2025, Vol. 36 
Organic luminescent systems have been benefiting tremendously from the donor-acceptor (D-A)/donor-π-acceptor (D-π-A) molecular construction, as it facilitates the transition from locally-excited (LE) state to intramolecular charge transfer (ICT) state, enabling bathochromic-shift with large Stokes’ shift to overcome the signal overlap between excitation and emission [1,2]. ICT state is conducive to fluorescence-based visualization and bio-imaging applications, because the resulting long-wavelength excitation/emission helps to deeply penetrate the living tissues without causing excess photodamage [3,4]. In addition, the molecular design with D-A/D-π-A construction offers the feasibility to fine-tune the photophysical properties of the synthesized compound by tailoring the π-bridge length or D-A substituents, which contributes to the development of numerous chemical/biological sensing applications [5–7]. D-A/D-π-A luminescent systems generally have a charge-separation state susceptible to the surrounding chemical species, which endows them with high sensitivity to subtle changes in both chemical environment and aggregation structure [8]. Scientists have utilized this feature to fabricate a variety of tailor-made fluorescent probes with specific functionalities, e.g., imaging organelles by pH value [9], sensing volatile organic compounds (VOC) [10], and mapping specific disease marker [11]. The attachment of D-A substituents would polarize the resulting compounds to produce dipole-dipole interaction, which is a major driving force for molecular self-assembly [12,13]. With the assistance of intermolecular π-π and/or dipole-dipole interactions, these compounds are ideal building blocks for specific self-assembly structures down to micro/nano-scale [14–16]. The strength and directions of their dipole moments can strongly influence the morphology and architecture of the self-assembled aggregates, wherein the functions can be well-controlled for a given situation [17,18].
D-A/D-π-A systems are also promising agents for fingerprint imaging [19]. There are two major aspects in this domain, i.e., latent fingerprint visualization and fingerprint mapping. Latent fingerprint is formed by finger-contacting on various object surface and is generally invisible to naked eyes, which is vital for solving criminal cases [20]. The latter aims to record the biometric information of a human individual for the purpose of identity verification. Thus, it is widely used in many civil activities, such as access control [21], biometric lock [22] and information security applications [23]. Accuracy and precision are emphasized for both aspects. In order to make a standardization of precision, it is customary to categorize the fingerprint information into three levels [24]. Level-1 covers the macroscopic features of the fingerprint, which do not satisfy identification requirements due to lack of precision [25]. Level-2 information contain various feature details (e.g., loops, arches, whorls, ridge endings and hooks), which are helpful for similarity comparison [26]. Level-3 refers to the microscopic properties of fingerprints, including the shape/size of ridges and furrows, warts, creases, and most importantly, the sweat pores [27]. It covers a large number of detailed features that are extremely difficult to forge, thus becoming the most recognizable “signature” in personal identity verification (PIV) [28]. However, most of the reports only allow the visualization of level-1/2 information, whereas level-3 accuracy is difficult to achieve. From a security standpoint, level-1/2 fingerprint information has the drawback of easy counterfeiting, rendering the identification system vulnerable to forgery attacks [29,30]. Fortunately, this problem can be partially solved by visualizing level-3 fingerprint information, i.e., sweat pore distribution, based on tailer-made fluorescent probes [31]. Kim et al., for instance, developed a class of functionalized polydiacetylenes films that trigger a fluorescent turn-on effect in the presence of water, accompanied by a large bathochromic-shift from blue to red, and the active sweat pores within the fingerprint area can be mapped to obtain level-3 information [32]. It is worth mentioning that the aggregation-induced emission (AIE) probes are advantageous in this respect owing to their intrinsic high on-off ratio and low background noise [33,34]. For example, Zhu et al. demonstrated a series of fluorescent probes capable of in-situ visualization of latent fingerprints with high contrast and resolution. The individual identity of sweat pore distribution can be readily observed [35–37]. Similarly, Li et al. synthesized an amphiphilic luminescent compound, which exhibited excellent capability in probing level-3 details including sweat pores by targeting fatty-acid residues of fingerprints [38]. Generally, the aforementioned studies succeed in upgrading imaging accuracy, wherein level-3 information can be directly observed by eyes or microscopy. Nevertheless, this area of research is still facing challenges. On one hand, the working processes of the imaging probe, e.g., sample preparation, fingerprint imaging and subsequent image analysis, would take a certain period of time, which shows shortcomings in applications requiring real-time feedback. On the other hand, the related studies are staying at the mapping/imaging stage, wherein the comparison of individual difference relies on naked-eye observation. It is therefore not suitable for PIV in the case of large sample tests and big data sets. In addition, it is also a challenging work to achieve 100% accuracy in fingerprint matching as there are fuzzy bits in biometrics [39].
