Inkjet Printed Photonic Cellulose Nanocrystal Patterns

Naturally‐sourced cellulose nanocrystals (CNCs) are elongated, birefringent nanoparticles that can undergo cholesteric self‐assembly in water to produce vibrant, structurally colored films. As such, they are an ideal candidate for use as sustainable and cost‐effective inks in the printing of scalable photonic coatings and bespoke patterns. However, the small volume and large surface area of a sessile CNC drop typically leads to rapid evaporation, resulting in microfilms with a coffee‐stain‐like morphology and very weak coloration. Here, it is demonstrated that inkjet printing of CNC drops directly through an immiscible oil layer can immediately inhibit water loss, resulting in reduced internal mass flows and greater time for cholesteric self‐assembly. The color of each microfilm is determined by the initial composition of the drop, which can be tuned on‐demand by exploiting the overprinting and coalescence of multiple smaller drops of different inks. This enables the production of multicolored patterns with complex optical behaviors, such as angle‐dependent color and polarization‐selective reflection. Finally, the array can be made responsive to stimuli (e.g., UV light, polar solvent) by the inclusion of a degradable additive. This suite of functional properties promotes inkjet‐printed photonic CNC arrays for smart colorimetric labeling or optical anticounterfeiting applications.


Introduction
Structural color, which arises from the selective interference of light with periodic nanostructures, has attracted considerable interest due to the ability to generate vibrant films and coatings DOI: 10.1002/adma.202307563 with a complex optical response that can depend on the angle of illumination (e.g., iridescence), light polarization (e.g., circular dichroism), or external stimuli (e.g., mechanical deformation, temperature, humidity).[7] Upon drying, the relative orientations of the CNCs within this mesophase can be preserved, leading to a helicoidal nanoarchitecture in the solid state.Crucially for their application as a photonic material, the periodicity of this helicoid (referred to as the pitch, p) is typically on the order of the wavelengths of visible light, giving rise to vibrant, structural color.This cholesteric self-assembly process can be tuned by a range of methods, with common approaches to alter the final color including: tip sonication (which breaks apart the crystallite bundles that act as colloidal chiral dopants [8] ), addition of an electrolyte (e.g., NaCl, which screens the electrostatic repulsion between CNCs [9] ) or heat-induced desulfation (which reduces CNC surface charge [10] ).Alternatively, the replacement of water with a nonvolatile hydrophilic additive can lead to a redshifted film, [11,12] attributed to the reduced compression experienced by a kinetically arrested cholesteric domain upon complete loss of water. [7]hile the cholesteric self-assembly of CNCs into photonic films and their resultant optical properties have been extensively studied at laboratory scale, the development of photonic CNC coatings toward real-world applications necessitates moving beyond small-scale or batch-based casting methods.For example, while the dimensions of a dish-cast film are typically defined by the geometry of the container, meter-scale films can be continuously cast using Roll-to-Roll deposition. [13][17][18][19] However, this approach cannot be directly translated to dilute CNC suspensions, as the rapid drying of the picoliter-scale drops produced by ) schematic of the printing process; whereby: i) actuation of a solenoid within the translatable printhead dispenses a drop of aqueous CNC suspension, ii) the drops pass through a layer of dodecane oil and wet onto a glass substrate, iii) within the pinned drop the cholesteric domains merge and align to the substrate surface, iv) subsequent loss of water (through the oil layer) results in a photonic microfilm.The color reflected by each microfilm is principally determined by the initial formulation of the dispensed drop, but is also dependent on the angle and polarization of illumination and/or detection.For scale, the Petri dish in a) has a diameter of 9 cm.c,d) Printed "R G B" text using red, green, and blue inks.The text was printed using nominal drop volumes: c) 100 nL with a 2.4 mm inter-drop spacing and d) 10 nL with a 1.2 mm spacing, such that the font size is consistent.The white dashed circles indicate the microfilms reported in the high magnification images.These were collected in epi-illumination and through either a left-or right-circular polarized filter (LCP and RCP, respectively), confirming polarization-selective reflection.commercial printers does not allow sufficient time for cholesteric self-assembly to occur. [20,21]n general, the self-assembly of CNCs within an evaporating sessile drop is challenging, as pinning of the contact line can lead to the presence of evaporation-induced mass flows (known as the coffee stain effect [22] ) that can disrupt cholesteric ordering. [23,24]hese radial capillary flows can be suppressed by either: i) increasing the CNC suspension viscosity (e.g., via a thickener such as glucose, [23] or inducing gelation with electrolytes [25] ), which slows deposition at the contact line, [26] or ii) inducing a Marangoni counter-flow at the liquid-air interface to balance the outward mass transfer (e.g., by using two solvents of different volatility and surface tension [27] ).However, such disturbances typically lead to a polydomain texture and weak structural coloration in the resultant films.Conversely, pinning of the con-tact line can be avoided by reducing the surface tension via addition of a nonionic surfactant, which leads to more uniform, but significantly redshifted, films. [28]In our previous study, we instead proposed drying CNC drops under a layer of immiscible oil, which then undergo uniform water loss via diffusion across the oil-water interface and thus induce negligible radial capillary flow. [29]This results in an extremely well-aligned cholesteric architecture within the drying drop, leading to the optimum photonic response from the resultant CNC microfilms.
