Cholesteric Liquid Crystals Based Micro‐Fingerprints Generator for Anti‐Counterfeiting Labels

In this study, a method is presented to fabricate optical fingerprint‐like patterns. The method relies on the use of commercially available chiral nematic liquid crystals (CNLCs) confined in microspheres. The peculiar optical texture is obtained by applying a high‐frequency voltage to the micrometric objects able to distort the molecular director orientation. The texture can be stabilized by doping the CNLC with photosensitive materials. Each microsphere shows a different fingerprint‐like pattern that is generated in a completely random manner making this procedure suitable to create physical unclonable functions (PUFs) keys. The optical patterns can be stored to create an artificial fingerprints database or they can be used to fabricate electroluminescent labels to be exploited as complex anti‐counterfeiting devices. An authentication software is developed and used to test the robustness of the proposed anti‐counterfeiting system.


Introduction
Today, there is a growing need for simple, effective, and affordable novel technical solutions for the authentication, identification, and traceability of goods.Historically, cryptographic primitives, algorithms used to build cryptographic protocols for computer security, have provided the most efficient protection for data exchange.Nowadays, cryptographic primitives need to be anchored to the physical world to resist physical attacks and provide physical security.An example of these primitives are the physically unclonable functions (PUFs). [1,2]PUFs are based on unique physical keys generated randomly.A PUF key is a physical manifestation of a cryptographic key and it is impossible to counterfeit. [3]Different types of PUFs exist and some of them are based on pattern recognition.In recent years, several methodologies have been developed to produce artificial fingerprint-like structures.These structures are appealing since fingerprints are rich in minutiae, furrows, and crests that bifurcate, curve, end abruptly, or cross, then they can be used to encode a high volume of information.In 2014, J. Kim et al. obtained nanoscopic fingerprints using randomly distributed silver nanowires coated with fluorescent dyes embedded in a PET film. [4]In 2015 Bae et al. obtained artificial micro fingerprints following the random drying procedure of a silicon film encapsulating discotic polymeric particles few hundreds of microns in size. [5]From an encoding point of view, the fingerprint approach has a huge capacity that can be further increased if the pattern is enriched with other unique features.As an example, in 2016 Tian et al. obtained a fingerprint-like structure by doping a random folded ultrathin gelatinous polymer with plasmonic nanostructures in the form of silver-coated gold nanorods. [6]Then, in addition to the random pattern, the structure exhibited a unique surface enhanced Raman spectroscopy signal.The possibility to superimpose random patterns of different nature (Raman and fluorescence emission, melting points, and molecular masses) dramatically improves the encoding capacity of the PUF key.[9] Soft materials and, among them, liquid crystals (LCs) are suitable candidates for the creation of PUFs since they are excellent optical materials capable of self-assemble in complex photonic structures and self-repair.In particular, chiral nematic (CN) LCs show peculiar optical properties. [10]They possess a self-organized periodical helical arrangement of the molecules.In CNLCs the period of the structure is equal to half the pitch p of the helix, and light propagating along the helical axes undergoes Bragg-type reflection.These materials act as 1D photonic structures and exhibit a photonic bandgap.[18] In 2012 Nakayama et al. have proposed the use of chiral LCs as materials for optical security devices, developing a way to prepare random patterns with a fingerprint-like texture. [19]Recently, a lot of research has been carried out on the peculiar optical properties of LCs confined inside spheres, tens of microns in size. [20][23][24] Furthermore, topological transitions in chiral nematic microdroplets have been investigated as a function of external fields (temperature gradients, electric, and electromagnetic fields) obtaining a lot of complex free-standing metastable topological architectures at the micron scale. [25]Here, we demonstrate that chiral LCs encapsulated in micrometric droplets and immersed in an isotropic fluid are able to self-organize, in a strong electric field, in a structure that optically resembles a fingerprint texture.Each droplet immersed in the field shows a unique and randomly generated structure easily revealed through microscopy techniques.These micro objects can be used as PUFs keys in anti-counterfeiting applications.The coding capability of these patterns can be greatly increased by doping chiral LCs with materials such as fluorescent dyes or metallic nanoparticles that can provide them with additional optical properties.Further, we present a procedure to rescale the micron sized textures using confocal microscopy and screen printing techniques.This allows both to create a library of digital fingerprint images and to fabricate electroluminescent labels.The latter has been tested as anti-counterfeiting systems.

