VIS and VUV spectroscopy of 12 C 17 O and deperturbation analysis of the A 1 P , y ¼ 1 – 5 levels

High-accuracy dispersive optical spectroscopy measurements in the visible (VIS) region have been performed on the less-abundant 12 C 17 O isotopologue, observing high-resolution emission bands of the B 1 S + ( y ¼ 0) / A 1 P ( y ¼ 3, 4, and 5) ˚ Angstr¨om system. These are combined with high-resolution photoabsorption measurements of the 12 C 17 O B 1 S + ( y ¼ 0) ) X 1 S + ( y ¼ 0) and C 1 S + ( y ¼ 0) ) X 1 S + ( y ¼ 0) Hop ﬁ eld – Birge bands recorded with the vacuum-ultraviolet (VUV) Fourier transform spectrometer, installed on the DESIRS beamline at the SOLEIL synchrotron. The frequencies of 429 observed transitions have been determined in the 15 100 – 18 400 cm (cid:2) 1 and 86 900 – 92 100 cm (cid:2) 1 regions with an absolute accuracy of up to 0.003 cm (cid:2) 1 and 0.005 cm (cid:2) 1 for the B – A, and B – X, C – X systems, respectively. These new experimental data were combined with data from the previously analysed C / A and B / A systems. The comprehensive data set, 982 spectral lines belonging to 12 bands, was included in a deperturbation analysis of the A 1 P , y ¼ 1 – 5 levels of 12 C 17 O, taking into account interactions with levels in the d 3 D i , e 3 S (cid:2) , a 0 3 S + , I 1 S (cid:2) and D 1 D states. The A 1 P and perturber states were described in terms of a set of deperturbed molecular constants, spin – orbit and L -uncoupling interaction parameters, equilibrium constants, 309 term values, as well as isotopologue-independent spin – orbit and rotation-electronic perturbation parameters.


Introduction
Carbon monoxide (CO) is one of the most thoroughly studied molecules, bearing signicance to astronomy and cosmology. Aer H 2 , it is the second most abundant molecule in the interstellar medium (ISM), where it is investigated as a tracer of gas properties, structure and kinematics. 1,2 In such astrophysical environments CO controls much of the gas-phase chemistry, 3 and is a precursor to complex molecules. 4 The CO spectrum has been observed in comets, cool dwarfs, quasars, supernova remnants, and interstellar molecular clouds as well as in atmospheres of planets and transiting exoplanets. 5,6 Emissions originating from the B 1 S + (y ¼ 0), B 1 S + (y ¼ 1), and C 1 S + (y ¼ 0) vibrational levels were recorded from the Martian and Venusian atmospheres by the Hopkins Ultraviolet Telescope, 7 the FUSE satellite, 8,9 and the Cassini UVIS instrument. 10 Large CO abundances produce detectable signals even for the rare isotopologues, including 12 C 17 O. [11][12][13] Investigations of minor isotopologues are applied to unravel 'depth effects' in the interstellar absorptions 14 13,15 The CO vacuum ultraviolet absorption spectrum is of astrophysical relevance due to the photodissociation of VUVexcited states, e.g. the C 1 S + , B 1 S + and E 1 P states. 16 Isotopedependent photodissociation effects, due to self-shielding in high-column density environments, 15,17 lead to isotopic fractionation of CO. 13,18 The less-abundant 12 C 17 O isotopologue was detected in the ISM for the rst time in 1973 in the Orion Nebula 19 and has been studied in the laboratory in a number of investigations. [20][21][22][23][24][25] Table 1 Transition frequencies (in cm À1 ) of the 12 C 17 O B 1 S + / A 1 P emission bands from the high-accuracy dispersive optical spectroscopy measurements a J 00 B 1 S + / A 1 P (0, 3) B 1 S + / A 1 P (0, 4) B 1 S + / A 1 P (0, 5) P 11ee (J 00 ) Q 11ef (J 00 ) R 11ee (J 00 ) P 11ee (J 00 ) Q 11ef (J 00 ) R 11ee (J 00 ) P 11ee (J 00 ) Q 11ef (J 00 ) R 11ee (J 00 ) Hakalla and co-workers have investigated the visible spectrum of 12 C 17 O, comprising the B 1 S + -A 1 PÅngström system, 26,27 as well as the C 1 S + -A 1 P Herzberg system. 28 The VUV spectrum of the C 1 S + -X 1 S + system was investigated by laser excitation 22,29 and the B 1 S + -X 1 S + system by absorption of synchrotron radiation. 25 The A 1 P state is subject to some of the most extensive and complex perturbations among all the states that are known in the carbon monoxide molecule. [30][31][32][33][34][35][36][37][38] The d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D electronic states are responsible for all of the existing irregularities. A systematic classication of the perturbations of  High resolution emission spectra, recorded with the highaccuracy dispersive optical spectroscopy setup 66 at an instrumental resolution of 0.15 cm À1 , of the 12 C 17 O B 1 S + / A 1 P (0, 5) band with the perturber lines associated with the B 1 S + / d 3 D (0, 11) transition. The ratio of the gas compositions used to obtain the molecular spectra was 12 C 17 O : 12 C 16 O ¼ 1 : 0.35. (Panel (a)) An overview of the observed B 1 S + / A 1 P (0, 5) and B 1 S + / d 3 D (0, 11) spectra (upper trace) together with the final branch assignments, calibrating Th atomic lines (going beyond the scale), as well as simulated spectra 67 (lower trace). The empty circles indicate spectral lines of undetermined location due to overlap with much more intense atomic lines of carbon, hydrogen, and deuterium. (Panel (b)) Expanded view of the B / A (0, 5) band head region in 12 C 17 O at an enlarged scale.  2 High resolution emission spectra, recorded with the highaccuracy dispersive optical spectroscopy setup 66 at an instrumental resolution of 0.15 cm À1 , of the 12 C 17 O B 1 S + / A 1 P (0, 4) band, with the perturber lines associated with the B 1 S + / e 3 S À (0, 7), B 1 S + / I 1 S À (0, 6), and B 1 S + / a 03 S + (0, 14) transitions (upper trace) together with the final branch assignments, calibrating Th atomic lines (going beyond the scale), as well as simulated spectra 67 (lower trace). The ratio of the gas compositions used to obtain the molecular spectra was the A 1 P state in the main 12 C 16 O molecule was carried out by Krupenie. 