1. Introduction
Three-dimensional (3D) percolated porous nanostructures made of noble metals and having large surface-to-volume ratio have drawn significant attention due to their remarkable physicochemical properties allowing to use them for various important applications ranging from the photo- or electro-catalysis, water splitting, hydrogen storage to bio- and chemosensing via surface-enhanced effects [
1,
2,
3,
4,
5]. For most of the suggested applications, the surface-to-volume ratio defined by the distribution, density and size of the pores within the 3D nanostructure is of crucial importance.
Recently, optical properties of 3D porous nanostructures have become a hot topic [
6,
7,
8,
9,
10]. Specifically, the porous Au nanoparticles (also referred to as nanosponges) were shown to demonstrate polarization-dependent scattering as well as to support long-lived electron emission associated with localized and propagating surface plasmon modes having remarkably high quality factors. These optical properties make such structures appealing for various optical and nonlinear optical applications including random lasing, enhanced photo-emission, harmonic and supercontinuum light generation, as well as single-molecule biosensing based on metal-enhanced fluorescence, surface-enhanced Raman scattering (SERS), infrared absorption (SEIRA), etc. Multiple sensing applications are benefited from the random (but highly dense) arrangement of the plasmon-mediated electromagnetic (EM) hot spots within the structure allowing to obtain the spectrally broadband signal enhancement over the entire visible and near-IR spectral range [
11,
12,
13,
14]. Indeed, incorporation of the nanopores into the bulk plasmonic nanostructures with a sub-wavelength overall size provides more intense SERS signal due to multiple hot spots and enlarged surface area increasing probability for analyte molecules to reach these hot spots [
15,
16,
17,
18]. Nevertheless, upon excitation with EM radiation, both the arrangement of the nanosized pores and the overall nanostructure geometry govern the resulting response [
19,
20]. From this point of view, the plasmonic properties of the resulting 3D porous nanostructures, their general geometric shape as well as porosity are to be adjusted simultaneously that still remains challenging.
State-of-the-art methods for porous nanostructure fabrication generally require complicated multi-step fabrication protocols as dealloying and soft- or hard-template synthesis [
21,
22,
23,
24,
25,
26,
27], where accurate management of the reaction conditions (temperature, composition, precursors, etc.) is crucial. Furthermore, the minimization of the surface free energy typically leads to generally spherical-shaped nanostructures. Seed-mediated growth provides simple and versatile method allowing to produce arbitrary-shaped porous nanostructures [
28]. However, it’s often problematically to obtain only one desired geometry because of internal structural variations of the seeds as well as local variations of the reaction environment. Alternatively, liquid-free lithography-based approaches as electron- or ion-beam milling [
29] are suitable for high-precision formation of geometrically-diverse nanostructures. However, the need for upscaling of the fabrication procedure creates an economically justified barrier for lithography-based techniques being applied for porous nanostructure fabrication and large-area replication. In addition, post-processing is also required to impart porosity into the nanostructures.
Herein, we applied scalable easy-to-implement nanosecond (ns) laser ablation of nitrogen-rich Au films to fabricate parabola-shaped nanobumps containing multiple nanopores inside. The nanopores were found to originate from laser-induced explosive boiling and coalescence of the nucleation sites formed by nitrogen-rich areas of the Au film, while the nanopore density can be controlled by amount of nitrogen used as discharge gas for magnetron sputtering of the Au films. Most of the nanopores hidden beneath the nanobump surface can be further uncapped using gentle etching of the nanobumps by an Ar-ion beam to form 3D plasmonic nanosponges promising for various nonlinear optical and sensing applications.
