Plasmonic nanopapers: flexible, stable and sensitive multiplex PUF tags for unclonable anti-counterfeiting applications†

Hongrui Cheng ^a, Yongfeng Lu ^a, Dongyan Zhu ^a, Lorenzo Rosa ^bc, Fei Han ^a, Mingguo Ma ^d, Wenyue Su ^a, Paul S. Francis ^e and Yuanhui Zheng *^a
aCollege of Chemistry, Fuzhou University, Fuzhou, Fujian 350116, China. E-mail: Yuanhui.Zheng@fzu.edu.cn
^bDepartment of Engineering “Enzo Ferrari”, University of Modena and Reggio Emilia, via Vivarelli 10, I-41125, Modena, Italy
^cApplied Plasmonics Lab, Centre for Micro-Photonics, Mail H74, P.O. Box 218, Hawthorn, VIC 3122, Australia
^dCollege of Materials Science and Technology, Beijing Forestry University, Beijing 100083, P.R. China
^eSchool of Life and Environmental Sciences, Faculty of Science, Engineering and Built Environment, Deakin University, Waurn Ponds, Victoria 3216, Australia

First published on 3rd April 2020

---

---

Introduction

Counterfeited and pirated goods pervade all levels of our modern society, causing serious economic damage and posing security threats to individuals and companies.^1–3 An effective way to combat counterfeiting is to apply security labels that carry two-dimensional (2D) patterns to products.³ These patterns are generally fabricated on plastic films, which inevitably cause white pollution. Nanopaper based on biodegradable cellulose nanofibres (CNFs) may be an ideal alternative support substrate for security labels. Firstly, it is sufficiently transparent to allow visible light to reach embedded encoded materials.^4–6 Secondly, it is an ultra-light, earth-abundant, and renewable material that can be derived from plant waste.⁷ Thirdly, CNFs with a nanosize width and high aspect ratio can lend mechanical flexibility to colloidal encryption materials, such as fluorescence quantum dots and plasmonic nanoparticles.

Traditional security labels are generally fabricated through inkjet printing, but they are vulnerable to counterfeiting through ongoing technical advances due to their predictable and deterministic encoding and decoding mechanisms.^8–10 The anti-counterfeiting performance of such security labels largely depends on restricted access to the inks. Widely investigated ink materials include stimuli-responsive molecules,^11–14 organic phosphorescent materials,^15–18 photonic crystals,^19,20 fluorescent materials (e.g., carbon dots,^21–23 g-C₃N₄ dots,^24–26 perovskite quantum dots,^27–29 II–VI group quantum dots³⁰ and up-conversion nanocrystals³¹), and plasmonic nanoparticles.^32–34 Among these materials, plasmonic nanoparticles have attracted great interest for the next generation of anti-counterfeiting labels, owing to their unique optical properties (i.e., they possess specific colours depending on their shapes and at the same time they dramatically enhance the fluorescence and Raman signals of surrounding molecules).³³ These properties originate from surface plasmon resonance (SPR): the resonant coupling of electromagnetic waves to collective oscillations of free electrons in metals. This enables control of light properties at the nanometre scale, enhancing and confining light at dimensions much smaller than the incident wavelength. SPR excitation in metallic nanoparticles involves localised surface plasmon resonance (LSPR) phenomena, which occur when conduction electrons under an applied electric field undergo collective harmonic oscillations, resulting in a dipolar response. The maximum displacement amplitudes are found at resonant frequencies in analogy to a damped, forced harmonic oscillator. LSPRs occur in nanoparticles with dimensions comparable to or smaller than the wavelength of the exciting radiation, and provide enhancement in the optical field intensity in close vicinity to their surface (also referred as to near-field intensity). These highly localised extreme field enhancement spots (i.e., hot spots) occur in the gaps between nanoparticles or at sharp corners of certain nanoparticle shapes (i.e., nanocubes and nanotriangles). The strength of the local fields provided by a given metallic nanostructure can be fine-tuned by tailoring its size, morphology, interparticle gap, porosity, and external environment (e.g., the refractive index of the surrounding medium).^35–38 Nanofabrication techniques provide access to many (nearly identical) plasmonic nanostructures with control of the position of hot spots.^35,37,39,40 The enormous local fields are capable of dramatically amplifying the fluorescence and Raman scattering signals of molecules in the vicinity of the hot spots,³³ where the metal nanostructures are used as optical signal magnifiers. This phenomenon allows us to spectrally and graphically encode multiple pieces of optical information by exploiting plasmonic nanostructures (i.e., LSPR wavelengths) and the surrounding molecules (i.e., fluorescence and Raman scattering) into a security label.³³

Physically unclonable functions (PUFs) could offer a practical solution to the drawback of conventional security labels being easily duplicated.^30,41–46 A significant progress in PUF encryption, relying on the undetermined random structures formed naturally during sample manufacturing, has been introduced in the anti-counterfeiting field.^30,41–46 Typical PUF features that have been developed up to now include rough surfaces,^42,43 discrete nanoparticle arrays,^33,44–46 and printed random dot patterns.³⁰ Each of these physical entities provides an optical fingerprint response, such as reflection/far-field scattering and fluorescence, which can be used as a unique identifier under external stimuli. Yet, surface-enhanced Raman scattering (SERS) has not been applied to PUF codes, although it has been widely used in conventional anti-counterfeiting applications.^47–53 Since SERS intensity is proportional to the fourth power of the localised optical field magnitude at the location of the Raman probe molecule, it is extremely sensitive to the metal nanostructures, which usually show a wide distribution of SERS enhancement factors.^35–37,54 For instance, even a uniform silver film-on-nanosphere (AgFON) SERS substrate exhibited variable SERS enhancement factors ranging from 2.8 × 10⁴ to 4.1 × 10¹⁰.⁵⁴ The strongest hot-spots, with SERS enhancement factors greater than 10⁹, only accounted for 63 out of a million of total Raman-active sites. This results in a random SERS intensity distribution across the whole surface of the sample, thus providing us with an opportunity to fabricate SERS-based PUF codes.

