High-pressure time-resolved spectroscopy has gained widespread applications in various fields, including materials science, chemistry, and geophysics1–8. It is an essential tool for studying dynamic processes occurring in materials under high-pressure conditions, such as those encountered deep within the Earth’s interior or during shock compression experiments. Particularly, high-pressure time-resolved infrared spectroscopy is a powerful method for investigating the kinetics and mechanisms of chemical reactions, phase transitions, and structural transformations under extreme pressure9. Its capability to capture ultrafast molecular dynamics provides valuable insights into energy flow, structural changes, and intermolecular interactions.

Despite its significance, there have been only a few prior studies of time-resolved infrared spectroscopy under high pressure. In 2009, nanosecond time-resolved transient infrared spectroscopy in the mid-infrared region was used to investigate the photolysis of TATB under high pressure9. In 2014, sub-picosecond time-resolved two-dimensional infrared spectroscopy in the spectral range of 2200-2700 cm-1 with hole-burning methods was employed to study water dynamics under high pressure10. In 2018, sub-picosecond time-resolved optical pump infrared probe measurement was conducted to study the phase transition of VO2 probed with 10-μm infrared pulses under high pressure11. In 2021, sub-picosecond time-resolved THz spectroscopy was used to study the dynamics of photo-induced carriers in GaAs under high pressure12. However, these infrared experiments were constrained by the narrow bandwidth of the infrared pulses used, allowing only single bands of vibrations to be probed. This limitation precluded a global view of dynamics.

Furthermore, due to a shortage of sensitive detectors currently available, there are constraints in conducting time-resolved spectral measurements in the far-infrared spectral region (~50–600 cm⁻¹, ~1.5–18 THz) 13. This spectral region holds unique importance in the fields of the material science of physics14, chemistry15 and biology16, given the presence of various excitations such as molecular vibrations, molecular rotations, phonons, carriers, Cooper pairs, excitons, spin-orbit coupling resonating within this energy range17. For instance, lattice modes and doorway modes of energetic molecular crystals play crucial roles in the shock detonation initiation of explosives, located in the 0-200 cm-1 and 200-800 cm-1 range, respectively 18,19. Changes in the population distribution of these modes during shock-induced detonation are expected to lead to time-dependent changes in the far-infrared vibrational spectrum. While conventional two-dimensional time-resolved terahertz spectroscopy (2D TRTS) can be employed for acquiring time-resolved spectra in far-infrared 14, it is time-consuming, necessitating the collection of hundreds of data points along the THz detection delay for each pump-probe delay20–22. Additionally, due to the absorption from phonon modes of electrooptic crystals, real-time detection of the full-spectrum far-infrared spectrum poses a challenge in single-shot terahertz time-domain spectroscopy (THz TDS) using electro-optics sampling (EOS) detection, with most of work being done between 0.5 and 3 THz (between 20 and 100 cm-1)14.

In this paper, we present the first high-pressure ultrafast time-resolved full-spectrum far-infrared spectroscopy spanning from <50 to >1800 cm-1. The generation of the air-plasmon-based infrared continuum is effectively achieved by utilizing two-color laser filaments with a commercial femtosecond laser. This ultrabroadband continuum serves as the probe light source, providing sub-cycle coherent infrared pulses that benefit time-resolved pump-probe measurements23. In contrast to previous works, where the mid-infrared continua were obtained 24–27, the spectrum of the infrared continuum in this work is extended to the far-infrared using 100-fs laser pulses. A single-shot nonlinear upconversion detection scheme28 is employed to detect the entire spectrum of the infrared continuum in real-time. To avoid light absorption by the nonlinear crystal, air is selected as the nonlinear medium. Given the isotropy of air, the third-order nonlinear optical process is applied in this method 29. This process results in the upconversion of infrared light to visible light, capturing the complete spectrum of the infrared continuum by simply shifting the spectrum of the upconverted visible light, recorded using a high-performance visible dispersive spectrometer. Consequently, nonlinear spectral upconversion is achieved in a single shot, allowing direct detection across the entire source spectral range in the frequency domain without the need for Fourier transformation as required in conventional THz TDS. Compared to commercial FTIR spectrometers and infrared dispersive spectrometers with an MCT array detector, this method not only enables simultaneous detection of both mid-infrared and far-infrared without the need for additional mid- and far-infrared detectors, but also allows for measurement of the entire infrared spectrum in a single shot. To showcase the advantages of this system, we present the full transient far-infrared absorption spectrum of the complete set of vibrational fingerprint modes in the energetic molecular microcrystal of HMX, a typical energetic molecular material, following excitation of the stretching vibration of nitro groups under the conditions of greatest interest: the regime of relatively weak shock waves (pressure, p = 1-10 GPa) characteristic of accidents30.

