Coherent Vibrational Oscillation in Gold Prismatic Monolayer Periodic Nanoparticle Arrays

We studied the ultrafast laser-induced coherent phonon oscillation in prismatic shaped gold nanoparticles assembled in monolayer periodic arrays by using the nanosphere lithographic technique. The amplitude and phase of the oscillation observed by ultrafast pump−probe transient spectroscopy is monitored as the wavelength of the dipolar surface plasmon absorption decreases. At a certain wavelength, the oscillation could not be observed. As the monitoring wavelength decreases further, the sign of the amplitude changes. From the wavelength at which the oscillation is not detected, the dependence of the absorption maxima on the size of the nanoparticles, the changes in the nanoparticle size are estimated during its oscillation. This large change in the size of the prismatic nanoparticle compared to the small change reported previously for the nanosphere oscillations is discussed. Ultrafast laser-induced phonon oscillations have been observed for different shaped noble metal nanoparticles and also in different environments. They have been observed in spherical nanoparticles such as silver nanoparticles in glass matrix1 and gold nanoparticles in colloidal solution.2-4 Phonon oscillations have also been seen in ellipsoidal silver nanoparticles in glass matrix5 and cylindrical gold nanoparticles in colloidal solution.6,7 Oscillations have also been found in aggregated gold particles in colloidal solution.8 These coherent oscillations have been attributed to the symmetric acoustic lattice vibration (periodic size change) of the metal nanoparticles caused by ultrafast laser-induced heating. The lattice vibration of the metal nanoparticles induces a periodic shift of the plasmon absorption band, which causes a corresponding oscillation in the transient absorption intensity if it is monitored at a fixed wavelength. The frequency of this oscillation is found to be inversely proportional to the dimension of the nanoparticles, which can be predicted accurately by classical mechanics calculations.9 To observe these oscillations and their dependence on the size of the metal nanoparticles, a very monodisperse sample has to be used, since a polydisperse sample gives oscillations with different periods, which would “wash out” the oscillatory signal. A narrow particle size distribution can be achieved with special techniques such as radiation chemistry,10,11 tensile deformation,12 and improved wet chemistry methods.7 Another way to make nanoparticles of the narrow size and shape distribution is by using lithographic techniques.13 Nanosphere lithography, originally referred to as “natural lithography”, was introduced by Fischer and Zingsheim in 198114 and improved by Deckman et al.15 This technique is an inexpensive, inherently parallel, and highthroughput technique. It is an ideal technique to use for studying size-dependent properties. This technique was renamed nanosphere lithography (NSL) by the Van Duyne group,13,16-18 who made great contributions. They extended the original monolayer NSL technique to double layer NSL16 and angle-resolved nanosphere lithography (AR NSL) techniques.18 Van Duyne et al. also used atomic force microscopy (AFM) to study these structures.16 They also made different sized truncated tetrahedral silver nanoparticles that absorb in different wavelength ranging from the visible to the infrared region.17 Recently, large-scale 2D periodic carbon nanotubes19,20 and aligned ZnO nanorods21 were also successfully synthesized using NSL technique. In the present study, we report on the ultrafast laserinduced coherent phonon oscillation in a periodic monolayer of truncated tetrahedral gold nanoparticle arrays prepared by NSL technique. Due to the strong absorption of the dipolar surface plasmon oscillation and the sensitivity of the plasmon resonance frequency to the size of the truncated tetrahedral nanoparticles with sharp tips, it was possible to obtain excellent oscillation amplitudes from a monolayer array of these nanoparticles with good signal-to-noise ratio. In addition, due to the sharp tips of the prismatic shaped nanoparticles, large changes in the size are observed during the coherent lattice oscillations. * Corresponding author. E-mail: [email protected]. NANO LETTERS 2004 Vol. 4, No. 9 1741-1747 10.1021/nl048875p CCC: $27.50 © 2004 American Chemical Society Published on Web 08/10/2004 Experimental Section. Periodic arrays of gold nanoparticles are prepared using the method of NSL that was developed by Van Duyne and his group16 and slightly modified by Wang et al.21 The commercially available monodispersed polystyrene (PS) sphere