Unraveling the dynamics of multiple excited states in a single-molecule transistor Authors: Hao Zhang, Lijue Chen, Ziheng Yuan, Jingyu Yang, Yu Zhou, Yixuan Gao, Zhixin Qiu, Wei Xu, Zhexiong Zheng, Cankun Zhang, Haojie Liu, Kai-Qiang Lin, Ye Yang, Jing Li, Junyang Liu, Shixuan Du, Wenjing Hong Published: 2026 ABSTRACT Characterizing charge transport through single molecules provides a fundamental route to explore correlated quantum states, forming the basis for nanodevices in which quantum effects govern operation. During the non-equilibrium electron transfer, a rich manifold of excited states emerges, whose dynamics encode many-body interactions at the single-molecule level. Disentangling these dynamics is crucial for understanding such interactions, however, it has remained elusive due to the experimental challenge of simultaneously achieving both high temporal and energy resolution. Here, we resolved the dynamics of multiple excited states during non-equilibrium charge transport through a single-molecule radical junction using nanosecond differential conductance spectroscopy. The participation of singlet and triplet states, mediated by doublet states, is revealed in both time and energy domains occurring within ~150 ns. In addition, the transient resonance positions evolve by up to ~440 meV over hundreds of nanoseconds, consistent with a dynamic realignment at the molecule–electrode interface during charge dissipation. We interpret this behavior as a signature of Coulomb-renormalized transport resonances rather than as direct evidence for isolated molecular eigenstates shifting in energy. These results provide insight into transient intermediates in single-molecule charge transport and suggest potential routes for controlling quantum states in molecular-scale devices. Introduction Charge transport through a single molecule represents a fundamental model for understanding quantum effects at the nanoscale1–6. It enables the observation of quantum interference, Kondo resonance and Coulomb blockade, reflecting the wave-particle duality of electrons in single molecules7–15. However, charge transport is always accompanied by the formation of short-lived excited states via energy exchange of electron and molecule. Since time-averaged characterization of charge transport hinders deeper understandings, a promising route to attaining this goal is to incorporate time-resolved technique to investigate the nonequilibrium charge transport through single molecules16–19. The state-of-the-art studies of electronic pump-probe coupled scanning probe microscopy (SPM) offer promising insights into non-equilibrium quantum transport with nanosecond time resolution17–22. These techniques provide a powerful platform for probing transient dynamics mediated by electrically induced excited states in single-molecule tunnel junction that are weakly coupled to their environment. However, the typically orthogonal alignment between electron transport direction and dipole moments of excited states renders the weak electric-dipole interaction, which is essential for electrically controlling the intramolecular interactions of excited states. In single-molecule transistors, the electric field is parallel to the dipole of molecule leading to strong interaction23–28, and the external gate potential provides an additional electrostatic field to control the charge state and dynamics of single molecule. Here, we demonstrate the evolution of excited states in the charge transport through a single-molecule transistor by developing nanosecond differential conductance spectroscopy. The concept of our approach is to trigger the formation of charged excited states using a pump voltage pulse and detect the transient states with a subsequent probe voltage pulse as illustrated in Fig. 1. This technique enables the resolution of the energy evolution of intermediate states during quantum transport. Specifically, the molecule is first tuned into a radical state by a static gate voltage. A pump pulse with amplitude Vpump is then applied to raise the molecular energy and induce charge extraction, allowing the molecule to relax into accessible excited states. Subsequently, a probe pulse with amplitude Vprobe, arriving after a delay time Δt, brings the system into resonance with intermediate states, enabling relaxation back to the initial state. By sweeping the delay time Δt, we capture the transient species during quantum transport.Fig. 1Schematic of probing the multi-excited states dynamics in the charge transport through a single-molecule transistor using all-electronic pump-probe method.The delay time Δt is defined by the time difference between the falling edge of pump pulse and the rising edge of probe pulse. The process is divided into three stages. Before pumping, the single-molecule junction is initialized to be in a single-electron state by a static gate voltage. α, gate coupling efficiency. During pumping, the co-tunneling mechanism triggers charge and state transitions36, resulting in the formation of