols for drug screening. By controlling the printing cycle, Ma et al. [9] used screen printing to load "ink" containing graphene oxide and silver nanoparticles on cellulose paper to detect pesticide residues on fruit. Flexible and disposable SERS substrates still usually require various laboratory facilities that increase costs. Polavarapu et al. [10] used a pen-on-paper approach to write nanoparticle sols on plain photocopy paper with a fountain pen to prepare lowcost SERS test strips. They reported that gold nanoparticles, gold nanorods, or silver nanoparticles prepared via aqueous chemical reductions could be used as inks, and those containing Ag nanoparticles exhibited high SERS activity over a broad range of excitation wavelengths. The handwriting method significantly reduced the cost of flexible SERS test strips. In 2017, Han et al. [11] used pen-writing to load aqueous silver nanoparticles on modified cellulose paper for flexible SERS arrays that were used to detect melamine in liquid milk at a 0.27 mg/mL detection limit. However, all the strips noted above reported low laser damage thresholds and were readily ablated during applications. Moreover, the nanoparticle inks were made via chemically reducing Au (III) or Ag(I) in aqueous solutions. These nanoparticles were protected by adsorbed ionic agents that were unstable, which resulted in aggregation, clogged pens, and non-uniform writing. Polytetrafluoroethylene (PTFE) membranes have high laser damage thresholds [12] and stable chemical properties, and thus could be ideal flexible substrates. However, the super-hydrophobicity of PTFE makes it difficult to load aqueous dispersions of metal nanoparticles.
Here, we used oil-dispersed gold nanoparticles synthesized via a modified Brust-Schiffrin method [13] to handwrite, via a ballpoint pen, SERS strips on a PTFE membrane. We also synthesized gold-silver nanoparticle alloys with various Au and Ag ratios to optimize the SERS sensitivity. In this way, the localized surface plasmon resonance (LSPR) and the oxidation resistance of the metal nanoparticles were significantly improved. Overall, the flexible SERS test strips on PTFE substrates exhibited high performance, good stability, and large laser damage thresholds.

Preparation and Characterization of Metallic Nanoparticles
The metallic nanoparticles were prepared via a modified Brust-Schiffrin method [14]. A total of 1.5 g (2.75 mmol) of tetra-octylammonium bromide was added to a two-neck flask with 80 mL of toluene and magnetically stirred. Then 0.066 g (0.38 mmol) of AgNO 3 dissolved in 2 mL of ultrapure water was added and stirred. The toluene solution became milky white with a pale-yellow precipitation. Then 0.088 g (1.52 mmol) of NaCl was added, and the paleyellow flocculent precipitate disappeared within 5 min. A total of 0.16 g (0.38 mmol) of HAuCl 4 ⋅4H 2 O was dissolved in 2 mL of ultrapure water and added to the solution, which then turned orange-red. After stirring for 5 min, 0.36 g (1.88 mmol) of hexanethiol was added, which caused the solution to become colorless. After stirring for 15 min, 0.28 g (7 mmol) NaBH 4 dissolved in 20 ml of water was poured into the flask, and the solution became black. The reaction was finally stopped after stirring at room temperature for 4 h, and the aqueous phase was discarded. The black organic phase was distilled in a rotary evaporator at a temperature below 50 °C to evaporate the solvent. The remaining black product was suspended in 30 mL of methanol, briefly sonicated to ensure complete dissolution of byproducts, collected via centrifugation, and then washed five times with 20 mL of methanol. Upon drying at room temperature in vacuo, alloy nanoparticles with a Au:Ag mole ratio of 1:1 were obtained. Various Au and Ag mole ratios, such as Au, Au 1 Ag 1 , Au 1 Ag 2 , Au 1 Ag 4 , Au 1 Ag 8 , and Ag, could be obtained by changing the AgNO 3 and HAuCl 4 ⋅4H 2 O stoichiometry. The metal nanoparticles synthesized by the Brust-Schiffrin method could be purified and dissolved in most organic solvents for their surfaces were modified by an alkylthiol monolayer through coordinate bond. Therefore, the metal nanoparticle ink concentrations could be precisely adjusted and the colloids did not aggregate.
The gold or gold-silver alloy nanoparticles were dispersed in chloroform to prepare colloidal concentrations of 15 mg/ ml, 35 mg/mL, and 55 mg/ml that were used to fill a ballpoint pen. A 0.5×0.5-cm 2 SERS array was then written on the surface of the PTFE membrane. During the writing, the colloidal nanoparticles diffused along the membrane fibers via capillary effects and spread on the fiber surfaces as the solvent evaporated. Subsequently, the membrane was placed in a 200-°C muffle furnace for 20 min to remove the 1-hexanethiol ligands from the nanoparticle surfaces. At the same time, the nanoparticles fused into 10-50-nm diameter particles.
