Standard measurement procedures and data processing

Mass spectra and CID mass spectra were recorded in enhanced resolution mode (8100 m/z /s at a mass resolution of 0.2 fwhm / m/z). The mass range was adjusted to the respective compound under study; however, the lower boundary was typically left at m/z 50. Sample solutions were prepared at a concentration of 5·10-7 _{to 5·10}-6 _{mol/L and} continuously infused into the spraying chamber by a Hamilton® _{syringe and a syringe} pump at a set flow rate of 120 µL/h. The temperature of the nitrogen drying gas was set to 180 °C at a flow rate of 4-5 L/min. Nebulizer pressure and needle potential were adjusted to yield the highest and most stable precursor ion signal. Typical values were 5 psi (345 mbar) and 4.4 kV for the pressure and needle potential, respectively. The “target mass” option within the instrument software was used to tune the transfer parameters. The “target mass” was in general set to the m/z ratio of the compound under investigation. Accumulation time of the ions was adjusted so as not to exceed a total ion count of ~1-1.5·105_{. The helium buffer gas partial pressure inside the trap was} approximately 3·10-3 _{mbar. The mass spectrometer was controlled by the} BrukerTrapControl 7.0 software and data analysis was performed using BrukerDataAnalysis 4.0.

PF and tPF measurements were performed using the modified Bruker Daltonics amaZon Speed ion trap mass spectrometer in ultrascan mode (32.500 m/z /s at a mass resolution of 0.4 fwhm / m/z). Two non-collinear parametric amplifiers of white-light continuum (TOPAS-C, Light Conversion) pumped by a CPA ultrafast regenerative amplifier (Wyvern 1000™, KMLabs, 4 µJ, 780 nm, Δt=50 fs, controlled by Dragon Master V3.00 software) provided a source of tunable radiation in the UV/Vis and NIR region in order to record static (multiple-photon) PF and tPF spectra. The output from the T7 was used for generation of pump pulses (UV/Vis) in both static and transient experiments, whereas the T6 provided exclusively probe pulses in the near infrared (NIR) region (1200 nm).

Steady-state (multiple-photon) PF spectra were obtained by recording the PF mass spectra at an individual set wavelength one at a time. The output wavelength of the T7 was set by the WinTOPAS v3.2.35 software, checked using an AvaSpec USB high resolution fiber optics spectrometer (ULS3648, Avantes) and, if necessary, corrected by adjusting the Delay2 in the WinTOPAS software. Pulse energies (1-2µJ) were kept

35 constant across the whole recorded spectral range. For this purpose, the energies at a given wavelength were monitored on an energy meter (Vega, Ophir) equipped with a pyroelectric sensor, and attenuated using a neutral density filter to the desired pulse energy, prior to recording the fragment spectrum. The mass spectrometer was operated with the BrukerIonTrap 7.0 software in ultrascan mode, averaging ten cycles for a single mass spectrum. Accumulation time was set appropriately so that the ion count did not exceed a value of 1-1.5·105_{. If not stated otherwise, the fragmentation time was set to} 150 ms, which, in combination with the 981 Hz repetition rate of the amplifier, amounted to ~150 laser pulses per ion cloud. The fragment mass spectra were recorded for at least 2 minutes for a set wavelength. The procedure was repeated for each set wavelength in the recorded spectral region, spacing the data points equidistantly with a step size of 3-5 nm. Parent ion and fragment ion intensities were extracted as ion chromatograms from the mass spectra by using a home-made VisualBasic script.[26] _The total fragment ion yield Y(λ) at a given wavelength λ was calculated according to

Y(λ)=∑Fi/[(∑Fi+∑Pj)·λ], where Fi and Pj are the fragment and parent ion intensities,

respectively. Division by the numerical value of λ was performed to account for different photon energies at constant laser pulse fluence.

