Ultrafast Raman Thermometry

• Ultrafast Raman thermometry is a technique that uses time-resolved Raman scattering to infer transient, mode-specific temperatures from Raman observables.
• The method spans spontaneous, coherent, and stimulated regimes, analyzing Stokes/anti-Stokes asymmetry, spectral shifts, and motional sidebands to reveal phonon and quasiparticle dynamics.
• Practical implementations balance time-energy trade-offs in diverse systems—from quantum materials to gases and ion traps—highlighting its role in subsystem-specific thermal characterization.

Ultrafast Raman thermometry is the use of time-resolved Raman scattering to infer transient temperatures, or subsystem-specific effective temperatures, from Raman observables measured after ultrafast excitation. In current practice, the term spans several experimentally distinct regimes: spontaneous Raman thermometry based on Stokes/anti-Stokes asymmetry of phonons or electronic continua, coherent Raman-scattering thermometry based on model fits to rotational or vibrational spectra, and stimulated Raman thermometry based on motional sidebands or Doppler-broadened Raman lines. Across these regimes, the central objective is the same: to convert a Raman observable into a temperature-like quantity for a selected degree of freedom—lattice, quasiparticles, magnons, molecular rotation, or trapped-particle motion—on timescales ranging from sub-picosecond dynamics in quantum materials to kHz single-shot acquisition in gases and plasmas (Chou et al., 2023, Lu et al., 2024, Wang et al., 2018, Versteeg et al., 2018, Gatuingt et al., 6 Aug 2025, Mielke et al., 2021).

1. Thermometric principles

The most direct ultrafast Raman thermometer uses detailed balance between Stokes and anti-Stokes scattering. For a Raman-active phonon of frequency $\omega_{\text{ph}}$, the phonon occupation is

$$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$

and the spontaneous Raman intensity ratio can be written as

$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$

with the cubic frequency factor reflecting photon-counting detection rather than energy detection (Chou et al., 2023). In time-resolved experiments, repeating this measurement versus pump–probe delay yields a transient phonon temperature $T_{\text{ph}}(t)$, provided the probed mode remains describable by a Bose–Einstein occupation.

An analogous detailed-balance logic applies to low-energy electronic Raman continua. In the equilibrium limit,

$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$

so that

$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$

In sub-picosecond Raman thermometry with finite-bandwidth probe pulses, this relation must be corrected; in Bi$_2$Sr$_2$CaCu$_2$O$_{8+\delta}$ the extracted quasiparticle temperature is obtained from the pulse-bandwidth-corrected Stokes/anti-Stokes ratio rather than from the monochromatic formula (Gatuingt et al., 6 Aug 2025).

A second thermometric class uses Raman peak shifts rather than intensity asymmetry. In steady-state Raman thermometry, the Raman shift $$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$0 is calibrated against temperature and the measured Raman temperature is an effective, power-weighted average over the illuminated volume,

$$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$1

for strongly surface-absorbed cases with approximately constant Raman cross section (Stoib et al., 2014). Time-resolved implementations inherit the same spatial-averaging structure. This suggests that even when the temporal resolution is ultrafast, the measured temperature is generally an effective probe-volume average rather than the local maximum temperature.

2. Observable classes and inversion procedures

In spontaneous ultrafast Raman thermometry, the observable is usually either a Stokes/anti-Stokes ratio or a time-dependent Raman intensity integrated over a selected spectral window. A flexible formulation for spontaneous Raman in pump–probe experiments writes

$$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$2

which makes explicit that both population factors and the Raman tensor enter the measured signal (Versteeg et al., 2018). In silicon, this permits direct extraction of a time-dependent LO-phonon temperature from anti-Stokes and Stokes intensities; in cuprates, the same structure underlies lattice thermometry from a spectator phonon and quasiparticle thermometry from the low-energy electronic continuum.

In coherent Raman-scattering thermometry, temperature is inferred from the full spectral envelope rather than from a single asymmetry ratio. In single-shot single-beam coherent Raman scattering thermometry, the third-order polarization is modeled as

$$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$3

with

$$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$4

The rotational populations entering $$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$5 follow Boltzmann statistics, so the measured coherent rotational spectrum can be least-squares fitted to determine $$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$6 on a shot-by-shot basis (Lu et al., 2024).