TEM images of Au, Ag and Au-Ag alloy nanoparticles are shown in Fig. 1a-j. The particles are all spherical with 1-5-nm diameters except for the Ag nanoparticles. The diameter of Ag nanoparticles is 3-10 nm, indicating that Ag tends to grow into large particles under the same synthesis conditions. Au and Ag both have face-centered cubic (fcc) crystal structures and have similar lattice constants; hence, they can form perfect alloys in any proportion. From the high-resolution TEM images in Fig. 1 we calculated approximately 0.24-nm interplanar spacings of adjacent lattice fringes, which corresponded to Au (111) and Ag (111) planes. The absorption maxima for nanoparticles and the color of the colloidal metallic nanoparticles with various proportions of gold and silver are illustrated in Fig. 2. The wavelength of the LSPR extinction peak was linearly blue-shifted with increased Ag content, which indicates that we have obtained the Au-Ag alloy nanoparticles instead of the mixture of Au and Ag nanoparticles. Specifically, the Au 1 Ag 1 absorption peak was at 502 nm, the Au 1 Ag 2 peak was at 482 nm, the Au 1 Ag 4 peak was blue-shifted to 470 nm, the Au 1 Ag 8 peak was centered at 458 nm, and the Ag peak was blue-shifted to 449 nm.

Preparation of Handwritten Flexible SERS Strips
The process of writing on the surface of the PTFE membrane with a ballpoint pen filled with colloidal ink is shown in Fig. 3a. As noted above, the colloidal metal nanoparticles wet and spread on the PTFE surface. By maintaining the approximately 1-mm distance between pen strokes, the metal nanoparticles could be smoothly coated on the fiber surface. Hence, the method was simple and convenient.
Distributions of nanoparticles (35 mg/mL) written on the PTFE membrane were characterized by SEM, as shown in Fig. 4b. Compared with the blank PTFE membrane that had cracks caused by film stretching during fabrication (Fig. 4a), the surface of the written fiber was smoother in Fig. 4b, due to the metal nanoparticles had filled these cracks. After annealing, the nanoparticles fused into larger particles with 10-50-nm diameters. A large number of gaps less than 10 nm formed between adjacent nanoparticles, as shown in Fig. S1. These gaps were SERS "hot spots," where adsorbed analytes exhibited Raman spectra with significantly increased intensities.

SERS Performances of Handwritten Flexible SERS Strips
SERS characterization was performed with a WITec Alpha300A confocal Raman spectrometer. Vibrational Raman spectra of the analyte were acquired in "mapping" mode to characterize the properties of the flexible SERS substrate. The Raman spectrometer was equipped with a synapse charge-coupled device (CCD) detector and a power tunable 50-mW incident 532-nm laser. A 50× objective (NA=0.8) was used, and the laser focal spot was approximately 811 nm in diameter. The scanning area was 25×25 μm 2 , and 2500 spectra were collected from the mapping image. The integration time for each point was 0.5 s. An averaged spectrum over the entire mapping area was calculated as the final test result.
We firstly investigated the effects of ink concentrations (15 mg/mL, 35 mg/mL, and 55 mg/mL) on SERS performance. For Au 1 Ag 1 ink as shown in Fig. 5a, we found that the SERS performance increased as the concentration increase from 15 to 35 mg/mL, but slightly decreased when the concentration increased to 55 mg/mL. For the Au 1 Ag 8 ink as shown in Fig. S2, we found that the SERS intensity of R6G increase slightly with the ink concentration increased from 15 to 35 mg/mL. It can be seen that the ink concentrations ranging from 15 to 55 mg/mL did not significantly affect the SERS performance of the strips. This is because good ink wettability enables the PTFE membrane to store excess nanoparticles. When the colloid concentration was low, the nanoparticles were mainly distributed on the surface as the solvent evaporated. When the colloid concentration increased, there were particles both on the surface and inside the membrane, which prevented the formation of continuous metallic film after annealing. Further increases in concentration (>65 mg/mL) greatly decreased the performance because of excessive fusion of the metallic nanoparticles into solid blocks during annealing that prevented formation of "hot spots". Therefore, the 35 mg/mL concentration was used to characterize the various Au-Ag alloys.
