Probing emission of a DNA-stabilized silver nanocluster from the sub-nanosecond to millisecond timescale in a single measurement

A method for measuring emission over a range of sub-nanosecond to millisecond timescales is presented and demonstrated for a DNA-stabilized silver nanocluster (DNA-AgNC) displaying dual emission. This approach allows one to disentangle the temporal evolution of the two spectrally overlapping signals and to determine both the nano- and microsecond decay times of the two emission components, together with the time they take to reach the steady-state equilibrium. Addition of a second near-infrared laser, synchronized with a fixed delay, enables simultaneous characterization of optically activated delayed fluorescence (OADF). For this particular DNA-AgNC, we demonstrate that the microsecond decay times of the luminescent state and the OADF-responsible state are similar, indicating that the OADF process starts from the luminescent state.


Introduction
DNA-stabilized silver nanoclusters (DNA-AgNCs) are a class of emitters with intriguing photophysics dictated by the protecting DNA scaffold. [1][2][3][4][5] Depending on the DNA sequence, DNA-AgNCs with vastly different spectroscopic properties can be obtained. As such, DNA-AgNCs, with emission spanning the visible 6,7 and near-infrared (NIR) 8,9 range, have been prepared with decay times in the nano-10 and microsecond regimes. 11,12 While nanosecond decay times have been reported for DNA-AgNCs using time-correlated single photon counting (TCSPC), longer lived microsecond luminescence has only been very recently reported using pulsed Xe ash lamps or burst mode measurements. 11-13 DNA-AgNCs have long been known to possess microsecond lived dark states, which manifest themselves as uorescence blinking in single molecule and uorescence correlation experiments. [14][15][16] Considered to be non-emissive, these dark states have been utilized to generate background-free imaging capabilities by modulating the uorescence intensity output or by exploiting a process called optically activated delayed uorescence (OADF). [17][18][19] OADF is similar to thermally activated delayed uoresce (TADF), 20 but instead of thermal repopulation, a second NIR excitation source is used to repopulate the uorescent state. Time gating allows one to extract the OADF signal and generate a background free Anti-Stokes signal. 17 The recent discovery of microsecond emission from DNA-AgNCs, 11,12 led us to wonder whether these states are also able to generate OADF.
Since some DNA-AgNCs seem to emit photons over a large time domain, a single experiment covering this entire range will simplify characterization and provide the needed information to create a more complete picture of the excited state processes. The concept of simultaneously determining uorescence and phosphorescence has previously been implemented for lifetime imaging (FLIM/PLIM) with a confocal microscope and a TCSPC counting module. 21,22 In this application, a micro-to millisecond burst of high repetition rate excitation light (which enables to determine the nanosecond lived uorescence) builds up a population of microsecond lived states that, aer turning off the excitation source, allows any long-lived luminescence to fully decay without interfering uorescence. This cycle is repeated until sufficient photons are collected for constructing both the nano-and microsecond decay curves. The burst approach is a convenient way to build up a population of species with a decay time that is longer than the pulse repetition time. 23,24 We present here a similar and generally applicable method for studying systems exhibiting both nano-and microsecond decay times. As an exemplary system, we studied DNA-Ag 16 NC, 25 which shows both nanosecond and microsecond emission at room-temperature (RT). 13 The pulsed burst mode measurements were combined with secondary NIR excitation to elucidate the origin of the state responsible for OADF. By using time gating schemes, we were able to extract the decay time of the state responsible for OADF, which was similar to the decay time of the microsecond lived luminescent state.

