Ultrafast dynamics of formation and autodetachment of a dipole-bound state in an open-shell π-stacked dimer anion

Formation and mode-specific autodetachment from a dipole-bound state in a radical anion dimer is observed in the frequency and time-domains.


Introduction
The interplay between p-stacking and hydrogen-bonding interactions in large anionic systems is central to a range of physical processes in chemistry, biology and technology. These interactions govern, for example, anion-recognition processes in supramolecular assemblies, 1,2 protein structure and function, 3,4 and solution-phase anion aggregation. 5,6 Excited states of these complexes are important in molecular electronics, [7][8][9][10][11] may be involved in facilitating electron transfer in biological systems, [12][13][14] and play a key role in the interaction of low energy ballistic electrons with DNA. [15][16][17][18][19][20] However, gaining a detailed molecular level understanding of anion excited states and their non-adiabatic dynamics is generally hampered by the complex nature and environment of such systems. It is therefore appropriate to consider small p-stacked aggregates as simple models. The gas phase offers an idealised environment in which the desired intra-and intermolecular dynamics can be probed without the added complexity of surroundings. To date, most studies of open-shell p-stacked anion systems have focused around characterising the localised vs. delocalised character of ground and excited states. [21][22][23][24][25][26][27] However, the dynamics and timescales available to an excess electron in electronically excited states of a prototypical p-stacked system, or the competition between localised (intramolecular) and delocalised (intermolecular) dynamics, have not been experimentally characterised despite their importance in determining their photochemistry and physics.
Here, we consider the coenzyme Q 0 dimer radical anion, (CQ 0 ) 2 À , the calculated minimum energy structure of which is shown in Fig. 1. (CQ 0 ) 2 À is a useful prototype system because of the strong gas phase stability of one conformer, the wellunderstood dynamics of the isolated monomer radical anion, CQ 0 attached to the para-quinone ring that serves to enhance lipid miscibility. It is thought that the synergetic co-operation of hydrogen bonding and p-stacking of the para-quinone ring in biological systems play important roles in facilitating electron transfer. 12,[33][34][35] Anionic excited states of para-quinones have been proposed as possible bypasses of the inverted Marcus region, [36][37][38][39][40] thus providing a more efficient electron transfer route. Recent gas phase studies on a series of monomer paraquinone radical anions have shown that photoexcitation crosssections to quasi-bound p*-resonances can be larger than direct photodetachment. 28,[40][41][42] These resonances are valence-localised excited states that are embedded in the detachment continuum with inherent autodetachment lifetimes ranging from tens to hundreds of femtoseconds. [43][44][45][46] Despite their transient existence, internal conversion can compete with prompt autodetachment. 28,[40][41][42]47 Resonances can generally be divided into two types: shape resonances, in which the extra electron occupies an unlled valence orbital of the neutral ground electronic state conguration ($10 À14 s lifetimes); and Feshbach resonances, in which the corresponding neutral core is predominately in an electronically excited state (typically 10 À14 to 10 À12 s lifetimes). We have recently shown that anion photoelectron (PE) imaging is ideally suited to probe the dynamics of anion resonances, 28,[40][41][42]48,49 allowing identication of signatures from: direct photodetachment into the continuum; prompt autodetachment from a photoexcited resonance; delayed autodetachment from an excited state following internal conversion or extensive nuclear motion; and thermionic emission. Because the spectral contributions from direct photodetachment and prompt autodetachment cannot be easily resolved, they are jointly labelled as prompt detachment.
