Phase-Resolved Ultrafast Spectroscopy Overview

• Phase-Resolved Ultrafast Spectroscopy is a set of techniques that preserve phase information in ultrafast measurements to analyze complex optical fields and material responses.
• It employs methods like spectral interferometry, heterodyne detection, and time-frequency analysis to extract real and imaginary components with femtosecond resolution.
• These techniques enable detailed studies of molecular dynamics, chiro-optical responses, and transient material phases, driving advances in ultrafast and nanoscale spectroscopy.

Searching arXiv for the specified papers and closely related work to support the article. Phase-resolved ultrafast spectroscopy denotes a family of ultrafast measurement strategies in which the experimentally relevant phase information is retained rather than averaged away, so that the detected signal carries access to a complex field, a complex material response, a relative interferometric phase, or a phase-dependent distinction between material phases. In contemporary usage, however, the phrase spans several distinct meanings. In a narrow optical sense it refers to measurements that recover the emitted or transmitted electric field and its phase, as in spectral-interferometric or heterodyne schemes. In a broader materials-spectroscopy sense it can also denote ultrafast measurements resolved across different structural, electronic, or topological phases of matter, or measurements in which coherent beatings, interference, and relative timing constrain phase-sensitive interpretation. The literature therefore includes both strict field-phase-resolved methods and adjacent approaches—such as time-frequency analysis, tr-ARPES, high-harmonic spectroscopy, and ultrafast EELS—that do not retrieve optical phase directly but resolve transient matter phases, coherent oscillations, or complex-response quadratures that are essential for mechanistic interpretation (Pandey et al., 7 Mar 2025).

1. Definitions and scope

A strict optical definition is given by methods that measure the complex emitted field $E_S(\omega)=|E_S(\omega)|e^{i\varphi_S(\omega)}$ and reconstruct the temporal field $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$ by Fourier transform. In this class, phase resolution means access to the spectral phase $\varphi_S(\omega)$, the temporal phase $\phi_S(t)$, or both, usually relative to a calibrated reference field (Pandey et al., 7 Mar 2025). This is the sense used in femtosecond temporal phase-resolved nonlinear optical spectroscopy, where the observable is the emitted real-time electric field from an ultrafast third-order nonlinear optical interaction in molecules and the signal is related to the induced third-order polarization $P^{(3)}(t)$ governed by $\chi^{(3)}_{ijkl}$ (Pandey et al., 7 Mar 2025).

A second, still genuinely phase-sensitive, meaning arises in broadband chiro-optical pump-probe experiments, where the measured quantity is the complex chiro-optical susceptibility $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$. Here phase resolution means separation of the real and imaginary parts of the transient chiral response, so that transient optical rotatory dispersion and transient circular dichroism are obtained simultaneously from one interferometric measurement (Gucci† et al., 13 Nov 2025). A closely related visible-wavelength time-domain perspective appears in asymmetric double-pulse interferometric frequency-resolved optical gating, where the objective is reconstruction of the complex optical response—equivalently the complex dielectric function $\varepsilon(\omega)$, refractive index $\tilde n(\omega)$, or conductivity $\sigma(\omega)$—with ultrafast temporal resolution (Chan et al., 2022).

A broader usage includes methods that do not retrieve a full optical field phase but nonetheless retain phase-sensitive or phase-constraining structure. Time-frequency resolved ultrafast spectroscopy using the continuous wavelet transform is an example: it does not perform explicit phase retrieval of ultrafast spectroscopic fields, nor does it present a dedicated heterodyne or phase-cycling formalism, but it reveals when distinct oscillatory components appear, coexist, interfere, and exchange dominance, thereby sharpening phase-resolved interpretation of non-stationary beatings (Prior et al., 2013). tr-ARPES studies of $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$0, In/Si(111), and related systems provide another broader usage, where “phase-resolved” refers to spectroscopy resolved across different structural, electronic, or topological phases rather than the optical phase of a pulse (Crepaldi et al., 2017, Nicholson et al., 2018).

This range of meanings produces a recurring terminological ambiguity. Some works are phase resolved in the strict interferometric sense; others are phase resolved with respect to a complex response function; others are phase resolved only in the sense of material-phase discrimination. A precise reading therefore depends on the observable: electric field, complex susceptibility, coherent beating structure, or phase-specific band dynamics.

