Three-orders-of-magnitude variation of carrier lifetimes with crystal phase of gold nanoclusters

See allHide authors and affiliations

Science  19 Apr 2019:
Vol. 364, Issue 6437, pp. 279-282
DOI: 10.1126/science.aaw8007

Atomic packing controls exciton lifetime

Like semiconductors, small metallic clusters can absorb light and create excitons (electron-hole pairs). In ligand-capped gold clusters of 30 to 40 atoms (Au30 to Au40) that adopt the usual face-centered cubic packing, the lifetime of these excitons is ∼100 nanoseconds. Zhou et al. found that atomic packing and molecular orbital overlap can greatly affect carrier lifetimes. Despite having similar bandgaps to those of face-centered cubic clusters, a hexagonal close-packed Au30 cluster had a much shorter lifetime (∼1 nanosecond), and a body-centered cubic Au38 cluster had a lifetime of ∼5 microseconds, which is comparable to bulk silicon.

Science, this issue p. 279


We report a three-orders-of-magnitude variation of carrier lifetimes in exotic crystalline phases of gold nanoclusters (NCs) in addition to the well-known face-centered cubic structure, including hexagonal close-packed (hcp) Au30 and body-centered cubic (bcc) Au38 NCs protected by the same type of capping ligand. The bcc Au38 NC had an exceptionally long carrier lifetime (4.7 microseconds) comparable to that of bulk silicon, whereas the hcp Au30 NC had a very short lifetime (1 nanosecond). Although the presence of ligands may, in general, affect carrier lifetimes, experimental and theoretical results showed that the drastically different recombination lifetimes originate in the different overlaps of wave functions between the tetrahedral Au4 building blocks in the hierarchical structures of these NCs.

Light-harvesting nanomaterials in solar energy utilization (1, 2) convert absorbed light into excitons (electron–hole pairs). Excitons can dissociate productively to form free charge carriers or recombine unproductively, so the relative rates of these processes are important in energy storage and conversion. Correlating the carrier recombination dynamics and the structure of nanomaterials is of great importance. Carrier lifetimes are dependent on the band gap energy (Eg), overlap between the wave functions of the ground state and the excited state, temperature, and other conditions (3, 4). Manipulation of carrier lifetimes can greatly alter the functionalities of nanomaterials for different applications.

Metal nanoclusters (NCs) hold promise in a variety of applications (58) owing to their versatile functionalities that can be tailored by size, structure, and composition (9, 10). Unlike plasmonic gold nanoparticles (Au NPs), ultrasmall Au NCs (<2 nm in diameter) show discrete electronic energy levels and multiple peaks in their ultraviolet-visible (UV-vis) absorption spectra (11). Achieving a fundamental understanding of the optical properties and photophysics of metal NCs (including the electron and phonon dynamics) is of great importance to the exploration of their applications (12, 13). Ultrafast spectroscopy has revealed that Au NCs typically show fluence-independent electron dynamics (14, 15), which is different from the behavior of plasmonic Au NPs (16, 17) and semiconductor quantum dots (18, 19).

With respect to the effect of NC size on photophysics, a general trend is that the larger NCs have shorter carrier lifetimes because of a smaller Eg. Such an energy gap trend was recently reported in Au NCs (20). Although excited-state lifetimes generally follow the Eg law, Au NCs with Eg > 1 eV may show deviations (21). Therefore, the structure of NCs should play an important role in their carrier lifetimes because the quantum confinement of electrons is dictated by the shape of the potential well.

Bulk gold adopts the face-centered cubic (fcc) structure, but Au NCs can adopt many different types of structures (2224), including hexagonal close-packed (hcp) (25) and body-centered cubic (bcc) (26) structures. The different packing of Au atoms gives rise to different electronic structures and UV-vis absorption. The fcc series (Au28, Au36, Au44, and Au52, which are protected by the same thiolate ligand) adopted a layer-by-layer growth pattern and thus showed a uniform evolution in UV-vis absorption (21, 24). Upon photoexcitation in the fcc NCs, electron cooling occurred only in the metal core and there was no core–shell charge transfer (21), unlike similarly sized icosahedral NCs (27). Recently, excited-state electron localization was observed in the linear triicosahedral Au37 NC largely due to its anisotropic shape (28).