a series of charged excited states upon the application of a high pump voltage. After pumping, these transient states relax (dissipate) over a timescale of hundreds of nanoseconds. These distinct charge states are detected as resonant peaks in the nanosecond differential conductance spectra, where a transient probe pulse drives the system from the intermediate excited states to higher-lying energy levels. Finally, following the dissipation of the excess charge, the single-molecule radical fully relaxes back to the ground state. Schematic of probing the multi-excited states dynamics in the charge transport through a single-molecule transistor using all-electronic pump-probe method. The delay time Δt is defined by the time difference between the falling edge of pump pulse and the rising edge of probe pulse. The process is divided into three stages. Before pumping, the single-molecule junction is initialized to be in a single-electron state by a static gate voltage. α, gate coupling efficiency. During pumping, the co-tunneling mechanism triggers charge and state transitions36, resulting in the formation of a series of charged excited states upon the application of a high pump voltage. After pumping, these transient states relax (dissipate) over a timescale of hundreds of nanoseconds. These distinct charge states are detected as resonant peaks in the nanosecond differential conductance spectra, where a transient probe pulse drives the system from the intermediate excited states to higher-lying energy levels. Finally, following the dissipation of the excess charge, the single-molecule radical fully relaxes back to the ground state. Results Excited state dynamics of single-molecule transistor To investigate charge transport through a single molecule and the dynamics of charged excited states, we perform differential conductance measurement using a constant voltage and a pulse voltage with pulse width at nanosecond scale (Fig. 2a), respectively. As shown in the upper panel of Fig. 2a, the single 1,4-bis(hexyloxy)-2,5-bis(ethynyl-pyrene) benzene (BP) junction is fabricated via π-π stacking between the pyrene anchors and the graphene-based electrodes, with the charge state tuning of the BP molecular junction via a gold back gate insulated by a 10 nm HfO2 layer (see Supplementary Note 2)29. In direct current (DC) measurements (Methods), Fig. 2b shows the presence of a Coulomb diamond in a typical stability diagram at 30 K, indicated by the blue dashed lines. The observation of single resonant transport regime indicating sequential tunneling through single molecular orbital30. The current variations among different single-molecule junctions are evaluated to be within one order of magnitude due to their different contact geometries and gap sizes (Supplementary Fig. 8)31–34. The large sequential tunneling region defined by the Coulomb diamond is supported by a bandgap of 2.71 eV observed from the ultraviolet-visible spectrum (Supplementary Fig. 10), suggesting that the right region is a neutral BP junction (N state) and the left region is the positively charged BP+(N−1 state)35.Fig. 2Charge stability diagram and non-equilibrium dynamics of single-molecule transistor.a Schematic representation of the single-molecule transistor: a single molecule is wired into graphene nanogap to form a molecular circuit, and a local gate electrode separated from the molecule by a thin layer of HfO2 (10 nm thick) is used to shift the molecular energy levels. AWG, arbitrary waveform generator. b Charge stability diagram showing the differential conductance (dISD/dVSD) as a function of bias voltage (VSD) and gate voltage (VG) at 30 K. N (N-1), the neutral (positive) charge state. c The differential conductance of pulsed voltage at a continues VG = 0 V (red lines) and −2.5 V (blue lines), corresponding to N state and N−1 state, respectively. The orange arrow is a guide to resonance peak. d Schematic representation of relaxation dynamics in pump-probe measurements. ΔQ = 0 indicates the initial charge state as ground state. The ΔQ = +1(−1) indicates injecting one hole (electron) from the single molecule. The inset illustrates that the D0+ can be set to a high energy state (dashed line) being able to populate the excited states in other charge states. e Corresponding dependence of the probe current Iprobe on the delay time Δt as a function of pump voltage Vpump at a continuous VG = −2.5 V. The amplitude of Vprobe is fixed at 1 V, with the same polarity as the pump pulse. f Selected Iprobe traces as a function of Vpump, extracted from (e) for different value of Δt. Charge stability diagram and non-equilibrium dynamics of single-molecule transistor. a Schematic representation of the single-molecule transistor: a single molecule is wired into graphene nanogap to form a molecular circuit, and a local gate electrode separated from the molecule by a thin layer of HfO2 (10 nm thick) is used to shift the molecular energy levels. AWG, arbitrary waveform generator. b Charge stability diagram showing the