Enhanced toxic trace element detection in water using LIBS combined with a femtosecond laser-engineered hydrophobic–hydrophilic structured substrate

Abstract

Laser-induced breakdown spectroscopy (LIBS) offers significant advantages in the rapid, sensitive, and environmentally friendly detection of toxic elements in water. However, its sensitivity for liquid samples remains a critical limitation, hindering broader practical applications. This work introduces a novel method combining femtosecond laser selective irradiation and chemical modification to construct a hybrid superhydrophobic and hydrophilic surface structure on an aluminum substrate, effectively forming a hydrophobic–hydrophilic enclosure structure. This structure facilitates the stable accumulation and uniform deposition of droplets within the hydrophilic area, significantly suppressing the “coffee ring” effect and enhancing both the concentration efficiency and detection sensitivity. Compared to traditional LIBS, the proposed method achieves limits of detection for Cr, Pb, and As at the ppb level (<3 μg L⁻¹), with determination coefficients (R²) exceeding 0.98. Furthermore, by incorporating the Partial Least Squares Regression (PLSR) model, this method further enhances the accuracy and reliability of the quantitative analysis, maintaining low root mean square errors (RMSEs) in both the training and test datasets. Overall, this innovative method holds considerable potential for water quality monitoring and trace element analysis, offering a novel strategy for high-sensitivity, simultaneous multi-element detection.

---

1. Introduction

Monitoring toxic elements in water is crucial for effective environmental management and protection. Elements such as chromium (Cr), lead (Pb), and arsenic (As) are persistent and bio accumulative, leading to their gradual accumulation in the food chain, thereby posing significant risks to human health. Exposure to these toxic elements has been associated with various health issues, including but not limited to cancer,¹ renal impairment,² neurotoxicity,³ and cardiovascular diseases.⁴ Therefore, precise identification and quantification of these toxic substances in water are essential for safeguarding public health.

Currently, analytical techniques such as inductively coupled plasma optical emission spectrometry (ICP-OES),⁵ inductively coupled plasma mass spectrometry (ICP-MS),⁶ and atomic absorption spectroscopy (AAS)⁷ are widely utilized for detecting toxic elements in water. Although these methods provide high accuracy and a low limit of detection (LOD), they typically require complex and labor-intensive sample preparation procedures, which increases the risk of secondary contamination. In contrast, laser-induced breakdown spectroscopy (LIBS) is characterized by minimal sample preparation requirements, rapid and straightforward operation, and an environmentally friendly profile. These advantages have led to widespread applications in agriculture,⁸ industry,⁹ and biomedicine.¹⁰ Although extensive studies have focused on the LIBS analysis in water, its LOD and stability remain insufficient for meeting environmental monitoring standards. LIBS analysis of liquid samples is susceptible to splashing and surface ripples, which can interfere with data acquisition and compromise the reliability and reproducibility of spectral signals. Additionally, surface waves¹¹ generated during plasma ablation further destabilize LIBS detection, while ejected materials may also contaminate adjacent optical components.

To address these challenges and enhance signal quality, researchers have explored various strategies, including modifications to experimental setups and the use of different substrate-assisted enhancement techniques for applications in diverse monitoring environments. Techniques such as dual-pulse laser ablation,¹² resonance-enhanced LIBS,¹³ and magnetically confined LIBS¹⁴ have demonstrated potential for enhancing plasma emission and signal collection. Additionally, jet flow analysis, single-droplet analysis, aerosol generation,^15,16 and liquid-to-solid conversion^17,18 have been employed to enhance LODs and reproducibility. However, many of these approaches increase system complexity and cost. Notably, liquid-to-solid conversion stands out as a particularly viable method. It requires no modifications of the existing experimental setups while transferring the advantages of LIBS in solid sample analysis to liquid samples, thereby significantly improving sensitivity and stability. Another notable advantage of this approach is its low sample volume requirement, enhancing LODs without additional experimental setup modifications.

Liquid-to-solid conversion typically involves depositing and drying the liquid sample on an appropriate solid substrate, such as graphite,¹⁹ metal plates,²⁰ wood, or paper.²¹ Sample deposition methods include immersing the substrate in a solution or directly drying the solution on the substrate surface. Zhao et al.¹⁹ utilized graphite substrates for automatic enrichment, achieving LODs of 4.8, 36.1, 12.2, 3.1, 34.3, and 32.6 μg L⁻¹ for Cu, Zn, Cr, Cd, Ni, and Pb, respectively, demonstrating LIBS's effectiveness in elemental detection. Similarly, Yu et al.²¹ employed standard printing paper as a liquid absorbent, achieving LODs of 26 μg L⁻¹ for Cr and 33 μg L⁻¹ for Pd, slightly surpassing previous results using wood slices as absorbents. Although liquid-to-solid conversion has enhanced detection stability in these studies, challenges remain, particularly regarding achieving uniform sample distribution and further enhancing detection accuracy.