suspensions are purchased from Duke Scientific Corp. and used as received. The diameters of the spheres used in our experiments are 0.45, 0.60, and 0.74 μm. Quartz slides (Technical Glass Products, Inc.) are used as the substrate and are cleaned in piranha solution (3:1 H2SO4/30% H2O2) at 80 °C for 1.5 h. The quartz slides are placed in a solution of 5:1:1 H2O/NH4OH/30%H2O2 and are sonicated for 1 h. Five microliters of the PS spheres suspension solution is deposited on the quartz slides, which are tilted to disperse the suspension solution evenly on the substrate. The substrate is slowly immersed in water. Extra PS spheres are dispersed in water. One drop of 2% dodecylsodiumsulfate surfactant is then added to water to separate the excess floating PS spheres from the substrate, and the substrate is taken out from the clean area of the water in order to prevent additional deposition of the PS spheres. After the evaporation of the water, the PS sphere monolayer forms and covers several square millimeters. Scanning electron microscopy (SEM) is used to check the formation of the PS sphere monolayer mask, and a typical image is shown in Figure 1a. The mask is then mounted in a thermal evaporator (Denton DV-502A) to deposit 50 nm thick gold (99.999%, Alfa Aesar) in the spaces between the spheres. The thickness of the deposited gold is monitored by a quartz crystal thickness monitor (Inficon). After deposition, the PS spheres are dissolved in a tetrahydrofuran (THF) solution. The gold arrays are checked by using SEM, and the resulting images are shown in Figure 1b-d. Figure 1. SEM image of (a) highly ordered self-assembled monolayer of PS spheres on quartz substrate (the sphere has a diameter of 0.60 μm). (b-d) Periodic gold particle monolayer arrays (shown in gold color) produced with 0.45 μm, 0.60 μm, and 0.74 μm PS spheres, respectively (black spheres represent the positions of the PS spheres mask used during the Au deposition). (e) AFM image of the gold array particles made with 0.74 μm PS spheres. (f-h) Absorption spectra of the periodic array samples made with the 0.45 μm, 0.60 μm, and 0.74 μm PS spheres, respectively. Red line in (f) is microabsorption spectrum of a 3.75 × 3.75 μm2 area on the array sample made with the 0.45 μm PS spheres. The difference between the red and black spectra in (f) is a result of a small amount of broadening in the macroabsorption spectrum due to inhomogeneity. The spectra shown in blue dots in (f-h) are fits of the original absorption spectra to Lorentzian shapes. 1742 Nano Lett., Vol. 4, No. 9, 2004 A Shimadzu UV-3101-PC spectrophotometer is use to record the absorption spectra of the arrays. The incident light is perpendicular to the array substrate and the incident light bandwidth is 2.0 nm. The examined area of the nanoparticle array samples is several square millimeters, which allows us to check the absorbance of the nanoparticle arrays on a larger scale. The overall sample quality can be evaluated in this way because the intensity of the surface plasmon resonance absorption peak reflects the number of nanoparticles in the checked area of the array. The microabsorption spectrum is taken by a SEE 1100 microspectrometer in the transmission mode, and the examined area is 3.75 × 3.75 μm2. With the microspectrometer, the sample area can be viewed directly, which allows us to record the absorption of the array of nanoparticles and avoid most of the large defects. In the femtosecond transient absorption experiment,22 a frequency-doubled Nd:vanadate laser (Coherent Verdi) was used as the pump for the Ti:sapphire laser system (Clark MXR CPA 1000). This generates laser pulses of 100 fs duration (fwhm) with energy of 1 mJ at 800 nm at a repetition rate of 1 kHz. The second harmonic of the 800 nm fundamental at 400 nm was used as the excitation (pump) source. The pump beam was mechanically chopped with a light beam chopper (HMS 221). The diameter of the laser focus spot on the sample was 250 μm. The laser pump pulse energy used in our experiment is reduced to less than 250 nJ with neutral density filters. A white-light continuum probe is generated by focusing a small portion (4%) of the 800 nm fundamental beam of the Ti:sapphire laser into a 1 mm sapphire plate. The differential transmission signal S(t) is recorded with a pair of silicon photodiodes (Thorlab) and a lock-in amplifier (Stanford Research Systems). The recorded signal S(t) can be expressed as where ∆T/T is the % change in the transmission of probe light, Iλ,t is the intensity of the probe light at wavelength λ after a delay time t from the pump laser heating pulse, and Iλ,0 is the intensity of the probe light at λ without the pump. As a result, the