Here, we report the scalable deposition of photonic "dot-matrix" images by direct inkjet printing a cholesteric CNC suspension through a thin layer of immiscible oil (Figure 1a,b).By this method, drops of aqueous CNC suspension are individually dispensed (5-100 nL), which upon drying produce vivid, iridescent structurally colored microfilms ("dots") with diameters ranging from 0.4 to 1.2 mm, respectively.In contrast to our previous study, [29] where large numbers of identical drops were deposited by blade coating a CNC suspension onto a hydrophilic-hydrophobic patterned substrate, here the location, volume, and composition of each individual drop within the printed array is controlled, allowing for the production of full-color dot-matrix patterns and bespoke images (Figure 1c,d).Furthermore, through in situ mixing of different inks within each printed drop, the formulation can be tuned on demand to enable direct control over the final optical appearance.This approach to tuning the color is explored for both electrolyte and polymer additives, with the visual appearance and colorfastness of the different inks assessed.Finally, we demonstrate the features of CNC microfilm arrays that can be relevant for, e.g., anticounterfeit motifs, and propose how a solid varnish could be used instead of oil to unlock high-throughput printing.

Results and Discussion
To explore the viability of inkjet printing as a scalable method to deposit photonic CNC microfilm arrays, it was first necessary to formulate a suitable ink.Using a single aqueous CNC suspension (see the Experimental Section), it was found that a full spectrum of colored films could be produced by varying the ionic strength of the initial formulation within the range [NaCl]/[CNC] = 0-150 μmol g −1 ( Figure S1, Supporting Information).As such, 0, 25, and 55 μmol g −1 of NaCl were added to an 8 wt% CNC suspension to formulate "photonic inks" that would respectively produce red, green, and blue colors once dried.Moreover, to minimize the time required for cholesteric self-organization during the printing process, these relatively highly concentrated suspensions were allowed to phase separate, with the denser, fully-cholesteric phase isolated for further use.To print CNC microfilms, discrete drops of these inks (<1000 nL) were deposited onto a glass substrate using a noncontact solenoid-controlled dispenser (Biodot AD1520, see Figure 1a; and Video S1, Supporting Information), which allowed for precise control over both the x-y position and volume of each dispensed drop.
To understand the importance of the drying process upon the ability for the sessile CNC drops to self-assemble into structurally colored microfilms, three different environments were investigated, as summarized in Figure 2. First, as a control, CNC drops ("green" ink: [NaCl]/[CNC] = 25 μmol g −1 ) with volumes spanning 5-100 nL were printed and allowed to dry in situ within the chamber of the printer at room temperature and elevated humidity (20-25 °C, >80% RH).In this case, the cross-sectional profile of the resultant microfilms showed a significant mass transfer to the periphery, with an associated gradient in color.Moreover, the color intensity was very weak, which was attributed to both the very thin microfilm height (<3 μm) and the radial shear that arises from internal flows within the rapidly evaporating drop (drying time 1-12 min for 5-100 nL, respectively).In a second configuration, the drying time was dramatically increased (approx. 2 days) by increasing the relative humidity postprinting to near saturation (i.e., >99% RH).While smaller microfilms produced under these conditions were still only weakly discernable under the microscope, those produced from larger drops displayed visible structural color, albeit with significant radial color gradients.This suggests that a degree of unavoidable drying occurs during the printing process, which for smaller drops results in near-complete drying before a saturated environment can be attained.In contrast, larger drops do benefit from a reduced drying rate, which offers sufficient time for self-assembly to occur.This transition is consistent with the change in microfilm crosssection, which switches from a typical coffee stain profile to that of a truncated dome for drops ≥25 nL.However, the thinness of even the largest microfilms (approx.3 μm) still results in an overall low color intensity.