These reported examples pertain to the fcc (21) or icosahedral (13, 28) NCs. Here, we report unusual carrier dynamics of Au NCs with hcp and bcc crystalline phases. Specifically, the carrier dynamics of hcp Au30(S-Adm)18 (hereafter Au30, where “S-Adm” stands for 1-adamantanethiolate) and bcc Au38S2(S-Adm)20 (hereafter Au38) NCs exhibit drastic differences compared with the icosahedral Au25 and fcc Au36/Au44/Au52 NCs, although all six NCs possess comparable band gaps (1.3 to 1.77 eV). Surprisingly, hcp Au30 had a substantially shorter lifetime (1 ns) than the typical lifetime (~100 ns) of NCs with similar Eg values, whereas the bcc Au38 had a substantially longer lifetime (4.7 μs). We argue that the ~4700 times difference in lifetime between the hcp Au30 and bcc Au38 originated from the different arrangements of local Au4 motifs within the hcp and bcc cores.

The syntheses of hcp Au30 and bcc Au38 NCs followed our previous methods (25, 26), and their structures are shown in Fig. 1, A and B. The hcp Au30 comprises an Au18 kernel in which four layers of atoms (Au3/Au6/Au6/Au3) are arranged in the a/b/a/b manner (inset of Fig. 1C). The Au18 kernel is protected by six dimeric staples (-S-Au-S-Au-S-). The bcc Au38 comprises an Au30 kernel in which bcc unit cells can be seen clearly (inset of Fig. 1D) and the kernel is protected by four dimeric staples, two sulfidos and eight bridging thiolate (-S-).

Fig. 1 X-ray structures and steady-state UV-vis absorption spectra of Au30(S-Adm)18 and Au38S2(S-Adm)20 NCs.

(A) Core–shell structure of Au30. (B) Core–shell structure of Au38. (C and D) UV-vis absorption spectra of Au30(S-Adm)18 and Au38S2(S-Adm)20. The arrows indicate their lowest-energy absorption bands. The inset in (C) shows an Au18 kernel in four layers of an hcp structure, and the inset in (D) shows an Au30 kernel in a bcc structure with the two bcc unit cells highlighted in red and green. Color labels: yellow, S atoms; gray, C atoms. H atoms are omitted for clarity; all other colors are for Au.

The steady-state absorption spectrum of hcp Au30 shows prominent absorption peaks at 370 and 545 nm, as well as a hump at 480 nm and a shoulder at 680 nm (Fig. 1C), whereas bcc Au38 shows prominent peaks at 640 and 740 nm and a hump at 580 nm (Fig. 1D). By extrapolating absorbance to 0, the energy gaps of hcp Au30 and bcc Au38 were determined to be 1.55 and 1.45 eV, respectively (fig. S1).

To further understand the role of core and surface in the excited states, we compared the transient absorption (TA) spectra pumped at 360 nm and probed between 430 and 810 nm for both NCs (Fig. 2, A and B). By the time (t) of 10 ps after the pump, hot carriers in both samples had cooled down and three ground-state bleaching (GSB) peaks were seen in both cases. In hcp Au30, these peaks were at 480, 545, and 680 nm, together with excited-state absorption (ESA) at 600 nm (Fig. 2A). The 680-nm GSB observed in Au30 corresponds to the shoulder band at 680 nm in the UV-vis absorption spectrum (Fig. 1C). In bcc Au38, GSB peaks were at 580, 640, and 760 nm, which were overlapped with ESA at 500 and 700 nm (Fig. 2B). The TA spectra for both NCs showed somewhat similar profiles, with TA in Au38 redshifted compared with that of Au30. There was no real ESA peak in the TA spectra; all of the ESA peak positions agreed with those minima in the steady-state absorption spectra. The very broad ESA spanned the entire visible region and even into the near-infrared, and this broad ESA overlapped with multiple GSB peaks. Such features have been widely observed in other thiolate-protected Au NCs (21). The broad ESA is helpful in solar cell and photocatalysis applications that require continuous white light excitation so that reexcitation of excited states helps to maintain a longer excited-state lifetime.