2. Field-resolved nonlinear optical spectroscopy

A direct implementation of phase-resolved ultrafast spectroscopy is spectral-interferometric retrieval of weak nonlinear emission. In the lock-in-enabled molecular method demonstrated on impulsively aligned $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$1, the signal and a known reference interfere on a spectrometer, and the detected spectrum is written as

$E_S(t)=|E_S(t)|e^{i\phi_S(t)}$2

with $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$3. Fourier-transform spectral interferometry then retrieves the signal spectral phase from the interference sideband, and inverse Fourier transformation yields the temporal electric field (Pandey et al., 7 Mar 2025). In the demonstrated non-degenerate four-wave-mixing experiment, the method measures the full emitted electric field of an ultrafast nonlinear optical signal even when that signal is extremely weak and sits on top of a large coherent background (Pandey et al., 7 Mar 2025).

The technical obstacle addressed in that work is coherent background contamination. With a background field $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$4, the measured spectrum contains signal–reference, background–reference, and signal–background interference terms,

$E_S(t)=|E_S(t)|e^{i\phi_S(t)}$5

The key instrumental advance is a software lock-in imaging spectrometer that demodulates the detector image both at the chopper frequency and at 0 Hz, thereby separating the desired signal–reference interference into the lock-in AC channel and the background–reference fringes into the DC channel (Pandey et al., 7 Mar 2025). Because the background fringes and signal fringes share the same interferometric drift $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$6, subtracting the DC fringe phase from the AC fringe phase passively removes slow interferometric drift. This yields a drift-corrected relative signal phase without active stabilization or a tracer beam (Pandey et al., 7 Mar 2025).

The phase recovered in such schemes is explicitly relative. Since the reference pulse is independently characterized by FROG, the signal spectral phase can be reconstructed relative to that calibrated reference; however, the drift correction itself is a common-mode subtraction. This means that the method guarantees a stable relative phase and preserves pump-induced phase shifts that would otherwise wash out, but it is not carrier-envelope-phase metrology (Pandey et al., 7 Mar 2025). The molecular demonstration shows that the retrieved temporal phase distinguishes a positively chirped signal pulse at negative pump–probe delay, a negatively chirped signal pulse near pump–probe overlap due to free-electron-induced negative dispersion, and a reduced or modulated chirp at the anti-aligned delay $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$7 ps (Pandey et al., 7 Mar 2025). The result is therefore a genuine temporal phase-resolved nonlinear spectroscopy of molecular dynamics.

A visible-wavelength analogue appears in asymmetric double-pulse interferometric frequency-resolved optical gating. Standard SHG-FROG is insensitive to time direction, zero-order phase, and first-order phase. The asymmetrically chirped calibration–probe double pulse creates diagonal interference terms in the FROG trace whose phase depends on the relative zero-order phase,

$E_S(t)=|E_S(t)|e^{i\phi_S(t)}$8

with $E_S(t)=|E_S(t)|e^{i\phi_S(t)}$9. This phase-locking mechanism preserves the relative low-order phases needed for visible time-domain spectroscopy (Chan et al., 2022). After calibration-based alignment of sample and reference pulse trains, the reconstructed probe fields are Fourier transformed and inserted into the Fresnel relation

$\varphi_S(\omega)$0

up to sign convention, enabling direct extraction of the complex dielectric function in the visible (Chan et al., 2022).

Phase-resolved transient absorption in silicon provides a complementary case in which the measured phase is not the emitted field phase of a nonlinear signal but the pump-induced phase shift of a transmitted probe. The probe is split into reference and probe pulses; the reference traverses the unexcited sample and the probe the excited sample, so the pump-induced refractive-index change $\varphi_S(\omega)$1 generates a temporal shift $\varphi_S(\omega)$2, measurable as a spectral-interferometric phase shift $\varphi_S(\omega)$3 (Wörle et al., 2021). Combined with transient optical density

$\varphi_S(\omega)$4

the phase data permit a Drude analysis of the complex free-carrier response: $\varphi_S(\omega)$5 From this the collision time and transient optical effective mass are extracted, including a collision-time rise with $\varphi_S(\omega)$6 fs timescale and a bi-exponential decrease of $\varphi_S(\omega)$7 from about $\varphi_S(\omega)$8 to about $\varphi_S(\omega)$9 with time constants of 4 fs and $\phi_S(t)$0 fs (Wörle et al., 2021). This is phase resolved in the sense of measuring both absorptive and dispersive components of the transient optical response.