Considering the above limitations, we herein design and synthesize a D-A type fluorescent probe with hydrophilic groups. It can self-assemble into interlaced textures down to nanoscale and can be disassembled by trace amount of water in fingertip area, which reveals level-1/2/3 fingerprint information, including sweat pore details. The sweat pore imaging shows high accuracy and long-term stability. The corresponding sweat pore distribution can be collected and digitized by polar coordinate, allowing machine-learning-based analysis for PIV application. We further plow forward on the construction of a prototypical PIV system based on the fluorescent probe. It combines fingerprint imaging, sweat pore analysis, data processing, machine-learning-based recognition into one platform, enabling 100% recognition accuracy under finite-sample condition.
The target compound, namely (E)−4‑chloro-3-((2-hydroxybenzylidene)amino)benzoic acid (CHBA), was synthesized by a one-step reaction under mild condition (Fig. 1A). The chemical structure was confirmed by 1H nuclear magnetic resonance (1H NMR) spectrum (Figs. S1 and S2 in Supporting information), 13C NMR spectrum (Figs. S3 and S4 in Supporting information) and high-resolution mass spectrometry (HRMS, Fig. S5 in Supporting information). The procedure for the synthesis together with the structural characterization data are reported in detailed in the Experimental section (Supporting information). From the molecular design, the imine group serves as the electron donor, whereas the benzoic acid and chlorine are electron acceptors. Quantum chemical calculations were performed to verify the push-pull effect intrinsic to D-A units. As shown in Fig. 1B, CHBA molecule adopts twisted conformation according to the energy minimum optimization, which is conductive to the charge transfer process. The highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) are not completely separated, but exhibit shifting from imine group to phenol group. The partial separation implies a hybrid transition containing both LE state and ICT state. The charge difference density (CDD) together with the charge centroids (C+/C−) of CHBA was calculated and plotted in Fig. 1C. If comparing the CDD of S1 to that of S0, a charge transfer distance (DCT) up to 2.39 Å was generated (Table S1 in Supporting information), providing direct evidence for the ICT process. The attachment of D-A pairs would also polarize the molecule to generate dipole-dipole interactions. According to quantum chemical calculations, there is a notable dipolar moment of 6.33 Debye (Table S1 in Supporting information), which serves as a major driving force for molecular self-assembly. The hybrid transition of CHBA is further confirmed by the absorption spectrum. As shown in Fig. 1D and Table S1, there is a distinct dual absorption band in tetrahydrofuran (THF), locating at 320 and 435 nm, respectively. The former corresponds to the LE state while the latter indicates ICT state. Upon addition of water, the ICT band drops rapidly until it disappears, leaving the LE band as the main peak. The time-dependent emission spectrum in THF-water mixtures shows similar variation trend (Fig. 1E). The transition from hybrid state to the resulting LE state driven by water can be ascribed to the hydrophilic groups, i.e., carboxylic acid and phenolic hydroxyl. Their water binding activity would weaken the electron-donating/accepting ability, further causes a dampening effect to ICT [40]. When fabricating CHBA into a drop-casting film, it exhibits a fluorescence lifetime of 0.47 ns (Fig. S6 in Supporting information). Immersing CHBA film into water leads to an erosion effect. Concretely speaking, the CHBA film would be invaded by water molecules and partly dissolved, leading to a destruction of the initial aggregation structure. The erosion effect was confirmed by time-dependent absorption spectroscopy (Fig. 1F). Before measurements, CHBA was deposited on a quartz plate and fixed at the bottom of cuvette. With addition of water, there is a negligible absorption peak, implying slight solvation. Then the spectrum shows gradual increase with time and displays an absorption band resembling the final-state spectrum in Fig. 1D, which indicates a transfer of CHBA molecules from solid phase to aqueous phase. In addition, the transmittance of CHBA film undergoes a step-by-step increase in water, which is also an indication of water-induced erosion effect (Fig. 1G). After erosion, the film exhibits a significant decrease in emission intensity, easily discernible to the naked eyes (Fig. 1H). Besides, this erosion only occurs at the solid-liquid interface. Fumigation in water vapor exerts no effect to the film even in high relative humidity (Fig. 1I and Fig. S7 in Supporting information). In other words, CHBA film is highly resistant to water vapor, which is advantageous in practical use.