39 Simmons et al. 40 made a critical analysis of this study as well as completed it. A conclusive analysis and deperturbation calculations were carried out by Field et al. 30,32,41 Next, Le Floch et al. 31 conducted a comprehensive study of perturbations in the lowest A 1 P, y ¼ 0 vibrational level. In his next works 42,43 he analysed perturbations occurring in the A 1 P, y ¼ 0-4 levels, and calculated very precise term values for the A 1 P, y ¼ 0-8 states, respectively. Recently, the A 1 P state of the main 12 C 16 O isotopologue has been studied in the A-X transition [44][45][46] by the Amsterdam group by means of highly accurate twophoton Doppler-free excitation using narrow band lasers 47 with relative accuracy up to Dl/l ¼ 2 Â 10 À8 , as well as by vacuum ultraviolet Fourier-transform spectroscopy (VUV-FTS) at the SOLEIL synchrotron. [48][49][50] An improved deperturbation analysis of A 1 P in ordinary CO has recently been performed by Niu et al. 44,51 Far fewer deperturbation analyses of the A 1 P state have been performed in other isotopologues of CO ( 12 C 18 O and 13 C 18 O). 33,52,53 A considerable contribution to the identication and classication of the A 1 P state perturbations has been made by Kępa and Rytel in a number of investigations over the years. [54][55][56][57][58] Here, the focus is on a deperturbation analysis of the A 1 P (y ¼ 1, 2, 3, 4, and 5) levels in the 12 C 17 O isotopologue. The deperturbation is based on new observations of the 12 C 17 O B / A (0, 3), (0, 4), (0, 5) bands recorded in visible emission at high resolution and previously published studies of theÅngström 26,27 and Herzberg bands. 28 The deperturbation analysis prompted some reassignment of lines in the B-A and C-A systems. New, highly accurate measurements of the 12 C 17 O B ) X (0, 0) and C ) X (0, 0) transitions with VUV-FTS were performed and included in the study in order to (i) establish and verify that B (y ¼ 0) and C (y ¼ 0) levels are unperturbed, and that our perturbation analysis of A-state is not affected by shis in the upper states, (ii) include an independent set of improved constants, therewith level energies, of B (y ¼ 0) and C (y ¼ 0), as well as (iii) determine level energies of A-state with respect to ground state of CO. The comprehensive t on B-A, C-A, B-X, and C-X systems allowed us to perform the most accurate deperturbed rotational constants of the states under consideration.

Experimental details
2.1. Emission spectra of the B 1 S + / A 1 P system In this study, a water-cooled, hollow-cathode lamp with two anodes 65 and a high-accuracy dispersive optical spectroscopy method were used for a high-resolution spectroscopic investigation of the 12 C 17 O B 1 S + (y ¼ 0) / A 1 P (y ¼ 3, 4, and 5) bands in the visible region. The lamp was initially lled with a mixture of helium and acetylene 12 C 2 D 2 (Cambridge Isotopes, 12 C 99.99%) under the pressure of approximately 6 Torr. An electric current was passed through the mixture for about 200 h, aer which a small quantity of 12 C carbon became deposited on the electrodes. Subsequently, the lamp was evacuated and Table 2 Transition frequencies of the (B 1 S + , C 1 S + ) / (d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D) extra-line bands in 12  a The estimated calibration 1s uncertainty was 0.002 cm À1 . The absolute accuracy of the signicant majority of extra-lines should be assumed as not better than approximately 0.01 cm À1 due to their weakness. b Lines marked with 'w' were weak and with 'b' were blended in the spectra. The superscript o, p, r, s, or q preceding the main notation P, Q, R of the branch indicates the change in total angular momentum excluding spin for transition to the perturber state. 68 oxygen containing the 17 O 2 isotope (Sigma-Aldrich, 17 O 2 60%) was admitted at a static gas pressure of 2 Torr. The anodes were operated at 2 Â 650 V and 2 Â 50 mA dc. During the discharge process the 17 O 2 molecules decay into atomic oxygen, which then combine with 12 C-carbon atoms, ejected from the outer layer of the cathode, thus forming the 12 C 17 O molecules in the gas phase. The temperature of the plasma formed at the centre of the cathode was about 600-700 K. These conditions were found to be optimal for the production of CO molecular spectra under control of isotopic composition. The experimental equipment of the Rzeszów laboratory, where these measurements were conducted, has been described in detail by Hakalla et al. 66 Spectroscopic measurements were made by means of a 2 m Ebert plane-grating spectrograph equipped with a 651.5 grooves per mm grating with a total of 45 600 grooves, blazed at 1.0 mm in 3 rd and 4 th order, giving reciprocal dispersion and resolving power in the ranges 0.11-0.19 nm mm À1 and 182 400-136 800, respectively. Discharge emission signals were recorded by means of a photomultiplier tube (HAMAMATSU R943-02) mounted on a linear stage (HIWIN KK5002) along the focal curve of the spectrograph. The input and exit slits were 35 mm in width. The intensities of the lines were measured by means of photon counting (HAMAMATSU C3866 photon counting unit and M8784 photon counting board) with a counter gate time of 200-500 ms (no dead time between the gates). The position of the exit slit was measured by means of a He-Ne laser interferometer (LASERTEX) synchronized with the photon counting board. During one exposure of the counter gate, the position was measured 64 times. Simultaneously recorded thorium atomic lines, 69 obtained from an auxiliary water-cooled, hollowcathode tube lled with Th foil were used for absolute CO wavenumber calibration.
