letters to nature prisms. Third, detailed time-dependent UV-vis -NIR measure- and results in exclusive formation of the smaller type I nanoprisms ments show that the onset of the growth of the band at 1,065nm (72+ 8 nm), as evidenced by UV-vis. -NIR spectra and TEM (assigned to type 2)is significantly delayed in comparison with the analysis(Fig. 3b, spectrum 4, and Fig. 3e). We further investigated indicates that the fusion of nanoprisms occurs only after type 1 found that a 550-nm/340-nm coupled beam, in which the 340-nm lanoprisms have accumulated. Fourth, a small population of dimer light coincides with the out-of-plane quadrupole plasmon of the 2 and trimer 3 intermediates(Fig. ld)is observed during the early type 1 nanoprisms, also can inhibit the formation of type 2 stages of type 2 particle growth(see Supplementary Information). nanoprisms and result in unimodal growth. However, in the cases We also performed electrodynamics calculations on the optical of 550-nm/395-nm, 550-nm/610 nI nm properties of possible intermediate species involved in the fusion coupled beams, in which the secondary wavelengths fall within growth process. The results show that intermediates 2 and 3 have the dipole resonances of the Ag nanospheres(395 nm)and type 1 dipole plasmon excitations close to 600 and 1,065 nm(see Sup- nanoprisms(610 and 650 nm), respectively, bimodal growth is plementary Information). Type I particles and the intermediates 2 observed(see Supplementary Information). These data strongly and 3 can thus all absorb light at 600 nm, which can lead to the indicate that only secondary wavelengths that can excite quadrupole excited state needed for particle fusion to occur. However, the type 2 plasmon modes can inhibit bimodal growth. Indeed, it is this particles do not show dipole plasmon excitation explaining why photo-cooperativity that leads to the results observed with a ontext that removal of surface ligands has, in the case of CdTe(ref. of a fluorescent tube exhibits bands at 546 nm and 440 nm, and has 27)and PbSe(C. B. Murray, personal communication), resulted in the appropriate intensity ratio(100%: 40%)to effect photosynthet the fusion of spherical particles into nanowire structures; similar cooperativity and hence unimodal growth. Consistent with this examples involving spherical particle fusion have also been conclusion, when a 550 20 nm band filter is used with a fluor escent tube to effect the photosynthetic conversion, bimodal growth At first glance, the observed bimodal growth appears to contra- is observed. dict previous results in which unimodal nanoprism growth was This observation of photo-cooperativity provides a way of con- observed when visible light(white) from a conventional fluorescent trolling particle size with light. By supplementing the primary light tube was used as the excitation source. By careful analysis of the source(450-700 nm)with a fixed secondary beam(340 nm,corre- ptical properties of these nanostructures and the effects of photo- sponding to out-of-plane quadrupole plasmon excitation), we can lysis on them, we have identified a type of surface plasmon intentionally effect unimodal growth and generate a solution of ooperativity in the photochemistry of Ag nanoprisms. To demon- nanoprisms of a desired average size. Using this approach, we have trate this cooperative effect on nanoprism growth, we excited a been able to synthesize nanoprisms with in-plane dipole plasmon solution of Ag nanoparticles(4.8+ 1. I nm)at two wavelengths, resonances that track with particle size from 30 to 120 nm by using 550# 20 nm(primary)and 450+ 5nm(secondary)(Isso: 145o= primary excitation wavelengths of 450+ 20 nm, 490 t 20 nm, 2: 1, Fig. 3a). The 450-nm wavelength was selected to excite the 520# 20 nm, 550+ 20 nm, 650 20 nm and 750 20 nm, colloid AAAA Wavelength(nm) Primary beam wavelength (nm A罗 ▲ Figure 3 The unimodal growth of nanoprisms. a, Schematic diagram of dual-beam with a secondary wavelength(340 nm; width, 10 nm). c, The edge lengths as a function of excitation. b, The optical spectra (normalized) for six different-sized nanoprisms(1-6 the primary excitation wavelength. d-f, TEM images of Ag nanoprisms with average edge edge length:38±7mm,50±7mm,62±9m,72±8mm,95±11 nm and lengths o38±7nm(d,72±8 nm (e) and120±14m(. Scale bar applies to 20+ 14 nm) prepared by varying the primary excitation wavelength(central wavelength at 450, 490, 520, respectively, width, 4 NaturEVol42512october2003www.nature.com/nature e 2003 Nature Publishing Groupprisms. Third, detailed time-dependent UV–vis.