Voigt Profile Research Articles

The physical nature of circumgalactic medium absorbers in Simba

ABSTRACT We study the nature of the low-redshift circumgalactic medium (CGM) in the Simba cosmological simulations as traced by ultraviolet absorption lines around galaxies in bins of stellar mass ($\mbox{$M_\star $}\, > 10^{10}{\rm M}_\odot$) for star-forming, green valley and quenched galaxies at impact parameters r⊥ ≤ 1.25r200. We generate synthetic spectra for H i , Mg ii , C ii , Si iii , C iv , and O vi , fit Voigt profiles to obtain line properties, and estimate the density, temperature, and metallicity of the absorbing gas. We find that CGM absorbers are most abundant around star-forming galaxies with $\mbox{$M_\star $}\, < 10^{11}\,\,{\rm M}_{\odot }$, while the abundance of green valley galaxies show similar behaviour to those of quenched galaxies, suggesting that the CGM ‘quenches’ before star formation ceases. H i absorbing gas exists across a broad range of cosmic phases [condensed gas, diffuse gas, hot halo gas, and Warm-Hot Intergalactic Medium (WHIM)], while essentially all low ionization metal absorption arises from condensed gas. O vi absorbers are split between hot halo gas and the WHIM. The fraction of collisionally ionized CGM absorbers is $\sim 25{\text{--}}55{{\ \rm per\ cent}}$ for C iv and $\sim 80{\text{--}}95{{\ \rm per\ cent}}$ for O vi , depending on stellar mass and impact parameter. In general, the highest column density absorption features for each ion arise from dense gas. Satellite gas, defined as that within 10r1/2,⋆, contributes $\sim 3{{\ \rm per\ cent}}$ of overall H i absorption but $\sim 30{{\ \rm per\ cent}}$ of Mg ii absorption, with the fraction from satellites decreasing with increasing ion excitation energy.

Line shape parameters of air-broadened 12CO2 transitions in the 2.0 µm region, with their temperature dependence

The 20013–00001 and 20012–00001 bands of 12C16O2 in the 2 µm region are used by many instruments on-board satellites and in ground-based networks to monitor the CO2 column-averaged dry air mole fraction in the Earth's atmosphere. An accurate knowledge of the corresponding spectroscopic parameters and their temperature dependence is thus needed to map sources and sinks of carbon dioxide. In this work, we retrieve the line-shape parameters and their temperature dependence exponents/coefficients for the P(16) and P(28) transitions of the 20012–00001 band, and the R(24) and R(30) transitions of the 20013–00001 band. These parameters are obtained from a multi-spectrum fit procedure using a speed-dependent Nelkin-Ghatak profile including the line-mixing effect in its first-order approximation. High signal-to-noise spectra (quality factor between typically 3000 and 6000) are recorded at different conditions of pressure (from 50 to 750 Torr) and temperature (from 245 to 330 K) for mixtures of CO2 in air using a comb-assisted cavity ring down spectrometer with a temperature stabilized high finesse cavity. Air-broadening coefficients are measured with an estimated uncertainty better than 0.1% and absolute frequencies of the transitions are obtained with uncertainty better than 100 kHz. The determined line-shape parameters are consistent with previous values derived in the 1.6 µm region confirming the absence of significant vibrational dependence (except for the air-pressure shift coefficient). The retrieved parameters agree well with literature values by Fourier transform spectroscopy and classical molecular dynamic simulations. The parameters of the speed-dependent Voigt profiles provided by the HITRAN2020 database are mostly validated. The comparison to the CO2 line parameters adopted for the OCO missions is discussed.

Pressure dependence of the measured line intensity and super-Lorentzian effects in the absorption spectra of pure HCl.

Super-Lorentzian effects in the troughs between lines and the pressure dependence of the line intensities retrieved from fits of absorption spectra of pure HCl have been investigated both experimentally and theoretically. For that, spectra of pure HCl gas in the 2-0 band were recorded with a Fourier Transform spectrometer at room temperature and for pressures ranging from 1 to 10 atm. The line intensities, retrieved from fits of the measurements with the Voigt profile using a single spectrum fitting technique, reveal large decreases with increasing pressure - up to 3% per atm - with a relatively weak rotational dependence. We also show that the absorptions in-between successive P and R transitions are significantly larger than those predicted using Voigt profiles. Requantized classical molecular dynamics simulations were made in order to predict absorption spectra of pure HCl matching the experimental conditions. The pressure dependence of the intensities retrieved from the calculated spectra as well as the predicted super-Lorentzian behavior between lines are in good agreement with the measurements. Our analysis shows that these effects are essentially due to incomplete collisions, which govern the dipole auto-correlation function at very short times.