The peak positions of spectral lines were derived by means of a least-squares procedure assuming a Gaussian line-shape for each spectral contour (30 points per line), with a tting  [26][27][28]. b The estimated calibration 1s uncertainty was 0.002 cm À1 . The absolute accuracy of the signicant majority of the lines should be assumed as not better than approximately 0.01 cm À1 . c Lines marked with 'w' were weak and with 'b' were blended in the spectra. uncertainty of the peak position for a single unblended line in the range 0.1-0.2 mm, that is 2.5-8 Â 10 À4 cm À1 in the observed region. To determine the 12 C 17 O B 1 S + / A 1 P wavenumbers, 5 th -and 6 th -order interpolation polynomials were used for the (0, 3), (0, 4), and (0, 5) bands. The absolute wavenumber calibration at 1s uncertainty is 0.002 cm À1 . The strong and unblended lines exhibit a full-width half-maximum (FWHM) of 0.15 cm À1 , maximum signal-to-noise ratio of about 100 : 1 as well as count rates of up to about 16 000-60 000 photons per s for the 12 C 17 O B 1 S + (y ¼ 0) / A 1 P (y ¼ 3, 4, and 5) bands. The absolute accuracy of the frequency measurements was 0.003 cm À1 , corresponding to a relative accuracy of Dl/l ¼ 2 Â 10 À7 , for the 15 180-18 400 cm À1 spectral region. However, weaker or blended lines have lower accuracy, at worst 0.07 cm À1 or Dl/l ¼ 4 Â 10 À6 . Preliminary identication of the B 1 S + (y ¼ 0) / A 1 P (y ¼ 3, 4, and 5) bands was carried out by means of the information provided in our recent works on the 12 C 17 O molecule. 26,27 For the frequency measurements of the lines investigated, blending effects of the 12 C 16 OÅngström system were taken into account. They occur as a result of using oxygen 17 O 2 with spectral purity of only 60%. In total, 283 emission lines belonging to the B 1 S + / A 1 P band system in 12 C 17 O were identied and rotationally assigned. The transition frequencies are provided in Table 1.
The observed 12 C 17 O B 1 S + (y ¼ 0) / A 1 P (y ¼ 3, 4, and 5) spectra, together with extra-lines, assignments, calibrating Th atomic lines, and nal simulated spectra are shown in Fig. 1-3. By "extra-lines", we refer to the spectral emission lines terminating on perturber states and gaining intensity from mixing with the A 1 P state. An additional impediment was the appearance of four atomic lines overlapping the region of the 12 C 17 O B / A (0, 5) band with signicantly higher intensities and broader FWHMs. They were identied by means of the Atomic Spectra Database (ASD) of NIST 59-63 as the C lines at 15186.739 cm À1 and 15197.891 cm À1 , as well as the H Balmer-alpha line at 15233.157 cm À1 and deuterium D line at 15237.272 cm À1 . As a result, it was not possible to measure the positions of the P(11), P(15), and R(6) B / A (0, 5) lines (marked with empty circles in Fig. 3a).
Our deperturbation analysis allowed us to assign 24 rotational lines from 14 bands of the B 1 S + / d 3 D i , B 1 S + / e 3 S À , B 1 S + / a 03 S + , B 1 S + / I 1 S À , B 1 S + / D 1 D, C 1 S + / e 3 S À , C 1 S + / a 03 S + , and C 1 S + / D 1 D systems in 12 C 17 O. The transition frequencies and assignments are presented in Table 2. Since most of them are weak their accuracy is not better than 0.01 cm À1 . The deperturbation included some lines from the 12 C 17 O B 1 S + / A 1 P (0, 1) and (1, 5) bands which we have measured with an improved accuracy 26-28 and reassigned. Lines in the B / A (1, 5) band originate from above the rst dissociation limit of CO located at 90679.1 cm À1 , 64 and have low intensities due to the competition of emission with predissociation. 70 The wavelengths for these lines are collected in Table 1. All high-J lines located in the perturbation regions, previously analysed 26-28 in 12 C 17 O, were checked carefully with regard to their quality, because these lines are usually weak. Those lines that were too weak and/or blended were removed from the deperturbation analysis. Also, we extended and corrected the assignment of some heavily perturbed or extremely weak lines located in the region of strong and multistate interactions. They are collected in Table 3.
