Femtosecond photoelectron and photoion spectrometer with vacuum ultraviolet probe pulses
Markus Koch, Thomas J. A. Wolf, Jakob Grilj, Emily Sistrunk, Markus Gühr
aa r X i v : . [ phy s i c s . a t m - c l u s ] A p r Femtosecond photoelectron and photoion spectrometer with vacuum ultraviolet probepulses
Markus Koch,
1, 2, ∗ Thomas J. A. Wolf, Jakob Grilj,
1, 3
Emily Sistrunk, and Markus G¨uhr Stanford PULSE Institute, SLAC National Accelerator Laboratory, Menlo Park, CA 94025, USA Institute of Experimental Physics, Graz University of Technology, Petersgasse 16, A-8010 Graz, Austria Laboratoire de Spectroscopie Ultrarapide, Ecole Polytechnique F´ed´erale de Lausanne EPFL, 1015, Switzerland (Dated: September 5, 2018)We describe a setup to study ultrafast dynamics in gas-phase molecules using time-resolved photo-electron and photoion spectroscopy. The vacuum ultraviolet (VUV) probe pulses are generated viastrong field high-order harmonic generation from infrared femtosecond laser pulses. The band passcharacteristic in transmission of thin indium (In) metal foil is exploited to isolate the 9 th harmonicof the 800 nm fundamental (H9, 14 eV, 89 nm) from all other high harmonics. The 9 th harmonic isobtained with high conversion efficiencies and has sufficient photon energy to access the completeset of valence electron levels in most molecules. The setup also allows for direct comparison ofVUV single-photon probe with 800 nm multi-photon probe without influencing the delay of exci-tation and probe pulse or the beam geometry. We use a magnetic bottle spectrometer with highcollection efficiency for electrons, serving at the same time as a time of flight spectrometer for ions.Characterization measurements on Xe reveal the spectral width of H9 to be 190 ±
60 meV and aphoton flux of ∼ · photons/pulse after spectral filtering. As a first application, we investigatethe S excitation of perylene using time-resolved ion spectra obtained with multi-photon probingand time-resolved electron spectra from VUV single-photon probing. The time resolution extractedfrom cross-correlation measurements is 65 ±
10 fs for both probing schemes and the pulse durationof H9 is found to be 35 ± I. INTRODUCTION
Time-resolved photoelectron spectroscopy (TRPES) isa unique method to investigate photoinduced ultrafastrelaxation processes in molecules [1–4]. As in other ul-trafast methods, molecular dynamics in covalent statesare induced by the interaction with an excitation pulsetypically in the visible or ultraviolet range. The ex-cited state molecular wavepacket is projected onto dif-ferent electronic continua by an ionization pulse. Theenergy flow in a molecule after photoexcitation can befollowed in time as transient changees in the intensityand energy of photoelectron (PE) peaks. In order toaccess the excited state dynamics, the probe photon en-ergy has to exceed the excited state ionization potential.TRPES has allowed the deduction of ultrafast photoex-cited dynamics of many molecules ranging from simplediatomic systems [5] to polyenes [6, 7] and more complexclusters [4, 8, 9]. For the polyenes, TRPES has allowedthe separation of nuclear relaxation from electronic relax-ation by applying Koopmans’ single electron propensityrules [6, 7]. Probe pulses are generally obtained by crystalbased nonlinear optics, limiting the probe pulse photonenergy to < ∗ [email protected] will also result in a decay of the photoelectron signal dueto a) changing dipole matrix elements and b) further nu-clear relaxation in a different electronic state. In general,it is difficult to separate nuclear from electronic relax-ation with low photon energy probe pulses, since bothprocesses lead to similar decay patterns. In addition, thismethod cannot follow further molecular dynamics afterthe initial decay processes, since the ability for probingis lost. Strong field multi-photon ionization provides theopportunity to ionize from excited states and from theground state. However, its interpretation requires exten-sive strong field ionization calculations [14] and may befurther complicated by transient shifts of intermediateresonances.Vacuum ultraviolet (VUV) probe photons, on the otherside, have large enough photon energies to follow relax-ation of the excited non-equilibrium state all the wayto the vibrationally hot ground state. Paired with theshort pulse characteristics obtained with high-order har-monic generation (HHG), they are advantageous as probepulses in TRPES. A difficulty that arises with probing inthe vacuum ultraviolet and extreme ultaviolet range isa persistent background PE signal from the (unexcited)electronic ground state. This requires a high signal tonoise ratio to isolate excited state dynamics. We useHHG, a tabletop generation mechanism for VUV pulsesbased on strong field interaction of near infrared fem-tosecond laser pulses with noble gas atoms. At intensi-ties of ∼ W/cm