Electron localization in rod-shaped triicosahedral gold nanocluster

Significance

Understanding the carrier dynamics in ultrasmall (<2-nm) gold nanoclusters is fundamentally important for their applications in solar energy storage and conversion. This work tackles how the electrons of gold nanoclusters flow after photoexcitation. The electron localization/delocalization is commonly observed in polymers and other molecular aggregates, but little is known about this phenomenon in gold nanoclusters. Here, we observed ∼100-ps excited-state electron localization in atomically precise rod-shaped Au₃₇ nanoclusters made up of three icosahedra. The activation energy was further obtained by temperature-dependent experiments. The excitation localization observed in rod-shaped clusters of clusters will advance their applications in nonlinear optics and energy harvesting.

Abstract

Atomically precise gold nanocluster based on linear assembly of repeating icosahedrons (clusters of clusters) is a unique type of linear nanostructure, which exhibits strong near-infrared absorption as their free electrons are confined in a one-dimensional quantum box. Little is known about the carrier dynamics in these nanoclusters, which limit their energy-related applications. Here, we reported the observation of exciton localization in triicosahedral Au₃₇ nanoclusters (0.5 nm in diameter and 1.6 nm in length) by measuring femtosecond and nanosecond carrier dynamics. Upon photoexcitation to S₁ electronic state, electrons in Au₃₇ undergo ∼100-ps localization from the two vertexes of three icosahedrons to one vertex, forming a long-lived S₁* state. Such phenomenon is not observed in Au₂₅ (dimer) and Au₁₃ (monomer) consisting of two and one icosahedrons, respectively. We have further observed temperature dependence on the localization process, which proves it is thermally driven. Two excited-state vibration modes with frequencies of 20 and 70 cm⁻¹ observed in the kinetic traces are assigned to the axial and radial breathing modes, respectively. The electron localization is ascribed to the structural distortion of Au₃₇ in the excited state induced by the strong coherent vibrations. The observed electron localization phenomenon provides unique physical insight into one-dimensional gold nanoclusters and other nanostructures, which will advance their applications in solar-energy storage and conversion.

Excitons in nanomaterials are highly dependent on size, shape, and atomic structure (1–5). By designing different types of nanomaterials, it is possible to control their carrier dynamics and facilitate their applications. Steady-state and transient spectroscopy on nanomaterials reveals the details of their excited states, where excitation experiences redistribution. Ligand-protected, atomically precise metal nanoclusters have emerged as a model system for investigating the correlation between structure and electronic properties of nanostructures (6, 7). Among the nanoclusters with known structures, a unique type is comprised of small building blocks, that is, the “clusters of clusters” (8), such as the rod-shaped Au₂₅ nanocluster made up of two Au₁₃ units via vertex sharing (9–11), as well as the Au₂₈(SR)₂₀, Au₃₆(SR)₂₄, Au₄₄(SR)₂₈, and Au₅₂(SR)₃₂ nanoclusters built up by double-helical assembly of Au₄ tetrahedral units (12). Some of these gold nanoclusters retain the electronic properties of the building blocks while new electronic transition emerges due to “conjugation” of Au₁₃ units (10, 13–15). Excitation localization and delocalization is widely observed in light-harvesting molecular systems (16, 17), conjugated polymers (18, 19), and other molecular aggregates (20–22). However, little is known about the localization/delocalization effect in metal nanoclusters. Understanding the electron localization/delocalization will advance the application of metal nanoclusters in solar-energy storage and conversion.

The optical properties of semiconductor quantum nanorods have been extensively investigated due to their applications in solar-energy conversion and H₂ generation (4, 23–25). Unlike zero-dimensional quantum dots where excitons are confined in all three dimensions (26), excitons in one-dimensional nanorods are only confined in the radial direction, whereas in the axial direction electrons are relatively free (27). These nanorods therefore possess the properties of both quantum dots and bulk materials. Moreover, exciton localization was observed by ultrafast spectroscopy in the nanorods due to the protrusion morphology of the nanorod (28, 29). Similarly, in ultrasmall metal nanoclusters (<2 nm), shape and size also largely influence the electronic structure and electron dynamics (30, 31). It is thus of great interest to investigate the potential of electron localization in rod-shaped gold nanoclusters.

