Plasmonic devices based on the unique optical properties of metallic nanoparticles (NPs) have been a topic of interest in the last decade, casting a wide net of promising applications in fields such as chemical and biomedical sensing , , , , , photochemistry , , , colouring , , , , , , , ,  and surface-enhanced spectroscopy , . Localised surface plasmon resonances (LSPRs) are due to collective oscillations of free electrons when metal nanoparticles are exposed to optical radiation. They are sensitive to the shape and size of the nanoparticles, the proximity of any surrounding nanoparticles and the refractive index of the local surrounding medium , , . The use of plasmonic nanoparticles in sensing applications exploits the latter for the detection of target analyte. Research on optimising the sensitivity of plasmonic nanoparticles by seeking the largest possible LSPR shift has been an important driver in the development of plasmonics. Researchers have explored the fabrication of nanowires , nanoshells , , nanodisks  and other distinctive shapes and arrangements  to achieve the best LSPR sensor. Typical fabrication methods of such structures include lithography , , , ,  and the growth of nanoparticles via chemical means . While nanoimprint lithography ,  has promising scale-up potential, the initial step of mould fabrication is lengthy and often limited to small surface areas.
Recent advances in ultrafast laser technology and the creation of single-step plasmonic surfaces via the fast direct laser writing , ,  of large areas ,  show great promise in increasing production and cutting down of the fabrication time and cost of plasmonic devices. In addition, compared to ordered surfaces commonly fabricated using lithographic methods, surfaces coloured by ultrafast lasers are random. For colouring applications, this implies that under certain conditions such surfaces do not exhibit angle-dependent colours typical of ordered surfaces . However, using different laser parameters, the exposure of metals to laser light can also result in angle-dependent colours via the creation of laser-induced periodic surface structures (LIPSS)  that originate from the interference of the incoming light with a scattered surface wave , . Recently, Guay et al.  demonstrated that these LIPSS, combined with nanoparticles, enhance plasmon resonances, and that they could be tuned via different laser burst arrangements. Colour arising from light interacting with LSPRs on noble metals is sensitive to the local environment, so colour changes can be monitored in colourimetric sensor applications .
Plasmonic colours created via laser pulses on Ag require a coating layer to passivate the surface . Here, we investigate the optical response of plasmonic colours coated by alumina layers of different thicknesses formed via atomic layer deposition (ALD) on laser-coloured silver surfaces. The colour palettes investigated were created via the burst and nonburst colouring methods , . We observed, with increasing alumina thickness, a broadening of the spectrum of burst colours and witnessed an initial fading of the colours followed by the recovery of the original chroma in the case of the burst colours. Finite-difference time-domain (FDTD) simulations were conducted to understand the role of an alumina layer in colour recovery, and to identify that the difference in behaviour between burst and nonburst coloured surfaces originates from the LIPSS that are formed using burst. We also consider our laser-coloured Ag surfaces as colourimetric and radiometric sensors, based on their responses to different alumina thicknesses.
2 Experimental methods
Silver surfaces were coloured by exposing large areas of the samples to laser irradiation by raster scanning the surfaces in a unidirectional top to bottom fashion (Figure 1). The machining of the surfaces was carried out using a wavelength of 1064 nm emitted by a Duetto mode-locked laser (Nd:YVO4, Time Bandwidth® Products, Zurich, Switzerland) outputting 10 ps pulses. Two laser colour marking schemes are investigated: the ‘burst mode’ composed of a train of closely time-spaced pulses and the standard ‘nonburst mode’. The colouring of silver using these two marking techniques is described in previous work , . The time separation between each of the pulses within the burst train is governed by the oscillator and is set to tib=12.8 ns. One to eight pulses within a laser burst can be selected and the energy distribution among the pulses can be adjusted using Flexburst™. The burst pattern (i.e. energy distribution) used to create the colours is shown in Figure 1B. The repetition rate, f=1/Tb, of the laser was set to 25 kHz using a pulse picker of 2 kHz for the burst colours and 50 kHz for the nonburst colours. After going through the XY galvanometric mirrors (TurboScan 10, Raylase, Webling, Germany), the light was focused onto the silver surface using an F-theta lens (f=254 mm, Qioptic, Goettigen, Germany). For accurate focusing, the surface of a sample was located using a touch probe arrangement. The silver samples were of 99.99% purity with a diameter of 38 mm and a thickness of 3 mm. For machining, the silver samples were placed on a three-axis stage (HF-MP23, Mitsubishi Electric, Tokyo, Japan) with a translation resolution of 1 μm in the lateral and axial directions. The laser power was controlled by a user interface and monitored using a power meter (3A-P-QUAD, Ophir, Jerusalem, Israel). The spot size was measured to be ~28 μm following the semi-logarithmic plot approach of the modified region following previous work .
The colours were quantified using a Konica Minolta CR-241 Chroma meter (Tokyo, Japan) in the CIELCH colour space, 2-observer and illuminant C (North Sky Daylight). The instrument outputs LCH values of the measured colours, where L is lightness, C is chroma (i.e. colour saturation) and H is hue (colour value associated with a 360° polar scale). The LCH values were converted to the XYZ tristimulus colour space using Matlab (Natick, MA, USA) for the plotting of the Commission de l’éclairage (CIE) diagrams. The reflectance measurements were taken via an in-house setup comprising a spectrometer (CCS200, Thorlabs, Newton, NJ, USA) and a cold LED source (MCWHF2, Thorlabs, Newton, NJ, USA). The light was sent and collected via a fibre-optic reflection probe bundle (RP28, Thorlabs, Newton, NJ, USA).