The schematic of the experimental setup is shown in Fig. 1. A commercial Ti:sapphire amplifier (800nm, 1kHz) is employed to provide 100-fs pulses with an energy of 6mJ, which is subsequently split to three beams with an energy ratio of 1:1:4. The first beam is used to induce the infrared continuum generation by driving the air plasmon. The nonlinear crystal BBO is used for second harmonic generation (SHG) of the fundamental laser, achieving a 5% conversion efficiency. A collinear configuration, where the fundamental and SHG beams are not separated, is used to stabilize the output intensity. The time dispersion of the fundamental and SHG pulses is compensated by carefully rotating a phase retarder. A 3-mm length plasma is generated by focusing the pulses of the fundamental and SHG laser in air. A conical infrared continuum emission is produced through the light-matter interaction of the two-color laser field and air plasma31. A pulse energy of 50 nJ is measured using a pyroelectric joulemeter (J-10mb-e, Coherent) behind a silicon window, which filters the continuum from the residual light. The silicon window reflects half of the energy, indicating that the total energy of the produced infrared continuum is approximately 100 nJ. The continuum is focused onto a DAC with a culet size of 600μm by a parabolic mirror, with a focal spot size of 300 μm. The continuum passing through the DAC is then collimated by a parabolic mirror, and its spectrum is detected by the second beam.

The second beam, used as detection light, is initially stretched to generate chirped pulses with a time duration of 10ps using a grating stretcher, ensuring high detection spectral resolution. Subsequently, it is focused, by a lens to the air in the detection cell through the 3-mm hole in the center of a parabolic mirror. The infrared continuum is focused to overlap the focal spot of the chirped beam by the parabolic mirror. The field of the infrared continuum induces a four-wave different frequency (FWDFG) process, generating a visible mixture light with a pulse energy of several picojoules. The frequency of the mixture light is expressed as \({\omega }_{vis}=2{\omega }_{nir}-{\omega }_{con}\), with *ω**nir* and *ω**con* denoting frequencies of the chirped light and the infrared continuum, respectively. The spectrum of the visible mixture light is recorded using an optical spectrometer and a commercial EMCCD detector after filtering out the residual chirped light with a broadband short-pass filter. Then, the entire spectrum of the infrared continuum can be obtained by simply shifting the spectrum of the visible light. The delay line τ1 is used to find out the temporal overlap of the chirped pulses and continuum pulses, and it doesn’t need to scan during detection, which means that the spectrum of the infrared continuum can be obtained by a single pulse. According to the four-wave mixing generation theory29, the derived FWDFG efficiency *F**2* is governed by an integration given as

$$\begin{array}{c}{F}_{2}={\left|{\int }_{-l}^{l}dx\frac{\text{exp}\left(-i\varDelta kx\right)}{1+{\left(2x/k{}_{0}^{2}\right)}^{2}}\right|}^{2} \left(1\right)\end{array}$$

where 2*l* is the length of the FWDFG region, *k* and *ζ*0 is the wave vector and beam-waist radius of the generated beam, respectively, and Δ*k* ~ 0 is the wave-vector mismatch in air. In accordance with Eq. (1), it is evident that efficiency is only dependent upon geometric parameters and remains independent on the frequency and intensity of infrared light. This characteristic implies the potential of FWDFG for detecting transient changes in infrared light intensity for the entire infrared range.