|
Download:
|
| Fig. 1. (A) Synthetic route of CHBA and the indication of hydrophilic groups. (B) Top- and side-view of CHBA's conformation optimized and HOMO/LUMO distribution calculated at B3LYP/6–31+G* level in tetrahydrofuran solvent with isosurface of 0.02 au. (C) CDD of S0/S1 with isosurface value of 0.001 au, DCT and the corresponding charge centroids (C+(r)/C−(r)) with isocontour value of 0.001 au calculated at the PCM(THF)/TD-B3LYP/6–31+G* level for CHBA. The green and blue regions correspond to density increment and decrement, respectively. (D) Time-dependent ultraviolet–visible (UV–vis) spectrum of CHBA in THF-water binary solvent system (50 µmol/L, 70% water fraction). (E) Time-dependent emission spectrum of CHBA in THF-water binary solvent system (0.1 mmol/L, 70% water fraction), measured every minute for 10 min. (F) Changes of the UV–vis spectra of the aqueous phase over time after the immersion of the CHBA film in water, measured every minute for 10 min. (G) Time-dependent transmission spectrum of CHBA film with continuous water erosion, measured every minute for 4 min. (H) Emission spectrum of CHBA film before and after water erosion. (I) The maximum emission intensity of CHBA film with increased humidity. Excitation wavelength for all the emission measurements: 375 nm. | |
CHBA film and finger would form a similar solid-liquid interface upon physical contact, as there is trace amount of water on the surface of the finger skin. With gentle finger pressing, a fingerprint image was imprinted. It exhibits clear ridges and furrows, revealing level-1/2 fingerprint information in every detail, which is visible under both ambient light and UV light (Fig. 2A). Under fluorescence microscope, CHBA is well-distributed on the film, exhibiting yellow-orange emission, which is consistent with its emission spectrum. Following finger pressing, light and dark textures appear on the film, reflecting the furrows (uncontacted) and ridges (contacted), respectively. It is worth mentioning that there are groups of fluorescent dots distributed on the ridges, which represent the sweat pores, namely the most important level-3 information (Fig. 2B). Although the patterns can be observed clearly in both fluorescence and bright-field modes, we tend to use the former in naked-eye observation owing to its large signal contrast and high resistance to variegated noise. The morphology of CHBA film was clarified by scanning electron microscope (SEM). At low magnification, the as-prepared film exhibits granular morphology with high homogeneity resembling the fluorescence microscope images (Fig. 2C). At high magnification, the “granular aggregates” are actually self-assembled interlaced textures with porous structure down to nanoscale. It affords large specific surface area, rendering the film sensitive to liquid water. The driving forces of self-assembly stem from the dipole-dipole interactions intrinsic to the D-A pairs as well as the π-π interactions resulted from the conjugate core. Upon finger pressing, physical contact occurs between ridge and film, and the interlaced textures would be disassembled and/or dissolved by the wet skin, resulting in disordered molecular arrangement. Thus, the rotational freedom of CHBA molecule would be increased to facilitate the ICT state featuring nonradiative decay, and the emission is quenched as the consequence [8]. In contrast, there is no physical contact between invaginated furrows and sweat pores, thus the self-assembled CHBA film keeps the original state in these regions, where the interlaced textures could maintain the morphology to a certain extent. As indicated in Fig. 2D, the furrow and ridge are separated by a clear outline, enabling accurate measurements of their size and shape at the nanoscale. From Fig. 2E, the size and shape of sweat pores can be revealed in the same way. The borders are clear and well-defined, allowing precise localization of the sweat pores. It is essential for determining the sweat pore distribution for PIV application. Furthermore, we could observe every single sweat pore and conduct a comparative analysis of their shape feature. It is actually beyond level-3 information and can be promising as alternative biometric information when level-1/2/3 information is unobtainable, or in the case of fuzzy bits and skin disease. In short, a clear fingerprint image can be obtained via finger-touching, which covers comprehensive details of level-1/2/3 fingerprint information in high resolution. Cell counting kit-8 (CCK-8) colorimetric assay was performed to assess the toxicity of CHBA, using its aqueous solutions with increasing concentration. The resulting histogram in Fig. S8 (Supporting information) indicates that the HeLa cells maintain high viability (>96%) in CHBA solutions with lower concentration ranging from 0.1 µmol/L to 5 µmol/L. With 7.5–10 µmol/L, the cell viability slightly decreases by ca. 12%, suggesting a relatively low toxicity.
