In Situ Raman Monitoring of Trace Antibiotics in Different Harsh Water Environments

In situ surface‐enhanced Raman scattering (SERS) is a widely used operando analytical technique, while facing numerous complex factors in applications under aqueous environment, such as low detection sensitivity, poor anti‐interference capability, etc., resulting in unreliable detectability. To address these issues, herein a new hydrophobic SERS strategy has been attempted. By comprehensively designing and researching a SERS‐active structure of superhydrophobic ZnO/Ag nanowires, we demonstrate that hydrophobicity can not only draw analytes from water onto substrate, but also adjust “hottest spot” from the bottom of the nanowires to the top. As a result, the structure can simultaneously concentrate the dispersed molecules in water and the enhanced electric field in structure into a same zone, while perfecting its own anti‐interference ability. The underwater in situ analytical enhancement factor of this platform is as high as 1.67 × 1011, and the operando limited of detection for metronidazole (MNZ) reaches to 10−9 M. Most importantly, we also successfully generalized this structure to various real in situ detection scenarios, including on‐site detection of MNZ in corrosive urine, real‐time warning of wrong dose of MNZ during intravenous therapy, in situ monitoring of MNZ in flowing wastewater with particulate interference, etc., demonstrating the great application potential of this hydrophobic platform. This work realizes a synergistic promotion for in situ SERS performance under aqueous environment, and also provides a novel view for improving other in situ analytical techniques.

In situ surface-enhanced Raman scattering (SERS) is a widely used operando analytical technique, while facing numerous complex factors in applications under aqueous environment, such as low detection sensitivity, poor antiinterference capability, etc., resulting in unreliable detectability.To address these issues, herein a new hydrophobic SERS strategy has been attempted.By comprehensively designing and researching a SERS-active structure of superhydrophobic ZnO/Ag nanowires, we demonstrate that hydrophobicity can not only draw analytes from water onto substrate, but also adjust "hottest spot" from the bottom of the nanowires to the top.As a result, the structure can simultaneously concentrate the dispersed molecules in water and the enhanced electric field in structure into a same zone, while perfecting its own anti-interference ability.The underwater in situ analytical enhancement factor of this platform is as high as 1.67 × 10 11 , and the operando limited of detection for metronidazole (MNZ) reaches to 10 −9 M. Most importantly, we also successfully generalized this structure to various real in situ detection scenarios, including on-site detection of MNZ in corrosive urine, real-time warning of wrong dose of MNZ during intravenous therapy, in situ monitoring of MNZ in flowing wastewater with particulate interference, etc., demonstrating the great application potential of this hydrophobic platform.This work realizes a synergistic promotion for in situ SERS performance under aqueous environment, and also provides a novel view for improving other in situ analytical techniques.
Experimental and theoretical results demonstrate hydrophobicity can not only facilitate the aggregation of analytes from water onto the SERS structure via the strong adsorption effect between low surfaceenergy groups and target molecules but also assemble hotspot and target molecules into a same zone, achieving multiplicative breakthrough in integration between analytes and enhanced localized electrical fields (Figure 1a).Benefitting from these, the superhydrophobic ZnO/Ag platform exhibits excellent in situ SERS sensitivity, and its operando limited of detection (LOD) for metronidazole (MNZ) reaches to 10 −9 M, which is comparable to and even far lower than many known results obtained by ex situ analytical techniques. [2,8,26,27]Moreover, the hydrophobicity also endows the ZnO/Ag NWs with outstanding robustness.As a proof of application, we successfully generalized this structure to various in situ detection scenarios, such as on-site detection of MNZ in corrosive urine, real-time warning of wrong dose of MNZ during IV therapy, in situ monitoring of MNZ in flowing wastewater with particulate interference.This work realizes a synergistic promotion for in situ SERS performance under aqueous environment, and also provides a novel view for improving other in situ analytical techniques.