Stimulated Raman thermometry in trapped-ion systems uses yet another inversion procedure. In the sideband-resolved regime, the red and blue motional sidebands obey

$$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$7

with

$$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$8

In the Penning-trap $$n(\omega_{\text{ph}},T)=\frac{1}{e^{\hbar\omega_{\text{ph}}/k_B T}-1},$$9Be$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$0 implementation, the experiment reported is not yet sideband-resolved; instead, the Doppler width $$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$1 of a Raman line is fitted by a Gaussian and converted via

$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$2

This is a Raman thermometer, but in the CW, spectrally narrow limit rather than the ultrafast limit (Mielke et al., 2021).

3. Temporal and spectral implementations

Ultrafast Raman thermometry is constrained by the usual time–energy trade-off. In a dedicated time-resolved spontaneous Raman setup, the transform-limited Gaussian relation

$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$3

sets the balance between spectral resolution and temporal resolution (Versteeg et al., 2018). A Yb:KGW-based platform demonstrates both regimes: a narrow-band 512 nm probe with $$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$4 cm$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$5 and $$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$6 ps for high spectral resolution, and a broader femtosecond probe for sub-ps dynamics.

In driven YBa$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$7Cu$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$8O$$\frac{I_{\text{AS}}}{I_{\text{S}}} = \frac{(\omega_0+\omega_{\text{ph}})^3}{(\omega_0-\omega_{\text{ph}})^3} e^{-\hbar\omega_{\text{ph}}/k_B T_{\text{ph}}},$$9, ultrafast spontaneous Raman thermometry is implemented with 20 THz mid-infrared pump pulses of 800 fs duration and a 405 nm Raman probe of 600 fs duration in backscattering geometry, with both incoming probe and scattered photons polarized along the crystallographic c-axis (Chou et al., 2023). The thermometer is an undriven 15 THz A$T_{\text{ph}}(t)$0 apical oxygen phonon whose anti-Stokes/Stokes ratio tracks the lattice temperature during and after coherent phonon driving.

In Bi$T_{\text{ph}}(t)$1Sr$T_{\text{ph}}(t)$2CaCu$T_{\text{ph}}(t)$3O$T_{\text{ph}}(t)$4, the pump–probe architecture instead targets electronic Raman thermometry. The pump is 1030 nm, $T_{\text{ph}}(t)$5 fs, at 250 kHz; the probe is 515 nm with pulse-shaper-controlled time resolution of 0.36 ps or 0.43 ps (Gatuingt et al., 6 Aug 2025). Polarization-resolved $T_{\text{ph}}(t)$6 Raman scattering then selectively addresses the antinodal region of the cuprate Fermi surface, enabling simultaneous tracking of quasiparticle temperature and superconducting pair-breaking spectral weight.

For gases and plasmas, a distinct ultrafast architecture uses self-generated air lasing as the Raman probe. A Ti:sapphire regenerative amplifier at 800 nm, 40 fs, 6 mJ, 1 kHz produces N$T_{\text{ph}}(t)$7 air lasing at 428 nm in a 5 mbar N$T_{\text{ph}}(t)$8 cell; the air-lasing pulse is delayed by about $T_{\text{ph}}(t)$9 ps relative to the femtosecond pump and serves as a narrowband probe for coherent rotational Raman scattering in O$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$0 (Lu et al., 2024). The delayed picosecond probe intrinsically suppresses non-resonant four-wave-mixing background while preserving single-beam alignment.

These implementations show that “ultrafast” in Raman thermometry can refer either to femtosecond-to-picosecond intrinsic excitation and probing, or to high-speed acquisition such as 1 kHz single-shot operation. The instrumental design is therefore not uniform across the field; it is dictated by whether one prioritizes sub-ps nonequilibrium dynamics, shot-to-shot monitoring, symmetry selectivity, or sideband resolution.