Recent advances in laser-based surface modification and substrate patterning offer further potential for enhancing LIBS performance in liquid-to-solid conversion.²² Bae et al.²³ employed a laser-patterned substrate to facilitate the enrichment of potassium ions dissolved in small water droplets, resulting in a significant enhancement of the LIBS signal and achieving a LOD of 0.23 × 10⁻⁹ mol. Niu et al.²⁴ developed an innovative enrichment method using a laser-pretreated aluminum (Al) substrate to collect LIBS spectra. Due to the reduced interfacial tension, the deposited solution spread rapidly, achieving uniform distribution and enabling homogeneous sample enrichment. Using a deposition volume of 1 mL, standard calibration curves were established for trace levels of Ni, Cr, and Cd, ranging from sub-mg L⁻¹ to several mg L⁻¹.

In our previous work, femtosecond laser processing enabled the fabrication of micro–nanostructures on Al substrates, effectively mitigating the “coffee ring” effect. This enhancement notably improved sample deposition uniformity and detection reproducibility, and minimized signal fluctuations, achieving LODs of 6.33 μg L⁻¹ for Cr and 2.53 μg L⁻¹ for Pb.²⁵ This approach demonstrates that the incorporation of surface micro/nanostructures has significantly improved the sensitivity and accuracy of LIBS for detecting trace elements in water. The textured surface remained essentially uniform in wettability (i.e., without deliberate wetting contrast or confinement), and quantitative performance was ultimately constrained by the lack of active droplet confinement and preconcentration. Similarly, Ta et al.²⁶ reported laser-induced superhydrophobic modification of stainless steel surfaces, suggesting potential applications in enhancing elemental detection sensitivity for liquid samples. These studies indicate that by designing and tuning micro-structured regions (e.g., wettability), further advancements in solute enrichment and improved detection capabilities for trace elements in water can be achieved.

In this work, a structured surface with a superhydrophobic perimeter and a hydrophilic interior was prepared on a pure Al substrate by integrating femtosecond laser processing with chemical modification. By restricting droplet spreading within the superhydrophobic boundary, liquid samples are confined to a reduced area during drying, facilitating stable enrichment within the hydrophilic region. This approach substantially increases local elemental concentration and significantly suppresses the “coffee ring” effect, enabling high-sensitivity detection of toxic elements (Cr, Pb, and As) in aqueous samples.

2. Materials and methods

The hydrophobic–hydrophilic enclosure structure developed in this work features a superhydrophobic perimeter while retaining the hydrophilicity of the untreated central region, as illustrated in Fig. 1. Initially, femtosecond laser processing was applied to a pure Al substrate (contact angle (CA) = 81°) to fabricate a superhydrophilic enclosure pattern (CA close to 0°), leaving the central region unprocessed to retain the original Al surface structure. Subsequently, the laser-patterned Al surface was chemically treated, converting the enclosure region into a superhydrophobic surface (CA = 155°), while the hydrophilicity of the untreated central region remained unaffected after chemical treatment.²⁷ This design effectively confines droplet spreading within the superhydrophobic barrier while leveraging the hydrophilic central region for stable droplet placement, thereby facilitating efficient analyte concentration during deposition and drying.

Fig. 1 Schematic illustration of the fabrication and experimental process of the hydrophobic–hydrophilic enclosure structure.

2.1 Laser irradiated surfaces

All experiments used pure Al substrates with a purity exceeding 99.99%, measuring 50 mm in length and width and 1 mm in thickness. Before laser processing, substrates underwent no additional treatments apart from removing the protective film. A femtosecond laser (YSL Photonics, FemtoYL-20, China) with a wavelength of 1036 nm, pulse duration of 400 fs, and repetition rate of 300 kHz was utilized. The laser beam was directed through a galvanometric scanner and focused onto the surface using a lens with a focal length of 100 mm, resulting in a spot diameter of approximately 20 μm at the focal plane. The laser output power was maintained at 1.6 W during processing, with a scanning speed of 100 mm s⁻¹ and a scanning interval of 30 μm. The substrate was held stationary, and the scanning process was repeated twice, with the scanning direction perpendicular to the laser polarization direction.

The surface morphology of the original and laser-patterned Al substrates was observed by scanning electron microscopy (SEM) (HITACHI, SU8220, Japan). As shown in Fig. 2a, the original Al substrate exhibited a relatively smooth surface. During laser irradiation, multiple laser pulses induced localized melting and rapid resolidification of the material, resulting in micro/nanoscale grooves and protrusions, as shown in Fig. 2b. The formation of these micro–nanostructures is primarily governed by the laser parameters and the properties of the material.²⁸ Ultimately, periodic subwavelength structures covered with nanoparticles were generated on the Al surface,²⁹ as illustrated in Fig. 2c.