part of the valence electrons areionized from the atoms and accelerated in the laser field.Upon reversal of the electric field the electrons can re-combine with the remaining ion and their kinetic energyis converted into photon energy [15–18], resulting in theemission of a comb of odd numbered harmonics of thefundamental laser frequency. Femtosecond VUV pulsescan also be obtained from free electron lasers at highfluences but also high costs [19–21].To perform PE spectroscopy, a single harmonic is usu-ally selected from the HHG spectrum by monochromiz-ing beamlines using gratings [22–27] or multilayer mir-rors [28–32]. Grating monochromators even allow abandwidth reduction below the inherent width of the re-spective harmonic at the cost of photon flux [23, 25].Pulse broadening in the time domain due to pulse fronttilting of diffracted pulses can be corrected with timecompensated dual grating setups [24] preserving the pulseduration of high-order harmonics which is much shorterthan that of the driving pulse [33]. However, these se-tups reduce the beamline transmission and increase thealignment effort of the setup. Multilayer mirrors are tai-lored to reflect a single harmonic. Other harmonics arereflected as well, albeit with reduced efficiency, due toa finite bandwidth and finite specular reflectivity of thematerials used. A combination of two multilayer mirrorsallows to reduce the reflection of undesired harmonics toas low as ∼
1% [30]. VUV pulse durations around 10 fsand below achieved with both grating [24] and multi-layer [30, 34] monochromatization have been reported.Here we present a very simple and cost efficient way toselect a single harmonic as pump or probe pulse for TR-PES with VUV light. We exploit the bandpass structurein transmission of a 150 nm thick indium (In) metal filterwhich is centered at harmonic nine (H9, 89 nm, 14 eV)of a 800 nm fundamental. In combination with a focus-ing optic (MgF protected Al, or unprotected Au mirror)the transmission contrast is 20 : 1 to higher harmonicsand >
70 : 1 to lower harmonics. We expect negligibletemporal broadening of the VUV pulse, as it is the casewith multi layer mirrors [30] or time-delay compensatedmonochromators [24]. While the photon energy of H9is sufficient to ionize many molecules from the groundstate it has also the advantage of a higher conversion ef-ficiency in the HHG process because it is closer to theperturbative region [35].VUV light was used in gas phase TRPES as a probepulse to study molecular dynamics of valence electronicstates such as the dissociation of Br upon 400 nm ex-citation to the lowest dissociative state [36–39]. TheVUV light can also be used to excite a system to Ry-dberg states which are transiently ionized by visible orUV pulses [28, 40–47]. Time-resolved dynamical studiesin cationic states, for example in ethylene, have also be-come possible with VUV pump excitation and IR [29] orVUV [34] probe.In the following we describe the pump-probe setup, inparticular how harmonic 9 is spectrally isolated from theharmonic spectrum (Experimental Setup). The probepath of our setup allows for changing from VUV tothe 800 nm fundamental without alteration of the pump-probe delay and without need for realignment, simplyby removing the In filter and deactivating HHG. This is useful for alignment purposes but, more importantly,also enables multi-photon ionization as a complemen-tary ionization mechanism. Direct comparison of VUVsingle-photon probing with multi-photon 800 nm ioniza-tion is thus possible. We use a time-of-flight (TOF) spec-trometer to record mass spectra for multi-photon probewhich is also described in the Experimental Setup. ForH9 single-photon probe we detect the electron kineticenergy and operate the TOF spectrometer as magneticbottle spectrometer to benefit from the higher collectionefficiency. We present calibration measurements with Xegas to estimate the spectral purity, width and photon fluxof H9. Finally, in the Results and Discussions section wepresent time-resolved pump-probe studies on the fluores-cence dye perylene. For 400 nm pump excitation we com-pare 800 nm multi-photon probe with H9 single-photonprobe and determine the spectrometer time resolution forboth schemes. II. EXPERIMENTAL SETUP
Fig. 1 shows a scheme of the VUV time-resolved pho-toelectron/photoion spectrometer. This beamline wasadded to a previously existing setup for time-resolvedVUV and soft x-ray (SXR) spectroscopy, which has re-cently been characterized [48]. An amplified laser sys-tem (Mantis oscillator and Legend Elite Duo amplifierfrom Coherent) generates 800 nm, 8 mJ, 25 fs pulses at a1 kHz repetition rate, up to 4.5 mJ of which are availablefor HHG to obtain VUV probe pulses. The rest of the800 nm fundamental can be frequency doubled or tripledfor UV pump excitation. A time delay between pumpand probe pulses is introduced by a translation stage inthe pump arm.