|
Download:
|
| Fig. 2. (A) Schematic illustration of the preparation of CHBA film and fingerprint imaging under finger pressing. (B) Bright-field and dark-field fluorescence microscope images of CHBA film before and after finger pressing (365 nm excitation). (C) SEM images showing the self-assembly structure of the as-prepared CHBA film. (D) SEM images of the CHBA film after finger pressing. The ridge and furrow regions are separated by the dotted lines. (E) Representative SEM images of sweat pores. | |
We examined a fingerprint segment by optical microscope and fluorescence microscope, respectively. The optical microscope image reveals eighteen sweat pores as indicated in Fig. 3A, which all appear in the same locations in the fluorescence microscope image in Fig. 3B. The sweat pores in two photographs match well in terms of quantity and position, indicating a high degree of precision in one-to-one correspondence. We further performed a stability test according to a well-established method [41]. As indicated in the fluorescence microscope image in Figs. 3C and D, we selected four sweat pores to form a quadrilateral, and measured its side lengths (Ls) and included angles (θ) using image processing software (Figs. S9–S13 in Supporting information). After 7 days, the four sweat pores remain unchanged from naked-eye observation comparing with the as-prepared fingerprint image (Figs. 3E and F). The variations of Ls and θ were measured once a day for a week (Table S2 in Supporting information), which are summarized in Figs. 3G and H, respectively. From the histograms, Ls and θ values exhibits minor variations but are basically stable. The variances can be attributed to the fluctuations in the sweat-pore activity at different times, or different levels of finger pressure. The areas of the quadrilateral were measured as well, which show larger variations than the former two (Fig. 3I). Variation coefficient (Cv) was introduced to characterize the stability of sweat pore imaging. According to literature report, Cv within <10% implies high stability and can be used for fingerprint comparison analysis. After calculation through the formula in Fig. 3J, Cv,Ls ranges from 1.8% to 2.5%, while Cv,θ falls into a scale of 2.8%–5.2%, suggesting high stability. However, Cv of the area (Cv,A) is much higher, as it varies greatly due to the activity of sweat pores (e.g., open/close, keratosis and sweating). Although it is currently within an acceptable limit, it should be avoided as a basis for comparative analysis due to perturbations in different states. The above experiments demonstrate the reliability of sweat pore imaging, which paves the way for PIV application.
|
Download:
|
| Fig. 3. (A) Micrograph of a real fingerprint segmentation and (B) fluorescence microscope image of CHBA film in the same region after finger pressing (365 nm excitation). The sweat pores within this area are circled one by one and suggest 100% matching. The fluorescence microscope images of the fingerprint in the as-prepared state (C, D) and that of the same region (E, F) after 7 days (365 nm excitation). Four sweat pores were selected on the magnified fluorescent images for Cv-based reliability test. Scale bar: 200 µm. Weakly change of (G) Ls, (H) θ and (I) areas shown in histograms. (J) Histograms showing the Cv values of Ls, θ and areas. | |
The reliability of the similarity comparison using CHBA-based fingerprint imaging is verified in Fig. S14 (Supporting information). In practical applications, however, there would be an extremely large number of samples involved, and it is impossible to rely on the naked-eye observation or simple algorithms for similarity analysis. We therefore try to use machine-learning methods for similarity analysis and identity recognition, by virtue of sweat pore information. As described in Fig. 4A, fingerprint images of volunteers were obtained using CHBA film, wherein sweat pores can be located to obtain a distribution map. After a polar coordinate transformation, they are readily transformed into distances (l) and angles (θ), which can be easily recognized by machine-learning methods (Figs. S15–S26 in Supporting information). In the PIV stage, a variety of supervised machine-learning models are employed, including decision tree, back propagation, support vector machines, random forest, naive Bayes model and K-nearest neighbor. In the above methods, the training set (Fig. S27 in Supporting information) and the test set account for 50% each. Figs. 4B–G indicate the resulting confusion matrices of the above models, in which the ground truth and predicted values are denoted by the row and the column, respectively. Accurate predictions would fall into the diagonal cells, resulting in color deepening. From the results, the diagonal cells all turn into dark state, suggesting that the final accuracy rate reaches 100% for all the involved machine-learning models, with no erroneous judgements. In the above process, the machine-learning algorithms verified the identification with 15 groups of polar coordinates input (Tables S3–S8 in Supporting information). Theoretically, all the sweat pores in the fingerprint region can be digitized by polar coordinates to obtain complete sweat pore information of an individual. Even with the fuzzy bits and high approximations of sweat pore features that occasionally appear in fingerprint imaging, the machine-learning methods can be trained heavily to deal with such datasets that are ill-defined and difficult to segregate. Thus, we believe that it is highly possible to carry out the PIV work in real cases containing large-scale samples. An unsupervised model, namely hierarchical clustering, is employed to classify and recognize sweat pore information, which achieves satisfactory classification results (Figs. S28–S41 in Supporting information).
