Multimodal STM Pump-Probe Technique

• The paper demonstrates that the technique uses a controlled delay between pump and probe pulses to extract transient tunneling and force signals with a resolution of around 70–80 ps.
• The paper details the integration of optical and electrical timing architectures, enabling nanoscale spatial localization while suppressing thermal artifacts via delay-time modulation.
• The paper shows applications across GaAs carrier dynamics, layered WSe2 force detection, and ESR-assisted spin-relaxation measurements, highlighting versatile, multimodal analysis.

Searching arXiv for recent and foundational papers on multimodal STM-based pump–probe techniques. arxiv_search(query="multimodal STM pump-probe optical pump-probe scanning tunneling microscopy AFM ESR waveform sequencing", max_results=10) Multimodal STM-based pump-probe technique denotes a family of time-resolved scanning probe methods in which a tip-sample junction is driven out of equilibrium by a pump excitation and interrogated after a controlled delay by a probe, while a common timing and delivery architecture is reused across several readout channels. In its optical pump-probe scanning tunneling microscopy (OPP-STM) form, short optical pulses excite carriers beneath the STM tip, and the subsequent relaxation is reconstructed from the delay dependence of the time-averaged tunneling current measured by a conventional low-bandwidth preamplifier. Recent implementations combine nanometer spatial resolution, tens-of-picosecond time resolution, stable long-term illumination, and electrical delay control, and the same pump-probe logic has been extended to atomic force microscopy, all-electrical STM pump-probe spectroscopy, and ESR-STM platforms (Iwaya et al., 2023, Iwaya et al., 2024, Weerdenburg et al., 2020).

1. Physical principle and observables

In STM, the tunneling current $I_\mathrm{t}$ between a sharp metallic tip and a sample is exponentially sensitive to the tip-sample separation and linearly sensitive, to first order, to the local density of states and the effective barrier set by band bending. OPP-STM exploits this locality under nonequilibrium conditions. A pump pulse optically excites carriers in the sample under the tip, transiently changing the band bending, local potential, and occupation of electronic states. A probe pulse then arrives after a controlled delay $t_\mathrm{d}$ and again excites carriers. Because some states may still be occupied or band bending may still be modified at $t_\mathrm{d}$, the probe-induced transient current differs from the pump-induced one, so the delay dependence of the tunneling current encodes carrier recombination, trapping, escape, and related relaxation channels (Iwaya et al., 2023).

The directly measured quantity is not the instantaneous current but the time-averaged current $\langle I(t_\mathrm{d}) \rangle$. A standard observable is the delay-dependent contrast

$$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$

where $t_\mathrm{max}$ is a long reference delay for which pump-induced excitation has fully decayed. When $t_\mathrm{d}$ is long, the probe sees a relaxed system and generates a large transient tunneling current. When $t_\mathrm{d}$ is short, optical absorption is partially bleached, band bending is partially screened, and the probe-induced transient is smaller. The time-averaged current therefore decreases as $t_\mathrm{d}$ becomes comparable to or shorter than the relaxation time (Iwaya et al., 2024).

The same pump-probe logic generalizes beyond tunneling-current detection. In OPP-AFM, the observable is the resonance-frequency shift of a tuning-fork-type AFM sensor in frequency-modulation mode. In all-electrical STM pump-probe and ESR-STM, the pump and probe are voltage or RF pulse sequences applied through the bias line, and the response is extracted from the spin-polarized or otherwise nonlinear tunneling current. The multimodal character therefore resides not in a single contrast mechanism but in a shared delayed-excitation formalism applied to distinct local observables.