Results and discussion
Simultaneous detection of ns uorescence and ms luminescence The principle of simultaneous uorescence and luminescence decay determination by TCSPC is depicted in Fig. 1 and the setup used is described in Fig. S1. † The method relies on assigning a 'macro-time' and a 'micro-time' to each detected photon. The macro-time (T-T 0 ) is the arrival time of a photon from the start of the experiment (T 0 ) to the detection event (T) and has a time resolution of 50 ns from an internal timer. Note that this time is actually the time of the synchronization signal (sync signal) of the next laser pulse and not the actual photon detection event. The micro-time (t) is the photon arrival time compared to the next sync signal and has a sub-ns time resolution.
The assignment of a micro-and macro-time to every photon brings many advantages. Histogramming the micro-times yields the nanosecond uorescence decay curve, while long lived luminescence will appear as an increased background in the micro-time domain (Fig. 1c). To determine the luminescence decay (Fig. 1a), the high repetition rate of the 520 nm excitation source (f Micro ¼ 11 MHz) is kept on for a certain period (T on ) to build up a population of the luminescent state, followed by a period where the excitation source is turned off until the next burst cycle starts (T off ). 11 MHz was chosen since it is a suitable time window to capture the ns decay curve. The sync signal of the high repetition rate laser allows for the determination of the macro-times of the luminescence photons during the T off period and to construct a decay curve by histogramming these photons. While the T on period will contain a combination of uorescence and luminescence, the decay curve during T off can be considered as luminescence only (if the photons in the nanosecond range aer switching off the last pulse are ignored). Knowing both the micro-and macro-times, it is possible to gate the photons to extract specic temporal evolutions (see Fig. S2 † for a representation of all gating schemes used). For instance, the temporal evolution of the luminescence during T on and T off in the macro-time domain can be constructed by only using the photons corresponding to the background level in Fig. 1c. A similar temporal evolution of the uorescence can be constructed in the macro-time domain, allowing to investigate if the equilibrium times for the uorescence and luminescence are alike. The latter was previously demonstrated by Petty et al. and must be the case when all states are connected in a classic three-level system. 12 To verify whether the microsecond lived luminescent state of DNA-Ag 16 NC is capable of OADF, experiments were conducted by co-illuminating with a primary (520 nm) and a secondary (850 nm) excitation source able to regenerate the uorescent state (Fig. 1b). Owing to their different optical path lengths, there is a temporal delay (t delay ) between the two pulses on the nanosecond timescale (Fig. 1c). The temporal evolution of OADF in the macro-time domain can be reconstructed by The micro-time domain represents the photon arrival times compared to the sync (t) and contains fluorescence decay information with sub-ns time resolution; ms luminescence will appear as a constant background together with detector dark counts and after pulsing (Bkg). The temporal dynamics of OADF can be studied by adding a secondary excitation source with delay, t delay .
continuously illuminating with the secondary excitation source and micro-time gating the OADF-related photons ( Fig. S2c †), as is done for all experiments. Note, however, that Fig. 1b shows the general realization of the experiment, where one can control independently the duration (T on2 ) and the macro-time delay (T delay ) of the secondary excitation source with regard to the primary.

DNA-Ag 16 NC
Details on the synthesis and HPLC purication of the collected DNA-Ag 16 NC fraction can be found in the ESI and Fig. S3. † As reported previously, 13 DNA-Ag 16 NC has no signicant long-lived emission in H 2 O at RT, but is clearly dual emissive in D 2 O with a nano-and microsecond emission component (Fig. S4 †). 13 While the long lived emission bears the hallmarks of phosphorescence, we currently do not feel condent enough to assign it to a spin-forbidden transition. The individual emissive components are spectrally resolved when frozen at 77 K with maxima at around 690 nm and 840 nm, but become hard to separate at RT as the combined emission appears as a single broad feature with a maximum around 750 nm. However, through time-resolved measurements, it is possible to disentangle the spectrally overlapping nanosecond emission from the microsecond emission at RT. 13 To maximize the luminescence-to-uorescence ratio, we used an emission lter centered around 850 nm (Fig. S4 †). A 10 mM NH 4 OAc D 2 O solution of DNA-Ag 16 NC was illuminated with 520 nm (1084 W cm À2 , f Micro ¼ 11 MHz, f Macro ¼ 500 Hz, T on ¼ 0.5 ms, and T off ¼ 1.5 ms), and micro-and macro-times of the collected photons were recorded. By binning the micro-times of all the photons, the nanosecond decay is immediately obtained (Fig. 2a), with a background composed of luminescence, detector dark counts and aer pulsing counts.
Binning all the macro-times directly, yields an ungated intensity trace that represents the combination of both uorescence and luminescence contributions (see also Fig. S5; † note that Fig. S5 † is an example at 77 K, while the data in Fig. 2 is recorded at RT). Since each photon has a micro-and macrotime assigned, it is possible to gate photons based on whether they are in or out of the uorescence window (approximately from 12 to 35 ns, see Fig. 2a). Furthermore, the fraction of luminescence in the uorescence window can be calculated (see ESI and Fig. S6 † for further details on the disentanglement of both signals). Thus, the macro-time evolution of uorescence and luminescence can be determined and represented separately (Fig. 2b). Fig. 2b shows that the uorescence intensity drops with time until the steady-state equilibrium is reached. Similarly, the luminescence intensity increases until it also reaches a steady-state equilibrium (see Fig. S7 † for the IRF prole of the burst period). Aer the last 520 nm laser pulse (at 0.5 ms), a steep drop can be observed, as would be expected for the ns-lived uorescence. The decay of the luminescence is slower, spanning from tens to hundreds of microseconds at RT and 77 K, respectively. The constant intensity level from ca. 1 to 1.5 ms in the luminescence trace (Fig. 2b) is due to detector dark counts. This contribution could be subtracted, but it was neglected here given the low number of counts.
With the uorescence and luminescence disentangled, it is possible to extract four parameters from the data in