In this work, we developed a highly flexible and stable plasmonic nanopaper (PNP) composed of AgNCs and CNFs for unclonable anti-counterfeiting labels. The AgNCs offer chemical anchoring sites for the direct attachment of thiolated Raman molecules to the silver surface via silver-thiol chemistry, while the CNFs provide physical adsorption sites for dye molecule loading through inkjet printing. The direct binding of Raman molecules to AgNCs guarantees the highest Raman enhancement, as the near-field enhancement decays exponentially with the increase of the molecule-nanoparticle distance.⁵⁵ However, the printing of fluorophores on the CNFs avoids fluorescence quenching caused by the energy transfer from the excited molecules to the AgNCs. The reason for choosing AgNCs as plasmonic components is that they possess sharp corners (i.e., hot spots), which add additional near-field enhancement over spherical Ag nanoparticles.⁵⁶ They also show at least an order of magnitude higher maximum near-field enhancement than their gold counterparts at the corresponding resonance frequency.⁵⁷ The fabricated PNPs show excellent mechanical properties with a tensile strength of 63.92 MPa and strain fracture of 9.0%. They were used as multiplexed tags that carry LSPR, fluorescence, and Raman spectral and graphical information for anti-counterfeiting applications, thus increasing the security level. Significant enhancement of fluorescence emission and SERS enabled by the AgNCs is observed, in good agreement with simulated near-field enhancement surrounding the AgNCs. It is also found that the SERS intensity varies greatly from spot to spot, which enables us to produce PUF codes based on SERS mapping. Both LSPR and fluorescence encryption are used as the first layer of security and can be easily verified by the naked eye, while the SERS mapping based PUF encryption is used as the second layer of security and can be readily authenticated with the help of Raman spectroscopy using a computer vision technique previously developed by our group.⁵⁸

Results and discussion

The fabrication procedure of the PNP-based security labels is schematically illustrated in Fig. 1. AgNCs and CNFs were chosen as building blocks. The former offers rich optical properties and an extremely strong electromagnetic enhancement around the nanoparticles, especially at the eight corners with a theoretical near-field enhancement of 1.15 × 10⁵,⁵⁷ while the latter was adopted to protect the AgNCs from being removed during the deployment of the plasmonic security labels and provide mechanical flexibility to the nanopaper. The AgNC building blocks were synthesised using a well-established chemical route,⁵⁹ where cetyltrimethylammonium bromide (CTAB) and 5-bromosalicylic acid (5-BA) were used as stabilising agents. The ammonium groups of CTAB anchored on the AgNCs via the Ag–Br bond^60,61 conferred a net positive surface charge (ζ = +25.0 mV, Fig. S1 ESI†). The CNFs were derived from the renewable cellulose of garlic husk and were negatively charged (ζ = −28.2 mV, Fig. S1 ESI†) due to the abundant carboxylate groups on the surface of the nanofibres.⁶² When the colloidal AgNC solution was added dropwise to the CNF solution (step 1), the positively charged AgNCs were electrostatically assembled onto the negatively charged CNFs (step 2). Following the self-assembly, the mixed suspension was subjected to filtration under vacuum, forming a free-standing and flexible PNP (step 3). Our previous work on solar hydrogen evolution nanopapers using g-C₃N₄ and CNFs for photoactive and mechanical functions showed that, during the vacuum filtration process, the CNFs are stacked together through hydrogen bonds.⁶² The thickness of the PNP was found to be 51 μm using a Vernier caliper. The produced PNP was then immersed in a benzenethiol Raman reporter solution to allow the thiol molecule to bind to the silver surface,⁶³ encoding Raman signals in the PNP (step 4). After loading the Raman reporter, a fluorescein ink was printed into 2D patterns on the PNP, forming an invisible security label under ambient light (step 5).

Fig. 2 shows the typical transmission electron microscopy (TEM) images of the AgNC and CNF building blocks and scanning electron microscopy (SEM) images of the fabricated PNP. The TEM images confirm the cubic shape of the plasmonic nanoparticles with a mean edge length of 60 ± 10 nm (Fig. 2a) and the 1D nanofibre morphology of the nanocellulose with a width of ∼5 nm and length ranging from a few hundred nanometres to several micrometres (Fig. 2b). The top-view SEM image of the PNP (Fig. 2c) shows that individual silver nanoparticles with cubic and other shapes are randomly distributed on the surface of the nanopaper. It is believed that the synthesised AgNCs have rounded corners, which is supported by the simulation results shown in Fig. 5 (see the later section for detailed discussion). For TEM characterisation, the sample was prepared by loading the AgNCs on a super flat carbon film supported by a copper grid. Consequently, a perfect cubic shape was observed (Fig. 2a) because we were looking at the projection of the NCs lying down on the carbon film. However, the SEM images, obtained by detecting the secondary electrons produced by the interactions between the electron beam and the sample, provide the 3D spatial topography of the sample. It is well-known that SEM characterisation requires good conductivity of the sample to achieve clear images. Yet, the CNFs in the nanopaper are non-conductive, resulting in charge accumulation on the NCs during the SEM measurement. The accumulated charges can alter the emission direction of the secondary electrons from the different exposure faces of the NCs, causing poor contrast between these faces (i.e., the cubic shape in Fig. 2c is lost). Furthermore, since the NCs were randomly attached to the CNFs and subsequently vacuumed on a filter paper (step 2–3 of Fig. 1), they were distributed on the surface and in the bulk of the nanopaper with various orientations. They show different shapes depending on their orientations as schematically illustrated in Fig. S2 of the ESI.† This explains the observation of other shapes in the SEM image. Fig. 2c also shows the meshwork feature of the nanopaper, in good agreement with the previous reports on CNF-based nanopapers.^7,62 The nanopores are also irregularly distributed in the whole nanopaper, which stems from the random packing of CNFs during the vacuum filtration. The cross-section view SEM image of the same nanopaper (Fig. 2d) shows that the CNFs self-assembled into layered microstructures with wrinkle features. Randomly dispersed individual AgNCs were also observed on the cross-section of the nanopaper (inset of Fig. 2d), further confirming that the AgNCs were embedded in the nanopaper. This also explains the uniform colour across the nanopaper shown in the last panel of Fig. 1.

