Ultrafast Pump-Probe Photocurrent Spectroscopy

• Ultrafast pump-probe photocurrent spectroscopy is a time-resolved method that monitors current responses to capture nonequilibrium electronic dynamics directly in functioning devices.
• It employs diverse experimental setups—ranging from graphene junctions to polymer solar cells—to resolve processes like hot-carrier cooling, THz transients, and exciton dissociation on femtosecond-to-picosecond scales.
• The technique differentiates multiple current-generation pathways, including displacement, thermoelectric, and exciton-mediated currents, providing actionable insights into charge transport and interfacial behavior.

Ultrafast pump-probe photocurrent spectroscopy is a family of time-resolved spectroscopies in which an ultrashort pump pulse creates a nonequilibrium electronic state and a delayed probe pulse reads out the resulting optoelectronic response through a current observable rather than through transmission or reflection alone. Depending on the implementation, the measured signal can be a probe-induced photocurrent, a sampling current that gates a propagating electrical transient, or a nonlinear photocurrent generated by a multi-pulse sequence. Across graphene junctions, GaAs photoswitches, single InAs nanowires, nanoscale field-effect devices, polymer solar cells, and CNT/$\mathrm{C}_{60}$ photovoltaics, these methods have been used to resolve hot-carrier cooling, displacement and transport currents, exciton dissociation, interfacial charge generation, and state-to-state couplings on femtosecond-to-picosecond timescales (Sun et al., 2011, Prechtel et al., 2010, Erhard et al., 2015, Vella et al., 2015, Faitz et al., 26 Jun 2025).

1. Core measurement concept

In its conventional form, ultrafast pump-probe photocurrent spectroscopy measures how a pump-induced nonequilibrium population modifies a later probe-generated current as a function of pump-probe delay. In the graphene pn-junction implementation, the pump and probe are focused onto the same spot, the probe beam is mechanically chopped at about $1.7$ kHz, and the lock-in amplifier detects the probe-induced photocurrent in the presence of the pump pulse; the primary observable is the probe-induced photocurrent versus pump-probe delay time, and the photocurrent response time $\tau$ is defined as the half-width at half-maximum of the dip centered at zero delay (Sun et al., 2011). In on-chip sampling architectures for GaAs photoswitches and InAs nanowires, the pump launches an electromagnetic or current transient into a coplanar stripline and a delayed probe pulse gates a field probe, so that the measured quantity is the sampling current $I_{\text{Sampling}}(\Delta t)$ as a function of optical delay (Prechtel et al., 2010, Erhard et al., 2015).

A second class of experiments uses spatial separation between the pump and probe. In femtosecond scanning photocurrent microscopy, the pump is fixed near one contact, the probe is scanned elsewhere in the device, and the delay-dependent change in probe photocurrent,

$$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$

acts as a reporter of carrier migration between different points in an operating nanowire or nanotube device (Son et al., 2014). Here the delay of the remote suppression maximum is interpreted as the most probable arrival time of the transported carrier packet.

A multidimensional extension replaces the single pump-probe pair with a phase-controlled four-pulse sequence. In two-dimensional coherent photocurrent excitation spectroscopy, four collinear femtosecond pulses with independently controlled delays $t_{21}$, $t_{32}$, and $t_{43}$ generate a nonlinear photocurrent that is demodulated at the rephasing and non-rephasing frequencies

$$\Omega_{43}-\Omega_{21} \qquad \text{and} \qquad \Omega_{43}+\Omega_{21},$$

respectively (Vella et al., 2015). This converts photocurrent detection from a one-dimensional transient into a two-dimensional frequency-frequency correlation measurement.

2. Experimental architectures and signal channels

The literature represented here spans several distinct instrumental realizations, all of which retain the same basic pump-probe logic while differing in how the probe converts the nonequilibrium state into a current observable.

