Introduction

The spectacular optical properties of nanoparticles (NPs) formed from noble metals have long been a subject of theoretical and experimental interest [1–4]. The optical properties of these NPs are ascribed to localized surface-plasmon resonance (LSPR). The LSPR arises from collective oscillations of the surface charge density, which are induced by incident electromagnetic radiation. The position and width of the LSPR depend on the particular material, its size, shape, as well as its interactions with the environment and the proximity of other NPs [5–8]. The excitation of a LSPR leads to the confinement of electromagnetic energy in regions at the particle surface, whereby strong and localized electromagnetic fields are created. One interesting property of these localized electromagnetic fields is their ability to enhance radiative rates in various linear and nonlinear optical processes at metal surfaces. This phenomenon has been utilized for the development of various “surface-enhanced” techniques, such as surface-enhanced Raman scattering (SERS) [9, 10], surface-enhanced one- and multi-photon fluorescence [11, 12] and surface-enhanced second-harmonic generation (SESHG) [9, 10, 13]. Gold and other metal NPs have been used in different technological, biological and medical applications. For instance, in solar energy conversion, reaction catalyses, biosensors, imaging microscopy and in medical diagnostics [9, 10].

Multi-photon absorption induced luminescence (MAIL) from noble metals was first observed in 1981 by Chen et al [13]. The luminescence was attributed to a radiative recombination of the electrons excited from the d bands to the s-p conduction levels above the Fermi level [14–16]. Other mechanisms have been suggested, such as radiating plasmons of NPs [17, 18] and a localized IR surface plasmon enhanced s-p intraband radiative transition [19]. It has also been discussed whether the observed luminescence is due to an interplay between these processes [20]. At present, however, the mechanism of luminescence from metal NPs is still a question under debate.

It has been observed that the MAIL from roughened surfaces is several orders of magnitudes stronger than that from flat surfaces. This is ascribed to a local-field enhancement caused by nano-sized curvatures at the metal surface [13, 14]. It has also been suggested that the main contribution to the observed nonlinear luminescence spectra emanates from surface atoms localised where the local fields are strongest [14]. This implies that the MAIL is much more sensitive to properties of the LSPR than its linear analogue. This theoretical model of local-field enhancement, was applied successfully for explaining the observed luminescence enhancement from silver nanoparticles of different size, excited at different wavelengths [21]. The enhanced MAIL was also observed from AuNPs embedded in SiO2 films, single AuNP aggregates on glass surfaces, as well as from AuNPs in solutions [17, 18, 22, 23].

Strong luminescent enhancement has been observed from aggregated and agglomerated noble metal NPs [6–8, 24–26]. In agglomerates, the near-field interaction between NPs leads to strong local electric fields in the vicinity of particle surface, so called “hot spots” [22]. These cause several multi-polar surface plasmon resonances, which are red-shifted as compared to the individual NP spectrum [6, 8, 21, 24, 26]. The near-field interaction is extremely sensitive to the interparticle distance, as well as the particle size and shape. One can therefore manipulate these parameters in order to maximize the luminescence enhancement. A notable example is the plasmonic interactions within a dimer of NPs, which is the simplest agglomerate system [6, 8, 17]. It was shown theoretically and experimentally that LSPR energies can change considerably with interparticle distances and particle sizes [6, 8, 24–26].

This article reports on the first observation of a MAIL enhancement from agglomerates of AuNPs by two orders of magnitude, as compared to the luminescence from individually suspended spherical nanoparticles in solution. The enhancement is reported as a function of excitation wavelength in the interval of 630–900 nm. Our observations support the theory, which states that the electromagnetic field enhancement is due to the coupling between the incident radiation and the LSPR of NP agglomerates.

Experimental Section

Sample Preparations

Chemicals

NH4OH (25% aqueous solution) from Scharlau, 2-propanol and N-cetyl-N, N, N-trimethylammoniumbromide (CTAB) from Merck, tetraethylorthosilicate (TEOS, 99.999%) and sodium borohydride (99,9%) from Sigma-Aldrich, trisodium citrate dihydrate (Na-cit, ≥99,5%) from Riedel-de Haёn, HAuCl4.aq (≥49% Au) from Fluka. All chemicals were used as received. The water used for all preparations were run through Milli-Q water purification system with a resistivity of at least 18 MΩ·cm.