3. Results and Discussion
In this paper we considered three types of glass-supported Au films produced via magnetron sputtering in various discharges gases—argon, nitrogen and purified air (≈80% of nitrogen). Produced films had the same thickness (
nm) and showed identical diffuse reluctance as well as AFM verified surface roughness of about
nm. This guarantied identical coupling of the incident ns-laser pulse energy to all types of metal films under study (
Figure 1a). Such laser pulse induces thermalisation of charge carriers in the metal film that results in its local melting accompanied by detachment from the substrate via relaxation of the thermal-generated stress or evaporation at the interface between the film and the substrate [
32,
33]. At a pulse energy that is smaller than the ablation threshold (F
0.17 J/cm
[
30,
33,
34,
35]), detached metal shell resolidifies before its rupture forming parabola-shaped surface protrusion (also referred to as nanobump;
Figure 1b,c). Typical laser-printed nanobump on the surface of 150-nm thick Au film sputtered with nitrogen discharge gas is illustrated by corresponding top- and side-view SEM images. The former image combines the signals from the SE and BSE detectors giving useful information regarding either a surface morphology or difference in chemical composition/density. The latter allowed to reveal nanoscale pores under the surface of the nanobump that can be also visualized by producing its FIB cross-sectional cuts (
Figure 1d).
At elevated laser fluence (
J/cm
), rupture of the nanobump led to formation of the through hole in the metal film. In this case, the multiple nanopores can be identified within the resolidified rim surrounding the microhole as revealed by SEM visualization of the FIB cuts (see
Figure 2a). Generally, for the fixed composition of the Au film, higher laser fluences produced the nanopores of the larger size. Similarly, the larger nanopores typically appeared closer to the nanobump center (see
Figure 2d–e). Taking into account the Gaussian-shaped intensity profile of the irradiating laser beam, the nanopore formation appears to be driven by the local temperature (that will be discussed later in the text) that is higher in the metal film section coinciding with the beam center. To enrich information regarding the density and geometry of the nanopores, multiple FIB cuts were merged to build exact 3D model of through hole and the surrounding rim (
Figure 2b,c). This 3D model clearly indicates broad size distribution of the nanopores ranging from 10 to 150 nm. Also, average size of the nanopores in the rim increases towards the center of the microhole. The larger nanopores can have irregular shape and reach the size ≈150 nm (
Figure 2c), while the smaller nanopores far from the rim walls preserve spherical-like geometry.
In part, broad size distribution of the nanopores could be explained by merging (coalescence) of the closest nanopores growing from neighbouring randomly distributed nucleation centers. Origin of such nucleation centers will be discussed somewhat later in the paper in the context of chemical composition of the metal film fabricated with different discharge gases. Earlier coalescence was suggested as a leading mechanism of the water bubbles growing on dispersed gold nanoparticles heated by incident laser radiation [
36,
37]. As two bubbles with radii
r merge, the gain in the surface energy
compensates the work of viscous forces needed to move the melt over a distance
at a velocity, which can be estimated as
. Here,
t is the time needed for this melt relocation and
is the surface tension [
38]. Using Newton’s law of viscosity to calculate the energy dissipation force
, where
—the surface area of the bubble,
—viscosity of liquid gold [
38], and
—the characteristic length scale of the flow, we estimate the energy loss as
. Equalizing
and
we estimate the time for two bubbles to merge
ps (estimated for
nm), which is less than the time gold film remains liquid.
This time can be estimated assuming that the heat accumulated by the illuminated spot is removed mostly by the thermal conductivity. The heated volume can be estimated as a cylinder of the radius
m (the typical lateral size of the nanobump) and the height
m (the film thickness). The areas of the side
and the bottom
surfaces are comparable, but the thermal conductivity of gold
W/(Km) is larger than that of glass
W/(Km), so that only the heat flow through the side wall decreases the temperature of the molten surface. This heat flow can be estimated as
, where
K is the melting point of Au and
K. This heat flow should remove the energy accumulated by the surface in one pulse
, hence the characteristic cooling time scale is
, i.e., is also comparable to the laser pulse duration. This estimation also agrees with previously reported studies [
39]. Hence, coalescence of the bubbles provides a plausible explanation of the observed distribution of the pore sizes.