SERS spectra of 10 -6 -mol/L R6G analyte on SERS strips with various Au-Ag alloys are shown in Fig. 5b. The SERS intensity increased with Ag content in the alloys. The intensity from the Au 1 Ag 8 alloy strips was approximately ten times that from the Au strips. We attributed the enhancement to two mechanisms, (1) Electromagnetic (EM) enhancement mechanisms: as shown in Fig. S3, with the increased Ag content on the SERS substrate, the LSPR wavelength blueshifted, which was better for 532-nm excitation to generated a strong localized EM field between adjacent nanoparticles. According to the Finite-Difference Time-Domain (FDTD) calculating results shown in Fig. S4a, the Raman enhancement is proportional to the fourth power of the EM field intensity, so that the R6G Raman scattering was greatly enhanced from the Au 1 Ag 8 alloy strips. (2) charge-transfer enhancement mechanisms: the Fermi level of Au-Ag alloy nanoparticles increases with the increase of the Ag content [15], which facilitated the transfers of hot electrons from Au 1 Ag 8 alloy to the lowest unoccupied molecular orbital (LUMO) of the adsorbed R6G. The electron transfers leaded to increased vibration of the R6G molecules, and subsequently enhancing the Raman scattering on Au-Ag alloy SERS strips.
In Fig. 5b, the Ag strips did not obtain the strongest SERS intensity, which is different from our previous report [14]. We then characterized the samples by SEM, as shown in Fig. S5, we found that the Ag nanoparticles were aggregated into large particles on the PTFE strip. Considering the poor wettability of Ag colloids on PTFE membrane during handwriting, we believed that it was difficult for colloidal Ag nanoparticles to spread smoothly on the membrane because of their larger sizes [shown in Fig. 1e, j]. Therefore, it was difficult to form high-quality "hot spots" on Ag strips.
Using the Au 1 Ag 8 alloy, we characterized the SERS performance for various R6G concentrations (10 -6 -10 -12 M), as shown in Fig. 6. The SERS intensity decreased with decreasing of R6G concentration; characteristic peaks were still visible at 10 -11 M. However, when the concentration was 10 -12 M, the SERS spectrum was not uniformly detected at all mapping locations. Therefore, for the Au 1 Ag 8 alloy and a 10-μL loading volume, 10 -11 M R6G was the detection limit. With Eq. (1), we calculated the number of R6G molecules loaded on a single mapping point at any concentration. Surface-enhanced Raman scattering spectra for various R6G concentrations (10 -6 -10 -11 M) on Au 1 Ag 8 alloy test strips. (Spectra obtained at 10 -11 M were amplified 10-fold for clarity). All spectra were acquired with 532-nm excitation where S l was the area of the laser spot, S A was the distribution area of the solution to be detected, N E was the number of molecules detected in a single laser spot, and N A was the total number of molecules detected. For the 10 -11 -M R6G concentration, there were 6 molecules in a single laser spot; and for 10 -12 M R6G, it was 0.6 molecules, and the detailed calculations have been provided as text S1 in the supporting information. This explained why the 10 -12 -M solution could not be uniformly detected. The flexible SERS strips had a high R6G sensitivity. From Eq. (2), we can also calculate the enhancement factor (EF) of the substrate.
where I SERS was the intensity of a specific analyte Raman band, and N SERS was the number of molecules contributing to I SERS . Similarly, I REF was the intensity of the same Raman band from a reference solution, and N REF was the number of molecules contributing to I REF . The EF was 5.4×10 8 for the 10 -11-M R6G 613-cm -1 peak from Au 1 Ag 8 alloy strips, and the detailed calculation process can be found in text S2 in the supporting information. According to recent results on flexible SERS substrates [9][10][11], the enhancement factor of our flexible SERS strip was one of the best results.
To demonstrate the uniformity of the Au 1 Ag 8 alloy SERS strips, spectra for more than 2500 points were collected in a 25×25-μm 2 mapping area, with 0.5-s integration times at each point. The intensity of the 613-cm -1 Raman peak is plotted in Fig. 7a, and the average SERS spectrum over the entire mapping area is shown as red in Fig. 7b. The maximum and minimum spectral intensities were plotted as blue and black curves, respectively, in Fig. 7b. The high-intensity (bright) areas and the low-intensity (dark) areas were randomly distributed in the 2D mapping image, and the minimum-intensity spectrum still exhibited characteristic R6G peaks. Fig. 7c shows cross-sections of the mapping image in Fig. 7a, which reflect intensity fluctuations of 613-cm -1 and 1362-cm -1 peaks. From Fig. 7c, we calculated 18% and 31% relative standard deviations (RSD) for the 613-cm -1 and 1362-cm -1 peak intensities, respectively. The uniformity of the PTFE SERS strips was mainly determined by the membranes. Here we used the membranes with a pore size of 0.2 µm. In Fig. S7, we also provided the results for a membrane with 0.1µm pore size, which had the optimizing performances on an area of 50×50-μm 2 .