Finally, when CNC drops were dried under an immiscible oil layer over several days, much more uniform and vibrant coloration was achieved.The presence of oil changes the principal mechanism of water loss from that of evaporation to diffusion. [29]ince oils typically become saturated by water at very low concentrations, water loss from the CNC drop is dramatically slowed down and thus occurs uniformly across the surface, minimizing the hydrodynamic flows that lead to the coffee stain effect.By optimizing the oil thickness and dispensing speed of the printhead (see the Experimental Section) it is possible to directly print through an oil layer, as demonstrated here with an approximately 1 mm thick layer of dodecane (or food-grade sunflower oil, see Figure S2, Supporting Information).The immediate encapsulation of the sessile drop also allows for the drying rate to be decoupled from the printing process, which enables the production of smaller microfilms with visible structural color or the printing of much larger arrays (see later).Moreover, the presence of oil dampens the force of the impact of the drop on the substrate and resists surface spreading, resulting in drops with a much higher contact angle to the surface.This leads to thicker microfilms with a domed profile and consequently more intense structural coloration for a given initial volume.Finally, the slow drying rate allows for the relaxation of shear and for neighboring domains to orient and merge, leading to monodomain microfilms with visible disclinations across the surface.Notably, for all drying environments and drop volumes, the reflection from the resultant microfilm is almost entirely left-circular polarized (LCP), with the absence of RCP reflectance evidence that the microfilms contain well-ordered cholesteric domains (Figure S3, Supporting Information).Macroscopically, the uniform vertical alignment of the nanostructure within the CNC microfilms arising from planar anchoring results in a highly iridescent reflection that is visible only at specular angles (see Figure S8, Supporting Information).
The microfilms produced from drops dried under dodecane are more uniform in color than those that were dried in air, but also show evidence of surface buckling, with dark folds visible in the optical micrographs (Figure 2a) and sharp dips in the microfilm profile (Figure 2b).This phenomenon can be attributed to initial rapid water loss across the drop surface prior to the local saturation of the oil phase.This fast initial drying first pins the drop, which inhibits lateral contraction upon subsequent drying and second, it leads to the formation of a more rigid "shell" that resists compression upon further volume reduction (as the concentration front propagates inwards from the drop interface).Therefore, when combined with the higher aspect ratio of oildried drops with respect to those dried in air, there is an increasing excess of surface which prevents the drops from maintaining a domed shape.Together these effects promotes the formation of wrinkles upon drying, similarly to emulsified CNC droplets. [30]igure 2. The impact of drying conditions upon the morphology and visual appearance of "green" CNC microfilms produced from a range of dispensed drop volumes (V = 5-100 nL).a) Optical micrographs collected through a left-circular polarized filter for microfilms dried: under dodecane oil, in a saturated atmosphere (>99% RH) and at elevated humidity within the chamber of the printer (80% RH).b) Cross-sectional profiles of the microfilms in (a).All plots have the same axes.c) Micro-spectroscopy of the region indicated by the white circle in (a), but collected using a higher magnification objective.The spectra were recorded using either a left or right-circular polarization filter (LCP and RCP, respectively).All plots have the same axes.
This "buckling" of the helicoidal architecture is expected to locally result in tilted domains that will both deflect light away from the collection cone of the microscope objective and also act to depolarize the light, which can explain the small amount of blueshifted RCP reflection and nonoptimal intensity for the thicker microfilms.Such buckling is even more apparent when the CNC drops were printed onto a more hydrophobic substrate, such as polyethylene terephthalate (PET).When compared to analogous drops printed onto glass, the microfilms printed onto PET were much smaller in diameter (Figure S4, Supporting Information), which is attributed to a higher contact angle with the substrate and the absence of pinning.Moreover, the corresponding thicker drop profile and high degree of buckling leads to a significant amount of redshifted, tilted domains (as evidenced by darkfield microscopy), [31] which results in a weaker, but less iridescent, macroscopic visual appearance.
Printing CNC drops through an oil layer offers the best optical performance in the resultant microfilms, while also decoupling the printing step from the self-assembly process.As such, this method was chosen to explore the printing of patterned arrays, as demonstrated by three-color printing of the text "R G B" in Figure 1c,d.This image was printed as a single automated process by sequentially dispensing one of the three photonic inks detailed above for each letter and was repeated at two different resolutions: i) 7 × 24 array of 100 nL drops with a center-to-center spacing of 2.4 mm (i.e., 11 DPI), and ii) 14 × 48 array of 10 nL drops with a center-to-center spacing of 1.2 mm (i.e., 22 DPI).For all three inks, the final microfilms were comparable in size for a given drop volume, with those from 100 nL drops approximately 5x larger in area than for 10 nL drops.As such, despite doubling the resolution, the printed area of the glass substrate is comparable in both cases.However, the higher intensity of larger microfilms (arising from their greater thickness) results in an overall more vibrant image for the 100 nL sample.This demonstrates the trade-off between increasing resolution and decreasing visibility that arises from printed photonic structures, where for thin films the intensity scales quadratically with the number of cholesteric pitch repeats. [7,32]It is noteworthy that even this low DPI is comparable to that of dot matrix printers, or commonly used in large-scale signage, such as billboards.Additionally, it is apparent that the smaller drops are more severely impacted by printing through an oil layer with a greater degree of misalignment due to deflection by the air-liquid interface.This was attributed to a combination of the thickness of the oil layer relative to the drop volume and the presence of surface ripples caused by the rapid printing of adjacent drops within a row.This issue can be mitigated by printing drops as several layers of staggered arrays (as later employed for the printing of larger images) or by optimization of the thickness of the oil layer to the specific drop volume.