A. Optical setup
To generate high harmonics, we focus the 800 nmpulses with a f = 1680 mm lens into a gas cell (27 mmlength, 4 mm inner diameter) inside a vacuum chamber.The cell contains argon gas at a pressure of about 10 torr.The laser perforates the copper foil seals of the cell, whichare replaced every couple of days. The fundamental (p-polarized) and high-order harmonics co-propagate afterthe gas cell and are reflected from a silicon wafer (pur-chased from University Wafer) in order to reduce a majoramount of the fundamental [49, 50]. The angle of inci-dence is chosen to be 81 degrees as a compromise betweenreduction of the 800 nm light ( ∼ ∼
85% for H9), as shown in Fig. 2a.This reduction of the fundamental is crucial to preventdestruction of the metal filters that follow in the beampath.To select the 9 th harmonic (H9, 14 eV, 89 nm) we use acombination of an In filter and an Al mirror. Fig. 2 showsreflectance (a) and transmittance curves (b) of elements FIG. 1. Setup for time-resolved photoelectron/photoion spec-troscopy based on HHG. In the probe path high-order har-monics are generated in a gas cell. Reflection from a Si waferat grazing incidence removes the major part of the 800 nmfundamental and H9 is spectrally isolated by a combinationof an In filter and a MgF coated Al mirror. Multiple har-monics can be used as probe by using an Al filter and 800 nmmulti-photon probing is possible without filter and deacti-vated HHG. In the pump path the fundamental is frequencyupconverted and a pump-probe time delay is introduced. Thesample is evaporated into the interaction region by an oven. ATOF spectrometer is used to detect electrons in a magneticbottle configuration or ions by applying a positive repellervoltage. that are useful for the selection of a variety of harmonics.Additional information for all elements is listed in Tab. I.A MgF coated Al mirror with a radius of curvature of1000 mm ( f = 500 mm) focuses the diverging beam intothe ionization region of the TOF spectrometer. The Almirror can be replaced by an unprotected gold mirror forbetter reflectance at higher photon energies. Ultra-thinIn foil is characterized by a bandpass structure centeredat H9 (dashed red curve in Fig. 2b). In combinationwith the Si wafer and the Al mirror the total filter curve(solid red curve) spectrally isolates H9 with an estimatedtransmittance of 5%. The transmission contrast on theblue side to H11, H13, H15 is ∼ > O layer due to oxidation, in which case the trans-mittance is reduced by 30-60% [48]. Although not usedfor this work we prepare for the isolation of H6 at 9.3 eV.Even harmonics can be generated by using a two colorfield consisting of the fundamental and the second har-monic [51] and H6 can be isolated with a bandpass filter(A133, Fig. 2, black lines).The fundamental is blocked to 2 · − by the In filter(3 · − with the Al filter) [52]. After separation fromtheir fundamental the diverging harmonic(s) traverse thespectrometer chamber before they are focused by the Al(or Au) mirror into the ionization region of the time offlight (TOF) spectrometer (Fig. 1). To avoid backgroundsignal due to ionization by the incoming beam the ion-ization region and the TOF entrance is located slightlyabove the incoming beam. r e f l e c t an c e R −2 photon energy / eV t r an s m i tt an c e T Si R p Al−MgF AuInIn−Si−AlAlAl−Si−AlA133A133−Al H3 H5 H7 H9 H11 H13 H15 H17 H19 H21 a) b)
