Last year at this conference several approaches to utilize machine learningBishop, C M. “Neural networks for pattern recognition.” Oxford university press, 1995.o train a computer to recognize the patterns inherit in rotational spectra were presentedZaleski, D. P.; Prozument, K. Identifying Broadband Rotational Spectra with Neural Networks, International Symposium on Molecular Spectroscopy, June 21; 2017. It was shown that the recognized patterns could be used to identify (or classify) a rotational spectrum by its Hamiltonian type, but at the time, the rotational constants were not recovered. Here, we describe a feed forward artificial neural network that has been trained to identify different types of rotational spectra and determine the parameters of the molecular Hamiltonians. The network requires no user interaction beyond loading a “peak pick”, and can return fits within a fraction of a second. The rotational constants are typically deduced with the accuracy of 1–10 MHz. We will describe how the network works and provide benchmarking results.

---

Footnotes:

Bishop, C M. “Neural networks for pattern recognition.” Oxford university press, 1995.t Zaleski, D. P.; Prozument, K. Identifying Broadband Rotational Spectra with Neural Networks, International Symposium on Molecular Spectroscopy, June 21; 2017..

Chirped-pulse Fourier-transform microwave (CP-FTMW) spectrometers such as the instrument at the University of Virginia can acquire spectra with high sensitivity in a short amount of time. This necessitates new approaches to spectroscopic analysis to ensure identification of all species in each recorded spectrum. With an intensity range covering four orders of magnitude after averaging 1 million free induction decays, a recent spectrum of a fluoroethylene (FE)/CO₂ mixture in the 2 – 8 GHz range has over 11,000 lines with signal-to-noise ratio above  ∼ 2.5. These transitions may arise from a combination of monomer, dimer and larger cluster species, including low abundance isotopes and complexes with carrier gas, water or other contaminants. Our current focus is identifying spectra of FE(CO₂)_n clusters, containing progressively larger numbers of CO₂ molecules. Several methods have been used to facilitate identification of lines for these spectra, which are expected to lose 1-2 orders of magnitude of intensity for each increase in cluster size. These approaches include subtraction of transitions that are observed in the FE-only spectrum from the spectrum of the FE/CO₂ mixture, visual identification of patterns characteristic of asymmetric molecules, and application of extended cross correlation (XCC) techniques.N.P. Jacobson, S.L. Coy, R.W. Field, J. Chem. Phys., 107 (1997) 8349.n the XCC approach, several spectra with systematically varied conditions (such as pressure or concentration) are compared, and a combination of graphs and computerized algorithms is used to identify transitions that behave similarly under the changing conditions. In addition to applications for fundamental spectroscopic studies, this approach has potential application to identification of the components of complex mixtures.

---

Footnotes:

N.P. Jacobson, S.L. Coy, R.W. Field, J. Chem. Phys., 107 (1997) 8349.I

Although there is a vast amount of van der Waals complexes containing small molecular species, there have only been a small number of studies containing an O₂ binding partner. Within this grouping, only five are known to have been attempted using high-resolution microwave techniques with only two- O₂-HFS. Wu, G. Sedo, E. M. Grumstrup, K. R. Leopold, J. Chem. Phys., 127 (2011) 204315-1-204315-11.nd O₂-H₂OY. Kasai, E. Dupuy, R. Saito, K. Hashimoto, A. Sabu, S. Kondo, Y. Sumiyoshi, Y. Endo. Atmos. Chem. Phys., 11 (2011) 8607–8612. appearing in the literature. This void is presumably due to the ³Σ ground state of O₂ and how this coupling complicates spectral assignment. However, these sorts of couplings determined from high-resolution analysis add rich and important information useful in complex structure determination. Until now, the only analyses of such complexes have been done utilizing Hund’s case a asymmetric modelsW. M. Fawzy, J. Mol. Spec., 191 (1998) 68-80.^,H.−B. Qian, S. J. Low, D. Seccombe, B. Howard, J. Chem. Phys., 107 (1997) 7651−7657.r Hund’s case b linear Hamiltonians. However, further study either resulted in inaccurate predictions of close transitions (O_2−H_2O) or difficult adjustments to similar systems due to extra complexity (O_2−HCl from O_2−HF). This made a more standardized approach, such as using the ubiquitous program suite of SPFIT/SPCAT by PickettH. M. Pickett, J. Mol. Spec., 148 (1991) 371−377.eem like a more flexible solution. However, SPFIT/SPCAT uses a Hund’s case b approach which needed to be sorted out and the problems and pitfalls of these analyses will be discussed. This talk will include how to be predictive using O_2−H_2O and O_2−HF as examples as well as discuss work currently being pursued on the molecules O_2−HCl and O_2 H.-B. Qian, S. J. Low, D. Seccombe, B. Howard, J. Chem. Phys., 107 (1997) 7651-7657.oH. M. Pickett, J. Mol. Spec., 148 (1991) 371-377.s