Neutral molecules or clusters, such as (CQ 0 ) 2 , with an electric dipole moment, |m| > 2.5 D, can support a dipole-bound state (DBS), 46,50 in which an excess electron is weakly bound (10-100's meV) in a highly diffuse (non-valence-localised) orbital located at the positive end of m. Due to the diffuse nature of a DBS, the direct photoexcitation cross-section from a valence-bound state is usually very small. In contrast, photodetachment crosssections of a DBS can be very large and increase with decreasing photon energy. 51 DBSs are believed to play an important role in the formation of anions in the interstellar medium and low energy electron capture biochemical systems such as DNA. [15][16][17][18][19][20]46,52,53 Here, a combined frequency-, angle-, and time-resolved photoelectron imaging (FAT-PI) 41 and electronic structure study on the spectroscopy and dynamics of excited states of pstacked (CQ 0 ) 2 À is presented. The frequency-and angleresolved dimensions involve recording single-photon PE images (spectra) at many different photon energies (hn) to identify trends and ngerprints of resonances and their associated dynamics. A selected resonance can then be photoexcited and the resulting dynamics monitored in real-time using time-resolved PE imaging. Supporting ab initio calculations using multi-state XMCQDPT2 theory with a large CASSCF reference space allow clear assignment of the experimental dynamics. 54 Complete experimental and theoretical details are given in the ESI. † We show that the excess electron in the ground electronic state exists as a localised charge on one monomer. Photoexcitation of a selected resonance at hn ¼ 3.10 eV with high intermolecular charge-transfer character leads to the formation of a DBS on a $60 fs timescale, which then undergoes vibration-mediated autodetachment on a 2.0 AE 0.2 ps timescale. This determination represents the rst direct observation of the conversion of above-threshold valencelocalised population to a DBS, and the rst real-time characterisation of vibration-mediated autodetachment of a DBS. At slightly higher hn, a competition between non-adiabatic dynamics leading to the DBS and autodetachment dynamics has been observed. The competition is assigned to the interplay between dimer and monomer-like dynamics, and further highlights the role of p-stacking in inuencing the excited state dynamics.

Results and analysis
Frequency-resolved photoelectron imaging Three example PE spectra of (CQ 0 ) 2 À are shown in Fig. 2, 13 eV spectra are also compared with selected PE spectra of CQ 0 À that extend to the same maximum in electron kinetic energy (eKE). 28 In both cases there is a difference in the adiabatic detachment energy (ADE) between CQ 0 À and (CQ 0 ) 2 À of $1.2 eV. The spectra for (CQ 0 ) 2 À and CQ 0 À broadly exhibit similar spectral features, except the dimer generally shows an increased yield of low-eKE electrons. The PE spectrum in Fig. 2(c) is the average of a number of spectra at low hn, all of which are identical within noise. These spectra show three (perhaps four) reproducible partially resolved features that have the appearance of vibrational structure. It is remarkable that vibrational-like structure is discernible for such a large molecular system at 300 K, and implies a mode-specic detachment process. , adopting a distorted sandwich geometry. The excess electron is localised on the left (planar) monomer in the ground electronic state. The right (nonplanar) monomer has the carbonyl (C]O) groups bent out of the ring plane by $10 . Key: charcoalcarbon; redoxygen; whitehydrogen; hydrogen bondsblue dashed. Fig. 3(a) shows the frequency-resolved PE spectra (37 in total) as a two-dimensional intensity plot. Each spectrum has been normalised to have unit total area for clarity. A number of trends are immediately evident. For hn < 3.0 eV, denoted as region (i), all PE spectra are essentially identical (see Fig. 2(c)). Between 3.3 $ hn $ 3.7 eV, there are two PE features: a high-eKE feature that is broadly consistent with prompt detachment; and the low-eKE feature observed for hn < 3.0 eV. Region (ii) in Fig. 3(a) shows a modulation in intensity between these high-eKE and low-eKE features. Two representative PE spectra from region (ii) are shown in Fig. 3(b), which illustrate that depletion of high-eKE signal is concomitant with an increase of low-eKE signal. Between 4.0 $ hn $ 4.5 eV the PE spectra resemble that of the isolated monomer ( Fig. 2(a)), 28 although (CQ 0 ) 2 À has an increased yield of PE signal in the eKE # 0.2 eV range.