|
Download:
|
| Fig. 4. (A) Schematic illustration showing the digitization of sweat pore distribution via polar coordinate conversion and data training/prediction under supervised learning models. Confusion matrices trained with various supervised models including (B) decision-tree, (C) back propagation, (D) support vector machines, (E) random forest, (F) naive Bayes model and (G) K-nearest neighbor, with 50% training set and 50% test set. | |
The above sweat pore images cover a small area and do not provide complete fingerprint information. To obtain a full-size fingerprint image, the fingerprint area can be scanned sequentially by CHBA film. Thus, a series of fragmentation maps would be obtained, whereby all the sweat pore information can be extracted after integration. However, it leads to a huge workload and data volume, which is not favorable for practical applications. To solve this problem, convolutional neural network (CNN) is employed to process complete fingerprint images without segmentation and integration. It is designed to deal with data with similar grid structure, such as pixel image [42–44]. Five image groups containing complete fingerprint patterns were collected from volunteers (with 10 repeated trials), which form a data set containing 50 images (Fig. S42 in Supporting information). Fig. S43A (Supporting information) illustrates the working process of CNN. Firstly, these images are separated into RGB channels and transformed into characteristic patterns by using a convolution kernel with fixed pixels. Then they are extracted by a “pooling” (down-sampling) process, where the secondary features of the fingerprint images are deleted, retaining the primary features that are highly recognizable. CNN would repeat the convolution-pooling process for pre-set cycles using gradually diminishing convolution kernel, which yields a feature pattern for each fingerprint image. By comparing the characteristic differences of the feature patterns, the image set is classified into five categories to match the volunteers in the final stage. The confusion matrix in Fig. S43B (Supporting information) indicates the identification results, wherein the correct predictions fall into the diagonal cells. Clearly, there is only one incorrect judgment out of 10 predictions, suggesting a high accuracy up to 90%. In addition, CHBA was integrated with artificial intelligence hardware, which facilitates the construction of a PIV system for real-world application (Fig. S44 in Supporting information). Based on commercial machine-learning platforms, the system can execute all the above-mentioned machine-learning algorithms for data/image analysis and identification-classification.
In summary, a D-A fluorescent probe CHBA is designed and synthesized. The attachment of phenolic hydroxyl and carboxylic acid endows CHBA with hydrophilicity. It self-assembles into interlaced textures with porous structure down to nanoscale, providing large specific surface area to make the film sensitive to liquid water. Upon finger pressing, there is physical contact between ridge and film, and the interlaced textures would be disassembled and/or dissolved by the wet skin, resulting in disordered molecular arrangement to the disadvantage of fluorescence emission. Owing to the selective water-erosion effect on the solid-liquid interphase, fingerprint image can be revealed in both bright field model and fluorescence mode. It covers level-1/2/3 fingerprint information, wherein the sweat pores can be mapped in detail as well, enabling differentiation of different fingerprint patterns based on naked-eye comparison or analysis software. More importantly, the sweat pore distribution can be digitized via polar coordinate conversion and further analyzed by multiple machine-learning models, which reaches 100% accuracy in PIV application. A prototype of PIV system is constructed by integrating CHBA with artificial intelligence hardware. It includes the acquisition module, the data processing module and the identification module into one hardware platform, allowing real-time imaging and analysis of sweat pores. It is believed that CHBA could be further developed into a machine vision-haptic material, which has a broad application prospect in the era of big data and artificial intelligence.