Measurements of the fragment ion yield dependence on the pump pulse energy were performed analogously. For a set wavelength, consecutive PF mass spectra were recorded at increasingly higher pulse energies. Fragment ion chromatograms were extracted and processed according to the procedure described above. Energy E dependence of the fragmentation yield Y was evaluated applying a polynomial fit function ( = ∙ ), where n is an estimate for the number of photons needed to induce fragmentation.[47]

tPF spectra were recorded by resonant (one-photon) excitation of the isolated ions using the output from TOPAS-C T7 (usually in the visible region) and subsequent time- delayed probing by non-resonant (multiple-)photon (1200 nm) absorption of the photoselected ion ensemble. Pump intensities were, if applicable, usually kept as low as possible to avoid multiple-photon excitation. If not stated otherwise, the initial repetition rate (981 Hz) of the pump and probe pulse trains was reduced to 327 Hz by an optical chopper, resulting in irradiation of a single ion cloud with ~50 pump-probe pairs. The mutual polarization plane orientation of the linearly polarized pump and

36

probe beams was adjusted by rotating the polarization plane of the pump with respect to the (vertically polarized) probe pulses using a Berek polarization compensator. The retardation for a selected pump pulse wavelength was adjusted accordingly by tilting the compensator plate, so that the compensator functioned as a λ/2-waveplate, allowing for easy control of the polarization plane orientation by rotation of the Berek compensator. The probe pulses were continuously delayed in time with respect to pump pulses by means of a silver retroreflector mounted on a single axis delay stage (M- 531.DD, PI). The DC motor controller (C-863.11 Mercury, PI) operating the delay stage was controlled using the PIMikroMove Software. The pump and probe pulses were spatially overlapped and focused onto the ion cloud inside the trap with a f=50 cm lens. Pointing of the beams was adjusted for maximum PF yield at zero time delay (t0). The time-delay was continuously scanned (from negative to positive pump-probe delay) at constant scan speed, which was adjusted for equidistant steps between each data point of typically 4 ps (for a total time delay of 1.6 ns), 2 ps (total delay of 800 ps), 100 fs (total delay of 40 ps) and 20 fs step size for total delays shorter than 40 ps. Each scan took 10 min to complete and was repeated 12-15 times, depending on signal quality. Fragment ion chromatograms were extracted from the raw data using a VisualBasic script,[26] _{which was run in the BrukerDataAnalysis 4.0 software. The yields Y}

i of the

individual fragment ion channels were evaluated according to Yi=Fi/[(∑Fi+∑Pj)]

(Fi: intensity fragment i and Pj: intensity parent j) and additionally converted from the

internal time-scale of the mass spectrometer to a pump-probe delay dependency with a template file developed in LibreOffice Calc. The so obtained kinetic traces were averaged over all repeated scans for every single fragment ion channel and either summed up (total fragment yield) or subjected to a fitting routine individually.

To evaluate the system response at any given pump-probe wavelength combination, the cross correlation function (ccf) was determined by measuring the pump-probe dependent multiple-photon ionization of neutral furan inside the ion trap with the same step size and at the same scan speed used in the respective experiment.[27-28] _{For that} purpose, furan at a partial pressure of ~1-3·10-7 _{mbar was introduced into the trap} according to the procedure described above. The overlap of the pump and probe pulse electric fields give a rise to a quasi-instantaneous, short-lived ion signal resulting from laser induced electron detachment from neutral furan molecules, which are then detected mass spectrometrically. Time-zero and the temporal resolution were estimated

37 from the center and the fwhm of the Gaussian ion signal (Figure 20), respectively and used in fitting of the kinetic traces.

Figure 20 Temporal resolution estimated from a Gaussian approximation of

the cross correlation measured by coherent (multiple-)pump+probe photoionization of neutral furan; shown examplatorily for a wavelength combination of: λpump=400 nm + λprobe=1200 nm.

Transient fragment spectra were usually fitted by a convolution of a sum of exponentials augmented with the system response obtained from the laser pulse cross correlation according to S(t)=∑iAi(exp(-t/τi)*g(t,t0,tp)) (τi decay time constants, t0: time zero, tp:

fwhm of ccf), using Origin 9.0G or a fitting software[48] _{based on the MINUIT} optimization package.[49] _{Sinusoidally modulated transients were evaluated with} Origin 9.0G or an open source software (DecayFit)[50] _{running within the MATLAB} (R2013a) environment.