Preparation of Gold Nanoparticles

In the preparation of pure gold nanoparticles, gold salt was reduced by sodium citrate, as previously described [27], but with the difference that the surfactant CTAB and the base NaOH was present during the preparation. Shortly, 18 mL of a 0.1 M CTAB solution was mixed with 0.5 mL 0,01 M HAuCl4·H2O, 0.5 mL of 0.01 M trisodium citrate and 0.5 mL 0.1 M NaOH, all aqueous solutions. To this, 0.5 mL of freshly prepared 0.1 M NaBH4 solution was added all at once and mixed before left undisturbed for some time. The solution was kept in a water bath maintained at 25°C. This preparation resulted in gold nanoparticles with a diameter of 4–6 nm.

The gold nanoparticles used for silica coating was prepared by another method [28]. Shortly, 200 mL of 0.24 mM HAuCl4 in H2O was heated to 80°C before 0.94 mL of 0.34 M Na-Cit in H2O was added under vigorous stirring. The colour turned wine red within 15 min and the solution was left to cool after 30 min. This method yields gold nanoparticles of average diameter 15 nm.

Preparation of Silica Coated Gold Nanoparticles

The silica coating of the gold nanoparticles was performed as previously described by E. Mine et al. However, 2-propanol was used instead of ethanol. The synthesis was performed using a method where a mixture of water/gold particles/2-propanol was stirred vigorously before the addition of first TEOS and second aqueous ammonia, which initiated the silica coating. The samples were left with stirring for 24 h in room temperature. The suspensions were then washed and centrifuged three times with 2-propanol and water, respectively, before re-suspended in water. Centrifugation was performed at 20°C, 10 000 rpm and 10 min. The thickness of the silica layer was controlled by varying the amount of TEOS.

Sample Characterisation

Properties of the four samples (I–IV) studied are given in Table 1. The transmission electron microscopy (TEM) images of the NPs are shown in Fig. 1. Samples I and II consist of spherical naked AuNPs (diameter 5 ± 1 nm) and SiO2 coated AuNPs (diameter 12 ± 4 nm), respectively. The TEM image of II reveals the presence of a small number of dimers (cf. Fig. 1b). However the concentration of dimers is negligible in comparison to the abundance of individually SiO2coated NPs. Samples III and IV contain suspended SiO2-coated agglomerates of AuNPs with the mean core diameter of 12 ± 2 nm and a coating thickness of 170 ± 20 nm and 300 ± 35 nm, respectively. The average number of AuNPs in a silica particle core is about two times higher in sample IV, as compared with sample III. It can be assumed that the studied agglomerates are intrinsically 3D structures, although this cannot be observed from an inspection of the 2D TEM images.

Table 1 The shorthand notation and description of the AuNP samples studied in this work
Full size table
Fig. 1
figure 1

TEM images of the studied samples I (a), II (b) III (c) and IV (d). The characteristic properties of the samples are summarised in Table 1

Full size image

Due to the sample heterogeneity, it was difficult to determine the exact AuNP concentrations. In order to avoid concentration dependence effects, the samples were diluted to the same maximum OD(520 nm) = 0.1 in 1.0 cm cuvettes. The detected MAIL intensity from all samples depends linearly on the particle concentration.

TEM

Formvar-coated 200 mesh Cu/Ni grids were placed on Parafilm and 20 uL of the sample was put on top of the grids for 30 min to sediment particles by gravity to the film. The grids were then removed, excess sample was dried by pointing the edge of the grid to a filter paper and final drying was done for at least 6 h in room temperature before microscopy. All samples were analyzed in a JEOL JEM 1230 transmission electron microscope and appearances of the samples were assessed with magnifications between 80.000× and 150.000×. Measurements of particle sizes were performed both directly in the microscope and from saved images.

Apart from the four samples discussed in this article, other samples were examined during the course of experiments. Those samples were composed of naked and silica coated individual spherical AuNPs. The MAIL properties of these samples were similar to those discussed in this article (cf. Table 1). For clarity reasons, the results obtained from these samples have not been reported here.

