Exploiting the Thermotropic Behavior of Hydroxypropyl Cellulose to Produce Edible Photonic Pigments

Hydroxypropyl cellulose (HPC) is a widely commercialized cellulose derivative. While it is typically used as a binder or stabilizer for foods and pharmaceuticals, it can also form a cholesteric liquid crystal in aqueous solution. Moreover, at high HPC concentrations this lyotropic and thermotropic mesophase is known to reflect structural color. However, it remains a challenge to retain this vibrant coloration into the solid state. Herein, by combining the emulsification of a HPC mesophase with drying at elevated temperature, solid microparticles are produced that can reflect color across the visible spectrum, from blue to green and red. This method provides a facile and scalable pathway to fabricate structurally colored, edible pigments, which can displace existing synthetic additives used in a wide range of foods and cosmetics.


DOI: 10.1002/adsu.202200469
Although HPC allows access to different colors by simply altering its concentration in water (approx. 60-70 wt%), this tunability comes at the cost of color stability. [14] In particular, the reflected wavelength will blueshift into the ultraviolet as the mesophase dries, ultimately resulting in transparent films. [15] As such, the photonic applications of HPC have typically focused upon sealing an aqueous mesophase within nonpermeable media, [16][17][18] for use as, e.g., a mechanochromic sensor. [3,17,19,20] However, to enable its use as a colorant it is preferable to employ HPC in the solid state, where the pitch is fixed to reflect a specific color. To date, several methods have been developed to retain the photonic properties of HPC in a film, however these tend to rely on covalent cross-linking of the concentrated mesophase (either via difunctional linkers or through direct functionalization of the HPC). [21,22] Consequently, upon complete loss of solvent, the shrinkage of the cross-linked mesophase can result in a distorted or inhomogeneous cholesteric structure, which alters both the final color and visual appearance. [23,24] Lastly, it is questionable whether the desirable natural properties of HPC are retained upon the introduction of synthetic cross-linking groups, potentially limiting its suitability as a biocompatible and/or edible photonic material. [25] It is known that the equilibrium color of a cholesteric HPC mesophase is dependent upon both its concentration [26,27] and the local temperature, [28,29] with an increase in each parameter resulting in a blueshift [19] or a redshift, [21,30] respectively. In this study, we exploit these orthogonal behaviors to produce solidstate HPC pigments with visible coloration. Specifically, by emulsifying a relatively low concentration of HPC (45 wt%) in oil, followed by loss of water from the microdroplets at elevated temperature (68.5-77.5 °C), we produced visibly red, green, and blue microparticles of pure HPC. We show that arising from their polydomain structure, the reflected color is relatively size invariant, allowing for scalable emulsification methods to be employed. Crucially, a resting procedure is introduced as a method to improve the cholesteric order of the viscous HPC microdroplets, which increases the intensity of the resultant photonic microparticles. Overall, as "photonic pigments" with controlled visual appearance are produced via a scalable method from purely food-grade HPC, they represent a good alternative for the replacement of current synthetic colorants in the food and cosmetics sectors. As such this study contributes to the Hydroxypropyl cellulose (HPC) is a widely commercialized cellulose derivative. While it is typically used as a binder or stabilizer for foods and pharmaceuticals, it can also form a cholesteric liquid crystal in aqueous solution. Moreover, at high HPC concentrations this lyotropic and thermotropic mesophase is known to reflect structural color. However, it remains a challenge to retain this vibrant coloration into the solid state. Herein, by combining the emulsification of a HPC mesophase with drying at elevated temperature, solid microparticles are produced that can reflect color across the visible spectrum, from blue to green and red. This method provides a facile and scalable pathway to fabricate structurally colored, edible pigments, which can displace existing synthetic additives used in a wide range of foods and cosmetics.