Construction and Characterization of Superhydrophobic ZnO/Ag NWs
In order to ensure the photopermeability of the SERS platform in in situ detection, the superhydrophobic ZnO/Ag NWs were prepared on a transparent cover glass (Figure 1b).The ZnO/Ag NWs are gently slant and distribute randomly (Figure 1c and Figure S1a,b, Supporting Information), and the width and length are ca.60-80 nm and 0.3-0.5 μm, respectively (Figure S1d,e, Supporting Information).[30] In this work, the optimal ZnO/Ag NWs can maintain a static contact angle (CA) of 159.8°for a water droplet of 4 μL (inset in Figure S1a, Supporting Information).The Ag nanofilm with thickness of 35 nm (Figure S1f, Supporting Information) is cap-liked and uniformly wraps part of the ZnO top (Figure 1c,d and Figure S1b, Supporting Information).According to XRD patterns (Figure 1f), the Ag and ZnO are, respectively, cubic and hexagonal phases and are both with high crystallinity.The high crystallinity is further verified by the clear crystal lattices in HRTEM image of the ZnO/Ag NW (inset in Figure 1d), and the lattice intervals are ca.2.87 Å and 2.34 Å, corresponding to lattice planes (100) of ZnO and (111) of Ag, respectively.Figure 1g 1 -g 4 and Figure S2, Supporting Information, respectively, show the energy dispersive spectrum (EDS) mappings of different elements collected in single ZnO/Ag NW and corresponding ZnO/Ag NWs substrate, which both exhibit uniform distributions, and the existence of element F stems from the hydrophobic group of perfluorodecyltriethoxysilane.All these results demonstrate the successful construction of the superhydrophobic ZnO/Ag NWs.
On account of the wide optical band gap (3.4 eV), ZnO NWs show no absorption in visible region (Figure 1e).Therefore, the optical photograph of ZnO NWs is transparent (Figure S1c, Supporting Information).In actual in situ SERS detection, excited laser will penetrate from back of the glass and enter into the ZnO/Ag NWs (cartooned in Figure 1a), and thus transparent ZnO NWs can be used as an ideal framework without optical loss.When Ag is covered, the substrate turns into khaki (Figure S1c, Supporting Information).The absorption peak is found at 498 nm and the full width at half maximum is from 451 to 544 nm, just covering the Raman exciting laser (532 nm) in our test and ensuring the high sensitivity of in situ SERS.In a noteworthy manner, the ZnO/Ag NWs before and after hydrophobic modifications (AM, BM) exhibit same absorbance and phase (Figure 1e,f), implying that hydrophobic modification has less influence on these two properties.

Influence of Structure on Wettability
However, hydrophobic modification is closely related to the contact way between the water and the substrate, [31][32][33][34] directly determining the aqueous in situ detection sensitivity and reliability.Therefore, before further applications, correlation between the wettability and the ZnO/Ag NWs should be clarified.To figure this out, we take growth time of ZnO and deposition time of Ag as variable parameters.Figure 2c (red marks) firstly shows the CA changes of ZnO NWs (AM) with different growth time (GT), in which three wettability states are observed.For the bare glass (GT = 0 s), that is, the state I, the CA value is ca.106.4°(Figure 2b 1 ), indicating that the water droplet is in Wenzel state (Figure 2a 1 ).To further investigate the dynamic interaction between the water and the substrate, we then tested the water droplet motion behavior, as exhibited in Figure 2d 11 -d 15 : A syringe with 4 μL of water droplet suspended on head firstly moved down to the surface of the substrate followed by a slight press (Figure 2d 11 -d 13 ), then the syringe was slowly raised (Figure 2d 14 ,d 15 ).During this process, a large contact area at the interface between the liquid and the solid was observed (Figure 2d 14 ), and the droplet was even pulled out from the syringe finally (Figure 2d 15 ), indicating that the surface is very sticky.The contact angle hysteresis (CAH) [35,36] in this case is ca.41.7°, also verifying this conclusion (Figure S3, Supporting Information).Although the structure is hydrophobic in this state, the viscous surface will easily cause particulate blockage when it is in complex water environment, [37,38] thereby lowing the aqueous in situ detection performance of this structure.With more ZnO NWs covering glass (GT from 10 to 60 s, Figure 2b 2 and Figure S4, Supporting Information), the maximal CA value is increased to 159.8°and the corresponding contact model transitions to Cassie-Baxter state, that is, the state II (Figure 2a 2 ).Water droplet motion behaviors (Figure 2d 21 -d 25 and Figure S5, Supporting Information) demonstrate that the adhesion phenomenon fades away at this time and the minimum CAH value is only 3.05 o as well (Figure S3, Supporting Information), heralding that viscous interaction between the moving water and the substrate will be very weak during in situ detection.While, when ZnO grows further (GT = 90 s), the adhesion effect re-appears (Figure 2d 31d 35 ), and the CAH exceeds 18.5°(Figure S3, Supporting Information).According to the SEM image (Figure 2b 3 ), we find that the NWs become thicker and adhere to sidewalls between each other.