In order to estimate the strength of the MAIL from AuNPs, one can compare with a standard two-photon reference. Here a 1 μM methanol solution of Rhodamine 6G was used, which has a two-photon absorption cross-section of 65 Göppert-Mayer [29]. In the present study of samples III and IV (λ exc = 800 nm) the MAIL intensity was weaker by two orders of magnitude.

Experimental System

The laser source was a 1 kHz Ti:Sapphire amplifier (Legend, Coherent) pumped by a frequency doubled Q-switched Nd:YLF laser (Evolution-30, Coherent) and seeded by a mode-locked Ti: Sapphire oscillator (Vitesse-800, Coherent). The output at 800 nm was directed to an optical parametric amplifier (TOPAS, Light Conversion). The excitation wavelengths between 630 and 900 nm were obtained by frequency doubling of the TOPAS signal which was tuneable in the range of 1260–1800 nm. The excitation beam was linearly polarised; the pulse duration and the spectral width were dependent on the wavelength, and was measured to be in the range of 160–210 fs, and 9–15 nm, respectively. Care was taken to eliminate the IR idler and signal beams generated by the TOPAS and propagating along the excitation beam. Colour glass filters were used to block the IR and VIS super-fluorescence of the TOPAS output.

The excitation beam was focused into the sample, which was contained in standard fused silica cuvette. The emission from the samples was collected at right angle to the excitation beam by a bi-convex lens and using a monochromator (ARC SpectraPro-150, 1800 g/mm density and blaze wavelength of 500 nm) equipped with a photomultiplier tube detector (R928, Hamamatsu). A BG42 glass filter was used to suppress scattered laser light. In order to improve the signal-to-noise ratio the PMT output was analyzed by means of gated integrator and boxcar averager (Stanford Research System). White light generation from a particle suspension could be observed when the pulse energy was about 200 nJ and higher. Hence all the measurements reported here were performed at excitation energies lower than 200 nJ/pulse.

The time-resolved measurements of the MAIL were carried out by means of a streak camera detection (C5680, Hamamatsu). The excitation source was a 200 kHz amplified Ti:Sapphire operating at 800 nm with pulses width of 120 fs [30]. Extinction spectra were recorded on a UV-VIS-NIR spectrophotometer (Cary 5000, Varian).

Results

Extinction Spectra

The extinction spectra of samples I–IV are displayed in Fig. 2. Sample I shows a distinct surface plasmon extinction band, with a peak maximum at approximately 523 nm. The rising part of the spectra at shorter wavelengths is due to the onset of interband transitions between the 5d and 6s-p bands near the L point of the band structure [18]. Sample II exhibits a similar extinction spectrum as sample I, but also a weaker band centred at approximately 700 nm, which can be attributed to the presence of a small amount of agglomerated AuNPs in the core (cf. Fig. 1b). The broadening of the plasmon absorption band linewidth in sample I in comparison to II is likely a manifestation of plasmon oscillation damping due to scattering of electrons at the NP interface [31].

Fig. 2
figure 2

Extinction spectra of the studied samples: I (dashed curve), II (dotted curve), III (dash-dotted curve) and IV (solid curve). All the curves were normalised to unity at the maximaum of the corresponding LSPR peak centred at approximately 520 nm

Full size image

The extinction spectra of samples III and IV are dominated by Rayleigh scattering. For sample III, two LSPR bands are observed, with peak maxima centred at approximately 520 and 800 nm. These bands can be explained by an agglomeration, as was also detected in sample II. The thicker SiO2-coating (300 ± 35 nm) in Sample IV causes a stronger Rayleigh scattering, which makes it difficult to localise the positions of the LSPR peaks. This and the large diversity of the agglomerated structure in sample IV could account for the lack of additional features in its spectrum.