To analyse the detachment channel contributions in Fig. 3(a), all frequency-resolved PE spectra were tted with a three channel global model that is detailed in the ESI. † The three channels are labelled as DBD, DA, and PD. The DBD channel describes the low-eKE vibrational distribution shown in Fig. 2(c). The DA feature is centred at eKE ¼ 0.22 AE 0.04 eV regardless of hn, while the centre of the PD feature increases linearly with hn. The relative contributions of the three channels are shown in Fig. 3(c). An example t of each channel is given in the in hn ¼ 4.66 eV PE spectrum in Fig. 2(a). Fig. 3(c) shows that the DBD channel is dominant for hn < 3.0 eV, while the PD channel becomes dominant for hn > 3.0 eV. The modulation in region (ii) is reproduced between the PD and DBD channels. For hn > 3.5 eV, the DA channel becomes available and the contribution of the DBD channel is minimal ($5% at hn ¼ 4.66 eV).  The adiabatic detachment energy (ADE) was determined in the global t by extrapolating the rising edge of the PD feature for all frequency resolved PE spectra. Similarly, the vertical detachment energy (VDE) was determined from the maximum of the PD feature in the global t. These data are tabulated in Table 1. Note that some PE signal can be observed below ADE due to the nite temperature of the ions ($300 K), which corresponds to $0.6 eV of internal energy.
PE angular distributions associated with Fig. 3(a), quantied in terms of the conventional b 2 parameter (À1 # b 2 # 2), 55,56 are given in the ESI. † b 2 values of À1 and +2 correspond to electron ejection perpendicular and parallel to the laser polarisation, 3, respectively. In contrast to CQ 0 À , 28 the angular distributions do not exhibit changes in anisotropy that could reect changes in detachment channel contributions or dynamics.

Photodetachment yield spectroscopy
To further investigate the DBD and PD channel modulation in region (ii) ( Fig. 3(a)), the total photodetachment yield spectrum spanning 3.2 < hn < 4.0 is given in Fig. 4. The yield spectrum shows a broad maximum centred at hn $ 3.75 eV, with its red edge overlapping with the modulations. While the total PE yield shows some reproducible closely-spaced oscillations, their extent does not account for the observed channel modulation. Instead, the modulations appear to result from a competition between processes yielding the two detachment channels rather than any sharp changes in the total photodetachment crosssection. Hence, the presentation of normalised PE spectra in Fig. 3(b) is representative of their relative intensity.

Ground state calculations
The calculated minimum energy geometry of (CQ 0 ) 2 À is shown in Fig. 1(b). The structure exhibits a distorted p-stack involving ve hydrogen bonds of length < 2.8Å. The separation between the centres of each para-quinone ring is 3.9Å, which is the same as the neutral p-stacked benzene dimer. 57 The two monomers in the (CQ 0 ) 2 À equilibrium geometry ( Fig. 1 The extent to which the anionic ground state is localised or delocalised can be estimated from population analysis of each monomer. The MP2//GEN1 calculations produced MP2-density Mulliken (natural bond order, NBO, 58 in parentheses) charges of À0.94 (À0.96) and À0.06 (À0.03) for the planar and non-planar monomers, respectively, implying a ground state anion in which the electron is localised on the right hand (non-planar) monomer in Fig. 1(b). The uB97XD/GEN2 and CASSCF calculations gave similar populations. The non-planar geometry of one of the monomers in the dimer anion but not in the dimer neutral appears to result from differing dispersion interactions between the two species.
