Vacuum-ultraviolet frequency-modulation spectroscopy U. Hollenstein,1 H. Schmutz,1 J. A. Agner,1 M. Sommavilla,1 and F. Merkt1 Laboratorium fu¨r Physikalische Chemie, ETH Zu¨rich, 8093 Zu¨rich, Switzerland (Dated: 10 January 2017) Frequency-modulation (FM) spectroscopy has been extended to the vacuum- ultraviolet (VUV) range of the electromagnetic spectrum. Coherent VUV laser radi- 7 ation is produced by resonance-enhanced sum-frequency mixing (ν = 2ν +ν ) VUV UV 2 1 0 in Kr and Xe using two near-Fourier-transform-limited laser pulses of frequencies 2 ν and ν . Sidebands generated in the output of the second laser (ν ) using an n UV 2 2 a J electro-optical modulator operating at the frequency ν are directly transfered to mod 6 the VUV and used to record FM spectra. Demodulation is demonstrated both at ] h ν and 2ν . The main advantages of the method compared to VUV absorption mod mod p - spectroscopy is its background-free nature, the fact that its implementation using m e table-top laser equipment is straightforward and that it can be used to record VUV h c absorption spectra of cold samples in skimmed supersonic beams simultaneously with . s c laser-induced-fluorescence and photoionization spectra. To illustrate these advan- i s y tages we present VUV FM spectra of Ar, Kr, and N in selected regions between 2 h p 105000cm−1 and 122000cm−1. [ 1 PACS numbers: 32.30.Jc, 32.80.Ee, 33.20.Ni, 33.20.Sn, 42.60.Fc v 3 Keywords: VUV absorption spectroscopy; frequency modulation spectroscopy 2 2 2 0 . 1 0 7 1 : v i X r a 1 I. INTRODUCTION Frequency-modulation (FM) spectroscopy is a very sensitive and powerful method to record atomic and molecular laser spectra in the gas phase1–7. Originally, the method was introduced to exploit the advantages of narrow-band lasers for spectroscopic investi- gations in the infrared (IR) and the visible (VIS) ranges of the electromagnetic spectrum. Eyler and coworkers later applied the method in combination with nonlinear frequency- upconversion methods to extend the wavelength range to the UV down to 214.5nm and achieved an absorption sensitivity of 8 10−4 in this range.7 The high sensitivity results from · the background-free nature of the detection of the absorption signals. FM spectroscopy relies on the modulation at frequency ν of a narrow-band laser operated at frequency mod ν , resulting in an amplitude spectrum schematically depicted in Fig. 1 for the modulation L indices β = 0.5 and 1.25 relevant for the present investigation (β defines the sideband-to- carrier ratio, see Ref. 8 for additional details). Maximal sensitivity is reached when the modulation frequency is larger than the intrinsic linewidth of the transitions. This property prompted Janik et al.6 to demodulate the signals at a frequency 2ν instead of ν to mod mod improve the contrast in the case of a linewidth Γ comparable to ν . mod In this article, we present an extension of FM spectroscopy to the vacuum-ultraviolet (VUV, λ < 200nm) range of the electromagnetic spectrum and explore the sensitivity of de- modulation at both ν and 2ν . High-resolution absorption spectroscopy below 200nm mod mod is notoriously difficult. In early high-resolution spectroscopic work in the VUV, the ab- sorption spectra were recorded using high-pressure lamps in combination with large VUV monochromators9–13, enabling resolution of about 0.5cm−1 between 10eV and 20eV. By re- placingthehigh-pressurelampsbysynchrotronradiation, spectroscopicmeasurementscould be extended beyond 20eV. VUV radiation from synchrotron sources, however, also needs to be monochromatized, which limits the bandwidth of the radiation to at best 0.1cm−1.14 VUV Fourier-transform absorption spectroscopy using synchrotron radiation was recently extended to the range below 105nm and offers the multiplex advantage, but so far the best resolution achieved with this method is 0.33cm−1.15 Pulsed VUV laser systems based on four-wave mixing enable a higher resolution (better than 200MHz, see Refs. 16–19), but