Photoluminescence Properties

Typical MAIL emission spectra from the samples are displayed in Fig. 3. These spectra were not corrected for the wavelength dependent sensitivity of the detectors. The spectra were broad with a peak maximum centred approximately at 440 nm. None of the samples exhibited any signal arising from hyper-Rayleigh scattering, or second-harmonic generation. This suggests that contributions from these second-order nonlinear processes are much weaker than the MAIL. A shoulder observed at ca. 600 nm was caused by Rayleigh scattering of the excitation light and transmission properties of the BG42 glass filter. No significant change in the shape of MAIL spectra obtained from different NPs was observed. Neither the excitation power nor the excitation wavelength did significantly influence the MAIL spectra. Previously an emission spectrum identical to those shown in Fig. 3 was reported by Zhu et al [16]. In their study AuNPs were excited with the 1064 nm laser beam, and the observed emission was attributed to three-photon absorption induced process.

Fig. 3
figure 3

MAIL emission spectra obtained for sample I (dashed curve), II (dash-dotted curve), III (dotted curve) and IV (solid curve) and the excitation wavelength 630 nm

Full size image

The dependence of the integrated MAIL intensity (I MAIL) on the averaged excitation power (W A) suggests a quadratic dependence for low excitation power in the wavelength region 450–530 nm. At higher excitation power, the slope level off, as shown in Fig. 4. A similar behaviour was reported on similar systems by others [16]. Power dependent measurements at shorter wavelengths (290–390 nm) indicate a slope of 3.4 ± 0.9 upon exciting at 850 nm. This clearly indicates that the observed luminescence is due to a three-photon process (cf. Fig. 4).

Fig. 4
figure 4

Dependence of the integrated MAIL intensity, I MAIL, on the averaged incident power, W A. The data correspond to the excitation wavelength 850 nm and for sample III. Data points denoted with squares correspond to the intensity I MAIL, calculated for the wavelength range 450–590 nm, while data points with circles refer integrated intensities for the range 290–390 nm. I 1 is the reference intensity. Next to each straight line, its slope is given

Full size image

The temporal profile of the MAIL spectrum was measured by means of streak camera detection. For all studied AuNPs, the time-resolved MAIL intensity resembles the response function of the streak camera, as obtained from Rayleigh scattering of AuNPs (FWHM = 12 ps). This suggests that the MAIL is practically instantaneous. Indeed, the MAIL from AuNPs has been suggested as a very versatile instrumental response function for multi-photon excited luminescene lifetime experiments with ps time resolution, such as TCPSC, streak camera detection and fluorescence lifetime imaging [30]. A very short lifetime of the MAIL has also been reported by researchers [18, 20, 32].

The obtained MAIL efficiency (E MAIL(λ exc)) at different excitation wavelengths (λ exc) is exemplified in Fig. 5. The efficiency was calculated from the measured integrated MAIL intensity (I MAIL) and the incident excitation power (W A) according to:

Fig. 5
figure 5

Normalised MAIL efficiencies for samples I (squares), 2 (triangles), III (circles) and IV (stars). All the curves were normalised to unity at their corresponding maxima at 630 nm. Error bars correspond to two standard deviations of measured data points

Full size image
$$ {E_{\rm{MAIL}}}\left( {{\lambda_{\rm{exc}}}} \right) = {I_{\rm{MAIL}}}W_{\text{A}}^{ - 2}. $$
(1)

In Eq. 1, the efficiency is normalized to an assumed two-photon excited process by dividing the luminescence signal by the square of the incident power. Here the used value of I MAIL corresponds to the integrated intensity over wavelengths between 450 and 550 nm. For this range, the observed logarithmic power dependence is compatible with a two-photon process (Fig. 4). In order to examine different wavelength dependencies of the MAIL efficiency (for samples I–IV), the calculated efficiency curves were normalized to the sample´s maximum luminescence at λ exc = 630 nm, thereby removing the scaling with respect to particle concentration and MAIL intensity (cf. Fig. 5). No particular physical mechanism motivates this normalization, which was only introduced for the sake of clarity. Depending on the samples compared, absolute MAIL intensities from the agglomerated structures were in average up to 1 order of magnitude higher than those from individually suspended coated or uncoated AuNPs for the excitation wavelength 630 nm.