Fig. 2 SEM images of (a) the original Al plate (1000×) and (b) the laser-irradiated Al plate (1000×), and (c) high-magnification SEM image of an individual unit structure (5000×). White arrows indicate the scanning direction, and red arrows denote the laser polarization direction.

2.2 Preparation of superhydrophobic substrates

As shown in Fig. 3, an enclosed micro/nanostructure array was fabricated on the cleaned Al surface via femtosecond laser irradiation. The central unprocessed region measured 2 × 2 mm. Initially, the laser-induced micro/nanostructures demonstrated superhydrophilicity (SHL). Subsequently, the laser-patterned Al substrate was immersed for approximately 30 minutes in a solution containing 1 wt% 1H,1H,2H,2H-perfluorooctyltriethoxysilane (POTS, C₈F₁₃H₄Si(OCH₂CH₃)₃, purchased from Shanghai Titan Scientific Co., Ltd, China) and dissolved analytically pure ethanol (Chengdu Changlian Chemical Reagent Co., Ltd, China). After immersion, the Al substrate was dried on a heating platform at 80 °C (BANGYUAN, BY-3030, China) for about 10 minutes to eliminate residual ethanol. After cooling, a hydrophobic–hydrophilic enclosure substrate exhibiting superhydrophobic (SHB) properties was successfully obtained.

Fig. 3 Schematic illustration of the fabrication process of the hydrophobic–hydrophilic enclosure substrate.

2.3 LIBS experimental setup

The LIBS system utilized in this study is shown in Fig. 4a. A Q-switched Nd:YAG laser (InnoLas, SpitLight 400, Germany) operating at a wavelength of 1064 nm served as the light source for the experiments. The laser pulses were focused onto the sample surface using a plano-convex lens with a focal length of 150 mm, delivering approximately 8 mJ per pulse. Each pulse had a duration of 12 ns and a repetition rate of 1 Hz, creating an ablation spot with a diameter of approximately 300 μm. To enhance experimental repeatability and reliability, the sample was translated using an X–Y motorized stage, ensuring that each pulse targeted a fresh location. LIBS spectral signals were collected at a 45° angle relative to the incident laser. After collimation using an 8.7 mm focal length lens, the signal was coupled into an optical fiber and transmitted to a spectrometer for analysis. The spectrometer (Avantes, AvaSpec-ULS2048, Netherlands) covered a wavelength range of 200–750 nm. During the experiment, the spectrometer integration time was fixed at 2 ms, and a 1 μs spectral acquisition delay was introduced to minimize the continuous background radiation from the plasma at the early stage.

Fig. 4 Schematic illustration of the (a) LIBS system and (b) sample deposition process.

2.4 Sample preparation

For the preparation of standard solutions of Cr, Pb, and As at various concentrations, commercially available standard solutions of Cr (GSB 04-1723-2004), Pb (GSB 04-1742-2004), and As (GSB 04-1714-2004) were obtained from Guobiao (Beijing) Testing & Certification Co., Ltd. These solutions were diluted and homogenized using ultrapure water to prepare 15 different concentrations for each element, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90 and 100 μg L⁻¹.

Following the liquid-to-solid conversion pretreatment, the Al substrate was placed on a heating platform set at 70 °C. A micropipette was employed to sequentially deposit 20 μL of test solutions at different concentrations onto designated areas of the substrate. Subsequently, the substrate was heated until the solution fully evaporated and dried (Fig. 4b).

2.5 LIBS spectral data acquisition and analysis

To obtain spectral data for the 15 different concentration gradients of Cr, Pb, and As, a mapping scan was conducted on the dried substrates. For the hydrophobic–hydrophilic enclosure structure, where the enriched sample region was relatively small, spectral data were collected from 30 dried droplets per sample, arranged in a 3 × 10 grid. In comparison, for the original Al substrate, spectral data were collected from 5 dried droplets per sample, arranged in a 1 × 5 grid. In total, 45 spectra were acquired from the central region of each dried droplet, carefully avoiding the “coffee ring” edges. These spectra were subsequently processed and analyzed. The scanning parameters were optimized based on the LIBS system's laser spot diameter and the deposition area of the standard solutions, ensuring that each laser pulse targeted a fresh location while fully covering the sample deposition area.