Based on this fact, we speculate that the adhesion effect mainly originates from the change of the three-phase contact line (TCL) of gas/ liquid/solid from discontinuous to continuous (Figure 2a 2 ,a 3 ), that is, the state III.To demonstrate it, we measured the surface energy (σ) of different ZnO NWs. Figure 2c presents that the σ values of short GT (20-60 s) are only from 0.41 to 0.71 mN m −1 , while the one of long GT (90 s) is as high as 4.85 mN m −1 .At microscopic scale, the change in TCL mainly reflects the alteration in contact area between the droplet and the substrate.[41] As a proof, the CA value in continuous TCL case is ca.135.2°(Figure 2b 3 ), declining 24.6°in contrast with the best one, further demonstrating our speculation.Therefore, this kind of hydrophobic structure is not suitable for in situ detection as well.Figure S6, Supporting Information, shows the SEM images of ZnO/ Ag NWs with different deposition time of Ag (from 0 s to 100 s).It can be observed that these NWs are essentially the same, with a slight increase in thickness of the NWs as the Ag deposition time increases.Compared with the big change of CA values with different growth time of ZnO, the CA values with different deposition time of Ag only fluctuate in a small range from 156.2°to 158.8°, indicating that the Ag has weaker effect on hydrophobicity.

Influence of Structure on In Situ SERS Performance
Based on above results, the in situ Raman enhancement of these samples was subsequently tested in order to determine an optimal structure suitable for on-site Raman detection, during which R6G aqueous solution was chosen as the probes.For ZnO NWs with and without hydrophobic modifications, both of them can only recognize R6G at concentration of 10 −2 M (Figure 2f).The corresponding spectrum intensity is barely ca.4-times higher than that of the bare glass, implying the Raman enhancement of ZnO is weak.By contrast, ZnO/Ag NWs show strong Raman enhancement.Even for the R6G of 10 −8 M, the characteristic peaks can be clearly observed via ZnO/Ag NWs (Figure 2e).Experimental results indicate that the ZnO/Ag NWs with ZnO growth time of 25 s and Ag deposition time of 80 s show the best in situ Raman enhancement.Electrical field distributions (EFDs) in above structures were also simulated to illustrate the difference between them during in situ Raman detection, and the corresponding results are shown in Figure 2g 1 -g 8 .As observed, changes in both the maximum and average intensity of simulated electrical field exhibit the same trend as the Raman spectral change in Figure 2e,f.The strongest EFD is achieved in the structure of ZnO/Ag NWs with ZnO growth time of 25 s and Ag deposition time of 80 s, which is in highly consistent with the experimental conclusion.According to the fourth power law, [42,43] the theoretical maximum and average values of the electromagnetic enhancement factor (EF) can be estimated as 5.47 × 10 6 and 6.09 × 10 3 (Figure 2 g 9 ), respectively, predicting the fine in situ SERS sensitivity of the superhydrophobic ZnO/Ag NWs.

