Broadband cavity-enhanced ultrafast spectroscopy
Myles C. Silfies, Grzegorz Kowzan, Neomi Lewis, Thomas K. Allison
JJournal Name
Broadband cavity-enhanced ultrafast spectroscopy
Myles C. Silfies a , Grzegorz Kowzan ab , Neomi Lewis a , and Thomas K. Allison ∗ a Broadband ultrafast optical spectroscopy methods, such as transient absorption spectroscopy and 2Dspectroscopy, are widely used to study molecular dynamics. However, these techniques are typicallyrestricted to optically thick samples, such as solids and liquid solutions. In this article we discuss acavity-enhanced ultrafast transient absorption spectrometer covering almost the entire visible rangewith a detection limit of ∆ OD < × − , extending broadband all-optical ultrafast spectroscopytechniques to dilute beams of gas-phase molecules and clusters. We describe the technical innovationsbehind the spectrometer and present transient absorption data on two archetypical molecular systemsfor excited-state intramolecular proton transfer, 1’-hydroxy-2’-acetonapthone and salicylideneaniline,under jet-cooled and Ar cluster conditions. The spectra of polyatomic molecules that undergo ultrafast dy-namics are inherently broad, due both to the energy-time uncer-tainty principle and also the large number of degrees of free-dom usually involved in the dynamics. Thus, in general thespectral "blobs" observed in the so-called linear spectra of poly-atomic molecules in the visible and ultraviolet are not particu-larly informative regarding the underlying dynamics. Ultrafastspectroscopy techniques attempt to address this problem by ob-serving dynamics directly in the time domain. Put another way,by using a nonlinear spectroscopy, in which the molecule inter-acts with multiple photons, one tries to “parse the blob" into sub-components which may distinguish themselves with different ki-netics, orientational dynamics, or spectral correlations in the caseof 2D spectroscopy. However, even the parsed blob, for exam-ple broken down into constituent parts by global analysis , stilloften leaves much room for interpretation in assigning the com-ponents of ultrafast spectra and extracting the relevant physicalquantities.Spectral assignments aside, even interpreting the seeminglysimplest aspect of ultrafast spectroscopy data—extracting ki-netic time constants from the decay of signals with increasingpump/probe delay—is not simple. Although thousands of such“lifetimes" are published every year, the measurement does notactually give an excited-state lifetime. Rather, when a time-dependent excited state Ψ ( t ) is probed in a pump/probe exper-iment, what is actually recorded are projections : S f ( t ) ∝ (cid:12)(cid:12) (cid:104) Ψ f | ˆ µ | Ψ ( t ) (cid:105) (cid:12)(cid:12) (1) a Departments of Chemistry and Physics, Stony Brook University, Stony Brook, NY11790-3400, USA; E-mail: [email protected] b Institute of Physics, Faculty of Physics, Astronomy and Informatics, Nicolaus Coperni-cus University in Toru´n, ul. Grudziadzka 5, 87-100 Toru´n, Poland where the Ψ f are final states and ˆ µ is the dipole operator thatconnects the molecule to the electromagnetic field. For exam-ple, “energy windowing" effects can have a large impact on thekinetic time constants observed in time-resolved photoelectronspectroscopy experiments . Even without windowing effects,in general, any ultrafast experiment in polyatomic molecules nec-essarily projects dynamics with many active degrees of freedomonto an observable with far fewer dimensions, resulting in signif-icant information loss. The nature of the chosen projection canthen have a profound impact on interpreting the results.Ideally, to get maximum information, one would project thestate of interest Ψ ( t ) onto as many final states as possible andmake comparisons between systems prepared differently usingthe same observables. It has also recently emerged that it is criti-cal to compare the experimental signals with theory that directlysimulates the experiment by calculating the relevant observables,performing the same projections (Eq. (1)) in silico that are doneby the experiments in the lab . Comparisons with ab ini-tio theory are most robustly done for gas-phase systems, but thebulk of ultrafast spectroscopy is done in solutions, and almostalways with very different observables than gas-phase studies.The current paradigm is illustrated in figure 1. Optical meth-ods, such as transient absorption (TAS) and 2D spectroscopy, arewell-established for solution-phase work ( Ψ f = neutral states).In contrast, gas-phase experiments, particularly in the physicalchemist’s playground of molecular beams, rely almost exclusivelyon photionization methods such as time-resolved photoelectronspectroscopy (TRPES) – action spectroscopies that project thesystem onto the 1-electron continuum ( Ψ f = free electron +cation).Comparing results using these different observables can bevery difficult. A good example is the problem of internal con-version processes in nucleo-bases, responsible for the UV photo- Journal Name, [year], [vol.] , a r X i v : . [ phy s i c s . op ti c s ] F e b ell-established methods Well-established methodsLiquid microjet TRPES This Work
Molecular beam expt.Optical probesPhotoionization probes Solution
Expt.
Fig. 1 Overview of ultrafast spectroscopy methods. By measuringthe same observable as most solution-phase ultrafast spectroscopystudies, but on jet-cooled molecules and clusters, cavity-enhancedultrafast transient absorption spectroscopy (this work) establishesa link between gas-phase methods based on photoionization andsolution-phase ultrafast dynamics studies based on optical probes. protection of DNA . Researchers studying isolated gas-phasemolecules using TRPES report quite different dynamicsthan those studying solution-phase molecules with TAS or 2Dspectroscopy . Undoubtedly, the dynamics are different in solu-tion, but the very different observables and disparate data setsobtained with the different experimental techniques cloud thecomparison, to the point where the practitioners of the differentmeasurements almost form separate communities with their ownseparate review papers .In a previous article, Reber, Chen, and Allison described theextension of ultrasensitive direct absorption techniques to fem-tosecond time-resolved experiments, reporting cavity-enhancedall-optical measurements in a dilute molecular beam that are si-multaneously ultrasensitive and ultrafast. Using frequency combsand optical resonators, cavity-enhanced transient absorption (CE-TA), or pump-probe, measurements were demonstrated with atime resolution of 120 fs and a detection limit for changes in sam-ple absorbance of ∆ OD = × − , an improvement over the pre-vious state of the art by nearly four orders of magnitude. Thislarge advance in sensitivity can enable many measurements pre-viously thought impossible. However, this previous demonstra-tion operated at only one wavelength (530 nm). One wavelengthdoes not make a spectrum, and the inherently broad spectra ofchemically-relevant molecules undergoing ultrafast dynamics de-mand wide spectral coverage.In this article, we report the development of a broadbandcavity-enhanced ultrafast transient absorption spectrometer (CE-TA S ) operating across the wavelength band of 450-700 nm—abandwidth greater than 7900 cm − (240 THz) covering almostthe entire visible spectral range. To go from CE-TA to CE-TAShas involved considerable innovation, since many of the necessarycomponents did not exist prior to our work. We have previouslypublished