Plasmonic colors are structural colors that are determined by the plasmonic resonances of metallic nanostructures or nano-devices. Tuning the geometries and sizes of plasmonic nanostructures is an effective means of tailoring the absorption spectra in the visible frequency range. This has been utilized to realize highly saturated color with narrow bandwidth^[1–3], color filtering with extraordinary optical transmission^[4–6], high-resolution color pixels for imaging^[7,8], polarization-dependent^[9,10] or -independent^[11–13] color filtering and imaging, and angle-insensitive structural color printing^[8,13].

Plasmonic color generation has aroused intense research interest for years, but most previous works rely on complex and expensive electron beam lithography (EBL)^{[7,12,14–18]} or focused ion beam (FIB) ^[19,20] to achieve demanded fabrication accuracy. These approaches cannot be modifiable once the structure is fabricated. Alternatively, a tightly focused pulsed laser beam opens a new route, tailoring light matter interaction with high efficiency and low cost^[16,21,22]. The pulsed laser provides an important method for optical storage and color printing by precisely controlling the geometry and size of plasmonic nanostructures. It is generally believed that laser induced temperature rising above the threshold is necessary for melting and shape transition of nanostructures^[18,21,23]. Kristensen etal. reported using nanosecond pulses to induce the photothermal reshaping of Al nanodiscs to realize color printing^[18]. However, this method that is dependent on particle ablations due to high pulse energy above the ablation threshold^[24] leads to generated nanostructures with less controllable morphologies. Later, it was reported that a high repetition rate femtosecond (fs) pulsed beam can locally generate enormous heat for photothermal reshaping of gold nanorods with low powers^[25]. Thermal reshaping of gold nanorods at the temperature that is below the bulk melting point is well dependent on the surface curvature^[26]. The dynamic thermal method can also reshape other high-ε metal materials such as Ti and W with various nanostructures^[27,28].

In this Letter, we demonstrate vivid color generation of Al nanodiscs induced by fs pulsed beams and propose a curvature-induced surface diffusion mode to explain the mechanism of shape transition of Al nanodiscs. The nanodisc is consequently reshaped below the melting threshold by using fs multi-pulses. The reshaped nanodisc geometries as well as the printed colors rely on the power of the tightly focused laser. The curvature-driven surface diffusion model is adopted to study the mechanism of morphologic evolution of Al nanodiscs. Our work proposes a method to print high-density color information, which holds promising application prospects in data storage, color printing, and information security^{[1–3,7,20,29,30]}.

The concept of our fs pulsed-beam-induced reshaping of Al nanodiscs for color printing and the curvature-driven diffusion model is shown in Fig. 1(a). For the large-scale demonstration, the initial structure can be achieved by cheaper technologies such as nanoimprinting^[31] and roll-to-roll processing^[32]. The color pixels are composed of arrays of different morphological changes from Al nanodiscs to Al nanospheres that are printed by fs pulses. Each of the individual plasmonic nanodiscs can display a deterministic color. We choose Al as the plasmonic material because it supports surface plasmon resonances (SPRs) from visible light to ultraviolet light. The structure of a unit is illustrated in Fig. 1(b). Arrays of 30 nm thick and 130 nm diameter Al discs are designed as the top metal. Furthermore, a thin (2–3 nm) oxide layer formed at the nanostructure increases the retention and durability of printed plasmonic pixels without significantly affecting spectral movement. A 200 nm thick silver layer is chosen as the grounding metal for the reason that the silver film has the largest reflectance in the visible wavelength range. The insulator layer is a 20 nm thick SiO2 film, and the substrate is also SiO2.

The MIM configuration forms an FP resonator to enhance absorption of the incident laser energy^[33]. By exciting the FP mode with localized electric field confinement, the electromagnetic energy and, hence, the photothermal effect are dominantly concentrated in the top Al nanodisc layer. Figure 2(a) shows the intensity distribution of the electric field, magnetic field, and heat after multi-pulse laser irradiation at the wavelength of 800 nm.

The transient absorbed electromagnetic energy density can be described by a formula related to the electric field distribution^[33]: Q(x,y,z,t)=ω2ε″(ω,x,y,z)ε0|E(ω,x,y,z,t)|2,(1)where ω is the angular frequency, and ε″ and ε0 represent the imaginary part of the relative permittivity and the vacuum permittivity, respectively. |E(ω,x,y,z,t)| is the time-varying electric field.

The two-temperature model (TTM) is used to describe the ultrafast heating behavior and temperature evolution. Transient thermoplasmonic heating can raise the temperature of Al nanodiscs through the electron lattice relaxation^[34]: δTeδt=−g(Te−Tl)/Ce(Te),(2)δTlδt=g(Te−Tl)/Cl−(Tl−T0)/τs,(3)where g is the electron–phonon coupling constant (g≈5.69×1017W·m−3·K−1 for Al^[35]), τs is the characteristic time for heat dissipation to the environment, Ce(Te)=γTe is the temperature-dependent heat capacity of the electron gas, and γ is the electron–phonon coupling constant (γ=134.5J·m−3·K−2 for Al^[35]). The high repetition frequency fs pulse laser (80 MHz, 140 fs) is selected to irradiate the Al disc. Laser power is decreased to prevent discs ablating away under ultra-high temperature. Depending on the laser fluence in experiment, the temperature of Al nanodiscs will rise after multi-pulse irradiance and then reach a dynamic equilibrium situation, as shown in Fig. 2(b). The nanodiscs thus transform their shapes into thicker discs or spheres in a crystallographic phase-transition process.