the large pulse-to-pulse fluctuations resulting from the nonlinearity of the VUV generation pro- cess limits the sensitivity of absorption measurements. Consequently, only few VUV-laser- 2 (a) (b) 5 5 (c) (d) 0 0 ts)−5 −5 ts) ni 5 5 ni u (e) (f) u b. b. ar 0 0 ar ( ( al al sign−55 −55 sign d d e (g) (h) e t t a a ul 0 0 ul d d o o m m de−5 −5 de 1f 5 5 2f (i) (j) 0 0 5 5 − − 10 5 0 5 10 10 5 0 5 10 − − − − (ν−νL)/νmod (ν−νL)/νmod Fig. 1. (a): Amplitude spectrum of a laser of frequency ν modulated at frequency ν with L mod a modulation index β = 0.5. The sign indicates the phase of the corresponding sideband. (c), (e), (g) and (i): Simulations of demodulated signals for Lorentzian line profiles with full width at half maximum of Γ = 1 ν , ν , 10ν and 50ν , respectively, and phase φ = 0. (b): 10 mod mod mod mod Amplitude spectrum of a laser of frequency ν modulated at frequency ν with a modulation L mod index β = 1.25. (d), (f), (h) and (j): Simulations of the demodulation signals at a phase of φ = 0 obtained at the same values of Γ. absorption spectra of atoms and molecules in supersonic beams have been reported.17,19–21 Supersonic beams enable one to cool the internal degrees of freedom of molecules and to re- duce spectral congestion and Doppler broadening, however, at the cost of a reduced column density. To improve the low sensitivity resulting from the large pulse-to-pulse fluctuations of the VUV laser radiation, Sommavilla et al.22,23 have used a dispersion grating and exploited the beam diffracted in the negative first order to normalize the VUV laser intensity pulse by pulse and were able to reliably measure relative changes of the transmission of 10−4, which is sufficient to record VUV absorption spectra of molecules in cold supersonic beams. Although (1+1(cid:48)) resonance-enhanced multiphoton-ionization (REMPI) spectroscopy can 3 Fig. 2. Schematic representation of the experimental setup. Upper part: optical setup of the VUV absorption experiment. Lower part: modulation/demodulation components (see text for details). be used to record VUV spectra with high sensitivity, the line intensities may be reduced, compared to an absorption spectrum, by predissociation. Here, we present an alternative methodtorecordVUVabsorptionspectrawithhighsensitivitythatreliesonFMtechniques. TunableVUVradiationisproducedfromtheoutputoftwoFourier-transform-limitedpulsed lasers (pulse length 5ns, obtained by pulse amplification of cw single-mode ring laser ra- diation) by two-color resonance-enhanced four-wave mixing (ν˜ = 2ν˜ + ν˜ ) in Kr or VUV UV 2 Xe. The modulation of the VUV laser frequency is achieved by generating sidebands on the output of the second laser (ν˜ ) using an electro-optical modulator. These sidebands 2 are automatically transferred to the VUV, because the four-wave-mixing process is linearly dependent on the intensity of the second laser. II. EXPERIMENTAL SETUP AND PROCEDURE The experimental setup used in this work is depicted schematically in Fig. 2 and consists of a laser system (Section IIA) and vacuum chambers where the absorption experiments are carried out on cold gaseous samples in skimmed supersonic beams. Critical components for the success of the experiment such as the (de)modulation setup and the home-built VUV photodetector are described in Sections IIB and IIC, respectively. Section IID presents the different detection schemes used to characterize the VUV absorption and Section IIE 4 provides details of the analysis of the lineshapes. A. VUV Laser System The near-Fourier-transform-limited VUV laser source18,22 used in the present work is depicted schematically in the upper part of Fig. 2. VUV radiation at the wave number 2ν˜ +ν˜ is produced in a resonance-enhanced four-wave mixing process in atomic krypton UV 2 and xenon using the two-photon resonances Kr (4p)5 5p[1/2] Kr (4p)6 1S at 2ν˜ = 0 0 UV ← 94092.863cm−1 (Ref. 24) and Xe (5p)5 6p(cid:48)[1/2] Xe (5p)6 1S at 2ν˜ = 89860.015cm−1 0 0 UV ← (Ref. 25). Fourier-transform-limited