Inkjet printers typically achieve a full gamut of colors by using a sufficiently high resolution such that individual dots are no longer distinguishable by eye for a typical viewing distance (e.g., >150 DPI for documents).This results in the perceived color from a given area being an average of all the dots of each distinct ink.However, for printed dot-matrix images, where each individual dot is clearly resolved by eye, it is necessary to instead tune the CNC formulation at the single drop level to expand from "RGB" to full-color printing.This was achieved through in situ mixing, whereby a larger drop was produced from the overprinting and coalescence of multiple smaller drops.This is exemplified in Figure 3a, where a rainbow of colored microfilms was created by overprinting each location with 10 × 10 nL drops and systematically varying the ratio of either "red" or "blue" ink drops (i.e., respectively [NaCl]/[CNC] = 0 and 55 μmol g −1 ).Note that for this coalescence approach, a well-mixed biphasic CNC suspension was employed, such that the concentration and size distribution of the CNC suspension remains constant across the array of drops irrespective of the specific combination of inks.Microfilms dried from coalesced drops have the same overall color as expected from those printed from a single drop, suggesting that there is sufficient diffusion of electrolytes within the final drop during self-assembly.However, microscopically it is apparent that they no longer exhibit buckling (Figures 1c vs 3a), but instead display an increase in the number of concentric surface defects, which could be explained by in situ phase separation of the biphasic suspension during drying (rather than the existence of a single interfacially pinned domain), combined with the presence of a small number of trapped oil droplets during overprinting that distort the oil-water interface.
Full-spectrum color tuning of CNC microfilms was also achieved with nonvolatile additives, such as the hydrophilic polymers: poly(ethylene glycol) and hydroxypropyl cellulose (respectively PEG and HPC). [11,33,34]As shown in Figure 3b,c, incorporation of either additive leads to a redshifted color (i.e., a larger final pitch), which is attributed to a reduction in the amount of compression the cholesteric domains experience upon drying.As such, biasing the drop ratio toward the additive-containing ink results in a smooth gradient from blue to red.However, the microscopic appearance of the two series is quite different.The incorporation of up to 25 wt% of PEG in suspension results in comparable coloration to those tuned by electrolytes, with the extent of the redshift consistent with a simple geometric expression for the change in pitch as a function of additive volume fraction (Figures S5 and S6, Supporting Information).In contrast, microfilms containing over 14 wt% of HPC display multiple blueshifted domains in the center, with multiple peaks also detected by microspectroscopy.This suggests that the much higher molecular weight of the HPC results in it acting as a depletant, with HPC accumulating at the domain boundaries where it can inhibit the fusion of tactoids into larger domains.The increasing coexistence of HPC-rich and HPC-poor domains also explains why the reflectance peaks progressively deviate away from that expected by the geometric model, with a greater loading of HPC required to achieve the same macroscopic redshift as observed for PEG.Finally, to demonstrate the scalability of the overprinting approach, a polychromatic picture of a flower was printed using just blue ([NaCl] = 55 μmol g −1 ) and red photonic inks ([NaCl] = 55 μmol g −1 & [PEG] = 25 wt%).As shown in Figure 4a, each position in the array was first encoded into the ratio of these inks required to produce the desired color upon drying.This was then realized in a single, automated process by overprinting multiple differently patterned layers of drops for each ink.
While all three methods give a similar initial appearance, the composition of the microfilms enables a controllable response to post-treatments.For example, pure CNC films are resistant to photobleaching, as the color is intrinsic to the nanostructure of chemically-stable cellulose, but this is not the always the case when other additives are incorporated. [35]To investigate the colorfastness of the printed microfilm arrays, the long-term environmental exposure was mimicked using an accelerated weathering system.In this test, the samples were exposed to 200 h of broadband UV light at 50 °C, which corresponds to a blue wool rating of 6-7.As shown in Figure 3a,b, the series of microfilms tuned with either electrolyte or HPC did not show substantial color change, with a mild blueshift observed across all microfilms (e.g., Δ ≈ 10-15 nm for the 5:5 ratio microfilm), suggesting a small decrease in residual water content upon baking at elevated temperature (corresponding microspectroscopy is reported in Figures S5 and S6, Supporting Information).In contrast, microfilms containing PEG showed a much larger blueshift, resulting in a blue-cyan coloration across the series (Figure 3c).This was attributed to the irreversible photooxidation of PEG into volatile components (e.g., formates [36] ), the loss of which results in vertical collapse of the microfilm and thus compression of the helicoidal pitch.This effect can be particularly striking for films with different compositions, but similar initial coloration.As shown in Figure 3d, a series of red films were prepared by combining an increasing gradient in electrolyte (i.e., blueshift) with an increasing gradient in additive (i.e., redshift).Upon UV exposure, the degradation of PEG revealed the underlying pitch gradient imparted by the electrolyte during cholesteric self-assembly, which results in a spectrum of colored microfilms.Importantly, this series demonstrates that as the cellulose itself is not degraded, there is no associated reduction in color intensity, i.e., the loss of PEG does not cause the microfilms to fade (Figure S5, Supporting Information).