Low-temperature vibronic spectroscopy of condensed chromophore exhibiting inhomogeneous distribution of vibrational frequencies in a mixed quantum-classical environment.

This work has been motivated by the recent paper by the author [M. Toutounji, Phys. Chem. Chem. Phys., 2021, 23, 21981] whereby a mixed quantum-classical Liouville equation was used to probe the spectroscopy and dynamics of a spin-boson system. A mixed quantum-classical Liouville equation treats the system of interest quantum mechanically, the bath classically, and the coupling term mixed quantum-classical mechanically. This paper offers a two-fold advantage: correcting the treatment of the electronic transition decay (width in frequency domain) and assessing the local heterogeneous vibrational structure. The homogeneous linear absorption spectrum of a chromophore embedded in a mixed quantum-classical environment at low temperature is composed of a sharp peak called a zero-phonon line (ZPL) and a broad phonon sideband (PSB), whereby the ZPL and the PSB are assimilated by a Lorentzian function and Voigt profiles, respectively. The PSB, in this case, is characterized by a local heterogeneous structure due to a dispersive medium of vibrations, modeled by vibrational Gaussian distributions to represent the arising inhomogeneous broadening and Lorentzians to model the homogeneous vibrations. This description seems to model proteins and amorphous solids exhibiting a local heterogeneous structure as both electronic and vibrational inhomogeneous broadening seems to be large in these media. This work provides a derivation of linear absorption lineshape and vibronic transition dipole moment time correlation functions, both of which account for pure electronic dephasing (ZPL width) the Voigt profile description of the phonon profiles (PSB) in dispersive media.

Line-shape parameters and line mixing for the low-intensity high-J oxygen transitions

We present the results of investigations of the oxygen B-band R-branch transitions with intensities in the range from 1.5×10−26 to 2.5×10−28 cm−1/(molecule/cm2) and the total angular momentum quantum number for the lower state, J”, up to 38. High resolution CRDS laboratory spectra were fitted with the Voigt profile as well as with advanced line-shape functions consistent with the Hartmann–Tran profile, that is the speed-dependent Voigt profile and speed-dependent Nelkin–Ghatak profile using the multispectrum fitting technique. The line positions, line intensities and other line-shape parameters, together with our previous data, were collated to show trends for the whole R branch. Using the determined frequencies of the B-band transitions, the Dunham fit was performed to obtain spectroscopic parameters for both ground and excited states. Newly recorded spectra at the band head, with the signal-to-noise ratio above 20000, were analyzed with line profiles including also the line mixing. First-order line mixing determined directly from the line shape fitting at relatively low pressure (32 Torr) is in good agreement with values calculated by Sung et al. [1].

High precision measurement of spectroscopic parameters of H<sub>2</sub>S in 6320—6350 cm<sup>–1</sup> band

As a highly corrosive and highly toxic gas, hydrogen sulfide (H<sub>2</sub>S) is an important intermediate product or pollutant in many fields such as chemical industry, energy and environment. Accurate online measurement of its concentration is of great significance for process control and production safety. Tunable diode laser absorption spectroscopy (TDLAS), as a quantitative absorption spectroscopy technique, is suitable for high-precision on-line measurement of H<sub>2</sub>S concentration in atmospheric environmental monitoring and industrial processes control. Considering that most of the spectroscopic parameters of H<sub>2</sub>S in the HITRAN2020 database are mainly calculated based on semi-empirical theoretical model and the experimental data to support them are lacking. In this work, direct absorption spectroscopy (DAS) method is firstly used to measure the absorption spectra of H<sub>2</sub>S in the band of 6320–6350 cm<sup>–1</sup>. Six groups of characteristic lines with strong absorption and relative independence are selected as the target transitions for experimental measurement. Then, the wavelength modulation-direct absorption (WM-DAS) method with no calibration and high signal-to-noise ratio is used to measure the absorbances of the six groups of transitions under different pressures. Voigt, Raution and quadratic speed-dependent Voigt profiles fit the measured absorbances by least squares method in order to obtain the spectroscopic parameters such as the collision broadening coefficient, line strength and Dicke narrowing coefficient. And the minimum standard deviation of residual error of absorbances is 7×10<sup>–5</sup>. The measurement uncertainty of each line strength is less than 2%, and the uncertainty of collision broadening coefficients, Dicke narrowing coefficients and the speed-dependent coefficients are all less than 10%. This work is helpful in improving the H<sub>2</sub>S spectral database and providing the spectral data basis for the high-precision measurement of H<sub>2</sub>S concentration.