2. Delay control, modulation, and readout architectures

A central methodological distinction within STM-based pump-probe work is between optical delay generation and all-electrical waveform synthesis. In the externally triggerable OPP-STM architecture, the pump and probe pulses are generated electrically with commercial picosecond lasers that accept external triggers. The timing electronics comprise a digital delay/pulse generator (Stanford Research DG645), a high-speed RF switch (Analog Devices HMC-C011), and a frequency divider (74HC4040). The pump laser is triggered at $f_\mathrm{rep}=1~\mathrm{MHz}$, while the probe channel alternates between two trigger trains, one delayed by $t_\mathrm{d}$0 and the other by $t_\mathrm{d}$1. A 1 MHz clock divided by 1024 produces $t_\mathrm{d}$2, which toggles the RF switch. The selected trigger sequence goes to the probe laser, so the probe delay is modulated between $t_\mathrm{d}$3 and $t_\mathrm{d}$4 at a low frequency while the pump-probe configuration is quasi-static over many pulses (Iwaya et al., 2023).

This delay-time modulation scheme is crucial because direct low-frequency modulation of optical intensity would introduce periodic heating of the tip and sample. Since tunneling current depends exponentially on distance, even sub-Å thermal expansion can dominate the electronic signal. By modulating the relative delay rather than the optical power, the average optical load remains constant: each repetition period contains one pump and one probe pulse, and only their relative timing changes. A lock-in amplifier referenced to the modulation frequency then detects the differential signal $t_\mathrm{d}$5, strongly suppressing thermal artifacts and slow drift (Iwaya et al., 2023, Iwaya et al., 2024).

The all-electrical branch uses a different timing architecture but the same differential logic. A standard implementation retrofits an existing STM with a single transfer relay in the bias line and a commercial arbitrary waveform generator with waveform-sequencing capability. The demonstrated setup uses a Keysight 33600A, a simple mechanical relay switch, and optional RF delivery through a bias-tee. Short arbitrary waveforms define pump and probe pulses; the AWG alternates between probe states A and B and supplies a sync signal to the lock-in, which measures the difference $t_\mathrm{d}$6. In this architecture, the junction bias can be written as

$t_\mathrm{d}$7

with $t_\mathrm{d}$8 controlled by phase modulation of the waveform sequence. The demonstrated implementation achieves $t_\mathrm{d}$9 time resolution, limited by the available electronics, with measured pulse edge times of $t_\mathrm{d}$0 and timing precision of $t_\mathrm{d}$1 in cable-delay diagnostics (Natterer, 2019).

A dilution-refrigerator realization extends the same waveform-sequencing concept to mK temperatures and vector magnetic field. There the pump-probe implementation uses a Keysight 33622A, a mechanical relay, a variable attenuator set to $t_\mathrm{d}$2, and the same high-frequency tip line used for ESR. Phase modulation shifts the pump relative to the probe within a programmed span, while polarity modulation of the probe at tens to hundreds of hertz provides the lock-in reference (Weerdenburg et al., 2020).

3. Optical integration and the emergence of literal multimodality

The optical OPP-STM platform that has become a prototype for multimodal operation is built around a UHV, low-temperature STM based on a UNISOKU USM1400. Reported operating conditions include a base pressure of $t_\mathrm{d}$3 at room temperature and $t_\mathrm{d}$4 when cold, with operation demonstrated at 78 K and 6 K. The tip and sample are horizontally mounted, coarse approach and lateral positioning are done with shear piezo stacks, and the head is mechanically isolated by springs and eddy-current damping, giving low vibrational noise with no peaks up to 1 kHz in the distance-noise spectrum (Iwaya et al., 2023).

The optical subsystem uses two independent, externally triggerable picosecond lasers, one for pump and one for probe. The reported implementation employs two KATANA 05 laser heads at 532 nm with pulse width $t_\mathrm{d}$5 and repetition rate 1 MHz. Each beam passes through a beam-stabilization unit, steering optics, beam splitters, and neutral-density filters. The beam stabilization system uses a position-sensitive detector and active mirrors in each beam path to compensate slow drift and mechanical perturbations. The total optical-system footprint is about $t_\mathrm{d}$6, markedly smaller than earlier optics-heavy OPP-STM systems based on long mechanical delay lines (Iwaya et al., 2023).

