The effect of deuterium on the photophysical properties of DNA-stabilized silver nanoclusters

We investigated the effect of using D2O versus H2O as solvent on the spectroscopic properties of two NIR emissive DNA-stabilized silver nanoclusters (DNA–AgNCs). The two DNA–AgNCs were chosen because they emit in the same energy range as the third overtone of the O–H stretch. Opposite effects on the ns-lived decay were observed for the two DNA–AgNCs. Surprisingly, for one DNA–AgNC, D2O shortened the ns decay time and enhanced the amount of µs-lived emission. We hypothesize that the observed effects originate from the differences in the hydrogen bonding strength and vibrational frequencies in the two diverse solvents. For the other DNA–AgNC, D2O lengthened the ns decay time and made the fluorescence quantum yield approach unity at 5 °C.


Introduction
DNA-stabilized silver nanoclusters (DNA-AgNCs) were rst introduced by Petty et al. 1 and consist of a limited number of silver atoms and cations embedded in one or more DNA strands. A comprehensive introduction to the structure/ property relationship of this class of emitters can be found in a review by Gonzàlez-Rosell et al. 2 Recent ndings show that the interplay between the DNA host sequence and the emissive properties of the stabilized silver nanoclusters is intricate and complex. For example, replacing a single guanine with an inosine resulted in a longer uorescence decay time and a quantum yield (Q) increase from 0.25 to 0.63. 3,4 This is remarkable since the only difference between these two nucleobases is a single amino group, which seems to control the amount of non-radiative decay. 4 Inspired by these results, we decided to evaluate the effect of exchanging H 2 O with D 2 O on the photophysical properties of DNA-AgNCs. It is well-known that substituting H 2 O with D 2 O lengthens the excited state decay time and increases the luminescence quantum yield of both ns-lived uorescence from organic uorophores, 5,6 as well as msand ms-lived emission from trivalent lanthanide ions. 7,8 The main reason for this is the lower vibrational frequency of the O-D stretch with respect to the O-H stretch, resulting in less solventmediated non-radiative decay. 5,9 Maillard et al. showed that H 2 O and alcohols can act as weak quenchers and proposed a model of resonant energy transfer from the electronic excited state to vibrational overtones of the O-H bond. 5 Previous studies on a green emissive DNA-AgNC reported no signicant effect of exchanging H 2 O with D 2 O on the ns-lived excited state decay. 3,10 To assess the effect of D 2 O as a solvent, we chose two well-characterized DNA-AgNCs that emit close to 750 nm, which is where the absorption bands of the third overtone of the antisymmetric, symmetric O-H stretch and combination bands thereof are located. 5 Both DNA-AgNCs are stabilized by multiple DNA decamers. One DNA-AgNC contains 16 Ag atoms embedded in two 5 0 -CACCTAGCGA-3 0 strands (further dened as DNA-Ag 16 NC), 11,12 and its structure can be found in the PDB database (6JR4). The structure of the second DNA-AgNC has not been determined yet, but the hydrodynamic volume 13 suggests it is likely wrapped in two or three 5 0 -CCCGGAGAAG-3 0 strands (further referred to as DNA721-AgNC). At 25 C in a 10 mM ammonium acetate (NH 4 OAc) aqueous solution, DNA-Ag 16 NC has a moderate Q of 0.26 and a decay time that is very temperature-dependent. 12 On the other hand, DNA721-AgNC is characterized by a high Q of 0.73 and a decay time that is largely independent of temperature. 13 Surprisingly, and against all expectations, the ns decay time of DNA-Ag 16 NC was found to be shorter and Q lower upon using D 2 O as solvent versus H 2 O. Additionally, D 2 O enhanced the red-shied ms-lived emission, which was negligible in H 2 O. Dual emissive DNA-AgNCs, featuring both ns-and ms-lived emission, have been only recently reported in literature, 14,15 and it was surprising to discover ms-lived emission for this wellcharacterized DNA-Ag 16 NC upon addition of D 2 O. For DNA721-AgNC, a more expected behavior was observed using D 2 O as a solvent: the ns-lived decay lengthened and Q increased, reaching unity at 5 C. To the best of our knowledge this is the highest reported Q value for a NIR-emitting DNA-AgNC.