Intensity dependence of the equilibrium times
Fluorescence and luminescence intensity time traces in the macro-time domain were measured as a function of 520 nm excitation intensity (see Fig. 3 and S9 † for exemplary traces) to further our understanding of the photophysics of DNA-Ag 16 NC.
So far, the phenomenological electronic structures of DNA-AgNCs have been described by either a three-12 or a four-level model. 19 The values s Fl and s Lu,D represent post-laser pulse and post-burst processes on the micro-and macro-timescale, respectively, and hence should be excitation intensityindependent. Indeed, in the entire investigated excitation intensity range, s Fl is constant and centered around 2.17 ns. s Lu,D is fairly constant, however, a minor drop from 76 ms to 73 ms can be observed upon going from 1084 W cm À2 to 23 W cm À2 . We believe that this drop is the result of an environmental change, rather than an intensity-dependent phenomenon. During the  13 Note that the experiment was performed from high to low excitation intensity.
Unlike s Fl and s Lu,D , the equilibrium times of the uorescence (s Fl,E ) and the luminescence (s Lu,E ) should be excitation intensitydependent. For a classic three-level system, s Fl,E would be expected to be equal to s Lu,E , as was recently shown for a greenemitting DNA-AgNC, where the excitation intensities were kept below 50 W cm À2 . 12 s Fl,E and s Lu,E are indeed similar and approach the value of s Lu,D at low excitation intensities. 11 Shorter equilibrium times are obtained for increasing excitation intensities (Fig. 3). It is worth noticing that both equilibrium times could have been satisfactorily tted with a mono-exponential function at low excitation intensities, given the limited number of counts. However, at higher excitation intensities, especially above 400 W cm À2 , s Fl,E was signicantly better tted with a biexponential model (see Fig. S8 †), thus we chose to t also s Lu,E biexponentially. It is unclear why the equilibrium time s Fl,E is biexponential, but a possible explanation could be that there are subsets of populations with different sets of photophysical rates. This suggestion is not unreasonable, since we previously demonstrated for a red-emitting DNA-AgNC that a small fraction of the population had a signicantly higher dark state formation yield in poly vinyl alcohol (PVA). 26 However, this explanation should be considered suggestive at this point and further studies are needed to unravel the origin of the bi-exponential nature of s Fl,E , but this is beyond the scope of this paper.