A particularly consequential design choice is the placement of the focusing optics. An aspheric lens inside UHV, near the tip and sample, focuses the combined pump and probe beam onto the junction at an incident angle of about $t_\mathrm{d}$7 from the sample normal. This lens is mounted on a custom three-dimensional piezoelectric actuator capable of stick-slip motion of $t_\mathrm{d}$8 in $t_\mathrm{d}$9 and $\langle I(t_\mathrm{d}) \rangle$0 and $\langle I(t_\mathrm{d}) \rangle$1 in $\langle I(t_\mathrm{d}) \rangle$2. A second aspheric lens on another three-dimensional piezo actuator serves as an in-vacuum objective for imaging; together with an external imaging lens it yields a total magnification of about $\langle I(t_\mathrm{d}) \rangle$3. Because the focusing lens is mounted on the same mechanical stage as the STM head and sample, relative motion between the sample and optical focus is minimized. Optical images before and after 16 hours show negligible drift of the tip, its mirror image, and the laser spot, and direct monitoring at 6 K under continuous illumination of 0.25 mW for 12 hours shows only about $\langle I(t_\mathrm{d}) \rangle$4 RMS variation in current with stable tip height (Iwaya et al., 2023).

The multimodal extension to AFM is not merely conceptual. The same optical system and electrically controlled timing scheme are reused for OPP-AFM, where pump-probe pulse pairs are focused onto the apex of an AFM probe and the delay is modulated between $\langle I(t_\mathrm{d}) \rangle$5 and $\langle I(t_\mathrm{d}) \rangle$6 exactly as in STM. The AFM implementation uses a tuning-fork-type sensor in frequency-modulation mode, and the lock-in extracts the frequency-shift modulation induced by the pump-probe delay sequence. The reported force signal is attributed to dipole-dipole interactions between optically induced dipoles,

$\langle I(t_\mathrm{d}) \rangle$7

at the tip and sample. This extension allows time-resolved measurements in conductive and insulating materials within one compact optical-timing infrastructure (Iwaya et al., 2024).

4. Temporal resolution, spatial locality, and nanoscale carrier-dynamics mapping

In externally triggerable OPP-STM, the temporal resolution is set by the cross-correlation of the pump and probe pulses, including pulse width and timing jitter, rather than by the STM preamplifier bandwidth. The reported theoretical estimate combines a laser pulse width of $\langle I(t_\mathrm{d}) \rangle$8, laser-trigger jitter of $\langle I(t_\mathrm{d}) \rangle$9, and electrical-trigger jitter of $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$0: $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$1 A sum-frequency-generation cross-correlation measurement gives a full width at half maximum of $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$2, in good agreement with the estimate. This is the basis for the reported tens-of-picosecond regime and for the observation of a decay time of $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$3 on low-temperature-grown GaAs (Iwaya et al., 2023).

A recurrent misconception is that the diffraction-limited optical spot fixes the spatial resolution. In OPP-STM the optical field is global at the micrometre scale, but the measured signal is the tunneling current at the immediate tip location. The spatial resolution is therefore fundamentally that of STM: the tunneling current is dominated by states within a few Å of the tip apex, and the pump-probe contrast can be localized to nanoscale or, in principle, atomic-scale features even though the optical beam itself is diffraction-limited (Iwaya et al., 2023, Iwaya et al., 2024).

The semiconductor benchmark system is GaAs(110). On $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$4-type GaAs(110) at positive sample bias, illumination generates electron-hole pairs; holes accumulate near the surface, electrons are driven into the bulk, and the resulting surface photovoltage screens the tip-induced field and shifts the bands by about $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$5 at $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$6. Delay curves at 78 K under 532 nm illumination are fitted with a bi-exponential form and yield $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$7 and $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$8, interpreted respectively as bulk-side decay and surface-side decay. In low-temperature-grown GaAs, the sharp feature near $$\Delta I(t_\mathrm{d}) = \langle I(t_\mathrm{d}) \rangle - \langle I(t_\mathrm{max}) \rangle,$$9 is the pump-probe cross-correlation rather than a material relaxation; fitting the tails for $t_\mathrm{max}$0 gives $t_\mathrm{max}$1, consistent with fast defect-mediated recombination (Iwaya et al., 2023).