4. Effective temperatures, thermalization, and nonthermal response

Ultrafast Raman thermometry does not always measure a thermodynamic temperature in the strict sense. A microscopic nonequilibrium theory for nonresonant time-resolved Raman scattering in the Hubbard model formalizes this by defining

$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$1

When the Stokes/anti-Stokes ratio is approximately frequency independent over a chosen window, $$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$2 behaves as a meaningful effective temperature; when it becomes strongly frequency dependent, the Raman response is dominated by coherent many-body dynamics or nonthermal distributions rather than thermalization (Wang et al., 2018). The same theory identifies weak-pump regimes where effective heating dominates and strong, near-resonant regimes where Floquet renormalization and nonthermal many-body effects invalidate a single-temperature description.

Experiments in quantum materials confirm this distinction. In driven YBa$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$3Cu$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$4O$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$5, time-resolved spontaneous Raman scattering yields a lattice temperature rise of approximately 80–85 K from the A$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$6 spectator phonon, whereas the equivalent quasiparticle temperature extracted from THz conductivity rises by approximately 200 K (Chou et al., 2023). The superfluid response is transiently enhanced only during the pump pulse, even as quasiparticles are hot and the lattice is only moderately heated. This is direct evidence that lattice, quasiparticle, and superconducting responses need not share a common temperature.

In Bi$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$7Sr$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$8CaCu$$I_S(\Omega)\propto [1+n_B(\Omega,T)]\,\chi''(\Omega),\qquad I_{AS}(-\Omega)\propto n_B(\Omega,T)\,\chi''(\Omega),$$9O$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$0, ultrafast Raman thermometry of the antinodal quasiparticles shows a rapid rise of the effective quasiparticle temperature from about 60 K to about 105 K near threshold fluence, while the superconducting pair-breaking spectral weight depletes with a finite delay and recovers on a different timescale (Gatuingt et al., 6 Aug 2025). The observed dynamics “cannot be framed in terms of a single effective electron temperature,” because the condensate and the quasiparticle sector evolve differently during both destruction and recovery.

Graphene under 3 ps excitation provides a further cautionary case. There, hot photoluminescence indicates electronic temperatures in the $$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$1–3100 K range, while the lattice in the probed volume remains below 700 K by an upper-bound estimate (Ferrante et al., 2017). The anti-Stokes/Stokes ratio is explicitly not a straightforward thermometer in this resonant, nonequilibrium regime because the Raman excitation profile is strongly wavelength dependent and correlated Stokes–anti-Stokes processes contribute. This establishes a common misconception: the mere availability of Stokes and anti-Stokes spectra does not guarantee a valid Raman temperature unless the underlying assumptions of detailed balance and time-independent matrix elements are satisfied.

5. Representative regimes and performance

In condensed-matter ultrafast thermometry, the technique has been used to resolve mode-specific and sector-specific dynamics. In silicon, time-resolved spontaneous Raman spectroscopy finds an LO-phonon temperature rise from about 310 K to about 355 K and a decay time of $$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$2 ps, directly revealing hot-phonon generation and anharmonic cooling (Versteeg et al., 2018). The same platform tracks unequal softening rates in the two-phonon overtone spectrum at different high-symmetry points, and in Cu$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$3OSeO$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$4 it measures magnon softening and broadening with a characteristic $$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$5 ps, effectively probing spin-lattice thermalization rather than lattice thermodynamics.

In driven cuprates, spontaneous ultrafast Raman thermometry has become a mode-resolved probe of nonequilibrium superconductivity. In YBa$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$6Cu$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$7O$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$8, the equilibrium anti-Stokes/Stokes ratio of the 15 THz A$$\frac{I_{AS}(-\Omega)}{I_S(\Omega)}=e^{-\hbar\Omega/k_B T}.$$9 mode is about 0.2 at about 295 K, and the transient Raman analysis gives a peak phonon temperature of about 370 K at room-temperature base conditions before deconvolution, corresponding to a $_2$0 K rise and to about 80–85 K after temporal deconvolution (Chou et al., 2023). In Bi$_2$1Sr$_2$2CaCu$_2$3O$_2$4, the threshold fluence for superconducting destruction is $_2$5J/cm$_2$6, and the antinodal quasiparticle temperature rises above $_2$7 while the condensate depletion follows a delayed trajectory (Gatuingt et al., 6 Aug 2025).