Ethical statementThis work contains fingerprint samples from six volunteers, which were collected according to the Declaration of Helsinki and approved by ethical standards of the Ethics Committee of Capital Normal University. All the volunteers have signed the informed consent form.
Declaration of competing interestThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
CRediT authorship contribution statementXinyi Zhao: Writing – original draft, Investigation. Yuai Duan: Software, Methodology, Investigation. Zihan Liu: Methodology. Hua Geng: Software. Yaping Li: Methodology. Zhongfeng Li: Resources, Data curation. Tianyu Han: Writing – review & editing, Supervision, Conceptualization.
AcknowledgmentsWe acknowledge the financial support from National Natural Science Foundation of China (No. 51703135) and the technical support from Beijing Key Laboratory of Optical Materials and Photonic Devices. Authors are thankful to all the volunteers for their participation in the experiments. Authors are also thankful to the staff in the Analysis and Testing Center of Capital Normal University.
Supplementary materialsSupplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2024.110617.
| [1] |
Y.C. Deng, Y. Shi, L. Li, et al., Appl. Catal. B: Environ. 352 (2024) 124043. |
| [2] |
A. Pal, M. Karmakar, S.R. Bhatta, A. Thakur, Coord. Chem. Rev. 448 (2021) 214167. |
| [3] |
Y.Q. Wang, Y. Deng, Y.X. Wang, et al., J. Mol. Struct. 1308 (2024) 138107. |
| [4] |
Y.L. Zou, Y.T. Liu, Spectrochim. Acta A: Mol. Biomol. Spectrosc. 312 (2024) 124040. |
| [5] |
J.Q. Han, Y.P. Li, J. Yuan, et al., Sens. Actuators B: Chem. 258 (2018) 373-380. |
| [6] |
B.K. Dwivedi, P. Bhowmik, D.S. Pandey, Dyes Pigm. 204 (2022) 110444. |
| [7] |
L.L. Bu, M. Remond, P. Colinet, et al., Dyes Pigm. 202 (2022) 110258. |
| [8] |
C. Wang, W.J. Chi, Q.L. Qiao, et al., Chem. Soc. Rev. 50 (2021) 12656-12678. DOI:10.1039/d1cs00239b |
| [9] |
K.P. Jethava, P. Prakash, P. Manchanda, H. Arora, G. Chopra, ChemBioChem 23 (2022) e202100378. |
| [10] |
Y. Zhang, X.Y. Zhao, L.F. Xu, et al., Sens. Actuator. B: Chem. 399 (2024) 134834. |
| [11] |
H.J. Yang, C. Zhou, G.Y. Xu, et al., Results Chem. 7 (2024) 101325. |
| [12] |
Y.J. Li, T.F. Liu, H.B. Liu, M.Z. Tian, Y.L. Li, Acc. Chem. Res. 47 (2014) 1186-1198. DOI:10.1021/ar400264e |
| [13] |
L. Wang, T. Wei, T. Meng, Q.D. Zeng, H.J. Xu, J. Mol. Struct. 1320 (2025) 139730. |
| [14] |
I. Matsumoto, R. Sekiya, T. Haino, Angew. Chem. Int. Ed. 133 (2021) 12816-12821. DOI:10.1002/ange.202101992 |
| [15] |
Q.Q. Yang, J.Y. Wang, L.P. Huang, W. Song, L. Ding, Polymer 310 (2024) 127475. |
| [16] |
G.S. Nair, R.K. Lekshmy, K. Szaciłowski, M. Takeuchi, B. Balan, J. Lumin. 276 (2024) 120857. |
| [17] |