Q-Gaussian Tsallis Line Shapes and Raman Spectral Bands

q-Gaussians are probability distributions having their origin in the framework of Tsallis statistics. A continuous real parameter q is characterizing them so that, in the range 1 < q < 3, the q-functions pass from the usual Gaussian form, for q close to 1, to that of a heavy tailed distribution, at q close to 3. The value q=2 corresponds to the Cauchy-Lorentzian distribution. This behavior of q-Gaussian functions could be interesting for a specific application, that regarding the analysis of Raman spectra, where Lorentzian and Gaussian profiles are the line shapes most used to fit the spectral bands. Therefore, we will propose q-Gaussians with the aim of comparing the resulting fit analysis with data available in literature. As it will be clear from the discussion, this is a very sensitive issue. We will also provide a detailed discussion about Voigt and pseudo-Voigt functions and their role in the line shape modeling of Raman bands. We will show a successfully comparison of these functions with q-Gaussians. The role of q-Gaussians in EPR spectroscopy (Howarth et al., 2003), where the q-Gaussian is given as the

L-band dual-comb based on gain-switched lasers for CO2 monitoring

This paper reports the development of a dual-comb spectroscopy system based on gain-switched single-mode laser diodes emitting at 1572 nm. The switching is performed by a train of short electrical pulses at a repetition rate of 100 MHz while the lasers are subjected to external optical injection. Under these conditions, broad and Hat optical frequency combs (OFC) are generated, showing 500 tones within 10 dB and a line spacing fix ed by the repetition rate of the swi tching pulses. The dual-comb gene rated by th e interference of two such combs with slightly different line spacing, is used for measuring the absorption line of CO2 at 1572.02 nm. The transmission spectrum is analyzed and fitted to a Voigt profile. A deviation of 3% in the residuals of the fitting is obtained, denoting the high spectral performance of the spectroscopy system.

Single Spectral Line Method for TDLAS Imaging of Temperature and Water Vapor Concentration

A single spectral line method for gas temperature as well as concentration extraction was proposed by using iterations with a fuzzy proportional–integral-derivative (PID) controller in tunable diode laser absorption spectroscopy (TDLAS). Usually, two or more spectral lines are essential to extract temperatures from coupled gas concentration and pressure. The proposed method utilized the line shape of a single absorption spectrum to decouple gas parameters along the laser path simultaneously and simplified the optical implementations. Differential equations of the Voigt profile were introduced at multiple wavenumbers to extract gas parameters from a single spectral line. A fuzzy PID controller was utilized to iteratively achieve the single spectral line from widely existed overlapping spectra. A relative temperature error within 2.5% at reference temperatures ranged from 300 to 2300 K was achieved by using the spectral line of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$7444.35\,\,{\mathrm {cm}}^{-1}$ </tex-math></inline-formula> from noise free data. At low temperatures below 350 K, the proposed method using the single spectral line of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$7181.14\,\,{\mathrm {cm}}^{-1}$ </tex-math></inline-formula> yielded smaller errors in noisy cases, and the temperature errors were verified within 5 °C on the water-based thermostat below 100 °C using the single spectral line. Experiments of axisymmetric counterflow flames at different heights were also implemented. The proposed method effectively reconstructed distributions of flame temperature and water vapor concentration inside, by only using a single spectral line of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$7444.35\,\,{\mathrm {cm}}^{-1}$ </tex-math></inline-formula> .