The ALD of alumina was performed using a Picosun R150 (Picosun Oy, Espoo, Finland) thermal deposition tool. Trimethylaluminium (TMA, >98%) and distilled water were held at 18°C in stainless steel bubblers for all of the depositions on the silver surfaces. Purges and line flows used 99.998% N2 (150 sccm for TMA, 200 sccm for water). For all the depositions, there were six initial 1 s pulses of TMA followed by a 30 s purge. After the initial cycle, pulses and purge times for TMA and water were 1 s and 90 s, respectively. The depositions were performed with the temperature of the chamber set to 60°C. In addition, Si (100) “witness” wafers were placed in the chamber to confirm the expected growth rate of the alumina films. The silver samples were placed on a stainless-steel mesh during deposition in order to support the samples and ensure complete coverage.
The alumina films were probed by electron X-ray dispersive microscopy (EDX) (Oxford Instruments, Abingdon, UK). The alumina layer thicknesses were estimated by entering the values of the EDX probe, voltage, takeoff angle, element, element k-ratio, and estimated density in the software GMFilm (Cambridge University, Cambridge, UK) . The EDX alumina thickness measurements were confirmed by ellipsometry (FUV-NIR, Horiba Uvisel, Kyoto, Japan), which was also used to measure the refractive index of the alumina films at an incident angle of 70° over the spectral range of 235–2000 nm in steps of 5 nm. Grazing angle measurements were performed with an FTIR spectrometer (Nexus 870, Thermo Nicolet, Waltham, MA, USA) with the incident light hitting the surface of the sample at an angle of 80°.
3 Results and discussion
3.1 Colourimetric response to the atomic layer deposition of alumina
In a previous study, a multi-layer coating method was developed to minimise morphological and colourimetric changes on the sample surface while protecting the surface against chemical attacks in order to target a specific industrial application . The mechanism of island formation and film growth on the silver surface during ALD was discussed . Here, a layer of alumina is deposited over pre-existing silver colours to investigate its effect on colour changes in order to compensate or predict the final image canvas. Figure 2 is a side-by-side comparison of a full colour palette (a) without and (b) with a 90-nm-thick film of alumina deposited via ALD. The colours on the surface of the silver samples were obtained using the nonburst (top in Figure 2A and B) and burst (bottom in Figure 2A and B) laser colouring methods, discussed in a previous publication , . The nonburst and burst laser parameters used were a marking speed of 100 mm/s with a laser fluence of 5.73 J/cm2 and a marking speed of 150 mm/s with a fluence of 12.12 J/cm2, respectively. The different colours were obtained by changing the line spacing, Ls, between successive lines, controlling the density of nanoparticles re-deposited onto the metal’s surface .
As shown in Figure 2A and B, the colours are observed to change significantly for both nonburst and burst colours following the application of the 90 nm layer of alumina. For the burst case, intense colours are still observed after deposition of the 90 nm alumina film. Conversely, for nonburst, the colours are observed to be greyish and lighter. In industrial applications where visual colours are required, a smaller change in colour (i.e. ΔE) is preferred , thereby conserving the vivid colour palette. Figure 2B suggests that burst coloured surfaces are the preferred choice in applications where protective coatings or tagging agents are to be placed on the coloured surface. Conversely, in colourimetric sensing applications, large changes in colour are desired. Figure 2C shows the evolution of colour with increasing alumina thickness in steps of ~7 nm for a few selected line spacings. Twelve identical colour palettes on 12 silver samples were placed in the ALD chamber and one sample was removed after each deposition of ~7 nm. The nonburst colours are observed to gradually lose their chroma with increasing alumina thickness. This decrease in the visual quality of the colours is largely consistent with groups who created colours by lithographic fabrication techniques and covered their surfaces with a passivation layer , . Alternatively, for the burst case, the chroma of the colours is mostly recovered at larger film thicknesses. Indeed, the burst colours are observed to rotate in hue by almost 360° going from blue (H=249.2) to purple (H=291.4) with increasing alumina thickness (as discussed further below). Green colours can be distinctively observed during the colour transition with increasing ALD thickness, e.g. burst colours in Figure 2C for the case Ls=4 μm. For nonburst, only blue is observed to rotate in colour; however, the chroma of the colours is significantly reduced and does not recover with increasing alumina thickness. Green colours with this direct laser colouring method have, in the past, been difficult to obtain . Thus, the alumina layer can serve as a tuning element to obtain colours over a wider range of hue values. Furthermore, techniques for controlling the thickness of ALD in selective areas were previously demonstrated via the use of self-assembled monolayer (SAM) films. The selective deposition of the ALD layer was controlled by etching the SAM layers at the metal surface , . In a previous publication, we demonstrated that our surfaces could be used to cleave SAM molecules by a plasmon-assisted mechanism, making the deposition of selective thicknesses of alumina a real possibility .