It is noticed that, given the utilization of chirped pulses for upconverting the infrared, consideration must be given to the cross-phase distortions within the spectrum of the visible mixture light. These cross-phase modulations are addressed using algorithms as detailed in literatures 32–34. Briefly, the compensation involves adjusting the phase of the Fourier transform of the measured spectrum through the time-dependent phase of the chirped pulse 33 given as

$$\begin{array}{c}\left(t\right)=\frac{{\omega }^{\left(1\right)}{t}^{2}}{2} \left(2\right)\end{array}$$

where, the chirped rate parameter *ω**(1)* is 4 rad/ps2 in this work, corresponding to a second-order spectral-phase parameter of 0.25 ps2.

The third beam with a pulse energy of 4mJ is used to pump an optical parameter amplifier, the outputs of which are used to generate excitation pulses. The excitation pulses are used to excite the sample in transient spectrum measurement. The pump beam is focused to the DAC by a lens through the 3-mm hole in the center of PM2. The focal spot size of the pump beam is 500µm, slightly larger than that of the infrared continuum. The Delay line is used to vary the time delay between the excitation and probe pulses. The optical path for infrared pulse is purged with dry air or nitrogen gas, shown as the light grey area in Fig. 1.

First, the spectrum of the continuum light source was measured. The spectra of the infrared continuum through no sample cell and through a DAC are shown in Fig. 2(a). Here, an enhancement in the far-infrared region of the spectrum extends the infrared continuum, spanning six octaves and covering nearly the entire mid- and far- infrared range (< 30 to > 2400 cm− 1 or < 1 to > 72 THz). Compared with the previous work25, the whole spectrum in this work is red-shifted, achieved by employing a laser pulse with a longer duration because the center frequency of the infrared continuum is determined by the duration of laser pulses. According to the infrared continuum theory35, the contribution of the four-wave mixing process dominates the generation of the infrared continuum in the long-filament regime. Therefore, the electric field can be expressed as

$$\begin{array}{c}{E}_{ir}\left(\omega \right)\propto {\omega }^{2}F(E\left(t{)}^{3}\right) (3)\end{array}$$

where *F*(*E*(t)3) is the Fourier transform of *E*(t)3, *E*(*t*) denotes the electric field with the fundamental and its SHG 36. Given that:

$$\begin{array}{c}E\left(t\right)={E}_{0}\left(\sqrt{1-r}\text{exp}\left(-\frac{{t}^{2}}{{{t}_{p}}^{2}}\right)\text{cos}\left({\omega }_{0}t\right)+\sqrt{r}\text{exp}\left(-\frac{2{t}^{2}}{{{t}_{p}}^{2}}\right)\text{cos}\left(2{\omega }_{0}t\right)\right) \left(4\right)\end{array}$$

where *t**p*, r and ω0 denote the pulse duration parameter, the SHG intensity fraction, and the fundamental central frequency. Thus, *E**ir*(ω)∝ω2exp(-ω2/(4/*t**p*)2), and we expect a maximum at ω = 4/*t**p*, which is inversely proportional to the pulse duration. A longer pulse duration results in a higher contribution of low-frequency component, consistent with previous simulation work37. For 100-fs pulses with *t**p*=85 fs, the central frequency is around 250 cm− 1, which agrees well with the experimental result. The broadening of the high-frequency component of the continuum is likely a result of the spectral self-broadening of the fundamental pulses.