The icosahedron motif is widely observed in gold nanoclusters, such as monoicosahedral [Au₁₃(dppe)₅Cl₂]³⁺ [dppe: 1,2-bis(diphenylphosphino)ethane)] (Au₁₃, monomer) (32) and rod-shaped [Au₂₅(PPh₃)₁₀(SC₂H₄Ph)₅Cl₂]²⁺ nanocluster (Au₂₅, dimer) (9) consisting of two Au₁₃ icosahedra sharing one common vertex—which can be viewed as a dimer of Au₁₃ units. Recently, the structure and optical properties of a rod-shaped [Au₃₇(PPh₃)₁₀(SC₂H₄Ph)₁₀X₂]⁺ (X = Cl/Br) nanocluster (15) (Au₃₇, trimer), which consists of three Au₁₃ units, have been reported. The Au₁₃, Au₂₅, and Au₃₇ nanoclusters possess 8, 16, and 24 valence electrons, respectively, all satisfying Mingo’s electron-counting rule (33). Moreover, all these nanoclusters composed of icosahedral units contain no surface staple motifs as in Au_n(SR)_m nanoclusters (34), which leads to simplified electronic structures compared with Au_n(SR)_m nanoclusters. Theoretical simulations predicted that a linear long chain of Au₁₃ icosahedra fused together by vertex sharing can be either semiconducting or metallic by tuning the charge state, whereas the long chain of Au₁₃ icosahedra via face sharing is always metallic (35). Understanding the interaction between the individual building blocks is thus beneficial to future development of larger cluster-assembled materials and their applications (36–38). On the other hand, electron and phonon dynamics provides important information of how electrons flow in and out of the nanostructures, which has greatly improved the understanding of energy flow in gold nanoclusters (31, 39–43). It is therefore of paramount importance to investigate the ultrafast electron dynamics of these clusters of clusters and compare them with the individual Au₁₃ unit.

Here, we report the ultrafast electron dynamics of Au₁₃ (monomer) and Au₃₇ (trimer) by femtosecond-transient absorption (TA) spectroscopy. Unlike simple molecular-like electron dynamics observed in Au₁₃ and rod-shaped Au₂₅, an additional electron localization (∼100-ps) process was observed in the triicosahedral Au₃₇, which is a unique observation in metal nanoclusters. Temperature-dependent analysis further confirmed that the observed excitation localization process is thermally driven and two activation energy values (E_a) were obtained. Furthermore, two phonon frequencies (20 and 70 cm⁻¹) were observed from the TA kinetic traces, which can be assigned to the axial and radial breathing modes, respectively. These results provide fundamental insight into energy flow in the gold nanoclusters and will stimulate future work on other types of nanostructures composed of repeating building blocks.

Results and Discussion

Optical Absorption Features.

