FL particles were synthesised using an emulsion polymerisation based on a receipt detailed by Zupkauskas et al. [42]: 2 mg sodium dodecyl sulphate (SDS; purchased from Sigma-Aldrich and used without further purification) was dissolved in 12 ml of deionised (di-)water. Next, 325 mg of 2,2,3,3,4,4,4-Heptafluorobutyl methacrylate monomer (HFBMA; obtained from Alfa Aesar and used without further purification) was added and emulsified at 70°C under nitrogen atmosphere while stirring at 600 rpm for 1 h. A solution of potassium persulfate (KPS; purchased from Sigma-Aldrich and used without further purification; 7.75 mg dissolved in 500 μL deionised water) was injected into the mixture while stirring at reduced speed. The reaction continued for 18 h at 70°C. The synthesised particles were purified by dialysis and then stored in a refrigerator. Negatively charged polystyrene (PS) particles used here were obtained from Cambridge Bespoke Colloids LLC.

The FL particles were imaged using a Zeiss scanning electron microscope (SIGMA VP field emission SEM), while the PS particles were imaged with a transmission electron microscope (TEM, TECNAI 20 from FEI). Size analysis based on the SEM measurements was carried out using ImageJ software by calculating the diameter of 100 particles.

We used a Malvern ZetaSizer Nano ZSP to measure the particle size and size-distribution. In order to determine the Zeta potential, we dispersed the particles in 10 mM phosphate buffer, while all other results were obtained from suspensions in deionized water.

The transmission measurements were taken using a Cary UV-Vis spectrometer. A cuvette whose path length is 1 cm was filled with the sample and in-line transmission results were recorded starting with the sample of 40% concentration. Measurements at lower concentrations were taken by diluting the sample.

The reflectivity measurements were taken using a reflection-probe bundle equipped with different LED light sources and a spectrometer. Illumination and collection were done at normal angle. The green and red samples were illuminated with a fibre-coupled white LED (Thorlabs MCWHF2) and the blue sample was illuminated with a cyan LED emitting at 490 nm (Thorlabs M490F3). The samples were illuminated by using a collimator to provide unidirectional lighting. The spectra were measured using a Thorlabs CCS 100 spectrometer. Reference measurements were taken using a Thorlabs BB1-E02 dielectric mirror. The angle-dependent scattering measurements were taken by illuminating the samples at normal angle and changing the angle of the optical fiber collecting the reflection.

SAXS measurements were carried out at the ID02 beamline of the European Synchrotron Radiation Facility (ESRF) in Grenoble, France [43] and at the B21 beamline of the Diamond Light Source, United Kingdom. The 2D spectra measured at the ESRF were obtained using X-rays with a wavelength of 0.1 nm and a beam size of 50 μm by 50 μm on the sample. The sample to detector distance was 10 m and the camera used was a Rayonix MX170 detector. The setup at Diamond used the same wavelength X-rays but a beam size of 1 mm by 1 mm, a sample to detector distance of 4 m and a Pilatus 300 K detector. The scattering vector was set to a range of q = 0.0032 to 0.38 Å^-1. All spectra were background corrected and normalized to the intensity of the transmitted primary beam. The scattering of the empty cell was subtracted. Flat, 1 mm-wide capillaries with a thickness of 200 nm were employed for the measurements. All of the samples were placed such that the beam hits the sample at normal angle.

Electromagnetic simulations were carried out using a commercially available finite-difference time-domain simulator in three dimensions (Lumerical Inc., FDTD). Here, latex particles were modelled as dielectric spheres with a refractive index of 1.37 while the background index was set to 1.33 to mimic the water. The colloidal lattice constant was varied between 300 and 450 nm in 25 nm intervals. The structure was illuminated at normal angle by a broadband plane-wave source. Bloch-boundary conditions were applied at the x- and y-boundaries of the simulation region while the z-boundary was set as a perfectly matched layer. The effect of the refractive index was tested by changing the refractive index of the particles in the photonic crystals between 1.33 and 1.54.