The X⁻·(H₂O) (X=I, Cl, F) clusters ion provide a microscopic window into the intracluster energy relaxation dynamics that ultimately lead to dissociation when excited above threshold (Do). Here we explore the spectra of simple binary complexes of one water molecule with an iodide ion when the system is excited to vibrational levels that span the energy range through Do. This is accomplished by recording the vibrational photodissociation spectra of different vibrational excited states of X⁻·(H₂O) using a two-color, IR-IR photodissociation technique. We first quantify the ground state spectra by recording the single photon absorption spectra of cryogenically cooled cluster ions with messenger tag technique. Then we fix the pump laser frequency on a transition known to the ground state cluster ion and scan the probe laser to obtain the excited state photodissociation spectra. Owing to the long lifetime of the vibration excitations in this class of clusters, each pump laser excitation frequency yields a different spectrum (the traces on the right in the figure). This provides an opportunity to obtain a kind of 2DIR spectroscopy (left bottom figure) of cryogenically cooled gas phase ions. The result is a cluster variation of the vibrationally-mediated photodissociation experiments pioneered in the 1990s by Crim and Rizzo on polyatomic molecules. In the cluster regime, we find remarkably long lived (greater than 50 μs!) vibrational levels 300 cm⁻¹above the dissociation threshold and surprisingly localized excitations for the bound OH stretches. This feature allows us to follow many pathways up the vibrational landscape far beyond the dissociation limit.

The water-iodide monomer (I⁻H₂O), nominally the simplest of the halide-water complexes, challenges our current understanding of ion hydration. Most notably, this seemingly simple complex displays multiple resonant vibrational transitions, a low tunneling barrier, and a strong transition dipole moment along the bound OH-I coordinate. These effects combine to yield spectroscopic signatures that deviate strongly from traditional harmonic analyses and are even difficult to qualitatively reproduce with anharmonic methodologies. Among these signatures is a quartet of peaks in the 3300 – 3500 cm-1 range that is unexplained using traditional single-photon spectroscopy. Challenging both experiment and theory alike, this situation required the interpretation of newly developed IR-IR 2-color photodissociation spectroscopy that probes well below the dissociation threshold. In this work, we use both exact eigensolver techniques and a newly developed vibrationally adiabatic model, along with a new potential energy surface^a to computationally explore the excited-state spectra. The resulting analyses identify the source of the strength of the resonant transitions, directly assign the vibrational and rotational spectroscopic signatures, and discern the electronic origin of these surprising effects in this fundamental model of ion hydration. ^aBajaj, P. Gotz, A.W., Paesani, F. Journal of Chemical Theory and Computation, 2016, 12 (6), 2698-2705.

The fundamental Ritz combination principle W. Ritz, On a new law of series spectra, Astrophys. J. 28 (1908) 237.riginally found for atoms has also been applied to molecules as a method to reconstruct the energy states from measured lines without relying on any model Hamiltonian. In 2006 Nesbitt and coworkers C. Savage, F. Dong, D.J. Nesbitt, Toward a quantum-mechanical understanding of the high-resolution infrared spectrum of CH₅⁺, in: Contribution TA05, 61st International Symposium on Molecular Spectroscopy, Columbus, OH, USA, 2006.roposed to apply it to protonated methane, CH₅⁺. We used this idea to reconstruct a part of its ground state energies employing spectra of combination differences (CDs) determined from very high resolution ro-vibrational data Oskar Asvany, Koichi M. T. Yamada, Sandra Brünken, Alexey Potapov, Stephan Schlemmer. Experimental ground-state combination differences of CH₅⁺. Science, 347(6228):1346–1349, 2015. Since then the method has been significantly improved S. Brackertz, S. Schlemmer, O. Asvany, Searching for new symmetry species of CH₅⁺ – From lines to states without a model, J. Mol. Spectrosc. 342 (2017) 73–82.s the CD lines essentially represent kernel density estimations, a well-known tool in mathematics. Furthermore, a combinatorial approach has been developed to reconstruct vibrational ground states as well as vibrationally excited states from the CD spectra without relying on measurements at different temperatures. As a result, 1063 of the 2897 measured lines of CH₅⁺ being part of four different symmetry species could be assigned. This allowed for a comparison of the measurements with the analytical model of Schmiedt et al. H. Schmiedt, P. Jensen, S. Schlemmer, Rotation-vibration motion of extremely flexible molecules – the molecular superrotor, Chem. Phys. Lett. 672 (2017) 34–46.s well as with the ab initio calculations of Wang and Carrington X.-G. Wang, T. Carrington, Calculated rotation-bending energy levels of CH₅⁺ and a comparison with experiment, J. Chem. Phys. 144 (2016) 204304.