Figure 3 shows CIE xy chromaticity diagrams for nonburst and burst colours as the alumina film thickness increases. The data points in the CIE diagram were obtained from the LCH measurement of each colour using the unpolarised light of the chromameter. Initially, the area covered by the colours on the CIE diagrams for each of the colouring methods is observed to decrease with increasing alumina thickness. This observation is consistent with other studies where a passivation layer was deposited on nanostructures responsible for colours . In a CIE xy chromaticity diagram, the chroma (i.e. saturation) increases from the centre outwards. For burst, the area covered by the colours extends over more of the green region after the initial deposition of alumina. In addition, the decrease in covered area was observed up to a thickness of 55 nm before gradually recovering up to the maximum applied thickness of 90 nm. The reduction in the area is caused by a reduction in the chroma. The point of recovery for the burst colours appears to occur around ~55 nm of alumina. At 90 nm, the chroma is largely recovered. In the nonburst case, the chroma decreased and never recovered, yielding colours that are mostly grey. In addition, the area spanned by the colours continuously decreased with increasing thickness, collapsing into a small region of the CIE diagram.
Figure 4 shows polar plots of hue (θ) versus total accumulated fluence (r) for the nonburst and burst colours with increasing alumina thickness. For the nonburst colours, the evolution of the hue values is observed to follow a brushing/sweeping motion (red arrow) with increasing alumina thickness. The brushing/sweeping motion initially moves to the right, until the thickness of 55 nm, before reversing direction to sweep left. At the thickness of 90 nm, the hue values have almost recovered their original values. It should be pointed out that throughout the motion in the nonburst case, new hue values are never added. For burst colours, however, the evolution of the hue values is different, exhibiting a counter-clockwise rotation with increasing alumina thickness. This counter-clockwise motion allows access to the green hue range (90–150°) that is usually very difficult to achieve with direct laser machining alone. A chroma gain of up to 61% can be observed in the green colour region (H=90 to 150) compared to the best green colours obtained to date . Hence, the burst colours are tunable to cover hue values outside the normal range depending on the thickness of the alumina layer. Similarly, to the nonburst colours, the original hue range is mostly recovered after deposition of 90 nm of alumina, with a distinctive crossing point at 55 nm.
It is widely known that changes in the local permittivity can be used to tune the plasmonic response of nano-textured surfaces . While both colouring methods recovered their original hue values (Figure 4), recovery of the chroma values was observed solely in the burst case (Figure 3). The results were observed to be reproducible in multiple trials demonstrating the deterministic process of the colour recovery of the burst surfaces. However, since the burst and the nonburst surfaces have similar nanoparticle statistics  and alumina film thicknesses, it would be expected that both should present a similar behaviour. The only difference between the two lies with the underlying topography of the burst surfaces . The LSFL and HSFL structures found on the burst coloured surface is the key to understanding the different behaviours with increasing alumina thickness. The structures would serve to increase the area covered by the nanoparticles and cause field enhancement in the crevices, both increasing absorption . The colours produced by the burst method are largely angle independent; however, if the colours are observed under intense white light, it is possible to see faint diffraction. The suppression of diffraction by the underlying periodic structure was previously explained as selective absorption dominating over diffraction . We observed experimentally, for burst arrangements that produced the best-defined LSFL and HSFL structures, a dichroism in the perceived colours (e.g. blue-purple; purple-red; red-yellow) by simply rotating the coloured samples by 90°. The effect of diffraction is ruled out due to the absence of the majority of the colour spectrum while rotating the sample. The dichroism of the surface could originate from selective absorption by the defined LSFL and HSFL periodic structures. It is important to note that the burst colours used in this paper did not exhibit any dichroism.
In a colouring application, the passivation of the nonburst coloured surfaces would necessitate the deposition of a thickness lower than 7 nm in order to minimise the change in the colours and retain the original visual appeal of the colours (see Figure 3). For burst, the alumina layer could either be below 7 nm or around 90 nm (see Figure 3B), the latter providing better passivation. Furthermore, while the coloured surfaces have been previously shown to be sensitive to heat and to irreversibly degrade with prolonged exposure , the recovery of the colours with increasing film thickness rules out heat as being responsible for the colour change. In a previous publication, it was demonstrated that alumina films deposited at 60°C required a minimum layer thickness of ~60 nm in order to protect coloured surfaces against chemical attack . In an application, it is important that the protective layer has a minimal effect on the colours. For example, a colour design featuring a blue sky should still be blue following the deposition of the protective layer. The introduction of a protective layer usually shifts and shrinks the colour gamut rendering some colours inaccessible . This has been a significant problem for plasmonics in decorative applications. A strategy to counter these issues has been to pre-compensate for the protective layer by modifying the nanostructure responsible for the colours . This approach is certainly good in practice; however, coating techniques often demonstrate certain variability between runs which can counteract and negate the effect of the pre-compensated nanostructure. For instance, in our trials, we observed a significant change in colour for just a few nm of alumina. Alternatively, metal-insulator-metal (MIM) structures supporting gap plasmons and used to render colour are largely insensitive to changes in the local permittivity above the metal layer . However, like most approaches based on nanolithography, the fabrication of MIM structures can be lengthy, complex and is limited to small planar areas. Therefore, recovery of the original colours of the burst surfaces reduces the difficulty of conserving the original colour design and provides an elegant solution to the problem. It does not require any pre-compensation of the nanostructure for the application of the protective layer while providing the necessary protection against the environment, but tight tolerances are required in the formation of the layer.