The FL particles were synthesised using an emulsion polymerisation technique reported previously [43]. Our FL particles have a refractive index of 1.37 [44], which is very close to the refractive index of water (1.33 at 20°C). We measured a zeta potential of (-82 ± 4) mV and an average hydrodynamic diameter of 196 nm with a polydispersity index of 1.6% (Figures 1A,B). Note that SEM images of the dried colloidal samples suggested an average particle size of (160±25) nm (inset in Figure 1A). In Figures 1C,D, we present the characterisation of the PS particles used: Their Zeta potential was (-30 ± 4) mV and their hydrodynamic diameter, determined in DLS measurements of highly dilute samples, was 220 nm with a polydispersity index of 1.4%. TEM images give a dry particle size of (190 ± 5) nm. These PS particles were synthesised by Cambridge Bespoke Colloids LLC using KPS as initiator and were surfactant free.

Both dilute aqueous suspensions of FL and PS particles appear white, with the dilute solutions of PS particles (2% v/v) showing some iridescent colour when left at rest (see inset in Figure 1C). Figures 2A,B show that upon concentrating the solutions by centrifugation, the FL suspensions became significantly more transparent and strongly coloured throughout, while both pure PS and mixed PS + FL suspensions remained predominantly white and opaque, showing intense colours only within a relatively narrow window of angles, which were due to Bragg reflection. The concentration process via centrifugation was fully reversible. For better visualisation and for the SAXS measurements shown later, we filled the concentrated samples into flat glass capillaries (25 × 4 × 0.2 mm) that were subsequently sealed with a two-component epoxy glue. In Figure 2A the images of FL samples display a strong blue reflection at 40% (v/v), green at volume fractions of around 26–33% (v/v) and red for 20–23% (v/v). Their spectra showed that the transmission decreased with decreasing concentration in the range of 40–20% (v/v) (Figure 2B). Below around 15% (v/v), the FL samples turned into whitish suspensions again with very little transmission in the visible. For comparison, the aqueous PS suspensions with a refractive index difference Δn_w-PS ≈  0.23 remained predominantly opaque at concentrations between 2% and 40%. In fact, for PS volume fractions of 26% and higher the transmission dropped almost to zero in the visible range. Spectra of the 1:1 FL + PS particles mixtures in Figure 2B showed slightly enhanced transmission with respect to the pure PS suspensions but remained opaque. Despite the opaqueness of the PS and the PS + FL suspensions they displayed strong Bragg reflections within a very narrow angular range (not shown here).

Following the reflectivity measurements indicating that the charge-stabilised FL and PS + FL colloidal suspensions form ordered structures, we performed SAXS measurements at the ID02 beamline at the ESRF. In Figure 3 we present the two-dimensional SAXS patterns and their corresponding structure factors, S(q), for FL and PS + FL particle suspensions with a volume fraction of ∼ 30% (v/v); the PS + FL samples were prepared with a 1:1 volume-fraction ratio to obtain an AB-type crystal. The S(q) curves were obtained by azimuthally integrating the two-dimensional scattering intensities, I(q), and dividing these with their form factor, F(q). Latter was obtained form scattering images taken for FL suspensions with a volume fraction of 0.1%. The indexing of the diffraction peaks in Figures 3B,C reveal that the FL particles packed in a face centred cubic (FCC) symmetry, while the 1:1 FL + PS suspension has a CsCl-like crystal.

Based on the SAXS measurement results, we carried out electromagnetic simulations to show that changing the lattice parameter of the FCC crystal sets the reflection colour observed. For this purpose, we formed FCC lattices of dielectric spheres having a refractive index of 1.37, mimicking the FL colloids. The simulations were done for various diameters of the particles and for various lattice parameters. Our simulations prove that despite the low refractive index contrast between the medium and the colloids, this system can still reflect the incoming radiation at a certain range of wavelengths, which is determined by the lattice constant (Figure 4). We observed that with increasing lattice constant there is a clear red-shift in the reflection peaks. Increasing the particle diameter also caused a slight red-shift in the reflection spectra which is a direct consequence of the Bragg condition. These results indicate that in our samples there are crystal grains with varying lattice parameters that leads to slightly different reflection peak wavelengths. Furthermore, different lattice parameters together with different grain sizes in the sample contribute to different absolute reflectance levels (Supplementary Figure S6). We also show that the whole visible regime can be covered using our colloids, which further proves the feasibility of using our materials in colour filters, reflectors, and optical coatings [45]. We have also analysed the effect of the particle refractive index on the reflection levels for perfect photonic crystals (Supplementary Figure S7). We see that the reflection becomes stronger with increasing refractive index contrast, which explains the difference of the transmission levels in FL and PS samples.