Gold nanoclusters made up of linear combinations of Au₁₃ building blocks (Au₁₃, Au₂₅, and Au₃₇) exhibit an interesting trend in the optical spectra (9, 15, 32). In the short-wavelength range, the absorption spectra of the Au₂₅ and Au₃₇ resemble that of the Au₁₃ building block (Fig. 1), whereas in the long-wavelength range, the absorption peaks are red-shifted due to the interaction between Au₁₃ building blocks. The absorption spectrum of the rod-shaped [Au₂₄(PPh₃)₁₀(SC₂H₄Ph)₅X₂]⁺ (X = Cl/Br) (Au₂₄ for short) (44) nanoclusters was also shown for comparison. Compared with Au₂₅, Au₂₄ has a similar crystal structure but lacks the absorption around 670 nm due to the absence of the gold atom on the vertex position. Theoretical calculations revealed that the absorption peak around 670 nm in Au₂₅ and 800 nm in Au₃₇ originate from the dimeric interactions between Au₁₃ units, whereas the additional peak around 1,230 nm in Au₃₇ originates from trimeric interaction (10). The absorption peaks at shorter wavelengths originate from the individual Au₁₃ icosahedrons. The highest occupied molecular orbital (HOMO)–lowest unoccupied molecular orbital (LUMO) gap was determined by extrapolating optical absorption to zero, that is, 1.96 eV for Au₁₃, 1.73 eV for Au₂₅, and 0.83 eV for Au₃₇. The general trend—that is, the longer the aspect ratio, the more red-shifted the longer wavelength peak—resembles that of plasmonic gold nanorods (45), but the nature of the observed trend in Au₁₃, Au₂₅, and Au₃₇ is due to quantum confinement in a one-dimensional box. Based on the previous theoretical calculation results, the distributions of LUMO of Au₂₅ and Au₃₇ are localized on the one and two vertexes of Au₁₃ icosahedron (10), respectively (Fig. 1).

Fig. 1.

Steady-state UV-Vis-NIR absorption spectra and structures of the metal cores in Au₁₃ (monomer), Au₂₅ (dimer), Au₂₄ (hollow rod), and Au₃₇ (trimer) nanoclusters (surface-capping ligands are omitted for clarity). The red color indicates the distribution of 6s electrons of the LUMO orbital based on theoretical calculations (9).

Electron Dynamics of Au₁₃ Monomer.

Before looking into the interactions between Au₁₃ building blocks in Au₃₇, we first probed the photophysics of the monomer, [Au₁₃(dppe)₅Cl₂]³⁺ nanocluster (Au₁₃ for short). Femtosecond TA with excitation of 360 nm (3.4 eV) and 560 nm [2.2 eV; HOMO–LUMO transition (46)] were carried out and compared (Fig. 2). Upon excitation at 360 nm, one can observe prominent excited-state absorption (ESA) centered at 630 nm and ground-state bleaching (GSB) at 500 nm (Fig. 2A). In about 1 ps, the 630-nm ESA decays to give rise to an additional ESA centered at 450 nm. In the case of excitation at 560 nm, the ESA around 450 nm can be directly excited and the 630-nm ESA is absent (Fig. 2B). The ESA at 630 nm can thus be ascribed to S_n←S₂, whereas ESA at 450 nm should be assigned to S_n←S₁. The kinetic traces of femtosecond TA with excitation at 360 nm exhibit an ultrafast decay plus a nondecay plateau, whereas the TA decay traces with excitation at 560 nm lack the ultrafast component (Fig. 2 C and D). Nanosecond TA and time-correlated single-photon counting (TCSPC) measurements suggest that the long-lived relaxation has only one decaying component (SI Appendix, Fig. S1). Global analysis on TA excited at 360 nm gives rise to two decay components (0.4 ps and 1.72 μs), which can be assigned to LUMO+n-to-LUMO internal conversion and relaxation back to the ground state, respectively (SI Appendix, Fig. S2). The monoicosahedron Au₁₃ exhibits similar carrier dynamics to that of the dimer Au₂₅, which also exhibits two decay components (0.8 ps and 2.3 μs), as previously reported (31). The relatively simple relaxation pathway in both the monomer and dimer suggests that there is no excited-state interaction between the icosahedron units in these two structures (Au₁₃ and Au₂₅).

Fig. 2.

Electron dynamics of Au₁₃ nanoclusters. (A and B) Femtosecond-TA spectra at all time delays of Au₁₃ with excitation of 360 and 560 nm, scattering around 560 nm because the pump laser was cut off. (C and D) Selected decay traces and corresponding fit of Au₁₃ with excitation at 360 and 560 nm. The slow decay of 1.72 μs was fixed based on nanosecond-TA measurement in SI Appendix.