2.2. VUV-FTS of the B 1 S + ) X 1 S + and C 1 S + ) X 1 S + systems We have measured photoabsorption spectra for two bands of 12 C 17 O: B 1 S + ) X 1 S + (0, 0) and C 1 S + ) X 1 S + (0, 0). Their spectra, shown in Fig. 4 and 5, respectively, were recorded at the SOLEIL synchrotron utilising the tunable-undulator radiation source of the DESIRS beamline and its permanently-installed vacuum-ultraviolet Fourier-transform spectrometer. The High resolution absorption spectra of the C 1 S + / X 1 S + (0, 0) Hopfield-Birge band system in the less-abundant 12 C 17 O isotopologue recorded with the VUV-FTS setup at the SOLEIL synchrotron at an instrumental resolution of 0.20 cm À1 . We used two scans at different column density for the lower (red spectrum) and higher (green spectrum) J to get the final list of transition wavenumbers. The estimated absolute calibration 1s uncertainty was 0.005 cm À1 . The 1s uncertainty due to fitting errors of measured wavenumbers (exclusive of calibration uncertainty) was estimated from the least-squares optimisation algorithm and varies between 0.002 and 0.1 cm À1 for the strongest and weakest lines, respectively. The ratio of the gases used in the experiment was 12  characteristics of the beamline and spectrometer are described by Nahon et al. 50 and de Oliveira et al. 48,49 Two roomtemperature spectra were recorded with approximate column densities of 2 Â 10 15 and 6 Â 10 13 cm À2 , and have spectral resolutions of 0.32 and 0.21 cm À1 FWHM, respectively. The lower column density measurement was necessary to avoid saturation of the strongest rotational transitions of C 1 S + ) X 1 S + (0, 0) (as indicated in Fig. 5), and was also used by Stark et al. 25 to determine the oscillator strength of this band.
There is signicant admixture of the 12 C 16 O and 12 C 18 O isotopologues in our gas sample 25 and lines from these isotopologues frequently overlap the transitions of 12 C 17 O. Despite this, we were able to t wavenumbers with an accuracy better than 0.01 cm À1 for many 12 C 17 O transitions by modelling the sinc-function line broadening inherent to Fourier-transform spectrometry, as previously implemented and shown with multiple independent codes. 25,71-73 A brief summary of the steps involved in our spectral modelling is as follows: An initial wavenumber and integrated cross section was assigned to every observed rotational transition in a recorded B ) X or C ) X band, and assuming a column density for each isotopologue component of our spectrum.
A Gaussian wavelength-dependent cross section for each simulated line was calculated from these values, assuming a Doppler width characteristic of the known experimental temperature (FWHM of 0.20 cm À1 for the case of 12 C 17 O and 295 K). The summation of all lines provided a total cross section.
The total cross section was converted into a transmission spectrum by the Beer-Lambert law, then convolved with a sinc function to represent the known instrumental broadening of the FTS, and multiplied by the slightly wavelength dependent synchrotron beam intensity, giving a completely simulated absorption spectrum.
The simulated spectrum was compared with the raw experimental data and model line wavenumbers and cross sections, and isotopologue column densities, were adjusted to minimise the model-to-experiment difference in a pointwise least-squares sense.
The wavenumbers of 12 C 16 O and 12 C 18 O B ) X (0, 0) and C ) X (0, 0) transitions were determined by the analysis of separate spectra recorded with pure samples of those gases. Additionally, the oscillator strengths of the two bands were shown to be independent of isotopic composition and have the rotational dependence of unperturbed 1 S + ) 1 S + transitions. 25 Thus, we could x all details of the individual 12 C 16 O and 12 C 18 O lines in our mixed-gas spectrum while tting the 12 C 17 O lines. The nal assessment of column densities allowed us to estimate the admixture of isotopologues in our mixed sample to be 12 C 17 O : 12  The estimated absolute calibration 1s uncertainty was 0.005 cm À1 . The 1s uncertainty due to tting errors of measured wavenumbers (exclusive of calibration uncertainty) was estimated from the least-squares optimisation algorithm and varies between 0.002 and 0.1 cm À1 for the strongest and weakest lines, respectively. A listing of 122 measured transition wavenumbers is given in Table 4.