Introduction
Hydroxypropyl cellulose (HPC) is a derivative of natural cellulose which is known to be biocompatible, [1] biodegradable, [2] and even edible. [3] The latter has enabled this water-soluble polymer to be employed in a wide range of commercial applications; from a bulking [4] and drug-release agent [5,6] in pharmaceuticals, to a thickener [7] and stabilizer [8] in food. A less widely exploited property of HPC is its ability to organize into a cholesteric (i.e., chiral nematic) liquid crystalline phase in water, which under suitable conditions can lead to selective reflection of visible light, i.e., structural color. [9] By exploiting the lyotropic [10] and thermotropic [11] properties of the cholesteric HPC solution, the periodicity of the underlying helicoidal nanostructure (referred to as the "pitch", p) and therefore the reflected wavelength can be dynamically tuned, enabling access to a full spectrum of colors. As such, HPC is a strong candidate for exploitation in the pursuit of next generation sustainable and biocompatible pigments. [12,13]

Results
In order to understand the effect of drying the HPC mesophase at elevated temperatures, it is necessary to first understand its thermotropic behavior under equilibrium conditions. For a fixed concentration, the reflected color of the HPC mesophase dramatically redshifts with increasing temperature, which can be approximated to a pseudolinear relationship of Δλ/T = 12 nm °C −1 ( Figure S1, Supporting Information). While this thermotropic effect is typically reversible, when combined with evaporation beyond the point of kinetic arrest (estimated as 76 wt%, which corresponds to the complete loss of unbound water [31] ) this redshift can be locked-in, allowing for solid-state HPC to reflect in the visible regime.
To demonstrate this concept, we exploited the thermotropic effect to produce microparticles of HPC with vibrant colors, using a scalable methodology (see schematic in Figure 1). In short, an aqueous HPC solution (45 wt%) was emulsified via planetary centrifugation in paraffin oil to obtain HPC/oil microdroplets. In contrast to previous studies, where HPC is typically formulated directly as a colorful mesophase (60-70 wt%), [9] this lower concentration was selected as it has a sufficiently low enough viscosity to be emulsified, while still being able to form a cholesteric mesophase. Once emulsified, the polydisperse HPC microdroplets were allowed to rest at room temperature for 1 day, during which they settled to the bottom of the oil-filled dish, which impeded water loss at the expense of a degree of coalescence (Video S1, Supporting Information). Importantly, during this time the cholesteric phase within the microdroplets, which is disrupted by shear during the emulsification process, can relax prior to the onset of the drying process. [32] Examining the coalesced microdroplets by cross-polarized optical microscopy revealed that they were birefringent, but evidence of the Maltese cross that is indicative of a radially aligned Frank-Pryce structure was only observed near the droplet interface ( Figure S2, Supporting Information). [33][34][35] As such, this appearance indicates that the microdroplets have a polydomain morphology, with the large diameter and high viscosity inhibiting significant reordering into a monodomain structure. Once rested, the microemulsion was heated at elevated temperature, which accelerated the loss of water from the microdroplets (via diffusion of surfactant micelles and an increased water miscibility in the oil phase), resulting in the formation of solid HPC microparticles with visible color (see photograph in Figure 1). Importantly, since the color of the dried microparticles is primarily determined by the thermotropic properties of HPC, different colors can be achieved by varying the drying temperature alone. As shown in Figure 2a, dispersions of blue, green, and red pigments could be prepared from the same starting HPC microemulsion (45 wt%, 1 day of rest) with small changes in the temperature of the oven, respectively, 68.5 °C, 71.0 °C, and 77.5 °C.