results regarding aspects of the optical technology crit-ical to CE-TAS, namely the development of widely tunable, lownoise, high-power frequency combs and the enhancement ofthese widely-tunable combs in a femtosecond enhancement cav-ity with custom mirror coatings . Achieving reliable and repro- ducible transient absorption spectroscopy data with ∆ OD < − using this optical technology has also required significant inno-vation which we detail in this paper, where we present the firstspectroscopy results from this system. To our knowledge, this alsothe first cavity-enhanced comb spectroscopy of any kind (ultrafastor otherwise) using a widely tunable platform.This work establishes CE-TA S as a new broadly applicabletechnique for gas-phase chemical physics, and creates a missinglink between gas-phase and solution-phase studies shown in fig-ure 1. For gas-phase molecules, UV-visible CE-TAS provides an-other projection of the dynamics complimentary to gas-phase TR-PES, with a dataset that is directly comparable to solution-phasework via the common observable. Cluster studies enabled by CE-TAS also allow probing intermediate levels of solvation. We notethat others pursue a similar linking path via attempting TRPES onmolecules in solution via the liquid micro-jet approach , as alsoillustrated in figure 1. Sensitivity is also the challenge in theseexperiments to move beyond neat liquids or ultra-concentratedsolutions to pump/probe experiments on more chemically rele-vant systems .In addition to filling the gap illustrated in figure 1, we believethese broadband CE-TAS methods can also be adapted for workon solids, ultra-dilute solutions, or sparsely covered surfaces thatwould benefit from improved sensitivity. In the sections belowwe describe the many unique aspects of the spectrometer and adetailed analysis of its performance. The optical setup is illustrated in figure 2a). We derive the ini-tial frequency comb at 1064 nm from a 1550 nm Er:fiber oscil-lator (Menlo Systems Ultra-Low-Noise variant), shifted to 1064nm using dispersive-wave generation in a short highly nonlin-ear fiber . We then amplify the shifted Er:fiber comb to 10 Waverage power in a home-built large-mode-area Yb-doped pho-tonic crystal fiber amplifier previously described . The 100 MHzrepetition-rate ( f rep ) amplified pulse train from this laser is fre-quency doubled and tripled (2+1) in critically phase-matchedlithium trioborate (LBO) and beta barium borate (BBO) crys-tals, respectively. We use the third harmonic at 355 nm fromthis setup, with approximately 500 mW of average power, for thepump in the CE-TAS measurements presented here. In the presentmeasurements, working with molecules with relatively large ex-citation cross sections, we obtain sufficient signal to noise with-out employing an enhancement cavity for the pump to boost thepump power, as was done in Reber et al. , but it would be rela-tively straightforward to implement a pump enhancement cavityif even higher sensitivity were needed. We use the residual secondharmonic (4.5 W) to pump a home-built tunable synchronously-pumped optical parametric oscillator (OPO) with subsequent in-tracavity doubling for both the signal and idler . Using the 532nm pump as well to cover the gap near OPO degeneracy, this pro-vides tunable combs over the range of 420-720 nm, as describedin ref. 15.We couple the tunable combs from the OPO to a broadband Journal Name, [year], [vol.][vol.]
Fig. 1 Overview of ultrafast spectroscopy methods. By measuringthe same observable as most solution-phase ultrafast spectroscopystudies, but on jet-cooled molecules and clusters, cavity-enhancedultrafast transient absorption spectroscopy (this work) establishesa link between gas-phase methods based on photoionization andsolution-phase ultrafast dynamics studies based on optical probes. protection of DNA . Researchers studying isolated gas-phasemolecules using TRPES report quite different dynamicsthan those studying solution-phase molecules with TAS or 2Dspectroscopy . Undoubtedly, the dynamics are different in solu-tion, but the very different observables and disparate data setsobtained with the different experimental techniques cloud thecomparison, to the point where the practitioners of the differentmeasurements almost form separate communities with their ownseparate review papers .In a previous article, Reber, Chen, and Allison described theextension of ultrasensitive direct absorption techniques to fem-tosecond time-resolved experiments, reporting cavity-enhancedall-optical measurements in a dilute molecular beam that are si-multaneously ultrasensitive and ultrafast. Using frequency combsand optical resonators, cavity-enhanced transient absorption (CE-TA), or pump-probe, measurements were demonstrated with atime resolution of 120 fs and a detection limit for changes in sam-ple absorbance of ∆ OD = × − , an improvement over the pre-vious state of the art by nearly four orders of magnitude. Thislarge advance in sensitivity can enable many measurements pre-viously thought impossible. However, this previous demonstra-tion operated at only one wavelength (530 nm). One wavelengthdoes not make a spectrum, and the inherently broad spectra ofchemically-relevant molecules undergoing ultrafast dynamics de-mand wide spectral coverage.In this article, we report the development of a broadbandcavity-enhanced ultrafast transient absorption spectrometer (CE-TA S ) operating across the wavelength band of 450-700 nm—abandwidth greater than 7900 cm − (240 THz) covering almostthe entire visible spectral range. To go from CE-TA to CE-TAShas involved considerable innovation, since many of the necessarycomponents did not exist prior to our work. We have previouslypublished results regarding aspects of the optical technology crit-ical to CE-TAS, namely the development of widely tunable, lownoise, high-power frequency combs and the enhancement ofthese widely-tunable combs in a femtosecond enhancement cav-ity with custom mirror coatings . Achieving reliable and repro- ducible transient absorption spectroscopy data with ∆ OD < − using this optical technology has also required significant inno-vation which we detail in this paper, where we present the firstspectroscopy results from this system. To our knowledge, this alsothe first cavity-enhanced comb spectroscopy of any kind (ultrafastor otherwise) using a widely tunable platform.This work establishes CE-TA S as a new broadly applicabletechnique for gas-phase chemical physics, and creates a missinglink between gas-phase and solution-phase studies shown in fig-ure 1. For gas-phase molecules, UV-visible CE-TAS provides an-other projection of the dynamics complimentary to gas-phase TR-PES, with a dataset that is directly comparable to solution-phasework via the common observable. Cluster studies enabled by CE-TAS also allow probing intermediate levels of solvation. We notethat others pursue a similar linking path via attempting TRPES onmolecules in solution via the liquid micro-jet approach , as alsoillustrated in figure 1. Sensitivity is also the challenge in theseexperiments to move beyond neat liquids or ultra-concentratedsolutions to pump/probe experiments on more chemically rele-vant systems .In addition to filling the gap illustrated in figure 1, we believethese broadband CE-TAS methods can also be adapted for workon solids, ultra-dilute solutions, or sparsely covered surfaces thatwould benefit from improved sensitivity. In the sections belowwe describe the many unique aspects of the spectrometer and