Elucidating the Near-Infrared Photoluminescence Mechanism of Homometal and Doped M25(SR)18 Nanoclusters

More than a decade of research on the photoluminescence (PL) of classic Au25(SR)18 and its doped nanoclusters (NCs) still leaves many fundamental questions unanswered due to the complex electron dynamics. Here, we revisit the homogold Au25 (ligands omitted hereafter) and doped NCs, as well as the Ag25 and doped ones, for a comparative study to disentangle the influencing factors and elucidate the PL mechanism. We find that the strong electron–vibration coupling in Au25 leads to weak PL in the near-infrared region (∼1000 nm, quantum yield QY = 1% in solution at room temperature). Heteroatom doping of Au25 with a single Cd or Hg atom strengthens the coupling of the exciton with staple vibrations but reduces the coupling with the core breathing and quadrupolar modes. The QYs of the three MAu24 NCs (M = Hg, Au, and Cd) follow a linear relation with their PL lifetimes, suggesting a mechanism of suppressed nonradiative decay in PL enhancement. In contrast, the weaker electron–vibration coupling in Ag25 leads to higher PL (QY = 3.5%), and single Au atom doping further leads to a 5× enhancement of the radiative rate and a suppression of nonradiative decay rate (i.e., twice the PL lifetime of Ag25) in AuAg24 (hence, QY 35%), but doping more Au atoms results in gold distribution to staple motifs and thus triggering of strong electron–vibration coupling as in the MAu24 NCs, hence, counteracting the radiative enhancement effect and giving rise to only 5% QY for AuxAg25–x (x = 3–10). The obtained insights will provide guidance for the design of metal NCs with high PL for lighting, sensing, and optoelectronic applications.


■ INTRODUCTION
Photoluminescence (PL) is one of the most important properties of atomically precise metal nanoclusters (NCs).−3 However, the PLQY of most metal NCs is quite low (i.e., PLQY < 1%); thus, PL enhancement is of critical importance.−15 Nevertheless, the detailed PL mechanism has not been fully elucidated in many cases.
Among the thiolate (−SR)-protected metal NCs, the icosahedral [M 25 (SR) 18 ] − NCs (M = Au, Ag) often serve as the model systems for investigating the physical and chemical properties of metal NCs because of their easy syntheses and superatomic closed shell electron configuration (1S 2 |1P 6 ). 16A full understanding of the excited-state relaxation in [M 25 (SR)  18 ] − NCs can provide guidance for the fundamental studies of other NCs toward the tailoring of their optical functionality, but this is still hampered by the complex electronic dynamics 16 in such NCs caused by the intricate interactions between the kernel and the surface staple motifs. 4he structure of [Au 25 (SR) 18 ] − NC was solved as early as 2008 with phenylethanethiolate (abbrev.PET) as the protecting ligand. 17,18Single-crystal X-ray diffraction revealed that [Au 25 (PET)  18 ] − (denoted Au 25 hereafter) consists of an icosahedral Au 13 kernel (or core) and six Au 2 (PET) 3 staple motifs on the surface.The study on the structure−PL correlation of Au 25 has been carried out ever since its structure was determined, 7 but the full emission spectrum was not obtained for a long period.With a broadband detector that covers both the visible and near-infrared range, Liu et al. recently determined the full emission profile of Au 25 to be a single broad peak centered at around 1100 nm with a PLQY of 1%. 4 For the silver counterpart (i.e., [Ag 25 (SR) 18 ] − ), its structure was reported in 2015 and also comprises an icosahedral Ag 13 kernel and six Ag 2 (SR) 3 staple motifs, 19 being similar to Au 25 but with slight differences in bond lengths and bond angles.For [Ag 25 (SR) 18 ] − , 2,4-dimethylbenzenethiol (2,4-DMBT) is the most commonly used ligand and hereafter [Ag 25 (2,4-DMBT) 18 ] − is denoted as Ag 25 .The emission peak of Ag 25 is located at 970 nm, and both Au 25 and Ag 25 exhibit a large Stokes shift (∼0.5 eV), indicating structural changes in the excited state. 16Although much effort has been devoted to revealing the PL mechanism of the Au 25 NC, 6,16 a universal picture that can rationalize all of the experimental results is still lacking; similarly, the case of Ag 25 also requires more elucidation.Meanwhile, their low PLQY (<5%) makes the study of PL mechanisms quite challenging, especially in timeresolved PL analyses.
−23 Similarly, in the case of Ag 25 , the structures of Pd-, Pt-, and Au-doped M 25 NCs have also been successfully determined. 19,24,25The PL properties of these M 25 nanoclusters have drawn great attention, but the understanding of the PL mechanism is still limited due to the very complex excited-state dynamics.In addition, because of the limitations of PL instruments (e.g., the cutoff of the near-infrared part of the emission by visible range detectors), the reported emission spectra of many M 25 NCs may not be complete, which results in some deficient conclusions.
Herein, we select two series of M 25 NCs to study the PL mechanism, (i) the Au 25 -based series (including Au 25 , CdAu 24 , and HgAu 24 ) and (ii) the Ag 25 -based series (including Ag 25 , AuAg 24 , and Au x Ag 25−x ).To make a better comparison with the parent [Au 25 (SR) 18 ] − and [Ag 25 (SR) 18 ] − , all of the doped NCs are chosen to have a closed shell of 8-electron configuration; thus, the Pd and Pt doping cases are not included since they typically lead to 6e systems.We performed temperature-dependent steady-state and time-resolved PL measurements to understand how the doping atoms affect both the radiative and nonradiative decays.The monoheteroatom (Cd or Hg) doping into Au 25 is found to significantly affect the electron−acoustic phonon interaction (i.e., the nonradiative decay), while gold atom doping into Ag 25 significantly affects both radiative relaxation and nonradiative decay processes.The obtained insights not only provide some guiding principles for the design of NCs with high photoluminescence but also will promote the research on the electroluminescence (in light-emitting diodes 3 ) and electrochemiluminescence as well as solar concentration applications 2 of the NCs.