Robust Frequency Estimation Under Additive Symmetric α-Stable Gaussian Mixture Noise

Here the estimating problem of a single sinusoidal signal in the additive symmetric α-stable Gaussian (ASαSG) noise is investigated. The ASαSG noise here is expressed as the additive of a Gaussian noise and a symmetric α-stable distributed variable. As the probability density function (PDF) of the ASαSG is complicated, traditional estimators cannot provide optimum estimates. Based on the Metropolis-Hastings (M-H) sampling scheme, a robust frequency estimator is proposed for ASαSG noise. Moreover, to accelerate the convergence rate of the developed algorithm, a new criterion of reconstructing the proposal covariance is derived, whose main idea is updating the proposal variance using several previous samples drawn in each iteration. The approximation PDF of the ASαSG noise, which is referred to the weighted sum of a Voigt function and a Gaussian PDF, is also employed to reduce the computational complexity. The computer simulations show that the performance of our method is better than the maximum likelihood and the -norm estimators.

In Ukrainian

The study of the microstructure of nanocrystalline substances by the method of powder diffractometry based on the physical broadening of diffraction lines involves the use of diffraction data of standard polycrystalline samples, preferably, one nature with the test samples with the size of crystallites exceeding 100 nm. In the absence of such standards, researchers resort to the existing dependence of the width of diffraction peaks on the angle of diffraction for the standard sample or the construction of theoretical instrumental profiles due to the collimation parameters of x -rays used. In this paper a comparative study of the microstructure of nanocrystalline titanium oxide (anatase), tin oxide iron oxide (magnetite), synthesized in various ways, using several methods of analysis of powder diffractograms, was carried out. To evaluate the average crystallite sizes of the studied oxides, the Sherer equation with a graphical method of determining the width of instrumental profile and the influence of dublet radiation was chosen. Methods of profile analysis of diffraction spectra, such as a method of whole profile modeling of powder diffractograms (WPPM) and the chord method, were used to construct crystallite size distribution functions and determine the average size of crystallites of the oxides. Modeling of instrumental diffraction profiles of titanium, tin and iron oxides was performed using X -rays collimation parameters determinated using a polycrystalline silicon as standard and pseudo -Voigt function, which best describes the form of diffraction peak. The crystallite size distribution functions were constructed by means of WPPM and chords methods based on the instrumental profiles. It has been found that the values of average size of the crystallites, obtained by the methods of Sherer, WPPM and chords, differ within the one order of magnitude for each oxide. Thus, for titanium oxide this value is within 12-18 nm, for tin oxide within 7-10 nm, the iron oxide of iron within 9-12 nm. Analysis of size crystallite distribution functions and average sizes of the crystallites of the studied oxides showed the advisability of using different methods of studying microstructure to clarify the true type of the size crystallites distribution and establish its connection with the conditions and the synthesis method.

Searching for dilaton fields in the Lyman- α forest

Dilatons (and moduli) couple to the masses and coupling constants of ordinary matter, and these quantities are fixed by the local value of the dilaton field. If, in addition, the dilaton with mass ${m}_{\ensuremath{\phi}}$ contributes to the cosmic dark matter density, then such quantities oscillate in time at the dilaton Compton frequency. We show how these oscillations lead to broadening and shifting of the Voigt profile of the Lyman-$\ensuremath{\alpha}$ forest in a manner that is correlated with the local dark matter density. We further show how tomographic methods allow the effect to be reconstructed by observing the Lyman-$\ensuremath{\alpha}$ forest spectrum of distant quasars. We then simulate a large number of quasar lines of sight using the lognormal density field, and forecast the ability of future astronomical surveys to measure this effect. We find that in the ultra low mass range ${10}^{\ensuremath{-}32}\text{ }\text{ }\mathrm{eV}\ensuremath{\le}{m}_{\ensuremath{\phi}}\ensuremath{\le}{10}^{\ensuremath{-}28}\text{ }\mathrm{eV}$ upcoming observations can improve over existing limits to the dilaton electron mass and fine structure constant couplings set by fifth force searches by up to five orders of magnitude. Our projected limits apply assuming that the ultralight dilaton makes up a few percent of the dark matter density, consistent with upper limits set by the cosmic microwave background anisotropies.