Figure 5 shows reflectance measurements of selected coloured squares with different alumina thicknesses. The reflectance measurements using unpolarised light in Figure 5 are for blue (Ls=5 μm) and purple (Ls=8 μm) colours prior to the deposition of the alumina film, produced using the nonburst and burst colouring methods . The features in the reflectance curves appear to follow similar trends for the different colours. The burst colours showed more defined features and larger variations between the peaks and valleys, explaining the higher chroma. With increasing film thickness, various features in the reflectance spectra emerge or shift (dotted lines); some of these features are more active than others. The shifts in the peaks/troughs are observed to be colour dependent with the blue colours being the most active.
3.2 Surface chemical analysis
FTIR measurements of coated surfaces, plotted in Figure 6A, clearly show the presence of Al2O3 on the surface of coloured silver. A peak at 952 cm−1 representative of Al–O vibration can be distinguished , . The alumina film was deposited at the low temperature of 60°C in order to limit any colour change caused by thermal damage to the nanostructure during deposition (the melting temperature of small silver nanoparticles is lower than the bulk , ). Silicon witness wafers were placed in the deposition chamber and one was removed systematically at the same time as a silver sample. The thickness of each film was measured using a multi-point measurement approach and determined using GMRFilm software using K-ratios from electron dispersive X-ray spectroscopy . The thickness of each film was also determined by ellipsometry measurements, along with their refractive index dispersion. Figure 6B compares the refractive index of thin films of Al2O3 deposited via pulsed laser deposition , to the refractive index of our Al2O3 layers deposited by ALD on our witness wafers. The thicknesses obtained by ellipsometry on the silicon witnesses are in agreement with the EDX measurements obtained on the silver samples. In every case, the measured refractive index of the deposited films is lower than the films measured by Boidin et al. . The low refractive index of our alumina is explained by the low deposition temperature used . The refractive index of the deposited alumina layer can be modified by changing the deposition temperature; e.g. see the case for 250 cycles deposited at 100°C in Figure 6B.
3.3 Numerical simulations
Numerical simulations using the FDTD method were conducted to give insight into the experimental observations. We use a 3D model with a silver surface in the xz plane, which can be a flat surface or a surface with a ripple structure, representing nonburst and burst morphologies, respectively. In Figure 7, we show the setup for a burst simulation and highlight the geometric parameters. We model burst by a sine-modulated surface f(x, z)=A·sin(2πx/Lp), where Lp is the spatial period of the sinusoid and A its amplitude. The nonburst simulation is obtained for A=0. Silver nanoparticles of radius R are distributed on the burst silver surfaces following a square lattice with centre-to-centre distance Dm and an embedding length into the surface Remb expressed as a fraction of the nanoparticle radius. We consider a y-propagating plane wave excitation (normal incidence) and periodic boundary conditions are applied in the x- and z-direction. We excite the system with a broadband pulse to get results in the optical spectrum in a single run of the code. For the simulations, we used in-house parallel FDTD software . An alumina layer of thickness tALD is applied conformally to cover the surface of the silver samples . The optical parameters of alumina are modelled in FDTD by fitting the data in Figure 6B (case of 250 cycles at 100°C) to a Lorentz model with relative permittivity where εs=2.6872, ε∞=1, ω0=1.9833×1016 rad/s, and δ0=1.325× 106 rad/s. The Ag is modelled using the Drude model with two critical points (Drude+2CP) . A space step of 0.5 nm was used to discretise the simulation domain.
In particular, Figure 7 shows the electric field magnitude for the x-polarised excitation at λ0=500 nm for Lp=1000 nm, R=35 nm, Remb=0.3R, Dm=100 nm, A=100 nm and tALD=50 nm. In the figure we see a surface wave propagating in the x-direction directly above the alumina layer due to grating coupling. This figure is also extracted from Media 1, which shows the field distribution as a function of λ0 for x-polarization, while Media 2 shows the field distribution as a function of λ0 for z-polarization. In Media 1 and Media 2 we observe that for shorter wavelengths the field is more diffused in the alumina layer, while for longer wavelengths the field is confined in the vicinity of the nanoparticles. In Media 1 we also see the complexity of the resonant modes for nanoparticles arranged on a sinusoidal surface. In Media 3 we show the time domain evolution of the electric field in the case of an x-polarised plane wave pulse interacting with the surface for Lp=1000 nm, Dm=100 nm, A=100 nm and tALD=100 nm. This movie helps in understanding the simulation setup and the physics involved. We consider here only LSFL (Lp=1000 nm), but simulations for HSFL (Lp=200 nm) were also conducted and they show similar trends (see Figure S3 and Media S1).
As shown in Figure 2, the colours obtained with the burst technique are more saturated than the colours obtained with nonburst , and this is valid also when an alumina layer is applied. In Figure 8A and B, we show the simulation results for nonburst and burst (Lp=1000 nm, A=150 nm), respectively. The simulations were performed for R=35 nm, Remb=0.3R and Dm=75 nm for increasing tALD. In our simulations, we find that the colours in the burst case (when the burst surface is excited through a plane wave linearly polarised in the z-direction) are more saturated than in the nonburst case for most alumina thicknesses. This is due to the underlying ripples which produce field enhancement in the crevices  and to the alumina layer which further confines the field close to the nanoparticles.