differential conductance (dISD/dVSD) as a function of bias voltage (VSD) and gate voltage (VG) at 30 K. N (N-1), the neutral (positive) charge state. c The differential conductance of pulsed voltage at a continues VG = 0 V (red lines) and −2.5 V (blue lines), corresponding to N state and N−1 state, respectively. The orange arrow is a guide to resonance peak. d Schematic representation of relaxation dynamics in pump-probe measurements. ΔQ = 0 indicates the initial charge state as ground state. The ΔQ = +1(−1) indicates injecting one hole (electron) from the single molecule. The inset illustrates that the D0+ can be set to a high energy state (dashed line) being able to populate the excited states in other charge states. e Corresponding dependence of the probe current Iprobe on the delay time Δt as a function of pump voltage Vpump at a continuous VG = −2.5 V. The amplitude of Vprobe is fixed at 1 V, with the same polarity as the pump pulse. f Selected Iprobe traces as a function of Vpump, extracted from (e) for different value of Δt. To validate the reliability of pulse response, we present the pulsed differential conductance (dIpump/dVpump) spectra in Fig. 2c for each charge state, corresponding to continuous VG = −2.5 V and 0 V, respectively (details in Methods, Single-pulse measurement). A resonant conductance peak at Vpulse = −0.7 V and shoulder feature from +0.5 to +1.0 V after the controllable formation of N−1 state under VG = −2.5 V (lower panel of Fig. 2c), whereas no resonant conductance peak is observed at VG = 0 V (upper panel of Fig. 2c). Since the ground state of BP molecule at VG = −2.5 V is D0+, this resonance is associated with the transition from D0+ to other charged excited states. Compared with the DC measurement, the absence of the peak at VSD = ± 0.1 V and VG = −2.5 V under pulsed measurement may originate from the reduced signal-to-noise ratio as evidenced by the raw pulsed I–V data shown in Supplementary Fig 11. Additionally, the relaxation of transient species, which is highly sensitive to the pulse duration, may further affect the differential conductance, as revealed by the pulse-width-dependent current measurements (Supplementary Fig. 12). To gain the temporal and energy information about the charge injection into single molecule, we probe the dynamics of charged excited states by the all-electronic pump-probe measurement as presented in Fig. 2d (details in Methods, Pump-probe measurement)19–21. When the pump voltage is sufficiently high to interact with the single molecule, it triggers charge transitions as well as the formation of excited states in the new charge state36. After the formation of the charged state, the excited states evolve continuously due to dissipation and eventually relax back to the initial charge state. During this dissipation process, a probe voltage is applied to interrogate the charge state of the single-molecule radical, enabling the acquisition of temporal dynamics. We first obtain the pump-voltage-dependent dynamics under the fixed VG = −2.5 V (Fig. 2e). A reduced-current regime with a lifetime of hundreds of nanoseconds appears when the pump voltage is set around Vpump = ±0.8 V, indicating the formation of transient species. This phenomenon disappears when the gate voltage is set at VG = 0 V (Supplementary Fig. 13). The trigger voltage is consistent with the resonant peak and shoulder in pulsed differential conductance spectroscopy in Fig. 2c. In the negative pump voltage regime presented in Fig. 2f, a clear fast decay component appears in the low pump voltage regime (Vpump < −0.4 V), followed by a slow decay component at higher pump voltage regime. A similar feature appears in the positive pump regime; however, it exhibits a faster decay. The fine features of the pump–probe data are highly sensitive to the molecular orientation and gap size, which affect the charge transfer rates and result in the incomparability of different devices (Supplementary Note 4). Despite this, we successfully reproduce similar dynamic behavior in device 2 of BP (Supplementary Fig. 14), where the transient relaxation dynamics emerge with increasing pump voltage and persist for hundreds of nanoseconds. Moreover, we demonstrate control of the relaxation dynamics by using the gate voltage to tune the energy level alignment in device 3 of BP (Supplementary Fig. 15). This demonstrates that gate coupling provides direct electrostatic control of molecular relaxation dynamics, independent of transport bias. Hidden states revealed by nanosecond dynamics We quantitatively extract the pump-voltage-dependent lifetime from Fig. 2e and Supplementary Fig. 13 to reveal the hidden states after pump excitation, as presented in Fig. 3a. The central energy of its first order derivative indicates the resonance energy of −0.68 eV associated with the excited state, which is consistent with the resonance position observed in the single-pulse differential conductance spectrum. The observed broadening originates from the coupling between the molecule and