DNA-Ag 16 NC
Details on the synthesis, HPLC purication and the collected fraction of the DNA-Ag 16 NC sample can be found in the ESI and Fig. S1. † Fig. 1A shows the normalized absorption and emission spectra of DNA-Ag 16 NCs in the DD condition (synthesized and measured in a 10 mM NH 4 OAc D 2 O solution) and HH condition (synthesized and measured in a 10 mM NH 4 OAc H 2 O solution) at room temperature. Note that we did not use deuterated ammonium acetate (ND 4 OAc) because the concentration of H 2 O present as impurity in D 2 O (0.1% of 55 M D 2 O) is in the same order of magnitude as the concentration of NH 4 OAc (10 mM). The absorption features of the main 525 nm peak are identical for the DD and HH conditions, and only a minor offset can be seen at wavelengths below 470 nm. The emission spectra at room temperature look also similar, with a slightly more pronounced red edge in the DD condition.
While at rst sight this minor deviation might look unimportant, it manifests itself at À196 C as an additional emission band centered around 850 nm (Fig. 1B). When performing timecorrelated single photon counting (TCSPC) measurements in the DD condition at 5, 25, and 40 C, the background amplitude in the decay curves increases from 600 nm to 850 nm (dashed traces, Fig. 2B). This indicates the presence of a long-lived and red-shied luminescence. 14 The ns-lived emission intensities are reported as solid traces in Fig. 2B. At 5 and 25 C the decay curves can be tted satisfactory with a mono-exponential function and only a very small nanosecond (slow) spectral relaxation 12,17,18 is observed at 40 C ( Fig. 2A). Moreover, the emission intensity decreases by increasing the temperature, as shown in the steady-state spectra reported in Fig. S3 (see also Fig. S9A in ref. 12 for HH condition). † TCSPC data, e.g. in Fig. 2B, were recorded with different laser repetition rates and emission attenuations, therefore the intensities of hsi were normalized and the background amplitudes were rescaled dividing by the corresponding hsi emission intensity maxima (the same normalization was also performed for Fig. 3, S4 and S5 †). The background amplitude seems to reach a maximum around 830 nm, but it should be noted that given the drop in the detector sensitivity above 800 nm, the spectral shape might not be represented correctly. However, it still gives a good indication of where the long-lived emission is spectrally located.
Similar results were found when DNA-Ag 16 NCs were rst synthesized in H 2 O and then the solvent was changed to D 2 O for the measurements (see Fig. S4 and Section 3 in the ESI †). Fig. 2C shows decay curves detected at 810 nm in the DD condition, exciting at 531 nm with a Xe ash lamp (repetition rate ¼ 300 Hz). A bi-exponential tail t was used to determine the ms decay time (hs ms i) in the 5 to 40 C range (Table 1). When DNA-Ag 16 NCs are measured in H 2 O, the amount of ms-lived emission at 25 C can be considered negligible, as shown in Fig. S5B and S6. †  As mentioned above, freezing DNA-Ag 16 NCs with liquid nitrogen (À196 C) makes the ms-lived emission band (l max z 850 nm) much more pronounced. The ns-lived uorescence is blue-shied for both DD and HH conditions with a maximum around 690 nm. When performing time-resolved measurements with a Xe ash lamp (repetition rate ¼ 300 Hz) at À196 C (Fig. 3C, S7 † and Table 1), ms-lived emission can be observed for both the DD and HH conditions. The long-lived emission is more pronounced for the DD condition with hs ms i ¼ 447 ms, in line with the steady-state results presented in Fig. 1B. A decay time of 245 ms was instead found for DNA-Ag 16 NCs in the HH condition. In addition, based on a recent article by Petty et al., 15 we decided to measure the ms-lived emission of DNA-Ag 16 NCs in the DD condition at À196 C using the burst excitation mode with a picosecond-pulsed laser (IRF z 150 ps). 