X.T. Ji, N. Wang, J.K. Wang, et al., Chem. Sci. 15 (2024) 3800-3830. DOI:10.1039/d3sc06527h |
| [18] |
L.Z. Fang, Y. Dai, Y.Q. Bai, et al., Chem. Commun. 60 (2024) 7646-7649. DOI:10.1039/d4cc01987c |
| [19] |
P.J. Zhang, Q.F. Shen, Y. Zhou, et al., Chin. Chem. Lett. 34 (2023) 107910. |
| [20] |
D. Peng, Z.H. Zhao, Spectrochim. Acta A: Mol. Biomol. Spectrosc. 299 (2023) 122827. |
| [21] |
N.A. Alam, M. Ahsan, M.A. Based, J. Haider, Technologies 9 (2021) 9-26. DOI:10.3390/technologies9010009 |
| [22] |
A.J.M.A. Cader, J. Banks, V. Chandran, Sensors 23 (2023) 6591-6617. |
| [23] |
D. Sero, I. Garachon, E. Hermens, R.V. Liere, K.J. Batenburg, J. Comput. Cult. Herit. 4 (2021) 51-71. |
| [24] |
C. Champod, Phil. Trans. R. Soc. B 370 (2015) 20140259. DOI:10.1098/rstb.2014.0259 |
| [25] |
D.L. Gao, F. Gao, Q.Q. Kuang, et al., ACS Appl. Nano Mater. 5 (2022) 9929-9939. DOI:10.1021/acsanm.2c02163 |
| [26] |
Y.Q. Li, C.Y. Xu, C. Shu, X.D. Hou, P. Wu, Chin. Chem. Lett. 28 (2017) 1961-1964. |
| [27] |
J. Kaur, M. Dhall, Heliyon 9 (2023) e17927. |
| [28] |
J.T. Hu, M.M. Wu, X.Y. Zhao, et al., Dyes Pigm. 198 (2022) 110034. |
| [29] |
P. Kieseberg, S. Schrittwieser, Fingerprinting using database steganography, in Proc. IEEE International Conference on Software Security and Assurance, St. Pölten, Austria, 2019, pp.16–20.
|
| [30] |
V. Soum, S. Park, A.I. Brilian, et al., ACS Omega 4 (2019) 8626-8631. DOI:10.1021/acsomega.9b00936 |
| [31] |
C. Suresh, G.P. Darshan, S.C. Sharma, et al., Opt. Mater. 100 (2020) 109625. |
| [32] |
D.H. Park, B.J. Park, J.M. Kim, Acc. Chem. Res. 49 (2016) 1211-1222. DOI:10.1021/acs.accounts.6b00128 |
| [33] |
J. Mei, N.L.C. Leung, R.T.K. Kwok, J.W.Y. Lam, B.Z. Tang, Chem. Rev. 115 (2015) 11718-11940. DOI:10.1021/acs.chemrev.5b00263 |
| [34] |
M.K. Ravindraa, K.M. Mahadevana, R.B. Basavarajb, et al., Mater. Sci. Eng. C: Mater. Biol. Appl. 101 (2019) 564-574. |
| [35] |
Y.L. Wang, C. Li, H.Q. Qu, et al., J. Am. Chem. Soc. 142 (2020) 7497-7505. DOI:10.1021/jacs.0c00124 |
| [36] |
R. Tian, Q. Yu, L.J. Mei, et al., Sens. Actuator. B: Chem. 396 (2023) 134634. |
| [37] |
R. Tian, F.Y. Zhu, R.L. Ma, et al., Biosens. Bioelectron. 263 (2024) 116572. |
| [38] |
Y. Chen, A. Li, X.N. Li, et al., Adv. Mater. 35 (2023) 2211917. |
| [39] |
H.Y. Chen, R.L. Ma, M.Q. Zhang, ChemistryOpen 11 (2022) e202200091. |
| [40] |
J.Y. Sun, J. Yuan, Y.P. Li, et al., Sens. Actuator. B: Chem. 263 (2018) 208-217. |
| [41] |
J. Kaur, M. Dhall, Egypt J. Forensic Sci. 12 (2022) 7-15. |
| [42] |
W. Fan, J.W. Fang, G. Huang, Phys. Commun. 66 (2024) 102395. |
| [43] |
S. Mohapatra, P.S. Jeji, G.K. Pati, M. Mishra, T. Swarnkar, Biomed. Tech. 8 (2024) 1-16. |
| [44] |
S.Y. Huang, W.J. An, D.S. Zhang, N.R. Zhou, Opt. Commun. 533 (2023) 129287. |