Noteworthy, different growth times of the pores can be also considered as alternative way to explain the large size variation of the nanopores. In particular, the pores started growing short after the onset of the melting had more time to develop than that ones started just before the resolidification. This assumption cannot be accepted because of the low growth velocity of the bubbles. Wang et al. [
37] reported that during the initial bubble nucleation phase the bubble size
R grows with time as
(we notice that at later stages in degassed water this dependency drops to
, whereas in air-equilibrated water
). Hence, to get one order of magnitude difference in the nanopore radii, the times should differ by six orders magnitude. If the earliest possible nucleation starts after ten picoseconds after the start of the laser pulse (time comparable with the electron-phonon coupling times in metals), then the smaller pores formation must start several microseconds later, which is impossible because the resolidification happens on a time scale of several nanoseconds.
In a similar way, the broad size distribution can not be explained by local temperature fluctuations (pores at hot spots grow quicker) as the
[
37]. Hence, three-orders-of-magnitude fluctuation of the local temperature in gold separated by a distances much less than one micrometer is obviously not possible. Spontaneous merging (coalescence) of several nanopores into the larger one can also stimulate rupture of the nanobumps and can explain formation of the through nanoholes previously reported for AuPd films processed with ns laser pulses [
40]. The average pore size was observed to grow with the laser intensity, as it was shown in [
15]. This can be explained by cooperative action of two mechanisms: (1) each pore grows more rapidly, since the temperature is higher and the pore radius
. (2) The metal film remains molten for a longer time and the coalescence-based growth is stopped by the resolidification after a longer time interval
, so that more pores can merge together. Assuming that the surface temperature is proportional to the intensity, we expect the average observed pore size to grow with the intensity.
Remarkably, the density and the average size of the nanopores started to grow from randomly distributed nucleation centers were found to depend on the type of discharge gas used for the film fabrication. More specifically, negligible amount of nanopores was found for the nanobumps produced on the surface of Au films sputtered with Ar (see
Figure 2d). The maximal density of the nanopores was observed for the Au film evaporated with nitrogen discharge gas, decreasing for the Au film produced in purified air (
Figure 2e,f). Taking into account same morphology and light absorbing characteristics of all mentioned Au films produced with different discharge gases, the amount of nitrogen in the film could be considered as a driving parameter that allows to control density of the nanopores. The molecular nitrogen can be bounded to the metal film surface as well as form chemically stable AuN phase upon magnetron sputtering as it was reported by several previous studies [
41,
42,
43].
XPS measurements were performed to understand the effect of the discharge gas used for magnetron sputtering on the resulting chemical composition of the Au films (see Methods for details). Analyses of the deconvolved high resolution spectra taken in the N1
s region revealed near-zero signal for films produced with Ar discharge gas. However, a pronounced signal for Au films fabricated with purified air and nitrogen (
Figure 3b–d) were also observed. Deconvolution of the latter two spectra revealed several peaks with their binding energies at ≈401, 399 and 397.3 eV. The first high-energy peak can be assigned either to carbonitride [
43] or to molecular nitrogen (that can be adsorbed on the metal film surface having multiple grains and cracks as well as be trapped beneath the surface [
44,
45]). The two remaining peaks can be attributed to chemically stable gold nitride phases AuN
as well as to oxynitrides [
41,
42,
43]. All of the different Au films demonstrated similar Au4
f XPS spectra with low-intense shoulders (marked by blue areas in the
Figure 3a) shifted towards larger binding energies near both low- (Au4
) and high-energy (Au4
) peaks. These shoulders can be indicative of stable chemical compounds like surface bonded carbon (AuC) or nitrogen (AuN
) as shown in previous studies [
43]. However, the similarity of the signals obtain from different Au films does not clarify the exact chemical nature of these low-intense shoulders where signals from different compounds could overlap. To confirm such peak assignment, the Au film sputtered with nitrogen was further annealed at 200
C for 2 h. Thermal annealing was expected to remove thermodynamically unstable compounds (such as oxynitrides) as well as most of the molecular nitrogen from the near surface layer probed by XPS. N1
s core-level spectra of the annealed Au film demonstrates the only remaining peak with a binding energy of 398.8 eV that can be presumably attributed to AuN
.