the electrodes, leading to phonon-induced broadening of the electronic states. Notably, prior to reaching this resonance energy, we observe a fast decay component with a lifetime of ~68 ns that exhibits a plateau-like feature. This indicates that one non-equilibrium state is accessible even at low applied voltages. We attribute this behavior to the gate voltage being tuned near the charge degeneracy point (cross point) of the N and N−1 states37, where a slight voltage applied can trigger the charge filling and the charge transition is expected to be faster. Therefore, we attribute the short lifetime plateau observed in the low pump voltage regime to the formation of S038. Interestingly, the non-quantized correlation between lifetime and pump voltage around Vpump = −0.8 V is distinct from the quantized excited state dynamics in an individual naphthalocyanine molecule where the lifetime against the voltage exhibits the plateau-like feature39. Such a discrepancy may suggest the involvement of multiple excited states, potentially with coherent interactions, extending beyond a simple picture solely based on charge-state transitions.Fig. 3Excited-states dynamics of single molecules with different charge modulated by pump voltage.a The lifetimes as a function of trigger voltage Vpump. The black sphere represents the transition lifetime τ with error bar obtained through single exponential fitting in the function of Iprobe and delay time Δt from Fig. 2e and Supplementary Fig 13, along with the fitted functions derivative (red line). Red and blue areas indicate two distinct regimes corresponding to D0+ to S0 and D0+ to S02+/T12+. The error bars represent the uncertainty associated with the single-exponential decay fitting. b Many-body schematics of the possible excited state in N−2, N−1 and N electrons. The intrastate energy space is obtained by TDDFT calculation. The y-axis (E) represents the absolute energy relative to the vacuum level. Brown crosses indicate the excluded transitions. The red arrow indicates the excitation from N−1 to N state, and the blue arrows indicate the N−1 to N−2 transition, corresponding to the red area and blue area in (a). Single-particle illustration of N − 1 → N (c) and N − 1 → N − 2 (d) transition via co-tunneling. The orbital energies of the N − 1 state relative to the vacuum level are obtained from our calculations. With the gate tuned near the N − 1/N degeneracy point, the Fermi level (EF ≈ − 7 eV) is approximately aligned with the SUMO orbital (−6.85 eV). Under negative bias in a nearly symmetric junction, the drain chemical potential shifts upward while the source shifts downward. At low Vpump, resonance between the drain and the α-SUMO enables electron injection, forming S0 (c). At higher Vpump, deeper occupied orbitals become accessible for electron extraction (d): removal from the β-orbital is associated with the S02+ state, while removal from the α-orbital corresponds to the T12+state. Excited-states dynamics of single molecules with different charge modulated by pump voltage. a The lifetimes as a function of trigger voltage Vpump. The black sphere represents the transition lifetime τ with error bar obtained through single exponential fitting in the function of Iprobe and delay time Δt from Fig. 2e and Supplementary Fig 13, along with the fitted functions derivative (red line). Red and blue areas indicate two distinct regimes corresponding to D0+ to S0 and D0+ to S02+/T12+. The error bars represent the uncertainty associated with the single-exponential decay fitting. b Many-body schematics of the possible excited state in N−2, N−1 and N electrons. The intrastate energy space is obtained by TDDFT calculation. The y-axis (E) represents the absolute energy relative to the vacuum level. Brown crosses indicate the excluded transitions. The red arrow indicates the excitation from N−1 to N state, and the blue arrows indicate the N−1 to N−2 transition, corresponding to the red area and blue area in (a). Single-particle illustration of N − 1 → N (c) and N − 1 → N − 2 (d) transition via co-tunneling. The orbital energies of the N − 1 state relative to the vacuum level are obtained from our calculations. With the gate tuned near the N − 1/N degeneracy point, the Fermi level (EF ≈ − 7 eV) is approximately aligned with the SUMO orbital (−6.85 eV). Under negative bias in a nearly symmetric junction, the drain chemical potential shifts upward while the source shifts downward. At low Vpump, resonance between the drain and the α-SUMO enables electron injection, forming S0 (c). At higher Vpump, deeper occupied orbitals become accessible for electron extraction (d): removal from the β-orbital is associated with the S02+ state, while removal from the α-orbital corresponds to the T12+state. To understand the origin of the coherent transition at Vpump around −0.8 V, we calculate the energy distribution of excited states in N (BP), N−1 (BP+), and N−2 (BP2+) states using the time-dependent density functional theory (TDDFT) and map them onto a many-body