Field-resolved optical spectroscopy in the mid-infrared

The most common linear mid-IR spectroscopy techniques rely purely on measuring either transmitted power or spectrum of the utilized light source. This only provides information about the imaginary part of the complex refractive index of the material under test, in other words on (frequency dependent) absorption^380,381. A coherent light field, however, imprints a complex polarization on the ensemble of quantum mechanical oscillators which results in a characteristic temporal response of the sample. Field-resolved spectroscopy exploits this additional temporal information to exceed the measurement sensitivity of conventional spectroscopy methods⁶⁴. The generic setup of the technique is sketched in Fig. 5.2. The electric field of the free-induction decay of the sample is measured by EOS.

More details on the experimental technique and details of the setup can be found for instance in refs. 64, 198, 452. Field resolution may be also extracted from other pulse characterization methods, like FROG³⁸⁷, but EOS does not require a time-consuming

mid-IR pulse

IR active sample

EOS crystal

transmitted mid-IR pulse

free-induction decay

EOS signal

Fig. 5.2. Generic setup of field-resolved spectroscopy. A short mid-IR pulse passes the sample under test and is reshaped through dispersion and absorption. The free-induction decay of the vibrationally excited sample follows. It is up-converted via sum-frequency generation with an ultrashort (known) near-IR pulse. The up-converted signal generates together with the (filtered) near-IR pulse the EOS signal. By varying the delay between mid-IR and near-IR pulses, the full electric field of the free-induction decay is recovered.

100 fs pulses from oscillator after compressionFaraday rotator

1.8 –3.5 mm

3.5 –7 mm

7 –12 mm Beam splitter

Beam splitter Delay stage LMA 25

ANDiPM l/2l/2 Beam combiner

Beam combiner LGS

PPLN GaAs

ZGPDelay stage Beam splitter

Beam splitter

Beam dump AchromatChirped mirrors

Chirped mirrors

Chirped mirrors

Chirped mirrors

Chirped mirrors Beam splitter l/2 Parabola Parabola

Parabola Achromat

l/2 Polarizer

Polarizer active stabilization l/2

Achromat

to sample & EOS

Folding / ChirpedMirrors

Folding / ChirpedMirrors

Fig. 5.3. Multi-channel OPA setup for field resolved infrared spectroscopy. Oscillator, near-IR compres-sion stage, sample chamber and EOS setup are not shown. The LGS spectral broadening stage in the box with the blue dashed lines is optional and depends on the targeted extension into the mid-IR. Addi-tional feedback loops for timing control and noise reduction are not shown. They may become necessary, however.⁴⁴¹(λ/2: half-wave plate)

160 Field-resolved optical spectroscopy in the mid-infrared

(especially for high resolution) retrieval algorithm and lock-in amplification as well as balanced detection can be readily implemented.

EOS is conventionally understood as a time domain technique⁴⁵³ where an ultrashort pulse with a duration of less than half an optical cycle of the measured pulse directly sam-ples field oscillations⁴⁵². Interpreting the method, however, in frequency-domain allowed measuring much higher frequencies, and using clearly longer sampling pulse durations, re-spectively⁴⁵². In this picture, EOS can be described as a broadband 0-to-f-interferometer with a variable delay. The sum-frequency of the CEP-stabilized probe pulse and the long-wave spectral components of the sampling pulse is generated and interferes after proper polarization adjustment with the short-wave spectral components of the sampling pulse. In this way, the phase of the probe field can be directly accessed which makes the measurement technique fast, but presumably also more vulnerable to fluctuations than characterization methods that rely on measuring up-converted delay-dependent spectra (X-FROG).