Synergistic Promotion for In Situ SERS Performance by Wettability and Localized Electric Field
In order to further evaluate the comprehensive in situ SERS detectability of the superhydrophobic ZnO/Ag NWs, a series of Raman tests were made in this work.The Raman spectra of R6G with different concentrations were firstly in situ acquired via the superhydrophobic ZnO/Ag NWs, and the results are exhibited in Figure 3a.As observed in this figure, the LOD of the superhydrophobic ZnO/Ag NWs reaches to 10 −14 M, approaching to the single-molecule level, and the analytical EF is as high as 1.67 × 10 11 .These results are superior to many know ones of the in situ SERS detection, as listed in Table S1, Supporting Information.One interesting phenomenon is that the characteristic peaks of R6G acquired on superhydrophobic ZnO/Ag NWs are always stronger than those obtained on hydrophilic ones (Figure 3d), Energy Environ.Mater.2024, 7, e12517 3 of 10 implying that hydrophobic modification contributes to the improvement of in situ SERS sensitivity of the ZnO/Ag NWs.To verify this result, we further tested three other analytes including 4-ATP, MG, and CV and obtained same conclusions (Figure S7, Supporting Information).To the best of our knowledge, this phenomenon has not been mentioned in previous reports on aqueous in situ SERS detection.Energy Environ.Mater.2024, 7, e12517 Why hydrophobic substrate is better in aqueous in situ SERS detection?We suspect that one possible reason is the hydrophobic functional groups (perfluorodecyltriethoxysilane), which makes ZnO/Ag NWs with low surface energy and thus shows higher affinity to analytes dispersed in solution. [44,45]To verify this view, the adsorption energies and charge density differences between interfaces of R6G/perfluorodecyltriethoxysilane (R/P) and R6G/Ag (R/A) were, respectively, estimated by the density functional theory simulations.As exhibited in Figure 3e, adsorption energy at R/P interface is only 0.925 eV, while the one at R/A interface is 2.23 eV.In addition, the R/P interface Energy Environ.Mater.2024, 7, e12517 exhibits more obvious charge transfer than R/A interface at the same isosurface (0.001 eV), especially around the atom F, illustrating that perfluorodecyltriethoxysilane has stronger interaction with R6G molecule.This fact demonstrates the superhydrophobic ZnO/Ag NWs is easier to capture molecules from aqueous solution onto substrate and verifies our conjecture.Beyond this, another possible reason we think is the re-distribution of the electric field caused by hydrophobic modification.For superhydrophobic ZnO/Ag NWs under Cassie-Baxter state, the nanowires can only partially contact with water, [46] thus microscopically causing a sudden change in refractive index around ZnO/Ag NWs and in turn affecting the electric field distribution in corresponding space (cartooned as Figure 3f).By comparing the simulated EFDs in ZnO/Ag NWs with different wettability (Figure S8, Supporting Information), we found the electric field inside the structure that partially contact with water (Figure S8c, Supporting Information) shows stronger intensity than the one fully in contact with water.The electric field on two sides of the water/air interface is discontinuous (marked as P, i.e., the phase line, in Figure S8, Supporting Information), and the electric field intensity in water side is 2-3 times stronger than that in air side.In actual detection, the analytes are all dispersed in water, thus this result means the superhydrophobic ZnO/Ag NWs can focus electric field into the region with most molecular distribution.In contrast, above phenomena cannot be observed in the hydrophilic ZnO/Ag NWs (Figure S8a,b, Supporting Information), hydrophobic and hydrophilic ZnO NWs (Figure 2g 2, g 3 ), which are, respectively, due to the continuous medium environment (hydrophilic ZnO/Ag NWs) and loss of LSPR effect (hydrophobic and hydrophilic ZnO NWs). Figure 3g exhibits the EFDs at different x-y cross sections (marked as 1 and 2 in Figure S8a-c, Supporting Information), in which the strongest EFD are all found in the superhydrophobic ZnO/Ag NWs, illustrating the enhanced effect of hydrophobicity on the electric field in water.The above mechanism demonstrates the unique advantage of the superhydrophobic ZnO/Ag NWs on in situ SERS detection under aqueous environment, which can not only facilitate the aggregation of the analytes from water onto ZnO/Ag NWs, but also improve the matching level between analytes and hotspots, ensuring the high in situ SERS sensitivity of this structure.Figure 3b exhibits a good linear relationship between the spectral intensity and the analyte concentration, and the mean regression coefficients (R 2 ) is up to 0.957, indicating that the structure is with good