adetailed analysis of its performance. The optical setup is illustrated in figure 2a). We derive the ini-tial frequency comb at 1064 nm from a 1550 nm Er:fiber oscil-lator (Menlo Systems Ultra-Low-Noise variant), shifted to 1064nm using dispersive-wave generation in a short highly nonlin-ear fiber . We then amplify the shifted Er:fiber comb to 10 Waverage power in a home-built large-mode-area Yb-doped pho-tonic crystal fiber amplifier previously described . The 100 MHzrepetition-rate ( f rep ) amplified pulse train from this laser is fre-quency doubled and tripled (2+1) in critically phase-matchedlithium trioborate (LBO) and beta barium borate (BBO) crys-tals, respectively. We use the third harmonic at 355 nm fromthis setup, with approximately 500 mW of average power, for thepump in the CE-TAS measurements presented here. In the presentmeasurements, working with molecules with relatively large ex-citation cross sections, we obtain sufficient signal to noise with-out employing an enhancement cavity for the pump to boost thepump power, as was done in Reber et al. , but it would be rela-tively straightforward to implement a pump enhancement cavityif even higher sensitivity were needed. We use the residual secondharmonic (4.5 W) to pump a home-built tunable synchronously-pumped optical parametric oscillator (OPO) with subsequent in-tracavity doubling for both the signal and idler . Using the 532nm pump as well to cover the gap near OPO degeneracy, this pro-vides tunable combs over the range of 420-720 nm, as describedin ref. 15.We couple the tunable combs from the OPO to a broadband Journal Name, [year], [vol.][vol.] , ig. 2 a) Optical layout of the broadband cavity-enhanced transient absorption spectrometer. Tunable frequency combs are derived from a syn-chronously pumped optical parametric oscillator (OPO) and coupled to a 4-mirror broadband dispersion-managed enhancement cavity. The thirdharmonic of the Yb:fiber comb at 355 nm is used for molecule excitation in the current experiments. More details regarding the optical componentsare in the main text and references 15,16. b) OPO (dashed) and cavity-enhanced (solid) spectra across the OPO 450-700 nm tuning range. Broadbandspectra are assembled from pump/probe traces recorded with different OPO wavelengths. enhancement cavity with custom mirror coatings optimized tomanage group-delay dispersion (GDD) over a wide tuning range,as described in detail in ref. 16. Figure 2b) shows representativeOPO and enhanced intracavity spectra across the tuning range.The intracavity spectrum are narrower that the OPO spectra dueto residual GDD of the enhancement cavity. This sets the limitto the simultaneous intracavity bandwidth, and thus intracavitypulse duration, that can be attained irrespective of the incidentcomb bandwidth . The cavity is in a bow-tie configurationwith two plane mirrors for the input and output couplers (nomi-nal 0.3% transmission), and two high reflectors with 50 cm radiusof curvature. With most of the cavity loss coming from the inputand output couplers, the cavity is close to the impedance-matchedcondition . We calculate the beam size ( / e radius) to be w probe = 65 µ m at 532 nm using the ABCD matrix formalism, and thisonly scales weakly with probe wavelength as w probe ∝ √ λ . Thecavity has a nominal finesse ( F ) varying from 600 to 1400 acrossthe range of 450-700 nm. OPO output wavelengths outside thisrange are not used due to the limits of the cavity mirror high-reflectance band. We focus the 355 nm pump beam to a waistsize of approximately w pump =150 µ m and overlap the pump fo-cus with the enhancement cavity focus above the molecular beamsource, as illustrated in figure 2. The pump beam is chopped ata frequency between 3 and 4 kHz, well inside the enhancementcavity’s minimum linewidth of 70 kHz (above which the cavitywould low-pass filter the CE-TA signal unless higher-order modesare used ), but above the lab’s / f noise.Although the residual OPO pump (532 nm), doubled signal( s ), and doubled idler ( i ) combs follow the same optical path,there are substantial differences to the setup for using each ofthese three combs. First of all, the OPO optical-phase transfer re-lations we discovered in ref. 15 necessitate that the three differ-ent combs are frequency-locked to the enhancement cavity using three different schemes with different actuators, as detailed inref. 16. Furthermore, there are substantial differences in the rel-ative intensity noise (RIN) spectra of the intracavity light beforethe common-mode noise rejection scheme described below is ap-plied . Also, we change the mode-matching optics between theOPO and the enhancement cavity when changing between out-put combs to account for different spatial modes and divergencefrom the OPO. Despite all these differences, comparable CE-TASperformance can be obtained using all three combs as we show insection 2. The enhancement cavity is mounted on a 60 cm ×
120 cm bread-board inside a rectangular vacuum chamber. The breadboard issupported via legs that protrude through the bottom of the cham-ber via bellows down to the optical table. In this way the bread-board is isolated from vibrations of the vacuum chamber or flex-ure of the vacuum chamber upon pump out.Molecules are introduced at the common focus of the probecavity and the pump beam using a continuously-operating slitnozzle. A planar expansion, as opposed to an axisymmetric ex-pansion from a pinhole, is used to attain a higher column densityof molecules and also facilitate cluster studies . The gas loadof the continuous planar expansion is handled by a three-stagepumping system consisting of two Roots pumps (5000 m /hr and1400 m /hr) in series backed by a two-stage 100 m /hr oil-sealedrotary vane pump.To prevent cavity mirror contamination, the supersonic ex-pansion takes place in a small inner chamber inside the main vac-uum chamber, as shown in figure 3. The inner chamber is main-tained at ∼
100 mTorr via the Roots pumping system. The innerchamber is connected to the main chamber via two 3 mm holesthat allow the laser beams to pass through. We then flow Argon
Journal Name, [year], [vol.] , MT CameraLensHeatedmirror Inner chamberand nozzleLaser beams
Detail of innerchamber/nozzleassembly
Aperture To Pumps
Fig. 3
Molecular beam setup and fluorescence monitor. The fluorescencedetection scheme is described in detail in the main text. Inset showsa cutaway of the nozzle assembly. The inner chamber surrounding theheated nozzle contains the sample molecule near the pump/probe overlapregion. gas into main chamber which creates a flow of Ar into the innerchamber via these 3 mm holes. This steady purging flow pre-vents sample molecules from exiting the inner chamber. Argonis used instead of nitrogen to avoid possible artifacts due to non-resonantly excited rotational coherences . Typical argon pres-sures in the main chamber are ∼
10 mbar, which is sufficient toprevent mirror contamination, but small enough that it does notproduce enough group delay dispersion to narrow the enhancedcomb bandwidth . A small flow of oxygen is also directed ateach cavity mirror to further help mitigate hydrocarbon contami-nation.For introducing non-volatile molecules to the experiment,molecules are sublimed at temperatures up to 150 ◦ C in a cellexternal to the vacuum chamber and then entrained in a flow ofnoble carrier gas. The supersonic nozzle assembly and associatedgas feedline are also heated to prevent molecule condensation.Typical sample consumption rates are 0.5-3 g/hr.