■ RESULTS AND DISCUSSIONS
Optical Characterization of MAu 24 (M = Cd, Hg, and Au) NCs.For the Au 25 -based series of NCs, Au 25 (PET) 18  − was synthesized following a previously developed one-pot synthesis method. 26The Cd or Hg monodoped NCs (denoted as CdAu 24 and HgAu 24 ) were synthesized by a reaction between Au 25 (PET) 18  − and metal thiolate complexes (i.e., Cd(PET) 2 , Hg(PET) 2 ). 22,27Thin-layer chromatography (TLC) was employed to separate CdAu 24 and HgAu 24 from the crude products.Crystals of CdAu 24 and HgAu 24 were obtained by diffusion of acetonitrile into toluene solutions of the NCs.All measurements used the crystallized NCs that were redissolved in solvents in order to ensure the highest purity of the NCs.
Based on the reported single-crystal X-ray diffraction (SCXRD) results, the structure of Au 25 comprises a Au 13 icosahedral core and six Au 2 (PET) 3 staple motifs 17 (Figure 1a).Due to the insufficient difference of atomic mass between Hg/Cd and Au atoms, it is not easy to reliably determine the Cd or Hg doping site by SCXRD.−30 However, recent NMR results and theoretical simulations proved that the actual doping site of the Hg or Cd atom is on the Au 12 icosahedral shell, resulting in an Au@MAu 11 kernel (Au@MAu 11 is denoted as MAu 12 below, M = Au, Cd, or Hg, Figure 1a−c). 27,31The difference in the six staple motifs of the three NCs is negligible since all are gold-containing Au 2 (PET) 3 staples.
Here, we use deuterated chloroform (CDCl 3 ) as the solvent for all the room-temperature optical measurements in the nearinfrared (NIR) region.Compared to other solvents, CDCl 3 has weaker NIR absorption (i.e., vibrational overtones), which can alleviate the solvent absorption-induced distortion of the NIR PL spectra.Figure 1d−f (solid lines) shows the roomtemperature optical absorption spectra of MAu 24 in CDCl 3 (M = Au, Cd, or Hg).The characteristic peak of Au 25 at 678 nm is slightly shifted to 657 nm in CdAu 24 and 703 nm in HgAu 24 .The ∼800 nm shoulder peak of Au 25 experienced a blue shift to ∼760 nm in CdAu 24 but vanished in HgAu 24 , suggesting that the single heteroatom affects the energy levels and the electronic transitions.The PL spectra of MAu 24 NCs are measured in deaerated CDCl 3 (with N 2 ) under 500 nm excitation (Figure 1d−f, shaded areas).On a note, the existence of O 2 or higher excitation energy could potentially oxidize or damage the NCs and result in a side-product with emission at around 800 nm.All three NCs exhibit a single emission peak in the range of visible to NIR (detector's range: 500−1700 nm).The emission peak of Au 25 is centered at 1080 nm, whereas CdAu 24 and HgAu 24 are at 1045 nm and 1170 nm, respectively.A comparison of PL spectra in DCM and CDCl 3 is shown in Figure S1, demonstrating a negligible solvent effect.The PL excitation spectra (Figure 1d−f, dashed lines) of all three NCs match well with their UV−vis absorption peaks, suggesting that the emission originates from the first excited state after the hot exciton relaxes.
The photoluminescence quantum yield (PLQY) of Au 25 was previously determined to be 1% via a relative method using the IR-1061 dye as the standard. 4In this work, the PLQY of Au 25 is further confirmed using the PLQY of AuAg 24 (2,4-DMBT) 18 − as the standard.The PLQYs of CdAu 24 and HgAu 24 were obtained by comparing their integrated peak areas with that of Au 25 and were determined to be 1.7 and 0.3%, respectively.The PL lifetimes were measured by a multichannel scaling (MCS) single-photon counting technique.As shown in Figure S2, the decay curves of all three NCs can be well fitted by monoexponential functions.The fitting results suggest that the PL lifetime is 53 ns for HgAu 24 , 155 ns for Au 25 , and 257 ns for CdAu 24 .The measured lifetimes are consistent with the previous results extracted from the transient absorption spectroscopy (TAS) analyses. 32he nature of the PL of Au 25 has been discussed for a long time, but it is still under debate.The existence of a triplet state in Au 25 was earlier proved by singlet oxygen generation experiments conducted by Kawasaki et al. 33 In later work, Agrachev et al. 34 also confirmed that all of the MAu 24 NCs (M = Au, Cd or Hg) can be singlet oxygen sensitizers, and their activity was found to be related to their excited-state lifetimes (determined by TAS). 32The NIR emissions of the three NCs have the same lifetimes as the TAS results, 32 suggesting that the transient absorption and PL decay involve the same excited state.Since the longest component in the TAS comes from the triplet state, the ∼1100 nm emission should arise from the same triplet state.Therefore, we can ascribe the emission peak at ∼1100 nm to the phosphorescence from the MAu 12 kernel.The lifetime of MAu 24 (i.e., ranging from 53 to 257 ns) is much shorter than the typical lifetime of phosphorescence (i.e., microseconds), but similar results have been observed in metal−organic complexes. 35In TAS analyses, a picosecondscale lifetime (2−5 ps) was observed in all three MAu 24 NCs, which is similar to the picosecond lifetime in the chargeneutral Au 25 . 32,36Previously, this picosecond lifetime was assigned to the transition between the excited state and the surface trap state.Now, we believe it is better to be assigned as the lifetime of the S 1 state.
Interestingly, we found that the PLQYs of the three NCs are proportional to their PL lifetime (Figure 1g).Theoretically, the PLQY of phosphorescence (Φ p ) is defined as