Fig. 2

Comparison of spectral features and carrier dynamics of the two NCs. (A and B) TA spectra (black) of (A) hcp Au30 and (B) bcc Au38 NCs at a time delay of 10 ps pumped at 360 nm. Steady-state absorption spectra (gray) are also shown for comparison. (C and D) TA data map with excitation of 360 nm and kinetic traces of the hcp Au30 NC. (E and F) TA data map with excitation of 360 nm and kinetic traces of the bcc Au38 NC. (G) Data map of ns-TA in bcc Au38 NCs between 0.01 and 20 µs with an excitation of 480 nm. (H) Kinetic traces probed at 620 nm and the corresponding fit. ΔA, change in absorbance; mOD, milli–optical density units.

Despite similar TA profiles at t = 10 ps for the two NCs, drastic differences in carrier dynamics were observed. In Au30, after photoexcitation at 360 nm, the broad ESA disappeared within 5 ps (Fig. 2, C and D), which we attribute to hot-carrier relaxation. In the subsequent 2 ns, most of the TA signal disappeared. Global fitting required three decay components to fit the dynamics (1.2 ps, 4 ps, 1 ns, fig. S2A). For excitation at 560 nm, the fast decay component was accelerated to 0.8 ps, but the slow components (3.4 ps and 1.1 ns) remained the same (figs. S2B and S3). The 1-ns lifetime in Au30 is very short considering its Eg of 1.55 eV; for comparison, the Eg of the Au25 NC is 1.3 eV (11) and its excited-state lifetime is ~100 ns, which is a typical value of thiolate-protected Au NCs (29). A recent study also reported a relatively short lifetime (~3 ns) in Au30(SR)18 (where R indicates t-butyl group) NCs with a different structure and ligand (30).

In bcc Au38, the broad ESA for excitation at 360 nm decayed within < 2 ps, giving rise to a net negative GSB at 640 and 760 nm (Fig. 2, E and F). In contrast to that in Au30, the TA signal showed almost no decay between 10 ps and 2 ns (Fig. 2F), which suggests that it has a substantially longer excited-state lifetime than that of Au30. To obtain the complete excited-state lifetime of bcc Au38, a nanosecond TA measurement with excitation at 490 nm was performed (Fig. 2G), which gave a single exponential decay lifetime of 4.7 μs (Fig. 2H). With the input of the 4.7 μs into the femtosecond dynamics, global fitting showed three decay components, 0.6 ps, 4.8 ps, and 4.7 μs (fig. S4A), for fitting the relaxation dynamics. With excitation at 730 nm, the fast-decay component disappeared and only the 5-ps and 4.7-μs components remained (figs. S4B and S5). The ~4700 times difference in excited-state lifetimes of the hcp Au30 and bcc Au38 NCs was unexpected given their similar Eg (~1.5 eV) according to the energy gap law.

The excited-state lifetimes of Au30 and Au38 NCs can be further compared with those of other NCs of similar Eg (Fig. 3, A to D), including the icosahedral Au25(SR)18 and fcc Au52(SR)28, Au44(SR)32, and Au36(SR)24 NCs with Eg between 1.3 and 1.77 eV. The excited-state lifetimes of the latter four NCs were all reported to be ~100 ns and did not follow the energy gap law (21). These differences in carrier lifetimes could arise from the differences in molecular orbital distribution, core–shell interactions, and metal core structures. Surface ligands and the overlap of the sulfur orbitals with the Au electronic structure can affect the carrier dynamics (31), but the bcc Au38 and hcp Au30 have the same type of ligands. Moreover, the metal cores of both NCs were protected by dimeric staples (-S-Au-S-Au-S-), which rules out surface differences in staple type.