Already early on, the synthesis of fluorinated colloids [46] was pursued in order to investigate the physics of dense colloidal solutions. The fluorination of the starting monomers provided the low refractive index of the colloids needed to match or be close to that of water. The group of Asher [45] refined the synthesis of these fluorinated latex particles using emulsion polymerization of 1H,1H-heptafluorobutyl methacrylate monomers to obtain highly charged, monodisperse colloids ranging in size from 50 to 250 nm. While using the same starting monomer as Pan et al. [45], we employed KPS as initiator and the surfactant SDS to control the particle size. This simple synthesis allowed us to use the FL colloids directly after washing with di-water. They showed consistent Zeta potentials around −82 mV and DLS measurements provided a single, average hydrodynamic diameter of 196 nm with a polydispersity index of 1.6% (Figures 1A,B). However, SEM images suggested that our synthesis produced a bimodal size distribution. Performing SAXS measurements on dilute suspensions of our FL particles provided us with the form factor (see Supplementary Figures S1, S2), which indeed revealed a bimodal distribution: 70% of the signal accounted for FL particles with a diameter of 183 nm while the remaining particles showed a rather narrow size distribution around 134 nm. The low fraction of the smaller colloids does not show up in the DLS measurements because of its relative lower intensity, thus falling inside the tail of the scattering intensity due to the larger particles. This bimodal size distribution had little effect on the overall crystallinity of the sample as will be discussed in the upcoming sections. At this point it needs to be stressed that the timing of the emulsification process in nitrogen atmosphere is important. If the nitrogen purging continued for too long before the polymerization starts, more monomer evaporated leading to smaller particles, and if the purging time was too short, larger particles formed. In both cases, the polydispersity was affected as well.

In Figures 1C,D we show that the PS particles had also a negative Zeta potential (-30 mV) and their average hydrodynamic diameter was 220 nm with a polydispersity index 1.4%. TEM images gave a dry particle size of (190 ± 5) nm suggesting that the PS particles “shrink” less strongly than the FL ones when dried. Hence in aqueous solution, the two average particle sizes were similar but still differed by more than 10%.

Ashers’ group [45] developed transparent near-UV notch filters with a narrow extinction band by embedding their 82 nm large FL particles in a photo-polymerised matrix. They obtained a diffraction peak at 436 nm for a colloidal volume fraction of 3.6%. The diffraction peak blue-shifted to 270 nm when the samples were concentrated to 17.8% accompanied by an increase in the transmission. Our data presented here for aqueous suspensions of ∼196 nm large FL particles shows the same trend in the transmission data upon concentration (Figure 2): First, the 40%, 32% and the 26% FL samples showed transparency for wavelengths between 500 and 800 nm, with the most concentrated sample showing the highest transparency at all wavelengths. Second, samples with smaller colloidal volume fractions became completely non-transparent for the lower wavelengths.

As Ashers’ group demonstrated [45] the transparency of a dielectric background decreases as more and more colloids are added, due to the increased scattering. However, when even higher colloidal volume fractions are reached, the average refractive index of the system also increases. In this case, the opposite, namely an increasing transparency, is to be expected owing to the ordering of the particles as discussed earlier [47–49], which is in agreement with our observations. Further, while Pan et al. [45] observed colloidal crystallinity even at volume fractions as low as 3.6% our samples turned turbid. This suggest that although our FL particles form highly charged colloidal Wigner crystals at larger ϕ, their Debye screening length was not long enough to maintain the crystallinity when diluted. Instead, the thermal motion of the FL particles lead to random motion and thus diffuse scattering, which we observe.