Electron Dynamics of Au₃₇ Trimer.

Femtosecond TA spectroscopy was then performed on the Au₃₇ nanocluster (trimer). Excitation at 1,200 nm (1.03 eV) was chosen because it lies close to the band gap of Au₃₇ (0.83 eV), which excluded the excess excited-state energy. Upon photoexcitation, ESAs at 515, 760, and 830 nm, and GSB at 480 nm were observed in the TA spectra (Fig. 3 A and B). Between 1 and 500 ps, the ESA around 830 nm decays prominently to zero, and notably the ESA around 760 nm experiences a significant growth at the same time. Population evolution based on a two-state model (solid line, Fig. 3C) closely resembles the femtosecond-TA kinetic traces (dots) at selected wavelengths, which indicates that the excited state follows a sequential A→B relaxation model. Nanosecond-TA experiments revealed that the lifetime of Au₃₇ is 28 ns (SI Appendix, Fig. S3). Global analysis can thus extract two species-associated spectra (SAS) with time constants of 115 ps and 28 ns. As the 1,200-nm excitation (1.03 eV) is only slightly above the band gap of Au₃₇, it can directly pump the electrons to the lowest excited state (S₁) so that the last long-lived state (defined as S₁*) lies between S₁ and ground state. The 115-ps component, which was not observed in the rod-shaped Au₂₅ (dimer) and Au₁₃ (monomer), can be tentatively ascribed to the coupling between three Au₁₃ units in the Au₃₇ nanocluster. Compared with the rod-shaped Au₂₅ nanocluster (2.3 μs) (31), the last long-lived state has a nearly 100 times shorter lifetime in Au₃₇ (28 ns). After repeating the experiment by bubbling argon into the solution and dissolving samples in solvents of different polarities, respectively, we further confirmed that the last long-lived state is neither a triplet state nor a charge transfer state (SI Appendix, Fig. S3).

Fig. 3.

TA spectra of Au₃₇ nanoclusters probed in the visible region. (A) Femtosecond-TA data of Au₃₇ with excitation of 1,200 nm. (B) TA spectra as a function of time delay between 1 and 300 ps. (C) Population evolution (solid line) based on a sequential model versus kinetic traces (dots) where decay (518 nm) and growth (710 nm) dominates, respectively. (D) Species-associated spectra (SAS) obtained from global fitting on the TA data. The second lifetime was fixed to 28 ns based on the nanosecond measurement.

The ∼100-ps process in Au₃₇ suggests the existence of additional electronic state between S₁ and S₀ states, which was not observed in both spherical Au₁₃ (Fig. 2) and rod-shaped Au₂₅ nanoclusters. By comparing the TA spectra between rod-shaped Au₂₅ (dimer) and Au₃₇ (trimer), it is found that the TA spectra of dimer and trimer at 1 ns have similar spectral features (Fig. 4); the ESA around 610 nm and GSB around 677 nm in the dimer were red-shifted to 760 and 800 nm in the trimer, respectively. Previous DFT calculations (9) indicated that the LUMO of the dimer is localized on the shared vertex of two icosahedra, whereas the LUMO of the trimer is distributed over the two vertexes of three icosahedra. By directly exciting the trimer to S₁ state, the additional S₁* state (similar to the S₁ state of dimer) is formed, indicating that the final excited states of dimer and trimer have similar electron distributions, that is, the electrons are finally localized at one vertex in the trimer. Therefore, the ∼100-ps process observed in Au₃₇ can be assigned to the excitation localization, in contrast with the simple relaxation pathways in the monomer and dimer cases. This assignment is reasonable considering that the electrons are less confined in the axial direction with increasing aspect ratio from Au₁₃ to Au₃₇.

Fig. 4.

Comparison between TA spectra of rod-shaped Au₂₅ and Au₃₇ nanoclusters. (A) Transient spectra at 1 ns of Au₂₅ rod. (B) Transient spectra at 1 ns of Au₃₇ rod. (Inset) The distribution of electrons in Au₂₅ and Au₃₇ rod based on the electron localization model.