To test the laser damage threshold of our SERS strips, we focused a 532-nm laser on the stripes using a 100× microscope objective (NA=0.9) and tested the changes on the strips using the "mapping" mode of a Raman spectrometer. Usually, the laser power we used in SERS testing is 1.8 mW. As we gradually increased the laser power to 8.5 mW, the changes of the SERS strip can be observed from the optical microscope, as shown in Fig. S8b. Subsequently, we characterized the morphology by the SEM, as shown in Fig. S8c, d, and we found that the metallic nanoparticles on the strips fused into large particles due to the laser-induced welding effect, while the PTFE membrane did not change significantly. Therefore, the laser damage threshold of PTFE flexible SERS strips was mainly depended on the laser welding threshold of the metallic nanoparticles.
The stability of the Au-Ag alloy SERS strips is also an important issue needs to be characterized. In our previous work [14], we investigated the stability of the Au, Ag and Au 1 Ag 8 alloy on rigid substrates by comparing the properties of freshly prepared SERS substrates with those of stored in air for several weeks. We found that the Au substrate had a very stable SERS performance and its SERS spectrum did not decrease in intensity with increasing storage time. However, after 4 weeks, the SERS intensity from the Ag substrate became very weak. The intensity of the SERS intensity from the Au 1 Ag 8 alloy substrate stabilized after 2 weeks and no longer decreased, which is attributed to the stabilizing effect of Au preventing further oxidation of the substrate. In this work we investigated the stability of the Au 1 Ag 8 and Ag strips by comparing the properties of freshly prepared SERS strips with those of stored in air for several days and the results are shown in Fig. 8. We found in Fig. 8a that in the first two days, the SERS intensity of the Au 1 Ag 8 sample did not decrease. On day 7, the intensity decreased to nearly half of that of the freshly prepared sample. However, on day 2, the SERS intensity of the Ag sample was reduced to nearly half of that of the freshly prepared sample, as shown in Fig. 8b, and reduced to one-sixth of initial intensity on day 6. It can be seen that the Au 1 Ag 8 alloy SERS substrates not only had better stability than Ag sample, but also had better SERS sensitivity than Au sample.

Applications to On-Site Sampling and Detecting
Malachite green (MG) was once widely used as a fungicide in aquaculture. However, because of its carcinogenic, teratogenic, and mutagenic hazards [16], most countries have banned it since the 1990s. Nevertheless, because of its low cost and bactericidal effects, MG can still be found in aquaculture.
Using Au 1 Ag 8 SERS strips and beakers, we simulated MG sampling of a fish tank surface. We added various concentrations of aqueous MG solutions to the beakers, and let them stand for 4 h. We then poured out the solutions and wiped the beakers with SERS test strips that were wetted with ethanol, to mimic sampling of MG residues on the surface of fish tanks in a market. The test results for various MG concentrations are shown in Fig. 9. The laser wavelength was 532 nm, the power was 0.6 mW, and the integration time was 1s. Residual MG on the inner surface of the beaker was still clearly detected when the 10 -7 -M MG solution was contained in the beaker. This demonstrated that the SERS substrates could be used as test strips for on-site sampling and detection.

Conclusions
We used plasmonic inks dispersed in oil to handwrite with a ballpoint pen high-performance SERS test strips on the surface of PTFE membranes. The test strips exhibited good reproducibility because the SERS nanoparticles were uniformly spread on the PTFE fibers. The membrane enabled flexible SERS strips with high chemical stabilities and large laser damage thresholds. The plasmonic nanoparticles of Au, Ag, and their alloys were prepared via a modified Brust-Schiffrin method. The substrates patterned with Au-Ag alloy nanoparticle inks exhibited increased SERS performance with Ag content, where the Au 1 Ag 8 alloy had the best performance. The wipeable capability of the flexible SERS strips has potential applications for on-site sampling. As SERS functionalized membranes, they are also expected to be applied to lateral flow strips or gas detectors.

Authors' Contributions
Honghao Tian, Youjian Qin and Hongmei Liu completed the experiments of the thesis. Tian Li and Yuting Li characterized the metal nanoparticles. Honghao Tian and Hongmei Liu completed the data collation and the draft paper. Hongmei Liu, Xiaohui Fang and Xinping Zhang discussed and revised the paper.