LIBS spectra were analyzed using the National Institute of Standards and Technology (NIST) spectral database. The characteristic emission lines selected for qualitative and quantitative analysis include Cr I at 425.50 nm, 427.79 nm, 429.21 nm, 520.84 nm; Pb I at 589.07 nm and As II at 432.57 nm, 443.16 nm, 445.84 nm, 611.01 nm, and 617.03 nm. Calibration curves were constructed by correlating normalized spectral intensities with the respective concentrations of Cr, Pb, and As. The quantitative performance of LIBS enhanced by the hydrophobic–hydrophilic enclosure structure was evaluated using the coefficient of determination (R²) and LODs. Here, R² reflects the linear correlation and accuracy of the results, while LODs represent the detection sensitivity of LIBS with an enclosed microdroplet on a hydrophobic–hydrophilic enclosure structured substrate. Additionally, the predictive capability of the enclosure-structured substrate combined with the quantitative model was further analyzed using the classical partial least squares regression (PLSR).

2.6 Calibration curves and LOD calculation

In most cases, the intensity I of LIBS elemental spectral lines and the concentration C of the element in the sample follow the Schiebe–Lomakin equation, I = aC^b, where a is a constant dependent on the plasma state and elemental concentration, and b is related to element self-absorption. Under conditions of low concentration and negligible self-absorption (b ≈ 1), this relationship simplifies to a linear form y = kx + b, where y represents the spectral intensity, x is the element concentration, k denotes the detection sensitivity, and b is the intercept.

The LOD for Cr, Pb, and As on the hydrophobic–hydrophilic enclosure substrate was calculated according to the definition of the International Union of Pure and Applied Chemistry (IUPAC) using the following equation:

In this equation, σ represents the standard deviation of the background signal and K denotes the slope of the fitted calibration curve.

2.7 PLSR quantitative analysis

PLSR is a chemometric-based multivariate analysis method for both qualitative classification and quantitative analysis of LIBS spectra.³⁰ This method reduces dimensionality by extracting high-information variables as principal components and establishing relationships between these principal components and elemental concentrations. In the PLSR model, the relationship between elemental concentration C and spectral intensity I_i is given by:where C represents the reference elemental concentration, I_i is the spectral intensity at different wavelengths, k_i signifies the regression coefficient, and E denotes the residual error. The root mean square error (RMSE) is used to evaluate the model's fitting performance and is given by:where n is the number of samples in the training set, y_i and ŷ_i represents the reference concentration and predicted concentration of the sample, respectively. A smaller RMSE indicates a better fit of the model.

3. Results and discussion

3.1 Hydrophobic–hydrophilic enclosure substrate characterization

The drying behavior of the droplets on both the original substrate and the hydrophobic–hydrophilic enclosure structure is illustrated in Fig. 5. CA measurements were conducted using a contact angle meter (KINO Scientific Instrument Inc., SL150, USA) on three types of substrates: the original Al plate, the laser-patterned Al plate, and the Al plate after POTS solution treatment. The original Al substrate exhibited a CA of 81°, indicating hydrophilic behavior. Following femtosecond laser irradiation and micro–nanostructuring, the patterned surface exhibited a CA approaching 0°, reflecting superhydrophilicity. Subsequent chemical modification with POTS increased the CA to 155°, demonstrating superhydrophobic characteristics of the treated regions, while untreated regions retained their initial hydrophilic state. To assess reusability and post-modification durability/chemical stability, we tracked the superhydrophobic region's wettability for six months (static contact angles of 155.0°, 152.1°, 152.0°, and 151.7° at 0, 2, 4, and 6 months), which remained >150° (∼98% retention), indicating a persistent wettability contrast, and long-term durability of the enclosure substrate.

Fig. 5 (a) Contact angle measurements of the original Al substrate and the hydrophobic–hydrophilic enclosure substrate. (b) Drying process of 20 μL droplets on the original Al substrate and the hydrophobic–hydrophilic enclosure substrate. Microscope images of the original substrate, captured at a (i) 30°, (ii) 90° and (iii) 90° angle after drying, and the enclosure substrate, captured at a (iv) 30°, (v) 90° and (vi) 90° angle after drying.

As shown in Fig. 5b, the original substrate displayed a pronounced “coffee ring” effect at the droplet edges. This phenomenon is further emphasized in Fig. 5b(iii), where substantial edge deposition is evident, consistent with the known mechanism of the coffee ring effect.³¹ Conversely, droplets on the femtosecond laser-processed hydrophobic–hydrophilic enclosure substrate retained a nearly spherical shape (Fig. 5b(iv)–(vi)). After drying, the residual traces were highly concentrated and confined within a minimal area, with negligible spreading. This localized enrichment of analytes substantially enhances the accuracy and reliability of subsequent LIBS analysis.