19,20 The results can be found in Fig. S8 † and very similar values (ranging from 445 to 457 ms for both the rise and decay times of the ms-lived emission) were obtained. So far, the presented data shows that the ms-lived emissive state is more clearly visible in D 2 O with respect to H 2 O. Is this because the longlived emissive state is less quenched in D 2 O than H 2 O, or that D 2 O promotes the formation of the long-lived state, or a combination of both? Qualitatively, the reduced quenching by D 2 O (hs ms i is 1.82 times longer at À196 C) is not enough to explain the much larger intensity increase of the 850 nm emission in DD versus HH condition (Fig. 1B). Hence, D 2 O must increase both the quantum yield of formation of the mslived state and its decay time compared to H 2 O. This seems to agree with the nding that the quantum yield of uorescence is lower in the DD condition versus the HH condition ( Fig. S9 and Table S2 †). The mechanistic origin of why deuterium would enhance the transition to the ms-lived state is currently not understood, but we speculate that differences in the vibrational frequency of the X-D versus the X-H bonds (X being nitrogen or oxygen) and/or changes in the strength of the hydrogen bonding network 9,21 promote the formation of a Frank-Condon (FC) state that increases the likelihood of mslived state formation.
The unusual effect of deuterium on DNA-Ag 16 NCs can also be seen in the ns-lived emission. Contrary to the vast majority of uorophores, 5,8 hsi in the DD condition is shorter than that in HH at all temperatures (see Table 1). To conrm that hydrogen and deuterium can dynamically exchange in the DNA-Ag 16 NC structure, two additional conditions were created and measured (HD and DH, see Table 1 for description). Table 1 shows that the solvent used during the synthesis of DNA-Ag 16 NCs is not important for the observed behavior; only the solvent in which the measurements are performed determines the spectroscopic properties. In addition, when DNA-Ag 16 NCs were measured in a 1 : 1 mixture of D 2 O and H 2 O (DH 50 /D 50 or HH 50 /D 50 ), hsi and Q were found to be in between the DD and HH values (Tables 1, S2 and Fig. S9-S11 †). While overall smaller, the ns decay time in D 2 O was signicantly less temperature-dependent than in H 2 O (Table 1).
Based on available data from organic uorophores, 5 it is rather unlikely that the shortening of the ns-lived state can be attributed to a more efficient non-radiative quenching by D 2 O versus H 2 O. Instead, the origin of the lower ns-decay time in D 2 O must be related to another effect. We have observed previously for a red-emissive DNA-AgNC that, when immobilized in a polymer lm and studied at the single molecule level, hsi is oppositely correlated with the ability to form long-lived (ms) dark states. 22 In the latter case, the dark state formation was probed by optically activated delayed uorescence (OADF). 22 Intriguingly, Fig. 3A shows that hsi at À196 C increases from 650 nm (2.19 ns) to 600 nm (2.72 ns). This might indicate the presence of a blue-shied population that has a longer decay time and perhaps less ms-lived state formation. Future excitation-wavelength dependent ratiometric studies of the nslived versus ms-lived emission could potentially shed light on this. When DNA-Ag 16 NC are frozen in the HH condition, hsi is signicantly longer (5.22 ns), as is usually observed when suppressing temperature-dependent non-radiative decay pathways.