recorded signal represents a transient bleach when S(λ,t) is larger than zero. Results and Discussion. Absorption Spectra. Figure 1b-d are SEM images of the hexagonal patterned gold array particles (shown in gold color) that are produced with different size PS spheres. The black spheres in Figure 1b-d represent the positions of the etched PS sphere masks that are used during the Au deposition and are etched away in tetrahydrofuran (THF) solution after Au deposition. In this planar view, the particles are approximately triangular. The AFM image (Figure 1e) shows the actual shape of the particles produced by NSL technique to be truncated tetrahedron.23 The size of the particle is represented by the bisector, a, of the triangle. The bisectors of the particles are 104, 138, and 166 nm for the sample made with the 0.45, 0.60, and 0.74 μm PS spheres, respectively. The size distribution of the particles is very narrow with a standard deviation between 5 and 8% (excluding the large defects when the long axis is larger than 2a). Although there are some defects on the sample caused by misalignment of the PS spheres that form stacking defects of the mask, the size of the defects is far away from the average size of the array particles examined. Jensen et al.23 have shown that the defects on the sample have negligible effect on the plasmon absorption band by comparing the microand macroabsorption spectra of their silver array samples. Figure 1f-h shows the absorption spectra of different sized gold nanoparticles. The red spectrum in Figure 1f is the microabsorption spectrum and the black spectrum is the macro spectrum of the array sample with the particle size a equal to 104 nm. For this array, the observed surface plasmon absorption maxima for the two spectra (the black and red) in Figure 1f are around 742 nm. In the macroabsorption spectra, there is a dip around 500 nm which result from the gold film24 or large defects formed during the sample preparation. This dip is not observed in the microabsorption spectrum since a spot that is free of defects has been selected for the absorption measurement. As shown in Figure 1f-h, as the size of the particle increases, the surface plasmon resonance absorption band red shifts. This type of dependence was used by Jensen and co-workers17 to make silver arrays with tunable surface plasmon resonance absorption maxima. The macroabsorption spectra also show a weak band around 550-600 nm on the blue side of the main plasmon resonance band, which is observed very clearly in the microabsorption spectra. Compared to the macroabsorption spectrum (in which the area examined was larger than several mm2), the microspectrum is much sharper and seems to eliminate most of the inhomogeneous broadening due to the defects in the array sample. This is supported by the fact that the fit to Lorentzian line shape (blue dots) is much better for the microthan the macrospectrum. Since the microspectrometer is capable of recording only the sample absorbance between 400 and 850 nm, we could not get the whole absorption spectrum for the large array samples (134 and 166 nm). These macrospectra are broader than the Lorentzian fit (blue dot) indicating the presence of some inhomogeneous broadening. The weak surface plasmon band at 550 nm observed in the microabsorption spectrum was also observed by Mohamed25 from triangular gold nanoplatelets. From the absorption spectrum and discrete dipole calculations, Jin and coworkers26 found that the plasmon resonance of silver prism nanoparticles, having an edge length of 150 ( 16 nm, had three absorption bands located at 340 nm (assigned to outof-plan quadrupole resonance), 600 nm (assigned to in-plane quadrupole resonance), and the strong band at 1065 nm (assigned to in-plane dipole resonance). Due to the similarity of the particle shape in our case, it is likely that the weak 550 nm plasmon resonance band in our gold array sample (104 nm particles) results from the in-plane quadrupole resonance transition, and the strong absorption bands shown in Figure 1f-h are the in-plane dipolar transitions. Figure 2 shows the dependence of the plasmon absorption maximum λmax on size for gold spherical nanoparticles3,27 and truncated tetrahedron array nanoparticles. This illustrates S(λ,t) ) ∆T T ) Iλ,t Iλ,0 Iλ,0 (1) Nano Lett., Vol. 4, No. 9, 2004 1743 a much more pronounced sensitivity of plasmon absorption maximum on size for the prismatic shaped nanoparticles compared to the spherical nanoparticles. For gold spherical nanoparticles, the plasmon absorption maximum increases by 0.64 nm for every 1 nm increase in the diameter of the sphere. For the truncated tetrahedron array nanoparticles, the