picture in Fig. 3b (details in Supplementary Fig. 17)40. Since cotunneling processes in the transport allow the access to excited states in single-molecule transistor when the applied voltage matches the transition energy between states36,41,42, correlating the energy difference between ground state and excited states with experimental observations is essential. We first rule out the involvement of the excited states in N state due to the 1.65 eV energy difference (>0.8 eV) between the first excited state T1 and the ground state S0. Although the direct excitation of an anion has been observed by light-STM experiments43 and the energy difference between D1+ and D0+ states of 0.82 eV is in principle accessible, the doublet excited state D1+ is unlikely to contribute due to its short lifetime of ~4.7 ns estimated by the Einstein equation (details in Methods, Computational details). The observed broadened peak around −0.8 V originates from the T12+ and S02+ states, indicating the coherent dynamics of N−2 states44. The incoherent transition at Vpump > −0.4 V corresponds to the electron injection to the singly unoccupied molecular orbital (α-SUMO) illustrated in Fig. 3c, enabling the transition from D0+ to S0. Before reaching the energy level of singly occupied molecular orbital (α-SOMO and β-SOMO), the dynamics exhibit a flat trend caused by S0. At Vpump around −0.6 V, the S02+ state gradually increases to be the dominant state, resulting from the removal of one electron from SOMO. As Vpump < −0.8 V, reaching the energy level of α-SOMO, the T12+ state replaces S02+ to be the dominant state. The shift in the branching ratio of transitions into S0 and S02+/ T12+ leads to the observed increase in τ with pump voltage. Based on the TDDFT and experimental results, we propose the possible schematic energy alignment of both states relative to the Fermi level in Supplementary Fig. 18, which showcases that the orbitals associated with T12+ are energetically closer to the Fermi level compared to those of S02+. This closer proximity implies a higher tunneling probability for T12,45, which accounts for Gaussian-shaped Iprobe regime around Vpump = ±0.8 V in Fig. 2f. Such an observation demonstrates that the selective excitation of excited states of a single molecule radical can be achieved by modulating the amplitude of the pump voltage. Disentangling the excited-state pathways To elucidate the role of excited states during discharging dynamics with following resonant excitation, we acquire probe-voltage-dependent dynamics with the pump voltage fixed at −0.8 V as presented in Fig. 4a (Details in Methods, Transient I–V measurement). A time-dependent Coulomb blockade region emerges over a 0−500 ns timeframe, with features distinct from those observed in the pump-voltage-dependent dynamics. The apparent motion of the conductance boundary with delay time indicates a time-dependent change in the transport resonance condition. We do not interpret this observation as a direct proof that isolated molecular eigenstates rigidly shift in energy. Rather, the data reflect the delay-dependent alignment between molecular many-body states and the electrode chemical potentials, which can be influenced by Coulomb renormalization, charge redistribution, local electric fields and state-dependent tunneling rates. Importantly, a blank graphene tunnel junctions without molecules show no comparable response to either pump or probe voltage pulses (Supplementary Fig. 22), indicating that the observed dynamics originate from the molecular junction rather than from trap states in graphene46.Fig. 4Transition pathways and Coulomb interaction among excited states.a Corresponding dependence of the probe current Iprobe on the delay time as a function of probe voltage Vprobe after a fixed pump voltage Vpump = −0.8 V at a continuous VG = −2.5 V. b Transient differential conductance spectra derived from (a). Each differential conductance curve is smoothed with a Savitzky–Golay filter with a window length of 11 for better recognition. c Selected differential conductance derived from (b) before (black) and after (blue) smoothing at Δt = 0 ns (lower panel) and 500 ns (upper panel). d Schematic of transition pathways of peaks 1–3. Peak 1 appears at long delay times, where the population of the D0+ state is higher, and is therefore attributed to a re-excitation process involving transitions from D0+ to S02+/T12+. Peak 2 is associated with the T12+ state and corresponds to the transition to Q0+. Peak 3 corresponds to the transition from S02+ to the Dn+ state. e Peak energy extracted from (b). The lines are guides to the eye. f Illustration of Coulomb interactions among excited states, showing possible global and differential energy shifts during charge dissipation. The y-axis (E) represents the relative energy with respect to the Fermi level of the electrode. Transition pathways and Coulomb interaction among excited states. a Corresponding dependence of the probe current Iprobe on the delay time as a function of probe