To further probe the origin of the ∼100-ps process in Au₃₇, excitation wavelength-dependent TA experiments on Au₃₇ were performed. Carrier dynamics of Au₃₇ with different excitation energies were compared. Fig. 5 A–D shows the TA data after excitation with pump pulse at 380 nm (3.2 eV), 490 nm (2.5 eV), 820 nm (1.51 eV), and 1,200 nm (1.03 eV), respectively. With excitation energy higher than 1,200 nm (1.03 eV), one can observe additional ESA signals at 520 and 680 nm in the initial few picoseconds (Fig. 5 A–C). The negative signal (deep blue color around 800 nm) with excitation at 360 and 490 nm should be ascribed to the stimulated emission from higher excited states. To look into the ∼100-ps process, we monitor the kinetic traces around 820 and 760 nm (Fig. 5 E and F) as a function of excitation wavelength. As shown in Fig. 5E, the decays of the kinetic traces probed at 820 nm (ESA) are independent of the excitation energy; all of the traces decay to zero within 600 ps. The ESA around 820 nm with 1,200-nm excitation grows immediately within the instrument response (∼100 fs), whereas at higher excitation energies the 820-nm ESA experiences slow growth in the first 5 ps. It suggests that the 820-nm ESA originates from S₁ to higher excited states. In the kinetics of ESA around 760 nm (Fig. 5F), one can observe prominent excitation-dependent growth: kinetics with near-band gap excitation exhibit single-exponential rise (115 ps), whereas at higher excitation energies, additional fast growth can be observed. Excited decaying time constants based on target global fitting are listed in Table 1 (for details of target global fitting, see SI Appendix, Figs. S4–S6).

Fig. 5.

Excitation-dependent electron dynamics. Femtosecond-TA spectra data between 0.1 ps to 3 ns of Au₃₇ with excitation of (A) 380 nm, (B) 490 nm, (C) 820 nm, and (D) 1,200 nm. (E and F) Kinetic traces probed at 820 and 760 nm at all excitation wavelengths and the corresponding fit.

Table 1.

Time constants obtained from target global fitting of TA spectra at different excitation energies

	Excitation energy
Time constant	3.2 eV	2.5 eV	1.51 eV	1.03 eV
τ1	1.9 ps	1.5 ps	3 ps	115 ps
τ2	99 ps	100 ps	108 ps	28 ns
τ3	28 ns	28 ns	28 ns	—

The TA spectra probed in visible region are complicated due to strong overlap between ESA and GSB. To further resolve the excited-state species, UV pump/near-infrared (NIR) probe TA was carried out to monitor the net bleaching dynamics around 1,200 nm. After excitation at 490 nm, strong and net GSB centered at 1,230 nm can be observed (Fig. 6A), which corresponds to the absorption peak due to the trimeric interaction in Au₃₇. During the first few picoseconds, the GSB grows slowly and broad ESA at other wavelength decays at the same time (Fig. 6 B and C). After global analysis, three decaying components (1.5 ps, 105 ps, >1 ns) are required for best fitting, and the time constants match well with those of TA probed at visible region. In Fig. 6D, one can observe that strong ESA spanning between 1,100 and 1,600 nm dominates the first decay-associated spectra (DAS) component, which explains the initial growth in the GSB around 1,230 nm. The 105-ps component, which has contribution all over the detection range, can be well explained by electron localization.

Fig. 6.

TA spectra of Au₃₇ nanoclusters probed in the NIR region. (A) TA data probed at NIR region with excitation of 490 nm. (B) Transient spectra as a function of time delay in the first 2 ps. (C) Kinetic traces and the corresponding fit at selected wavelengths. (D) Decay-associated spectra (DAS) obtained from global analysis on the TA data.