FIG. 2. (a) Reflectance R of the reflective elements used inthe setup. For silicon the angle of incidence is 81 degrees andthe reflectance is calculated for p-polarized ( R p ) light usingthe Fresnel equations and data from Ref. [52]. The reflectancecurves of the MgF coated Al and the unprotected Au mirrorare obtained from McPherson Inc. [53] (b) Transmittances T are shown for 150 nm thick In and Al filters (dashed lines) [54].The total transmittance function of the beam line (solid lines)is obtained as combination of each filter with the the reflec-tivity of the Si wafer and Al mirror. In addition, the trans-mittance of a bandpass filter (Acton Optics A 133) is shown.Vertical lines indicate the energies of the odd harmonics. Al metal filters are routinely used in HHG setups toeliminate the remaining fundamental radiation, as theytransmit harmonics in the 17 to 70 eV range (see Fig. 2).Sn metal filters exhibit a transmission window between16 and 24 eV which allows selection of harmonics 11 to15. This bandpass structure was exploited in one armof a split mirror interferometer to realize a femtosecondpump–probe experiment with a pair of VUV pulses (thesecond arm used a multilayer mirror) [34]. In anothersetup, a resonance energy that was included in the rela-tively broad multiband reflectivity of a multilayer mirrorwas optionally eliminated by the sharp Sn transmission optical element product coating, f = 500 mm Acton Optics & Coatings, Princeton InstrumentsAu mirror AU-B-MPC-1025-1000 f = 500 mm Lattice Electro Optics Inc.In filter 150 nm thick, Ni mesh support Lebow, LuxelAl filter 150 nm thick, Ni mesh support Lebow, LuxelA133 133-XN-1D 2.5 mm MgF substrate Acton Optics & Coatings, Princeton InstrumentsTABLE I. Optical elements used in the beamline to spectrally isolate and deliver harmonics to the electron/ion spectrometer. edge at 24 eV [28]. Semiconductor (Si) and metallic (Al,Zr) filters have also been used for amplitude and phaseshaping of high-order harmonics aiming at the pulse du-ration reduction of attosecond pulses [55].Our setup allows for optimization of the HHG conver-sion efficiency, i.e. phase matching of the HHG process,by measuring the flux of photoelectrons emitted from aAu surface with a low noise charge sensitive amplifier(SRS model SR 570) [48]. In the TRPES setup the Ausurface can be brought into the beam after the Si waferand Al filter (optional) but in front of the In filter. HHGis optimized by adjusting the focus position with respectto the cell, the cell gas pressure and the chirp of thefundamental laser pulse. However, we note that pureoptimization for maximum HHG power might result inincreased spectral width of the harmonics [56, 57]. In-stead, optimization on the width and strength of a pho-toelectron reference spectrum (e.g., that of Xe, see below)resulted in better spectral quality of the HHG source.The pump beam is branched off from the laser outputby a 20% beamsplitter (BS, Fig. 1) and traverses a de-lay stage for pump-probe delay adjustment. The 800 nmfundamental is then frequency doubled in a BBO crys-tal and the 400 nm light is separated by a dichroic beamsplitter. A halfwave plate in front of the BBO crystalallows to adjust the pump power from zero to a few hun-dred microjoules. A f = 1000 mm Al mirror is used tofocus the pump beam into the ionization region througha 3 mm thick CaF window. Spatial overlap of pump andprobe beam can be optimized with the help of a phosphorscreen that can be moved into the interaction region andviewed with a CCD camera. To find time overlap bothbeams can be guided to a photo diode. B. Focus size
We measure the excitation and probe beam diameterat the ionization region with a phosphor screen that canbe brought into the beam and viewed with a camera.The full width at half maximum (FWHM) of the beamdiameter is determined as average of two Gaussian fits tothe horizontal and vertical intensity distribution. For the800 nm fundamental we find a beam diameter of ∼ µ m.To guarantee good spatial overlap we use a bigger diam-eter of the 400 nm pump beam of 250 µ m. We determine the diameter of the harmonics using the Al filter andMgF coated Al mirror (the HHG spectrum consists pri-marily of H11 to H15, see below) and find a diameterof ∼ µ m. This diameter is the demagnified image ofthe HHG source region. With a (de-)magnification of M = − .