Platform	Measured current observable	Distinct capability
Graphene pn junction	Probe-induced photocurrent	Intrinsic junction response time and gate-dependent mechanism separation
GaAs photoswitch / CPS	Sampling current $I_{\text{Sampling}}(\Delta t)$	Separation of displacement and transport current pulses
Single InAs nanowire / CPS	Sampling current $1.7$0	Simultaneous access to photo-thermoelectric current, THz transient, and hole transport
fs-SPCM in nanowires / SWNTs	$1.7$1	Spatiotemporally resolved carrier transport imaging
Polymer solar cell 2D-PCS	Nonlinear device photocurrent	Rephasing/non-rephasing spectra and off-diagonal cross-peaks
CNT/$1.7$2 device	Photocurrent-detected pump-probe and 2D spectra	Same-position comparison of photoabsorption and photocurrent observables

In graphene, the experiment combines scanning photocurrent microscopy with femtosecond pump-probe timing. Laser pulses at $1.7$3 nm from a Coherent MIRA oscillator with $1.7$4 MHz repetition rate are focused to about $1.7$5, with a pulse width at the sample of about $1.7$6 (Sun et al., 2011). In the GaAs and InAs stripline implementations, a Ti:sapphire pump pulse excites the photoswitch or nanowire locally, while a delayed probe pulse temporarily short-circuits a remote field probe and samples the propagating electrical transient; this is effectively a photoconductive sampling gate (Prechtel et al., 2010, Erhard et al., 2015).

In polymer solar cells, the nonlinear photocurrent protocol is fully collinear and uses acousto-optic phase modulation rather than noncollinear phase matching. Each pulse acquires a distinct modulation frequency $1.7$7 near $1.7$8 MHz, pair-difference frequencies are on the order of $1.7$9 kHz, and dual lock-in detection isolates the nonlinear photocurrent pathways (Vella et al., 2015). In CNT/$\tau$0 photovoltaics, the four-pulse spectrometer measures both optical and electrical responses on the same device and suppresses incoherent photocurrent background through the polarization sequence

$\tau$1

combined with phase cycling. The reported pump pulse duration is $\tau$2 fs, the probe pulse duration is $\tau$3 fs, and the instrument response is $\tau$4 fs (Faitz et al., 26 Jun 2025).

These architectures imply an important methodological distinction. Some experiments read out the functional optoelectronic response directly from a working device, while others read out a current-like electrical transient launched into a circuit. Both belong to pump-probe photocurrent spectroscopy in the broad sense used in the cited literature, but they weight microscopic processes differently: device photocurrent emphasizes charge generation and collection, whereas stripline sampling is sensitive to field redistribution, transient transport, and THz-coupled electrodynamics.

3. Dynamical regimes and mechanisms resolved by photocurrent readout

A central contribution of ultrafast pump-probe photocurrent spectroscopy is that it separates current-generation mechanisms that are merged in steady-state measurements. In graphene pn junctions, the pump-probe dip width shows that the photocurrent response time increases from about $\tau$5 at $\tau$6 to about $\tau$7 at $\tau$8, implying a fundamental bandwidth near $\tau$9 and supporting the paper’s statement of $I_{\text{Sampling}}(\Delta t)$0 device operation speed (Sun et al., 2011). Gate-dependent measurements further show that both photo-thermoelectric and built-in electric field effects contribute, with suppression, polarity reversal, or enhancement at zero delay depending on junction configuration. The weak temperature dependence of pulsed photocurrent, contrasted with a factor of $I_{\text{Sampling}}(\Delta t)$1 increase in CW photocurrent on cooling, is interpreted as evidence that hot carriers rather than lattice phonons dominate energy transport at high frequencies (Sun et al., 2011).