As can be seen in Fig. 5, samples I and II with individually suspended AuNPs show a rapid decrease of E MAIL with increasing λ exc, resulting in a negligible emission at λ exc > 750 nm. On the other hand, the MAIL signal from the samples with agglomerates (III and IV) is significantly stronger over the whole range of excitation wavelengths. Furthermore, the MAIL intensity of sample IV remained rather constant over this spectral range. Thus, the MAIL efficiency is strongly dependent on the presence of NPs agglomerates. Moreover, the thickness of the silica coating did not affect MAIL intensity wavelength dependence. What is, however, the enhancement of the MAIL from agglomerated AuNPs relative to individual AuNPs at different excitation wavelengths? For illustrating this, a wavelength dependent enhancement factor η(λ exc) is introduced as:

$$ \eta \left( {{\lambda_{\rm{exc}}}} \right) = {{{E_{\rm{MAIL}}^{\rm{aggl}}\left( {{\lambda_{\rm{exc}}}} \right)}} \left/ {{E_{\rm{MAIL}}^{\rm{indiv}}\left( {{\lambda_{\rm{exc}}}} \right)}} \right.} $$
(2)

where \( E_{\rm{MAIL}}^{\rm{aggl}}\left( {{\lambda_{\rm{exc}}}} \right) \) denotes the normalized MAIL efficiency from either sample III or IV, and \( E_{\rm{MAIL}}^{\rm{indiv}}\left( {{\lambda_{\rm{exc}}}} \right) \) stands for the normalized MAIL efficiency from either sample I or II. Prior to calculating η(λ exc), the experimental curves were approximated by polynomials of the 11th degree, which smoothes data and compensates for different choices of excitation wavelengths during different measurements. The resulting η(λ exc) graphs are displayed in Fig. 6.

Fig. 6
figure 6

The wavelength dependence of the enhancement factor η(λ exc). The graphs were obtained as outlined in the text (Eq. 2). by dividing efficiencies from different samples in Fig. 5 as follows: \( E_{\rm{MAIL}}^{\rm{III}}/E_{\rm{MAIL}}^{\rm{I}} \) (triangles), \( E_{\rm{MAIL}}^{\rm{III}}/E_{\rm{MAIL}}^{\rm{II}} \) (squares), \( E_{\rm{MAIL}}^{\rm{IV}}/E_{\rm{MAIL}}^{\rm{I}} \) (stars) and \( E_{\rm{MAIL}}^{\rm{IV}}/E_{\rm{MAIL}}^{\rm{II}} \) (circles). Standard deviations are shown for selected data points

Full size image

Taken together, a MAIL intensity enhancement by at least two orders of magnitude was calculated for samples III and IV relative to sample II at excitation wavelengths between 750 and 850 nm. A somewhat weaker enhancement was obtained for samples III and IV when compared with sample I.

Discussion

For the samples composed of individual nanoparticles (I and II), the MAIL efficiency increases with decreasing excitation wavelength and reaches it highest value at λ exc = 630 nm (cf. Fig. 5), which is the shortest wavelength used in the present study. It is suggested that the near overlap between the laser excitation energy and the LSPR band of single AuNPs is the determining factor for the increased emission intensity at higher energies, i.e. at shorter excitation wavelengths. It is known that the spectral overlap increases the luminescence efficiency, particularly in the case of multi-photon excitation [9, 14, 19, 23]. In this case, the wider absorption band of sample I, relative to sample II, can explain the stronger MAIL from the former system (cf. Fig. 5).

The observed wavelength dependence of MAIL efficiency differ for the NPs agglomerates, as exemplified by samples III and IV. Similar to what was observed for samples I and II, the MAIL increases at shorter excitation wavelengths. However, there is no significant reduction in luminescence at longer excitation wavelengths, where the MAIL efficiency becomes almost wavelength independent (Fig. 5). Again it can be suggested that the nearly constant efficiency, as a function of wavelength, is due to the overlap of excitation energies with a modified LSPR of the agglomerated NP system [5–8, 23, 25, 26].