To investigate the chemical and mechanical stability of the fabricated PNPs, we carried out XPS and mechanical property measurements (Fig. 3). Fig. 3a shows the survey XPS spectrum of the PNP after exposure to air for two months, in which the three main peaks at 286.0 eV, 368.5 eV and 532.8 eV were ascribed to C 1s, Ag 3d, and O 1s. The carbon and oxygen species are believed to be mainly from CNFs, while the silver species are from the plasmonic NCs. Fig. 3b shows the high-resolution C 1s XPS spectrum of the PNP. There are four different types of carbon species with XPS peaks at 284.8 eV, 286.6 eV, 288.2 eV and 289.4 eV corresponding to the carbon in C–C, C–O, CO and O–CO groups, respectively.⁶²Fig. 3c displays the high-resolution O 1s XPS spectrum, where two different types of oxygen species, i.e., CO (531.8 eV) and C–O (533.0 eV) are observed. The C 1s and O 1s data match the chemical structure of CNFs (Fig. 1). The Ag 3d XPS spectrum (Fig. 3d) shows two symmetric peaks at 368.2 eV and 374.1 eV, which are assigned to metallic silver.⁶⁴ Generally, an ultrathin layer of silver oxide forms on the surface of metal silver when it is exposed to air. However, no oxidised silver is observed for the PNP (Fig. 3d) after being exposed to air for two months, indicating the good chemical stability of AgNCs in the nanopaper. It has been reported that the AgNCs synthesised in the presence of CTAB are capped with a dense double layer of CTAB,^60,61 conferring a net positive surface charge (ζ = +25.0 mV, Fig. S1 ESI†). The inner CTAB layer bound to the NCs through Ag–Br bonds,⁶¹ eliminating the unsaturated surface Ag atoms that can react with oxygen in air. The long carbon-chain double layers also suppressed the diffusion of oxygen molecules to the surface of the NCs. For the NCs embedded in the nanopaper, the CNFs offer additional isolation from air. In other words, both CTAB and CNFs can isolate the AgNCs from air, suppressing the oxidation reaction between silver and oxygen, which explains the Ag 3d XPS results shown in Fig. 3d. Fig. 3e shows the comparison of the stretching property of a commercial A4 paper, a CNF nanopaper (CNFNP), and the PNP, all of which show an elastic-plastic deformation behaviour. For the A4 paper, the tensile strength is 16.5 MPa and 40.7 MPa when stretched transversely (i.e., A4-T) and longitudinally (i.e., A4-L), respectively, while their tensile fracture strain is identical (i.e., 3.6%). This indicates that the cellulose fibres in the A4 paper are preferentially arranged along the longitudinal direction.⁶⁵ The tensile strength and the tensile fracture strain of the pure CNFNP is approximately 43.5 MPa and 9.7%, respectively. The former is almost the same as the longitudinal tensile strength of the A4 paper, while the latter is twice as high as that of the A4 paper, implying that the flexibility of the CNFNP is superior. After the AgNCs were embedded in the CNFNP, its tensile strength is further increased to 62.9 MPa but the tensile fracture strain slightly decreased to 9.0%. Compared with the A4 paper, the higher tensile strength and fracture strain of the PNP indicate that it can withstand higher stress (i.e. weight) without breaking and it exhibits better flexibility. The PNP with a thickness of 51 μm can withstand a stress (weight) of 32 N (∼3 kg) without breaking. It also remains intact even when immersed in water and subjected to 400 W ultrasonication for 30 min (see ESI video 1 and Fig. S3 ESI†). The high chemical and mechanical stability of the PNP is attributed to both the strong CNF-CNF hydrogen bonding and AgNC-CNF electrostatic attraction. Fig. 3f shows the comparison of the Young's modulus of the A4 paper, the CNFNP and the PNP, which is 486.8 ± 30 MP (transverse)/1141.5 ± 20 MP (longitudinal), 450.0 ± 60 MPa and 690.9 ± 40 MPa, respectively.