pulses at the wave numbers ν˜ = 1ν˜ and ν˜ are 1 3 UV 2 generated by pulsed amplification of the continuous-wave single-mode output of two ring dye lasers (Coherent 699-21 and 899-29, output power of about 500mW) using dye cells pumped bythesecondharmonicofapulsed,injection-seededNd:YAGlaser(Spectra-Physics,Quanta Ray Lab 170, pulse length of about 8ns, repetition rate 25Hz). To generate the desired wave number ν˜ , the amplified laser radiation is up-converted to ν˜ = 3ν˜ using two successive UV UV 1 β-barium-borate crystals (BBO), see Ref. 18 for details. The beam with wave number ν˜ is 2 the pulsed-amplified fundamental output of the second ring dye laser or its second harmonic. As long as the modulation index of the fundamental output remains below 0.5, the doubled output is characterized by approximately the same modulation index as the fundamental because the sidebands are too weak to be efficiently frequency doubled. The two two-photon resonances listed above enable the generation of VUV radiation over a broad spectral range of about 60000cm−1 to 135000cm−1. The sum-frequency VUV laser radiation with wave number ν˜ = 2ν˜ +ν˜ is separated VUV UV 2 from the fundamental laser beams and beams generated through other nonlinear processes in a vacuum monochromator. The separation is achieved with a toroidal grating, which also refocusses the VUV radiation at the exit slit of the monochromator chamber.26 The detected VUV-laser beam intersects one or more pulsed skimmed supersonic beams of the sample gas atrightanglesintheabsorptionchambers. TheVUV-laserpulseshaveaduration(fullwidth at half maximum) of approximately 2.5ns and a bandwidth of approximately 300MHz. The absorption chambers consist of two separate differentially pumped regions. The probe-gas beams are generated by supersonic expansion using pulsed valves (Parker, Gen- eral Valve, Series 9, valve orifice diameter 1mm) operated at a stagnation pressure of about 5 2bar. The opening time of the valves is typically 200µs. In the first region, referred to as photoionization chamber below, a single probe-gas beam is collimated at a distance of 2cm from the valve orifice by a skimmer (Beam Dynamics, orifice diameter 2mm). Photoexci- tation takes place on the axis of a linear time-of-flight photoionization mass spectrometer with which photoionization spectra can be recorded after the photoions are extracted with a pulsed electric field towards a microchannel-plate (MCP) detector. With this detector, the fluorescence induced by the VUV radiation can also be monitored. In the second region, referred to as absorption chamber below, up to ten nozzles located 2.0cm above their respec- tiveskimmers(orificediameter1mm)areused.23 ThetransmittedVUVintensityisdetected using a home-built VUV photodetector (see Section IIC). The detector signals are ampli- fied and processed using a digital oscilloscope (Teledyne LeCroy, WavePro 760Zi, 6GHz oscilloscope), transferred to a computer, and recorded as a function of the wave number ν˜ . 2 The fundamental wavenumber of the second ring dye laser is calibrated by recording a laser-induced fluorescence (LIF) spectrum of molecular iodine and the transmission signals through two ´etalons using a fraction of the cw laser output. The output of the frequency- fixed first ring dye laser (ν˜ = 1ν˜ ) is diffracted using an accousto-optical modulator 1 3 UV operated at 675MHz. In the case of sum-frequency mixing in Kr, the zero-order beam is transmitted to the amplification chain whereas the frequency of the first-order sideband is locked to the “t” hyperfine component of the B 3Π+ (v(cid:48) = 8, J(cid:48) = 129) X 1Σ+ (v(cid:48)(cid:48) = 4, 0u ← g J(cid:48)(cid:48) = 128) transition of molecular iodine at 15682.1648cm−1 (Ref. 27) so that the wave number 2ν˜ = 6ν˜ = 94092.906cm−1 ((4p)5 5p[1/2] (4p)6 1S ) is located within the UV 1 0 0 ← bandwidth of the two-photon resonance in Kr.24,28 In the case of sum-frequency mixing in Xe, the literature value of the resonance 2ν˜ = 89860.015cm−1 ((5p)5 6p(cid:48)[1/2] (5p)6 1 0 ← 1S ) is used for the calibration.29 The calibration uncertainties of the VUV wave numbers 0 are 0.015cm−1 for experiments carried out with Kr as nonlinear gas and about 0.15cm−1 for those carried out with Xe, which includes the uncertainties caused by the ac Stark shift. The probe gases (Kr: purity grade 4.0; Ar: 4.8; N : 5.0; all from Pangas) were used 2 without further purification. 