Based on the excitation-dependent ultrafast TA experiments, the energy diagram of Au₃₇ can be constructed (Fig. 7). As discussed above, the 820-nm ESA arises from S₁ state, whereas the 760-nm ESA originates from S₁* state between S₁ and S₀. With 1,200-nm excitation, Au₃₇ is directly excited to the S₁ state, and relaxation from S₁ to S₁* occurs in 115 ps. With higher excitation energies, the relaxations from S_n to S₁ and from S_n to S₁* occurs in a similar timescale, which leads to the fast growth in ESAs around both 820 and 760 nm. Subsequently, relaxation occurs from S₁ to S₁* at a similar time constant to that at near-band gap excitation (Table 1). As proposed above, the relaxation from S₁ to S₁* can be ascribed to an electron localization process. The relatively slow excited-state process (∼100 ps) suggests that there is a significant energy barrier between the initial and final electronic states. It is therefore of great interest to probe the activation energy of this process by carrying out temperature-dependent TA measurements.

Fig. 7.

Energy levels (Left), relaxation pathway (Middle), and diagram of electron localization (Right) of Au₃₇ nanocluster.

Temperature-Dependent Electron Dynamics of Au₃₇.

As shown in Fig. 8 A and B, TA spectra of Au₃₇ film [clusters in poly(methyl methacrylate) (PMMA)] at 5 ps and 1 ns measured at different temperatures are compared. At room temperature, the transient spectra exhibit the same features to those measured in solution. With temperature decreasing from 330 K down to 77 K, three features were observed: (i) both ESA and GSB peaks became sharper and additional peaks emerged, (ii) both the maxima of ESA and GSB evolve to higher energies, and (iii) the intensities of both ESA and GSB increased significantly. The steady-state absorption spectra of Au₂₅(SR)₁₈, Au₃₈(SR)₂₄, and rod-shaped Au₂₅ were reported to exhibit similar behaviors as the temperature decreased (47, 48), which was explained by electron–phonon (e–p) coupling and lattice expansion interactions. Here, the blue shift and enhancement of ESA at low temperatures suggest that the ESA is also strongly affected by the e–p coupling effect. To look into the thermal effect on the electron dynamics, we compared the normalized kinetic traces measured at 290 and 77 K and the corresponding fits in Fig. 8 C and D. As the positions of ESAs around 760 and 820 nm shifted to 740 and 780 nm when the temperature decreased from 290 to 77 K, the maximum positions of ESAs were compared. It is found that the electron dynamics measured at 77 K is significantly slowed down compared with that measured at 290 K, which suggests that the electron localization slows down as the temperature drops. The temperature dependence of the electron localization indicates that it is a thermally driven process.

Fig. 8.

Temperature-dependent electron dynamics of Au₃₇. (A and B) Temperature-dependent TA of Au₃₇ film at time delay of 5 ps and 1 ns with 100-fs, 1,200-nm (1.03-eV) excitation. (C and D) Normalized kinetic traces at selected wavelengths measured at 290 and 77 K. (E) Normalized kinetics of concentration of the first component as a function of temperature obtained from global fitting. (F) lnk versus 1/T.

To resolve the activation energy of this process, we performed the global fitting on the data of a series of temperatures and plotted the concentration dynamics of the first decaying component (Fig. 8E and SI Appendix, Fig. S7). According to the Arrhenius relationship:

$$k=A\exp \left(-\frac{E_{\text{a}}}{k_{\text{B}}T}\right),$$ [1]

$$\ln k=\ln A-\frac{E_{\text{a}}}{k_{\text{B}}T},$$ [2]

where k is the reaction rate, A is the constant, E_a is the activation energy, k_B is the Boltzmann constant, and T is the absolute temperature, we plot lnk as a function of 1/T (Fig. 8F) to extract the activation energy (E_a). It is interesting to see that the linear fitting gives rise to two slopes (−464 and −41), which indicates two activation energies (E_a), that is, 3.85 and 0.34 kJ⋅mol⁻¹. The emergence of two activation energies suggests that there might be two deactivation pathways, the origin of which has not been fully resolved yet. TA anisotropy measurements (SI Appendix, Figs. S8 and S9) were further performed to help identify the origin of the 115-ps process. Meanwhile, it is found that the rod-shaped Au₂₄ nanocluster, which lacks the central gold atom, does not show e-localization process after photoexcitation (SI Appendix, Fig. S10).