plasmon absorption maximum increases by 4.4 nm for every 1 nm increase in the bisector of the particle. Phonon Oscillation Period. Figure 3a shows the laserinduced phonon oscillations for different sized gold nanoparticle arrays. According to this simple monolayer packing geometry (P6mm) and laser focus point area (4.91 × 10-4 cm2), we can calculate how many nanoparticles are excited by the pump laser pulse. About 5.2 × 104 gold array nanoparticles are irradiated by the pump laser for the nanoparticle array with a ) 166 nm. Even for the smallest nanoparticles array sample (a ) 104 nm), only 1.4 × 105 particles are irradiated. The concentration of the nanoparticle arrays is several orders of magnitude lower than that in the common experiment carried out in colloidal nanoparticles solutions. There are two reasons explaining the fact that we can observe these strong coherent oscillation signals. The first is the strong absorption coefficient and the high sensitivity of the absorption spectrum to size (Figure 2). The second is the very narrow size distribution of the nanoparticles in the arrays. The experimental oscillation signals in Figure 3a are fitted to a damped cosine function (dash line) in order to obtain the frequency and the phase of the modulations. The damped cosine function used has the form where A0 is the initial amplitude, τd is the damping time, T is the period of the oscillation and æ is the initial phase of the oscillation signal. In addition, a one component exponential decay is used to fit the slowly varying background in the decay trace. For the traces in Figure 3a, the initial oscillation amplitudes A0 are 0.090 for 104 nm particles, 0.135 for 138 nm particles, and 0.104 for 166 nm particles. The damping time τd are 102, 258, and 419 ps for the three different size nanoparticles, respectively. The initial phase æ of the three oscillation traces are 7.7°, 8.7°, and -8.3° for the three different size nanoparticles, respectively. The plot of the oscillation period T vs the nanoparticle size a is shown in Figure 3b. The period of the oscillation increases Figure 2. Dependence of the wavelength maxima (λmax) of the surface plasmon absorption spectra on the size of the gold spherical nanoparticles3,27 (b) and truncated tetrahedron (prismatic) array nanoparticles (2). This illustrates the much more pronounced sensitivity of λmax on size for the prismatic shape compared to the spherical nanoparticles. For gold spherical nanoparticles, the plasmon absorption maximum increases by 0.64 nm for every 1 nm increase in the diameter of the sphere. For the truncated tetrahedron array nanoparticles, the plasmon absorption maximum increases by 4.4 nm for every 1 nm increase in the particle bisector, a. Figure 3. (a) Lattice phonon oscillations induced in the prismatic gold nanoparticle monolayer arrays with a 100 fs laser pulse of 250 nJ at 400 nm and monitored close to the absorption maximum of each nanoparticle (solid line). The oscillation was fitted with a damped cosine function (dash line), from which the oscillation period is calculated. (b) Oscillation period verses particle size. The solid line is a linear fit of the experimental data. The dash line is the calculated linear relation between the period and the particle size (calculated from τ ) 2a/νl, where νl is the longitude sound velocity in bulk gold and a is the particle bisector). The longer observed period compared to the calculated one might result from damping due to the attachment of the particle to the substrate, from the presence of sharp tip that could make the vibration amplitude larger, and/or from particle-particle interaction. S(t) ) A0 exp(-t/τd)cos(2πt/T + æ) (2) 1744 Nano Lett., Vol. 4, No. 9, 2004 almost linearly with the size of the array particles (Figure 3b). For spherical nanoparticles, the vibration periods can be accurately calculated from the continuum elastic theory for both homogeneous28,29 and core-shell particles.30 However, for our truncated tetrahedron particles, there is no such analytical expression available. Perner et al.5 probed the aligned silver ellipsoid nanoparticles along the short (40 nm) and the long (100 nm) axes and found the period to be 22 ps for the short axis and 52 ps for the long axis. This is very close to the calculated values of 23 and 56 ps for the short and long axes respectively, which is obtained by using the approximation that the vibration mode of the silver ellipsoids is a one-dimensional standing acoustic wave with free boundary conditions (with period τ ) 2a/νl, where a is the width or length of the silver ellipsoids and νl is the longitude sound velocity in bulk silver). Using the same approximation, we can