A similar visual effect can also be achieved by washing the microfilms with a nonaqueous polar solvent.).The photographs in (c) and (d) were recorded under specular illumination.e) The rapid colorimetric response of patterned CNC microfilms to increasing relative humidity (as achieved using a series of saturated salt solutions).
Notably, in contrast to UV exposure, washing with ethanol can remove both HPC and PEG.This sensitivity to polar solvents is exploited in Figure 4b, where comparable "red" arrays were printed that upon subsequent exposure to ethanol revealed the text "C N C" in different color combinations.As such, by tuning both the initial composition and the nature of the post-treatment, information can be encoded into the printed CNC arrays.This behavior can be further combined with the polarization-selective reflection (Figure 4c; and Video S2, Supporting Information) and angle-dependent hue (Figure 4d; and Figure S8, Supporting Information) that is inherent to well-aligned CNC microfilms, [7] or the real-time colorimetric response to humidity that arises from their thin dimensions (Figure 4e; and Video S3, Supporting Information). [29]Together, this suite of functional properties promotes photonic CNC arrays as being highly relevant for smart labeling or anticounterfeit applications.
The slow concentration rate typically required for lyotropic self-assembly of CNCs is often considered a fundamental limitation in the scale-up of these photonic materials. [13]In the above studies, the CNC drops were allowed to dry slowly under a relatively thick oil layer at ambient conditions to allow for direct comparison between samples.However, to probe the limits of cholesteric self-assembly, CNC drops were also dried under a much thinner oil layer and at elevated temperature (T = 20-98 °C).As reported in Figure S9 (Supporting Information), it was found that reducing the drying time to 2 h still allowed for the production of vibrant microfilms, albeit with a strong redshift in their appearance.This can be attributed to the lyotropic suspension no longer being able to maintain an equilibrium pitch upon drying (i.e., the time scale of pitch relaxation is outpaced by the rate of concentration increase). [7,13,37]Accelerating the drying process further allowed for colored microfilms to be produced in as little as 20 min, but with reduced color uniformity.This can be rationalized by considering that at these rapid drying rates, the reemergence of a gradient of water loss across the drop cross-section can introduce convective mass flows (cf.drying in air).
Although drying under oil over several hours is sufficient for a well-aligned cholesteric architecture to self-assemble within the printed drops, it is less practical for real-world printing scenarios.As such, we explored whether a solid varnish could instead be employed as a semipermeable coating.This would allow for the rate of water loss from the CNC drops to be reduced, mimicking the role of the oil, while also enabling handling or processing of the sample directly after printing.To validate this approach, drops of a biphasic CNC suspension (100 nL, [NaCl] = 55 μmol g −1 ) were printed through a 15 wt% solution of cellulose acetate butyrate (CAB 551-0.01) in n-propyl propionate.Upon loss of organic solvent (approx.5 min) a solid cellulosic layer formed above the still wet CNC drops.Beneath this transparent varnish the drops continued to slowly dry over several hours to form colored microfilms.As shown in Figure 5a, this resulted in microfilms that are similar to those produced when dried in a saturated environment, in terms of both a radial color gradient and overall reflectivity (cf. Figure 2a).This appearance can be rationalized by considering that initially there is substantial water loss into the surrounding n-propyl propionate (resulting in radial mass flows).However, once this solvent is lost, the hydrophobic varnish significantly slows down further water loss from the drops, allowing sufficient time for cholesteric selfassembly to occur.Finally, to demonstrate the scalability of inkjetprinted CNC microfilm arrays, this approach was translated to the deposition of a commercial CNC source (University of Maine) using an industrial inkjet printer (Domino C-Series Macrojet).As shown in Figure 5b, arrays of CNC drops could still be successfully deposited, despite the relatively high viscosity of this CNC suspension (13.6 cP) compared to the typical commercial aqueous inks formulated for this printer (<3 cP).Moreover, when these drops were immediately coated after printing with the same cellulosic varnish, the rate of water loss could again be reduced resulting in improved cholesteric self-assembly over equivalent drops dried directly in air, leading to visible structural coloration.