Cholesteric Microfingerprints
In this work, LCs are confined in a spherical geometry.The fabrication of liquid crystal microdroplets is a straightforward procedure.It has been shown that a LC , confined in an immiscible fluid matrix, after a shaking procedure, will separate in form of droplets having a spherical shape and diameters ranging from few millimeters to few microns. [26,27]CLCs are prepared by adding a precise amount of chiral dopant to a nematic matrix.These materials self-assemble in a chiral structure, with the pitch of the cholesteric helix decreasing with increasing the chiral dopant amount. [28]In this case, the pitch of the mixture is 2.5 μm.A small amount of a fluorescent dye is added to the chiral mixture: this procedure is necessary to visualize the optical texture using Confocal Laser Scanning Microscopy (CLSM).Glycerol is chosen as fluid matrix since it is able to provide a planar alignment of the LC molecules at the interface.When the cholesteric pitch is smaller than the droplet's radius, the formation of an internal radial configuration of the cholesteric helices is obtained.Cholesteric layers are bent in concentric spherical surfaces forming the typical Frank-Pryce texture (Figure 1a), with the Maltese cross appearing when the spheres are observed using an optical microscope between crossed polarizers (Figure 1b). [27]Furthermore, glycerol is a viscous medium, necessary condition to hinder the microspheres migration toward the electrodes when an electric field is applied to the emulsion and to prevent coalescence.Emulsions are infiltrated in optical cells 200 μm thick.Microspheres' texture is investigated while applying an electric field of variable intensity at 1MHz perpendicular to the cell plates, to prevent the movement of microspheres (Figure 2a). Figure 2b shows a CLSM image of a single droplet, 58 μm in diameter, possessing a Frank-Pryce texture when no voltage is applied to the cell.The optical pattern starts to distort at around 0.4V μm −1 .Below this voltage value, the Frank-Pryce texture is completely recovered when the electric field is switched off.A fingerprint-like pattern appears at 0.6V μm −1 applied to the cell and this texture stabilizes at 0.7V μm −1 .The fingerprint texture formation is not the only observable phenomenon, in fact, also a variation of the droplet shape is recorded (Figure 2b).Since the droplet volume is fixed, an increase in the direction perpendicular to the electric field should lead to a shrinkage in the parallel direction.Then, the droplet is supposed to assume the form of an oblate spheroid.We could infer that the shape variation is associated with the electric stress at the interface of the droplets with the dielectric matrix that may induce a flow with associated hydrodynamic stresses.The sum of the electric and hydrodynamic stresses may determine the deformation of the drop: the total stress at the drop surface can be balanced by interfacial forces through a change in curvature of the interface.Surprisingly, this deformation has not been observed on nematic droplets in the same experimental conditions (Figure S1, Supporting Information).The droplet deformation and, consequently, the fingerprint-like texture formation are strictly linked to the cholesteric phase.Usually, when a CLC is confined between two glass plates, the planar anchoring provides the well-known Grandjean texture.It is known that, when a weak electric field is applied perpendicularly to the cholesteric planar layers, the dielectric torque tends to reorient the layers along the electric field but free rotation is hindered by limiting surfaces.As a result the layers undergo some periodic sinusoidal deformations, called undulations, that strongly depend on the cell thickness/cholesteric pitch ratio. [29]In an oblate ellipsoid type of confinement, the dielectric torque and the planar alignment could cause a 3D formation of undulations that could lead to this peculiar fingerprint-like texture. [30]In general, for droplets with diameters between 40 μm and 70 μm, both distortion and fingerprint texture are observed at about the same voltages.In this work, the size variation is studied varying frequency and intensity of  the applied voltage.Three different experiments are carried out at fixed voltages: 0.3, 0.4 and 0.7V/ μm −1 and the frequency of the applied field is increased from 50Hz to 100 kHz and 1MHz (Figure S2, Supporting Information).Figure 2c summarizes the results obtained, reporting the increase in the imaged surface area as a function of increasing frequencies for the three different voltages.Applying a voltage of 0.3V μm −1 , a distortion is observed at 50Hz.At 0.4V μm −1 , the droplet is distorted at 10Hz.At 0.7V μm −1 , the droplet is immediately distorted.In general, low intensity electric fields need higher frequencies to induce the fingerprint texture, that better develops when the droplet dimension increases dramatically.Every droplet of the emulsion shows a unique fingerprint texture, which complexity depends on the droplet dimension and on cholesteric pitch (Figure 3a).Due to the random nature intrinsic to the creation process, microdroplets can be exploited to create PUF keys.In this perspective different issues have to be considered: droplets' dimension, cholesteric pitch, and texture stability.
To obtain droplets with an optical pattern that can be clearly imaged using a low-resolution microscope, or even a smartphone camera, they must have a diameter of several tens of microns and the distance between crests must be bigger than 1.5 μm.This last request can be easily achieved by varying the composition of the chiral mixture.The distance between two adjacent crests is related to the cholesteric pitch that, in turn, is related to the amount of chiral dopant.To increase the distance, a decreased amount of the dopant is necessary.We have prepared a CLC mixture, with a helical pitch of 3.5 μm.Applying an electric field at 0.7V μm −1 and 1MHz the fingerprint-like texture is obtained, using a smaller amount of chiral dopant helps to better visualize the features of the texture, as shown in Figure 3b.The droplet in Figure 4 has a diameter of about 150 μm.In general, the droplets diameter can be increased or decreased by changing the shaking frequency used during the emulsion preparation.Low frequencies produce large droplets.The droplet texture relaxes in a distorted radial texture when the electric field is switched off.For practical applications, the fingerprint texture has to be fixed and this can be obtained by doping the CLC mixture with a monomer and a photoinitiator.[33] Irradiating the emulsion with a 100 W Mercury lamp, with the electric field on, fixes the finger-print texture on each microdroplet.After 2 min the UV lamp (HG 100 AS) and the electric field can be switched off and the fingerprint remains stable.The presence of monomer and photoinitiator does not affect the electric field parameters necessary to obtain the fingerprint structure.Once microspheres are polymerized, the textures are preserved for months (Figure S3, Supporting Information).Compared to human fingerprints, cholesteric optical fingerprints are richer in minutiae, that are randomly distributed and lack the alignment and local orientation shown by human ones.Figure 4a shows a characteristic of human fingerprints known as "delta" that can also be found in a CLC fingerprint.Using short pitch cholesterics and producing large spheres can lead to an increased encoding capability of each single droplet that could be further enhanced using, for example, fluorescent dyes or nanoparticles with specific properties or using more complex materials as native chiral LCs or polymer and nanoparticlesstabilized blue phases. [34,35]Then, each fingerprint texture inside a microsphere, even if obtained through a precise protocol, has unique and unpredictable features.This randomness makes the fingerprint texture impossible to clone.For these reasons, the texture inside the single microsphere can be exploited as a PUF.It is possible to obtain simultaneously a high number of microfingerprints in the same sample allowing to produce extremely complex PUF keys.The proposed procedure can be used as a method to create an artificial fingerprints database (Figure S4, Supporting Information).This could represent an important step in the fight against counterfeiting of individuals sensitive data.