quantifiability during in situ detection.Figure 3c is the Raman mapping of peak 613 cm −1 in an area of 20 μm × 20 μm.In this pattern, the uniform color pixel distribution can be observed and relative standard deviation (RSD) value is only 8.63%, further implying the fine in situ detection uniformity of this superhydrophobic substrate.All these phenomena together demonstrate the excellent comprehensive in situ SERS detection capability of the superhydrophobic ZnO/Ag NWs.To make the conclusion more convincing, we also chose 4-ATP, MG, and CV as reference and did same test as above.The detection sensitivity, quantification, and homogeneity of the substrates to these molecules are consistent with R6G, for which the results can be found in Figure S9, Supporting Information.In addition, the SERS stability test of the superhydrophobic ZnO/Ag NWs was also made in this work.As shown in Figure S10, Supporting Information, the enhancement effect of the substrate continues to decline within 2 weeks, which is mainly attributed to the oxidation of Ag.However, the decrease becomes flat after 10 days, and the characteristic peaks of R6G can be still clearly detected on the substrate at this time.Based on these researches, we prove that the superhydrophobic ZnO/Ag NWs has excellent in situ SERS detectability under water environment.

Applications of MNZ Detection in Various Water Environments
Metronidazole is one of the most pharmaceutical-effective antibiotics, but usually overused in both human beings and animals, leading to increasing potential risk of cancer and genetic mutations. [2,4]In most cases, MNZ distributes in aqueous solution, and thus detecting MNZ precisely by in situ SERS under water environment is necessary and meaningful.However, complexities in real water environments, for example, analytes with low concentration, corrosion of acid/alkali, interference by particulate matters, moving media, etc., bring immense hardship for in situ SERS detection of MNZ in water.At this stage, there is no effective and general in situ SERS strategy for detecting MNZ under aqueous solution, aiming to solving all above difficulties.Herein, we tried to use this superhydrophobic Raman platform to achieve above goal.
In situ detectability of the superhydrophobic ZnO/Ag NWs for MNZ was firstly evaluated.
According to the numerical simulations of the adsorption energies and charge density differences between interfaces of MNZ/perfluorodecyltriethoxysilane and MNZ/Ag (Figure S11, Supporting Information), we found that the superhydrophobic ZnO/Ag NWs are beneficial to capture MNZ as well, which is in consistent with the conclusion achieved previously, indicating the fine in situ detectability of superhydrophobic ZnO/Ag NWs.In addition, we also examined the x-ray photoelectron spectra of the superhydrophobic ZnO/Ag NWs and the superhydrophobic ZnO/Ag NWs after immersing into MNZ or R6G solutions, respectively.We found the characteristic peak of CF2 (689.08 eV) undergoes a large shift for the substrate after soaking in solution, as shown in Figure S12, Supporting Information, in which the shift is ca.0.3 eV for R6G (688.78 eV) and ca.1.05 eV for MNZ (688.03 eV).This phenomenon further proves the strong adsorption effect of perfluorodecyltriethoxysilane for R6G and MNZ.Moreover, the shift between the substrates those, respectively, immersed into MNZ and R6G is ca.0.75 eV, corresponding well to the above simulated value of 0.581 eV. Figure 4a 1 shows the Raman spectra of MNZ solutions with different concentrations in situ acquired from superhydrophobic ZnO/Ag NWs.Characteristic peaks at 823 cm −1 , 1188 cm −1 , and 1283 cm −1 correspond, respectively, to the vibrational mode of δC-H rocking, in-plane bending, and twisting of MNZ molecule, [26] which can be easily identified in all the spectra in Figure 4a 1 .Even for Raman spectrum with concentration of 10 −9 M, its signal to noise ratio is higher than 4.2, confirming the outstanding Raman signal amplification capability of this SERS platform for MNZ.Average R 2 of fit lines between peak intensity and MNZ concentration exceeds 0.93 (Figure 4a 2 ), indicating this superhydrophobic platform can be utilized for quantifying MNZ.As a testament to these conclusions, we analyzed three MNZ-containing drugs, including metronidazole tablet, galculus bovis and metronidazole capsules, metronidazole gel.We, respectively, dissolved 1.71 × 10 −6 g of these medicines in 1000 mL of water and collected corresponding in situ Raman spectra, and the results are shown in Figure 4a 3 .According to the fit equations in Figure 4a 2 , MNZ dosages in these three drugs can be estimated as 1.41 × 10 −6 g, 3.01 × 10 −6 g, and 1.20 × 10 −6 g, respectively, basically equaling to the contents mentioned in the instruction manual.