3. Complex-response detection beyond scalar intensity

A major expansion of phase-resolved ultrafast spectroscopy has been the recovery of both quadratures of a complex response function rather than only a field or intensity. Broadband ultrafast self-heterodyned chiro-optical spectroscopy is exemplary. A linearly polarized probe generates a strong vertically polarized achiral free induction decay (AFID) and a weak horizontally polarized chiral free induction decay (CFID). A birefringent common-path interferometer scans the relative delay $\phi_S(t)$1 between these components, and a polarization bridge with balanced detection measures a chiral interferogram $\phi_S(t)$2 or pump-induced $\phi_S(t)$3 (Gucci† et al., 13 Nov 2025). Fourier transformation and normalization by an achiral calibration spectrum then yield the complex chiro-optical response, with the real part corresponding to $\phi_S(t)$4ORD and the imaginary part to $\phi_S(t)$5CD (Gucci† et al., 13 Nov 2025).

In the static modeling of the helicoid array, the circular-basis transmission coefficients $\phi_S(t)$6 and $\phi_S(t)$7 are converted to ORD and CD through

$\phi_S(t)$8

The experiment therefore distinguishes the dispersive and absorptive chiro-optical quadratures of the transient response in one measurement (Gucci† et al., 13 Nov 2025). This is a strict phase-sensitive measurement because the detected signal depends on the interference between AFID and CFID and therefore on their relative phase, yet it is not framed as direct electric-field waveform retrieval.

A more radical departure is ultrafast pump-probe phase-randomized tomography. Here the interferometric detection is phase sensitive at the quadrature level, but the optical phase is not stabilized; instead, random carrier-envelope-phase fluctuations are used to sample phase space uniformly (Glerean et al., 2024). The measured histogram is the phase-averaged quadrature distribution of a weak probe, and an iterative maximum-likelihood reconstruction yields the diagonal Fock-basis elements

$\phi_S(t)$9

from phase-averaged projectors $P^{(3)}(t)$0 (Glerean et al., 2024). This approach does not recover the full phase-sensitive optical state of a conventional homodyne tomography experiment. It is intentionally phase averaged. Its relevance lies in showing that ultrafast spectroscopy can target photon-number statistics and fluctuation observables rather than field phase itself (Glerean et al., 2024).

Single-photon pulse metrology occupies another adjacent territory. Variable electro-optic shearing interferometry measures the modulus squared of the self-gated short-time Fourier transform

$P^{(3)}(t)$1

using variable temporal delay $P^{(3)}(t)$2 and electro-optically applied frequency shear $P^{(3)}(t)$3 (Kurzyna et al., 2022). In the weak-light regime, the normalized second-order correlation

$P^{(3)}(t)$4

yields the modulus-squared STFT, from which spectral amplitude and spectral phase are recovered computationally by iterative phase retrieval (Kurzyna et al., 2022). This is not direct experimental phase measurement, but it provides weak-light-compatible access to pulse phase needed for phase-sensitive spectroscopy.

An even more infrastructural development is TSUBAME, a scan-free MIR pulse metrology scheme that combines MIR-to-NIR upconversion, time stretch, and spectral interferometry to recover the spectral phase of each MIR pulse at 1 MHz (Deng et al., 30 Jun 2026). The directly measured per-shot quantity is the spectral phase difference

$P^{(3)}(t)$5

from which the MIR spectral phase is reconstructed relative to a known reference (Deng et al., 30 Jun 2026). TSUBAME does not measure a sample response, but it supplies pulse-to-pulse phase information at the native repetition rate of a high-speed MIR source, an enabling capability for phase-sensitive strong-field and spectroscopy experiments (Deng et al., 30 Jun 2026).

4. Time-frequency analysis and phase-constraining interpretation

A narrower experimental definition of phase-resolved spectroscopy does not exhaust the practical problem of phase-sensitive interpretation. Many ultrafast signals are non-stationary: different oscillatory components emerge, overlap, shift, and decay on femtosecond to picosecond timescales. In such cases a global Fourier transform retains frequency content but loses the temporal sequence of events. Time-frequency resolved ultrafast spectroscopy based on the continuous wavelet transform addresses this problem (Prior et al., 2013).

The wavelet formalism uses translated and scaled atoms

$P^{(3)}(t)$6

and defines

$P^{(3)}(t)$7

The resulting scalogram maps wavelet intensity over time and scale or frequency (Prior et al., 2013). The method reveals temporal localization of oscillatory components, coexistence or succession of multiple frequencies, frequency evolution in time, coherence lifetime, transient mode switching, and relative phase-sensitive interference effects when overlapping components suppress or enhance one another (Prior et al., 2013).