In GaAs photoswitches, the time-domain signal contains two distinct components: a prompt displacement current pulse and a delayed transport current pulse (Prechtel et al., 2010). The inter-peak delay $I_{\text{Sampling}}(\Delta t)$2 varies linearly with excitation position, enabling a time-of-flight analysis that yields propagation velocities of $I_{\text{Sampling}}(\Delta t)$3 at $I_{\text{Sampling}}(\Delta t)$4 and $I_{\text{Sampling}}(\Delta t)$5 at $I_{\text{Sampling}}(\Delta t)$6. Because these velocities exceed typical Fermi and single-particle quantum velocities, the delayed transport signal is interpreted as a collective electron-hole plasma excitation rather than ordinary single-carrier drift (Prechtel et al., 2010).

Single InAs nanowires exhibit an even richer decomposition. The measured transient is fitted as

$I_{\text{Sampling}}(\Delta t)$7

and the three components are assigned to a prompt photo-thermoelectric current, a THz-emission-related transient interpreted mainly in terms of the photo-Dember effect, and a delayed broad peak attributed to transport of photogenerated holes to the contacts (Erhard et al., 2015). The contact-related photo-thermoelectric contribution has a characteristic timescale $I_{\text{Sampling}}(\Delta t)$8, a localized dielectric displacement-current signature at one contact has $I_{\text{Sampling}}(\Delta t)$9, and time-of-flight analysis of the delayed hole packet gives

$$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$0

The same experiment therefore accesses prompt thermal/contact physics, radiative THz electrodynamics, and slower charge transport in one waveform (Erhard et al., 2015).

Spatially separated fs-SPCM extends these ideas to micrometer-scale transport imaging. In a $$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$1-$$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$2m-channel Si nanowire with effective photocurrent-spot separation $$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$3, the remote peak transit time is $$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$4, corresponding to

$$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$5

The gate dependence is explicit: more negative $$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$6 strengthens the Schottky-barrier field, and the authors relate the initial drift velocity to

$$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$7

In SWNT devices the same method yields $$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$8 for a $$\Delta I_{pr}(t_{\mathrm{delay}})= I_{pr,\mathrm{with~pump}}- I_{pr,\mathrm{without~pump}},$$9 transit, corresponding to $t_{21}$0, more than $t_{21}$1 faster than in Si nanowires of similar length (Son et al., 2014).

4. Excitons, interfaces, and current-generating pathways

In excitonic and organic systems, ultrafast pump-probe photocurrent spectroscopy is especially valuable because not every optically excited state generates free charge. The polymer-solar-cell study on a PCDTBT:PC$t_{21}$2BM bulk heterojunction frames the problem around sub-$t_{21}$3-fs conversion of tightly bound intrachain excitons, with reported binding energy $t_{21}$4 eV, into charge-transfer or delocalized charge-separated states (Vella et al., 2015). The 2D photocurrent spectra measured at $t_{21}$5 fs show an off-diagonal structure approximately $t_{21}$6 meV above the diagonal signal. In 2D spectroscopy language, this is the key cross-peak: it indicates that optical excitation at one transition energy is correlated with photocurrent generation through a different electronic manifold or pathway. The authors interpret it as evidence for coupling between the polymer exciton and photocurrent-producing delocalized states, consistent with noise-induced resonant charge tunneling between intra-chain excitons and delocalized CT states (Vella et al., 2015).

The CNT/$t_{21}$7 photovoltaic study makes the same point in a different way by directly comparing photoabsorption-detected and photocurrent-detected ultrafast spectra on the same device (Faitz et al., 26 Jun 2025). Its central quantitative result is that 89% of photocurrent comes from a component with $t_{21}$8 fs timescale, while 11% of photocurrent comes from a $t_{21}$9 fs component. The $t_{32}$0 fs channel is assigned to excitons created adjacent to or very near $t_{32}$1, which dissociate almost immediately; the $t_{32}$2 fs channel is assigned to excitons that must diffuse and/or transfer between nanotubes before reaching a dissociating interface. By contrast, absorption observables show exciton diffusion and transfer for several picoseconds, even though transferred excitons no longer contribute to photocurrent after about $t_{32}$3 ps (Faitz et al., 26 Jun 2025).