Fig. 4a shows the absorption spectra of PNP, CNFNP and AgNC solution. The absorption spectra were recorded under either reflection (R) or transmission (T) mode. When measured under reflection mode, the CNFNP shows a maximum absorption at 251 nm with a shoulder at 356 nm. Upon the integration of CNFs with AgNCs, the resulting PNP displays a broad absorption band at 200–300 nm and two strong absorption peaks at 365 nm and 399 nm (PNP-R), which were attributed to the absorption of CNFs and AgNCs, respectively. When measured under transmission mode, the AgNC colloid exhibits three characteristic plasmon absorption peaks at 350 nm, 382 nm and 472 nm (AgNCs), in good agreement with those reported in the literature.^59,66 For PNP, the absorption spectrum (PNP-T) is similar to that of the AgNC colloid when measured under transmission mode. Compared with the AgNC colloid, all the characteristic plasmon resonance peaks are slightly red-shifted and wider. The peak broadening is due to the change of refractive index from the water medium to a mixture of air and CNF. Fig. 4b shows the comparison of the fluorescence spectra of fluorescein printed on a CNFNP and PNP. The fluorescence intensity of PNP increases by a factor of 3.6 compared with that of CNFNP. Fig. 4c shows the SERS spectra of benzenethiol from PNP and CNFNP exposed to 10⁻⁶ M benzenethiol solution overnight. Strong Raman peaks at 417 cm⁻¹, 691 cm⁻¹, 760 cm⁻¹, 999 cm⁻¹, 1022 cm⁻¹, 1075 cm⁻¹, 1450 cm⁻¹ and 1573 cm⁻¹ originating from benzenethiol were observed for the PNP,^36,37 while no distinct Raman peaks were observed for CNFNP. The fluorescence and Raman enhancements are due to the LSPR effect from AgNCs. Fig. 4d shows the spot-to-spot SERS intensity variation at 1075 cm⁻¹ from the PNP. Some of the SERS spectra including the strongest and the weakest spots are shown in Fig. S4 of the ESI.† The average signal intensity at the 1075 cm⁻¹ peak for the PNP is 1.3 × 10⁴ counts with a coefficient of variation of 16.2%. The random SERS intensity distribution allows us to encode Raman signals in the 2D pattern (inset of Fig. 4d), thus generating unbreakable Raman-based PUF codes.

To explain the observed spectral enhancements, we modelled the absorption spectrum and near-field enhancement of AgNCs in PNP using the 3D finite-difference time-domain (3D-FDTD) method (Fig. 5). Since the AgNCs are embedded in porous CNFs, the refractive index of the medium surrounding the NCs should be in-between 1.0 (n_air) and 1.53 (n_CNF). Fig. 5a shows the normalised absorption spectra of AgNCs in media with the refractive index varying from 1.0 to 1.53. When the medium is water (n = 1.33), the simulated absorption spectrum (pink curve) matches well with that of the experimental one (yellow curve of Fig. 4a). When the refractive index is 1.43, it gives the best match to the absorption spectrum of the PNP measured under transmission mode (red curve of Fig. 4a). The resonance broadening of the experimental absorption spectrum might be attributed to the inhomogeneous nanoparticle size (60 ± 10 nm) and dielectric environment surrounding the NCs originating from different spot-to-spot porosity of the PNP. The maximum electromagnetic field enhancement at the corner of the NCs in a medium with a refractive index of 1.43 as a function of the excitation wavelength is displayed in Fig. 5b. The AgNCs show two major electromagnetic field enhancement regions at 380–440 nm and 440–800 nm. The latter matches the fluorescence emission band of fluorescein (highlighted in orange) and Raman spectrum of benzenethiol (marked with green). Fig. 5c shows the spatial electric field intensity distribution (|E|²) in the vicinity of the NC surface at an excitation wavelength of 532 nm. As expected, the electric field enhancement is mainly localised near the corners of the NCs, resulting in an inhomogeneous electric field enhancement around the NCs.

The maximum SERS enhancement factor (|E|⁴) is calculated to be 7.2 × 10³. The inhomogeneous nanoparticle size of AgNCs, the variation of the refractive index of the surrounding media for each AgNCs and the inhomogeneous electric field enhancement around the NCs at a fixed refractive index are responsible for the observed spot-to-spot variations of SERS intensities. Fig. 5d shows that the electric field enhancement decays exponentially away from the metal surface, which explains the relatively low fluorescence enhancement.

It is well-known that metal nanoparticles such as AgNCs can interact with light in multiple fashions. This leads to the generation of multiple optical signals, such as far-field scattering, surface plasmon enhanced fluorescence and SERS, as discussed above. Encoding all this optical information in a PNP as illustrated in Fig. 1 can produce multiplexed 2D graphic security labels with an increased level of security (Fig. 6). Fig. 6a–c show photographs of a plasmonic multiplex security label with Fuzhou University logos printed on the back side using carbon ink (i.e., conventional security label) and the front side using fluorescein ink. There appears a uniform yellow-green (Fig. 6a) and wine-red (Fig. 6b) colour across the whole paper in reflected and transmitted light, respectively (note that the printed fluorescein pattern is invisible to the naked eye under the ambient environment). This phenomenon shares the same principle of the Lycurgus Cup displayed in the British Museum (Fig. S5 ESI†); that is, metal nanoparticles support LSPRs, giving them their colours. The security label also carries covert fluorescence and SERS patterns that can only be viewed under UV illumination (Fig. 6c) or laser excitation (Fig. 6d–f). Since AgNCs are randomly dispersed in the whole porous CNFNP, their local porosity (i.e., surrounding environments) are different. For example, some AgNCs expose more surface to air, especially those on the surface of the nanopaper, while others are covered with different degrees of CNFs. In other words, the refractive index of the surrounding media for each AgNCs is different (Fig. 5a), leading to random SERS intensity distributions. Furthermore, the inhomogeneous particle size and electric field enhancement around the NCs at a fixed refractive index (Fig. 5c) also contribute to the variation of spot-to-spot SERS intensities, as discussed above. This feature of randomness non-deterministically encodes graphical SERS information in the PNP, guaranteeing intrinsically unreplicable code outputs.