Simultaneous detection of optically activated delayed uorescence
Previous demonstrations of OADF have been able to infer the microsecond decay time of the dark state by continuously co-illuminating with a secondary CW laser and extrapolating to zero secondary excitation intensity. 18,19 In our method, we use the nanosecond response in the micro-time domain to create an OADF gate for constructing the temporal evolution in the macrotime domain, allowing us to directly extract three OADF-related quantities, s OADF , s OADF,E and s OADF,D , from a single measurement (see below). For OADF measurements on DNA-Ag 16 NC (see Fig. S1 and S7 †), a secondary delayed 850 nm pulsed laser (t delay ¼ 45 ns, T delay ¼ 0 ms, T on ¼ 2 ms) is co-illuminating with the primary 520 nm laser (T on ¼ 0.5 ms). To block the second excitation wavelength, the detection range was shied to 640 nm where the luminescence-to-uorescence ratio is lower (Fig. S4 †).
The luminescence of DNA-Ag 16 NC was collected during coillumination in D 2 O at RT, and the ungated micro-times of all photons were binned as shown in Fig. 4a. A uorescence response is observed when excited by the primary 520 nm laser at 12 ns, and a second smaller response arises at 57 ns due to the secondary 850 nm laser. Similar as for the disentanglement of uorescence from luminescence, it is possible to separate the additional OADF contribution by micro-time gating (Fig. S2c †). Thus, the primary uorescence, OADF, and luminescence signals are individually resolved in the macro-time domain (Fig. 4b).
From these measurements, three additional parameters can be extracted: the nanosecond decay of the OADF signal (s OADF ),     Fig. 4. In agreement with a previous report, 19 roughly the same nanosecond decay time is observed under primary (s Fl ) and secondary (s OADF ) excitation, alluding to the repopulation of the uorescent state upon 850 nm excitation. The equilibrium time of the state responsible for OADF (s OADF,E ¼ 49.8 ms) is similar but slightly lower than the equilibrium time of the luminescent state (s Lu,E ¼ 66.0 ms), while the macro-domain decay times under additional 850 nm excitation are also similar (s OADF,D ¼ 65.8 ms, s Lu,D ¼ 74.8 ms). Note that a previous report showed that s OADF,D can be multi-exponential, however, the amount of OADF counts in our case is rather low to justify going beyond a mono-exponential t. 19 The slight differences between the OADF and luminescence equilibrium and post-burst decay times could be due to the limited numbers of counts for the OADF signal. Despite these minor discrepancies, we feel condent to conclude that the state responsible for OADF is the luminescent state.
Embedment of DNA-AgNCs in a PVA lm has previously shown to enhance the overall OADF contribution (area under OADF decay divided by area under primary uorescence decay like in Fig. 4a). 19 Also for DNA-Ag 16 NC, the OADF contribution increases from 0.7% in a 10 mM NH 4 OAc D 2 O solution to 2.4% in PVA (Fig. S10 †). Furthermore, the ns uorescence decay lengthens to an intensity-averaged decay time, s Fl , of 3.73 ns; s OADF,D also increases to 107 ms. While other DNA-AgNCs have shown upconversion uorescence (UCF) during sole secondary illumination when embedded in PVA, 18 DNA-Ag 16 NCs within PVA exhibit negligible UCF at 6.3 kW cm À2 (Fig. S10 †).
Interestingly, while our results show a seemingly 0.7% OADF efficiency for DNA-Ag 16 NCs in a 10 mM NH 4 OAc D 2 O solution (Fig. 4a), it should be noted that this contains both steady-state and non-steady state contributions (Fig. 4b), and that during T on a signicant population of luminescent states is built up. One should also realize that the 0.7% is not a single 850 nm pulse efficiency as the OADF-responsible state is subjected to thousands of 850 nm pulses during its lifetime. In fact the value of s Lu,D with the secondary light on (74.8 ms) is basically identical to the value of s Lu,D reported in Fig. 3 (ranging from 76 to 73 ms), where no secondary light was present. These ndings are in line with previous work showing that signicantly higher secondary excitation intensities are needed to really observe a shortening of s OADF,D (preferably kW cm À2 to MW cm À2 ). 18,19 Another reason for the absence of any signicant shortening of s OADF,D is that we did not investigate the wavelength dependency of the OADF process for DNA-Ag 16 NCs and perhaps 850 nm is not the most ideal wavelength. 16

Conclusions
We have demonstrated a method for simultaneously measuring nano-and microsecond decay times of a dual emissive DNA-AgNC using TCSPC where each photon is assigned a micro-and a macro-time. Utilizing the combination of micro-and macro-times, we were able to disentangle the spectrally overlapping uorescence and luminescence of DNA-Ag 16 NC into individual time-resolved contributions. Additionally, the method allows one to determine the equilibrium times for establishing steady-state conditions. Co-illumination measurements with a secondary NIR laser enabled the characterization of the decay and equilibrium times of the state responsible for OADF. The macro-domain decay time of the OADF-related state is similar to the luminescent state decay time, demonstrating for the rst time OADF from a luminescent state. In addition to being valuable for characterizing dual emissive DNA-AgNCs in one measurement, we believe that this approach could be of interest for studying a myriad of other systems with emission spanning the nanosecond to millisecond timescale, or where subsequent time-resolved measurements of short-and longlived emission may alter the true picture, only obtainable from simultaneously measuring both quantities. For this method, the only essential parts are the standard TCSPC detection hardware and a pulsed MHz laser that can operate in a burst mode fashion, eventually complemented with a secondary laser for investigating OADF.

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