of 10
Notable is, differences between peaks of 639, 1008, and 1129 cm −1 are observed in Raman spectra of these three drugs, and further principal component analysis shows that the confidence ellipses of these Raman spectra highly separated (Figure 4a 4 ), indicating this superhydrophobic platform is strongly discriminative for MNZ, even for similar components.Based on these phenomena and results, it is reasonable to believe the superhydrophobic ZnO/Ag NWs are with excellent in situ detectability for MNZ.However, the detections above are all static, and in practical applications, dynamic detection of MNZ is also important.For example, intravenous antibiotics are commonly used in clinic, yet more than half of life-threatening errors during clinical treatment have recently been proved to be related to medication. [47]Continuously monitoring the contents or concentrations of antibiotics within an IV therapy line can provide an additional error-checking layer and thus avoid the occurrence of dosage errors.Thus, dynamic detectability for MNZ of this Energy Environ.Mater.2024, 7, e12517 superhydrophobic platform was tested here as well.To simulate emergency in IV process, we designed a microfluidic device, as shown in Figure 4b 1 ,b 2 .In this device, the superhydrophobic ZnO/Ag platform was put inside a microfluidic chip and monitored by Raman spectrograph.During in situ detection, the 1 # injection pump continuously injected MNZ solutions with concentration of 10 −8 M into the chip.While, the 2 # injection pump periodically injected the MNZ solutions with concentration of 10 −6 M at time interval of 30s, and the dose for each injection was doubled from the previous one.These two embranchments were connected via a T-junction.Within 30 min, we saw that the spectra intensity (Figure 4b 3 ,b 4 ) keep rising.Assuming that the real-time concentration cannot exceed 10 −7 M for patient during IV therapy, we found that at 15 min the MNZ in microfluidic line reaches to the setting safe dose, offering a timely warning for potential wrong use of MNZ.In addition, this in situ dynamic detectability can also be utilized in assessment of sustained drug release in vitro.
Monitoring drug concentration changes in body fluid for patients under antibiotic therapy is another very important application in clinic, which can reflect timely health indicators of patients. [27]However, in many cases, human body fluids are corrosive, for example, the urine.Human urine is weakly acidic, and its pH value is between 4 and 8.If the urine is detected by common SERS substrate directly, corrosivity may cause structural damage and reduce the detection accuracy of the substrate.As a proof of this fact, we found that there are no intact structures on surface of the ZnO/Ag NWs without hydrophobic modification after touching with acid liquor (pH = 4) for only 10 min (Figure S13, Supporting Information).In contrast, the structure of superhydrophobic ZnO/Ag NWs is well preserved (Figure S13, Supporting Information).We used this superhydrophobic ZnO/Ag platform to in situ collect Raman signals from MNZ aqueous solutions with different pH values (Figure 4c 1 , pH from 4 to 8), and no obvious differences between them are observed (Figure 4c 2 ), proving the fine anti-corrosion of this structure.Based on this property, we assembled this platform into drainage bag (Figure 4c 3 ), which is usually used for patients after surgical procedures, to tested its real in situ detectability for MNZ in urine.First, Raman spectrum of artificial urine without MNZ was on-site collected as reference (Figure 4c 4 ), which does not show any characteristic peaks.While, once MNZ (10 −6 M) was added, the peaks of 822 cm −1 and 1186 cm −1 were observed, demonstrating the successful identification of MNZ in urine.Thus, even under corrosive body fluid environment, this superhydrophobic ZnO/Ag NWs also exhibit fine in situ detectability for MNZ.Weak signal intensity here is mainly due to the optical loss by the frosted calendered polyvinyl chloride film of the drainage bag.