Photocurrent-detected 2D spectra in mixed $t_{32}$4 CNT devices add spectral selectivity to this kinetic picture. At short waiting times, asymmetric lower and upper cross-peaks are interpreted as the combined effect of downhill exciton transfer $t_{32}$5 and bidirectional hole transfer; after $t_{32}$6 ps the cross-peaks become symmetric, indicating that only hole transfer remains photocurrent-relevant. Fitting the ratio of lower to upper cross-peak amplitudes gives $t_{32}$7 fs and $t_{32}$8 fs, which are assigned to exciton transfer contributing to photocurrent (Faitz et al., 26 Jun 2025).

These results support a general methodological lesson already implicit in the polymer-solar-cell study: photocurrent detection selects the subset of excitations that actually produce charge. This suggests that long-lived excitonic signals in purely optical pump-probe measurements may overestimate the importance of diffusion-mediated pathways for device efficiency when most current is generated in a prompt interfacial dissociation channel (Vella et al., 2015, Faitz et al., 26 Jun 2025).

5. Interpretation, coherence, and common pitfalls

A recurrent interpretive problem is that pump-probe current transients are not always simple population measurements. In resonant monolayer MoSe$t_{32}$9, optical pump-probe spectroscopy shows that coherent polarization, local-field-induced renormalization, excitation-induced dephasing, and inter-valley scattering all contribute and mix strongly; the paper explicitly argues that the usual separation between population dynamics at positive delay and coherence at negative delay fails in this regime (Rodek et al., 2021). The minimal three-level model includes local field effect $t_{43}$0, excitation-induced dephasing $t_{43}$1, and inter-valley scattering $t_{43}$2, and the observed negative-delay oscillations are identified as perturbed free induction decay. Although the observable in that work is optical rather than electrical, the implication for resonant photocurrent spectroscopy is direct: transient line shifts, broadening, and sign-changing spectral shapes need not be attributed to population transfer alone (Rodek et al., 2021).

A second caution concerns delay-dependent oscillations. The theory of transient interference in band and Mott insulators shows that oscillatory pump-probe structure can arise from interference mediated by an electronic excitation continuum, with oscillation frequency set by

$t_{43}$3

In this picture the pump imprints spectral information into a continuum that acts as a memory medium, and the later probe reads it out, producing probe-energy-dependent oscillations in delay traces (Shinjo et al., 2018). For photocurrent spectroscopy, this means that oscillations in current versus delay are not automatically evidence for a collective mode, a transport resonance, or coherent population transfer.

A third misconception is that delayed current peaks always imply single-particle motion. The GaAs photoswitch study explicitly rejects several alternative explanations for its second peak, including THz reflections and stripline dispersion, and then argues that the inferred propagation velocity is too high for ordinary carrier drift, favoring a collective electron-hole plasma interpretation (Prechtel et al., 2010). Likewise, the CNT/$t_{43}$4 study emphasizes that direct photocurrent-detected nonlinear spectroscopy is usually obscured by a large incoherent background from exciton-exciton and/or exciton-charge interactions after the probe pulse; the reported four-pulse polarization sequence is introduced precisely to suppress that incoherent mixing background in isotropically ordered samples (Faitz et al., 26 Jun 2025).

Taken together, these studies establish an interpretive hierarchy. A pump-probe photocurrent trace may contain population dynamics, coherent phase-sensitive structure, saturation, displacement current, thermoelectric current, plasma propagation, and artifact-like incoherent mixing. Objective analysis therefore requires attention to geometry, polarization, modulation/demodulation scheme, spectral selectivity, and control experiments such as bias dependence, gate dependence, and waiting-time dependence (Prechtel et al., 2010, Rodek et al., 2021, Shinjo et al., 2018, Faitz et al., 26 Jun 2025).