When recording transient absorption measurements using any ofthe three combs, we couple a delayed reference pulse train to thecavity in a counter-propagating direction as shown in figure 2a).The resulting pulse sequence at the molecular beam is shown infigure 4a). The reference beam pulses arrive ∼ ( ∆ S ) at each OPO wavelength is recovered via autobalanced sub-traction (probe − reference) and lock-in detection at the pump modulation frequency, such that the CE-TAS signal is given by ∆ S ( τ ) = π F ∆ I ( τ ) − ∆ I ( τ + ns ) I probe ≡ β [ ∆ I ( τ ) − ∆ I ( τ + ns )] (2)where τ is the pump/probe delay, I probe is the intracavity lightintensity for the probe beam, the ∆ I are pump-induced changesin the intracavity light intensity, and the factor π / F is the in-verse of the cavity enhancement cavity for impedance-matchedcavity and our experimental geometry . The subtraction accom-plishes two critical tasks. First and most important is common-mode noise subtraction. The probe ( ∆ I ( τ ) ) and reference ( ∆ I ( τ + ns ) ≈ ∆ I ( ns ) ) share mostly the same noise, but have differentpump/probe delay-dependent signals due to their timing with re-spect to the pump pulse train. Since at τ + 5 ns all fast dynamicshave subsided, the subtraction retrieves the femtosecond-delaydependent signal from the noise. Figure 4b) shows the effect ofthis common-mode noise rejection scheme on the relative inten-sity noise (RIN) of the intracavity light. With autobalanced sub-traction, the noise floor of the measurement is within one orderof magnitude (20 dB in RIN) of the quantum noise limit.Second is that the ∆ I ( τ + ns ) reference signal also containsany signal due to repetitive pumping of the sample or molecu-lar excitation that lives longer than / f rep = 10 ns. Another wayto think of this is that due to the 100 MHz repetition rate, insteady state ∆ I ( τ = ns ) = ∆ I ( τ = − ns ) such that the subtrac-tion of the reference signal removes any signal due to precedingpump/probe pulse sequences. This is relevant since for a molecu-lar beam speed of 500 m/s (e.g. for an Ar supersonic expansion),each molecule sees approximately f rep × ( µ m / m/s ) = pump pulses. The problem can be exacerbated via velocity slipbetween the sample molecule and the carrier gas, and even formolecules with short-lived excited states, a ground-state bleachsignal may persist. Subtraction of any persistent signal enablesCE-TAS to work even with these complications. For most purposes ∆ S ( τ ) can be regarded as the femtosecond to picosecond compo-nent of the true TAS signal induced by a single pump pulse, simplywith a DC offset subtracted. However, one must be aware of sub-tleties. For example, since the absolute signal size is reduced viasubtraction of ∆ I ( ns ) , care must be taken in considering signalratios as discussed in section 2.The polarization of the probe light is horizontal ( p ). Thepump polarization is controlled to be either p or s (vertical)with a zero-order half wave plate to give pump/probe signalsfor both parallel ( ∆ S (cid:107) ) and perpendicular ( ∆ S ⊥ ) polarizationconditions, respectively. We construct magic-angle signals, in-sensitive to molecular orientation or rotational coherences, via ∆ S MA = ( ∆ S (cid:107) + ∆ S ⊥ ) / . Another interesting subtlety of CE-TAS is that magic-angle data cannot be recorded simply by ori-enting the pump polarization 54.7 ◦ to the probe, as is usuallydone in transient absorption spectroscopy. This is due to the factthat the non-zero angles of incidence on the enhancement cavitymirrors causes the p and s eigenmodes of the cavity to be non-degenerate. Thus light scattered into an s mode of the cavity bya magic angle pump would not be exactly on resonance, leadingto increased noise and also a different signal enhancement. Us- Journal Name, [year], [vol.][vol.]
When recording transient absorption measurements using any ofthe three combs, we couple a delayed reference pulse train to thecavity in a counter-propagating direction as shown in figure 2a).The resulting pulse sequence at the molecular beam is shown infigure 4a). The reference beam pulses arrive ∼ ( ∆ S ) at each OPO wavelength is recovered via autobalanced sub-traction (probe − reference) and lock-in detection at the pump modulation frequency, such that the CE-TAS signal is given by ∆ S ( τ ) = π F ∆ I ( τ ) − ∆ I ( τ + ns ) I probe ≡ β [ ∆ I ( τ ) − ∆ I ( τ + ns )] (2)where τ is the pump/probe delay, I probe is the intracavity lightintensity for the probe beam, the ∆ I are pump-induced changesin the intracavity light intensity, and the factor π / F is the in-verse of the cavity enhancement cavity for impedance-matchedcavity and our experimental geometry . The subtraction accom-plishes two critical tasks. First and most important is common-mode noise subtraction. The probe ( ∆ I ( τ ) ) and reference ( ∆ I ( τ + ns ) ≈ ∆ I ( ns ) ) share mostly the same noise, but have differentpump/probe delay-dependent signals due to their timing with re-spect to the pump pulse train. Since at τ + 5 ns all fast dynamicshave subsided, the subtraction retrieves the femtosecond-delaydependent signal from the noise. Figure 4b) shows the effect ofthis common-mode noise rejection scheme on the relative inten-sity noise (RIN) of the intracavity light. With autobalanced sub-traction, the noise floor of the measurement is within one orderof magnitude (20 dB in RIN) of the quantum noise limit.Second is that the ∆ I ( τ + ns ) reference signal also containsany signal due to repetitive pumping of the sample or molecu-lar excitation that lives longer than / f rep = 10 ns. Another wayto think of this is that due to the 100 MHz repetition rate, insteady state ∆ I ( τ = ns ) = ∆ I ( τ = − ns ) such that the subtrac-tion of the reference signal removes any signal due to precedingpump/probe pulse sequences. This is relevant since for a molecu-lar beam speed of 500 m/s (e.g. for an Ar supersonic expansion),each molecule sees approximately f rep × ( µ m / m/s ) = pump pulses. The problem can be exacerbated via velocity slipbetween the sample molecule and the carrier gas, and even formolecules with short-lived excited states, a ground-state bleachsignal may persist. Subtraction of any persistent signal enablesCE-TAS to work even with these complications. For most purposes ∆ S ( τ ) can be regarded as the femtosecond to picosecond compo-nent of the true TAS signal induced by a single pump pulse, simplywith a DC offset subtracted. However, one must be aware of sub-tleties. For example, since the absolute signal size is reduced viasubtraction of ∆ I ( ns ) , care must be taken in considering signalratios as discussed in section 2.The polarization of the probe light is horizontal ( p ). Thepump polarization is controlled to be either p or s (vertical)with a zero-order half wave plate to give pump/probe signalsfor both parallel ( ∆ S (cid:107) ) and perpendicular ( ∆ S ⊥ ) polarizationconditions, respectively. We construct magic-angle signals, in-sensitive to molecular orientation or rotational coherences, via ∆ S MA = ( ∆ S (cid:107) + ∆ S ⊥ ) / . Another interesting subtlety of CE-TAS is that magic-angle data cannot be recorded simply by ori-enting the pump polarization 54.7 ◦ to the probe, as is usuallydone in transient absorption spectroscopy. This is due to the factthat the non-zero angles of incidence on the enhancement cavitymirrors causes the p and s eigenmodes of the cavity to be non-degenerate. Thus light scattered into an s mode of the cavity bya magic angle pump would not be exactly on resonance, leadingto increased noise and also a different signal enhancement. Us- Journal Name, [year], [vol.][vol.] , ig. 4 a) Pulse sequence at the sample. The reference pulse records anysteady-state pump/probe signal ∆ I ( ns ) and contains nearly identicalnoise to the probe for common mode subtraction. b) Noise spectrumof the intracavity light and subtracted signal using 469 nm (2 s ) light.Rejection of common-mode noise using the autobalanced subtractionscheme allows for the ultrafast molecular signal to be detected at thepump modulation frequency of 4 kHz. Also shown is the shot-noise (orquantum noise) limit calculated from the measured photocurrent. ing only s and p pump polarizations ensures that the intracavityprobe light remains p -polarized by symmetry. The probe bandwidth of each individual CE-TA measurement de-scribe above, with the OPO output tuned to a particular wave-length, is less than 10 THz (figure 2b). We thus assemble broad-band transient absorption spectra by combining a collection ofmeasurements taken at different