The local minima in the reflectance curves are due to the fact that the random surface is approximated by a perfectly ordered array of nanoparticles. The reflectance becomes smoother by averaging the curves associated with the variation of a geometric parameter, e.g. Lp, but the overall trend of the reflectance is not altered. Thus, the result of a single simulation (without performing averages) is enough to describe the behaviour of the surface.
In simulations, the colour is strongly dominated by shifts in plasmonic resonances up to a certain alumina thickness tALD ~0.5Dee (Dee is the edge-to-edge distance, defined as Dm – 2R). In fact, a small increment of tALD between 0 and ~0.5Dee produces a quick filling of the gaps in between the nanoparticles and a rapid change in colour. For example, in the case of silver nanoparticles in Figure 8 (R=35 nm, Remb=0.3R and Dee=5 nm), the conformal growth of the alumina layer increases the optical thickness between nanoparticles and a transition from blue to yellow is observed (see Figure 8), which mimics the trend previously reported , where we found that increasing interparticle distance of surfaces without alumina changes the colour from blue to yellow. For Dm=90 nm (Dee= 20 nm, see Figure 9), we see a transition from yellow to blue due to an increase of the effective nanoparticle size. In simulations, when we increase the alumina layer thickness from tALD=0 nm to tALD ~0.5Dee, we do not consider any modification of the geometry and distribution of the nanoparticles and we observe a sharp colour transition. In experiments, the surface could be modified when the alumina deposition process starts due to a temperature increase . This may modify the geometry and distribution of nanoparticles, for example, the nanoparticles could embed further into the surface, which may cause a drastic colour change. Recently, it was observed that the nanoparticles may consist of a core/shell (Ag/Ag2O or Ag/AgOHAg2CO3) structure which could significantly alter its dielectric environment . The presence of the oxide and hydroxyls would help the initial growth of the alumina but at the same time chemical changes may be incurred at the surface . On the basis of these considerations, we will not try to compare our simulations to the experimental results for an alumina thickness less than ~0.5Dee, and we will rather focus on the colour evolution for tALD≥0.5Dee (~10 nm). Under this condition the colour evolution as a function of tALD becomes more gradual.
In the case of x-polarised excitation of a burst surface, the grating-coupling (see Figure 7) may produce narrowband dips in the reflectance spectrum at λ0~0.5Lp that modifies the colour with respect to z-polarised excitation . Figure 9A and B show this effect in burst simulations as a function of tALD for x-polarization and z-polarization, respectively, for Lp=1000 nm, A=150 nm, R=35 nm, Remb=0.3R and Dm=90 nm. The polarization dependence arising in burst simulations is observed also in experiments, though not as strongly, and only for burst arrangements with very defined and continuous ripple structures. In Figure 2C, despite the low saturation of the nonburst colours (the colours seem to bleach), we still notice similar colour trends for nonburst and burst for increasing tALD; this was also observed in our simulations over a wide range of parameters (see Figure 8 for example). Since the contribution of the ripples to the hue is somewhat marginal as they primarily contribute to the colour saturation, nonburst simulations would be sufficient to study the mechanism of the colour evolution with increasing tALD. For all these reasons, we consider nonburst surfaces in the following discussion.
In Figure 10A and B, we show the colour evolution as a function of tALD for Dm=75 nm (Remb=0.5R) and Dm=90 nm (Remb=0.6R), respectively; the nanoparticles are arranged in a hexagonal lattice, and we consider R=35 nm and A=0 nm (i.e. a flat surface) in both cases. The trend in Figure 10A is observed in Figure 2C for nonburst 18 μm and burst 13 μm. The trend in Figure 10B is observed in Figure 2C for nonburst 5 μm and burst 4 μm. While the experimental palettes are shown up to tALD=90 nm, we have performed our simulations up to tALD=200 nm to confirm the colour recovery and rotation. In fact, in Figure 10A, we see that the orange for tALD=40 nm (H=68, C=51) recovers at tALD=180 nm (H=70, C=51) with a rotation over an alumina thickness of 140 nm. In Figure 10B, we see that the green for tALD=20 nm (H=140, C=5) recovers at tALD=160 nm (H=134, C=25) with higher saturation and with a rotation over an alumina thickness of 140 nm. The rotation of the colour palette over tALD ~140 nm (which is ~0.5 λ/n, where n~1.65 is the refractive index of alumina measured in Figure 6B) suggests that it may be due to the interference between the reflected wave at the air/alumina interface and the reflected wave at the alumina/silver interface. While the colour is strongly influenced by the plasmonic dip above tALD=10 nm, which does not seem to change, it is also affected by the condition of maximum constructive interference, which depends on the alumina thickness as well as the phase of the plasmonic reflection coefficient that is highly dependent on frequency. This maximum constructive interference condition will then move to different colours as the thickness increases, eventually returning to the original colour when the thickness has increased by ~0.5 λ/n. To summarise, for a small tALD the colour is primarily dominated by changes in plasmonic resonances, and starting from a certain tALD threshold the plasmonic resonance is less perturbed by adding more alumina. Thus, the evolution of the colour is dominated primarily by interference between the wave reflected at the air/alumina interface and the wave reflected at the alumina/plasmonic interface, where we can imagine the plasmonic layer as a composite layer made by metal and alumina with all the gaps filled by alumina.