voltage Vprobe after a fixed pump voltage Vpump = −0.8 V at a continuous VG = −2.5 V. b Transient differential conductance spectra derived from (a). Each differential conductance curve is smoothed with a Savitzky–Golay filter with a window length of 11 for better recognition. c Selected differential conductance derived from (b) before (black) and after (blue) smoothing at Δt = 0 ns (lower panel) and 500 ns (upper panel). d Schematic of transition pathways of peaks 1–3. Peak 1 appears at long delay times, where the population of the D0+ state is higher, and is therefore attributed to a re-excitation process involving transitions from D0+ to S02+/T12+. Peak 2 is associated with the T12+ state and corresponds to the transition to Q0+. Peak 3 corresponds to the transition from S02+ to the Dn+ state. e Peak energy extracted from (b). The lines are guides to the eye. f Illustration of Coulomb interactions among excited states, showing possible global and differential energy shifts during charge dissipation. The y-axis (E) represents the relative energy with respect to the Fermi level of the electrode. To reveal the emerging states during the transition from the N−1 to the N−2 charge state, we derive nanosecond-resolved differential conductance spectra from Fig. 4a. This analysis allows us to resolve delay-dependent transient conductance resonances that are associated with different charged excited-state configurations, as shown in Fig. 4b. Three resonant peaks are resolved and labeled as Peaks 1–3. Their central energies are determined from the probe voltage, which directly reflects the energy of each state via the cotunneling mechanism with a lever arm close to unity36. Among them, Peak 1 (yellow dotted line) shows an energy downshift at its early period and becomes stabilized around −0.65 V since 200 ns, which matches the resonant peak indicated by the arrow in Fig. 2c. This suggests that Peak 1 originates from the re-excited electrons from D0+, analogous to ground-state bleaching in optical spectroscopy. Since the differential conductance associated with a given state is proportional to its density of state, the value of Gtransient presented in Supplementary Fig. 24 could indicate the population of electrically stimulated transient species. The population of D0+, as indicated by the intensity of Peak 1, exhibits similar increasing features to the above results, confirming that Peak 1 reflects charge injection process through D0+ ground-state bleaching22. In contrast to the single-pulse differential conductance spectra (Fig. 2c), Peak 2 (~0.2 eV at 0 ns) and Peak 3 (~0.5 eV at 0 ns) appear exclusively after the pump pulse. Although a spectral feature exists in this energy range in Fig. 2c, the transient signal manifests as a distinct peak rather than the broad shoulder observed in the steady state spectra. Moreover, Peak 2 and Peak 3 decay with delay time (Fig. 4c), confirming that Peaks 2 and 3 represent transient species generated by excited states following the pump pulse. Based on our above observation and TDDFT calculations, we attribute the Peaks 2 and 3 to T12+ and S02+. The energy alignment between electrode and single molecule is shown in Supplementary Fig. 18. Owing to the small energy offset of the T12+ β orbital (SUMO + 2), which lies 0.21 eV below the Fermi level, electron injection can occur and form Q0+ at Vprobe = 0.2 V (Fig. 4d). In contrast, the LUMO + 1 of the S02+ state is located 0.48 eV below the Fermi level, and thus Peak 3 is attributed to the S02+ contribution, corresponding to the transition from S02+ to the Dn+ state as summarized in Fig. 4d. The electronic pump-probe technique can, in principle, average discrete quantum transition events into smoothed Iprobe–Vprobe responses through repetitive excitation cycles. However, the observed dynamics suggest a more complex origin. We extract the peak energies of them at every delay time presented in Fig. 4e. Peak 1 (D0+) and Peak 3 (S02+) evolve in a correlated manner, suggesting that part of the spectral motion reflects a global change in the electrostatic alignment of the molecular junction during charge relaxation. However, these apparent resonance shifts should not be interpreted as direct measurements of rigid molecular-state energy shifts. In a pulsed transport experiment, the observed peak position reflects the voltage required to satisfy a transient tunneling condition and can therefore be affected by state population, tunneling rate, local electrostatic potential, charge redistribution and Coulomb renormalization47. Within this framework, Peaks 2 (T12+) and 3 (S02+) show different delay-dependent resonance trajectories, with apparent shifts of ~440 meV and ~220 meV within ~150 ns, respectively. Given the complexity of the junction environment, multiple factors may potentially influence the observed spectral dynamics. We therefore first assess whether structural changes induced by charge injection or electric fields could account for the observations through DFT analysis (Supplementary