Journal of the American Chemical Society
where Φ isc is the quantum yield of intersystem crossing (ISC), k r is the radiative rate, and k nr is the nonradiative rate.In principle, the PLQY can be enhanced by accelerating the radiative decay (i.e., large k r ) and/or suppressing the nonradiative decay (small k nr ).To reveal the mechanism, eq 1 is rewritten as where τ av is the measured PL lifetime.To simplify the calculation, we assume that Φ isc is close to unity, which is justifiable. 13Then, the linear relationship between the PLQY and the lifetime of the three NCs indicates that the radiative rates (k r ) of the three NCs are approximately the same.Following this logic, a linear fitting is given in Figure 1g and the common k r of the three NCs is determined to be (7 ± 0.1) × 10 4 s −1 .This radiative rate is close to the rate constant of phosphorescence in Au 13 NCs, 13,14 which further supports the phosphorescence nature in the MAu 24 NCs.Given the above results, we can conclude that the radiative rates of these three NCs are essentially the same; thus, the doping atom only exerts its effect through the nonradiative channel, which accordingly affects the observed PLQY.
Another interesting relationship is that the nonradiative rates of the three NCs are found to follow the energy gap law (Figure 1h).The energy gap law is widely applied to understand the multiphonon-assisted emission mechanism. 37he formula of the simplest version of the energy gap law 38 is given in eq 3 where ln(k nr ) is the natural logarithm of the nonradiative rate constant, γ is a molecular parameter, ℏω M is the highest energy of phonon that assists the nonradiative decay, and ΔE is the energy gap.Of note, here we take the PL peak position as the energy gap.Considering the similarity in the geometrical structure of these three NCs as well as their electronic structure, the molecular parameter γ can be treated as the same for the three NCs.Then, it means that the ℏω M is roughly consistent in all three NCs.The main phonon that affects the electronic transition in Au 25 NCs in the solution phase was determined as the vibration from the Au 2 (SR) 3 staple motifs. 39ince the motifs of these three NCs are the same, it is reasonable to observe such a linear relationship.Temperature-Dependent PL Analysis of MAu 24 (M = Cd, Hg, and Au) NCs.Since the nonradiative relaxation through vibrations of staple motifs is strong for all three NCs and may hinder the understanding of doping effects on the core vibration, we carried out temperature-dependent PL measurements from room temperature down to 20 K (liquid helium as the cryogen).For such measurements, MAu 24 NCs were embedded in a polystyrene film matrix.Our previous work demonstrated that the vibrations from staple motifs could be suppressed in thin films, which allowed us to obtain insights into the core vibrations pertaining to the MAu 12 icosahedron. 4,39In cryogenic experiments, the chamber of the cryostat was filled with helium gas during the measurements to avoid interference from oxygen.
At room temperature, as shown in Figure S3a−c, the emission profiles of all three NCs in polystyrene thin films are consistent with their solution counterparts but with higher intensity due to the suppression of the motif vibration.Specifically, the PLQY of Au 25 increases to 1.8%, whereas CdAu 24 and HgAu 24 increase to 7 and 2.2%, respectively (Figure S3a−c); generally, the enhancement is a few times greater.The lifetimes of all three NCs become prolonged when embedded in polystyrene thin films (Figure S3d−f).
The temperature-dependent PL spectra of the three NCs are given in Figure 2a−c, where the black traces and white insets represent the temperature-dependent peak positions.The emission peaks of all three samples become narrower and show a gradual blue shift as the temperature decreases.A monotonic increase of PL intensity (Figure S4a−c) is observed for all three MAu 24 NCs with temperature decreasing from room temperature to 20 K. Nevertheless, even with the suppression of vibrations, the PLQY for the three samples is still below 30% at 20 K (Figure S4a−c).The low-temperature PLE spectra (Figure S4d−f) are consistent with the previously reported low-temperature absorption spectra, 32 suggesting that the emission peak is only related to the first excited state (S 1 ) at different temperatures.The temperature-dependent lifetimes (Table S1) show a monotonic increase with decreasing temperature.Moreover, the lifetimes for the three NCs at 20 K all reach the microsecond scale, which further supports the phosphorescence nature of the observed NIR emission.
As shown in Figure 2d, all three NCs show higher radiative rates in the thin film (k r ranging from 9 × 10 4 to 2 × 10 5 at r.t.) than in the solution (k r = 7 × 10 4 s −1 at r.t.).The enhancement of the radiative rate in the solid state may originate from aggregation-induced slight overlapping of wavefunctions in the solid state, which may boost the transition from the T 1 to the S 0 state.Interestingly, the radiative rate of CdAu 24 and HgAu 24 in thin films are close at all temperatures (Figure 2d), and both also surpass that of Au 25 .Such a difference may originate from the different charge states of MAu 24 NCs, among which CdAu 24 and HgAu 24 are charge neutral, while Au 25 possesses a negative charge, which may result in electrostatic repulsion between NCs in the solid state, hence, less aggregationinduced enhancement in the case of Au 25 .Meanwhile, a decrease in radiative rate with temperature decrease is observed for all three NCs, suggesting activation energy is present for radiative recombination. 40Overall, the nonradiative rates (Figure 2e, i.e., of order 10 6 ) are found to be roughly an order of magnitude larger than the radiative rates.The k nr trend is HgAu 24 > Au 25 > CdAu 24 when the temperature is above 80 K, but the k nr of HgAu 24 becomes smaller than that of Au 25 when the temperature is lower than 80 K.The monotonic decrease of k nr as the temperature decreases is because of the suppression of the phonon population at low temperatures.The rapid nonradiative recombination (i.e., of order 10 6 ) is the main reason for the low PLQY of these three NCs.
To understand the origin of rapid nonradiative rates in three gold NCs, we carried out an analysis of temperature-dependent emission broadening.−43 The linewidth (Γ) is represented by the full width at half maximum (FWHM).For most semiconductor materials, the temperature-dependent PL broadening follows a weak electron−phonon coupling (e−ph) and it can be described by eq 4 below. 43,44In this weak regime, the product of the phonon field fluctuation amplitude (Δ) and fluctuation correlation time (τ c ) is smaller than ℏ, which typically results in a Lorentzian line shape at low temperatures. 45,46 In the equation, Γ 0 is the temperature-independent intrinsic linewidth of materials, γ ac and γ LO are the coupling coefficients of an electron with acoustic phonon (ac) and longitudinal optical (LO) phonon, respectively, and ) that describes the temperature-dependent population of phonon modes.The last term in eq 4 accounts for the contribution from ionized impurities.The temperature-dependent general trends of the four contributors are plotted in Figure 2f.
In the case of gold NCs, their formulas are precisely defined, and the samples are molecularly pure; thus, the contribution from impurities is irrelevant and can be ignored.It is worth noting that N LO (T) can be regarded as ∝T when the temperature is high (i.e., T > 100 K, Taylor expansion), suggesting that the sum over the remaining three contributors should result in a linear increase of FWHM with increasing T (when T > 100 K).The FWHM of the three NCs at different temperatures are extracted and plotted in Figure 2g−i.Interestingly, the temperature-dependent trends of all three NCs in the high-temperature region do not follow the expected linear relationship, suggesting the common weak e−ph coupling approximation (Figure 2g−i, blue dashed lines) is not applicable here.
In contrast to the linear relationship (∝T) for the weak coupling, Toyozawa theoretically demonstrated that the temperature-dependent evolution of the excitonic linewidth would be proportional to T under strong electron−phonon coupling, in which the linewidth broadening can be described as eq 5 45−47 The extracted parameters by fitting with the strong e−ph coupling model are listed in Table 1.−50 The average energy E ac of acoustic phonon modes for the three NCs are quite close (Table 1), suggesting that the doped heteroatoms have a trivial contribution to the periodical expansions/contractions of the MAu 12 kernel.Nevertheless, the electron−acoustic-phonon coupling strength S ac in Au 25 is found to be significantly stronger than the Cd-and Hg-doped MAu 24 .It means that heteroatom doping could weaken the exciton-vibration coupling within the core.Yu et al. theoretically simulated the Raman spectra of Cd-and Hg-doped MAu 24 NCs and found that the intensity of the low-frequency vibration (i.e., < 100 cm −1 ) is very low and much weaker than Pd-or Pt-doped MAu 24 NCs. 50Meanwhile, Krishnadas et al. experimentally measured the Raman spectra of Au 25 and found strong signals in the low-frequency region (<100 cm −1 ). 51Considering these two facts, it is reasonable for us to observe less coupling after Cd and Hg doping.The breathing mode is known as the main acoustic vibration in Au 25 and it is highly symmetric. 4,36,48The doping of Cd or Hg atom into Au 25 breaks the symmetry of the core due to the asymmetric distribution of electron density. 22,50Therefore, such a dissymmetry diminishes the scalar product of the breathing mode atoms' displacement and excitation profile integrated over the spherical volume, 48 which weakens the coupling between the exciton and acoustic phonon after Cd or Hg doping.
As shown in Table 1, the average energy of optical phonon E op in Au 25 and HgAu 24 are determined to be ∼40 meV (∼300 cm −1 ), which is consistent with previous cryogenic absorption measurements. 4The coupling strength of the optical phonon S op is high for all three NCs (Table 1), demonstrating that the coupling of the exciton with the optical phonon modes is strong.Due to the high energy of the optical phonon, their population is very small at low temperatures (i.e., <100 K); thus, the acoustic phonon becomes the main contributor to nonradiative relaxation.Since CdAu 24 and HgAu 24 have weaker coupling to an acoustic phonon, these two NCs should have smaller nonradiative rates (compared to Au 25 ) when the temperature is low, which is consistent with the results in Figure 2e.In contrast, at room temperature, the doping of the heteroatom suppresses the electron−acoustic-phonon coupling, but the scattering of electrons by optical phonons is extremely strong for all three NCs, which inevitably results in low PLQY.