DNA721-AgNC
DNA721-AgNC emits in the same range as DNA-Ag 16 NC, but has a distinctly different spectroscopic behavior. 13 Details on the synthesis, HPLC purication and the collected DNA721-AgNC fraction can be found in the ESI and Fig. S13. † Fig. 4A shows the normalized absorption and steady-state emission spectra of DNA721-AgNCs in the DD and HH conditions. Both emission spectra, as well as the 640 nm absorption features, are identical and overlap perfectly. The only difference is an extra absorption band around 400 nm in the DD condition, which was not previously observed in H 2 O. 13 This feature is absent in the excitation spectrum (Fig. S14 †), thus we can exclude that D 2 O induces an additional electronic transition at 400 nm. It is instead very likely that the absorption bump for the DD condition is due to the presence of an impurity collected during the HPLC run. Freezing the DNA721-AgNC sample with liquid nitrogen in the DD and HH conditions blue-shis the emission maximum to 706 nm (Fig. 4B).
TCSPC measurements yielded differences in hsi (see Table 2) for the DD versus HH condition. For example, at 25 C, hsi is 4.36 ns in the DD condition and 3.72 ns for the HH condition. Unlike DNA-Ag 16 NCs, D 2 O seems to affect the behavior of DNA721-AgNCs in a more expected way, lengthening hsi and increasing Q. 5 For both the HH 13 and DD conditions, hsi is rather temperatureindependent in the 5-40 C range, while the emission intensity and absorbance are temperature-dependent. This means that hsi and Q are not interdependent and the classic three-level model where Q ¼ k f hsi (k f being the radiative rate constant) does not apply unless one introduces static quenching. 23 We have previously suggested that for some DNA-AgNCs, a phenomenological four-level model, originally introduced by Patel et al., 24 can explain the observed relationship between Q and hsi. 17,25 In this model, the DNA-AgNC is excited into a FC state that evolves ultrafast (sub-ps) 24 either back to the ground state, a ms-lived state, or the ns-lived emissive state. The rst two pathways (to the ground state and to the ms-lived state) can be considered as a type of static quenching with regard to the emission from the ns-lived state. Intensity-weighted average decay times hsi, obtained from decay curves recorded at the indicated emission wavelength (l exc ¼ 531 nm). c Decays measured at 720 nm, since the emission maximum is blue-shied at À196 C. d Microsecond intensity-weighted average decay times hs ms i, obtained from decay curves recorded at the indicated emission wavelength, exciting at 531 nm with a Xe ash lamp (repetition rate ¼ 300 Hz). e <: amplitude too low to determine the decay time. f IRF: IRF-limited decay time. Graphical representations of the data can be found in Fig. S9-S11. Hence, Q becomes the product of Q S1 (quantum yield of nslived state formation) and Q f (the quantum yield of uorescence from the emissive state to the ground state). 17 Since hsi is largely temperature-independent, the main cause for changes in Q is the temperature-dependent change of Q S1 .
We tried to test this hypothesis by plotting hsi as a function of Q calculated for different solvent and temperature conditions. Q values were determined using the previously reported 0.73 (HH condition at 25 C) as reference value. 13 Fig. 5 shows that hsi vs. Q, for both H 2 O and D 2 O measurement conditions, follows a linear trend that does not intercept the origin (0,0).
It is also worth noticing that Q for the DD condition at 5 C reaches unity. While Fig. 5 illustrates that the "classic" threelevel model is not applicable for DNA721-AgNCs, Petty et al. have recently demonstrated that the dual emission of a greenand NIR-emitting DNA-AgNC can be described by this model. 15 The latter highlights the need for electronic structure calculations in order to help interpret the experimental data from different DNA-AgNCs. 2 Unlike DNA-Ag 16 NC, DNA721-AgNC displays no signicant ms-lived emission either in the liquid (e.g. 5 to 40 C) or the frozen (À196 C) state. This is in line with the steady-state data in Fig. 4 where no additional emission band up to 850 nm appears. However, we have previously reported 13 that DNA721-AgNCs can form long-lived states that can be optically depleted yielding OADF. 13,25,27 This means that the ms-lived states in DNA721-AgNC are either dark or emit in a NIR range signicantly beyond our detection window. While no crystal structure information is available for DNA721-AgNC, its hydrodynamic volume (19.6 nm 3 ) 13 is signicantly larger than that of DNA-Ag 16 NC (10.5 nm 3 ). 12 The two DNA-AgNCs could have different levels of solvent accessibility to the AgNC, but in both cases the nal measurement solvent determines the properties, indicating reasonable exchangeability for both.  S15 for more details). Note: changes in the refractive index (both temperature and isotope changes) were ignored since they are in the 1% difference range. A graphical representation of hsi data as a function of temperature can be found in Fig. S16. edata not measured. Intensity-weighted average decay time hsi of DNA721-AgNCs as a function of fluorescence quantum yield (Q), for different solvent and temperature conditions. The HH condition at 25 C was used as the reference quantum yield (0.73), 13 and the other Q values were determined from single emission and absorption spectra at the specified condition (Fig. S15 †). 26 Note that changes in the refractive index (both temperature-and isotope-dependent changes) were ignored since they are in the 1% difference range. The dashed lines represent the linear fit of the blue and red data points, respectively.

Conclusions
We have shown that exchanging H 2 O with D 2 O as solvent can have diverse effects for different DNA-AgNCs. For DNA-Ag 16 NCs, hsi shortened, which is in contrast with the behavior observed for DNA721-AgNCs and most organic uorophores. 5 D 2 O also enhanced the formation of the ms-lived state and lengthened hs ms i for DNA-Ag 16 NCs. While the mechanistic origin is not understood, we hypothesize that the difference in the vibrational frequencies of the X-D versus the X-H bonds and/or changes in the strength of the hydrogen bonding network affect the excited state pathways. For DNA721-AgNCs, D 2 O lengthened hsi and increased Q compared to H 2 O. At 5 C in D 2 O, Q even approaches unity. Furthermore, we have demonstrated for DNA721-AgNCs that the temperature-dependent changes of Q are not reected in the hsi values. This indicates that the variations in Q are not due to the changes in the non-radiative decay rates from the emissive state, but are mostly caused by the changes in the quantum yield of the emissive state formation (Q S1 ). 17 We hope that our results will stimulate further research in obtaining high uorescence quantum yield NIR-emitting DNA-AgNCs.

Data availability
Experimental data and details on the experimental procedures are provided in the ESI. †

Conflicts of interest
There are no conicts to declare.