In gases, the air-lasing-assisted single-shot coherent Raman method is optimized for high-speed diagnostics rather than sub-ps delay scanning. For O$_2$8 at 50 mbar, temperatures of 294 K, 373 K, 473 K, 573 K, 673 K, and 773 K were investigated, with 1000 single-shot spectra recorded at each set point (Lu et al., 2024). The reported precision remains better than 2% across the range, whereas the inaccuracy grows from about 3.2% at 294 K to about 17% at 773 K. The method therefore emphasizes high acquisition cadence and single-shot robustness over absolute accuracy at the highest temperatures.

A boundary case is provided by Raman thermometry in trapped ions. The 139 GHz UV phase-locked Raman system for $_2$9Be$_2$0 in a Penning trap uses 500 $_2$1s Raman probe pulses rather than ultrafast pulses, and the demonstrated thermometer is Doppler-width based rather than time-domain ultrafast (Mielke et al., 2021). With a measured Gaussian width

$_2$2

the inferred temperature is

$_2$3

This system is explicitly designed to support future sideband-resolved Raman thermometry and cooling. Its relevance to ultrafast Raman thermometry is comparative: it defines the opposite limit of extremely narrow linewidth, kHz-scale phase stability, and motional-mode selectivity.

6. Limitations, controversies, and emerging directions

Several limitations recur across implementations. Probe-induced heating is a major issue at low temperature: in YBa$_2$4Cu$_2$5O$_2$6, the measured baseline phonon temperature at a cryostat setting of 100 K is about 179 K because the anti-Stokes signal of the 15 THz mode is very weak and repeated probing heats the sample (Chou et al., 2023). In coherent rotational thermometry, accuracy degrades at high temperature because the experiment was performed at constant pressure, the gas density decreases as temperature rises, and the modeling neglects effects such as collisional broadening and fine structure, which becomes increasingly consequential by 773 K (Lu et al., 2024). In steady-state Raman shift thermometry, the measured quantity is an effective Raman temperature rather than the true maximum temperature, and finite optical penetration depth can bias extracted thermal conductivities (Stoib et al., 2014).

A second class of limitations concerns strong-field and resonant regimes. In graphene under ultrafast excitation, Raman line shifts and linewidths are influenced by hot-carrier broadening of the Dirac cones and by electron–phonon scattering, so equilibrium temperature coefficients cannot be used directly (Ferrante et al., 2017). In plasmonic hotspots, an explicit Maxwell–Bloch treatment predicts population inversion, gain saturation, and a switch-like transition to stimulated Raman spanning twelve orders of magnitude in conversion efficiency; under those conditions the Stokes/anti-Stokes ratio is not a Boltzmann thermometer because the populations are dynamically driven rather than thermally distributed (Scalora et al., 2013). More generally, the assumption of a time-independent Raman tensor can fail after photoexcitation, as demonstrated in silicon where a pump-induced hole density quenches the Raman tensor by about 5% (Versteeg et al., 2018).

Current extensions aim to improve both selectivity and interpretation. One proposed direction is entangled-photon ultrafast stimulated Raman spectroscopy, in which the signal depends on an entangled two-photon wavefunction $_2$7 and Hong–Ou–Mandel interference suppresses an entire dissipative Raman pathway (Fan et al., 2023). The proposal claims a time–frequency scale not attainable by classical light and direct sensitivity to selected population and coherence terms. This suggests a possible future form of Raman thermometry in which transient populations and coherences are separated more cleanly than in classical pump–probe Raman.

Taken together, these developments define ultrafast Raman thermometry not as a single technique but as a measurement class. Its unifying feature is the extraction of transient temperature-like information from Raman observables with explicit spectral selectivity. Its main conceptual challenge is equally unifying: in nonequilibrium matter, the measured “temperature” is often mode-resolved, momentum-selective, or otherwise subsystem specific, and its validity depends on whether the probed Raman response still obeys a detailed-balance or calibrated-shift relation.