–NIR measure￾ments show that the onset of the growth of the band at 1,065 nm (assigned to type 2) is significantly delayed in comparison with the growth of the band at 640 nm (assigned to type 1) (Fig. 2a). This indicates that the fusion of nanoprisms occurs only after type 1 nanoprisms have accumulated. Fourth, a small population of dimer 2 and trimer 3 intermediates (Fig. 1d) is observed during the early stages of type 2 particle growth (see Supplementary Information). We also performed electrodynamics calculations on the optical properties of possible intermediate species involved in the fusion growth process. The results show that intermediates 2 and 3 have dipole plasmon excitations close to 600 and 1,065 nm (see Sup￾plementary Information). Type 1 particles and the intermediates 2 and 3 can thus all absorb light at 600 nm, which can lead to the excited state needed for particle fusion to occur. However, the type 2 particles do not show dipole plasmon excitation explaining why they represent the end of the particle growth path. Note in this context that removal of surface ligands has, in the case of CdTe (ref. 27) and PbSe (C. B. Murray, personal communication), resulted in the fusion of spherical particles into nanowire structures; similar examples involving spherical particle fusion have also been reported28. At first glance, the observed bimodal growth appears to contra￾dict previous results in which unimodal nanoprism growth was observed when visible light (white) from a conventional fluorescent tube was used as the excitation source23. By careful analysis of the optical properties of these nanostructures and the effects of photo￾lysis on them, we have identified a type of surface plasmon cooperativity in the photochemistry of Ag nanoprisms. To demon￾strate this cooperative effect on nanoprism growth, we excited a solution of Ag nanoparticles (4.8 ^ 1.1 nm) at two wavelengths, 550 ^ 20 nm (primary) and 450 ^ 5 nm (secondary) (I 550:I 450 ¼ 2:1, Fig. 3a). The 450-nm wavelength was selected to excite the quadrupole plasmon of the type 1 prisms. Double-beam excitation at these wavelengths inhibits the formation of type 2 nanoprisms and results in exclusive formation of the smaller type 1 nanoprisms (72 ^ 8 nm), as evidenced by UV–vis.–NIR spectra and TEM analysis (Fig. 3b, spectrum 4, and Fig. 3e). We further investigated the effect of varying the wavelength of the secondary beam and found that a 550-nm/340-nm coupled beam, in which the 340-nm light coincides with the out-of-plane quadrupole plasmon of the type 1 nanoprisms, also can inhibit the formation of type 2 nanoprisms and result in unimodal growth. However, in the cases of 550-nm/395-nm, 550-nm/610-nm, and 550-nm/650-nm coupled beams, in which the secondary wavelengths fall within the dipole resonances of the Ag nanospheres (395 nm) and type 1 nanoprisms (610 and 650 nm), respectively, bimodal growth is observed (see Supplementary Information). These data strongly indicate that only secondary wavelengths that can excite quadrupole plasmon modes can inhibit bimodal growth. Indeed, it is this photo-cooperativity that leads to the results observed with a fluorescent tube as the excitation source23. The emission spectrum of a fluorescent tube exhibits bands at 546 nm and 440 nm, and has the appropriate intensity ratio (100%:40%) to effect photosynthetic cooperativity and hence unimodal growth. Consistent with this conclusion, when a 550 ^ 20 nm band filter is used with a fluor￾escent tube to effect the photosynthetic conversion, bimodal growth is observed. This observation of photo-cooperativity provides a way of con￾trolling particle size with light. By supplementing the primary light source (450–700 nm) with a fixed secondary beam (340 nm, corre￾sponding to out-of-plane quadrupole plasmon excitation), we can intentionally effect unimodal growth and generate a solution of nanoprisms of a desired average size. Using this approach, we have been able to synthesize nanoprisms with in-plane dipole plasmon resonances that track with particle size from 30 to 120 nm by using primary excitation wavelengths of 450 ^ 20 nm, 490 ^ 20 nm, 520 ^ 20 nm, 550 ^ 20 nm, 650 ^ 20 nm and 750 ^ 20 nm, respectively (Fig. 3b–f). The average edge lengths of the resulting nanoprisms correlate well with the wavelength of the primary Figure 3 The unimodal growth of nanoprisms. a, Schematic diagram of dual-beam excitation. b, The optical spectra (normalized) for six different-sized nanoprisms (1–6 edge length: 38 ^ 7 nm, 50 ^ 7 nm, 62 ^ 9 nm, 72 ^ 8 nm, 95 ^ 11 nm and 120 ^ 14 nm) prepared by varying the primary excitation wavelength (central wavelength at 450, 490, 520, 550, 650 and 750 nm, respectively; width, 40 nm) coupled with a secondary wavelength (340 nm; width, 10 nm). c, The edge lengths as a function of the primary excitation wavelength. d–f, TEM images of Ag nanoprisms with average edge lengths of 38 ^ 7 nm (d), 72 ^ 8 nm (e) and 120 ^ 14 nm (f). Scale bar applies to panels d–f. letters to nature NATURE | VOL 425 | 2 OCTOBER 2003 | www.nature.com/nature © 2003 Nature PublishingGroup 489