Systematic conformation searches (from both semi-empirical UPM6 geometries and hand-oriented starting geometries that were re-optimised at the uB97XD//GEN1 level of theory) support the structure shown in Fig. 1(b) should be statistically predominant (>95%) in the experiment, assuming electrospray and ion thermalisation (trapping) recovers thermodynamic structures. 49 Calculated photodetachment energetics (including zeropoint energy) are summarised in Table 1, and are overall in very good agreement with experiment. The ve hydrogen bonds identied in Fig. 1(b) are the main dimer cohesion force. 59 The calculated adiabatic bond dissociation energy (BDE) of $0.8 eV is in reasonable agreement with that determined from the experiment (1.0 AE 0.2 eV) using the approximate relation: ADE  needs to be approached with particular caution in order to disentangle continuum effects. 46 The resonance energetics and oscillator strengths allow the broad feature in the photodetachment yield spectrum (Fig. 4) to be assigned: there are two overlapping photoexcitation proles arising from the intermolecular 7 2 [S] (oscillator strength $ 0.05) and intramolecular 8 2 [S] (oscillator strength $ 0.22) resonances. The spectrum has therefore been modelled using two Gaussians, plus an underlying baseline for prompt detachment and 9 2 [F] contributions at high hn. 63 The tted Gaussian positions are in good accord with the calculated energetics, and tted widths of $0.2 eV ($40 fs) in good accord with other studies. [20][21][22][23][24][25][26][27][28][40][41][42]49 From the CASSCF wavefunction characters (see ESI †), Franck-Condon (FC) photoexcitation of the intermolecular 5 2

Dipole-bound state calculations
Calculated values of |m| in the ground electronic state of (CQ 0 ) 2 are $6.9 D and $5.5 D at the optimised (CQ 0 ) 2 À and (CQ 0 ) 2 geometries, respectively, supporting DBSs with binding energy of $150 meV and $50 meV. However, the oscillator strength for direct photoexcitation of the DBS is $10 À4 to 10 À5 , which is small compared with those for valence-localised resonances. Fig. 6 shows m at the (CQ 0 ) 2 À and (CQ 0 ) 2 geometries. At the (CQ 0 ) 2 À geometry, m is oriented between the p-stacked monomers and close to parallel with the planar monomer ring and the chord joining the two carbonyl groups on the non-planar monomer. In contrast, at the (CQ 0 ) 2 geometry the orientation of m is almost orthogonal to the monomer ring planes. Calculations connecting the ground electronic state (CQ 0 ) 2 À to neutral geometries reveal that changes of the carbonyl tilt angle, q, associated with wagging modes of the non-planar monomer (and the concomitant contraction of C]O bonds on the planar monomer), are the principal geometrical changes responsible for the large change in orientation of m. These C]O wagging modes are also FC active following intermolecular photoexcitation. The rst excited (CQ 0 ) 2 state, 1 1 A at hn $ 4.5 eV (not shown in Fig. 5), has |m| $ 7.1 D, which also supports a DBS with binding energy $ 180 meV.

Time-resolved photoelectron imaging
The dynamics and origin of the DBD channel were investigated in real-time using pump-probe femtosecond PE imaging  following photoexcitation of the charge-transfer 5 2 [F] resonance. In these measurements, a femtosecond pump pulse photoexcites at hn $ 3.10 eV, and a second probe pulse monitors the excited state population aer some time delay, Dt. Results of the 3.10 + 1.55 eV (pump + probe) time-resolved measurements are summarised in Fig. 7. Fig. 7(a) shows four selected pump-probe spectra, obtained by subtracting the background spectra (for Dt ( 0) from each of Dt $ 0 spectra. Fig. 7(b) shows the total integrated pump-probe PE signal as a function of Dt, which reveals two timescales. The total PE signal was tted with two functions: a Gaussian cross-correlation convoluted with an exponential decay for the fast component; and a cross-correlation function convoluted with an exponential rise and decay for the slow component. The fast component lifetime, t 1 $ 60 fs, is limited by the experimental cross-correlation. The slow component reaches a maximum contribution aer the fast component has decayed, and subsequently decays with a lifetime of t 2 ¼ 2.0 AE 0.2 ps. From Fig. 7(a), t 1 is associated with the development and evolution of a broad PE feature into two narrow features, peaking at low-eKE and eKE ¼ 1.6 eV. The high eKE feature is very close to the 1.55 eV probe energy and has a highly anisotropic PE angular distribution with b 2 $ +2 (see inset). The anisotropy suggests that the outgoing photoelectron has p-wave character, and therefore that the original orbital from which the electron is detached has s-character. 55 Timescale t 2 is associated with the concerted decay of both pump-probe features.