The synthetic three-sinusoid example in that work already establishes the point: two signals with identical global Fourier spectra are immediately distinguished by the scalogram because one contains non-overlapping bursts while the other contains partial temporal overlap (Prior et al., 2013). In a dimer–vibration model relevant to coherent excitonic energy transport, the Hamiltonian

$P^{(3)}(t)$8

with pure dephasing

$P^{(3)}(t)$9

generates a coherence trace

$\chi^{(3)}_{ijkl}$0

whose wavelet analysis resolves an early high-frequency excitonic oscillation near $\chi^{(3)}_{ijkl}$1 and a later lower-frequency mode-driven response near $\chi^{(3)}_{ijkl}$2 (Prior et al., 2013). The crucial information is not merely that both frequencies exist, but when each dominates and that the low-frequency component appears with a time delay, constraining causal interpretation (Prior et al., 2013).

In the resonant case, where the initial excitonic oscillation and later mode-driven oscillation occur at essentially the same frequency, the time trace exhibits a suppressed-amplitude interval followed by revival-like growth. The authors interpret this as destructive interference due to relative phase, arguing semiclassically that the vibrational response is approximately $\chi^{(3)}_{ijkl}$3 out of phase with the initial coherence and that the mode-driven excitonic component then becomes approximately $\chi^{(3)}_{ijkl}$4 out of phase with the original oscillation (Prior et al., 2013). The method does not extract a full phase trajectory from the wavelet coefficients; rather, it sharpens phase-sensitive interpretation by temporally separating overlapping components and identifying where interference occurs (Prior et al., 2013).

The same logic applies to transient absorption-like signals with a time-dependent Stokes shift. For

$\chi^{(3)}_{ijkl}$5

with

$\chi^{(3)}_{ijkl}$6

the wavelet scalogram shows a smooth migration of spectral weight from the initial unrelaxed transition frequency to the final relaxed one, whereas a Fourier transform may produce broad or misleading multi-peak structures (Prior et al., 2013). This suggests that in many ostensibly phase-sensitive ultrafast datasets, robust phase interpretation presupposes a prior time-frequency decomposition of non-stationary content.

5. Material-phase-resolved ultrafast spectroscopy

A significant body of ultrafast literature uses “phase-resolved” in a materials sense: the measurement distinguishes transient structural, electronic, or topological phases even though it does not resolve optical phase. tr-ARPES provides the clearest examples. In $\chi^{(3)}_{ijkl}$7, a comparison between the low-temperature orthorhombic Weyl semimetal phase and the high-temperature monoclinic trivial phase shows that relaxation of pump-induced conduction-band population is systematically faster in the low-temperature phase (Crepaldi et al., 2017). Difference maps

$\chi^{(3)}_{ijkl}$8

are integrated over small windows near the predicted Weyl-point region, and each trace is fitted by a step function multiplied by a single decaying exponential and broadened by a resolution-limited Gaussian (Crepaldi et al., 2017). The extracted relaxation times are about 30–50% larger in the trivial 1T'' phase than in the Weyl semimetal 1T' phase, which is interpreted as a fingerprint of local gap closure and an additional efficient interband electron-electron scattering channel (Crepaldi et al., 2017). The “phase” here is material phase, not optical phase.

Another tr-ARPES study on $\chi^{(3)}_{ijkl}$9-MoTe$$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$0 resolves the unoccupied Weyl-node dispersion and then tracks pump-induced hot-carrier dynamics near the node. The Weyl node is identified at $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$1 meV above $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$2, and two intrinsic recovery timescales are extracted: a fast $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$3 fs component attributed to hot-electron cooling by optical phonon cascade emission, and a slow $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$4 ps component attributed to anharmonic decay of hot optical phonons (Wan et al., 2017). The transient electronic temperature is well described by a two-temperature model, yielding $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$5 (Wan et al., 2017). Again, the measurement is time resolved and momentum resolved, but not optical-phase resolved.

In In/Si(111) nanowires, high-repetition-rate XUV trARPES resolves the unoccupied dispersions and population dynamics of two structural phases, the metallic $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$6 phase and the gapped $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$7 phase (Nicholson et al., 2018). “Excited state band mapping” at $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$8 fs shows unoccupied dispersions up to about 1 eV above $$\chi_\text{ch}(\omega)\propto \mathrm{ORD}(\omega)+ i\,\mathrm{CD}(\omega)$$9, and comparison to LDA and GW reveals a direct $\varepsilon(\omega)$0-point gap of 270 meV in the $\varepsilon(\omega)$1 phase, versus 60 meV in LDA and 240 meV in GW (Nicholson et al., 2018). The momentum-resolved transient population dynamics are modeled by a spectral function

$\varepsilon(\omega)$2

with $\varepsilon(\omega)$3 a Fermi-Dirac distribution at transient electronic temperature $\varepsilon(\omega)$4 (Nicholson et al., 2018). The work is phase resolved in the sense that the two structural phases and the photo-induced phase transition are spectroscopically distinguished, not because any optical phase is measured (Nicholson et al., 2018).