The optical properties of a cholesteric liquid crystal system are dependent on a combination of the cholesteric pitch and the alignment of the domains. To better understand the interplay of pitch and domain orientation in determining the microparticles' optical appearance, the blue, green, and red HPC microparticles were examined under the microscope in epi-illumination (Figure 2b,c). Color is reflected from across the surface of each microparticle, which is consistent with a polydomain structure (in contrast with the single spot of color expected for a wellaligned radial cholesteric droplet under the microscope [36] ). Furthermore, the inhomogeneity in hue within an individual microparticle, as is most obvious for the green sample, suggests that there is also a degree of pitch variation, which is manifested in the broadness of the reflected peaks in Figure 2d. Finally, it is interesting to note that while the microparticles appear angular independent under diffuse illumination, they show iridescence when viewed with a directional light source ( Figure S3, Supporting Information). This appearance is comparable to that reported for similar disordered photonic systems. [25,37] To understand the internal structure, cross-sectional scanning electron microscopy (SEM) was performed on blue, green, and red HPC microparticles and representative images are included in Figure 3. The images at lower magnification (upper row) clearly show the polydomain structure within the microparticle, which is consistent with the optical microscopy reported in Figure 2 and Figure S2 (Supporting Information). The high magnification images (lower row) show the characteristic Bouligand arches of the cholesteric structure. From this the pitch can be measured and correlated with the optical response. The pitch increases from 267 ± 17 nm for the blue sample to 329 ± 19 nm for the green and 376 ± 23 nm for the red, which are in good agreement with Bragg's law (λ = npcosθ), when the refractive index of HPC is taken to be n = 1.49. [38] Adv. Sustainable Syst. 2023, 7, 2200469 The intensity of the reflection from a well-ordered cholesteric domain is dependent on its thickness. Indeed, we observe that the intensity of an encapsulated HPC solution increases over time as the cholesteric domains align and merge, with a maximum recorded after 5 h ( Figure S4, Supporting Information). As such, it is expected that allowing time for the cholesteric phase to evolve within the HPC microdroplet should lead to a more intense coloration. To validate this approach, samples were prepared with the HPC microemulsion allowed to rest at room temperature for different durations (0-2 days) prior to drying at elevated temperature. As shown in Figure 4a-c, while the microparticle color does not significantly vary depending on whether a resting step was implemented, there is a significant enhancement in the intensity upon resting, as confirmed by microspectroscopy (Figure 4i). Interestingly, there is also a clear difference in geometry between microparticles that had been allowed to rest and those that are heated directly after emulsification. SEM micrographs of the dry microparticles (Figure 4d-f and Figure S5, Supporting Information), show that with a resting step the microparticles generally appear as truncated spheres, while without resting they are much flatter (and adhere to the substrate), which is attributed to much greater wetting on the polystyrene Petri dish during drying (see Section 3). As such, while a resting step can double the reflectivity of HPC microparticles with comparable diameter, the change in aspect ratio from a flattened, spherical cap to that of a truncated sphere, makes it challenging to confirm if this is due to improved internal ordering (i.e., larger domains) or the thicker cross-section (i.e., more domains). Finally, we note that an increase in the resting time from 1 day to 2 days did not result in further improvement to the optical performance, which is also consistent with our encapsulation study ( Figure S4, Supporting Information).

Discussion
The presence of visible color in the HPC microparticles arises from the balance during the drying process of the thermotropic redshift upon increasing temperature (i.e., increasing  Corresponding micrographs of individual HPC microparticles recorded in bright-field configuration. The microparticles were removed from the Petri dish prior to imaging. c) Corresponding reflectance spectra for the microparticles indicated with a white dotted circle in (c). d,e) Plots indicating the distribution of the peak reflected wavelength and intensity as a function of the diameter of the HPC microparticles (as defined by the maximum Feret diameter). The color of each datapoint represents the approximate microparticle color, which in turn corresponds to the three drying temperatures.
cholesteric pitch) with the lyotropic blueshift upon increasing concentration (i.e., decreasing cholesteric pitch). By fully drying into the solid-state at elevated temperature, this thermotropic redshift can be kinetically trapped, avoiding the typical issue of pure HPC mesophases appearing colorless once dry (i.e., reflection at ultraviolet wavelengths). By changing the temperature of the drying step, the thermotropic component can be tuned, allowing for different colors to be achieved from a single HPC formulation. While this strategy is successful, the precise relationship between the reflected wavelength and the drying temperature remains difficult to define. In particular, the ability to directly image and understand the evolution of the color as the heated microdroplet dries is complicated by the presence of a lower critical solution temperature (LCST) for all concentrations of HPC in water. While the LCST varies depending on the HPC source and concentration, [27,39] rheological measurements suggest that for our system the LCST is around 60 °C for 45 wt% HPC in water ( Figure S6, Supporting Information). Thus, it is likely that at the processing temperatures used to form the microparticles, the HPC solution is heated past its LCST. In this state, HPC desolvates, resulting in phase separation and a high degree of scattering (giving it a whitish appearance). [40] Although the compositions of the two phases are well-studied, the liquid crystalline behavior of HPC above the LCST is poorly understood. [27,40] Furthermore, due to this broadband scattering dominating the visual appearance, it is not possible to measure the reflection wavelength from the cholesteric structure above the LCST. Indeed, it is only when the water (which causes the scattering) is removed in the final stage of drying that the underlying cholesteric color can again be observed ( Figure S7, Supporting Information).