Similar time-resolved measurements were performed with a 1.05 eV probe, which are summarised in Fig. 8. Selected background-subtracted pump-probe spectra in Fig. 8(a) show an initial broad distribution in the eKE > 0.25 eV window that rapidly sharpens (sub-100 fs, t 1 ) to a narrow distribution. This PE feature is again situated at the probe photon energy with b 2 $ +2 angular character, and decays on a t 2 $ 2.0 ps timescale. These observations are in agreement with the 1.55 eV probe experiments. However, concerted with the changes at high-eKE, there is now a bleach of the eKE < 0.25 eV signal, which recovers on the same t 2 $ 2.0 ps timescale. Fig. 8(b) shows the integrated PE yield of the two features, which conrms that decay of signal in the high-eKE window is mirrored by recovery at low-eKE. The total time-resolved PE signal is invariant with Dt.

Discussion
Ground electronic state (CQ 0 ) 2 À Our calculations show that the dimer radical anion can be thought of as a molecular cluster of type (CQ 0 À )CQ 0 , composed of a localised planar monomer anion solvated by the second non-planar monomer. The frequency-resolved PE spectra broadly support this conclusion based on the similarity with CQ 0 À spectra when hn is red-shied by $1.2 eV (Fig. 2). At high hn, the eKE distribution that increases commensurate with hn can be assigned to prompt detachment, PD, which arises from the combination of direct photodetachment into the continuum and fast autodetachment from resonances without signicant nuclear motion.

Dipole-bound state: identication, formation and autodetachment
The DBD channel in Fig. 2 and 3 is characterised by partially resolved vibrational structure at low-eKE (Fig. 2(c)). The observation of discrete vibrations combined with an eKE distribution that does not change with hn indicates an indirect vibrationmediated detachment. Since no similar detachment channel was observed with CQ 0 À , 28 the channel must result from dimer formation. The time-resolved measurements provide an unequivocal assignment of the DBD channel to result from autodetachment of a DBS. That is, the sharp pump-probe feature suggests that the potential energy surfaces for the excited anion state and the neutral ground state are similar, while the close correlation between the probe energy and the eKE of the peak suggest a very weakly bound excited state. In addition, the PE angular distribution implies that the orbital from which the electron is detached has s-character, also supporting a DBS. The associated depletion of the DBD feature in Fig. 8(b) arises because population is removed from the DBS by the probe. However, the bleach of the DBD channel is not observed with the 1.55 eV probe, probably because this probe is close to the energy difference between the neutral X 1 A and 1 1 A states. Since both of these neutral states support a DBS, photoexcitation between the two DBS is probably efficient due to a large photoexcitation cross-section. Autodetachment from the DBS associated with the 1 1 A neutral will lead to low-eKE signal that overlaps with the X 1 A state DBD depletion. Before the DBS is formed, the time-resolved measurements showed a transient broad PE feature that sharpens to the DBS feature on a t 1 < 60 fs lifetime. At the initial pump energy of 3.10 eV, photoexcitation is resonant with the optically-active 5 2 [F] resonance, which in accord with our other studies on monomer para-quinone anions, 28,[40][41][42] should produce a broad pumpprobe feature. The evolution of this broad PE feature into the narrow DBS feature reects internal conversion of 5 2 [F] population to the DBS on a <60 fs timescale. The internal conversion lifetime of <60 fs is similar to those measured in our earlier para-quinone monomer anion studies. 28 state over the DBS, which facilitates a curve crossing. The subsequent t 2 ¼ 2.0 AE 0.2 ps lifetime of the DBS can be assigned to its autodetachment lifetime and leads to the vibrational structure observed in the DBD channel as seen in Fig. 2(c). The overall time-resolved dynamics are summarised schematically in Fig. 9(a), while the detailed interpretation of the 3.10 + 1.05 eV time-resolved measurements are summarised in Fig. 9(b). Note that the t 1 timescale corresponds to only a few vibrational periods, implying that extensive intramolecular vibrational relaxation away from FC modes is unlikely.