The evolution of the colours as a function of tALD can also be used as a fingerprint to retrieve the nanoparticle distribution, since the sequence of colours is intimately connected to the surface characteristics. Another interesting aspect about the use of a coating layer is the fact that the rotation of the colour can be exploited for tunability and creation of colours which are difficult to create on the bare surface, for example green. Tunability means that starting from a colour (hue), there is an alumina thickness which is able to create any other colour. This is not true for all the initial colours, as we can see in Figure 10. For example, we find that it is true for the case in Figure 10B, where we see that the palette covers the full hue spectrum, and it is not true for the case in Figure 10A, where the range of hues is limited.
3.4 Surface sensitivity
The visual differences in the colours and the changes observed in the reflectance spectra with increasing alumina thickness (Figures 2C and 5) suggest that the surfaces could be used in colourimetric sensing applications (e.g. colour change biosensors). To examine the potential for this application, we plot in Figure 11 the change in colour, ΔE, versus hue with increasing film thickness for nonburst and burst colours. The change in colour is defined by the following relation :
where a and b are given as
L is lightness, C is chroma and H is hue – the three components of the LCH colour space. In Eq. (1), subscripts 1 and 2 are used to identify the colours before (1) and after (2) ALD coating. ΔE~1 is the minimum colour change that the human eye can discern . A significant change in colour can be observed for both the nonburst and burst colours with increasing alumina thickness. ΔE is observed to depend on the starting colour before deposition of the alumina film. In both cases, the most sensitive hue region is in the purple colour range (H=270–330). The sensitivity of the burst colours is observed to be higher in the blue (H=210–270) and purple region (H=270–330), by 96% and 75%, respectively, compared to the nonburst colours of the same hue regions. The increase in ΔE for the burst colours is believed to come from the periodic structures in the burst coloured surfaces. The ability to tune these structures with different burst arrangements  could potentially serve to increase their sensitivity for better performance in colourimetric sensing applications. Yellow (H=30–90) and red (H=330–30) colours do not produce a significant enhancement in ΔE for burst, yielding 13% and 64% increases, respectively. The similar sensitivities of the nonburst and burst surfaces in the yellow and red hue regions is attributed to the disappearance of the periodic features present in burst with increasing line spacing . Interestingly, the improvement in sensitivity of the burst colours follows the improvement in chroma compared to the nonburst colours. These results suggest that burst purple (Figure 11B) is the ideal choice for colourimetric sensing applications with a sensitivity of 3.06/nm (2.02/nm for nonburst – ΔE is dimensionless), implying that less than 1 nm of material on the surface is required for the change in colour to be perceived by the human eye. The sensitivity of the coloured surfaces is, however, not linear with increasing alumina thickness. For a film thickness of 90 nm, the sensitivity of the purple colours drops to 0.56/nm and 0.30/nm for burst and nonburst, respectively, requiring about 2 nm of material to produce a perceivable change in colour. Blue colours are observed to change the least with increasing film thickness.
Interestingly, while blue surfaces do not perform well as a colourimetric sensor, the surface is best as a radiometric sensor (see Figure 5). For the blue surfaces, the highest sensitivities observed are for the resonant features identified as Fnb,1,blue and Fb,2,blue, yielding 2.58 and 3.19 nm/nm for the nonburst and burst cases, respectively. These sensitivities rival sensors fabricated via nanolithography techniques. For the purple cases, the features producing the highest sensitivities are Fnb,4,purple and Fb,4,purple, yielding 0.63 nm/nm and 1.02 nm/nm, respectively. The sensitivity of the different features identified in the reflectance spectra varies significantly with thickness. For example, Fnb,2,blue and Fb,3,blue produce sensitivities of 0.31 nm/nm and 0.35 nm/nm, respectively. The plasmon shift, with increasing alumina thickness, is observed to decrease for colours produced with larger line spacings (i.e. lower nanoparticle densities – blue to yellow) making the blue colours better for radiometric sensing. In addition, while the current colours were chosen because they were aesthetically pleasing, colours can be produced with marking speeds as fast as 3000 mm/s making this process viable for industrial applications and the marking of large surface areas. Furthermore, while silver cannot be used directly in biosensing applications due to its reactivity, a thin passivation layer could be formed by ALD on the surface.
Colour palettes produced on the surface of silver using nonburst and burst colouring methods were modified by alumina films produced via ALD. The colour and reflectance changes were characterised as a function of alumina film thickness. For burst, the colours first degrade with increasing film thickness but recover at larger thicknesses and the colour range is expanded, whereas for nonburst the colours keep degrading. Underlying periodic structures specific to burst are responsible for this behaviour. FDTD modelling of representative surfaces, including a conformal alumina layer, helps explain the colour rotation and recovery observed in experiments with increasing alumina thickness. For alumina thicknesses smaller than the nanoparticle gaps, the changes in the perceived colours are due to the perturbation of plasmonic resonances. For alumina thicknesses larger than the nanoparticle gap, the change in colours originates from the complex reflectance response of the alumina-coated structures due to plasmonic resonances and interference effects in the multi-layer system.