The color of the HPC mesophase is primarily a function of its concentration and the temperature. As such, understanding the interplay between these two variables is key to predicting the color of the dried microparticles. Yet, due to the presence of the LCST, it is difficult to directly ascertain whether the typical lyotropic and thermotropic scaling laws that govern the HPC pitch apply at these drying temperatures. However, we do observe that when HPC solutions with various starting concentrations were dried at the same temperature, the reflected wavelengths of the produced microparticles remained relatively constant ( Figure S8a, Supporting Information). This implies that the onset of kinetic arrest is independent of the starting concentration, and that the pitch can achieve some degree of thermodynamic equilibration above the LCST. Notably, structural coloration was observed even when the initial concentration was below the threshold for cholesteric phase formation (i.e., 20 wt%, Figure S8c, Supporting Information). This curious behavior suggests that HPC can transition from an isotropic solution into a cholesteric mesophase despite heating above the LCST, although the lower intensity of the resultant microparticles suggests that the reduced time for self-assembly to occur (i.e., upon concentrating to > 40 wt% during the drying step) results in smaller cholesteric domains ( Figure S8b, Supporting Information).
The complex factors governing the microstructure of the resulting HPC microparticles can also be observed in the variation in the optical appearance for microparticles with different  structural parameters. For example, the preparation method inherently produces polydisperse microparticles with sizes ranging from about 50 to 550 µm. The effect of microparticle size on the photonic properties can be explored by comparing the wavelength and intensity of peak reflection against the diameter of the microparticles. As reported in Figure 2d, the reflected color appears largely unaffected by the microparticle diameter. This is reasonable if we consider each microparticle to consist of many domains, each of which is small relative to the overall particle dimensions. Furthermore, the increase in reflected intensity with microparticle diameter is consistent with this structure, given that with an increase in size there would likely be a larger number of domains aligned to the observation direction, with the optical response a convolution of reflections from multiple domains (Figure 2e). It is worthwhile to note that due to the low birefringence of HPC, large domains are required for a strong reflection, with numerical simulations using the Barreman approach suggesting that a minimum of 100 pitch repeats are required for a perfect reflection of right circular-polarized light. To put this into context, for a pitch of 400 nm this corresponds to a 40 µm thick cholesteric domain. [41] However, while the HPC microparticles are typically much larger than this size in diameter, the constituent domains are much smaller, resulting in a submaximal reflection, which is further attenuated by scattering at the domain boundaries and particle interface.
The irregular, polydomain nature of the HPC microparticles is further amplified by any inhomogeneity in the geometry of the dry microparticle. The particles are not spherical ( Figure S9, Supporting Information), and range from truncated spheres for rested microemulsions to highly flattened spherical caps for those dried immediately. To understand these differences, we Adv. Sustainable Syst. 2023, 7, 2200469   Figure 4. a-c) Micrographs of hydroxypropyl cellulose (HPC) microparticles dried at 73.0 °C after resting for a) 0 days, b) 1 day, and c) 2 days. Images were recorded in bright-field configuration and the microparticles were removed from the Petri dish prior to imaging. d-f) SEM images of the dried microparticles corresponding to samples reported respectively in (a-c). g) Bright-field reflectance spectra of the dried microparticles indicated by the white dotted circles in (a-c). h-i) Average peak reflected wavelength and intensity of individual microparticles as a function of the resting time. Each value is an average of the reflectance peaks extracted from 150 microparticles, which were measured from two distinct samples (75 spectra each). The error bars correspond to the standard deviation of these 150 measurements. Note, microparticles were selected with comparable diameters to minimize any effects of size variance.
consider that (i) there is significant coalescence of the initial dense microemulsion as it settles to the bottom of the Petri dish (Video S1, Supporting Information), and (ii) the interfacial tension of the emulsion will decrease at elevated temperature. As such, when the sample was heated directly after emulsification, the quickly sedimenting microdroplets will both coalesce and wet the substrate, leading to a flattened geometry. In contrast, if the microdroplets are allowed to rest prior to heat treatment, coalescence and sedimentation occur at room temperature. As such, the reduced speed of sedimentation under gravity (likely due to the higher viscosity of paraffin oil at low temperatures) combined with a higher interfacial tension reduces the wetting of these microdroplets to the substrate. Furthermore, as drying commences near the droplet interface, water loss during the rest period may promote the formation a more viscous skin at the dropletinterface that can inhibit further wetting upon heating.