Coherent Phonons of Au₃₇.

Besides the electron dynamics, phonon dynamics can also reflect the excited-state energy deactivation. Upon photoexcitation, the distinct different equilibrium between the configurations of ground state and excited state will give rise to coherent vibration of the nanoclusters along the potential energy surface (49). As long as the laser pulse is shorter than the vibration period, the coherent motion will be reflected as temporal oscillations in the early time of TA kinetic traces, which has been observed in a series of atomically precise gold nanoclusters (30, 31, 50). A closer look at TA spectra of Au₃₇ pumped at 1.03 eV in the first 10 ps gives prominent oscillations at almost all wavelengths (SI Appendix, Fig. S11A). When probed around 750 and 820 nm, a 20 cm⁻¹ (0.6-THz) oscillation can be observed, whereas at shorter probe wavelength, we observed a second mode with frequency of 70 cm⁻¹ (2.1 THz). The relatively low vibration frequencies suggest that both modes arise from acoustic vibrations. The change in the phase in oscillations of different wavelengths (Fig. 9A) suggests that the vibration arise from the excited state rather than the ground state. We use the following equation to fit the oscillation superimposed on the kinetic decay:

$$\mathrm{\Delta }A=A{e}^{-t^2/2{\sigma }^2\ast }\left[B{e}^{-t/{\tau }_1}+C{e}^{-t/{\tau }_2}\text{cos}\left(\omega t+\phi \right)\right],$$ [3]

where σ is the SD of Gaussian impulsive function, τ₁ is the relaxation time (fixed to 100 ps), τ₂ is the damping time of the oscillation, ω is the frequency, and φ is the phase.

Fig. 9.

Coherent phonons of Au₃₇. (A) Selected kinetic traces which show coherent oscillations of 70 cm⁻¹. (B) Selected kinetic traces that show coherent oscillations of 20 cm⁻¹. (Insets) Schematic diagram of vibration modes.

Compared with the rod Au₂₅ dimer, which shows oscillations at 26 cm⁻¹ (31), Au₃₇ has a similar frequency mode (20 cm⁻¹) plus an additional higher frequency mode (70 cm⁻¹). Considering the anisotropic effect in their atomic structures, the low-frequency vibration in both Au₂₅ (dimer) and Au₃₇ (trimer) should be ascribed to the axial vibration mode (Fig. 9A, Inset). On the other hand, the higher frequency mode (70 cm⁻¹) observed in Au₃₇ is quite similar to that (80 cm⁻¹) observed in spherical Au₂₅(SR)₁₈ nanocluster (30). Therefore, the 70 cm⁻¹ mode in Au₃₇ should be assigned to the radial breathing mode (Fig. 9B, Inset). Moreover, at 77 K, one can observe enhanced oscillation as well as longer dephasing time compared with that at room temperature (SI Appendix, Fig. S11B). After comparing the data of solution and film at different temperatures, we found that the longer lifetime of oscillations occurs in film at both 290 and 77 K. Therefore, it is the surrounding matrix (PMMA) rather than low temperature that gives rise to the slower damping in the rod-shaped Au₃₇ nanocluster. It is worth noting that one of the two activation energies (0.34 kJ⋅mol⁻¹, equal to 28 cm⁻¹), which we obtained from low-temperature TA, is very close to the energy of the axial vibration mode (20 cm⁻¹). Thus, it is very likely that the axial coherent vibration (20 cm⁻¹) gives rise to the structural distortion and thus the e-localization process.