Summary
In summary, we report that polychromatic arrays of CNC microfilms can be produced by inkjet printing.The optical response of each microfilm is principally determined by the drying process, with rapid drying (i.e., seconds-to-minutes) leading to negligible reflectivity, while slower drying, (i.e., minutes-to-hours, e.g., by elevated humidity) results in microfilms with visible radial color shifts.In contrast, drying under an oil layer suppresses internal flows within the drop, resulting in uniform microfilms with coloration that is tunable via the formulation and reflectivity that correlates with the drop volume.This approach also had the advantage of decoupling the drying and printing steps, allowing for larger arrays to be printed.In comparison to colloidal photonic crystals, where the color is intrinsically linked to the precise dimensions of the individual nanoparticles, the inherent tunability of the cholesteric self-assembly process enables full color printing from only a pair of CNC suspensions.Moreover, by mixing different color-tuning methods, arrays with a similar appearance can have quite different responses to stimuli (e.g., light, humidity), which suggests relevance for smart labeling or anticounterfeit packaging.
Finally, as an oil layer could be considered impractical for highthroughput applications, a solid varnish was explored as an alternative method to impede water loss from the drops.This proof-ofconcept showed that a touchable surface could be formed within minutes, while the self-assembly within the encapsulated CNC drops could still occur over several hours.As such, with further optimization of the ink (e.g., size fractionation to access anisotropic suspensions at lower concentrations) [38] and the varnish formulation (e.g., more hydrophobic solvent to minimise the initial rate of water loss), it is envisaged that high-quality, full-color photonic patterns could be achieved that are compatible with the stringent processability requirements of commercial printing (e.g., packaging labels).

Experimental Section
Cellulose Nanocrystal Suspension: A colloidal CNC suspension was produced by the hydrolysis of cotton-derived filter paper (Whatman No. 1, 60 g) with sulfuric acid (64 wt%, 420 mL) at 64 °C for a duration of 31 min under high mechanical stirring, upon which it was quenched by rapid dilution with ice-cold deionized water (Millipore Milli-Q Synergy UV).The acidic supernatant, containing soluble cellulose residues, was removed by centrifugation (3 × 20 000 g), followed by dialysis against deionized water (regenerated cellulose dialysis tubing, MWCO 12-14 kDa), to afford a stable CNC suspension ([CNC] = 2.95 wt%, pH < 3).Conductometric titration against sodium hydroxide (0.01 m) indicated [H + ] (1) = [─OSO 3 − ] = 237 mmol kg −1 of CNC.The CNC suspension was tip-sonicated in aliquots of 41 g in an ice bath (Fisherbrand Ultrasonic disintegrator 500 W, amplitude = 30% max, tip diameter = 12.7 mm, duration = 17 s.This corresponds to an approximate delivered power [8] of 20 W and sonication dose 8.3 J mL −1 ).The suspension was then vacuum filtered twice (MFTM nitrocellulose filter membrane, 8.0 μm pore size) and concentrated by rotary evaporation at 45 °C.This resulted in a 9.72 wt% CNC suspension with 236.6 mmol kg −1 of sulfate half-ester surface charges.At this concentration, the CNC suspension self-organized into a biphasic cholesteric liquid crystal (77 vol% anisotropic phase).Imaging of individual CNC nanorods by atomic force microscopy and the conductometric titration curve are, respectively, reported in Figures S10 and S11 of the Supporting Information.Printed inks were prepared from this stock CNC suspension by dilution with deionized water and sodium chloride (0.1 m) to give a concentration of [CNC] = 8.0 wt%.The final color was controlled via the ionic strength, within the range [NaCl]/[CNC] = 0-150 μmol g −1 .Typically, prior to printing the CNC suspension was left to phase separate for 3-5 days and the anisotropic phase isolated for further use.Alternatively the color of the dried microfilms could be tuned by formulating with hydrophilic polymers poly(ethylene glycol) (Fluka 8000: M r = 7-9 kDa as re-ported by manufacturer) or hydroxypropyl cellulose (Nisso SSL-SFP: M w = 40 kDa as reported by manufacturer), which both act as nonvolatile additives.The commercial CNC suspension was purchased from the Process Development Center of the University of Maine.The as-received suspension was provided pH-neutralized (i.e., negatively charged CNCs due to ─OSO3 − groups, with Na + counter-ions).The measured concentration of the suspension was 10.5 wt% (A&D, MX-50 moisture analyzer).This was diluted with Milli-Q water to 5.99 wt% and tip-sonicated in aliquots of 40 g in an ice bath (Fisherbrand Ultrasonic disintegrator 500 W, amplitude = 40% max, tip diameter = 12.7 mm, duration = 240 s.This corresponds to an approximate delivered power [8] of 29 W and a sonication dose of 174 J mL −1 ).The final viscosity was measured to be 13.6 cP (Brookfield LVDVII+ viscometer using spindle 0 operating at 60 rpm).Films cast from 2 mL of this suspension in a Petri dish were green-red in coloration.