calculate the periods for our array particles. The calculated size dependence is shown by the dashed line in Figure 3b, using the value of νl ) 3240 m/s31 for gold and a ) 104, 138, and 166 nm, respectively. It can be seen that the observed oscillation period is longer than the calculated one. This is not unexpected since prismatic shapes with sharp tips are not like ellipsoidal shapes with smooth round ends. Theoretically, atoms on the sharp tips are found to give rise to large change in the plasmon absorption maximum with size.32 Another manifestation of the effect of these sharp tips is the larger amplitude of the vibration of the prismatic shape that is observed. This together with the fact that the particles are attached to a substrate, which is likely to damp the motion, will result in a vibration time that is relatively long, as observed. In addition, the observed oscillation is not from individual but rather coupled nanoparticles due to the interparticle interaction. Monitoring WaVelength Dependence of the ObserVed Amplitude and the Phase (sign) of the Oscillation. Figure 4 shows the dependence of the amplitude and phase of the observed oscillations of the monitoring wavelength on the blue side of their surface plasmon absorption spectra for both the 104 and 166 nm nanoparticles. As the wavelength decrease below the maximum, the amplitude of the oscillation decreases, reaches zero, then slowly increases again but with opposite phase. To understand this behavior and make use of it, we need to examine what determine the observed bleach intensity, ∆Iλ,t. In Figure 4, the percentage of change in the transmission ∆T/T is given by eq 1. The pump laser excites the surface plasmon electrons, creating an optical hole in its homogeneously broadened absorption. This increases the transmission intensity of the monitoring light with a profile that mimics the surface plasmon absorption spectrum. This is because a hole is created by reducing the full absorption profile as the absorption is homogeneously broadened.27 Very rapidly (in ∼ 1 ps) electron phonon relaxation occurs,27 which heats the lattice coherently. The transmission is reduced to the value observed at the beginning of the oscillation. The pulsed excitation of the lattice leaves it either in the extreme contracted or extreme expanded configuration of its phonon breathing mode (as well as in many other modes). Once the electron phonon processes rapidly (1 ps) prepare the hot coherently vibrationally excited lattice, it oscillates between two configurations, the contracted and expanded forms, with the frequency of the symmetric acoustic breathing mode. Thus, at any time (t), the wave function of the system can be written as a linear combination of the two forms This wave function suggests that the absorption spectrum is oscillating between the absorption spectrum of the contracted Figure 4. Wavelength dependence of the amplitude (the % transmission change of the monitoring light as a result of laser pump) and phase (the sign of the change) resulting from the coherent lattice phonon oscillation in a ) 104 nm (a) and a ) 166 nm (b) gold particle arrays. The observed oscillation is fitted to a damped cosine function (dash line). It is interesting to observe that as we decrease the monitoring wavelength of the bleach created by the 400 nm pump laser, the size of amplitude decreases, reaches zero value (i.e., the oscillation disappears), and then reappears with opposite sign (phase). From the wavelength at which the oscillation disappears, the absorption maximum, and its dependence on size, the change in the nanoparticle size during the oscillation is estimated (see text). Ψt ) acon(t)ψcon + (1 acon )ψexp (3) Nano Lett., Vol. 4, No. 9, 2004 1745 configuration (con) and that of the expanded (stretched) configuration (exp), going through that of the equilibrium configuration. This leads to changes in surface plasmon absorption spectra with the maxima at shorter wavelength for the contracted lattice and at longer wavelength for the expanded lattice, as compared to that of the equilibrium configuration of the nanoparticles. Under our experimental conditions, the observed light intensity change in transmission (recorded in the femtosecond bleach experiment) is less than 1%. So it is equal to the change of the sample absorption upon laser excitation. The observed signal at certain wavelength as a function of time thus results from the net absorption changes at this wavelength between a bleached equilibrium spectrum (induced by the pump laser) and the new oscillating absorption spectra of the two configurations. Thus the observed signal A, at a certain wavelength λ at time t, is given by