In any case, the up-conversion techniques avoid the use expensive mid-IR detectors that have either to be cooled to liquid nitrogen temperatures or suffer from thermal radiation noise^386,390. The upper frequency-domain description makes EOS intuitively well suited for mid-IR sources which have been presented in section4.1. In-fact, the down-conversion process of the OPAs and the DFG setup is reversed. Thus, two conditions for EOS are automatically fulfilled, namely the use of a waveform-stabilized probe pulse and the avail-ability of a near-IR bandwidth which is at least as high as the lowest probe frequency (to get interference). The second condition is not fulfilled anymore after spectral broaden-ing of the mid-IR. For instance, the supercontinuum shown in Fig. 4.25 would require a bandwidth of about 190 THz (corresponding to a wavelength of 1.6µm). For comparison, the 7.7 fs pulse spectrum spanning from about 750 nm to 1300 nm (Fig.2.13(c)) exhibits a bandwidth of about 170 THz. This would be enough for sampling all wavelengths longer than 1.75µm, i.e. sufficient for tracing the signatures of vibrational transitions. Whereas EOS was implemented early for the DFG source presented in section 4.1.3, an imple-mentation of a field-resolved spectroscopy setup for the OPA source and in particular the mid-IR continuum, which is most attractive for molecular fingerprinting, has not yet been realized.

Fig. 5.3 presents a possible scheme for a field-resolved spectrometer. It is based on the results presented in this dissertation. The input, not shown in the figure, would be the commercial-grade KLM TD oscillator briefly introduced in section 3.2.2. The compres-sion of the full power to 30 fs would not be necessary. But using the first cascaded χ⁽²⁾ stage for reducing the pulse duration to about 100 fs would facilitate the seed generation in the normal dispersive fibers. Possibly, a larger core fiber could be used for generating the LGS OPA seed. Consequently, more seed energy would be available for parametric amplification (cf. section 3.1.1). Moreover, the continuum generation in the highly non-linear ANDi would not need a careful consideration of the pulse-to-pulse noise. An LMA fiber for seeding the PPLN OPA can, however, not be utilized due to the tailored disper-sion of the ANDi fiber (cf. Fig. 4.12(b)). Using shorter pump pulses would also increase the idler bandwidth of the PPLN OPA, possibly making the mid-IR compression stage before SCG unnecessary. The LGS OPA idler could also be broadened further to enter the wavelength range from 10µm to 20µm. Such a stage is indicated in the box with the blue dashed frames in Fig. 5.3. Whereas in the reported experiments, mid-IR pulse

compression was achieved only by exploiting material dispersion, chirped mirrors⁴⁵⁴ are utilized in the scheme of Fig. 5.3 to obtain precise phase control and better compression quality.

The near-IR pulse compression stage for obtaining sampling pulses is not included in the setup of Fig.5.3. An option was to compress the LMA-25 fiber output that is also used for seeding the LGS OPA. Pumping the fiber with 100 fs pulses would lead to a FTL of 10 fs or less (cf. Fig.3.3). After another bulk broadening stage (cf. section2.2.2), the bandwidth could be further extended by a factor of two which would be sufficient for EOS of the three mid-IR channels. The setup of Fig.5.3recycles the pump pulses after the LGS OPA to use the pump power more efficiently. By doing so, special care should be taken of the beam reshaping resulting from the high nonlinear refractive index of LGS. Pumping the PPLN OPA first would not work due to the strong pump beam distortions (cf. Fig.4.6).

The probably most critical point of the setup is to run the individual channels at the highest possible stability in order to enable ultrasensitive mid-IR detection. An active stabilization of the fiber stage was included in Fig.5.3. It is commercially available. The timing jitter between the channels might, however, be even more decisive. Active timing synchronization is an issue which has been addressed in OPCPA development⁴⁵⁵, optical waveform synthesis⁴⁵⁶, laser-microwave⁴⁵⁷ and laser-electron synchronization⁴⁵⁸, leading to jitter clearly lower than half of the mid-IR wave cycle period. In particular, the all-optical synchronization techniques may be transferred to the presented setup. A further simplification in this regard could be the usage of only a single seed source, the ANDi fiber output. This would require multiple stage amplification of the LGS OPA which might also reduce the idler power fluctuations if the OPAs are well saturated. The signal derived from the sequential OPA stages might further serve as the basis for ultrashort near-IR pulses that analyze the free induction decay in the EOS setup.

In conclusion, the essential building blocks to realize a broadband field-resolved mid-IR spectrometer for vibrational spectroscopy have been demonstrated in this dissertation, namely the generation of few-cycle pulses in the near-IR and CEP-stable continua in the mid-IR. Nevertheless, there are many technical details which still have to be clarified until such an outstanding tool becomes a working horse which is as reliable as an FTIR.

162 Field-resolved optical spectroscopy in the mid-infrared