To explain the morphology evolution, we utilized the curvature-driven surface diffusion model to analyze the reshaping procedure^[34,36,37]. The main idea for this theory is that the surface energy of an object can be minimized by atomic surface diffusion. It leads to photothermal reshaping behavior of plasmonic nanomaterials that can be initiated below the melting points and are heavily dependent on the initial aspect ratio.

The surface flux J_s of atoms along an arbitrary surface under the assumption of isotropy of surface tension and surface self-diffusion can be written as Js=ΩVskTDs→·∇μ,(4)where Ω, Vs, and k represent an atomic volume, the number of diffusing surface atoms in the unit area, and Boltzmann’s constant, respectively. T is the temperature, and μ is the chemical potential. Ds→ is the interface diffusivity. The gradient notation ∇ is two-dimensional on the surface. Equation (1) should satisfy the equation of continuity, ∂n∂t+∇·Js=0,(5)where ∂n/∂t represents the outward diffusion speed of a point on the surface in the normal direction.

For an isotropic surface, the diffusivity tensor can be simplified as Ds(T), having typical Arrhenius behavior with an activation energy Ea and a constant D0 as Ds(T)=D0exp(−EakT).(6)

The chemical potential μ can be described as μ=μ0+γsΩK,(7)where K=1/Rx+1/Ry is the mean curvature of the surface, γs is the free energy, and μ0 is the chemical potential for a flat surface. Equation (4) becomes dndt=v=Ω4/3γsDskT∇2K.(8)

The equation describes relation between the movement speed of a surface point and the curvature-driven surface diffusion.

Figure 2(c) shows surface diffusion coefficients related to the temperature and aspect ratio. It is prone to knowing diffusion coefficients rising with temperature. On the other hand, surface atoms of a nanostructure with higher aspect ratios have higher diffusion coefficients at a fixed temperature. It leads to reshaping below the melting threshold. The reshaping behavior can be simulated by the finite difference method. Movement of points on the entire surface of an object is calculated using Eq. (8) for each time step Δt=ti to account for surface motion at given temperature T. For simplicity, the cross section of Al nanodiscs is approximated as an ellipsoidal shape. Figure 2(d) shows a morphology evolution of an ellipsoidal nanorod (30 × 130 nm) subjected to the temperature, which is given by TTM, as shown in Fig. 2(c). Figure 2(e) shows the experimental results obtained at variant incident laser fluences corroborated with the simulation with reasonable congruence. The scanning electron microscope (SEM) images of Al nanodiscs with different morphologies by increasing laser fluence are shown on the right side. The morphologies of the nanodiscs were imaged by an SEM (FEI, Apreo Hivac).

Apparently, the final profile with a reshaped aspect ratio depends on the total absorption of the incident laser energy. Figure 3(a) illustrates the experimental and simulated reflecting spectra of nanodiscs. It reveals that the reflectance valley shifts from 620 nm to 420 nm along with the increasing of laser powers, which is related to a decreasing diameter as manifested from the SEM images shown in Fig. 2(e). The reflectance spectra of the nanodisc were experimentally measured by custom dark field microscope. A supercontinuum laser (SC-PRO, YSL Photonics) was used as a white light source to couple into a microscope to illuminate the sample. A fiber spectrometer (USB4000, Ocean Optics) was employed to acquire the spectrum. The optical responses of the Al nanodiscs are simulated by finite difference time domain (FDTD) methods with commercial software Lumerical FDTD Solutions. The simulation was performed on one unit cell with periodic boundary conditions. A plane-wave source at wavelengths ranging from 400 nm to 800 nm was employed to illuminate the structure. A full library of nanodisc morphology by the continuous tuning of scattering colors across the entire visible spectrum is obtained with different laser parameters [Fig. 3(b)]. The International Commission on Illumination (CIE) 1931 chromaticity diagram based on the measured reflectance is shown in Fig. 3(c).

Employing this feature, we have demonstrated vivid laser color printing by precisely controlling the shape transition of Al nanodiscs. In the experiment, the sample is placed on a computer-controlled three-dimensional (3D) displacement platform. The fs laser pulse at the wavelength of 800 nm and the repetition frequency of 80 MHz is attenuated through the neutral density filter and past the optical shutter. The attenuator is used to adjust the pulse energy to the desired value, and the optical shutter controls the exposure time of the pulse. The beam is then focused on the prepared sample through an objective lens (NA=0.65). For the color image printing, the original color image was firstly divided into multiple parts of different color tones. Then, each color tone image was printed through the established relationship between color and laser fluence. Figures 4(a)–4(c) are a collection of images printed by fs laser beams, showing the capability of high-resolution printing with high color fidelity. Figures 4(d)–4(e) show the zoom-in view SEM images of part of Fig. 4(a). Figure 4(f) shows the comparison with Al nanodiscs after laser irradiation (left) and the initial condition (right). The color images were captured by an Olympus microscope (BX53). Our method shows accurate control of the morphology of nanodiscs due to morphology reshaping below the melting threshold so as to avoid nanoparticles ablation.

In conclusion, we have presented fs pulsed-beam-induced photothermal effects and the resultant shape transition of Al nanodiscs. We systematically study the mechanism of morphologic evolution of Al nanodiscs below the thermal melting threshold with the curvature-driven surface diffusion model. The shape transition and corresponding plasmonic resonances of nanodiscs can be independently and precisely modulated to achieve controllable color generation.