The low-frequency part of the continuum spectrum with a sub-mm-scale spot size is attenuated slightly by the DAC with a small clear aperture, while the high-frequency part is absorbed by the DAC due to the intrinsic absorption band of the diamond in the 2000 cm− 1 region, leading to narrowing of the measurable spectral range: from < 50 to > 1800 cm− 1. The spectrum exhibits a smooth overall profile, indicating that this source is well-suited for use as probe light. The fine structures due to the absorption of silicon window at ~ 610 cm− 1 and silica protective overcoat on mirrors at ~ 1250 cm− 138 are evidently observed. The absorption lines of water vapor in the 200 cm− 1 and 1600 cm− 1 regions have also been clearly observed in Fig. 2(b) and Fig. 2(c), in good agreement with the water vapor absorption spectrum from the HITRAN database39 broadened by a Gaussian function with a 4-cm− 1 bandwidth, which indicates that the spectral resolution of this system is around 4 cm− 1.

To accurately determine the zero delay time and instrumental response function (IRF) of this system, cross-correlation measurements using a undoped germanium (Ge) wafer was performed at ambient pressure in the DAC, as previously employed in mid-infrared range2640. The multi-photon absorption of mid-infrared excitation pulses generates the photo-induced carriers in the Ge wafer, leading to a transient absorption in the far-infrared region, as shown in Fig. 3(a). The recombination relaxation lifetime of photo-induced carriers is on the 1 ns time scale, in agreement with the experimental result ever reported 41. The broadband transient spectrum around t = 0 contributes to the Kerr effect, a nonlinear process caused by changes in refractive index due to the pump pulse propagation. The cross-correlation trace S for the Kerr effect can be fitted by convoluting a Gaussian function with a bi-exponential decay, given by

$$S=\text{exp}\left(-\frac{{\left(t-{t}_{0}\right)}^{2}}{2 {d}^{2}}\right)\otimes h\left(t-{t}_{0}\right)\left({a}_{1}\text{exp}\left(-\frac{t-{t}_{0}}{{t}_{1}}\right)+{a}_{2}\text{exp}\left(-\frac{t-{t}_{0}}{{t}_{2}}\right)+{a}_{3}\right)$$

$$={a}_{1}\left(1+\text{erf}\left(\frac{t-\left({t}_{0}+\frac{{d}^{2}}{{t}_{1}}\right)}{\sqrt{2}d}\right)\right)\text{exp}\left(-\frac{t-\left({t}_{0}+\frac{{d}^{2}}{2{t}_{1}}\right)}{{t}_{1}}\right)$$

$$+{a}_{2}\left(1+\text{erf}\left(\frac{t-\left({t}_{0}+\frac{{d}^{2}}{{t}_{2}}\right)}{\sqrt{2}d}\right)\right)\text{exp}\left(-\frac{t-\left({t}_{0}+\frac{{d}^{2}}{2{t}_{2}}\right)}{{t}_{2}}\right)$$

$$\begin{array}{c}+{a}_{3}\left(1+\text{erf}\left(\frac{t-{t}_{0}}{\sqrt{2}d}\right)\right) \left(5\right)\end{array}$$

where t0 denotes the zero delay time, d = IRF/2, h(t) is the Heaviside unit step function, and a1, a2, a3 and t1, t2 are pre-exponential factors and time constants of the exponential function. Figure 3(b) shows a representative cross-correlation trace at 700 cm− 1. As seen in Fig. 3(c), the change of t0 with frequency indicates that the relative chirp of the infrared continuum is less than 90 fs across the entire spectral range. This small chirp is attributed to the low dispersion of diamond. Figure 3(d) shows that IRF of each frequency component remains around 0.3 ps for all frequencies, mainly determined by the time duration of the pump pulse. Additionally, the signal observed in Fig. 3(a) at approximately 1280 cm− 1 is a coherent artificial signal resulting from the up-conversion signal of the scattered pump light.