Fig. 3 Correlation between structures and excited-state lifetimes of bcc, hcp, and fcc NCs.

(A to C) Tetrahedral Au4 networks in Au30, Au38, and Au36 NCs. (D) Excited-state lifetimes versus Eg of several gold NCs. (E) Excited-state lifetimes versus distance between the Au4 units in the cores of bcc Au38, hcp Au30, and fcc Au36/Au44/Au52 NCs. (F) Frontier orbitals and HOMO-LUMO centroid distances of Au30, Au38, and Au36 from DFT calculations. Color labels: yellow, S atoms; all other colors indicate Au. Carbon tails are omitted for clarity.

A core–shell relaxation model was previously proposed to explain the picosecond relaxation in the Au25 NC (27, 31). In hcp Au30 and bcc Au38, the picosecond decay was always observed independently of the pump wavelength (figs. S2 to S5). The 4- to 5-ps process in both NCs could be explained as core–shell charge transfer or energy relaxation within the metal core. In hcp Au30, the amplitude of the picosecond component was larger than that of the bcc Au38 (figs. S2 and S4), which indicates the stronger picosecond decay in hcp Au30.

For both hcp Au30 and bcc Au38 NCs, careful analyses of Au–Au bond-length distributions revealed that their cores can indeed be viewed as several locally segregated Au4 tetrahedral units (i.e., very short bond lengths within each Au4 versus longer distances between Au4 units, figs. S6 and S7). In the hcp Au30, the Au18 core consisted of six Au4 units (Fig. 3A) assembled by sharing two vertexes of each Au4, so the distance between Au4 units was zero (Fig. 3E). Such a conjugated arrangement of Au4 units leads to a large overlap of the wave functions of Au4 units. However, the bcc Au38 had four Au4 units in the core and the distance between Au4 units was ~2.86 Å (Fig. 3B and fig. S7, II). The nonconjugated Au4 units and the longer distance between them led to much less overlap of the wave functions of Au4 units and slower energy dissipation from the excited state to the ground state.

In the fcc series of NCs (Au36, Au44, and Au52), the distances between neighboring Au4 units was ~3.0 Å (Fig. 3C and fig. S8) and energy dissipation would be slow, but the Au4 units in the fcc series were arranged in a double-helix pattern (24, 32); that is, Au4 units shared one vertex between neighboring units within each chain (Fig. 3C). Theoretical analysis (33) also revealed 1s-like superatomic orbitals of Au4 units: 4-centered-2-electron bonds (hereafter 4c-2e). Such an arrangement should lead to relatively more efficient energy transfer within each helix. Therefore, a moderate carrier lifetime (~100 ns) is observed lying between the very short lifetime of the compact, ring-like, conjugated Au4 superstructure in the hcp Au30 (Fig. 3A) and the very long lifetime of the loose, square-like, nonconjugated Au4 network in the bcc Au38 (Fig. 3B).