6 B. Modulation/Demodulation Units To exploit the advantages of modulation techniques in experiments with nanosecond laser pulses, the modulation frequency has to be chosen carefully. It has to be higher than the inverse of the duration of a single laser pulse so that there are at least a few oscillation cycles within the typical laser-pulse duration of 2ns to 2.5ns. The argument can also be formulated in the frequency domain: The modulation frequency should exceed the bandwidth of the unmodulated pulse. The modulation frequency also has to be low enough so that the bandwidth of the VUV detector (see Section IIC) can resolve the modulation of the laser pulse resulting from absorption. We found a modulation of about 1.4GHz to be an optimal compromise and, therefore, used a resonantly driven electro-optical modulator (EOM). The fixed output of the modulation source (Anapico APSIN6010) is split into two parts using a coupler (Mini Circuit ZADC-6-2G). The main output of the coupler feeds the ref- erence arm of the demodulator used to demodulate the VUV signal, whereas the coupled output is used to drive the EOM (Qubig T1500M3-400/800). To adjust the intensity of the signal used to modulate the cw output of the second ring dye laser to a desired modulation index β, a step attenuator (Narda 4748-69) is used in combination with a power amplifier (Becker AMP 20280035). The resonance frequency of the EOM is optimized by minimiz- ing the reflected modulation intensity monitored using a circulator (RYT 300010) and a home-built diode detector and determined to be 1.3875GHz. Small variations of the EOM resonance frequency were observed and the necessary adjustments carried out daily. To demodulate the VUV signal, we investigated both the standard demodulation tech- nique at ν (referred to as 1f demodulation below) and the demodulation technique at mod 2ν (2f) used by Janik et al.6. For 2f detection the modulation signal is amplified (Tron- mod TechP23GA)andfrequencydoubled(WatkinsJohnsonFD25HC)todrivethedemodulation mixer. A small attenuator improves the impedance matching. In the case of 1f (2f) demodulation, the VUV signal is first amplified (INA 34063) and then split using a diplexer (1f: ALRCOM FC 8312A; 2f: Microwave Circuit D9002G61) into a low- and a high-frequency component. The low-frequency component below 1.05GHz (2.0GHz) is used to monitor the envelope of the VUV signal and its amplitude represents the transmission signal. The high-frequency component of the VUV signal from the diplexer 7 Fig.3. 1fdemodulatedVUVabsorptionspectraofthe(4p)5 (2P )7d[3/2] (4p)6 1S transition 3/2 1 0 ← of krypton measured at a modulation index β = 0.5 for different lengths of the phase shifter. The length of the phase shifter was increased by approximately 2cm for each successive trace. The spectra were recorded using only the skimmed supersonic beam in the first experimental chamber. in the range 1.2–5.0GHz (2.4–2.9GHz) is demodulated at 1.3875GHz (or 2.775GHz) using a double balanced mixer (Watkins Johnson WJ-M1G). The demodulated signal is then amplified (Q-Bit QBH-9-131) and recorded with the digital oscilloscope. The phase shift betweentheVUVsignalandthereferencesignalisadjustedusinga“trombone”phaseshifter (Spinner, 152254) in the reference signal path. On a strong absorption line, the position of the phase shifter for absorption (in-phase, φ = 0) and dispersion (φ = π) are determined 2 by comparison with calculated line shapes. As illustration, Fig. 3 displays FM spectra of the (4p)5 (2P ) 7d[3/2] (4p)6 1S