In fact, besides utilization in clinical medicine, MNZ monitoring in daily life is also necessary.Today, although many countries have banned the use of MNZ in food production, there is ample evidence that antibiotics are found in a variety of aquatic environments, such as rivers, groundwater, waste water and even drinking water, and thus MNZ is regarded as an emerging contaminant in water. [2,4]Monitoring MNZ in such environments faces many difficulties, one of which is the particulate matter interference, and thus the related SERS structure is usually required to have strong anti-pollution ability.To evaluate this potential, we respectively put the bare and superhydrophobic ZnO/Ag platforms into muddy water with strong stirring, and the stirring speed is ca.60 rpm min −1 (Figure 4d 1 ,d 2 ).After 30 min of stirring, there is no morphology change for the superhydrophobic structure (Fig- ure 4d 4 ), and the in situ SERS enhancement is the same as before (Figure 4d 3 ).However, for the bare ZnO/Ag, the SERS enhancement almost disappears (Figure 4d 3 ).At micro level, its surface is blocked by many particulate matter (Figure 4d 4 ) preventing light from entering the depths of the structure.This test indicates that the superhydrophobic ZnO/Ag platform is with strong robustness, which is suitable to utilization in detection of MNZ in living aquatic environment.
Thus, based on all above tests, it is reasonable to believe that this superhydrophobic ZnO/Ag NWs have excellent in situ SERS detectability for MNZ under various harsh water environments.

Conclusion
In summary, a novel hydrophobic SERS strategy has been proposed and researched in this paper, aiming to facilitate synergistic promotion for in situ SERS performance under aqueous environment.According to systematic experiments and numerical simulations, hydrophobicity has been proven can not only draw analytes from water onto substrate, but also adjust "hottest spot" from bottom of the structure to the top, improving the matching or integration degree between the molecule and the electric field.Using superhydrophobic ZnO/Ag NWs as a model, the in situ SERS AEF of this structure is as high as 1.67 × 10 11 , and the operando LOD for MNZ reaches to 10 −9 M, above of which were both on-site obtained under aqueous environment.It is worth mentioning that we further extended this strategy in detection of MNZ under various water environments successfully, including on-site detection of MNZ in corrosive urine, real-time warning of wrong dose of MNZ during intravenous therapy, in situ monitoring of MNZ in flowing wastewater with particulate interference, etc., all achieving satisfactory results.This work provides a brand new strategy for the improvement of in situ Raman sensing of trace analytes in various harsh water environments and are also instructive for other in situ analytical techniques.
was a laser of 532 nm with power of 0.48 mW and the laser spot diameter was ca. 1 μm.The integration time was set to 4 s and the diffraction grid was selected as 600 g mm −1 .To ensure the reliability of the data, each spectrum in this work was an average of five randomly measured spectra, and the error bars were obtained according to the standard formula of the measurement uncertainty.The in situ Raman detection described in the manuscript refers to float the SERS substrate on water surface with its back side upward, the incident light irradiates from the back side of the cover glass into water.