wavelengths. Controlling sys-tematics is then of the utmost importance to assemble reliableand reproducible CE-TAS spectra, as several parameters affectingthe signal size vary intrinsically with wavelength and can alsovary with time over the course of an experimental run.To control for cavity finesse variation, we periodically per-form in-situ cavity ring-down measurements at each wavelengthin between pump/probe delay scans. We do this by inserting anacousto-optic modulator (AOM) in the reference beam to quickly( ∼ ns) turn off the reference beam while the Pound-Drever-Hall lock between the comb and cavity is maintained using theprobe beam. To achieve 100% turn-off of the reference beam for clean ring-down signals, we use the first-order diffracted beamfrom the AOM. The AOM is driven by a 2 f rep radio-frequency sig-nal derived from the Er:fiber comb. Using an integer multiple of f rep to drive the AOM ensures that the frequency-shifted diffractedcomb is still resonant with the enhancement cavity.To control for potential variations in pump power and sam-ple molecule density at the focus, we record fluorescence fromthe pump/probe interaction region using the scheme shown infigure 3. A mirror in the supersonic expansion path reflects flu-orescence out of the chamber. The mirror is heated to preventsample molecule condensation. To eliminate scattered light back-ground, we then use an f = 10 cm lens to image the pump/probeoverlap region to an adjustable aperture which rejects light fromelsewhere. The remaining light from the pump/probe overlapregion is recorded with a photomultiplier tube (PMT) using lock-in detection at the pump modulation (chopper) frequency. Thisscheme produces a background-free fluorescence signal that isproportional to the column density of excited molecules in the fo-cal region, and this fluorescence signal can then used to normal-ize and combine pump/probe data accumulated over extendedperiods of time. For the present demonstration of the instrument, we presentresults on 1’-hydroxy-2’-acetonaphthone (HAN), and Salicyli-deneaniline (SA), two archetypal systems for excited-state in-tramolecular proton transfer (ESIPT) shown in figure 5. Thesemolecules have previously been studied using both solution-phaseTAS and gas-phase TRPES , so they serve as good sys-tems to benchmark the instrument.Figure 6a) and b) show typical pump/probe data recordedin HAN using the i and s combs respectively. Each trace isthe average of three scans. Near time zero, a large polarizationanisotropy is seen, but this rapidly decays as the many rotationalcoherences excited by the pump pulse dephase from each other in this asymmetric top molecule. Thinking about the problemclassically (which is also appropriate here given the large num-ber of rotational states involved) one can think that the pumppulse preferentially excites molecules with their transition dipoleoriented along the pump polarization, but then these moleculesfreely rotate in random directions leading to an isotropic dis-tribution. We note that if one attempts to use the CE-TAS sig-nal (equation 2) to calculate a normalized anisotropy parameter r (cid:48) ( τ ) = ( ∆ S (cid:107) − ∆ S ⊥ ) / ( ∆ S (cid:107) + ∆ S ⊥ ) = ( ∆ S (cid:107) − ∆ S ⊥ ) / ( ∆ S MA ) , this pa-rameter need not be bounded in the usual range of ( − . , . ) due to the subtraction of the long-lived TAS signal sampled by O CH OH N OH
HAN SA
Fig. 5
Molecules in the present experiments. HAN = 1’-hydroxy-2’-acetonapthone. SA = salicylideneaniline
Journal Name, [year], [vol.] , probe = 636 nm a) || polarizations polarizationsMagic angle S [ O D - ] probe = 494 nm b)0 100 200 300 400 500 600 700 Delay [ps] -5051015 c)
636 nm494 nm
Fig. 6
Example transient absorption traces for λ probe = 636 nm (a)and 494 nm(b) combs recorded from HAN excited at 355 nm. Thepositive signal in a) corresponds to excited-state absorption and the neg-ative signal in b) to stimulated emission. The parallel and perpendicularpolarization data are each the average of three scans with 1s integra-tion time per pump/probe delay. Magic angle data is constructed via ∆ S MA = ( ∆ S (cid:107) + ∆ S ⊥ ) / . c) Full 700 ps magic angle data showing thelong decay of the transient signal including single-exponential fits (dashedblack lines). the reference beam. The numerator (a simple difference) givesno artifact, but even if the anisotropy decays to zero at long de-lays, the denominator of the expression for r (cid:48) ( t ) is still reducedby β ∆ I MA , throwing off the ratio. This is particularly acute for amolecule like HAN, with a fluorescence yield of approximately 1%and a radiative decay rate of 1/(10 ns) , for which the steady-state excited state population in the focal volume can build upover multiple pump pulses. Indeed, r (cid:48) ( τ = ) for the data shownin figure 6b) is 0.6, suggesting a steady-state magic angle back-ground signal of ∆ I MA ( ns ) = . ∆ I MA ( τ = ) , which is reasonableunder our experimental conditions.Figure 6c) shows magic-angle pump/probe traces for HANover the full 700 ps delay range accessible with our delay stage.Fitting these data with a single exponential + offset gives a timeconstant of 70 ps for internal conversion in HAN, in agreementwith previous TAS measurements in cyclohexane and fluores-cence measurements in the gas phase . However, we notethat the observed time constant is quite different than the previ-ous gas-phase ultrafast spectroscopy measurement based on TR-PES , which reported 30 ps decay time constants even whenusing longer excitation wavelengths closer to the origin of the S → S transition. This shows the impact of the observable on themeasurement of the kinetic time constants discussed in the intro-duction.
450 500 550 600 650 700
Probe Wavelength [nm] T i m e R e s o l u t i on [f s ] Fig. 7
Spectrometer FWHM time resolution across the tuning rangefound by fitting the rising edge of HAN signal to an error function as-suming instrument-limited response. Error bars are from fit.
With the assumption that the enol-keto tautomerization andthe corresponding appearance of excited-state absorption andredshifted stimulated emission in HAN happen much faster thanour time resolution , we estimate the time resolution of the in-strument by fitting the rising edge of the CE-TA traces with an er-ror function. Figure 7a) shows the resulting extracted instrumentresponse FWHM as a function of wavelength. Impulse responsewidths less than 275 fs are attained across the tuning range, withsomewhat better time-resolution observed using the i comb.We now discuss the sensitivity of the instrument. There aretwo main sources of uncertainty (i.e. noise) to consider. The firstis the optical noise floor of the system (figure 4b) due to resid-ual un-subtracted noise on the intracavity light and uncorrelatedquantum noise in probe/reference detection. The second is driftsof the instrument over longer time scales required to assemblea full data set. Both can be quantified using an Allan deviationanalysis . Figure 8a) shows the Allan deviation calculatedfrom data sets where the same signals are scanned repeatedly.The intrinsic noise performance of the optical setup is cap-tured by data taken without any sample (triangles on figure8a). Without sample, the Allan deviation comes down with aslope of − / on the log-log plot which indicates white-noise-limited performance (i.e. no drift). We observe this behaviorfor as long as we have averaged for and have seen noise downto ∆ OD = . × − (off the chart) after 90 minutes of integra-tion without sample, with a corresponding normalized sensitivityof ∆ OD = × − / √ Hz. Similar results with the 2i comb (notshown) give a sensitivity of ∆ OD = × − / √ Hz. These resultsare consistent with the optical noise floor of the subtracted signalobserved in figure 4 and comparable to the single-color result ofReber et al. of ∆ OD = × − / √ Hz , despite the significantadditional complexity of the current setup, showing the intrinsicrobustness of CE-TAS method.When accumulating an actual molecular signal, the uncer-tainty in ∆ S (circles on figure 8a)) becomes dominated by driftfor long accumulation times. These data are accumulated by re-peated scanning of a pump/probe signal (parallel polarizations,469 nm, 88 points, 0.5 s/point) such that the same pump/probedelay is re-encountered every 44 seconds. In this case, the Allandeviation differs (circles in 8a) from white-noise performance and actually increases with averaging time for real accumulationtimes longer than 10 minutes. The main source of drift is varia- Journal Name, [year], [vol.][vol.]