The vibrational structure associated with the DBD channel in the frequency-resolved spectra (Fig. 2(c)) indicates that the DBS autodetachment is mode-specic. The most likely vibrational modes responsible are the original FC-active carbonyl wagging modes; these modes induce a large change in orientation of m (see Fig. 6), which is a primary condition that inuences the lifetime of a DBS. 64,65 That is, the carbonyl wagging motion strongly modulates the orbital dening the DBS, which induces a coupling between the neutral and DBS potential energy surfaces. The result is the wagging motion shakes off the weakly bound electron at a kinetic energy proportional to the wagging frequency in accord with the propensity rule for the vibrational quantum number to be reduced by one. 65,66 The calculated energies of the three relevant wagging modes (see ESI †) have been included in Fig. 2(c), which agree well with the DBD vibrational structure. The measured DBS lifetime of t 2 ¼ 2.0 AE 0.2 ps represents the average vibration-mediated autodetachment rate from all contributing vibrational modes. Taking the three groups of wagging modes shown in Fig. 2(c) to have wavenumbers of 200 cm À1 , 400 cm À1 , and 800 cm À1 , the DBS electron is shaken off over ten to forty vibrational wags. In principle, each mode may be expected to exhibit a unique lifetime, which may be observable by integrating the recovery of the DBD feature over specic vibrational modes. However, our data is unable to resolve such a situation for (CQ 0 ) 2 À .
Time-resolved dynamics involving DBSs have been implicated by the Neumark group in their femtosecond PE spectroscopy experiments following photoexcitation of iodine anions coordinated to CH 3 CN, CH 3 NO 2 , or nucleobases. 67-70 CH 3 CN (|m| $ 3.9 D), which does not support a valence-bound anion, exhibits DBS autodetachment on a 4-900 ps timescale. 67 In contrast, the DBS of CH 3 NO 2 (|m| $ 3.5 D) converts on a $400 fs timescale to a valence-bound anion situated $100 meV lower in energy, which is facilitated by a similar vibrational wagging and modulation of m to that for (CQ 0 ) 2 À . [71][72][73] Similar studies on nucleobases characterised DBS lifetimes of 0.3-11 ps before internal conversion to a valence-bound anion situated close in binding energy to the DBS. [68][69][70] Again, electronic structure calculations suggest internal conversion to be facilitated through ring puckering and wagging modes. [68][69][70][74][75][76] The dynamics characterised in the present study suggest that a valence excited state (the 3 2 [F] or 5 2 [F] resonance) can also evolve into a DBS. However, in contrast to the Neumark studies, (CQ 0 ) 2 À does not undergo internal conversion from the DBS to the lower-lying 1 2 A and X 2 A states, probably because any coupling would require very large geometrical distortions (that may not support a DBS), and will be unlikely on the $2 ps DBS autodetachment lifetime. It can therefore be concluded that internal conversion between a valance-localised state and a DBS, in either direction, requires near degeneracy. These trends provide further conrmation that the 3 2 [F] resonance is likely involved in formation of the DBS rather than direct internal conversion from the photoexcited 5 2 [F] resonance. One of the key outcomes from (CQ 0 ) 2 À is evidence of the extent to which the non-adiabatic dynamics are altered by p-stacking. Specically, although the electronic ground state of the dimer represents a localised monomer that is merely solvated, a range of new non-adiabatic dynamics in the continuum are accessed due to the availability of chargetransfer excitations and a cluster DBS. This situation is likely to be common to other cluster anions with similar chromophore/ electrophore groups, and means that the general extrapolation of monomer to cluster dynamics is not trivial. Nevertheless, as will be described next, monomer-localised dynamics can be observed in some circumstances.