Analytical enhancement factor: Experimental AEF of the superhydrophobic ZnO/Ag NWs was estimated as below: In this formula, the I SERS and I RS are intensities of SERS and normal Raman scattering, respectively, and the N SERS and N RS are numbers of analytes that contribute to inelastic scattering intensity in SERS and normal Raman scattering measurements, respectively.Because the Raman signals were in situ collected from the superhydrophobic ZnO/Ag NWs that reversely floated on water, the N RS /N SERS can be estimated as: Where the N A is Avogadro's constant, the C (C RS and C SERS ) represents analyte concentration and the V (V RS and V SERS ) is corresponding volume focused by laser spot.In this work, the V RS basically equals to the V SERS , Thus, the AEF can be estimated as: In this paper, the blank glass and the superhydrophobic ZnO/Ag NWs were respectively chosen as the reference and the experimental substrates, thus AEFs of the characteristic peaks of 613, 774, and 1365 cm −1 could be calculated as ca.

Figure 1 .
Figure 1.a) schematic of the synergistic promotion for in situ SERS performance by superhydrophobic ZnO/Ag Raman platform under aqueous environment; b) photograph of the superhydrophobic ZnO/Ag Raman platform; c) SEM and d) TEM images of the superhydrophobic ZnO/Ag NWs, and the inset is the corresponding HRTEM image marked in d); e) absorbance and f) XRD patterns of ZnO NWs, ZnO/Ag NWs before and after hydrophobic modifications; (g 1 -g 4 ) EDS mappings of the superhydrophobic ZnO/Ag NWs.

Figure 2 .
Figure 2. a 1 -a 3 ) different wettability states of the superhydrophobic ZnO NWs during growth; b 1 -b 3 ) SEM images of ZnO NWs with different growth time, and the insets are optical photographs of 4 μL of water droplets on surface; c) surface energy and CA values of ZnO NWs with different growth time; d 1 -d 3 ) water droplet motion behaviors on surface of ZnO NWs with different growth time; e) Raman spectra of R6G in situ collected from superhydrophobic ZnO/ Ag NWs with different deposition time of Ag and f) ZnO NWs before and after hydrophobic modifications, respectively (e: R6G, 10 −8 M, f: R6G, 10 −2 M); g 1 -g 8 ) simulated electric field distributions in different structures; g 9 ) average electric field intensity and EF in superhydrophobic ZnO/Ag NWs with different deposition time of Ag.

Figure 3 .
Figure 3. a) Raman spectra of R6G with different concentrations (10 −9 M to 10 −14 M) in situ collected by superhydrophobic ZnO/Ag Raman platform; b) linearity between peak intensity and R6G concentration; c) peak intensity mapping of 613 cm −1 in an area of 20 μm × 20 μm; d) in situ Raman enhancement for ZnO/Ag NWs with and without hydrophobic modifications; e) adsorption energies and charge density differences at R/A and R/P interfaces; f) schematic of hotspot modulation by wettability; (g 1 -g 3 ) simulated EFDs in different x-y cross sections of the effective detection areas (marked as 1 and 2 in Figure S8a-c, Supporting Information).

Figure 4 . a 1 )
Figure 4. a 1 ) Raman spectra of MNZ with different concentrations (10 −5 M-10 −9 M) in situ collected by superhydrophobic ZnO/Ag NWs; (a 2 ) linearity between peak intensity and MNZ concentration; a 3 ,a 4 ) MNZ identification in different drugs and corresponding principal component analysis; b 1 , b 2 ) simulated IV process upon microfluidic device; b 3 ,b 4 ) time-dependent Raman spectra and peak intensity changes of MNZ acquired from microfluidic device; c 1 , c 2 ) anti-corrosion test of superhydrophobic ZnO/Ag NWs in MNZ solutions with different pH values and corresponding Raman intensity changes of the MNZ; c 3 , c 4 ) MNZ detection in artificial urine and corresponding Raman spectra; d 1 , d 2 ) anti-interference test of superhydrophobic ZnO/Ag NWs in MNZ solutions with moving particulates and d 3 , d 4 ) corresponding changes of the Raman spectra and microstructures, respectively.