18 of our beamline we estimate a HHG sourcediameter of about 520 µ m, which is consistent with theharmonic generation geometry. C. The sample
A molecular beam of perylene (Sigma Aldrich, whitepowder, purity ≥ . ∼
40 mmlength the oven is able to produce a much narrower beamthan a conventional Knudsen cell. Depending on the re-quired sample density of the spectroscopic scheme theoven is heated to temperatures in the range of 100 ◦ C to170 ◦ C, corresponding to perylene vapor pressures of 10 − to 10 − torr [59]. Based on deposition measurements ona quartz micro-balance we estimate a sample density atthe exit of the capillary ranging from 10 to 10 cm − for the given temperature range [58]. As the cone angleof the molecular beam increases significantly from ∼ ◦ (FWHM) at lower vapor pressures to ∼ ◦ at higherpressures the distance between the laser and the capil-lary exit is a sensitive parameter, especially when highsample densities are required. D. Time of flight spectrometer
The 1 kHz repetition rate of the laser is ideal for atime of flight electron/ion spectrometer. With typicalflight times in the microsecond range data acquisitioncan easily be achieved. Whole traces with multiple hitscan be digitized and transferred to the computer memorybefore the following laser shot.In gas phase photoelectron experiments with limitedphoton flux, especially in pump-probe experiments, itis desirable to maximize the collection efficiency of theelectron spectrometer. While more sophisticated meth-ods such as velocity map imaging [60] or cold targetrecoil ion momentum spectroscopy [30, 61] reveal addi-tional information about the velocity vector of electronsand/or ion fragments, we are primarily interested in thekinetic energy of photoelectrons. Up to 4 π sr collectionefficiency without reduction of the energy resolution canbe achieved with a magnetic bottle electron spectrome-ter. Based on an electron time of flight spectrometer, itexploits a combination of an inhomogeneous and a ho-mogeneous magnetic field to guide photoelectrons fromthe ionization region to the detector [62–64]. We use arelatively short flight tube length of 460 mm, similar toa setup that was recently presented and well character-ized [65]. The inhomogeneous magnetic field is producedby a stack of permanent magnets mounted on a xyz ma-nipulator, providing more flexibility compared to electro-magnets. We use five sintered neodymium disk magnets(25.4 mm diameter, each 6.4 mm height, purchased fromGrainger Industrial Supplies) that are attached by theirmagnetic action to a soft iron cylinder (25.4 mm diame-ter, 42 mm height, low carbon magnetic iron, ASTM des-ignation A848-01, purchased from ESPI Metals). On theside of the magnets facing the flight tube a soft iron trun-cated cone (2.8 mm top diameter, 12.2 mm bottom diam-eter, 4.2 mm height) is attached to the magnets in orderto increase the magnetic flux density B z in the ioniza-tion region. Out of several different cone geometries thisone gave the highest B z values at close distances. Witha Hall probe we measure B z = 640 mT at the cone sur-face and B z,i = 430 mT at 2 mm distance, approximatelythe location of the laser beam. A repeller electrode (25x 25 mm , copper) is located on top of the cone whichcan be set to small negative voltages (typically ∼ − B z,f = 0 . dE/E = B z,f /B z,i ≈ .