For nanoparticles, the position and width of the LSPR band depend on the NP size, interparticle distance, number of interacting particles, as well as their relative positions in space. The appearance of LSPR broad bands in the red and near IR part of the spectrum for agglomerated NPs was predicted theoretically and observed experimentally [5–9, 23–26]. For instance, the broad band centred at about 800 nm in the extinction spectrum of sample III (cf. Fig. 2) is attributed to an agglomerate surface plasmon excitation. Therefore, a coupling between this band and the radiation field could contribute to the observed enhancement of the MAIL from agglomerates. Similar conclusions have been drawn for SERS enhancement by similar systems of AuNP aggregates [25].

A similar dependence of the MAIL intensity on λ exc and LSPR was observed from AuNPs embedded in a solid surface coated with SiO2. In these studies, the luminescence efficiency also showed a dependence on the excitation wavelength [23]{22}.

The wavelength dependence of MAIL efficiencies displayed in Fig. 5, and the calculated enhancement shown in Fig. 6 agree qualitatively with the calculations of the local electromagnetic field enhancement for various interparticle separation between two AuNPs within a dimer [25]. For particles separated by 5 nm, calculations predict a drop in the field enhancement by several orders of magnitude in the range from 630 to 900 nm. This is a result of a weaker coupling between the individual AuNPs and the incident field. For separations less than 1 nm, field enhancement as a function of λ exc exhibits broad peaks that corresponds to multipolar resonances in the extinction spectrum of the particles. An averaging with respect to a large number of agglomerated particles and interparticle distances might result in a wavelength independent local field enhancement in the red and near IR spectral regions. The local field enhancement of agglomerates is by several orders of magnitude stronger than that of individual AuNPs. This behaviour qualitatively agrees with the results presented in Fig. 6. The maximum obtained enhancement in the region 800–850 nm is likely correlated with the observed extinction band in sample III (cf. Fig. 2). Thus, the calculated local field enhancement [25] correlates with the experimental enhancement obtained in the present work.

Three photon absorption induced luminescence from AuNPs has been reported upon excitation by NIR radiation [20, 23]. The high-energetic part of the emission spectrum, i.e. for the luminescence of shorter wavelengths, can be explained by a three photon excitation (cf. Fig. 4). It has also been proposed by Mooradian [15] that the electron and hole distributions are thermally smeared out and that the photo-excited holes are scattered into a broad range of states resulting in luminescence at shorter wavelengths.

Metal NPs are known to be a source for various nonlinear optical processes. In addition to the luminescence reported in this article, other processes such as SHG, sum-frequency generation (SFG), third-harmonic generation (THG) and four-wave mixing (4WM) have been observed [33, 34]. These processes are also interesting and can find applications in bio-sensing and microscopy due to their directionality and large spectral shift from the fundamental beam(s). It has also been observed that NPs in contact can exhibit a very intense field enhancement where the localized field can give rise to a strong nonlinear response [32].

In our experiments, using the TOPAS as the excitation source, we have observed several nonlinear processes from AuNPs. For example, when the idler wavelength equals 2λ exc, the generated signal appeared at a wavelength that could correspond to several physical processes, such as degenerated 4WM, SHG (involving idler and fundamental beam), THG of the idler and higher diffraction orders of Rayleigh scattering of the idler. These processes were also observed to be stronger from samples with the thickest silica coatings and agglomerates. Since the main subject of this paper is MAIL, care was taken to eliminate these processes by blocking the idler and signal beams generated by TOPAS, which propagate along the excitation beam used to generate MAIL. In many cases, these nonlinear processes can be much stronger than the MAIL, and could consequently lead to wrong interpretations of experimental results.

Conclusions

The MAIL intensity has been studied from naked individual, silica coated individual and agglomerated AuNPs suspended in water as the function of excitation wavelength. A strong MAIL efficiency from silica coated agglomerates was observed, which was weakly dependent on the excitation wavelength, while a rapid decrease in the MAIL intensity with the increasing excitation wavelength was observed from individual AuNPs. From the obtained data it was possible to determine the MAIL enhancement from agglomerates relative to individual NPs, as a function of the excitation wavelength. The maximum enhancement factor of at least two orders of magnitude was observed at the excitation wavelength between 800–850 nm. The results suggest that agglomerates exhibit plasmon excitations at wavelengths above 700 nm, which correlates with stronger MAIL. The magnitude and wavelength dependence of the enhancement agrees qualitatively with recent theoretical models available.