Combining the discussions above, we summarize the mechanism in Scheme 1.Briefly, the ∼1100 nm NIR emission from the MAu 24 NCs is ascribed to phosphorescence based on the following facts: (1) all three MAu 24 NCs can serve as singlet oxygen sensitizers, suggesting the existence of tripletstate populations; (2) the time-resolved PL measurements showed a single lifetime for their emission; and (3) the lifetimes determined from time-resolved PL measurements are consistent with the longest lifetime component in previous transient absorption results.Additionally, earlier TAS experiments demonstrated the presence of a picosecond-scale lifetime at 730 nm excitation, 32 suggesting that this lifetime originates from the lowest excited state.Therefore, we assign the picosecond lifetime to the decay of the S 1 state (i.e., intersystem crossing and vibrational relaxation).Moreover, an unusually strong electron−phonon interaction is found in all three NCs.Both acoustic phonons and optical phonons contribute to the electron−phonon scattering in Au 25 as indicated by large S ac and S op values, while the acoustic phonon contribution is suppressed in CdAu .One may question why the strong electron−phonon interaction is not able to suppress the phosphorescence completely; this is because the nonradiative relaxation of the triplet state also needs to go through a slow spin-flipping step, which slows down the entire nonradiative process.Overall, the monoatom doping could significantly affect the electron−acoustic phonon coupling but will not promote the PLQY a lot because the nonradiative relaxation by optical phonons (i.e., motif vibrations) is extremely strong in MAu 24 NCs.Insight from the Silver Series of Ag 25 , AuAg 24 , and Au x Ag 25−x (x = 3−10) NCs.To further understand the metaldoping effects, we investigated a second system related to Au 25 and doped NCs, that is, Ag 25 and doped NCs, collectively as Au x Ag 25−x (2,4-DMBT) 18  − NCs (with the number of gold atoms x = 0, 1, and 3−10).Previous SCXRD results showed that all these silver-based NCs possess an M 13 kernel and six M 2 (SR) 3 staple motifs, being the same as the structure of Au 25 . 19,24,52Therefore, these silver-based NCs can also be   19 The doping position of the gold atom in AuAg 24 is determined to be at the icosahedral center, forming a AuAg 12 kernel. 24In Au x Ag 25−x , one gold atom occupies the icosahedral center and the rest of the gold atoms substitute silver atoms in the staple motifs.Therefore, Ag 25 and AuAg 24 have the same staple motifs but different kernels, while AuAg 24 and Au x Ag 25−x have the same kernel but different staple motifs.Of note, silver NCs are much less stable than gold NCs; thus, we carried out the measurements immediately after we synthesized the samples.
The UV−vis absorption spectra of Ag 25 , AuAg 24 , and Au x Ag 25−x NCs in CDCl 3 (Figure 3d−f, solid lines) are in good agreement with previous reports. 19,24,53For comparison with the MAu 24 series, the steady-state PL spectra (Figure 3d  excitation.The emission peak of Ag 25 is found to be at 1000 nm, and it blueshifts to 900 nm for AuAg 24 and slightly redshifts to 1050 nm for Au x Ag 25−x .As shown in Figure S6, the peak positions for the silver series in CDCl 3 are slightly redshifted compared to their counterparts in DCM.The dashed lines in Figure 3d−f represent the PL excitation spectra of the silver series, and they are consistent with the corresponding UV−vis absorption spectra, suggesting that all the emission is from the first excited state.The absolute PLQY of AuAg 24 in CDCl 3 was determined to be 35% using an integrating sphere (Figure S7), and the PLQYs of Ag 25 and Au x Ag 25−x were determined to be 3.5 and 5% by the relative method, respectively.Time-resolved PL measurements (Figure S8) showed that the decay curves of the silver series can be fitted by monoexponential functions.The PL lifetime of Ag 25 is determined to be 1100 ns, while the lifetime of AuAg 24 is prolonged to ∼2000 ns.Surprisingly, Au x Ag 25−x has a higher PLQY than Ag 25 , though the lifetime of Au x Ag 25−x is shorter (450 ns), indicating that the PL enhancing mechanism is not due to the suppression of nonradiative decay (which would exhibit a prolonged lifetime), but due to the accelerated radiative decay (exhibiting a shorter lifetime or a larger rate).The measured lifetimes are consistent with previous timecorrelated single-photon counting (TCSPC) measurements and TAS results. 24,53,54The long lifetime suggests the existence of phosphorescence in Ag 25 and AuAg 24 NCs, which was previously proved by triplet−triplet annihilation upconversion PL experiments by Niihori et al., 54 and the origin of phosphorescence was ascribed to the staple's triplet state, 54 but they provided no solid evidence to support that the emission was from the motif states, nor the charge transfer from the core triplet state to the motif triplet state.Previous density functional theory (DFT) calculations proved that the HOMO and LUMO of Ag 25 were mainly located at the core Ag atoms. 55Metal−ligand charge transfer typically requires The FWHM of the steady-state PL spectra as a function of temperature for (g) Ag 25 , (h) AuAg 24 , and (i) Au x Ag 25−x .The blue dashed line in panel (i) is the fitting result to eq 4, and the red line is the fitting result to eq 5.
higher excitation energy and results in a discrepancy between UV−vis absorption spectra and PL excitation spectra. 6,56onsidering the perfect match of the PL excitation spectra with the UV−vis absorption spectra (Figure 3d−f), we believe that the phosphorescence of the silver series of NCs is from the core state, just like the MAu 24 series, that is, both pertain to the first excited state.
A plot of PLQY versus the average lifetime for the silver series and Au 25 is given in Figure 3g.Unlike the gold series, the plot of the silver series does not follow any linear relationship, indicating that the doping of the gold atom(s) into the Ag 25 template not only affects the nonradiative process but also the radiative rate.Here, since the long lifetime is the main component in the decay curve of the silver series, we could still use to estimate the radiative rate k r and nonradiative rate k nr .As shown in Figure 3h, similar to the gold series, the silver series also shows a good agreement with the energy gap law, suggesting that the maximum energy of the phonon that assists the nonradiative relaxation is similar among the three silver-based NCs.
Temperature-Dependent PL Analysis of Ag 25 , AuAg 24 , and Au x Ag 25−x (x = 3−10) NCs.To further understand the PL mechanism and vibrations of the doped silver NCs, we embedded them into a polystyrene matrix and carried out the temperature-dependent PL measurements from room temperature to 20 K.As shown in Figure S9a−c, the PLQY of all three silver-based NCs is increased in polystyrene films.The emission profiles of Ag 25 and AuAg 24 in thin films showed a slight red shift (∼10 nm) compared to their solution counterparts, while the emission peak of Au x Ag 25−x shows an 80 nm redshift when embedded in a polystyrene matrix.It suggests that the main contributor (i.e., different numbers of doped gold atoms) to the emission profile of Au x Ag 25−x may vary between the solution and thin film.Nevertheless, it is hard to determine the exact species because our synthesis is not able to separate the mixture by the x number.The PL lifetimes of the silver series are given in Figure S9d−f, where Ag 25 possesses a longer lifetime in the thin film, while AuAg 24 and Au x Ag 25−x both have shorter lifetimes in thin films, indicating that both radiative and nonradiative rates are changed.
The temperature-dependent PL spectra of the three silver NCs are given in Figure 4a−c, where the black traces and white insets represent the temperature-dependent peak positions.Interestingly, unlike the monotonic trend in the gold series, the temperature-dependent peak position of AuAg 24 and Ag 25 show an unusual zig-zag trend as the temperature decreases.Initially, the emission peak undergoes a blue shift, which is followed by a red shift and, eventually, another blue shift.The case of Au x Ag 25−x exhibits a monotonic blue shift of the peak position as the temperature decreases; nevertheless, the temperature-dependent peak position presents abnormal flatness between 200 K and room temperature (Figure 4c inset).The plots of temperaturedependent PL intensity for the silver series are given in Figures 4d and S10.The integrated PL intensities of Ag 25 and AuAg 24 show a similar three-stage evolution as their temperaturedependent peak positions.In stages I and III (Figure 4d), the PL intensity increases as the temperature decreases, but in stage II (Figure 4d), the PL intensity decreases as the temperature decreases.Meanwhile, a monotonic increase of the integrated PL intensity with decreasing temperature is observed for Au x Ag 25−x .
Here, we ascribe the abnormal temperature-dependent trends of Ag 25 and AuAg 24 to the coexistence of thermally activated delayed fluorescence (TADF) with phosphorescence.This can be rationalized as follows.Taking Ag 25 as an example, in stage I, both TADF and phosphorescence exist and present a blue shift as the temperature decreases, so the observed peak showed an overall blue shift.The suppression of the nonradiative rate with the temperature decrease results in a net increase of PL intensity in this stage.Then, in stage II, the lower temperature suppresses the reverse intersystem crossing and the TADF starts to vanish, but phosphorescence is still increasing, so the overall peak position showed a red shift, and the integrated PL intensity decreases in this region.In stage III, the TADF is almost completely suppressed, and phosphorescence becomes the major contributor to the emission peak.Therefore, a monotonic blue shift and also a monotonic increase of PL intensity are observed.Compared to Ag 25 , AuAg 24 shows a less significant decrease of PL intensity in stage II, suggesting a less contribution from TADF in AuAg 24 .The temperature-dependent excitation spectra (Figure S11) of Ag 25 and AuAg 24 are consistent with the previously reported temperature-dependent absorption spectra, 57 suggesting that both TADF and phosphorescence are from the first excited states (S 1 and T 1 ).The case of Au x Ag 25−x is much more complicated since it has more than one emitter due to the mixed x values.The short lifetime of Au x Ag 25−x suggests that its electron dynamic is closer to the MAu 24 series.Since we do not observe a clear stage II in Au x Ag 25−x , the TADF component in Au x Ag 25−x should be negligible, and phosphorescence is the main contributor, from which we can infer that a full occupation of silver in the six M 2 (SR) 3 staple motifs is crucial to achieve TADF in the M 25 NCs.Au x Ag 25−x has some gold atoms in the staples, hence, the disappearance of TADF.
The temperature-dependent lifetimes of the silver series are listed in Table S2.Of note, due to the low sensitivity of NIR detectors and the long lifetime of Ag 25 (>25 μs) at low temperatures, we can only obtain the lifetime of Ag 25 when the temperature is higher than 80 K.The lifetimes of all three NCs become much prolonged and reach the microsecond scale at low temperatures, suggesting that the emission is from the triplet state.The temperature-dependent radiative rate (Figure 4e) and nonradiative rate (Figure 4f) of the silver series are calculated using the temperature-dependent quantum yields and lifetimes.As shown in Figure 4e, the radiative rates of AuAg 24 and Au x Ag 25−x are close at room temperature, and both are 5 times larger than Ag 25 , suggesting that the central gold atom significantly improves the radiative decay.The radiative rate of Ag 25 is suppressed to one-fifth at 80 K, which further proves the existence of TADF in Ag 25 . 58Meanwhile, the suppression of radiative rates for AuAg 24 and Au x Ag 25−x is much less significant, which is also consistent with our above discussion of the smaller contribution of TADF to these two NCs.The nonradiative rate of Au x Ag 25−x is found to be an order of magnitude stronger than the other two silver-based NCs, suggesting that the Au 2 (SR) 3 staple motif is the main origin of nonradiative relaxation and that partial substitution by Ag 2 (SR) 3 is not able to alleviate the fast nonradiative relaxation.As the temperature decreases, the nonradiative rate of the silver-based NCs is suppressed significantly.The nonradiative rates of Ag 25 and AuAg 24 are very close at all temperatures, indicating that the central gold doping has a trivial effect on the vibration.