Inkjet Printing of CNC Suspensions: The CNC suspension was printed using a BioJet noncontact solenoid dispenser (Biodot AD1520 with two BioJet Elite dispense heads and humidity control up to 80% RH).This system enabled on-demand droplet dispensing (2-1000 nL), with precise spatial control using a motorized x-y stage.To print the CNC suspension, the system was primed with an aqueous backing solution (0.05% BioTerge -sodium C14-16 olefin sulfonate).A small volume of the desired CNC suspension was aspirated for each print (<13 μL), with 2-3 μL consumed in the pre-pressurization and predispense steps.A 100 μm ceramic tip was used to print individual CNC droplets with nominal volumes of 5, 10, 25, 50, and 100 nL, while a larger 190 μm tip can be used for larger drops.The corresponding solenoid valve open times are 100, 200, 500, 1000, and 2000 ms, respectively.Two dispensing speeds were used: 4 μL s −1 for droplet volumes of 5-10 nL, and 5 μL s −1 for droplet volumes of 25-100 nL.The chamber was equilibrated at 80% RH prior to printing, with the ambient temperature typically 20-25 °C.The printhead was held at 18.5 mm above the stage during printing, such that the nozzle was 1-2 mm above the oil surface.The center-to-center spacing of the drops was typically 1-2 mm, which was selected to minimize any lateral merging that could occur upon slow drying under oil.The substrate was either a glass microscope slide or a polyethylene terephthalate sheet, which were cleaned immediately prior to use with ethanol or exposure to ozone (NanoBioAnalytics UVC-1014, t = 5 min), respectively.Commercial CNC suspensions were printed using a Domino C-Series Macrojet inkjet printer with a nozzle diameter of 120 μm.
Drying Methods: In the simplest case, the printed sample remained in the atmospheric chamber of the printer until the CNC microfilms were fully dry, with the humidity set to the upper limit of the controller, i.e., 80% RH.For samples dried in a saturated environment, the substrate was placed on top of wet filter paper (7 cm diameter, with 1 g of deionized water added) within an open Petri dish (9 cm diameter).The sample was printed at ≈80% RH and then the lid placed on the dish as soon as the print was complete.To print through oil, the substrate was placed in a Petri dish and covered with a layer of dodecane (5.4 g, corresponding to ≈1 mm thickness above the substrate).Once the print process was complete, further dodecane was added (6.0 g) to increase the thickness of the oil layer and thus reduce the drying rate.Note that to account for the change in density when printing through food-grade sunflower oil (a natural oil that is a mixture of palmitic, stearic, oleic, and linoleic acids), the amounts were, respectively, recalibrated to 6.9 and 7.2 g so that the thickness of the oil layers were consistent.It was found that if the oil layer was too thin then the drops may not be fully encapsulated, leading to premature or uneven drying.Conversely a thick oil layer could lead to greater splashing (leading to misaligned drops) and entrapment of air bubbles within the CNC drop (leading to disordered or color-shifted microfilms, see Figure S12 of the Supporting Information).Once dry, the majority of the oil could be decanted and reused, with residual oil removed by washing with an apolar solvent (e.g., hexane).For accelerated drying studies, 100 nL CNC drops were printed through dodecane oil as above, but only a further 1.0 g of oil was added to reduce the overall oil thickness while still ensuring the drops were fully encapsulated.After which, they were placed on a preheated hot plate (T = 60-98 °C).For varnish studies, the oil was replaced with a 15 wt% solution of cellulose acetate butyrate (CAB 551-0.01) in n-propyl propionate (viscosity = 9.78 cP).Upon loss of organic solvent (approx.5 min, depending on volume used), this would form a solid cellulosic layer, encapsulating any CNC drops beneath.
Optical Characterization: Polarized optical microscope images of individual microfilms were collected in epi-illumination using a Zeiss Axio Scope A1 microscope with EC Epiplan-Neofluar objective (5x NA0.13) and fitted with a CMOS camera (Eye IDS, UI-3580LE-C-HQ, calibrated with a white diffuser) using a halogen lamp (Zeiss HAL100) as a light source.The reflected light was additionally passed through a quarter-wave plate and linear polarization filter with adjustable mutual orientation, to differentiate left-circular polarized (LCP) and right-circular polarized (RCP) light.To perform microspectroscopy, the microscope was coupled to a spectrometer (Avantes, AvaSpec-HS2048) using an optical fiber mounted in confocal configuration (Avantes, FC-UV600-2-SR, 600 μm core size).Spectra were collected using an EC Epiplan-Apochromat objective (50x NA0.95) and the intensity was normalized to the reflection of a silver mirror in LCP polarization.Large area images of printed arrays were either produced by stitching multiple images collected using a Keyence VHX-7000 digital microscope using a VHX-E100 objective (100x) or using a digital camera (Samsung Galaxy S21 Ultra) and imaged under specular illumination.Circular polarization filtering in Figure 4c; and Video S2 (Supporting Information) was achieved using a pair of nonanaglyph 3D cinema glasses.