The colours of the PNP originate from LSPR and the fluorescent logo can be easily verified by the naked eye, while the Raman-based PUF codes were authenticated with the conventional algorithm for image processing (i.e., the technique used for the verification of quick response (QR) codes)^{33,44,46,67,68} machine⁴³/deep learning^30,69–71 or the computer vision technique.⁵⁸ The authentication technique using the conventional algorithm requires a physical mark similar to those on the corners of a QR code on the security label to define the positions of all the pixels within the security label.^33,44,46,67 Although deep learning approaches do not need physical marks for authentication, it requires a time-consuming image training process to learn the key features (i.e., keypoints) on the security labels.^30,69 Here, we adopted the computer vision approach, previously developed by our group, which does not require any physical marks and training steps for Raman-based PUF code authentication.⁵⁸ In a typical authentication process, twenty SERS mappings were randomly selected to establish a security label database (named d_n, n = 1, 2, …, 6) (Fig. 6d). These images are then sent to the computer vision engine to extract the keypoint features using the Hamming distance algorithm. Once the keypoint features of the security labels have been extracted, the original SERS mappings used to establish the database are deleted from the computer vision engine. Then, we offer the computer vision engine twelve SERS mappings consisting of real (50%) and fake (50%) security labels for authentication. A SERS mapping (d₁) that represented a genuine product was obtained at different focus levels (e_1–6) to make sure it could be validated in different authentication scenarios (Fig. 6e). Note that SERS mapping e₁ is identical to d₁ in the database. Then a brute-force matcher was used for matching the keypoint features between images e_n and d_n. We used the similarity coefficient (S) that can be calculated using eqn (1) to describe the matching score:

S = (Nm/N0 + Nm/Nt) × ½	(1)

where N₀ is the number of total keypoint features of the SERS mapping from the database; N_t is the number of total keypoint features of the SERS mapping that are about to be verified; and N_m is the number of the matched keypoint features between the two images.⁵⁸ The matching score outputs for the images shown in Fig. 6e to d₁ are 100% (e₁), 89.1% (e₂), 78.1% (e₃), 85.9% (e₄), 78.1% (e₅), and 89.1% (e₆), respectively (the first row of Fig. 6g). As expected, a variation in the imaging focus level results in a change of the matching score. Using a software application to control the SERS mapping quality during acquisition enables an increase in similarities, making them closer to 100%. However, the matching scores of the images shown in Fig. 6e to d_2–6 are very small and close to zero. The same phenomenon occurs with the six fake security labels (Fig. 6h). For example, the matching score of the image f_1–6 to d₁ is 5.2% (f₁), 5% (f₂), 4.4% (f₃), 5.3% (f₄), 5.6% (f₅), and 5.8% (f₆), respectively (the first row of Fig. 6h). By setting a threshold value of 40%, we can distinguish the real security labels from the fake ones with 100% accuracy. The computer vision engine is immediately able to provide the end users with the authentication outcomes.

Conclusions

We have demonstrated a new unclonable anti-counterfeiting system based on highly flexible and stable plasmonic nanopapers, which were fabricated via a vacuum filtration method using silver nanocubes and cellulose nanofibres as building blocks. The fabricated plasmonic nanopapers show high fluorescence and Raman enhancements, excellent mechanical properties, and a random distribution of Raman intensity across the whole nanopaper. Molecular information from any optically active molecules can be encoded in the nanopaper. As a proof-of-concept demonstration, the plasmonic nanopapers were coded with multiplex optical signals including surface plasmon resonance, fluorescence and SERS for anti-counterfeiting applications. The surface plasmon resonance and fluorescence information is used as the first layer of security and can be easily verified by the naked eye, while the unclonable SERS mapping with random Raman intensity distribution is used as a second layer of security that can be readily authenticated by Raman spectroscopy using the computer vision technique. Finally, coupling excellent mechanical robustness and stability, the low-cost security label developed here has great potential for commercialisation.

Experimental

Materials and chemicals

Reagent-grade sulfuric acid, sodium chlorite, silver nitrate, glucose, ammonium hydroxide, and cetyltrimethylammonium bromide (CTAB) were purchased from Sinopharm chemical reagent co. Ltd (Shanghai, China); analytical reagent 5-bromosalicylic acid (5-BA) and benzenethiol were purchased from Aladdin industrial corporation (Shanghai China); and analytical reagent fluorescein was purchased from Sigma-Aldrich (Shanghai, China).

Synthesis of AgNCs

The AgNC building blocks were synthesised using a previously reported method with minor modifications.⁵⁹ In a typical procedure, 2 mL of 20 mM [Ag(NH₃)₂]OH, 2.4 mL of 50 mM CTAB, 4 mL of 5 mM 5-BA, and 3 mL of 16 mM glucose were sequentially added into a Teflon-lined stainless steel autoclave with a capacity of 25 mL under stirring. Note that (1) the Teflon containers have to be treated with aqua regia for 1 h and then cleaned with ultrapure water prior to use; and (2) the [Ag(NH₃)₂]OH solution was prepared by adding ammonium hydroxide into an aqueous solution of AgNO₃ until the solution changed colour from brown to transparent. The solution was sealed and heated at 120 °C for 8.5 h. The solution was then cooled to room temperature. The colloid in the Teflon containers was carefully collected in 5 mL centrifugation tubes with a pipette and centrifuged at 862 rcf for 25 min. The sediment was collected and re-dispersed in ultrapure water. This process was repeated three times to remove the unreacted reagents.

Extraction of CNFs from garlic husk

The extraction of CNFs from garlic husk was conducted using previously reported methods with minor modification.^7,62 In a typical procedure, 10 g of dried garlic husk was immersed in 400 g of NaOH solution (2 wt%) at 140 °C for 5 h and then washed with ultrapure water to the neutral pH value. Subsequently, the obtained garlic skin with 200 g of H₂SO₄ solution (2 wt%) and 300 g of sodium chlorite (1.5 wt%) was stirred at 80 °C for 6 h to remove the lignin. The resulting cellulose slurry was then ultrasonicated at 600 W for 1 h and then stirred at 1200 rpm for 1 h to obtain a CNF suspension, which was washed with ultrapure water through centrifugation three times to remove the chemical reactants and finally re-dispersed in ultrapure water.