Journal of the American Chemical Society
To better understand the vibrational properties of the three silver NCs, we extracted the temperature-dependent linewidths of the silver series from their corresponding emission spectra, and the plots are given in Figure 4g−i.Due to the large component of TADF at room temperature for Ag 25 , its temperature-dependent linewidth (Figure 4g) also presents a three-stage evolution that shares similar transition temperatures as the temperature-dependent peak position and intensity.Interestingly, when the temperature is above 160 K (mainly TADF), we observe a linear relationship between the temperature and PL linewidth, suggesting a weak electron− phonon interaction in Ag 25 .Meanwhile, a similar linear relationship is observed when the temperature is lower than 80 K (mainly phosphorescence), suggesting that the acoustic phonon is the main contributor to the PL broadening at low temperatures. 43However, the overall zig-zag trend of Ag 25 prohibits us from doing further quantitative analyses.Fortunately, in the case of AuAg 24 , less TADF and higher transition temperature lead to less interference in the analysis of temperature-dependent PL linewidths.As shown in Figure 4h, the temperature-dependent linewidth of AuAg 24 can be well fitted by the weak e−ph coupling model given in eq 4 (the last term is neglected since no impurity exists), and Table 2 gives the extracted parameters.The energy of the optical phonon is determined to be 28 meV, which matches well with the vibration energy of Ag 2 (SR) 3 staple motifs determined by Raman spectroscopy. 59In Figure 4i, the temperature-dependent linewidth of Au x Ag 25−x is similar to the MAu 24 series, which shows a nonlinear relationship at high temperatures.However, the multiple emitters in Au x Ag 25−x due to different x make it not possible to fit the temperature-dependent linewidth by the strong coupling model.The clear nonlinear relationship in the high-temperature region is sufficient to prove the strong electron−phonon interaction in Au x Ag 25−x ; 45 hence, its less PLQY, albeit the radiative rate is equally accelerated by central gold doping as in AuAg 24 , that is, the strong e−ph coupling from the gold-containing staples significantly counteracts the role of k r enhancement and thus reduces the PLQY of Au x Ag 25−x as in the gold series.
By combining the discussions above, the emission mechanism of Ag 25 and AuAg 24 is summarized in Scheme 2a, whereas the case of Au x Ag 25−x is closer to Scheme 1 because it exhibits predominant phosphorescence.Since the excitation spectra of the silver series match well with their corresponding UV−vis absorption spectra, Kasha's rule is valid.The different electron dynamics between Au x Ag 25−x and Ag 25 /AuAg 24 indicate that a full occupation of M 2 (SR) 3 staple motifs by Ag atoms is critical for observing TADF.Of note, although the percentage of TADF contributing to the emission intensity of Ag 25 is greater than that in AuAg 24 , it does not necessarily prove that the reverse intersystem crossing is more efficient in Ag 25 because the overall PLQY of AuAg 24 is 10 times higher than Ag 25 .The reverse intersystem crossing should also exist in Au x Ag 25−x or even the gold series, but the strong e−ph interaction in the latter series leads to an efficient nonradiative relaxation that completely suppresses the radiative recombination.In other words, the change from strong e−ph coupling in the gold series to weak e−ph coupling in the silver series (except Au x Ag 25−x ) should be the real reason that TADF is observed in Ag 25 and AuAg 24 .
The doping effects of gold atoms in the Ag 25 template are summarized in Scheme 2b.When gold atoms replace silver atoms in the staple motifs, it introduces strong electron− phonon coupling and results in fast nonradiative relaxation.The correlation between Au 2 (SR) 3 and strong e−ph interactions is consistent with the afore-discussed results of the MAu 24 series.The strong coupling may originate from the heavy mass and rich orbitals of gold atoms.Interestingly, the central doping by the gold atom significantly increases the radiative decay in AuAg 24 and Au x Ag 25−x .Such a phenomenon cannot be explained by previous DFT simulations because no triplet state was taken into consideration.Therefore, further theoretical studies are needed in order to better understand the excited states of the silver series.In the case of AuAg 24 , the six Ag 2 (SR) 3 staple motifs help the decoupling between the electron and phonons (i.e., suppressing the nonradiative decay), and the central gold atom boosts the radiative decay.These two factors make AuAg 24 one of the most luminescent nanoclusters (QY = 35% at r.t.).Finally, we compile the PL data of the gold series in Table 3 and the silver series in Table 4 for ease of comparison.