The same system supports long-duration nanoscale mapping. A constant-current STM image of a $t_\mathrm{max}$2 region on GaAs(110) at 6 K was followed by a $t_\mathrm{max}$3 grid-point pump-probe measurement, giving 2500 delay curves with 3 ns delay steps and 0.15 s averaging per point for a total mapping time of about 21 hours. The OPP amplitude at $t_\mathrm{max}$4 is enhanced along a bump perimeter and step edge, and the fitted decay-time map shows significantly shorter decay times of $t_\mathrm{max}$5–$t_\mathrm{max}$6 inside the bump than on the surrounding terrace. Double-exponential fitting within the bump resolves a fast component narrowly distributed around 10–50 ns and a slow component spanning 100–550 ns (Iwaya et al., 2023).

In the AFM branch, multilayer WSe$t_\mathrm{max}$7 provides a complementary example. Time-resolved $t_\mathrm{max}$8 signals measured with high signal-to-noise ratio are fitted with two exponential decays, which are interpreted as a faster surface-recombination channel and a slower diffusion channel in the layered semiconductor. This broadens the technique from conductive systems addressable by STM to insulating or weakly conductive systems addressable by force detection (Iwaya et al., 2024).

5. Electrical pump-probe, pulsed ESR, and spin-dynamics platforms

The all-electrical branch of multimodal STM-based pump-probe replaces optical excitation with nanosecond voltage or RF pulse sequences applied through the bias line. A straightforward implementation uses waveform sequencing in a standard AWG together with a simple transfer relay that connects either the regular STM control electronics or the AWG/RF chain to the STM bias line. Short arbitrary waveforms encode pump and probe pulses, probe states A and B are alternated for lock-in detection, and RF bursts can be generated by driving the pulse-trigger input of a microwave source. Demonstrated modes include DC/DC pump-probe, RF/DC pump-probe, and RF/RF operation for pulsed ESR. Cross-correlation through a nonlinear junction shows clean DC/DC and DC/RF overlap traces and highlights a $t_\mathrm{max}$9 time resolution, here limited by the speed of the available electronics (Natterer, 2019).

A more elaborate realization integrates this electrical methodology into a UHV dilution-refrigerator STM with vector magnetic field. Reported base temperature is $t_\mathrm{d}$0, with out-of-plane field $t_\mathrm{d}$1 and in-plane field $t_\mathrm{d}$2, rotatable $t_\mathrm{d}$3 in one plane with magnitude up to 3 T. Two RF lines reach the microscope, and transmission to the junction is characterized from 1 to 30 GHz. At 10.94 GHz and $t_\mathrm{d}$4, the extracted RF amplitude at the junction is $t_\mathrm{d}$5 zero-to-peak; at 20 dBm the thermometer near the sample rises by only $t_\mathrm{d}$6, indicating that appreciable RF excitation is compatible with mK operation (Weerdenburg et al., 2020).

In this instrument, ESR-STM and pump-probe spectroscopy are not independent add-ons but two modes of one high-frequency architecture. ESR is demonstrated on individual TiH molecules on MgO/Ag(100) using both frequency-sweep and magnetic-field-sweep modes. The ESR condition is expressed as

$t_\mathrm{d}$7

and measured resonances give $t_\mathrm{d}$8 from low-frequency frequency-sweep data and $t_\mathrm{d}$9 from magnetic-field-sweep data. ESR is observed down to $t_\mathrm{d}$0, with mean linewidth $t_\mathrm{d}$1 and mean peak intensity $t_\mathrm{d}$2 (Weerdenburg et al., 2020).