Preparation of AgNC/CNF PNPs

AgNC/CNF PNPs were prepared through the self-assembly assisted vacuum filtration method previously established by our group.⁶² In a typical procedure, 5 mL of CNF suspension (4.56 mg mL⁻¹) were added into a reaction cup with pipettes. Subsequently, 500 μL of 0.79 mg mL⁻¹ AgNC solution was added to the CNF suspension under continuous stirring. The mixed solution was stirred for 30 min to form a uniform suspension. The mixture was then filtered to form a PNP under vacuum.

Theoretical modelling

The optical properties and near-field enhancement of a single AgNC were simulated with the well-established 3D-FDTD method^35,72 using Lumerical software (FDTD Solutions, Inc.). The 3D-FDTD simulations were performed on a single AgNC with the edge of 60 nm and a vertex radius of 5 nm in media with different refractive indexes varying from 1.0 (air) to 1.53 (CNF). The AgNC was enclosed in a domain with a size of 400 × 400 × 400 nm³, lined with perfectly matched layers to absorb stray radiation. The square mesh size was 1.5 nm and the AgNC was excited by a plane-wave total-field scattered-field source ranging from 300–800 nm, and the total and scattered fields were collected by sets of monitors surrounding the AgNC.

Characterisation

A transmission electron microscope (TEM, Tecnai Model G2 F20 S-TWIN) was used to examine the morphology of the AgNC and CNF. A high-resolution field emission scanning electron microscope (SEM, Verios G4) was used to examine the morphology of the PNP. The surface elemental states of the samples were analysed by X-ray photoelectron spectroscopy (XPS, VG ESCALAB 250) with a monochromatised Al Kα X-ray source (15 kV, 150 W, 500 nm, pass energy 20 eV). The binding energy of C 1s (284.8 eV) was used for the spectral calibration. The optical properties were measured using a UV-vis spectrophotometer (Cary 500, Varian Co.). Photoluminescence (PL) spectra were recorded using a home-made dark-field, fluorescence and Raman multi-functional microscope (i.e., DFR multi-functional microscope) or a fluorophotometer (Edinburgh FL/FS900). Raman spectra were recorded on a home-made DFR multi-functional microscope or an inVia Reflex Raman microscope (Renishaw OPTIMA 8000) with an incident laser wavelength of 532 nm.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work is funded by the National Nature Science Foundation of China (Grant No. 61605028 and 61775040), Program for Minjiang Scholar (YZ), and Program for Thousand Young Talent plan (YZ).