Time-resolved high-harmonic generation in VO$\varepsilon(\omega)$5 occupies a similar category. The measured observable is harmonic yield, not harmonic spectral phase, and the technique discriminates the equilibrium monoclinic insulating phase M1, the equilibrium rutile metallic phase R, and the metastable monoclinic metallic hidden phase M (Bionta et al., 2020). In the intermediate fluence regime, the HHG yield recovers on $\varepsilon(\omega)$6–$\varepsilon(\omega)$7 ps timescales matching the hidden-phase formation time from ultrafast electron diffraction, while at high fluence the yield remains quenched for 10 ps, consistent with rutile formation (Bionta et al., 2020). This is phase resolved with respect to material-phase pathways rather than emitted-field phase.

These examples clarify a common misconception. Ultrafast measurements described as “phase resolved” are not necessarily phase sensitive in the interferometric sense. In tr-ARPES and tr-HHG, the primary phase variable may be the crystallographic or topological phase of the sample, or a transient phase-transition pathway, rather than the optical phase of the probing field.

6. XUV, attosecond, and nanoscale extensions

At shorter wavelengths, phase-resolved ultrafast spectroscopy becomes technically more demanding because the relevant coherence periods are sub-femtosecond and the phase stability requirements correspondingly severe. A seeded free-electron-laser experiment on helium and argon overcomes this by transferring phase control from ultraviolet seed pulses to XUV harmonic pulses (Wituschek et al., 2019). The seed-pulse phase difference $\varepsilon(\omega)$8 is multiplied in high-gain harmonic generation so that the XUV pulse-pair phase becomes $\varepsilon(\omega)$9, and shot-to-shot modulation of $\tilde n(\omega)$0 implements XUV phase cycling without direct XUV optics (Wituschek et al., 2019). The wave-packet signal is written as

$\tilde n(\omega)$1

and with phase modulation and lock-in detection as

$\tilde n(\omega)$2

Referencing to a cw laser at optical frequency $\tilde n(\omega)$3 yields a downshifted rotating-frame frequency $\tilde n(\omega)$4, permitting direct reconstruction of the complex signal $\tilde n(\omega)$5 and therefore amplitude and phase (Wituschek et al., 2019). In helium, the $\tilde n(\omega)$6 coherence at 23.74 eV is tracked; in argon, a $\tilde n(\omega)$7 autoionizing resonance at 28.51 eV reveals a characteristic $\tilde n(\omega)$8-like spectral phase jump across the Fano profile (Wituschek et al., 2019). This is a direct XUV implementation of phase-cycled coherent spectroscopy.

Interferometric XUV imaging methods provide a different extension. Fourier-transform spectroscopic holography combines Fourier transform spectroscopy with Fourier transform holography by using two phase-locked HHG beams and delay-dependent multi-harmonic holograms

$\tilde n(\omega)$9

The phase-resolved quantity is the complex reference–sample interference field at each harmonic, encoded through the controlled delay term $\sigma(\omega)$0 (Strauch et al., 2024). This work is not a direct pump–probe spectroscopy of sample dynamics, but it provides a phase-stable XUV spectromicroscopy platform that could be extended to such measurements (Strauch et al., 2024).

At the nanoscale, time-resolved and ultrafast electron energy-loss spectroscopy constitutes another extension. The fundamental low-loss observable is the loss function

$\sigma(\omega)$1

which makes ultrafast EELS sensitive to transient dielectric response, plasmons, phonons, excitons, and local near fields (Lee et al., 6 Oct 2025). For nanostructures, the energy-loss probability is written as

$\sigma(\omega)$2

linking the measured signal to the induced field along the electron trajectory (Lee et al., 6 Oct 2025). Most ultrafast EELS beyond PINEM is not optical-phase resolved, but it is often phase resolved in the sense of tracking coherent lattice motion, oscillatory plasmonic responses, and local field sign on nanometer length scales (Lee et al., 6 Oct 2025).