Competition between local and non-local dynamics in the continuum
The modulation between the DBD and PD channels in region (ii) of Fig. 3(a), which was specically illustrated in Fig. 3(b), is now considered. From the photodetachment yield spectrum in Fig. 4 The observation of a modulation usually reects a vibrational-specic process. From the example spectra in Fig. 3(b), if the photoexcitation cross-section to the 8 2 [S] resonance is assumed to be constant (i.e., prompt detachment contributions equal), there should be photoexcitation features in Fig. 4 with a $20% enhancement in relative intensity that are concomitant with the modulations in Fig. 3(a). However, while the photodetachment yield spectrum exhibits some sharp features, they do not correlate with the observed modulation, nor do they show a $20% signal enhancement. The modulation between DBD and PD channels therefore appears to arise from a competition in non-adiabatic decay channels from the 7 2 [S] and 8 2 [S] resonances. This conclusion is supported by three additional observations. First, Fig. 3 Fig. 3(a) and (c) have a spacing of $0.1 eV, which is similar to the wavenumber of the main intermolecular FC wagging modes (and IR active) calculated at $800 cm À1 . For 3.55 < hn < 3.8 eV, the 8 2 [S] resonance is predominantly photoexcited since it has a much higher oscillator strength. As this resonance is predominantly localised on the planar monomer, photoexcitation will involve intramolecular FC modes. From Fig. 3(a) and (c), 8 2 [S] photoexcitation does not result in efficient internal conversion to the DBS, rather the PD channel dominates. For hn > 3.75 eV, in which 8 2 [S] is exclusively photoexcited, the competition between DBD and PD channels is no longer observed. Instead, the DA feature centred at eKE ¼ 0.22 AE 0.04 eV becomes available. The DA feature energetically correlates with a delayed autodetachment from the 4 2 [S] resonance, which is also predominantly localised on the planar monomer. The assignment of DA to the 4 2 [S] resonance suggests an internal conversion route from 8 2 [S] photoexcitation. Because extensive nuclear motion (or energy redistribution) cannot occur on the ultrafast lifetimes of resonances, the internal conversion will be facilitated through intramolecular FC modes. In accord, the non-planar monomer effectively acts as a spectator so the detachment dynamics are monomer-like in character. 28 This conclusion is also supported from the hn > 3.75 eV PE spectra, which have a similar appearance to those of CQ 0 À , 28 except blue shied by $1.2 eV due of the increased electron affinity of the dimer.
In summary, the modulation between the DBD and PD channels in the 3.4 $ hn $ 3.8 eV range likely results from a mode-specic competition between non-adiabatic decay pathways of the 7 2 [S] and 8 2 [S] resonances. Such dynamics can only be uncovered by frequency-resolved PE spectroscopy combined with relative cross-section (total PE yield) measurements. A detailed theoretical account of such dynamics will likely be very challenging, particularly due to the participation of the continuum. However, despite the interplay and competition of multiple channels, it is remarkable and encouraging that the technique of frequency-, angle-, time-resolved imaging can provide such rich insight. Overall, to the best of our knowledge, this study presents the rst characterisation of a non-adiabatic dynamics competition between resonances with varying inter-and intramolecular character.

Conclusions
We have demonstrated that intermolecular or charge-transfer photoexcitation can play a signicant role in the excited state non-adiabatic dynamics of a p-stacked dimer radical anion. We have provided the rst direct and real-time evidence of internal conversion of above-threshold resonances into a cluster-supported DBS, and its subsequent vibration-mediated autodetachment. Formation of the DBS is facilitated through chargetransfer photoexcitation by virtue of p-stacking. However, despite the additional complexity introduced by dimerization, monomer-like dynamics can also be observed following photoexcitation of resonances primarily localised on the monomer that supports the excess electron in the dimer ground electronic state. When both inter-and intra-molecular resonance photoexcitation proles overlap, a remarkable competition between 'dimer'and 'monomer'-like non-adiabatic dynamics has been observed. Such interplays between interand intramolecular non-adiabatic dynamics are likely to be common in other similar p-stacked cluster anions, and further illustrate the rich dynamics that can occur in the detachment continuum.