3.2 Qualitative LIBS analysis

As illustrated in Fig. 6, the emission line intensities of Cr, Pb, and As (100 μg L⁻¹) samples were compared between the original substrate and the hydrophobic–hydrophilic enclosure substrates. The acquired LIBS spectral data were preprocessed through denoising and feature extraction, and then normalized using the characteristic spectral lines: Cr I 520.84 nm, Pb I 589.07 nm, and As II 445.84 nm from the original substrate. On the hydrophobic–hydrophilic enclosure substrate, the intensity of the Cr I 520.84 nm emission lines was amplified by over 14.30-fold, with an average enhancement factor of approximately 11.33-fold across four characteristic spectral lines. Similarly, the intensities of Pb I 589.07 nm and As II 445.84 nm increased by approximately 10.42-fold and 10.47-fold, respectively. Additionally, an average enhancement of about 11.67-fold was observed across five characteristic As lines.

Fig. 6 LIBS spectral comparison of 100 μg per L Cr, Pb, and As on the hydrophobic–hydrophilic enclosure substrate and the original Al substrate. (a) Cr I 425.50 nm, Cr I 427.79 nm, Cr I 429.21 nm, and Cr I 520.84 nm. (b) Pb I 589.07 nm. (c) As II 432.57 nm, As II 443.16 nm, As II 445.84 nm, As II 611.01 nm, and As II 617.03 nm.

Evidently, the hydrophobic–hydrophilic enclosure structure, engineered through femtosecond laser irradiation and chemical modification, significantly enhanced LIBS performance. The superhydrophobic region, characterized by the Cassie–Baxter state, effectively restricted droplet spreading, preserving a near-spherical droplet shape. Meanwhile, the central hydrophilic region facilitated stable droplet positioning and efficient sample concentration into a compact area during drying. This concentration substantially increased local elemental concentration, thereby markedly improving the characteristic spectral intensities during pulsed laser irradiation.

3.3 Calibration curves and LODs

To verify that the hydrophobic–hydrophilic enclosure structure enhances LIBS spectral quality and facilitates quantitative elemental analysis, standard calibration curves for Cr, Pb, and As were established using both the original Al substrate and the hydrophobic–hydrophilic enclosure substrate, as shown in Fig. 7. Compared to the original substrate, the R² improved significantly from 0.876 to 0.986 for Cr, from 0.939 to 0.989 for Pb, and from 0.857 to 0.988 for As. These improvements indicate a strong linear correlation between LIBS spectral intensity and elemental concentrations on the structured substrate.

Fig. 7 Standard calibration curves for Cr I 520.84 nm, Pb I 589.07 nm, and As II 445.84 nm: calibration curves on (a–c) the original substrate and (d–f) the hydrophobic–hydrophilic enclosure surfaces. The X-axis denotes the sample concentrations of Cr, Pb, and As, while the Y-axis indicates the LIBS spectral intensity.

The calculated LODs for Cr, Pb, and As using the hydrophobic–hydrophilic enclosure structure were 2.83 μg L⁻¹, 2.16 μg L⁻¹, and 2.27 μg L⁻¹, respectively. These values meet the regulatory standards set by the World Health Organization (WHO) Guidelines for Drinking-Water Quality (GDWQ) and the Chinese National Standard (GB 5749-2022), which specify maximum permissible concentrations of 50 μg L⁻¹ for Cr, 10 μg L⁻¹ for Pb, and 10 μg L⁻¹ for As. Moreover, this method exhibits notable advantages over previously reported techniques, as summarized in Table 1.

Table 1 Comparison of LODs of this work and references

Method	LODs	Ref.
Coffee ring effect enrichment	LODs of Cr on Zn plate: 3.881 mg L^−1	33
	Al plate: 5.959 mg L^−1
	Ni plate: Cr 10.677 mg L^−1
	Polymethyl methacrylate (PMMA) plate: 11.174 mg L^−1
SHB substrate (coating polydimethylsiloxane (PDMS) and titanium dioxide nanoparticles (TiO2 NPs) onto the glass slide surface)	Cr 0.045 mg L^−1	32
Chelating resin enrichment	Cr 0.148 mg L^−1	34
	Cu 0.150 mg L^−1
	Pb 0.149 mg L^−1
	Ni 0.240 mg L^−1
With surface structures of the Al substrate	Cr 6.33 μg L^−1	25
	Pb 2.53 μg L^−1
Resonant surface-enhanced LIBS	Pb 2.1 μg L^−1	35
Double-pulse plasma grating-induced breakdown spectroscopy	Cr 6.40 mg L^−1	36
Target-enhanced orthogonal double-pulse laser-induced breakdown spectroscopy	Pb 180 μg L^−1	37
Hydrophobic–hydrophilic substrate (this work)	Cr 2.83 μg L^−1	—
	Pb 2.16 μg L^−1
	As 2.27 μg L^−1