Finally, time-resolved inner-shell photoelectron spectroscopy of CH$\sigma(\omega)$3I illustrates how ultrafast core-level observables can follow local electronic-structure transformation even without phase resolution. Above the iodine 4d edge, the molecular 4d line near 57.3 eV shifts toward an atomic-like 58.3 eV as dissociation proceeds, while fragment ion signals appear later because they are delayed by charge-transfer and Auger-driven Coulomb dynamics (Brauße et al., 2019). The method is time resolved and site specific, but it is not phase resolved in the interferometric sense.

7. Conceptual tensions, misconceptions, and future directions

A recurrent misconception is that all “phase-resolved ultrafast spectroscopy” is about optical phase retrieval. The literature shows at least four distinct meanings. First, there is strict field-phase retrieval, as in spectral interferometry and visible or MIR pulse characterization (Pandey et al., 7 Mar 2025, Deng et al., 30 Jun 2026). Second, there is complex-response quadrature measurement, as in transient ORD and CD or visible time-domain dielectric-function retrieval (Gucci† et al., 13 Nov 2025, Chan et al., 2022). Third, there is time-frequency analysis that does not retrieve phase directly but constrains phase-sensitive interpretation of non-stationary beatings (Prior et al., 2013). Fourth, there is material-phase-resolved spectroscopy, in which different structural or topological phases are distinguished in time and momentum space (Crepaldi et al., 2017, Nicholson et al., 2018, Bionta et al., 2020).

A second misconception is that intensity-only methods are necessarily sufficient if a signal oscillates. Several of the works considered here show the opposite. In silicon, combining transient absorption with probe phase reveals collision time and transient effective mass that cannot be disentangled from intensity alone (Wörle et al., 2021). In broadband chiro-optics, simultaneous access to ORD and CD avoids inferring one quadrature from the other by Kramers–Kronig relations (Gucci† et al., 13 Nov 2025). In molecular nonlinear optics, full-field retrieval separates chirp and phase shifts that would remain hidden in spectrally integrated intensity (Pandey et al., 7 Mar 2025). Conversely, the wavelet work shows that even before direct phase retrieval, non-stationary time-frequency structure must often be resolved to prevent mechanistic misassignment (Prior et al., 2013).

A third tension concerns the role of statistical observables. Ultrafast pump-probe phase-randomized tomography deliberately gives up phase-stable field reconstruction in order to access photon-number statistics and fluctuation observables in the few-photon regime (Glerean et al., 2024). This suggests a plausible implication: as ultrafast spectroscopy moves toward weak-light, quantum-light, and fluctuation-sensitive regimes, “phase resolution” may no longer mean preserving a deterministic optical phase but may instead refer to preserving the phase-space information required to reconstruct phase-averaged quantum optical observables.

Several future directions are already explicit in the cited work. Lock-in-enabled spectral interferometry is applicable to gas, liquid, and solid samples and can be combined with multidimensional measurement spaces for passive phase tracking (Pandey et al., 7 Mar 2025). Visible-wavelength ADI-FROG is proposed as a route to ultrafast-compatible complex dielectric-function measurements in spectral regions where electro-optic sampling is impractical (Chan et al., 2022). TSUBAME suggests that real-time pulse-to-pulse MIR spectral phase metrology at MHz rates may support closed-loop stabilization and data sorting in high-repetition-rate phase-sensitive experiments (Deng et al., 30 Jun 2026). XUV phase cycling via seeded FEL harmonics opens a route to multidimensional XUV coherent spectroscopy (Wituschek et al., 2019). Interferometric XUV spectromicroscopy could, with pump–probe extension, provide complex spectroscopic images $\sigma(\omega)$4 rather than scalar intensity maps (Strauch et al., 2024). Ultrafast EELS, especially with monochromation, time tagging, and improved nonequilibrium theory, points toward nanoscale spectroscopies of coherent collective modes that are local in both real and reciprocal space (Lee et al., 6 Oct 2025).

Taken together, these developments show that phase-resolved ultrafast spectroscopy is not a single technique but a convergent set of strategies for retaining complex information in the ultrafast limit. Depending on the implementation, that information may be the optical phase of a pulse, the phase of an emitted nonlinear field, the real and imaginary parts of a susceptibility, the temporal architecture of non-stationary beatings, or the transient spectral signatures of distinct electronic and structural phases. The unifying principle is that mechanistically decisive information is carried not only by intensity, but by the phase, quadrature, timing, or complex structure that intensity-only representations usually discard.