2% [63, 64]. The field ratio also dictates the magnifi-cation of the magnetic bottle configuration [63, 64], inother words, the acceptance area (perpendicular to themagnetic bottle center line) from which electrons reachthe MCP detector. The acceptance area depends cru-cially on the distance between the laser focus and themagnet tip and we estimate it to be in the range of thebeam diameter of about 90 µ m.Signal pulses are decoupled from the conical anode ofthe MCP detector and digitized by a high-speed analog-to-digital converter card (Gage Cobra, Dynamic Sig-nals LLC, 2 GS/s repetition rate, 500 MHz bandwidth,8 bit resolution). After transfer to the computer a con-stant fraction peak-finding algorithm is used to determineflight times. E. Calibration with Xe
We use Xe to calibrate the electron kinetic energy ofthe TOF spectrometer, optimize the HHG conversion effi-ciency and determine the H9 photon flux. Figure 3 showsXe photoelectron spectra obtained with the In filter (H9)and with the Al filter. With a H9 photon energy of 14 eVand binding energies of the Xe 5p / and 5p / valenceelectrons of 13.4 eV and 12.1 eV, respectively [67], thephotoelctron peaks appear at 0.5 eV and 1.8 eV. We findthat the spectral purity of H9 strongly depends on theHHG phase matching conditions. Maximizing the over-all yield of high-order harmonics by maximizing the pho-tocurrent from a gold surface seems to result in poor spec-tral purity of H9 (lower trace in Fig. 3a). Instead of twoisolated peaks the Xe PE spectrum is composed of threeor four peaks. Spectral shifts and splitting of high-orderharmonics depending on HHG phase matching conditionshave been documented in literature [56, 57]. In contrast,optimization of the HHG parameters by monitoring theshape of the Xe PE spectrum yields the expected doublepeak structure (upper trace in Fig. 3a). A strong vari-ation of the spectral purity of harmonics 11 to 15 canbe seen in the lower image of Fig. 3b. The HHG phasematching in this case was optimized for H9 (black curve)and not changed when switching from the In to the Alfilter. A successive increasing spectral distortion of har-monics with increasing harmonic number is immediatelyevident (gray curve). We attribute the cut-off after H15to a poor reflectivity of the MgF coated Al mirror aboveH15. We applied a small repelling potential to record theXe PE which we accounted for by subtraction of 1.7 eVfrom the actually measured spectrum to match the peakpositions to the literature values.Both PE peaks in Fig. 3a have the same FWHM of220 ±
10 meV. We estimate the energy resolution of ourmagnetic bottle spectrometer to be ∆
E/E ≈ ∼ .
3% and ∼ . E = 110 ±
30 meV (averaged over the two peaks, whichappear at 2.2 eV and 3.5 eV, respectively). We thus ob-tain a spectral width of H9 of 190 ±
60 meV. no r m a li z ed c oun t s no r m a li z ed c oun t s H9 H11 H13 H15
In filterAl filter a) b)
FIG. 3. Xe reference spactra: (a) PE spectrum of Xe atomsobtained with H9 using the In filter. The upper spectrum(offset by 1) was recorded after optimization of the HHG pro-cess to obtain narrow PE lines. The two PE peaks at 0.5 eVand 1.8 eV correspond to the Xe 5p / and 5p / valence elec-trons. The solid line is a fit of two Gaussian functions to thedata points. For the lower spectrum the HHG process wasoptimized for maximum overall HHG intensity. (b) Xe PEspectrum obtained with the In filter (black curve) and withthe Al filter (gray curve). The HHG phase matching wasoptimized in both cases with the In filter. We estimate the photon flux of H9 based on PEcount rates obtained from Xe background gas. For afundamental pulse energy of 3.5 mJ we obtain ∼ · photons/pulse in the TOF ionization region, corre-sponding to ∼ · photons/pulse in H9 from the high-order harmonic generation process. This estimate is ob-tained from the sum of Xe 5p / and 5p / photoelec-tron counts per pulse (c.f., Fig. 3), a photoionizationcross section of 64.2 Mb at 14 eV [69], an estimated over-all electron detection efficiency of 0.5 [65, 68] and anestimated magnetic bottle acceptance area diameter of180 µ m (twice the beam FWHM diameter). We thus ob-tain a H9 conversion efficiency of 2 · − , which can becompared to literature values of ∼ · − for lower-orderharmonics in Ar [70]. III. RESULTS AND DISCUSSION