The same platform performs pump-probe measurements of spin relaxation on individual Fe atoms on MgO/Ag(100). Typical pulse parameters are a 30 mV pump pulse of width $t_\mathrm{d}$3 or $t_\mathrm{d}$4 and a 5 mV or 1 mV probe pulse of the same width, with rise times of 5 ns. The probe reads out the nonequilibrium spin population prepared by the pump, and the delay dependence is fitted with

$t_\mathrm{d}$5

The extracted $t_\mathrm{d}$6 increases from about $t_\mathrm{d}$7 at $t_\mathrm{d}$8 to about $t_\mathrm{d}$9 at $t_\mathrm{d}$0, while adding an in-plane field component $t_\mathrm{d}$1 drastically reduces $t_\mathrm{d}$2. Multimodality in this setting means that conventional STM/STS, ESR-STM, and pump-probe relaxation spectroscopy all operate on the same atom with the same tip, sample, cryostat, and high-frequency wiring (Weerdenburg et al., 2020).

6. Limits, misconceptions, and likely directions of development

Several common misunderstandings are directly addressed by the existing literature. One is that ultrafast resolution in STM requires ultrafast current amplifiers. In optical OPP-STM, the bandwidth of the STM electronics is not the limiting factor because the measured quantity is the average current over many pump-probe cycles; the relevant resolution is the pump-probe cross-correlation set by pulse width and jitter. Another is that pump-probe contrast must be generated by intensity chopping. In STM this is often undesirable because periodic heating produces vertical expansion and spurious current changes, which is why delay-time modulation rather than intensity modulation has become the preferred differential scheme (Iwaya et al., 2023, Iwaya et al., 2024).

The current limitations differ across implementations. In externally triggerable optical OPP-STM, the present temporal ceiling is about $t_\mathrm{d}$3, with delay range up to about $t_\mathrm{d}$4 for a 1 MHz repetition rate. Reaching sub-10-ps or few-fs regimes would require much shorter pulses, lower laser and trigger jitter, or a different operating principle such as field-driven STM using THz or mid-IR pulses. The optical demonstration discussed here also uses fixed 532 nm lasers, whereas many target materials would benefit from wavelength flexibility. The same sources note that externally triggerable picosecond lasers exist in the 532–1550 nm range, and that dual-color excitation, circular-polarization modulation, and time-resolved $t_\mathrm{d}$5-$t_\mathrm{d}$6 or $t_\mathrm{d}$7 spectroscopy are natural extensions of the architecture (Iwaya et al., 2023, Iwaya et al., 2024).

In all-electrical systems, the limiting factors are different: AWG bandwidth, pulse edge times, impedance mismatch, bias-tee ringing, and pulse distortions on the high-frequency line. The 2019 implementation reports $t_\mathrm{d}$8 effective resolution set by the electronics, while the mK instrument verifies clean pulses down to about 250 ns at the junction but notes distortions for $t_\mathrm{d}$9, implying that active pulse shaping would be required for reliable sub-100-ns dynamics (Natterer, 2019, Weerdenburg et al., 2020).

Several expansion routes are already identified. One is multiprobe OPP-STM, in which one tip serves as an electrode on an insulating substrate and another tip performs STM; OPP-MP measurements have been demonstrated on monolayer transition metal dichalcogenides to investigate exciton dynamics. Another is the extension of the compact optical pump-probe module to other scanning probe modes such as Kelvin probe force microscopy and near-field optical microscopy. A plausible implication is that single-shot spatiotemporal division of probe pulses, demonstrated in a purely optical imaging system, may inform future multiplexed delay-encoding strategies for STM-based pump-probe measurements, although the STM junction remains a local single-channel detector rather than a spatially resolved camera (Iwaya et al., 2024, Yeola et al., 2018).

Taken together, these developments define multimodal STM-based pump-probe technique as a convergent instrumentation paradigm rather than a single apparatus. Its unifying features are local nonequilibrium excitation, electronically or optically controlled delay, lock-in extraction of small differential signals, and the reuse of one stable high-frequency infrastructure across tunneling-current, force, and spin-resonance observables. Within that framework, the technique spans carrier dynamics in GaAs, force-detected dynamics in WSe$f_\mathrm{rep}=1~\mathrm{MHz}$0, nanosecond electrical pump-probe spectroscopy, and ESR-assisted spin-relaxation measurements at 30 mK in vector magnetic field.