Fig. 19). The calculated results show that geometric modifications lead to only minor orbital-energy perturbations (<0.1 eV; Supplementary Fig. 19), which are much smaller than the apparent resonance-position changes observed experimentally. We therefore consider structural relaxation unlikely to be the dominant origin of the transient spectral evolution. The strong external electric field in single-molecule junction (~108 V/m) during the pump pulse could, in principle, alter the transient transport resonance conditions through coupling to molecular dipoles and local charge redistribution24,48,49. The calculated dipole moment (Supplementary Table 2) suggests that the coupling strength of D0+ and S02+ are nearly the same, whereas that of T12+ is significantly smaller. This difference may contribute to the distinct initial resonance positions of Peaks 2 and 3, although we do not assign the observed resonance evolution solely to a Stark shift of molecular eigenstates. Instead, the convergence of Peaks 2 and 3 in Fig. 4e, which approaches a similar probe voltage after ~200 ns, suggests that these transient conductance features are associated with the relaxation of charged configurations during the N − 2 → N − 1 recovery process. Thus, we propose a physically plausible interpretation summarized in Fig. 4f. The common component of the resonance-position evolution may be associated with a time-dependent electrostatic realignment of the molecular junction during charge dissipation, including possible contributions from Coulomb renormalization and local charge redistribution. Meanwhile, the time-dependent change in the energy separation between the Peak 2 (T12+) and Peak 3 (S02+) states suggests that state-specific relaxation and tunneling pathways also contribute to the transient conductance spectra. In particular, the initially observed ~220 meV gap may reflect a differential stabilization of the T12+ state relative to S02+, potentially arising from differences in their effective dipole responses to the local electric field. Such behavior may be more readily observable in single-molecule radical junctions where charge transport occurs along the molecular backbone, which can enhance the coupling between transient charges and molecular orbitals. For comparison, STM-based studies have reported that typically orthogonal alignment between the electric field and molecular dipole can lead to only microelectronvolt-scale energy shifts50. Overall, our results provide experimental indications of complex excited-state relaxation dynamics under electrical excitation in single-molecule transistors, offering a framework for further investigation of nonequilibrium many-body effects in molecular charge transport. Discussion In summary, we report experimental signatures consistent with the presence of multiple excited states in a single-molecule transistor. The nanosecond evolution of transient resonance positions suggests that Coulomb interactions, charge redistribution and local electrostatic fields strongly influence the transport resonances associated with excited-state configurations in single-molecule junctions. The interplay between strong local electric fields and molecular dipole moments under pump bias may contribute to the energy splitting observed during the relaxation dynamics of S02+ and T12+ states. These results provide insights into transient charged intermediates that can influence pathways in chemical reactions, photocatalysis, and electrochemical processes, offering a perspective on charge-transfer dynamics at the single-molecule level. More broadly, our approach presents a potential experimental route for probing nonequilibrium states in charge transport through molecular junctions. A comprehensive microscopic understanding of the underlying mechanisms will require further theoretical development and close collaboration with the theoretical community to accurately capture the transient electronic structure of these systems51,52. Methods Experimental setup The fabrication of single-molecule devices is described in detail in Supplementary Note 2. Experiments are carried out at a low temperature under vacuum conditions (<10−6 mbar) in a commercially available cryogenic probe station (Lakeshore Cryogenics, model CRX-6.5 K) at a setting temperature of 30 K. The gate-voltage dependent differential conductance dISD/dVSD (GSD) displayed in Fig. 2b is measured using lock-in detection by applying a constant voltage with 10 mVrms modulation at 873.1 Hz to the single-molecule device. For voltage pulse-related measurements, pulse current is induced by a series of short voltage pulses rather than by continuous DC bias or modulated AC bias. A significant number of parameters can vary for the measurement of transient species. Pulse amplitude (Vpump and Vprobe), pulse width (τpulse), and delay time between pump/probe pulse pairs (Δt) can be adjusted depending on the nature of the measurement. The role of the four measurement stages: Gate, Pump, Relaxation (Delay), and Probe