Level energies
Rovibronic term values of the B 1 S + (y ¼ 0) and C 1 S + (y ¼ 0) Rydberg states, with regard to the lowest X 1 S + (y ¼ 0) rovibrational level of the 12 C 17 O ground state, were calculated by using the B ) X (0, 0) and C ) X (0, 0) transition frequencies obtained from a VUV-FTS experiment and using the ground state molecular parameters by Coxon et al., 80 given for the 12 C 17 O Table 4 Transition frequencies (in cm À1 ) of the 12 C 17 O B 1 S + ) X 1 S + , and C 1 S + ) X 1 S + absorption bands from the VUV-FTS measurements a J 00 The estimated absolute calibration 1s uncertainty was 0.005 cm À1 . Lines marked with 'w' were weak, and with 'b' were blended in the spectra. Absolute accuracy of the line frequency measurements varies between 0.002 and 0.1 cm À1 for the strongest and weakest lines, respectively. Table 5 Rovibronic term values of the A 1 P (y ¼ 1, 2, 3, 4, and 5), isotopologue. These data were combined with the B / A (this work, and ref. 26 and 27) as well as C / A 28 transition wavenumbers to give term values of the A 1 P (y ¼ 1, 2, 3, 4, and 5) levels as high as J max ¼ 27-30. They were calculated as differences of values of the B 1 S + (y ¼ 0), C 1 S + (y ¼ 0) terms and B / A (0 À y 00 ), C / A (0 À y 00 ) transition frequencies. A similar procedure was adopted to determine terms of the D, I, e, a 0 , and d perturbers in 12 C 17 O using the B 1 S + (y ¼ 0) and C 1 S + (y ¼ 0) level energies and (B 1 S + , C 1 S + ) / (d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D) extra-lines (listed in Table 2). The A 1 P (y) high-J level energies were calculated by means of the deperturbed T y rotationless energies of the A 1 P (y) state from Section 3.2 and relative terms of A 1 P (y) calculated on the basis of B-A 26,27 and C-A 28 bands by means of the linear least-squares method in the version given by Curl and Dane 78 and Watson. 79 The nal values of the A 1 P energy levels are obtained using the weighted average method and are collected in Tables 5 and 6.
In order to display a visual presentation of perturbations occurring in the 12 C 17 O A 1 P (y ¼ 1-5) rovibrational levels, we determined reduced term values T(J) À B A J(J + 1) + D A J 2 (J + 1) 2 of the A 1 P state with the hypothetical unperturbed and crossing perturber levels, where B A and D A refer to deperturbed rotational constants of the corresponding A 1 P level. The reduced term values were calculated in relation to the lowest y ¼ 0 rovibrational level of the 12 C 17 O X 1 S + ground state by means of the term values given in Tables 5 and 6. Those among the reduced terms which we were not able to determine from the experimental data, were calculated on the basis of isotopically recalculated equilibrium molecular constants by Field 30 for d 3 D i , e 3 S À , a 03 S + , and I 1 S À states and by Kittrell et al. 81 for D 1 D state. The T e values were taken from ref. [81][82][83], and the G(y ¼ 0) value for the X 1 S + state in 12 C 17 O, 1068.0310 cm À1 , from Coxon et al. 80 The results are presented in Fig. 6. Identication of perturbers for both e and f L-doubling components of the A 1 P (y ¼ 3, 4, and 5) levels are summarized in Table 7.

Deperturbation analysis of the A 1 P state in 12 C 17 O
In total, 982 transitions from 12 B-A, C-A, B-X, and C-X bands and their extra-lines of 12 C 17 O were used in the global tting procedure. This results in 72 molecular parameters tted for this minor CO species. This analysis is performed, in analogy to deperturbation analyses of the main 12 C 16 O isotopologue, 44,51 using the Pgopher soware. 67 Applying this program we simulated each member of the B(y 0 ¼ 0, 1) À A(y 00 ) and C(y 0 ¼ 0) À A(y 00 ) progressions independently with a parameterised model of the A(y) levels, perturber levels, and their interactions. The computed level positions, line frequencies, and intensities are the result of a matrix diagonalization including all interacting levels. The assignment of perturber levels, the selection of which parameters and interactions could be discriminated from our spectra, and the values of these parameters were iteratively optimised. The Pgopher program 67 uses the effective Hamiltonian with matrix elements similar to Field, 30 Bergeman et al., 84 and Le Floch et al. 31 The model is presented in Table 8. The non-diagonal elements describe the interaction of the A 1 P state with its perturbers, that is the d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D states. Interactions between the perturbing states were neglected. For the A 1 P diagonal element the '+' and 'À' signs relating to L-doubling refer to the e-and f-symmetry states, respectively. T y denotes the rotation-less energies calculated relative to the lowest rovibrational level of the X 1 S + ground state, h i is the spin-orbit interaction parameter, x i is the Luncoupling interaction parameter.
The D 1 D and d 3 D states have nearly degenerate e and f Ldoublet components. The e 3 S À state has two ne structure levels of e type and one f type, while the a 03 S + state has two ne structure levels of f type and one e type. By contrast, the I 1 S À state has only f levels. The interactions between the A 1 P state and the e 3 S À , a 03 S + , and d 3 D triplet states are caused by spinorbit coupling, represented by J-independent matrix elements. Interactions of A 1 P with the I 1 S À and D 1 D singlet states result from L-uncoupling and, therefore, produce heterogeneous interactions with J-dependent matrix elements. 32 It was necessary to adopt some isotopically recalculated molecular constants, using Dunham's relationship within the Born-Oppenheimer approximation, 85 of 12 C 16 O d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D states from ref. 30 and 81, because there are insufficient term-value data for these levels in 12 C 17 O to determine these independently. These values were held xed during the calculations. We only tted molecular constants to those perturber states for which a sufficient number of transitions were observed in the present experiments. All possible vibrational levels of the perturbers which have a non-negligible inuence on the A 1 P, y ¼ 1, 2, 3, 4, and 5 levels were included in the calculation. Some of them do not have crossings with the A 1 P state but still result in recognisable A-state energy level shis.