Although the Ge system can serve as a good benchmark for our new technique, the carrier relaxation exhibits similar behavior at different frequencies and can be detected by conventional narrow-band infrared pump-probe spectroscopy using optical parametric amplifiers. To illustrate the capability of full-spectrum infrared spectroscopy, we conducted transient vibrational spectrum measurements of the complete set of vibrational fingerprint modes to study vibrational coupling dynamics in microcrystalline HMX under different pressures in DAC. In this case, both a broad spectral range and high spectral resolution are necessary for effectively detecting and analyzing the overall dynamics of the distinguishable vibrational modes under high pressure.

The center frequency of the excitation pulses is tuned to 1280 cm− 1 to resonate with the nitro group symmetric stretching vibration of the HMX crystal. The pressure-dependent transient spectra covering the entire fingerprint region of HMX at 400ps delay time under four pressures in the 200−1700 cm− 1 range following excitation of the stretching vibration of nitro groups are presented in Fig. 4(a). A negative transient absorption value is assigned to bleach features (resulting from stimulated emission or loss of ground state absorption), while a positive value is assigned to excited state absorption features. The transient absorbance peaks evidently exhibit a blue shift and decrease as the pressure increases, as shown in Fig. 4(a), as their steady-state absorbance peaks broaden, weaken and blue-shift with increasing pressure42. Attaining this broadband far-infrared time-resolved spectrum would be challenging and time-consuming by conventional 2D TRTS or infrared dispersive spectrometer, while here we implement the real-time full-spectrum far-infrared (or THz) transient spectrum measurement using single-shot detection by air-based upconversion method.

As regards the time evolution of these fingerprint modes, Figs. 4(b-e) display the dynamics traces of the representative doorway mode near 600 cm− 1, one of the three near-degenerate in-plane bending vibrations in the 600 to 700 cm− 1 range. These traces are fitted with a mono-exponential curve. The rising signals represent the vibrational energy transfer from the nitro stretching to the bending vibration. This process emerges after t = 0 and continues until thermal equilibrium is reached. Figure 4(f) indicates that the time constant, ~ 100 ps at ambient pressure, decreases with pressure and is reduced by a factor of 10 under 7.3 GPa. This observation reveals that the coupling is enhanced under compression, consistent with the early theory proposed by Fayer and Dlott30, which stated that the principal effect of pressure on the transfer rate was the increase in the anharmonic coupling. It is suggested that the pressure-enhanced coupling between vibrational modes and doorway modes determines the sensibility of energetics under relatively weak shock waves (pressure p = 1−10GPa).

In conclusion, we have introduced a novel high-pressure, real-time ultrafast time-resolved infrared spectroscopy capable of spanning multiple octaves (from < 50 to > 1800 cm− 1) and covering the entire far-infrared range. The far-infrared continuum is generated through two-color laser filaments using 100-fs pulses and is detected using single-shot detection with air-based up-conversion. Transient spectra in the 30–2400 cm− 1 range can be acquired with a resolution of a few cm− 1. The six-octave spanning spectrum covers the entire molecular fingerprint region (50–1500 cm− 1) and a significant portion of the functional group region (1500–4000 cm− 1). With a full-spectrum infrared source, this detection method can essentially cover the entire infrared range. Combining this spectroscopy with high-pressure DAC technology, we investigate the vibrational coupling of all fingerprint modes in microcrystalline HMX following infrared excitation under various pressures, revealing pressure-dependent coupling enhancement between the excited modes and doorway modes. This observation highlights the effectiveness of our method in capturing comprehensive vibrational changes within energetic molecular systems.

Furthermore, the ability to probe the entire infrared spectrum under high pressure in a single laser shot will facilitate the study of a broader range of phenomena beyond vibrational coupling in energetic materials. For example, it can be used to investigate energy transfer between high- and low-frequency vibrational modes in other molecular systems such as the vibrational coupling between the bending and liberating modes in water43–45 or the lattice modes of high-pressure phases of ice46. It can also be applied to study carrier or exciton relaxation with broad spectral changes and their coupling with phonons in quantum dots4,47, 2D materials48 and bulk semiconductors1,2. We believe that this novel spectroscopy approach holds great potential for future research.