Considering the chemical composition of the Au film produced with different discharge gases we can suggest the following scenario regarding formation of the nanopores upon ns-pulse laser ablation. Such laser pulse with a near-threshold fluence rapidly heats up the Au film to temperatures above 10
K. Such temperature jump is expected to remove all molecular nitrogen from the exposed area as well as induce an explosive boiling of the N- or AuN
-rich areas of the film. The data regarding melting/boiling temperatures of gold nitrides is weakly discussed in a literature. For example, annealing at 90
C was shown to remove AuN
from the film [
41], while present studies clearly showed signature of AuN
remaining even after annealing at 200
C. Anyway, both melting/boiling temperatures of gold nitrides are expected to be much lower comparing to those for the pure gold. The light absorption should be initiated at the Au grain boundaries where also nitrogen adsorption should occur. The ionisation potential of Au is 9.23 eV while that of nitrogen is 14.53 eV (for O—13.62 eV). The laser pulse driven avalanche ionisation of gold is seeding the energy deposition which is evolving into run away ablation (melting, evaporation, ionisation).
The general geometry of the surface structure produced by direct laser ablation is defined by the laser irradiation parameters (fluence, pulse width and beam profile [
46]) as well as by the thickness/composition of the metal film [
33,
47]. Our results clearly show that advanced chemical engineering of the metal film composition gives additional degree of freedom allowing to modify morphology of the laser-printed structures at the nanoscale, namely, incorporate the nanopores and control their density. However, a large amount of the nanopores is typically hidden beneath the top metal shell of the nanobump that limits the potential range of practically relevant applications. To make morphology of the laser-printed nanobumps more functional, we applied gentle etching of the laser-printed nanobumps with unfocused Ar-ion beam schematically illustrated on the
Figure 4a. The processing parameters were calibrated to ensure gradual removal of the Au film without melting and deformation of the nanobumps geometry (see Materials and Methods). Two representative SEM images in the
Figure 4b,c compare nanoscale surface morphology of the nanobump before and after its etching with Ar beam for 20 min. As can be seen, etching reveals the multiple hidden nanopores making the nanobump shell surface perforated with multiple through nanoholes. Besides the rather random arrangement of the nanopores (nanoholes), general geometry of the produced nanosponges reproduces well from pulse to pulse (see
Figure 4d). From the simple geometrical consideration it is clear that the incident electromagnetic radiation can be efficiently absorbed by such nanosponges that produce enhanced electromagnetic fields via coupling to propagating and localized surface plasmons. Being combined with the mentioned broad size distribution of the nanopores, the 3D nanosponges with their lateral size of about visible-light wavelength are expected to be efficient for local EM field enhancement within rather broad spectral range spanning from visible to near-infrared. To illustrate this, we calculated squared electromagnetic field amplitude E
/E
(E
is the amplitude of the incident EM field) near the isolated plasmonic nanosponge pumped at 532, 632 and 1030 nm (see Materials and Methods). The cross-sectional E
/E
maps shown in the
Figure 4e indicate that the nanosponge supports densely arranged EM hot spots upon broadband EM excitation. Multi-fold enhancement of the EM field amplitude comes from coupling of the incident radiation to the localized plasmon resonances of the isolated nanoscale surface features on the top side of the Au shell (like nanopores, spikes and cracks clearly seen in the
Figure 4c) as well as plasmons propagating within overall micro-scale sponge geometry. Propagating plasmons can couple to or re-excite the localized ones appearing as the localized EM hot spots on the opposite size of the opaque 150-nm thick Au shell. Owing to weak radiate decay, the certain localized plasmon modes in such plasmonic nanosponges were found to survive for a long time producing substantial field localization and enhancement [
8]. Alternatively, most of the modes can dissipate via heating of the nanosponge. These features will potentially allow to use such nanosponges for various applications as thermo and nonlinear plasmonics (enhanced higher harmonic and white light generation), bio-imaging and visualization, photothermal conversion, sensing based on nonlinear optical effects as well as SEIRA- and SERS-based sensing [
48,
49,
50,
51,
52].