The frequencies of strong and isolated lines were assigned relative weights of 1.0 during the tting. However, the Table 6 Rovibronic term values of the d 3 D i (y ¼ 11), e 3 S À (y ¼ 4), a 03 S + (y ¼ 10, 13), I 1 S À (y ¼ 3, 6), and D 1 D (y ¼ 1) levels in 12  a All values in cm À1 . Level energies were calculated relative to the lowest y ¼ 0 rovibrational level of the X 1 S + ground state of 12 C 17 O from the combined data sets of two experiments: the VUV-FTS study for the C 1 S + (y ¼ 0) and B 1 S + (y ¼ 0) levels, as well as VIS high-accuracy dispersive optical spectroscopy measurements for the e 3 S À (y ¼ 4), a 03 S + (y ¼ 10, 13), I 1 S À (y ¼ 3, 6), and D 1 D (y ¼ 1) levels. The nal values of the terms were obtained using the weighted average method.
frequencies of weak and/or blended lines have lower accuracy, so they were individually weighted between 0.5 and 0.1, according to the degree of their weakening and/or overlap. Initial ts were made by varying the B, D, H rotational constants and the q L-doubling constant of the A 1 P (y ¼ 1-5) levels. This means that all parity-dependent interactions were included explicitly in the interactions contained in our deperturbation. Any additional L-doubling from remote perturbers was aliased by the interactions included in our perturbation. During the deperturbation, the rotational B and D parameters of the X 1 S + (y ¼ 0) ground state were xed to the values given by Coxon 80 for 12 C 17 O.
The unweighted obs-calc residuals of the tting method are dominated by the uncertainties of the very weak and heavily perturbed lines that belong to the weakest B-A (1, 1) and (1,5) bands. The weighted contribution to the root-mean-square (rms) residual value of high-accuracy dispersive optical spectroscopy and VUV-FTS data is 0.006 cm À1 . This shows that the tting model acceptably reproduces such a comprehensive experimental data set.
In a few cases, tting of the interaction parameters was statistically unjustied because there was an insufficient quantity of experimental transitions in the vicinity of the avoided crossings of the perturbing states or because of the interaction of energetically remote states (for J < 0 or J > J max ) without any observed crossing points with the A 1 P state in 12 C 17 O. In such cases we estimated the semi-empirical interaction parameters making use of the quality suggested in ref. 31,41 Fig. 6 The reduced T(J) À B A J(J + 1) + D A J 2 (J + 1) 2 term values for the 12 C 17 O A 1 P (y ¼ 1-5) levels and for the hypothetical unperturbed crossing rovibronic levels of the perturbers. Filled and open circles indicate e and f electronic symmetry of the A 1 P state, respectively. The reduced level energies (in cm À1 ) were calculated in relation to the lowest y ¼ 0 rovibrational level of the X 1 S + ground state by means of terms calculated in this work (see Tables 5 and 6). Some reduced terms were calculated on the basis of isotopically recalculated equilibrium molecular constants given by Field 30 for d 3 D i , e 3 S À , a 03 S + , and I 1 S À states and by Kittrell et al. 81 for D 1 D state. The T e values were taken from ref. [81][82][83], and the G(y ¼ 0) value for the X 1 S + state in 12 C 17 O, 1068.0310 cm À1 , from Coxon et al. 80 B A and D A symbols refer to deperturbed rotational constants of the respective A 1 P rovibronic level, determined in this work (see Table 10). Note that different reduced-energy scales in cm À1 are used for different vibrational levels of A 1 P.
This journal is © The Royal Society of Chemistry 2016 and 88, which shows that for perturbation between vibronic levels of a given pair of electronic states, the perturbation matrix element (a, b) is the product of a vibrational factor and a constant electronic perturbation parameter (a, b). The effective perturbation parameters a and b, in the e/f basis set, are dened as follows: where H SO and H RE are the spin-orbit and rotation-electronic operators, respectively, and a ¼ h2p|al   a The values in bold correspond to perturbations observed for the rst time in 12 C 17 O. b Theoretically predicted interaction of energetically remote states (for J < 0 or J > J max ) without any observed crossing points with the A 1 P state but the deperturbation t shows that they have a noticeable inuence on the A 1 P (y ¼ 3, 4, or 5) levels (see Table 10). c See Table 10. d Perturbation difficult to identify on the basis of observations only (e.g. Fig. 6) due to much stronger interaction that exists in this region due to the a 03 S + (y ¼ 13) state. Its signicance can be evaluated only on the basis of results of deperturbation t provided in Table 10. e Perturbation difficult to identify on the basis of observations only (e.g. Fig. 6) due to stronger interaction that exists in this region deriving from the F 1 term of the a 03 S + (y ¼ 16) state. Its signicance can be evaluated only on the basis of results of deperturbation t provided in Table 10. f Perturbation difficult to identify on the basis of observations only (e.g. Fig. 6) because of uncharacteristic behaviour of the rovibrational e-parity terms at J ¼ 26-28 due to overlapping interaction with distant a substantially interaction with the F 2 term of the a 03 S + (y ¼ 16) state.