---

A concise comparison with related liquid-to-solid strategies further underscores the advance here. Superhydrophobic dried-droplet enrichment suppresses the coffee-ring effect and stabilizes LIBS detections for heavy-metal analysis (e.g., Cr), but without reaching μg per L LODs.³² By contrast, coffee-ring-assisted enrichment using gelatin and Zn substrates improves SNR yet yields LODs of 0.715 mg per L (Cu) and 3.881 mg per L (Cr)³³—above the ppb-level LODs achieved here for Cr (2.83 μg L⁻¹). Beyond liquid-to-solid strategies, resonance-assisted schemes also deliver strong sensitivity. For Pb in water, resonant surface-enhanced LIBS (R-SENLIBS) achieved R² = 0.996 with a linear range of 0.01–0.20 mg L⁻¹ and an LOD of 2.1 μg L⁻¹.

Against this backdrop, our study primarily demonstrates the substrate's preconcentration capability, and the hydrophobic–hydrophilic enclosure structured substrate delivers ppb-level LODs for Pb, Cr, and As (R² > 0.98) using a single-pulse, non-resonant setup. As summarized in Table 1, conventional enhancements (e.g., dual-pulse excitation or resonance assistance) are not mutually exclusive and can be integrated with this substrate in the future to further lower LODs and broaden the quantitative range.

3.4 Calibration by using LIBS spectra combined with the PLSR model

A smoothing process was applied to the spectral data for denoising, with no further preprocessing steps performed. Both the training and prediction datasets contained spectral information for samples spanning concentrations between 5 and 100 μg L⁻¹. The number of PLSR principal components (PCs) was selected by cross-validation to minimize RMSECV, yielding 3 PCs (Cr), 7 PCs (Pb), and 5 PCs (As). Fig. 8 presents the predictions of the PLSR model, where the x-axis represents the concentrations of Cr, Pb, and As, and the y-axis denotes the model-predicted concentrations. To better visualize the model's performance, especially at lower concentrations (5–10 μg L⁻¹), both axes are plotted on a base-10 logarithmic scale. As shown in Fig. 8, the PLSR models for all three elements exhibit exceptionally high R². Specifically, the R² for the training set is 0.998, 0.997, and 0.997 for Cr, Pb, and As, respectively. In the prediction set, the corresponding R² is slightly lower yet remains robust at 0.996, 0.984, and 0.981, respectively.

Fig. 8 PLSR calibration curves for reference and predicted concentrations in droplets on the hydrophobic–hydrophilic enclosure substrate. (a) Cr, (b) Pb and (c) As.

Training set data points closely align with the fitted curve, while most prediction set data points cluster around it, demonstrating the model's robust predictive performance. Moreover, the RMSE results indicate that despite minor deviations in the prediction set, the overall error levels remain low. Specifically, the RMSE values for the training set are 1.53, 0.89, and 1.82 for Cr, Pb, and As, respectively, while those for the prediction set are 2.19, 4.44, and 4.84. The low RMSE values confirm the model's accuracy and reliability in predicting concentrations, even at trace levels (μg L⁻¹). Evidently, the hydrophobic–hydrophilic enclosure structure significantly enhanced the intensity and stability of the LIBS signal, substantially improving the quantitative analytical performance of the PLSR model.

Integrating LIBS technology with the hydrophobic–hydrophilic enclosure substrate and the PLSR model significantly improves trace elements' detection sensitivity and quantitative accuracy. This method is particularly beneficial for applications demanding high sensitivity and precision, such as environmental monitoring, food safety testing, and trace element analysis in materials science. Thus, combining the hydrophobic–hydrophilic enclosure substrate with the PLSR model exhibits great potential in the quantitative analysis of trace elements, offering a reliable technological foundation for high-precision analytical applications.

4. Conclusion

In this study, a hydrophobic–hydrophilic enclosure structure was fabricated on an Al surface by combining femtosecond laser irradiation with chemical modification. This structure facilitated efficient droplet enrichment and uniform deposition, effectively suppressing the “coffee ring” effect, thereby enhancing LIBS detection sensitivity and signal stability. Compared to traditional LIBS, this method enables efficient enrichment of trace Cr, Pb, and As in water samples without modifying the experimental setup. Compared with our earlier micro/nanostructured (uniformly hydrophilic) substrate, the enclosure structure design achieves substantially lower LODs (≤3 μg L⁻¹) with R² > 0.98 for Cr, Pb, and As. Furthermore, integrating the PLSR model further enhances the accuracy and reliability of quantitative analysis. In conclusion, this method not only lowers system costs and reduces operational complexity but also offers a highly sensitive solution for rapid multi-element detection, particularly beneficial in drinking and environmental water monitoring applications, highlighting its substantial potential for widespread use.