Conclusions
In summary, by drying emulsified HPC at elevated temperatures, solid-state microparticles can be produced with visible structural color. It was found that the final hue was solely dependent on the drying temperature, allowing for blue, green, and red pigments to be produced from a single formulation using a simple and scalable method. Furthermore, these "photonic pigments" consist solely of food-grade HPC and thus retain the edibility and biocompatibility of this widely used ingredient. Beneficially, the similar refractive indices of HPC and saccharides, such as glucose, enables our edible pigments to be readily employed in food coatings, as exemplified in Figure S10 (Supporting Information). The ability to produce vibrant color across the visible spectrum is also particularly important, as conventional synthetic colorants are drawing increasing concern over potential side effects and natural blue hues are rare. [42,43] As such HPC-based pigments have significant scope in the food and cosmetic industries as a sustainable and safe alternative.

Experimental Section
Materials: Hydroxypropyl cellulose (dry powder, HPC grade: SSL, M W = 40 000 g mol −1 , as reported by manufacturer) was purchased from Nisso Chemical Europe. Paraffin oil (M W = 338.7 g mol −1 ) and Span 80 surfactant were purchased from Sigma-Aldrich. Water was filtered using a Milli-Q water purification system prior to use. All materials were used as supplied, with no further purification.
Preparation of HPC Microemulsions: Water (5.5 g) was mixed with HPC (4.5 g) using a THINKY ARE-250 planetary centrifuge mixer (2000 rpm, 2 min, 2 repeats) to produce an initial HPC solution at 45 wt%. Span 80 surfactant (sorbitan monooleate) was added into paraffin oil at 2.0 wt% and dispersed as before (2000 rpm, 2 min, 2 repeats). Subsequently, the HPC solution (1.0 g) was added to the solution of Span 80 in paraffin oil (10.0 g) and mixed to form a microemulsion of HPC droplets (2000 rpm, 4.5 min, 2 repeats). This water/oil microemulsion was then poured into a polystyrene Petri dish (35 mm diameter). The Petri dish was covered, sealed with Parafilm and left to rest for the required number of days before transferring into an oven. If no resting was required, the sample was directly placed into the oven. The samples were heated on a stone shelf within the oven at the desired temperature for 2 days. The stone was preheated at the required temperature for at least 3 h to ensure homogeneous heating within the Petri dish and between dishes. For other initial HPC concentrations, the preparation method is the same as above, except the ratio of HPC and water was varied to maintain a total solution mass of 10.0 g. This protocol to produce photonic HPC microparticles can be applied directly to other oils, such as food-grade vegetable oil, as exemplified in Figure S10 (Supporting Information).
Optical Characterization of HPC Microparticles: Optical microscopy (Zeiss, Axio Scope A1) was carried out in bright-field configuration at 20× magnification (Zeiss EC Epiplan-Apochromat 20×/NA 0.6 objective). Images were recorded using a CMOS camera (Pixelink P-LD725C-UT). To record reflectance spectra, the microscope was coupled to a spectrometer (Avantes AvaSpec-HS2048), using a 600 µm core optical fiber (Thorlabs FC-UV600) in the confocal configuration. The spectra were referenced against a standard white diffuser (Labsphere SRS-99-010). For each dataset, two samples for each protocol were prepared, and 75 spectra were recorded for each sample. The prominence of the spectral peak and its corresponding reflective wavelength was determined by fitting the spectrum with an adjusted Gaussian model using a Matlab code.
Scanning Electron Microscopy (SEM): The dried HPC microparticles were washed with n-hexane to remove residual oil and surfactant. The suspension of microparticles in n-hexane was then deposited on a glass cover slip, and the solvent evaporated by mild heating on a hot plate. The coverslip was mounted on aluminum stubs using conductive carbon tape and coated with Pt (10 nm) using a sputter coater (Quorum Q150T ES). SEM was performed using a Mira3 system (TESCAN), operated at 3 kV and a working distance of 3-5 mm. For cross-sectional imaging, the samples were prepared by mechanically fracturing the HPC microparticles using the edge of a glass cover slip.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.