6. Capabilities, limitations, and emerging directions

The field now spans from device-scale photocurrent readout to atomic-scale tunneling-current detection. A recent THz-STM study on photoexcited GaAs(110) explicitly formulates its method as a pump-probe photocurrent-type measurement in which an $t_{43}$5 nm optical pump creates carriers and a delayed THz near-field probe modulates the STM junction, producing a rectified tunneling charge/current $t_{43}$6 that acts as a localized photocurrent (Jelic et al., 28 Apr 2026). After deconvolving the oscillatory sub-cycle THz probe field using in situ THz time-domain spectroscopy, the extracted intrinsic photocurrent lifetimes are $t_{43}$7 fs on a defect, $t_{43}$8 fs on the surrounding bright ring, and $t_{43}$9 fs in a pristine region. This extends pump-probe photocurrent spectroscopy into the regime of atomic spatial resolution and local band-bending dynamics (Jelic et al., 28 Apr 2026).

At the same time, related optical methods continue to inform photocurrent experiment design. Two-color side-view pump-probe interferometry in glass, while not measuring current directly, resolves a fast electron-removal process with a $$\Omega_{43}-\Omega_{21} \qquad \text{and} \qquad \Omega_{43}+\Omega_{21},$$0 fs time constant and argues that multi-wavelength, multi-observable readout helps distinguish free-carrier plasma, trapping, and later heating or scattering (Hayasaki et al., 2017). A plausible implication is that combining photocurrent with simultaneous optical observables can separate mobile-carrier loss from mere persistence of optically active excited states. The CNT/$$\Omega_{43}-\Omega_{21} \qquad \text{and} \qquad \Omega_{43}+\Omega_{21},$$1 platform already implements such simultaneous optical and electrical detection on the same working device (Faitz et al., 26 Jun 2025).

First-principles theory is also becoming more directly useful to the interpretation of pump-probe photocurrent measurements. A recent framework for pump-probe spectroscopy in solids combines real-time TDDFT, constrained DFT, and a nonequilibrium BSE to show that photoinduced Coulomb screening is the primary electronic effect, that Pauli blocking plays a minor role, and that thermal lattice expansion red-shifts the spectra in WSe$$\Omega_{43}-\Omega_{21} \qquad \text{and} \qquad \Omega_{43}+\Omega_{21},$$2, CsPbBr$$\Omega_{43}-\Omega_{21} \qquad \text{and} \qquad \Omega_{43}+\Omega_{21},$$3, and TiO$$\Omega_{43}-\Omega_{21} \qquad \text{and} \qquad \Omega_{43}+\Omega_{21},$$4 (Qiao et al., 9 Sep 2025). Because the computed observable is transient absorption rather than current, this is not yet a direct photocurrent theory. Nevertheless, it provides the microscopic ingredients—nonequilibrium carrier populations, exciton resonance shifts, screening changes, and thermal contributions—that frequently determine spectrally resolved photocurrent response (Qiao et al., 9 Sep 2025).

The main limitations remain method-specific. Some device studies are preliminary and do not yet provide complete waiting-time series or microscopic lineshape decomposition (Vella et al., 2015). Some on-chip sampling experiments are bandwidth-limited by the stripline and Auston-switch circuitry (Erhard et al., 2015). Spatially resolved methods are often contact-dominated because probe photocurrent is largest near Schottky barriers or defects (Son et al., 2014). And several studies show that purely optical observables can overweight long-lived or nonproductive excitations relative to current-generating channels (Faitz et al., 26 Jun 2025).

The most durable conclusion is therefore methodological rather than platform-specific: ultrafast pump-probe photocurrent spectroscopy is distinguished not simply by femtosecond timing, but by its ability to project ultrafast electronic dynamics onto the electrically functional channels of a device. In simple implementations this means measuring intrinsic response time or transport. In multidimensional implementations it means identifying off-diagonal couplings and transfer pathways that actually feed current. In emerging local and nonlinear variants it means resolving how defects, interfaces, and symmetry-broken states reshape current generation on their native spatiotemporal scales (Sun et al., 2011, Vella et al., 2015, Faitz et al., 26 Jun 2025, Jelic et al., 28 Apr 2026).