Author contributions

Gangrong Fu conceived the work, curated the data, and prepared the original draft. Rubo Chen contributed to methodology development, data visualization, and manuscript review. Yue Li performed the investigation and formal analysis. Jie Wu contributed to visualization and investigation. Shutong Wang contributed to methodology and data visualization. Guoliang Deng and Hao Zhou acquired funding and managed the project. Hong Zhao contributed to conceptualization and methodology. Shouhuan Zhou provided resources and supervised the research.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

Acknowledgements

This work was supported by the National Key R&D Program of China (2023YFB3610800); Science and Technology on Solid-State Laser Laboratory funding (HG2024158); Chengdu Municipal Bureau of Science and Technology Innovation and R&D Projects (2024-YF05-00268-SN).

References

1. L. Parida and T. N. Patel, Environ. Monit. Assess., 2023, 195, 766,  DOI:10.1007/s10661-023-11399-z.
2. C. Jalili, M. Kazemi, H. Cheng, H. Mohammadi, A. Babaei, E. Taheri and S. Moradi, Crit. Rev. Toxicol., 2021, 51, 165–182,  DOI:10.1080/10408444.2021.1891196.
3. B. Vellingiri, A. Suriyanarayanan, P. Selvaraj, K. S. Abraham, M. Y. Pasha, H. Winster, A. V. Gopalakrishnan, S. G, J. K. Reddy, N. Ayyadurai, N. Kumar, B. Giridharan, S. P, K. R. S. S. Rao, S. K. Nachimuthu, A. Narayanasamy, I. Mahalaxmi and D. Venkatesan, Chemosphere, 2022, 301, 134625,  DOI:10.1016/j.chemosphere.2022.134625.
4. A. Shrivastav, Swetanshu and P. Singh, Curr. Probl. Cardiol., 2024, 49, 102120,  DOI:10.1016/j.cpcardiol.2023.102120.
5. S. R. Khan, B. Sharma, P. A. Chawla and R. Bhatia, Food Anal. Methods, 2022, 15, 666–688,  DOI:10.1007/s12161-021-02148-4.
6. T. Van Acker, S. Theiner, E. Bolea-Fernandez, F. Vanhaecke and G. Koellensperger, Nat. Rev. Methods Primers, 2023, 3, 52,  DOI:10.1038/s43586-023-00235-w.
7. M. F. Hasan, M. Nur-E-Alam, M. A. Salam, H. Rahman, S. C. Paul, A. E. Rak, B. Ambade and A. R. M. Towfiqul Islam, Sustainability, 2021, 13, 6832,  DOI:10.3390/su13126832.
8. P. Yang, G. Fu, J. Wang, Z. Luo and M. Yao, J. Anal. At. Spectrom., 2022, 37, 1948–1960,  10.1039/D2JA00149G.
9. M. Cui, S. Xiong, N. Yang, Y. Wang, Z. Wang, M. Luo, C. Yao and Y. Deguchi, Appl. Spectrosc. Rev., 2024, 1–49,  DOI:10.1080/05704928.2024.2398196.
10. R. Zhang, S. Hu, C. Ma, T. Zhang and H. Li, TrAC, Trends Anal. Chem., 2024, 181, 117992,  DOI:10.1016/j.trac.2024.117992.
11. V. Lazic and S. Jovićević, Spectrochim. Acta, Part B, 2014, 101, 288–311,  DOI:10.1016/j.sab.2014.09.006.
12. B. Xue, Y. Tian, N. Li, Q. Li, Y. Lu and R. Zheng, J. Anal. At. Spectrom., 2020, 35, 2880–2892,  10.1039/D0JA00139B.
13. Y. R. Wang, J. Kang, Y. Q. Chen and R. H. Li, J. Appl. Spectrosc., 2019, 86, 353–359,  DOI:10.1007/s10812-019-00825-1.
14. A. Hussain, G. Xun, H. Asghar, M. Azam, Q.-t. Ain and Z. Nawaz, Opt. Spectrosc., 2021, 129, 452–459,  DOI:10.1134/S0030400X21040068.