Statistics
To verify the uniqueness of artificial microfingerprints, we have examined 62 different patterns.A software app is developed for this purpose.A software platform based on Computer Vision tools is designed and exploited to analyze the images.This application allows to identify the unique patterns and features contained in fingerprint images.The recognition relies on the Scale Invariant Feature Transform (SIFT) algorithm, [36] which enables pattern recognition on 2D-images regardless of brightness changing, scale, and acquisition angle.Basically, the algorithm executes a set of predefined steps: first of all, it constructs a scale space to identify the points that form the image independent of the scale, then it localizes the key points of the image with appropriate features.After that, it assigns the orientation to the key points to guarantee their rotation-invariant and it describes all key points with exclusive characteristics to assure their uniqueness.Each single image is analyzed to eliminate key points that show a similarity higher than 75% within the same image.This step is necessary to avoid that too common key points, that repeat themselves often, could produce a false positive during the verification procedure.Usually, the number of these points is around 5% and the number of key points left in each image is typically around 300.The correspondence among key points between different images can be evaluated, using this algorithm.A Java tool based on software libraries implementing the SIFT algorithm is exploited to manage the images.During the process of comparing two different images, each point of the first pattern is sought within the second one, if it is found the point counter is incremented.The same procedure is applied to the 62 images that allows to build a confusion matrix that shows on its diagonal the number of matched key points when each images is compared with itself.In Figure 5a confusion matrix calculated comparing ten different patterns is shown.
Figure 6a shows the number of matched key points obtained when comparing all the patterns with themselves, while Figure 6b shows the number of recognized points when comparing different patterns with each other that is, in the majority of cases, less than five.This analysis points to the uniqueness of each pattern and also to the fact that the probability to produce a second pattern that possesses at least the same 100 characteristic points of the first one using the same random fabrication procedure, from a statistical point of view, is extremely low.