In addition, the surfaces are thought to be good candidates for colourimetric sensing. The sensitivity of the surfaces depends on the initial colour prior to deposition. The colourimetric response of the burst surfaces is increased by up to 96% relative to the nonburst surfaces. The purple colours are the most sensitive, yielding a sensitivity of 3/nm, which implies that less than 1 nm of alumina is needed to produce a colour shift perceivable by the human eye. The coloured surfaces could also perform well as radiometric sensors, yielding sensitivities as high as 3.19 nm/nm for the blue colours. The ability to tune the structures for grating-assisted coupling and improve colours using burst makes laser-written surfaces very versatile for specific sensing application.
We acknowledge the Royal Canadian Mint, the Natural Sciences and Engineering Council of Canada, the Canada Research Chairs program, the Southern Ontario Smart Computing Innovation Platform (SOSCIP), and SciNet. We would also like to acknowledge Dr. Jaspreet Walia, Dr. Fabio Variola and Maël Chow-Cloutier at the University of Ottawa.
Zhu C, Yang G, Li H, Du D, Lin Y. Electrochemical sensors and biosensors based on nanomaterials and nanostructures. Anal Chem 2015;87:230–49. Google Scholar
Kreno LE, Leong K, Farha OK, Allendorf M, Duyne RPV, Hupp JT. Metal-organic framework materials as chemical sensors. Chem Rev 2011;112:1105–25. Google Scholar
Xu X, Peng B, Li D, et al. Flexible visible-infrared metamaterials and their applications in highly sensitive chemical and biological sensing. Nano Lett 2011;11:3232–8. Google Scholar
Anker JN, Hall WP, Lyandres O, Shah NC, Zhao J, Duyne RPV. Biosensing with plasmonic nanosensors. Nat Mater 2008;7:442–53. Google Scholar
Kabashin AV, Evans P, Pastkovsky S, et al. Plasmonic nanorod metamaterials for biosensing. Nat Mater 2009;8:867–71. Google Scholar
Walia J, Guay J-M, Krupin O, Variola F, Berini P, Weck A. Visible light driven plasmonic photochemistry on nano-textured silver. Phys Chem Chem Phys 2018;20:238–46. Google Scholar
Kim C, Lee H. Light-assisted surface reactions on metal nanoparticles. Catal Sci Technol 2018;8:3718–27. Google Scholar
Zhang Y, He S, Guo W, et al. Surface-plasmon-driven hot electron photochemistry. Chem Rev 2018;118:2927–54. Google Scholar
Vorobyev AY, Guo C. Direct femtosecond laser surface nano/microstructuring and its applications. Laser Photon Rev 2013;7:385–407. Google Scholar
Fan P, Zhong M, Li L, et al. Sequential color change on copper surfaces via micro/nano structure modification induced by a picosecond laser. J Appl Phys 2013;114:083518. Google Scholar
Fan P, Zhong M, Li L, et al. Angle-independent colorization of copper surfaces by simultaneous generation of picosecond-laser-induced nanostructures and redeposited nanoparticles. J Appl Phys 2014;115:124302. Google Scholar
Guay J-M, Calà Lesina A, Côté G, et al. Laser-induced plasmonic colours on metals. Nat Commun 2017;8:16095. Google Scholar
Clausen JS, Højlund-Nielsen E, Christiansen AB, et al. Plasmonic metasurfaces for coloration of plastic consumer products. Nano Lett 2014;14:4499–504. Google Scholar
Duempelmann L, Casari D, Luu-Dinh A, Gallinet B, Novotny L. Color rendering plasmonic aluminum substrates with angular symmetry. ACS Nano 2015;9:12383–91. Google Scholar
Kristensen A, Yang JKW, Bozhevolnyi SI, et al. Plasmonic colour generation. Nat Rev Mater 2016;2:16088. Google Scholar
Duan X, Kamin S, Liu N. Dynamic plasmonic colour display. Nat Commun 2017;8:1–9. Google Scholar
Guay J-M, Killaire G, Gordon PG, Barry ST, Berini P, Weck A. Passivation of plasmonic colors on bulk silver by atomic layer deposition of aluminum oxide. Langmuir 2018;34: 4998–5010. Google Scholar
Camden JP, Dieringer JA, Zhao J, Van Duyne RP. Controlled plasmonic nanostructures for surface-enhanced spectroscopy and sensing. Acc Chem Res 2008;41:1653–61. Google Scholar
Haes AJ, Haynes CL, Mcfarland AD, Schatz GC, Van Duyne RP, Zou S. Plasmonic materials for surface-enhanced sensing and spectroscopy. MRS Bull 2005;30:368–75. Google Scholar
Mie G. Beiträge zur Optik trüber Medien, speziell kolloidaler Metallösungen. Ann Phys 1908;330:377–445. Google Scholar
Doyle W. Optical properties of a suspension of metal spheres. Phys Rev B Condens Matter 1989;39:9852–8. Google Scholar
Murray WA, Barnes WL. Plasmonic materials. Adv Mater 2007;19:3771–82. Google Scholar
Cvelbar U, Ostrikov KK, Drenik A, Mozetic M. Nanowire sensor response to reactive gas environment. Appl Phys Lett 2008;92:133505. Google Scholar
Tam F, Moran C, Halas N. Geometrical parameters controlling sensitivity of nanoshell plasmon resonances to changes in dielectric environment. J Phys Chem B 2004; 108:17290–4. Google Scholar