15. K. Keerthi, S. D. George, S. D. Kulkarni, S. Chidangil and V. K. Unnikrishnan, Opt Laser. Technol., 2022, 147, 107622,  DOI:10.1016/j.optlastec.2021.107622.
16. D. C. Zhang, Z. Q. Hu, Y. B. Su, B. Hai, X. L. Zhu, J. F. Zhu and X. Ma, Opt. Express, 2018, 26, 18794–18802,  DOI:10.1364/OE.26.018794.
17. H. T. Temiz, B. Sezer, A. Berkkan, U. Tamer and I. H. Boyaci, J. Food Compos. Anal., 2018, 67, 48–54,  DOI:10.1016/j.jfca.2017.12.032.
18. Z. A. Abdel-Salam, S. A. M. Abdel-Salam, I. I. Abdel-Mageed and M. A. Harith, J. Adv. Res., 2019, 15, 19–25,  DOI:10.1016/j.jare.2018.07.001.
19. N. J. Zhao, D. S. Meng, Y. Jia, M. J. Ma, L. Fang, J. G. Liu and W. Q. Liu, Opt. Express, 2019, 27, A495–A506,  DOI:10.1364/OE.27.00A495.
20. M. Ma, L. Fang, N. Zhao, X. Huang, D. Meng, C. Pan, J. Liu and W. Liu, J. Laser Appl., 2022, 34, 012021,  DOI:10.2351/7.0000572.
21. Y. Yu, W. Zhou, H. Qian, X. Su and K. Ren, Plasma Sci. Technol., 2014, 16, 683,  DOI:10.1088/1009-0630/16/7/09.
22. Y. Zhang, Y. Jiao, C. Li, C. Chen, J. Li, Y. Hu, D. Wu and J. Chu, Int. J. Extreme Manuf., 2020, 2, 032002,  DOI:10.1088/2631-7990/ab95f6.
23. D. Bae, S.-H. Nam, S.-H. Han, J. Yoo and Y. Lee, Spectrochim. Acta, Part B, 2015, 113, 70–78,  DOI:10.1016/j.sab.2015.09.005.
24. S. Niu, L. Zheng, A. Q. Khan, G. Feng and H. Zeng, Talanta, 2018, 179, 312–317,  DOI:10.1016/j.talanta.2017.11.043.
25. R. Chen, G. Fu, H. Li, Y. Li, J. Wu, G. Deng, H. Zhou, X. Sui and S. Zhou, Opt. Express, 2024, 32, 32849–32860,  DOI:10.1364/OE.529641.
26. V. D. Ta, A. Dunn, T. J. Wasley, J. Li, R. W. Kay, J. Stringer, P. J. Smith, E. Esenturk, C. Connaughton and J. D. Shephard, Appl. Surf. Sci., 2016, 365, 153–159,  DOI:10.1016/j.apsusc.2016.01.019.
27. M. Zhang, Z. Liu, N. Pan, H. Tao and J. Lin, Chin. J. Lasers, 2021, 48, 7–16,  DOI:10.3788/cjl202148.1802001.
28. X. Xu, L. Cheng, X. Zhao, J. Wang and X. Chen, Micromachines, 2022, 13, 976,  DOI:10.3390/mi13060976.
29. A. Y. Vorobyev, V. S. Makin and C. Guo, J. Appl. Phys., 2007, 101, 034903,  DOI:10.1063/1.2432288.
30. G. Fu, W. Hu, W. Xie, X. Yao, J. Xu, P. Yang and M. Yao, Anal. Methods, 2023, 15, 5867–5874,  10.1039/D3AY01340E.
31. A. D. Nikolov, D. T. Wasan and P. Wu, Curr. Opin. Colloid Interface Sci., 2021, 51, 101387,  DOI:10.1016/j.cocis.2020.08.012.
32. Y. Liu, J. Pan, G. Zhang, Z. Li, Z. Hu, Y. Chu, L. Guo and C. Lau, Anal. Chim. Acta, 2021, 1151, 338253,  DOI:10.1016/j.aca.2021.338253.
33. Y. Wang, H. Li, X. Yin, C. Zhang, X. Li and L. Fang, Measurement, 2024, 224, 113869,  DOI:10.1016/j.measurement.2023.113869.
34. J. Wang, G. Li, P. Zheng, S. Shata, H. I. Qazi, J. Lu, S. Liu, H. Tian and D. Dong, Chemosensors, 2023, 11, 228,  DOI:10.3390/chemosensors11040228.
35. K. Liu, Z. Tang, R. Zhou, W. Zhang, Q. Li, C. Zhu, C. He, K. Liu and X. Li, J. Anal. At. Spectrom., 2021, 36, 2480–2484,  10.1039/D1JA00250C.
36. F. Li, M. Hu, Y. Qiao, S. Xu, E. Wan and H. Zeng, Spectrochim. Acta, Part B, 2024, 220, 107019,  DOI:10.1016/j.sab.2024.107019.
37. Y. Jiang, Z. Lin, R. Li and Y. Chen, Spectrochim. Acta, Part B, 2021, 181, 106221,  DOI:10.1016/j.sab.2021.106221.

---