Anticounterfeiting Label
Microfingerprints can also be used to create complex anticounterfeiting labels.For this purpose, a procedure has been developed: the generated microfingerprints patterns are imaged and transposed on a scale suitable to be acquired with a low-resolution camera.The process consists in reproducing  fingerprints on a flexible substrate using electroluminescent ink to enhance the visibility of the pattern.In this type of device, a layer of electroluminescent materialis sandwiched between two conductive layers.When current flows, the electroluminescent material emits light.The procedure used for creating electroluminescent labels, is based on the subsequent deposition of the different layers using the screen-printing technique: [37,38] starting from a substrate where the conductive layer is printed first, then the electroluminescent layer, the dielectric one and finally electrodes.Initially the microfingerprint pattern is imaged using CLSM (Figure 7a).Then, the image is resized (7cm x 7cm), its resolution is increased to 1016 dpi (Figure 7b) and it is converted in grey scale.The final image is printed using a photoplotter which impresses it on a film that will be used as substrate or "fingerprint layer" (Figure 7c).Once the substrate is ready, the different inks are deposed as schematically represented in Figure 7d: each layer is deposited in a single step, followed by a curing process (Figures S5-S7, Supporting Information).The label is intrinsically resistant to temperature variations up to 110°C due to its fabrication procedure and it has also shown to be mechanically robust (Figure S8, Supporting Information).The label works perfectly when bent using different cylinders with decreasing diameters and also when completely folded on itself.Once the label is ready, it can be connected to a proper circuit (Figure 7e).The pattern is visualized with the "fingerprint layer" face up.To switch on the electroluminescent label an AC generator is connected to the system.Figure 7f shows the off (left) and on (right) states of the label.Once the labels are created, a preliminary authentication test is performed on two printed labels containing two different textures.Initially, each label is imaged using a smartphone and pictures are stored in a database.When the label is used in a real life experiment, the authentication procedure is performed through a smartphone, that acquires a picture of the used label, and the software, that compares the acquired image with the ones stored in the database (Figure 8I).
Each label is imaged at different angles, then images are compared through the software, described in the previous paragraph, in order to test if they are recognizable as similar (Figure 8   We will distinguish the two labels as first tag (Figure 8 IIa-d) and second tag (Figure 8 IIe-h).Given two images of the same label at different rotation angles, the software recognizes and associates the characteristic-key points connecting them with a colored line.The more lines there are, the more characteristic points are associated, the greater the certainty that it is the same image.In presence of many lines the matching is to be considered good (Figure 8 IIIa-d).The absence of lines corresponds to a mismatch between the analyzed tags, as shown in Figure 8 IIIe-h.In Figure 8 IV the graph shows the percentage related to the number of characteristic recognized key-points obtained when an image is compared to itself (blue dots) and to a different image (light-blue dots) at different angles.The recognition percentage is very good for each different rotation angle.When the image is compared with itself and no rotation is applied to the images, the recognition is total (100%).When images acquired from the same label but rotated are compared, recognition percentage is never lower than 40%.In case of different images, the recognition percentage is close to 0%.From this preliminary test, the performances of both the label and the smartphone based authentication system are promising in view of possible applications.