To demonstrate the versatility of our spectrometerwe show time-resolved pump–probe photoion and pho-toelectron spectra where ionization is either achieved bya multi-photon 800 nm transition or by a single 14 eVphoton (H9, 89 nm). As a first sample we investigatethe polycyclic aromatic hydrocarbon perylene (C H ,mass = 252 u). Perylene is a strong absorber at 400 nm(S ← S excitation) [71] with a gas phase excited statelifetime of 5 ns [72] and negligible Stokes-shift [71]. Thefluorescence yield of the S state in the gas phase is97% [73]. A. 800 nm multi-photon probe: photoion detection
Figure 4 shows a mass spectrum of perylene obtainedwith 800 nm photoionization (black curve) and 400–800 nm pump–probe photoionization at positive time de-lay (the 400 nm pump pulse arrives before the 800 nmprobe pulse, gray curve). Strong field multi-photonionization from the S ground state with 800 nm at1 . × W/cm leads to significant fragmentation andmultiple ionization, as expected by comparison to theionization of similar size polycyclic aromatic hydrocar-bons in this laser intensity range [74]. The regions aroundthe parent ion peak at 252 u (Fig. 4b) and the doublycharged parent ion peak at 126 u (Fig. 4c) are shownin more detail. We estimate the mass resolution of ourspectrometer in ion mode to be at least 200 for a repellervoltage of +3 kV. We note that the ionization volume isvery small due to a tight focusing of the 800 nm beam( ∼ µ m FWHM) and the non-linearity of multi-photonionization process. Interestingly, for both perylene + andperylene ++ detachment of H dominates over single Hatom detachment. The peaks at 253 u and 126.5 u resultfrom the 1.1% natural abundance of the C isotope.The 400 nm–800 nm pump–probe spectrum is shown ingray in Fig. 4. We chose the laser intensities such that weobtain a negligible signal from the pump or probe laseralone ( ∼
50 counts per second, cps) as compared to thepump–probe signal ( ∼ + signal at252 u seems to be saturated, indicated by an unrealisti-cally high signal at 253 u. The time-dependent molecularresponse as function of the pump–probe delay is shownin Fig. 5. Plotted are the time-dependent ion signalsof C H +10 (perylene minus H ion at 250 u) in red (cir-cles and dashed line) and the doubly charged parent ionperylene ++ (126 u) in blue (crosses and solid line). Theduration of the signal increase around time zero is propor-tional to the cross correlation of the temporal profiles ofthe 400 nm pump and the 800 nm probe pulse. Assumingthat pump and probe pulses have a Gaussian temporalprofile, we fit the data points with the cumulative distri-bution function of a Gaussian normal distribution
50 100 150 200 25000.511.52 m/q [u] no r m a li z ed c oun t s
120 122 124 126 128 13000.51 m/q [u] no r m a li z ed c oun t s
246 248 250 252 25400.51 m/q [u]800 nm400−800nm a) b) c)
FIG. 4. (a) Photoion spectra obtained with solely 800 nmmulti-photon ionization (black curve, offset by 1, 1 . × W/cm ) and 400–800 nm pump–probe photoionization atpositive time delay (gray curve, 800 nm: 5 . × W/cm ).The lower images show regions of the doubly charged parentat 126 u (b) and the parent at 252 u (c). y = 1 / (cid:20) (cid:18) x − µ √ σ (cid:19)(cid:21) , (1)where erf() is the errorfunction, µ is the mean and σ is the standard deviation. We obtain the FWHM of thecross-correlation as ∆ T = 2 p · σ = 62 ± ± ± B. H9 single-photon probe: photoelectrondetection
The photoelectron spectrum of perylene observed withH9 (14 eV) (Fig. 6a, black curve) agrees well with a ref-erence spectrum obtained with a He(I) emission lamp at21.6 eV [75]. The vertical ionization potential of pery-lene is 7 eV and the spectrum consists of five bands inthe energy range of 7 eV to 14 eV [75]. The observedtransitions could be assigned to cationic states based on −200 −100 0 100 20000.20.40.60.81 pump−probe delay [fs] no r m a li z ed c oun t s FIG. 5. Time-dependent 400 nm–800 nm pump–probe pho-toion signal measured at 126 u (blue crosses and blue solidline) and at 250 u (red circles and red dashed line). The ris-ing edge is fitted with an error function. time-depended density functional theory and calculatedvertical excitation energies agree within ∼ . no r m a li z ed c oun t s no r m a li z ed c oun t s d i ff e r en c e c oun t s He(I)H9400−H9H9−400400 a) b)c)
FIG. 6. (a) Perylene PE spectrum obtained with H9 (14 eV,black curve) compared to a reference spectrum (gray curve,shifted in energy to be comparable to our result) [75]. (b)Pump–probe signal (300 fs time delay, gray curve), probe–pump signal ( −
700 fs time delay, black solid curve) and400 nm pump-only (black dotted curve) spectra. (c) Differ-ence spectrum obtained as difference of the pump–probe sig-nal and probe–pump signal shown in (b).