This relation between the carrier lifetime and the Au4 network was further corroborated by the analysis of the frontier orbitals (34). As shown in Fig. 3F, the conjugated Au4 network in the hcp Au30 led to an almost zero distance between its highest occupied and lowest unoccupied molecular orbital (HOMO and LUMO, respectively) and the shortest carrier lifetime. By contrast, the nonconjugated Au4 network in the bcc Au38 showed the largest geometric separation of HOMO and LUMO centroids among the three NCs and thus took the longest time for the excited state to relax back to the ground state. Overall, the specific arrangements of Au4 units in the NCs explained the drastically different excited-state lifetimes resulting from the different extents of overlap of 4c-2e bonds. Recent work reported that the Au–Au distance in the metal core could be affected by surface functional group (35, 36). To understand how the functional group affects the core structure, we performed density functional theory (DFT) calculations and optimized the structure of the hcp Au30 NC with R=Adm in comparison with R=CH3. The ligand effect on the Au–Au distances is rather minor in this case (fig. S9). The simulated optical absorption spectra of hcp Au30 and bcc Au38 (see supplementary text and fig. S10) show good agreement with the experiment (fig. S1), further indicating that the optical gaps and oscillator strengths are not key factors in dictating their excited-state lifetimes.

A closer examination of the early part of the TA dynamics also revealed different behavior for hcp Au30 and bcc Au38 (Fig. 4, A to F). Strong oscillation behavior was observed in the hcp Au30 TA decay traces between 510 and 560 nm (Fig. 4, A and D), which originate from coherent phonons. Coherent phonons were previously observed in Au NCs (13, 28, 37) and were caused by ultrafast photoexcitation. The frequency of the phonons in hcp Au30 was determined to be 16.7 cm−1 by fast Fourier transform (fig. S11). The oscillation only persisted for two periods and was totally damped in <4 ps. The fast damping of oscillation suggested that the energy loss in Au30 was very fast, which agreed with its rapid excited-state relaxation. By contrast, no oscillatory feature was observed in the TA time profile of bcc Au38 (Fig. 4, B and E). The 16.7 cm−1 phonon frequency of hcp Au30 was ascribed to the acoustic phonon in the metal core. The giant Au246(SR)80 NC (15) had a similar phonon frequency of ~16.7 cm−1 (Fig. 4, C and F), whereas Au25, which has a size similar to that of hcp Au30, exhibited coherent phonons with much higher frequencies (40 and 80 cm−1) (27). In plasmonic Au NPs, the frequency of the coherent vibration was inversely proportional to the particle diameter (16), but in ultrasmall Au NCs, the above results indicate that the structure rather than the size plays an important role in the phonon frequency.

Fig. 4 Oscillations observed in NCs.

(A to C) TA data map of hcp Au30, bcc Au38 (pumped at 360 nm), and Au246 (pumped at 470 nm) NCs between –1 and 16 ps. (D to F) Kinetic traces probed at selected wavelengths. Strong oscillations were observed in Au30 and Au246 but not in Au38.

We have demonstrated a three-orders-of-magnitude variation of carrier dynamics with crystalline phases of hcp Au30 and bcc Au38 NCs that relates to the distance between the Au4 tetrahedral units and their connection modes. The extraordinarily long lifetime of 4.7 μs in bcc Au38 is comparable to that of bulk silicon and is much longer than that of semiconductor quantum dots, so this NC material may hold promise in boosting the NC solar-cell performance (8, 38). The correlation of the structure and photodynamics of these metal NCs may stimulate their future applications in solar energy conversion, photocatalysis, and other optoelectronic processes.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S11

References (3943)

References and Notes

Acknowledgments: Funding: R.J. acknowledges financial support from the National Science Foundation (DMR-1808675) and the Air Force Office of Scientific Research. D.J. was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division. This work used resources of the Center of Functional Nanomaterials, which is a U.S. DOE Office of Science Facility at Brookhaven National Laboratory under contract no. DR-SC0012704. Author contributions: M.Z. and M.Y.S. performed all TA measurements and M.Z. performed the data analysis. T.H. prepared Au30 and Au38 NCs and Y.C. performed some of the steady-state measurements. G.H. and D.J. performed the DFT calculations and analysis. R.J. designed the study and supervised the project. M.Z. and R.J. wrote the manuscript with contributions from all authors. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available in the manuscript and in the supplementary materials. All data needed to evaluate our conclusions are provided in the manuscript or in the supplementary materials.

Stay Connected to Science

Navigate This Article