---

Footnotes:

W. Ritz, On a new law of series spectra, Astrophys. J. 28 (1908) 237.o C. Savage, F. Dong, D.J. Nesbitt, Toward a quantum-mechanical understanding of the high-resolution infrared spectrum of CH₅⁺, in: Contribution TA05, 61st International Symposium on Molecular Spectroscopy, Columbus, OH, USA, 2006.p Oskar Asvany, Koichi M. T. Yamada, Sandra Brünken, Alexey Potapov, Stephan Schlemmer. Experimental ground-state combination differences of CH₅⁺. Science, 347(6228):1346–1349, 2015.. S. Brackertz, S. Schlemmer, O. Asvany, Searching for new symmetry species of CH₅⁺ – From lines to states without a model, J. Mol. Spectrosc. 342 (2017) 73–82.a H. Schmiedt, P. Jensen, S. Schlemmer, Rotation-vibration motion of extremely flexible molecules – the molecular superrotor, Chem. Phys. Lett. 672 (2017) 34–46.a X.-G. Wang, T. Carrington, Calculated rotation-bending energy levels of CH₅⁺ and a comparison with experiment, J. Chem. Phys. 144 (2016) 204304..

Broadband rotational spectroscopy has several experimental advantages as a technique for the analysis of complex chemical mixtures without the need for chromatography to separate the distinct chemical species prior to analysis. The technique has high spectral resolution so that mixtures with a large number of components can be analyzed without spectral overlap, the frequency accuracy is excellent so that library spectra can be transferred between instruments, and the measurement has high dynamic range so that low level impurities can be detected in the presence of dominant species like the solvent in a direct-from-flask reaction mixture. It also has the special feature of a spectroscopic signature that is dependent on the molecular mass distribution so that isomers can be resolved – a problem that can be a challenge for the high-sensitivity analytical chemistry methods based on mass spectrometry. The problem of decomposing a measured spectrum into the individual rotational spectra of each different sample molecule is common to many applications of broadband rotational spectroscopy including reaction product screening in laboratory astrochemistry, the identification of different molecular clusters in the study of weakly bound complexes, and the analysis of chemical samples from a variety of chemistry fields including pharmaceutical science. In this talk we will discuss strategies that have been developed by the spectroscopy community to solve the problem of decomposing a measurement into its constituent spectra. These approaches include traditional analytical chemistry approaches like the creation of large chemical libraries. Instrumental methods that exploit broadband detection to implement efficient double-resonance methods will also be summarized. Two approaches that may deserve additional consideration in the future will also be discussed. One approach measures the spectrum as a function of a continuously variable external parameter and then uses computer algorithms to group transitions with similar parametric dependence as a way to separate the measurement into molecule-specific spectra. The second approach uses the fact that the Hamiltonian for rotational spectroscopy is known and that only certain patterns of transitions are consistent with it. An early example of this idea is the AUTOFIT routine. The possibility of extending this approach into a fully automated computer analysis of the spectrum will be considered.

H₆⁺ is generated in a supersonic expansion via pulsed electrical discharge of hydrogen. H_n⁺ clusters are extracted into a reflectron time-of-flight mass spectrometer and probed with infrared photodissociation spectroscopy (IRPD) in the 2050 – 4600 cm⁻¹ region. H₆⁺ was mass selected and found to have three distinct photodissociation channels by loss of one hydrogen atom, one hydrogen molecule or both. Each channel results in different spectra as a result of mode specific dissociation channels. The ground ²D_2d state is 4 kcal/mol lower in energy than the ²c_S state with a 7 kcal/mol barrier. We believe we are probing the ²D_2d structure with the three H_m⁺ (m=3,4,5) fragment channels as a result of rapid interconversion between the two states after IR photon absorption.

Thin films of 1,8-diazafluoren-9-one (DFO) in titanium dioxide were synthesized using the sol-gel method. Particular attention will be given towards preparation of DFO in titanium dioxide as a potential luminescent probe of amino acids. The photophysical properties of DFO in titanium dioxide thin films were identified by a variety of spectroscopic methods including: stationary absorption and emission, the fluorescence intensity decay profiles. Atomic force microscopy (AFM), scanning electron microscopy (SEM), confocal microscopy, Raman spectroscopy, and X-ray diffraction (XRD) techniques were applied to obtain structural characteristic of the prepared films. This research has been supported by the grant NCN 2017/01/X/ST5/01541, MINIATURA 1, A. Lewkowicz.

We investigated the ethylene glycol, which is the crucial ingredient in the automotive antifreeze coolants, the content of engine oil at various levels of contamination using Fourier transform infrared (FT-IR) spectroscopy and ultraviolet-visual spectroscopy (UV-Vis). It is known that glycol in SAE 15W-40 diesel engine lubricating oil has relatively strong signatures in the infrared spectrum, some of which overlap with other molecular bonds that may already be present in engine oil. Therefore, our aim is to correlate this FT-IR data with a UV-Vis spectrograph such that detection of glycol’s presence can be improved significantly.