We obtain 400 nm–H9 pump–probe difference spectraby subtracting the probe–pump (negative time delay)signal from the pump–probe (positive time delay) sig-nal, as shown in Fig. 6b and c. A careful choice of thepump pulse intensity is very important in order to gener-ate sufficient population in the S state and at the sametime limit multi-photon excitation by the pump pulse. A400 nm pulse energy of 650 nJ (4 . · W/cm ) turnedout to be a good compromise (we estimate that about17% of the molecules are excited, assuming a 64 Mb ex-citation cross section [79, 80]). At these laser intensitiesthe pump pulse alone produces a relatively large amountof photoelectrons (Fig. 6b, black dotted curve), proba-bly because of resonant enhancement of the two photontransition from S into the continuum. To decrease theload of our MCP detectors we apply a negative poten-tial to the drift tube, which we account for in the spec-tra. The retardation is reflected in the PE spectra bya cut at 3.2 eV. Fig. 6b compares a 400 nm–H9 pump–probe spectrum (300 fs time delay, gray curve) with aH9–400 nm probe–pump spectrum ( −
700 fs time delay,black curve). The corresponding pump–probe differencespectrum is shown in Fig. 6c. Two new PE bands in thedifference signal appear at 8.2 eV and 10.1 eV and canbe attributed to ionization of the excited molecule in theS state to the D0 and D1 continuum, respectively. Theenergy of the emerging bands is consistent with excita-tion by the 3 eV (400 nm) pump pulse. Consequently,a transfer of population from S to S reduces the PEsignal originating from S resulting in a negative differ-ence signal. This is directly evident for the two highestenergy peaks at 5.1 eV and 7.0 eV, which shift to 8.2 eVand 10.1 eV, respectively (corresponding energy regionsare marked with vertical dashed and solid lines in Fig. 6).The difference signal maximum at 6.1 eV corresponds toa shift of the peak at 3.3 eV. In the energy range between5 eV and 7.5 eV positive contributions (population of S )and negative contributions (depletion of S ) to the differ-ence signal overlap, rendering the determination of bandshifts complicated. Below 5 eV the difference spectrumbecomes very noisy due to the strongly increasing 400 nmelectron signal. −100 0 100 200 300 400 50000.51 pump−probe delay [fs] no r m a li z ed c oun t s FIG. 7. Time-dependent rise of the transient peaks in thedifference spectrum at 8.2 eV (red circles and dashed red line)and 10.1 eV (black circles and solid black line, c.f., Fig.6c).
To obtain the time-dependent PE signal we integratethe difference signal in the energy regions of the shiftedPE lines (marked with vertical lines in Fig. 6 c) for dif-ferent pump–probe delays. The transient PE signals areshown in Fig. 7. We use a random delay sequence toavoid systematic errors and the data points are moreclosely spaced around time zero. By fitting equation 1to the data points we obtain a cross correlation FWHMof 68 ±
10 fs (averaged over both PE energy regions).With a 400 nm pulse duration of 58 ± ± IV. CONCLUSION
We have described a time-resolved pump-probe pho-toelectron and photoion spectrometer to study photoin-duced relaxation processes in gas-phase molecules. Ultra-short pulses in the VUV region obtained from high-orderharmonic generation are used as probe. A simple andcost efficient way has been presented to isolate the 9 th harmonic of a 800 nm fundamental from the harmonicspectrum by means of an In metal filter in combinationwith a normal incidence MgF coated Al mirror. Theprobe photon energy of 14 eV allows to address the wholeset of valence electron levels in many molecules. Due tothe fact that the probe pulse also ionizes the ground state,difference spectra between reversed time delays are usedto infer molecular transients. The H9 pulse duration of35 ± A. Acknowledgements