Notes and references

1. Z. Hu, J. Comeras, H. Park, J. Tang, A. Afzali, G. Tulevski, J. Hannon, M. Liehr and S. Han, Nat. Nanotechnol., 2016, 11, 559 CrossRef CAS PubMed .
2. P. Aldhous, Nature, 2005, 434, 132 CrossRef CAS PubMed .
3. R. Arppe and T. J. Sørensen, Nat. Rev. Chem., 2017, 1, 0031 CrossRef CAS .
4. H. Fukuzumi, T. Saito, T. Iwata, Y. Kumamoto and A. Isogai, Biomacromolecules, 2009, 10, 162 CrossRef CAS PubMed .
5. M. Nogi, S. Iwamoto, A. N. Nakagaito and H. Yano, Adv. Mater., 2009, 21, 1595 CrossRef CAS .
6. H. Abral, J. Ariksa, M. Mahardika, D. Handayani, I. Aminah, N. Sandrawati, A. B. Pratama, N. Fajri, S. M. Sapuan and R. A. Ilyas, Food Hydrocolloids, 2020, 98, 105266 CrossRef CAS .
7. W.-T. Cao, F.-F. Chen, Y.-J. Zhu, Y.-G. Zhang, Y.-Y. Jiang, M.-G. Ma and F. Chen, ACS Nano, 2018, 12, 4583 CrossRef CAS PubMed .
8. J. Andres, R. D. Hersch, J.-E. Moser and A.-S. Chauvin, Adv. Funct. Mater., 2014, 24, 5029–5036 CrossRef CAS .
9. B. Duong, H. Liu, C. Li, W. Deng, L. Ma and M. Su, ACS Appl. Mater. Interfaces, 2014, 6, 8909 CrossRef CAS PubMed .
10. R. Xuan and J. Ge, Langmuir, 2011, 27, 5694 CrossRef CAS .
11. Y.-D. He, Z.-L. Zhang, J. Xue, X.-H. Wang, F. Song, X.-L. Wang, L.-L. Zhu and Y.-Z. Wang, ACS Appl. Mater. Interfaces, 2018, 10, 5805 CrossRef CAS PubMed .
12. R. Thirumalai, R. D. Mukhopadhyay, V. K. Praveen and A. Ajayaghosh, Sci. Rep., 2015, 5, 9842 CrossRef PubMed .
13. K. Jiang, L. Zhang, J. Lu, C. Xu, C. Cai and H. Lin, Angew. Chem., Int. Ed., 2016, 55, 7231 CrossRef CAS PubMed .
14. K. Jiang, Y. Wang, C. Cai and H. Lin, Adv. Mater., 2018, 30, 1800783 CrossRef PubMed .
15. Z. An, C. Zheng, Y. Tao, R. Chen, H. Shi, T. Chen, Z. Wang, H. Li, R. Deng, X. Liu and W. Huang, Nat. Mater., 2015, 14, 685 CrossRef CAS PubMed .
16. L. Bian, H. Shi, X. Wang, K. Ling, H. Ma, M. Li, Z. Cheng, C. Ma, S. Cai, Q. Wu, N. Gan, X. Xu, Z. An and W. Huang, J. Am. Chem. Soc., 2018, 140, 10734 CrossRef CAS PubMed .
17. Y. Su, S. Z. F. Phua, Y. Li, X. Zhou, D. Jana, G. Liu, W. Q. Lim, W. K. Ong, C. Yang and Y. Zhao, Sci. Adv., 2018, 4, eaas9732 CrossRef PubMed .
18. Y. Su, Y. Zhang, Z. Wang, W. Gao, P. Jia, D. Zhang, C. Yang, Y. Li and Y. Zhao, Angew. Chem., Int. Ed., 2019, 131, 1 Search PubMed .
19. H. S. Lee, T. S. Shim, H. Hwang, S.-M. Yang and S.-H. Kim, Chem. Mater., 2013, 25, 2684 CrossRef CAS .
20. Y. Li, X. Zhou, Q. Yang, Y. Li, W. Li, H. Li, S. Chen, M. Li and Y. Song, J. Mater. Chem. C, 2017, 5, 4621 RSC .
21. L. Pan, S. Sun, A. Zhang, K. Jiang, L. Zhang, C. Dong, Q. Huang, A. Wu and H. Lin, Adv. Mater., 2015, 27, 7782 CrossRef CAS PubMed .
22. W. Li, W. Zhou, Z. Zhou, H. Zhang, X. Zhang, J. Zhuang, Y. Liu, B. Lei and C. Hu, Angew. Chem., Int. Ed., 2019, 58, 7278 CrossRef CAS PubMed .
23. P. Long, Y. Feng, C. Cao, Y. Li, J. Han, S. Li, C. Peng, Z. Li and W. Feng, Adv. Funct. Mater., 2018, 28, 1800791 CrossRef .
24. G. Li, Z. Lian, W. Wang, D. Zhang and H. Li, Nano Energy, 2016, 19, 446 CrossRef CAS .
25. X. Zhang, H. Wang, H. Wang, Q. Zhang, J. Xie, Y. Tian, J. Wang and Y. Xie, Adv. Mater., 2014, 26, 4438 CrossRef CAS PubMed .
26. Z. Song, T. Lin, L. Lin, S. Lin, F. Fu, X. Wang and L. Guo, Angew. Chem., Int. Ed., 2016, 55, 2773 CrossRef CAS PubMed .
27. C. Sun, S. Su, Z. Gao, H. Liu, H. Wu, X. Shen and W. Bi, ACS Appl. Mater. Interfaces, 2019, 11, 8210 CrossRef CAS PubMed .
28. L. Xu, J. Chen, J. Song, J. Li, J. Xue, Y. Dong, B. Cai, Q. Shan, B. Han and H. Zeng, ACS Appl. Mater. Interfaces, 2017, 9, 26556 CrossRef CAS PubMed .
29. H. Wang, W. Yao, Q. Tian, M. Li, B. Tian, L. Liu, Z. Wu and W. Wu, Adv. Mater. Technol., 2018, 3, 1800150 CrossRef .
30. Y. Liu, F. Han, F. Li, Y. Zhao, M. Chen, Z. Xu, X. Zheng, H. Hu, J. Yao, T. Guo, W. Lin, Y. Zheng, B. You, P. Liu, Y. Li and L. Qian, Nat. Commun., 2019, 10, 2409 CrossRef PubMed .
31. Y. Liu, K. Ai and L. Lu, Nanoscale, 2011, 3, 4804 RSC .
32. Y. Liu, Y. H. Lee, M. R. Lee, Y. Yang and X. Y. Ling, ACS Photonics, 2017, 4, 2529 CrossRef CAS .
33. Y. Zheng, C. Jiang, S. H. Ng, Y. Lu, F. Han, U. Bach and J. J. Gooding, Adv. Mater., 2016, 28, 2330 CrossRef CAS PubMed .
34. P. Nordlander, N. J. Halas and N. A. Mortensen, Nat. Rev. Mater., 2016, 2, 16088 Search PubMed .
35. Y. Zheng, A. H. Soeriyadi, L. Rosa, S. H. Ng, U. Bach and J. J. Gooding, Nat. Commun., 2015, 6, 8797 CrossRef CAS PubMed .