Conclusion

In summary, we have observed the excited-state exciton localization in the Au₃₇ nanocluster that is made up of three Au₁₃ building blocks. In contrast with the monomer (Au₁₃) and dimer (Au₂₅), the trimer (Au₃₇) exhibits strong coupling between Au₁₃ units after photoexcitation. The temperature-dependent dynamics suggests that the electron localization process is thermally driven. Two vibration modes are observed in the kinetic traces, and they are assigned to the radial and axial vibrations, respectively. The electron localization in rod-shaped Au₃₇ is attributed to the structural distortion induced by the strong excited-state coherent vibrations. Gold clusters with longer linear chains of Au₁₃ units are expected to exhibit similar coupling between units. Our work reports the observation of electron localization as well as coupling between building blocks in gold nanoclusters. These results are of fundamental importance for understanding the electronic properties of not only gold nanoclusters but also other one-dimensional nanostructures composed of repeating building blocks. The obtained insight will help to advance the application of metal nanoclusters in solar-energy conversion.

Experimental Procedures

Sample Preparation.

The syntheses of [Au₁₃(dppe)₅Cl₂]³⁺ [dppe: 1,2-bis(diphenylphosphino) ethane)], [Au₂₄(PPh₃)₁₀(SC₂H₄Ph)₅X₂]⁺ (X = Cl/Br), [Au₂₅(PPh₃)₁₀(SC₂H₄Ph)₅Cl₂]²⁺, and [Au₃₇(PPh₃)₁₀(SC₂H₄Ph)₁₀X₂]⁺ (X = Cl/Br) follow the protocols reported previously (15, 32, 44, 51).

Steady-State Absorption and Fluorescence Measurements.

The UV-Vis-NIR absorption spectra were measured on a Shimadzu UV-3600 plus spectrometer. The steady-state fluorescence was measured on a HORIBA Jobin Yvon SPEX Fluorolog-3 spectrometer. Fluorescence lifetimes were measured with a TCSPC technique; a pulsed LED source (376 nm, 1.1 ns) was used as the excitation source.

Femtosecond and Nanosecond TA.

Femtosecond-TA spectroscopy were carried out using a commercial Ti:sapphire laser system (SpectraPhysics; 800 nm, 100 fs, 3.5 mJ, 1 kHz). Pump pulse was generated using a commercial optical parametric amplifier (LightConversion). A small portion of the laser fundamental was focused into a sapphire plate to produce supercontinuum in the visible region, which overlapped in time and space with the pump. The diameter of the pump beam was 0.75 mm, and the pump power was adjusted to 0.5 mW using neutral density filter. Multiwavelength transient spectra were recorded using dual spectrometers (signal and reference) equipped with array detectors whose data rates exceed the repetition rate of the laser (1 kHz). Solutions of clusters in 1-mm path length cuvettes were excited by the tunable output of the OPA (pump). Nanosecond-TA measurements were conducted using the same ultrafast pump pulses along with an electronically delayed supercontinuum light source with a subnanosecond pulse duration (EOS; Ultrafast Systems). The polarization between pump and probe pulse was set to magic angle (54.7°) for all experiments. The Au₁₃ and Au₃₇ clusters were dissolved in methanol and dichloromethane, respectively, for time-resolved experiments in solution. For matrix measurements, Au₃₇ was dissolved in dichloromethane along with commercial PMMA, followed by drop-casting of the solution onto a sapphire window and then solvent evaporation. We carefully checked the UV-Vis-NIR absorption spectra of Au₃₇ before and after each femtosecond experiment, and the spectra always remained the same (SI Appendix, Fig. S12).

Data Analysis.

Chirp correction was made before the global analysis. Before global fitting, singular value decomposition was performed to evaluate the number of time constants needed. Subsequently, graphical interface Glotaran and TIMP (52) were applied to analyze the population dynamics based on the method reviewed by van Stokkum et al. (53).