With the assumption that the enol-keto tautomerization andthe corresponding appearance of excited-state absorption andredshifted stimulated emission in HAN happen much faster thanour time resolution , we estimate the time resolution of the in-strument by fitting the rising edge of the CE-TA traces with an er-ror function. Figure 7a) shows the resulting extracted instrumentresponse FWHM as a function of wavelength. Impulse responsewidths less than 275 fs are attained across the tuning range, withsomewhat better time-resolution observed using the i comb.We now discuss the sensitivity of the instrument. There aretwo main sources of uncertainty (i.e. noise) to consider. The firstis the optical noise floor of the system (figure 4b) due to resid-ual un-subtracted noise on the intracavity light and uncorrelatedquantum noise in probe/reference detection. The second is driftsof the instrument over longer time scales required to assemblea full data set. Both can be quantified using an Allan deviationanalysis . Figure 8a) shows the Allan deviation calculatedfrom data sets where the same signals are scanned repeatedly.The intrinsic noise performance of the optical setup is cap-tured by data taken without any sample (triangles on figure8a). Without sample, the Allan deviation comes down with aslope of − / on the log-log plot which indicates white-noise-limited performance (i.e. no drift). We observe this behaviorfor as long as we have averaged for and have seen noise downto ∆ OD = . × − (off the chart) after 90 minutes of integra-tion without sample, with a corresponding normalized sensitivityof ∆ OD = × − / √ Hz. Similar results with the 2i comb (notshown) give a sensitivity of ∆ OD = × − / √ Hz. These resultsare consistent with the optical noise floor of the subtracted signalobserved in figure 4 and comparable to the single-color result ofReber et al. of ∆ OD = × − / √ Hz , despite the significantadditional complexity of the current setup, showing the intrinsicrobustness of CE-TAS method.When accumulating an actual molecular signal, the uncer-tainty in ∆ S (circles on figure 8a)) becomes dominated by driftfor long accumulation times. These data are accumulated by re-peated scanning of a pump/probe signal (parallel polarizations,469 nm, 88 points, 0.5 s/point) such that the same pump/probedelay is re-encountered every 44 seconds. In this case, the Allandeviation differs (circles in 8a) from white-noise performance and actually increases with averaging time for real accumulationtimes longer than 10 minutes. The main source of drift is varia- Journal Name, [year], [vol.][vol.] , Integration time per point [s] -9 A ll an de v i a t i on a) Raw ScansFluorescence norm.No sample
Number of scans
Accumulation time [min]
Delay [ps] -15-10-50 S [ O D - ] b) Day 1Day 2Day 2 fluorescence scaled
Fig. 8 a) Allan deviation recorded using repetitive scans at 469 nm 0.5s of integration per pump/probe delay. Without sample (triangles), thenoise averages down with the inverse square root of the measurementtime for as long as we have recorded data, following the dashed linewith a slope of -1/2. With molecular signal, drift in the molecular col-umn density on the ∼
10 minute time scale causes the main limitation toaveraging (circles), but this drift can be remedied to some extend usingfluorescence normalization (diamonds). b) Two pump/probe traces takendays apart can largely be brought into coincidence using normalizationto the fluorescence signal. tion in the sample molecular column density, which can be miti-gated using the fluorescence monitor as we describe below.Figure 8b) demonstrates fluorescence normalization for anextreme case. The two scans, both at a probe wavelength of 455nm and at equivalent sample backing pressures, were recordedseveral days apart. On the second day, there was less HAN re-maining in the sample cell which resulted in reduced fluorescenceand CE-TA signal. Scaling the day 2 data by the ratio of fluores-cence signals brings the two TA signals back into coincidence. TheAllan deviation of CE-TA data normalized using the fluorescencemonitor is shown as diamonds on figure 8a)). With normaliza-tion, individual CE-TA pump/probe traces can be accumulatedwith a noise level of ∆ OD = × − with repetitive scans over areal accumulation time of 22 minutes. This corresponds to a S/Nof 167 for this particular data set. In figure 9a), we show a constructed magic angle transient ab-sorption map for HAN in a He-seeded supersonic expansion (0.25Bar stagnation pressure) working 3 mm from the nozzle. Thisspectrum is sampled at the same 12 discrete probe wavelengthsas in figure 7 and the same delay axis as figure 6c). For each a)0100200300 D e l a y [ p s ] -10-5051015 S [ O D - ] -10010 S [ O D - ] b) 500 550 600 650 Probe Wavelength [nm] M A N o i s e [ O D - ] c) Fig. 9 a) Magic-angle transient absorption map for jet-cooled HAN ex-cited at 355 nm constructed from 12 probe wavelengths. Stimulatedemission is observed on the blue side of the spectrum and excited stateabsorption on the red. b) Comparison of TA spectra from jet-cooledHAN (magenta) and HAN in cyclohexane (green) from reference 32 at1 ps (solid), 5 ps (long-dashed), and 50 ps (short-dashed) delays. Thesolution-phase data has been multiplied by one overall scale factor tomake the comparison. c) Noise-levels attained as a function of wave-length for this full TA-map measurement. wavelength, we take three scans for parallel and three scans forperpendicular pump/probe polarizations with an integration timeof 1 s/delay. With 260 points/scan distributed over pump/probedelays out to 700 ps, each scan then takes 260 s = 4.3 minutes.Thus, we are accumulating data for a total of 26 minutes perwavelength. The entire spectrum comprising 12 wavelengths iscollected over the course of a day. Figure 9c) shows the noiselevel for the magic angle signals, using fluorescence normaliza-tion, obtained under these practical conditions as a function ofwavelength.In figure 9b), we extract TA spectra of the molecule at 1, 5,and 50 ps delay and compare them to the TAS data reported byLochbrunner et al. for HAN in cyclohexane . The most obviousdifference between our results from the jet-cooled molecule andthe cyclohexane data is a solvatochromic blueshift of the TA databy ∼
25 nm going from cyclohexane to gas-phase, similar (but notidentical) to the 15 nm shift for fluorescence reported by Cata-lan et al. . Furthermore, the differences in the TA spectra arenot fully explained only by a solvatochromic shift. A more com-prehensive analysis of the full HAN data is beyond the scope ofthis paper but will be the subject of a future publication includingglobal analysis and comparison to ab initio theory . Journal Name, [year], [vol.] , .3 Clusters Working 3 mm from the 200 µ m slit nozzle, we can easily gener-ate clusters with sufficient column density for CE-TAS studies andwe demonstrate this here. Figures 10a) and b) shows an exampleof this for SA recorded at λ probe = 455 nm expanded in heliumand argon, respectively. Using He carrier gas, we observe therotational anisotropy to decay in ∼
10 ps, whereas for Ar carriergas, the parallel and perpendicular polarizations data do not con-verge to the same signal until ∼