What further sets our system apart from a large volume of research and development of large scale colloidal crystals for optical applications is the fact that our FL suspensions show a relatively angle-independent colouring in addition to the sharp Bragg reflection that only occurs for specific angles. Even when placed in conic pipette tips, the FL samples display structural colour over a broad range of angles (Supplementary Figure S4). The sharp Bragg reflections simply stem from the preferred ordering of spherical colloids near flat surfaces leading to several, hexagonally densely packed layers that form an FCC crystal. However, this packing usually does not reach deep into the colloidal bulk solution, which is why most PS or silica bead suspensions appear whitish. The colour we observed by eye in reflection varied only slightly in wavelength (see Figure 2C; SupplementaryTable S1), depending on the position measured in the sample. Such a weak dependence of the peak reflection wavelengths on position suggests the existence of multiple crystalline domains having similar but slightly different lattice constants, most likely originating from fractionation due to small variations in the surface charge from colloid to colloid and the bimodal size distribution of our FL particles.

As shown in Supplementary Figure S4, the samples preserve their structural colour in pipette tips indicating that the samples scatter the light at a broad range of angles while preserving the colour. We quantitatively analysed this with the samples placed in rectangular capillary tubes by illuminating them at normal angle and collecting the light at various angles (Supplementary Figure S4C). The results show that there is a very strong reflection at normal angle whereas at other angles the light is also strongly scattered leading to the perception of the sample’s intense colour at a broad range of angles. The presence of multiple crystal domains of similar lattice parameter but different orientation allows for the scattering to be observed at different angles. Nevertheless, the presence of multiple large crystal domains is expected to increase the light scattering that would cause the sample to appear whitish at off-angles. This can be avoided if the refractive index difference between the colloids and the background solvent is very small, which is the case here (Δn_w-FL ≈  0.04 as opposed to Δn_w-PS ≈  0.23).

In our initial attempts, we also obtained non-iridescent colours for suspensions of FL particles that are 150 nm or 250 nm in diameter (data not shown here). Their scattered colours were blue- or red-shifted, respectively, relative to the 196 nm case, as expected. However, the strength of the reflected colours was most pronounced for the 190–200 nm large particles. Our experimental observation was supported by the simulated reflection data shown in Figure 4. These show that the intensity of the reflected light is strongest for a lattice spacing that is about twice the particle size, which is for the concentrations we used at around 180–210 nm. Additionally, suspensions with the smaller sized FL particles showed even higher transparency, as the scattering strength of the particles decreases with their size.

Finally, we were also able to prepare quasi ‘inverted opal’ structures by increasing the refractive index of the continuous phase to 1.4 by saturating it with sucrose. These structures displayed similar characteristics of enhanced transparency and isotropic structural colour (see Supplementary Figure S5). Unusually, these samples were less likely to show additional sharp Bragg peaks to the bare eye. One possible reason for this weaker crystallinity in the high sucrose solution is the possibility that the Zeta potential and thus the Debye screening length is reduced allowing for the thermal motion to introduce more disorder. Indeed, SAXS measurements revealed a predominantly liquid-like structure (Figure 5). This suggests also that there may be small crystallites forming near the flat container walls while the bulk sample is in a more glassy state. For comparison, below we also present separate spectra from FL and PS particles in deionized water (Figure 6), which generally show more crystallinity.

We show that charged, fluorinated particle suspensions form regular Wigner crystals in the volume fraction range of 13% to about 40%. In this concentration range the suspensions show strong, almost angle-independent colours in reflection, that could be tuned continuously in the visible spectrum. We argue that the angle-independence of the colours seen is due to the presence of multiple crystal domains having different orientations. Since the low refractive-index contrast partly weakened the scattering from the crystal defects, the crystallinity and the low refractive index contrast together contributed to the samples being partially transparent. In addition these samples showed brilliant Bragg peaks, underlying the strong crystallinity near the container walls. Below 13% the FL suspensions became turbid and non-transparent, when the Debye screening length became shorter than the inter-particle distance and thus thermal motion causing colloidal disorder. Interestingly, mixing similarly sized and equally charged PS particles with the FL particles to form AB, A₂B, AB₂, A₁₃B and AB₁₃ type crystals we obtain only opaque samples that, nevertheless, showed Bragg reflections at specific angles. Interestingly, the binary mixtures containing a higher fraction of FL particles showed a greater degree of order. By playing with the surface charges and sizes of the colloids can obtain more complex structures than simple FFC symmetry. We conclude that such systems could be ideal for the development of filters with robust structure.