Conclusion
Chiral LCs, when confined in curved geometries exhibit a plethora of interesting optical textures.Here, we report on how a peculiar fingerprint-like optical texture can be obtained by applying an intense electric field oscillating at high frequency to an emulsion containing chiral liquid crystalline microdroplets.Chiral LCs microspheres exhibit a planar orientation at the interface that, in the absence of the electric field, favors the formation of the typical Frank-Pryce pattern.Once the voltage is applied, a complex topology is observed that is different in each microsphere of the emulsion.The complexity of the texture depends on the microspheres' dimension and on the pitch of the cholesteric helix but, as for fingerprint textures observed in planar geometry, the optical pattern is randomly generated.Each pattern can be used as a PUF key.In general, the texture formation is linked to a deformation on the microspheres that assume an oblate ellipsoid form and it is not observed in droplets containing a plain nematic liquid crystal that undergo the same experimental conditions.The origin of the phenomenon is not clear and it could be related to electrohydrodynamic effects.Overall, the droplet texture is metastable and it relaxes back into a distorted radial one once the electric field is switched off.The texture can be stabilized by photopolymerizing the chiral liquid crystal.Adding a monomer and a photo initiator to the liquid crystalline mixture does not change the quality of the observed textures that can be preserved for months.Since an emulsion contains different microspheres each with a complex fingerprint-like optical pattern different from the others, they can be imaged with a high-resolution technique as laser scanning confocal microscopy and used to generate a database of artificial fingerprints.Further, these patterns can be used to create anti-counterfeiting labels extremely complex to reproduce.CLSM images can be rescaled and used to create easy to read labels through screen printing techniques.Electroluminescent labels, easy to be read even in low ambient light conditions, are proposed together with a userfriendly authentication system based on the use of a smartphone.The combination of a complex fingerprint-like pattern, a robust electroluminescent label and a specially developed software app could make the proposed technology attractive for real-life applications.
Emulsions: Emulsions were prepared adding a small quantity of the dye doped chiral mixture, about (1%wt.)inside a glycerol matrix, at room temperature.The glass vial (diameter 1.0cm, height 2.5cm) containing the glycerol and a chiral LC mixture was, then, subjected to a shaking process at 20Hz for 40s, at a temperature of 40°C in a laboratory vortex mixer.Droplets had a diameter comprised in a range between few microns and tens of microns.
Optical Cells: To analyze the optical properties of chiral mixtures and emulsions, different optical cells of appropriate thickness were prepared.Two glasses covered by a thin layer of conductive material, indium tin oxide (ITO), were assembled using 200 μm thick mylar as spacer.Cells were sealed along the longer edge using gel glue (UHU), and left to dry overnight.Two tiny electric wires were, finally, welded on the glasses.
Experimental Techniques: The electric field was applied using a high frequency alternating voltage generator (Agilent), coupled with an electric signal amplifier (Krohn-Hite 7500).Images of the microdroplets texture were acquired using a polarized light microscope (DMRX, Leica) and a confocal laser scanning microscope (TCS-SP8, Leica).Data were analyzed using OriginPro 8 (OriginLab Corporation).
Printing Procedure: Screen-printing inks, Conductive EL 7080, Ciano fluo dye EL 7002AC, Dielectric EL 7030D, Silver ink CP 6662 were pur-chased from Bectron.Their description, curing procedure and benefits were summarized in Table S1 (Supporting Information).For the curing process a UVA lamp was used, a Elmi B2-UV professional bromograph with 12 lamps of 20W each and an emission wavelength of 350nm.A Maishi Monofilament Polyester Printing Mesh (Bolting Cloth) was used, mesh specifications were grouped in Table S1 (Supporting Information).A photoplotter FP-8000/8000 XL and a photographic DigiDot far red film (Alliance DigiDot HND), with a spectral sensitivity in the wavelength range of a HeNe laser and a red laserdiode (630-670nm), were used in this work.An AC generator (Inverter -KPTA-0144-18) with a Vpp = 400-600 V, f = 2-4 kHz is used to drive the electroluminescent label.

Figure 1 .
Figure 1.a) Single microsphere observed through CLSM and b) microspheres observed through a polarizing optical microscope between crossed polarizers.

Figure 2 .
Figure 2. a) Sketch of the experimental setup used to apply an electric field to the emulsion contained in an optical cell.b) Fingerprint texture formation applying to the emulsion from left to right: 0V, 0.4V μm −1 , 0.6V μm −1 , and 0.7V μm −1 at 1MHz.Scale bar is 25 μm.c) Imaged surface area as a function of the increasing frequency at three different applied voltages: 0.3V μm −1 (black squares), 0.4V μm −1 (red circles) and 0.7V μm −1 (blue triangles).

Figure 3 .
Figure 3. a) CLC droplets exhibiting fingerprint textures (the size of the pitch is 2.5 μm); b) Confocal microscopy image of a droplet prepared using a CLC with large pitch (the size of the pitch is 3.5 μm).

Figure 4 .
Figure 4. Comparison between (a) a human fingerprint and (b) a CLC fingerprint.

Figure 5 .
Figure 5. Confusion matrix calculated on ten different fingerprint-like patterns.

Figure 6 .
Figure 6.a) Number of recognized common key points occurring during the comparison of the same pattern with itself.b) Number of recognized common key points occurring during the comparison of the one pattern with all the others.

Figure 7 .
Figure 7. a) Image acquired using a confocal microscope (scale bar is 25 μm), b) processed image, c) fingerprint image printed on a photosensitive film (scale bar is 2cm).d) Schematic rappresentation of the screen printing process.e) Label created with ambient the screen-printing technique with light switched off and f) applying an AC voltage (scale bar is 2cm). II).

Figure 8 .
Figure 8. I: The sketch shows the authentication procedure of merchandise labeled with the proposed anti-counterfeiting label.During the authentication, the consumer can analyze the label by using a smartphone.All the data are stored in a secure server.II: Photographic fingerprint images acquired at different angles: 0°-90°-180°-270°.First tag from (a-d), second tag from (e-h) (scale bar is 0.5cm).III: Software recognition procedure: comparing couples of images shown in II: first tag with themselves, (a) a-a ; (b) a-b ; (c) a-c ; (d) a-d ; first tag with the second tag (IV): (e) a-e ; (f) a-f; (g) a-g; (h) a-h.V: Percentage of features recognized while rotating the sample: Image at 0°of the first tag is compared with the images acquired at different rotation angles for the same tag and for the second tag.