50 ps. In the optimized groundstate geometry calculated by Pijeau et al. , the rotational con-stants of SA are A = 0.066 cm − , B = 0.0091 cm − , and C =0.0082 cm − , making the molecule nearly a symmetric top. Fora symmetric top, the width of the rotational anisotropy transientscales as / √ BT , indicating a large change in the rotationalconstant is required to explain the 5x increase in the width of therotational anisotropy transient. It is important to note that forthe case of SA, with much faster internal conversion than HANand much smaller fluorescence yield of ∼ − , the rotationalanisotropy parameter r (cid:48) ( t ) constructed from the CE-TAS signals ∆ S is free from the aforementioned complications due to referencesubtraction and is bounded by ( − . , . ) . Assuming the tempera-ture is similar in the two expansions, from the increased width ofthe rotational anisotropy, we estimate that the SA molecules havegained on average 24 Ar atoms, although actually this numbershould be taken as a lower bound since excitation of the moleculemay promptly evaporate many Ar atoms, as commonly exploitedin tag-loss spectroscopy .The effect of Ar clustering on the internal dynamics of themolecule can be seen in the magic angle data shown in figure10c). For He expansions, where no clustering is expected, weobserve fast decays of the TA signal in agreement with previoussolution-phase TAS and gas-phase TRPES . However whenforming large Ar clusters, the internal conversion is shut off andthe excitation is long-lived, as shown in figure 10. A correspond-ing large increase in the fluorescence signal is also observed, fur-ther supporting a suppression of internal conversion pathways inthe Ar cluster. Also shown for direct comparison is TRPES datafrom Sekikawa et al. , which shows a much faster decay of theobservable than TAS, similar to what we have also observed inHAN. More detailed analysis of the SA data, along with full spec-tra are the subject of a forthcoming paper . In this article, we have described the performance of a broad-band ultrasensitive spectrometer for recording transient absorp-tion spectra with ultrafast time resolution. The overall perfor-mance of the spectrometer is comparable to a previous 1-colordemonstration of the main concept in molecular I , despitethe significant additional complexity of both the optical setupand molecular beam system necessary to go past demonstrationsand record data on chemically relevant systems. We have alsodemonstrated the linkages shown in figure 1 by directly compar-ing cavity-enhanced transient absorption data to solution-phaseTA measurements and gas-phase TRPES for two example systems.We expect a wealth of information can be extracted from suchcomparisons going forward, given the large body of high-quality ParallelPerpendicular
Delay [ps] S [ a . u .] a) 0 2 4 6 8 10 Delay [ps] S [ a . u .] c) Fig. 10
Parallel and perpendicular polarization CE-TAS data for SAexcited at 355 nm in a 0.25 bar He expansion a) and 2.2 bar Ar expansionb). The polarization anisotropy transient decays much more slowly in theAr data, indicating the formation of large Ar clusters. c) Magic angledata for the Ar expansion and He expansion compared to previous TRPESdata recorded in jet-cooled SA from ref. 35. data existing from these well-established techniques. Further-more, performing CE-TAS measurement on clusters can enablea detailed microscopic understanding of the effect of the solventon molecular dynamics, as has been done for linear spectroscopy.For electronically excited molecules, UV-VIS CE-TAS offers acomplimentary ultrafast observable to those provided by well-established TRPES methods, with the idea that via comparisonto theory more information can be extracted from the combina-tion than can be had from either observable alone. This multi-observable approach has recently been promoted by others forthe combination of diffraction and spectroscopy data .The methods described here can also be implemented in themid-infrared to study purely vibrational dynamics on the elec-tronic ground state, and we are actively working on develop-ing cavity-enhanced two-dimensional infrared spectroscopy (CE-2DIR) . It is important to note that in contrast to the currentwork, which provides a complimentary view of the dynamics ofgas-phase molecules after electronic excitation, for which otheraction-based spectroscopy methods exist, an action-based analogof 2DIR with ultrafast time resolution does not currently exist. Inmany ways, we expect CE-2DIR spectroscopy to be less techni-cally challenging than the current demonstration due to the re-duced bandwidth requirements of 2DIR and also less difficultieswith mirror contamination due to the absence of UV light crack-ing residual hydrocarbons in the vacuum system.Finally, we note that the methods demonstrated here can beadapted to solids, liquids, and sparsely covered surfaces, as hasbeen done for cavity-enhanced linear spectroscopy . For exam- Journal Name, [year], [vol.][vol.]
Parallel and perpendicular polarization CE-TAS data for SAexcited at 355 nm in a 0.25 bar He expansion a) and 2.2 bar Ar expansionb). The polarization anisotropy transient decays much more slowly in theAr data, indicating the formation of large Ar clusters. c) Magic angledata for the Ar expansion and He expansion compared to previous TRPESdata recorded in jet-cooled SA from ref. 35. data existing from these well-established techniques. Further-more, performing CE-TAS measurement on clusters can enablea detailed microscopic understanding of the effect of the solventon molecular dynamics, as has been done for linear spectroscopy.For electronically excited molecules, UV-VIS CE-TAS offers acomplimentary ultrafast observable to those provided by well-established TRPES methods, with the idea that via comparisonto theory more information can be extracted from the combina-tion than can be had from either observable alone. This multi-observable approach has recently been promoted by others forthe combination of diffraction and spectroscopy data .The methods described here can also be implemented in themid-infrared to study purely vibrational dynamics on the elec-tronic ground state, and we are actively working on develop-ing cavity-enhanced two-dimensional infrared spectroscopy (CE-2DIR) . It is important to note that in contrast to the currentwork, which provides a complimentary view of the dynamics ofgas-phase molecules after electronic excitation, for which otheraction-based spectroscopy methods exist, an action-based analogof 2DIR with ultrafast time resolution does not currently exist. Inmany ways, we expect CE-2DIR spectroscopy to be less techni-cally challenging than the current demonstration due to the re-duced bandwidth requirements of 2DIR and also less difficultieswith mirror contamination due to the absence of UV light crack-ing residual hydrocarbons in the vacuum system.Finally, we note that the methods demonstrated here can beadapted to solids, liquids, and sparsely covered surfaces, as hasbeen done for cavity-enhanced linear spectroscopy . For exam- Journal Name, [year], [vol.][vol.] , le, inclusion of a reflection off a glass/liquid interface into thecavity could be used to perform cavity-enhanced ultrafast atten-uated total reflectance spectroscopy on molecules at the inter-face. Translating the current sensitivity to a molecular film indi-cates that coverages below 10 − monolayer could be investigated.The challenges in adapting CE-TAS to condensed-phase contextsare 1) managing the dispersion and loss of additional intracav-ity elements and 2) managing sample excitation and refresh rate.While it is likely that compromises regarding 1) and 2) wouldreduce performance, CE-TAS methods could still find applicabil-ity for small-signal condensed phase measurements inaccessiblewith other techniques. This work was supported by the U. S. National Science Foun-dation under award number 1708743 and the U. S. Air ForceOffice of Scientific Research under grant number FA9550-20-1-0259. M. C. Silfies acknowledges support from the GAANN pro-gram of the U. S. Dept. of Education. G. Kowzan acknowl-edges support from the National Science Centre, Poland schol-arship 2017/24/T/ST2/00242. The authors thank S.A. Diddams,H. Timmers, A. Kowligy, N. Nader, and G. Ycas for assistance withdeveloping the dispersive-wave-shifted Er:fiber comb.