Table 8
Effective Hamiltonian and matrix elements for perturbation analyses of the A 1 P (y ¼ 1, 2, 3, 4, and 5) rovibronic levels and their perturbers in 12 The model is consistent with that of Pgopher soware. 67 b The matrix is symmetric, therefore, the lower le non-diagonal elements, which are not shown in the Hamiltonian, are equivalent to those of the corresponding upper right elements. The matrix elements set to zero are results of an approximation consisting in neglecting the mutual interaction between the perturbing states. For the A 1 P diagonal element the '+' and 'À' signs relating to L-doubling refer to the e-and f-symmetry states, respectively. c T y -denotes the rotation-less energies calculated relative to the lowest rovibrational level of the X 1 S + ground state, h i -spin-orbit interaction parameter, x i -L-uncoupling interaction parameter. The rest of the parameters used are dened in the open literature. 68,86,87 is then possible to calculate initial values of interaction parameters for any pair of levels whenever the relevant vibrational wavefunctions are known. 31 So, the missing perturbation parameters, which were xed during the deperturbation calculation, were estimated on the basis of the isotopologueindependent purely electronic perturbation parameters a and b of Le Floch, 31 as well as hy A |y d,e,or a 0 i vibrational overlap integrals and the hy A |B(R)|y I or D i rotational operator integral in 12 C 17 O, according to eqn (1)- (8). These parameters are presented in Table 9. The vibrational integrals were calculated on the basis of 12  Only those were used that led to noticeable improvements in the quality of the t within the accuracy obtained. A careful examination of the correlation matrix shows satisfactorily low correlations between tted model parameters.
The nal set of deperturbed molecular constants from the ts is presented mainly in Tables 10 and 11. The relationships between the h and a as well as x and b perturbation parameters result from their different denitions, 30,67,94,95 which affect the interaction matrix elements, are as follows: where subscript 'i' indicates A$d, A$e, as well as A$a 0 interactions. The spin-orbit and rotation-electronic parameters obtained from the 12 C 17 O A 1 P (y ¼ 1-5) deperturbation analysis are collected in Table 11. The isotopologue independent, electronic perturbation parameters a and b for the A 1 P$(d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D) interactions are in very good agreement with the values given by Le Floch 31 (see Table 9) Field, 30 and Field et al. 32,87 While performing the deperturbation calculations, we also obtained the rovibrational constants for the B 1 S + (y ¼ 0 and 1) and C 1 S + (y ¼ 0) Rydberg states in 12 C 17 O. The results are given in Table 12. The constants for the B 1 S + and C 1 S + states are compared with analogous values derived in previous studies. [26][27][28] 3.3. Equilibrium constants and transition probabilities in 12 C 17 O Equilibrium constants of the A 1 P state in 12 C 17 O were determined on the basis of the A 1 P (y ¼ 1-5) deperturbed constants summarised in Table 10, using a weighted least-squares method. The results are collected in Table 13 and expressed as Dunham coefficients. Despite the fact that Dunham parameters do not include the parameters that describe perturbations between the zero-order states and they are not expected to t the data to measurement accuracy, they are the most appropriate input to RKR and Franck-Condon Factors (FCF) calculations. It allowed for obtaining the FCF for theÅngström (B 1 S + -A 1 P), Herzberg (C 1 S + -A 1 P) and Fourth positive (A 1 P-X 1 S + ) systems using the deperturbed RKR potential energy curve parameters of the 12 C 17 O A 1 P (this work), B 1 S + (ref. 27), C 1 S + (ref. 28), and X 1 S + (ref. 80) states. The FCFs in 12 C 17 O are provided in Table 14. The strongest perturbations occur because of the spin-orbit interactions with the d 3 D i , a 03 S + , and e 3 S À triplet states. They lead to clearly visible splitting of the L-doublet components in regions of avoiding crossings. This phenomenon is most visible for A 1 P (y ¼ 1) at J ¼ 18-24 caused by a 03 S + (y ¼ 10) with term shis of $2.5 cm À1 , A 1 P (y ¼ 2) at J ¼ 25-32 caused by e 3 S À (y ¼ 4) with maximum term shis of $4 cm À1 , A 1 P (y ¼ 3) at J ¼ 24-30 caused by a 03 S + (y ¼ 13) with maximum term shis of $3 cm À1 , and for A 1 P (y ¼ 5) where we observe a complex perturbation pattern occurring at J ¼ 28-32 resulting from the interactions with the three spin components of d 3 D i (y ¼ 11) and a 03 S + (y ¼ 16) with maximum term shis of about 2.5 cm À1 . In Table 9 Perturbation parameters of the A 1 P$(d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D) interactions, fixed in the 12

Discussion
a The spin-orbit and rotation-electronic perturbation parameters a and b were taken from Le Floch et al. 31 (      contrast, for A 1 P (y ¼ 1) we observe distinct upward shis of only the lowest rovibronic levels, with no signicant effects on the L-doublings, despite the fact that the interaction is of a spin-orbit type. The reason is that this perturbation is caused by the lower lying d 3 D i (y ¼ 5) state, which rapidly diverges with increasing rotation from the 1 P partner. We deal with a similar situation for A 1 P (y ¼ 4), where the perturbation is caused by the D 1 D (y ¼ 5) level, but this is far less noticeable in the presented scale of the plot. It is worth considering the effect of L-doubling caused by a state of S symmetry. However, interactions with the D 1 D and d 3 D states induce perturbations of both e and fparity levels, so do not result in L-doubling. We should also notice the cases of spin-orbit interactions between A 1 P and its e 3 S À , a 03 S + , d 3 D i triplet perturbers, for which negligible L-doubling effects are observed, in spite of the   Table 3.4 in ref.