■ CONCLUSIONS
In summary, we have systematically studied the PL properties for a series of superatomic closed shell (1S 2 |1P 6 electron configuration) M 25 NCs.The ∼1100 nm NIR emission of the three Au 25 -based NCs is determined as phosphorescence.Heteroatom doping by the Cd or Hg atom is found to mainly affect the nonradiative process (i.e., suppressing the electron− acoustic phonon interaction) rather than affecting the radiative relaxation.An unusually strong electron−phonon interaction is found to exist in all three Au 25 -based NCs, and the optical phonon from the Au 2 (SR) 3 vibration is determined as the main contributor.In the second system (the three Ag 25 -based NCs), the electron−phonon interaction is found to be weakened from heavy gold doped Au x Ag 25−x to the original Ag 25 NC, which further supports that the Au 2 (SR) 3 staple motif is the main source for strong electron−phonon interactions.Due to the much weaker electron−phonon interaction in Ag 25 and AuAg 24 , the carrier lifetime is much prolonged, and we observe the existence of TADF in the emission profile of these two NCs.Interestingly, the center doping of gold atoms into Ag 25 significantly improves the radiative recombination.The efficient radiative relaxation from the central gold atom and the suppression of nonradiative relaxation by six Ag 2 (SR) 3 staple motifs make AuAg 24 one of the most luminescent NCs.These insights can be extended to other icosahedral NCs, such as Au 38 and bi-icosahedral Au 25 , and will facilitate the development of applications of luminescent metal nanoclusters.