Notes and references
Biochimica et Biophysica Acta (BBA) - Bioenergetics , 2004, , 82 – 104.2 H. R. Hudock, B. G. Levine, A. L. Thompson, H. Satzger,D. Townsend, N. Gador, S. Ullrich, A. Stolow and T. J.Martínez,
The Journal of Physical Chemistry A , 2007, ,8500–8508.3 H. Tao, T. K. Allison, T. W. Wright, A. M. Stooke, C. Khurmi,J. van Tilborg, Y. Liu, R. W. Falcone, A. Belkacem andT. J. Martinez,
The Journal of Chemical Physics , 2011, ,244306.4 S. Adachi, T. Schatteburg, A. Humeniuk, R. Mitri´c andT. Suzuki,
Phys. Chem. Chem. Phys. , 2019, , 13902–13905.5 Y. Liu, S. L. Horton, J. Yang, J. P. F. Nunes, X. Shen, T. J. A.Wolf, R. Forbes, C. Cheng, B. Moore, M. Centurion, K. Hegazy,R. Li, M.-F. Lin, A. Stolow, P. Hockett, T. Rozgonyi, P. Mar-quetand, X. Wang and T. Weinacht, Phys. Rev. X , 2020, ,021016.6 T. K. Allison, H. Tao, W. J. Glover, T. W. Wright, A. M. Stooke,C. Khurmi, J. van Tilborg, Y. Liu, R. W. Falcone, T. J. Martinezand A. Belkacem, The Journal of Chemical Physics , 2012, ,124317.7 M. S. Schuurman and A. Stolow,
Annual Review of PhysicalChemistry , 2018, , 427–450.8 A. Stolow, A. E. Bragg and D. M. Neumark, Chemical Reviews ,2004, , 1719–1758.9 C. T. Middleton, K. de La Harpe, C. Su, Y. K. Law, C. E. Crespo-Hernández and B. Kohler,
Annual Review of Physical Chem-istry , 2009, , 217–239.10 H. Saigusa, Journal of Photochemistry and Photobiology C: Photochemistry Reviews , 2006, , 197 – 210.11 S. Ullrich, T. Schultz, M. Z. Zgierski and A. Stolow, Phys.Chem. Chem. Phys. , 2004, , 2796–2801.12 C.-h. Tseng, P. Sándor, M. Kotur, T. C. Weinacht and S. Mat-sika, The Journal of Physical Chemistry A , 2012, , 2654–2661.13 M. A. R. Reber, Y. Chen and T. K. Allison,
Optica , 2016, ,311–317.14 C. Schriever, S. Lochbrunner, E. Riedle and D. J. Nesbitt, Rev.Sci. Inst. , 2008, , 013107.15 Y. Chen, M. C. Silfies, G. Kowzan, J. M. Bautista and T. K.Allison, Applied Physics B , 2019, , 81.16 M. C. Silfies, G. Kowzan, Y. Chen, N. Lewis, R. Hou, R. Baehre,T. Gross and T. K. Allison,
Opt. Lett. , 2020, , 2123–2126.17 T. Suzuki, The Journal of Chemical Physics , 2019, ,090901.18 D. L. Maser, G. Ycas, W. I. Depetri, F. C. Cruz and S. A. Did-dams,
Applied Physics B , 2017, , 142.19 X. Li, M. A. R. Reber, C. Corder, Y. Chen, P. Zhao and T. K.Allison,
Review of Scientific Instruments , 2016, , 093114.20 R. J. Jones and J. Ye, Opt. Lett. , 2004, , 2812–2814.21 R. J. Jones and J. Ye, Opt. Lett. , 2002, , 1848–1850.22 W. Nagourney, Quantum Electronics for Atomic Physics , OxfordUniversity Press, 2010.23 A. Siegman,
Lasers , University Science Books, 1986.24 T. K. Allison,
Journal of Physics B: Atomic, Molecular and Op-tical Physics , 2017, , 044004.25 D. Miller, in Atomic and Molecular Beam Methods , ed.G. Scoles, Oxford University Press, 1988.26 W. Jarzeba, V. V. Matylitsky, A. Weichert and C. Riehn,
Phys.Chem. Chem. Phys. , 2002, , 451–454.27 I. Pupeza, X. Gu, E. Fill, T. Eidam, J. Limpert, A. Tünnermann,F. Krausz and T. Udem, Opt. Express , 2010, , 26184–26195.28 P. C. D. Hobbs, Appl. Opt. , 1997, , 903–920.29 Cavity Enhanced Spectroscopy and Sensing , ed. G. Gagliardiand H.-P. Loock, Springer, 2013.30 A. Weiner,
Ultrafast Optics , Wiley, 2009.31 P. M. Felker and A. H. Zewail,
The Journal of Chemical Physics ,1987, , 2460–2482.32 S. Lochbrunner, A. Szeghalmi, K. Stock and M. Schmitt, TheJournal of Chemical Physics , 2005, , 244315.33 M. Ziółek, J. Kubicki, A. Maciejewski, R. Naskrecki andA. Grabowska,
Phys. Chem. Chem. Phys. , 2004, , 4682–4689.34 S. Lochbrunner, T. Schultz, M. Schmitt, J. P. Shaffer, M. Z.Zgierski and A. Stolow, The Journal of Chemical Physics , 2001, , 2519–2522.35 T. Sekikawa, O. Schalk, G. Wu, A. E. Boguslavskiy andA. Stolow,
The Journal of Physical Chemistry A , 2013, ,2971–2979.36 J. Catalan and J. C. del Valle,
Journal of the American ChemicalSociety , 1993, , 4321–4325.37 A. Douhal, F. Lahmani and A. H. Zewail,
Chemical Physics ,1996, , 477–498.38 D. W. Allan,
Proceedings of the IEEE , 1966, , 221–230. Journal Name, [year], [vol.] , in preparation , 2021.40 S. Pijeau, D. Foster and E. G. Hohenstein, The Journal of Phys-ical Chemistry A , 2018, , 5555–5562.41 P. M. Felker, J. S. Baskin and A. H. Zewail,
The Journal ofPhysical Chemistry , 1986, , 724–728. 42 N. Heine and K. R. Asmis, International Reviews in PhysicalChemistry , 2015, , 1–34.43 M. Silfies and et al., in preparation , 2021.44 C. Vallance and C. M. Rushworth, in Cavity-Enhanced Spec-troscopy and Sensing , Springer-Verlag, 2014.
10 | 1–10
Journal Name, [year], [vol.][vol.]