87). f Calculated on the basis of Field's data 30 Table 11 Spin-orbit and rotation-electronic parameters obtained from deperturbation analysis of the A 1 P, y ¼ 1-5 levels in 12   fact that the crossings occur within the observed 0 < J < 35 region. We deal with such a case for the A 1 P, y ¼ 3 and 5 levels where the perturbers are d 3 D i (y ¼ 8), and e 3 S À (y ¼ 8), respectively. The reduced effects are in this case caused by the very small values of the vibrational integrals for the interacting levels in 12 C 17 O (see Table 9). In turn, the L-uncoupling interactions between the A 1 P state and I 1 S À , D 1 D singlet states are usually much weaker. We can notice these interactions distinctly in Fig. 6b-d, where there are interactions of A 1 P (y ¼ 2) with I 1 S À (y ¼ 3), and A 1 P (y ¼ 3) with D 1 D (y ¼ 4) as well as A 1 P (y ¼ 4) with I 1 S À (y ¼ 6). In all these cases the largest term shis do not exceed 0.5 cm À1 , which can be classied as weak interactions.
In Table 12, with the high accuracy of the results obtained, we notice a slight inconsistency of rotational constants B y and D y of B 1 S + (y ¼ 0 and 1) and C 1 S + (y ¼ 0) in relation to those that were calculated in our previous works. [26][27][28] This could be caused by the fact that the linear least-squares method in the version given by Curl and Dane 78 and Watson 79 takes no account of the impact of the Q(J) branches in the singlet-singlet ts. Improvement in the assignment of some of the heavily overlapped and/or extremely weak lines located in the region of strong and multistate perturbations, which was described in Section 2.1, could also be a reason for this inconsistency. It is worth noticing here that the deperturbation analysis conducted in this work was based on a global, three times more extensive experimental data set than was used in other works concerning the less-abundant 12 C 17 O isotopologue. [26][27][28] The present work also allowed for verication and improvement in the observed perturbations of the A 1 P, y ¼ 1, and 2 rovibrational levels in 12 C 17 O presented in ref. 26. For the A 1 P, y ¼ 1 level, the A 1 P (y ¼ 1)$D 1 D (y ¼ 1) avoiding crossing occurs at J ¼ 26-27, both for the e-and f-symmetry levels (see Fig. 6a). However, in the case of the A 1 P, y ¼ 2 level, it turns out that in the perturbation analysis we must take into account small, but not negligible, impacts of the a 03 S + (y ¼ 11) and D 1 D (y ¼ 2) states on its band origin and the fact that the maximum of the A 1 P (y ¼ 1)$e 3 S À (y ¼ 4; F 3 ) interaction for the esymmetry levels falls at J ¼ 31-32, and not at J ¼ 30-31 as had been thought (see Fig. 6b).
It can be seen in Table 10 that the energy levels, T y , for A 1 P (y ¼ 1) and A 1 P (y ¼ 4) have larger uncertainties than the remaining rovibrational levels of this state. This could be due to uncertainties derived from interactions with the d 3 D i (y ¼ 5) and D 1 D (y ¼ 5) states, respectively. It is important to note that the rotational progressions of these states do not cross the A 1 P (y ¼ 1) and A 1 P (y ¼ 4) states. The effects of such interactions result in global energy shis of the A 1 P (y ¼ 1, and 4) states, just as in the case of vibrational perturbations. 86 Thus, these interactions translate directly into uncertainties in T y .
There is a very good agreement between the present and Le Floch's, 31 Field's, 30 and Field's et al. 32,87 values of the isotopologue independent electronic perturbation parameters a and b for the A 1 P$(d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D) interactions, highlighted in Tables 9 and 11. The obtained electronic perturbation parameters can be used to predict perturbations in other A 1 P levels of all CO isotopologues. These parameters may be helpful in interpreting laboratory and astrophysical spectra of higher levels of the A 1 P state.

Conclusion
Two different experimental methods, high-accuracy dispersive optical spectroscopy in the visible region and Fourier-transform spectroscopy in the vacuum ultraviolet region, were used to obtain high-resolution spectra of the B 1 S + / A 1 P, B 1 S + ) X 1 S + , and C 1 S + ) X 1 S + systems in the less-abundant 12 C 17 O isotopologue; a total of 429 high-accuracy transition Table 13 Deperturbed equilibrium molecular constants of the A 1 P state in 12 C 17 O a,b,c frequencies were measured. The combined current data and our recent results, [26][27][28] in total 982 lines in 12 bands (B-A, C-A, B-X, C-X) and 15 bands consisting of extra-lines, were used to perform deperturbation analysis of the A 1 P state in 12 C 17 O, taking into account the complete impacts of the d 3 D i , e 3 S À , a 03 S + , I 1 S À , and D 1 D states. As a result the accurate perturbation model describes our experimental ndings to the quantum level energies of accuracy. Table 14 Franck-Condon Factors (FCF) of the B 1 S + -A 1 P, C 1 S + -A 1 P, and A 1 P-X 1 S + band systems in the 12 C 17 O isotopologue A 1 P (y 00 )