Figure 1 .
Figure 1.(a) Atomic structure of Au 25 (carbon tails and Au(core)−Au(shell) bonds are omitted for clearity).(b) Structure of the CdAu 12 core.(c) Structure of the HgAu 12 core.The structures of the NCs were redrawn from the CIF. 27−30 Absorption (solid line) and PL (shadowed area) spectra of (d) Au 25 , (e) CdAu 24 , and (f) HgAu 24 in deaerated CDCl 3 (with N 2 ).Dashed lines represent the excitation spectra of PL at (d) 1085 nm, (e) 1045 nm, and (f) 1170 nm.(For PL measurements: excitation at 500 nm with 0.2 OD, slit width 8 nm, and emission slit 8 nm.) (g) Plot of PLQY versus average lifetime for the three gold-based NCs.The red line is a linear fit to Φ = τ av k r Φ isc (assume Φ isc = 1).(h) Plot of ln(k nr ) versus the photoluminescence gap for the three NCs.The red line is a linear fit to the energy gap law.

Figure 2 .
Figure 2. Temperature-dependent PL spectra of (a) Au 25 , (b) CdAu 24 , and (c) HgAu 24 with excitation at 500 nm.(d) Temperature-dependent radiative rates of Au 25 , CdAu 24 , and HgAu 24 .(e) Temperature-dependent nonradiative rates of Au 25 , CdAu 24 , and HgAu 24 .(f) Functional form of the temperature dependence of the contributions to the PL linewidth.The FWHM of the steady-state PL spectra as a function of temperature for (g) Au 25 , (h) CdAu 24 , and (i) HgAu 24 .The blue dashed lines fit the results to eq 4, and the red lines fit the results to eq 5 (see main text).
Journal of the American Chemical Societywhere Γ 0 is the temperature-independent intrinsic linewidth, S ac and S op are the coupling strengths for acoustic phonons and optical phonons, respectively, and E ac and E op are the average energy of acoustic phonons and optical phonons.The red lines in Figure2g−i show the fitting results based on eq 5. Apparently, the fitting by the strong coupling model is much better than the weak coupling model.Therefore, strong electron−phonon interactions are involved in the MAu 24 NCs.
24 and HgAu 24 due to asymmetric doping.The strong electron−phonon interactions result in fast nonradiative relaxation, hence, low PLQY in the three NCs.The fast nonradiative relaxation is also the reason for the complete suppression of fluorescence and the unusually short lifetime (50−257 ns) of phosphorescence at room temperature, as dictated by the relation = +