is shown in Supplementary Fig. 25. Single-pulse measurement The voltage pulses with various amplitudes are produced at a repetition rate of 873.1 Hz by an arbitrary waveform generator function (AWG, Zurich Instruments, UHF-LI) and fed to the source electrode using high-frequency cables. In the measurements, the widths of the pulses are 25 ns along with 25 ns rising and falling edges. The pulsed current Ipump is collected using a combination of a low-noise preamplifier (Femto, LCA-20K-200M) and a lock-in amplifier (Zurich Instruments, UHF-LI). The pulsed-voltage-related experiments are performed within the range from −1 to 1 V. The raw data of pulsed current-voltage characterization is shown in Supplementary Fig. 11. Pump-probe measurement In the measurements of the all-electronic pump-probe method, both pulses use the same waveform as in pulsed I–V characterization. Firstly, a variable pump pulse Vpump is sent to single-molecule junction triggering the charge injection and the current of non-equilibrium states is probed by a latter pulse with 1.0 V amplitude, with the same polarity as the pump pulse. The purpose of using high amplitude probe pulse is to maximize the signal-to-noise ratio, although it may have the possibility of secondary excitation. According to the time-dependent Coulomb blockade phenomenon in Fig. 4a, we notice that the small probe amplitude is in the blocking window, making it difficult to capture the non-equilibrium dynamics. Corresponding dependence of the probe current Iprobe on the delay time as a function of trigger voltage Vpump is obtained by subtracting the signal of the single pump pulse from the total signal of pulse pairs20. An obvious low-conducted regime is found at VG = −2.5 V (Supplementary Fig. 13a–c), while no dynamic phenomena can be observed under N state (VG = 0 V, Supplementary Fig. 13d–f). Since the probe current will not return to baseline at 500 ns at high trigger voltage, we speculate that the molecule has not returned to the ground state at this time or the molecular configuration has changed. A comparison of forward and reverse sweeping pump-probe experiments with a long enough delay time is performed to rule out the influence of molecular configuration in Supplementary Fig. 16. And the dynamical curves indicated the nature of the non-equilibrium states of the single-molecule junctions. Transient I–V measurement For the measurement of transient electronic spectroscopy Iprobe–Vprobe and Gtransient (dIprobe/dVprobe), we fix the pump voltage of −0.8 V to trigger the redox dynamics and detect its transient response using various amplitudes of probe pulse Vprobe. Extracting probe current Iprobe will be more complicated since it is difficult to extract the contribution of the latter pulse separately from recorded total net current Itotal. The raw data of Itotal and the extraction method of Iprobe are shown in Supplementary Fig. 21. After considering phase inversion and deducting the current contributed by the pump pulse at Vprobe = 0 V, and we can obtain the transient I–V response after molecules being excited. In order to compare the reliability of the smoothed differential conductance curve in Fig. 4b, we compare several pieces of data before and after smoothing in Supplementary Fig. 23. Also, to verify the reliability of pump-probe experiments, we carry out the same characterizations using graphene nanogap junction with the phenomenon after depositing molecules shown in Supplementary Fig. 22. Among them, there are no obvious coulomb blockage regions in the differential conductance diagram and no dynamic phenomena could be observed. Computational details Geometric relaxations, energy level, charge distribution and dipole moment under electric field of the ground and positively charged states of this BP molecule are calculated by using density functional theory (DFT) within the hybrid B3LYP functional53 in conjunction with the 6-31 G(d,p) basis set as implemented in the Gaussian 1654. The underlying DFT calculation of the positively charged molecule BP+ and BP2+ are performed using the Kohn-Sham scheme with spin-polarized unrestricted open-shell ansatz. In order to describe the excited-state properties of the molecule, the gas-phase time-dependent density-functional-theory (TDDFT) calculations are also performed by Gaussian 1654. The energy difference between excited state and ground state, spin eigenvalue 〈S2〉 are obtained in this process. States with 〈S2〉 = 2.000 and 0.000 are configured as triplet state and singlet states, respectively. While 〈S2〉 of the excited state of BP+, are distributed between 0.765 and 0.790, which is close to 0.75 as expected for a spin doublet state. Finally, the lifetimes τ of doublet states of BP+ are estimated by the Einstein equation τ = 3/(2fv 2)55, where f is the oscillator strength, v is the excited energy (cm-1). Supplementary information Supplementary Information Transparent Peer Review file Supplementary Information Transparent Peer Review file Source data Source Data Source Data