Angle Resolved Optical Spectroscopy: A laboratory-built goniometer was used to analyze the angular optical response of the CNC microfilms.Broadband illumination from a xenon lamp (Ocean Optics HPX-2000) was coupled into a reflective collimator (Thorlabs RC08SMA-F01) using an optical fiber (50 μm core, Avantes FC-UV050-2), producing a spot size of approx.5 mm.The sample was mounted on a motorized rotation stage to determine the angle of illumination.The detector consisted of a second collimator on an arm attached to a motorized rotation stage, which coupled the collected light into a spectrometer (Avantes AvaSpec-ULS2048LTEC) via an optical fiber (200 μm core, Avantes FC-UVIR200-2).The recorded light intensity was normalized with respect to a silver mirror (Thorlabs PF10-03-P01).By simultaneously rotating both the sample [−45°→ +45°, defined normal to the surface] and the detector [−90°→ +90°] with respect to the illumination source, the reflectance spectra from the microfilms could be collected under specular conditions.
Profilometry: The cross-sectional profiles of the drops were obtained with a stylus profilometer (Bruker DektakXT) with a 12.5 μm tip radius at 4 mg force, translated with a speed of 50 μm s −1 , resulting in a lateral resolution of 0.17 μm.
Long-Term Color Stability: To assess the colorfastness of the printed arrays they were exposed to an accelerated weathering system (Sevar Bandol Wheel H 400), using a mercury bulb running at a temperature range of 50 °C.The samples were exposed to two 100 h doses, corresponding to a blue wool rating of at least 6.

Figure 1 .
Figure 1.a) Photograph and b) schematic of the printing process; whereby: i) actuation of a solenoid within the translatable printhead dispenses a drop of aqueous CNC suspension, ii) the drops pass through a layer of dodecane oil and wet onto a glass substrate, iii) within the pinned drop the cholesteric domains merge and align to the substrate surface, iv) subsequent loss of water (through the oil layer) results in a photonic microfilm.The color reflected by each microfilm is principally determined by the initial formulation of the dispensed drop, but is also dependent on the angle and polarization of illumination and/or detection.For scale, the Petri dish in a) has a diameter of 9 cm.c,d) Printed "R G B" text using red, green, and blue inks.The text was printed using nominal drop volumes: c) 100 nL with a 2.4 mm inter-drop spacing and d) 10 nL with a 1.2 mm spacing, such that the font size is consistent.The white dashed circles indicate the microfilms reported in the high magnification images.These were collected in epi-illumination and through either a left-or right-circular polarized filter (LCP and RCP, respectively), confirming polarization-selective reflection.

Figure 4 .
Figure 4. a) Schematic illustrating the production of a polychromatic dot-matrix design via in-drop mixing of two CNC inks, with a photograph under specular illumination of the corresponding printed image.For scale, the center-to-center dot spacing is 1.9 mm in this image.b-d) Examples of the optical effects that can be achieved using printed CNC microfilm arrays: b) Patterns can be encrypted into CNC arrays by tuning the relative amount of PEG and NaCl-upon washing the nominally red arrays with ethanol "C N C" in different color combinations is revealed.The strong vertical alignment of the helicoidal nanostructure within such CNC microfilms results in (c) selectivity for LCP reflectionand d) angular-dependent color (i.e., iridescence, following Fergason's law[7] ).The photographs in (c) and (d) were recorded under specular illumination.e) The rapid colorimetric response of patterned CNC microfilms to increasing relative humidity (as achieved using a series of saturated salt solutions).

Figure 5 .
Figure 5. a) Micrograph of a microfilm produced by printing a drop of CNC suspension (V = 100 nL, [NaCl] = 55 μmol g −1 ) through a 15 wt% solution of cellulose acetate butyrate (CAB) in n-propyl propionate (which produces a transparent, solid varnish upon solvent loss).The images were collected in epi-illumination and through either an LCP or RCP filter.b) Arrays of CNC microfilms produced using a commercial CNC suspension (Univ.Maine) and a commercial inkjet printer (Domino C-Series Macrojet) and dried either in air (top) or under a layer of the same CAB varnish (bottom).