Raschke G, Brogl S, Susha AS, et al. Gold nanoshells improve single nanoparticle molecular sensors. Nano Lett 2004;4:1853–7. Google Scholar
Liu N, Mesch M, Weiss T, Hentschel M, Giessen H. Infrared perfect absorber and its application as plasmonic sensor. Nano Lett 2010;10:2342–8. Google Scholar
Shipway AN, Katz E, Willner I. Nanoparticle arrays on surfaces for electronic, optical, and sensor applications. Chem Phys Chem 2000;1:18–52. Google Scholar
Kumar K, Duan H, Hegde RS, Koh SCW, Wei JN, Yang JKW. Printing colour at the optical diffraction limit. Nat Nanotechnol 2012;7:557–61. Google Scholar
Roberts AS, Pors A, Albrektsen O, Bozhevolnyi SI. Subwavelength plasmonic color printing protected for ambient use. Nano Lett 2014;14:783–7. Google Scholar
Tan SJ, Zhang L, Zhu D, et al. Plasmonic color palettes for photorealistic printing with aluminum nanostructures. Nano Lett 2014;14:4023–9. Google Scholar
Cheng F, Gao J, Stan L, Rosenmann D, Czaplewski D, Yang X. Aluminum plasmonic metamaterials for structural color printing. Opt Express 2015;23:23279–85. Google Scholar
Liu J, Lu Y. Fast colorimetric sensing of adenosine and cocaine based on a general sensor design involving aptamers and nanoparticles. Angew Chem 2006;118:96–100. Google Scholar
Sipe JE, Young JF, Preston JS. Laser-induced periodic surface structure. I. Theory. Phys Rev B 1983;27:1141–54. Google Scholar
Huang M, Zhao F, Cheng Y, Xu N, Xu Z. Origin of laser-induced near-subwavelength ripples: interference between surface plasmons and incident laser. ACS Nano 2009;3:4062–70. Google Scholar
Guay J-M, Calà Lesina A, Baxter J, et al. Topography tuning for plasmonic colour enhancement via picosecond laser bursts. Adv Opt Mater 2018;6:1800189. Google Scholar
Jandeleit J, Urbasch G, Hoffmann HD, Treusch H-G, Kreutz EW. Picosecond laser ablation of thin copper films. Appl Phys A: Mater Sci Process 1996;63:117–21. Google Scholar
Waldo RA. An iteration procedure to calculate film compositions and thicknesses in electron-probe microanalysis. In: Newbury DE, ed. Microbeam Analysis. San Francisco, CA, San Francisco Press, 1988, 310–4. Google Scholar
Bunting F. The COLORSHOP color primer an introduction to the history of color, color theory, color measurement. Grandville, Michigan, Light Source Computer Images, Inc., 1998. Google Scholar
Fang M, Ho JC. Area-selective atomic layer deposition: conformal coatings, subnanometer thickness control, and smart positioning. ACS Nano 2015;9:8651–4. Google Scholar
Wojtecki R, Mettry M, Nathel NFF, et al. Fifteen nanometer resolved patterns in selective area atomic layer deposition-defectivity reduction by monolayer design. ACS Appl Mater Interf 2018;10:38630–7. Google Scholar
Miyata M, Hatada H, Takahara J. Full-color subwavelength printing with gap-plasmonic optical antennas. Nano Lett 2016;16:3166–72. Google Scholar
Yang H, Liu M, Ouyang J. Novel synthesis and characterization of nanosized γ-Al2O3 from kaolin. Appl Clay Sci 2010;47:438–43. Google Scholar
Fedel M, Deflorian F. Electrochemical characterization of atomic layer deposited Al2O3 coatings on AISI 316L stainless steel. Electrochimica Acta 2016;203:404–15. Google Scholar
Qi WH, Wang MP. Size and shape dependent melting temperature of metallic nanoparticles. Mater Chem Phys 2004;88:280–4. Google Scholar
Luo W, Hu W, Xiao S. Size effect on the thermodynamic properties of silver nanoparticles. J Phys Chem C 2008;112:2359–69. Google Scholar
Boidin R, Halenkovic T, Nazabal V, Benes L, Nemec P. Pulsed laser deposited alumina thin films. Ceram Int 2016;42: 1177–82. Google Scholar
Groner MD, Fabreguette FH, Elam JW, George SM. Low-temperature Al2O3 atomic layer deposition. Chem Mater 2004;16:639–45. Google Scholar
Calà Lesina A, Vaccari A, Berini P, Ramunno L. On the convergence and accuracy of the FDTD method for nanoplasmonics. Opt Express 2015;23:10481–97. Google Scholar
George SM. Atomic layer deposition : an overview. Chem Rev 2010;110:111–31. Google Scholar
Guay J-M, Walia J, Côté G, et al. Effect of ps-laser repetition rate on colour rendition, nanoparticle morphology and surface chemistry on silver. Opt Mater Express 2019;9:457–68. Google Scholar
The online version of this article offers supplementary material (https://doi.org/10.1515/nanoph-2018-0202).
About the article
Published Online: 2019-03-04
Citation Information: Nanophotonics, 20180202, ISSN (Online) 2192-8614, DOI: https://doi.org/10.1515/nanoph-2018-0202.
©2019 Jean-Michel Guay, Antonino Calà Lesina et al., published by De Gruyter, Berlin/Boston. This work is licensed under the Creative Commons Attribution 4.0 Public License. BY 4.0