36. Y. Zheng, T. Thai, P. Reineck, L. Qiu, Y. Guo and U. Bach, Adv. Funct. Mater., 2013, 23, 1519 CrossRef CAS .
37. X. Zhang, Y. Zheng, X. Liu, W. Lu, J. Dai, D. Y. Lei and D. R. MacFarlane, Adv. Mater., 2015, 27, 1090 CrossRef CAS PubMed .
38. S. Lal, N. K. Grady, G. P. Goodrich and N. J. Halas, Nano Lett., 2006, 6, 2338 CrossRef CAS PubMed .
39. Y. Zheng, L. Rosa, T. Thai, S. H. Ng, D. E. Gómez, H. Ohshima and U. Bach, J. Mater. Chem. A, 2015, 3, 240 RSC .
40. T. Thai, Y. Zheng, S. H. Ng, S. Mudie, M. Altissimo and U. Bach, Angew. Chem., Int. Ed., 2012, 51, 8732 CrossRef CAS PubMed .
41. R. Pappu, B. Recht, J. Taylor and N. Gershenfeld, Science, 2002, 297, 2026 CrossRef CAS PubMed .
42. J. D. R. Buchanan, R. P. Cowburn, A.-V. Jausovec, D. Petit, P. Seem, G. Xiong, D. Atkinson, K. Fenton, D. A. Allwoo and M. T. Bryan, Nature, 2005, 436, 475 CrossRef CAS PubMed .
43. H. J. Bae, S. Bae, C. Park, S. Han, J. Kim, L. N. Kim, K. Kim, S.-H. Song, W. Park and S. Kwon, Adv. Mater., 2015, 27, 2123 CrossRef .
44. A. F. Smith, P. Patton and S. E. Skrabalak, Adv. Funct. Mater., 2016, 26, 1315 CrossRef CAS .
45. J. Kim, J. M. Yun, J. Jung, H. Song, J.-B. Kim and H. Ihee, Nanotechnology, 2014, 25, 155303 CrossRef PubMed .
46. M. R. Carro-Temboury, R. Arppe, T. Vosch and T. J. Sørensen, Sci. Adv., 2018, 4, e1701384 CrossRef PubMed .
47. Y. Liu, Y. H. Lee, Q. Zhang, Y. Cui and X. Y. Ling, J. Mater. Chem. C, 2016, 4, 4312 RSC .
48. K. J. Si, P. Guo, Q. Shi and W. Cheng, Anal. Chem., 2015, 87, 5263 CrossRef CAS PubMed .
49. Y. Chen, K. J. Si, D. Sikdar, Y. Tang, M. Premaratne and W. Cheng, Adv. Opt. Mater., 2015, 3, 919 CrossRef CAS .
50. K. J. Si, D. Sikdar, L. W. Yap, J. K. K. Foo, P. Guo, Q. Shi, M. Premaratne and W. Cheng, Adv. Opt. Mater., 2015, 3, 1710 CrossRef CAS .
51. D. Li, L. Tang, J. Wang, X. Liu and Y. Ying, Adv. Opt. Mater., 2016, 4, 1475 CrossRef CAS .
52. Y. Cui, I. Y. Phang, Y. H. Lee, M. R. Lee, Q. Zhang and X. Y. Ling, Chem. Commun., 2015, 51, 5363 RSC .
53. Y. Cui, R. S. Hegde, I. Y. Phang, H. K. Lee and X. Y. Ling, Nanoscale, 2014, 6, 282 RSC .
54. Y. Fang, N.-H. Seong and D. D. Dlott, Science, 2008, 321, 388 CrossRef CAS PubMed .
55. K. L. Tsakmakidis, R. W. Boyd, E. Yablonovitch and X. Zhang, Opt. Express, 2016, 24, 17916 CrossRef CAS PubMed .
56. S. Ben-Jaber, W. J. Peveler, R. Quesada-Cabrera, C. W. O. Sol, I. Papakonstantinou and I. P. Parkin, Nanoscale, 2017, 9, 16459 RSC .
57. N. Hooshmand, D. O'Neil, A. M. Asiri and M. EI-Sayed, J. Phys. Chem. A, 2014, 118, 8338 CrossRef CAS PubMed .
58. F. Han, Y. Liu, F. Li, Y. Lu, H. Cheng, Y. Lin, T. Zhao, S. H. Ng, U. Bach and Y. Zheng, J. Mater. Chem. C, 2019, 7, 13040 RSC .
59. D. Yu and V. W.-W. Yam, J. Am. Chem. Soc., 2004, 126, 13200 CrossRef CAS PubMed .
60. S. Zhou, D. S. Mesina, M. A. Organt, T.-H. Yang, X. Yang, D. Huo, M. Zhao and Y. Xia, J. Mater. Chem. C, 2018, 6, 1384 RSC .
61. K. Liu, Y. Zheng, X. Lu, T. Thai, N. A. Lee, U. Bach and J. J. Gooding, Langmuir, 2015, 31, 4973 CrossRef CAS PubMed .
62. W. Sun, H. Cheng, N. Lin, Y. Lu, L. Chen, Y. Zhao, P. S. Francis, N. Zhuang and Y. Zheng, Appl. Catal., B, 2020, 260, 118083 CrossRef CAS .
63. O. M. Bakr, V. Amendola, C. M. Aikens, W. Wenseleers, R. Li, L. DallNegro, G. C. Schatz and F. Stellacci, Angew. Chem., Int. Ed., 2009, 48, 5921 CrossRef CAS PubMed .
64. Y. Zheng, L. Zheng, Y. Zhan, X. Lin, Q. Zheng and K. Wei, Inorg. Chem., 2007, 46, 6980 CrossRef CAS PubMed .
65. G. Chinga-Carrasco, J. Microsc., 2009, 234, 211 CrossRef CAS PubMed .
66. T. Jiang, B. Wang, L. Zhang and J. Zhou, J. Alloy. Compd., 2015, 632, 140 CrossRef CAS .
67. R. Arppe-Tabbara, M. Tabbara and T. J. Sørensen, ACS Appl. Mater. Interfaces, 2019, 11, 6475 CrossRef CAS .
68. K. Nakayama and J. Ohtsubo, Opt. Eng., 2012, 4, 040506 CrossRef .
69. X. He, Y. Gu, B. Yu, Z. Liu, K. Zhu, N. Wu, X. Zhao, Y. Wei, J. Zhou and Y. Song, J. Mater. Chem. C, 2019, 7, 14069 RSC .
70. M. H. S. Segler, M. Preuss and M. P. Waller, Nature, 2018, 555, 604 CrossRef CAS PubMed .
71. M. Ziatdinov, O. Dyck, A. Maksov, X. Li, X. Sang, K. Xiao, R. R. Unocic, R. Vasudevan, S. Jesse and S. V. Kalinin, ACS Nano, 2017, 11, 12742 CrossRef CAS PubMed .
72. A. Bottomley, D. Prezgot, J. P. Coyle and A. Lanoul, Nanoscale, 2016, 8, 11168 RSC .

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0nr01223h

This journal is © The Royal Society of Chemistry 2020