transition of Kr recorded for different lengths of the 3/2 1 0 ← phase shifter and compares them with calculated spectra for the phases indicated in the respective panels. The lengths of the phase shifter (medium: air) needed to achieve phase shiftsof2π forthemodulationfrequency1.3875GHz(2.775GHz)is21.6cm(10.8cm), which corresponds to the wavelength in ambient air. C. VUV Detector For the experiments presented in this article, we developed a fast vacuum photodiode following principles found in the literature.30–32 Fig. 4 (a) depicts a section through the 10-mm-diameter cylindrically symmetric detector. This home-built detector consists of (i) 8 Fig. 4. (a) Section through the cylindrically symmetric home-built VUV photodetector. (b) Cor- responding electronic-circuit diagram (see text for details). a 10-mm-diameter polished copper disk serving as photocathode,33 (ii) a positively biased anode consisting of a woven stainless-steel mesh with a transmittance of 84% located at a distance of 1mm from the cathode, (iii) a coaxially tapered transmission line with a brass cone as central conductor and an aluminum ISO-KF reduction flange as an outer conductor at ground potential. At the vacuum interface, the diameter of the cathode is reduced to the dimension of the coaxial signal connector with a characteristic impedance of 50Ω.34 Fig. 4 (b) shows a schematic circuit diagram containing most critical components. The 20MΩ resistor, the 100pF capacitor and the 10µH inductor are discrete components and filter the noise from the bias power supply. The capacitance of 32pF results from the dielectric properties of the copper-cladded circuit board (Rodgers 3020) holding the anode mesh and is used as charge storage for the signal-induced current. The 50Ω resistor serves as damping element and is composed by 10 resistors of 500Ω soldered in parallel onto the circuit board. The temporal response is governed by circuit properties. The 0–100% rise time corresponds 9 Fig. 5. Time profiles of the frequency-modulated VUV laser pulse recorded under conditions where thelowerfirst-ordersidebandwasresonantwiththe(4p)5(2P )7d[3/2] (4p)61S transitionin 3/2 1 0 ← kryptonwithprobegasnozzle(a)onand(c)off,andcorrespondingpowerspectrashowingmaxima at ν and 2ν ((c) and (d)). The measurements were carried out at ν = 1.3875GHz and a mod mod mod modulationindexofβ 0.5. ThewidthsoftheobservedbandscorrespondtotheFourier-transform ≈ limit of the envelope of the 2.2-ns-long laser pulse. The band at 5GHz ν corresponds to an mod − artefact of the digital oscilloscope. to the transit time of the photoelectrons (0.1ns under typical operation conditions), which is proportional to the distance between cathode and anode and inversely proportional to the square root of the bias voltage. The signal decays exponentially with a time constant (τ = 120ps) given by the load resistance and the capacitance between cathode and anode.35 Lacking an appropriate impulse source, we could not directly measure the time resolution of the detector but instead measured the reflection coefficient up to 6GHz, which should ideally be 1.0 for all frequencies. The first deviation occurred at 5.1GHz, where it dropped to 0.5. No resonances were detected in the vicinity of the modulation frequency (ν ) and mod its second harmonic (2ν ). Fig. 5 compares a time trace of the VUV-laser intensity with mod the laser frequency adjusted such that the lower sideband is resonant with the (4p)5 (2P ) 3/2 7d [3/2] (4p)6 1S transition in atomic krypton (trace (a)) and a time trace recorded 1 0 ← withoutprobegas(trace(c)). ThestrongmodulationoftheVUVintensityresultingfromthe resonant absorption, with maxima separated by the inverse modulation frequency (0.71ns), is clearly visible in Fig. 5(a) and illustrates the high temporal resolution of our home-built VUV detector. The corresponding power spectra obtained by Fourier transformation are depicted in Fig. 5(b) and (d), respectively. The weak residual modulation signal in trace (d) 10