Figure 3 .
Figure 3. (a) Atomic structure of Ag 25 (carbon tails and Au(core)−Au(shell) bonds are omitted for clarity).(b) Structure of AuAg 24 .(c) Structure of Au x Ag 25−x .The structures are redrawn from the previous data. 19,24,53Absorption (solid line) and PL (shadowed area) spectra of (d) Ag 25 , (e) AuAg 24 , and (f) Au x Ag 25−x in deaerated CDCl 3 (with N 2 ).Dashed lines represent the excitation spectra of PL at (d) 1010 nm, (e) 910 nm, and (f) 1050 nm.For PL measurements: excitation at 500 nm with 0.2 OD, a slit width of 8 nm, and an emission slit of 8 nm.(g) Plot of PLQY versus average lifetimes for the four NCs.(h) Plot of ln(k nr ) versus the photoluminescence gap for the three silver-based NCs.The red line is a linear fit to the energy gap law.

Figure 4 .
Figure 4. Temperature-dependent PL spectra of (a) Au x Ag 25−x (b) AuAg 24 , and (c) Ag 25 with excitation at 500 nm.(d) Temperature-dependent radiative rate constants of Au x Ag 25−x , AuAg 24 , and Ag 25 .(d) Normalized integrated PL intensity of Ag 25 at different temperatures.(e) Temperaturedependent radiative rate of Ag 25 , AuAg 24 , and Au x Ag 25−x .(f) Temperature-dependent nonradiative rate constants of Ag 25 , AuAg 24 , and Au x Ag 25−x .The FWHM of the steady-state PL spectra as a function of temperature for (g) Ag 25 , (h) AuAg 24 , and (i) Au x Ag 25−x .The blue dashed line in panel (i) is the fitting result to eq 4, and the red line is the fitting result to eq 5.

Scheme 2 .
Scheme 2. (a) Emission Mechanism of Ag 25 and AuAu 24 .(b) Summary of Doping Effects in the Ag 25 -Based Series of NCs

Table 1 .
Extracted Parameters for the Three MAu 24 NCs (Fitted by eq 5)

Table 2 .
Extracted Parameters for AuAg 24 (Fitted by eq 4 without the Last Term)

Table 3 .
PL Parameters of the Three Au-Based NCs at Different Temperatures

Table 4 .
PL Parameters of the Three Ag-Based NCs at Different Temperatures