Use of Fourier Series in X-ray Diffraction (XRD) Analysis and Fourier-Transform Infrared Spectroscopy (FTIR) for Estimation of Crystallinity in Cellulose from Different Sources

Cellulose crystallinity can be described according to the crystal size and the crystallinity index (CI). In this research, using Fourier-transform infrared spectroscopy (FTIR) and X-ray diffraction (XRD) methods, we studied the crystallinity of three different types of cellulose: banana rachis (BR), commercial cellulose (CS), and bacterial cellulose (BC). For each type of cellulose, we analyzed three different crystallization grades. These variations were obtained using three milling conditions: 6.5 h, 10 min, and unmilled (films). We developed a code in MATLAB software to perform deconvolution of the XRD data to estimate CI and full width at half-maximum (FWHM). For deconvolution, crystalline peaks were represented with Voigt functions, and a Fourier series fitted to the amorphous profile was used as the amorphous contribution, which allowed the contribution of the amorphous profile to be more effectively modeled. Comparisons based on the FTIR spectra and XRD results showed there were no compositional differences between the amorphous samples. However, changes associated with crystallinity were observed when the milling time was 10 min. The obtained CI (%) values show agreement with values reported in the literature and confirm the effectiveness of the method used in this work in predicting the crystallization aspects of cellulose samples.

Mechanical Amorphization of Chitosan with Different Molecular Weights.

Mechanical amorphization of three chitosan samples with high, medium, and low molecular weight was studied. It is shown that there are no significant differences between the course of amorphization process in a planetary ball mill of chitosan with different molecular weights, and the maximum degree of amorphization was achieved in 600 s of high intensity mechanical action. Specific energy consumption was 28 kJ/g, being comparable to power consumption for amorphization of cellulose determined previously (29 kJ/g) and 5–7-fold higher than that for amorphization of starch (4–6 kJ/g). Different techniques for determining the crystallinity index (CrI) of chitosan (analysis of the X-ray diffraction (XRD) data, the peak height method, the amorphous standard method, peak deconvolution, and full-profile Rietveld analysis) were compared. The peak height method is characterized by a broader working range but provides deviated CrI values. The peak deconvolution method (with the amorphous Voigt function) makes it possible to calculate the crystallinity index of chitosan with greater accuracy, but the analysis becomes more difficult with samples subjected to mechanical processing. In order to refine the structure and calculation of CrI by the Rietveld method, an attempt to optimize the structure file by the density functional theory (DFT) method was performed. The averaged profile of amorphous chitosan approximated by an eighth-order Fourier model improved the correctness of the description of the amorphous contribution for XRD data processing. The proposed equation may be used as a universal standard model of amorphous chitosan to determine the crystallinity index both for the amorphous standard method and for peak deconvolution of XRD patterns for arbitrary chitosan samples.

Calculation of the absorption coefficient of air at different heights using the HITRAN and ACE-FTS databases

Subject of study. Increasing requirements for the accuracy of modeling the brightness characteristics of the atmosphere at different heights motivate the development of more detailed methods for the calculation of the absorption coefficient of air. Method. Modern databases of satellite sounding of the atmosphere that contain information on the volume fractions of air components at different heights in combination with a database of the parameters of the spectral lines of gases can be used in this task to calculate the spectral absorption coefficient. The possibility of specifying the geographic position of the height profiles of the air components (by specifying latitude and longitude) is an additional advantage of using the databases of satellite sounding. This method enables the differences in air compositions above different regions of the Earth to be considered. These databases usually provide information starting from a certain height; thus, databases of satellite sounding must be combined with the data obtained either from ground-based meteorological stations or using model distributions. The ACE-FTS database was used in this study, and it was combined with the standard atmosphere MODTRAN models at low heights. A filter for the ACE-FTS database was implemented using C++. The HITRAN database was used to calculate absorption coefficients. The ACE-FTS database provides data for 43 components of air, and the HITRAN database contains data for line-by-line calculation for 27 of these components and data on spectral absorption cross-sections for the remaining 16 components. Main results. A program for the calculation of the absorption coefficient of a multicomponent gas mixture was developed using C++. A Voigt profile was used in the line-by-line calculation, the effects of pressure and temperature on the line shift were considered, and line broadening owing to the following processes was considered: self-broadening, broadening resulting from interaction with water and carbon dioxide molecules, and broadening owing to interaction with other components of the air. The software explicitly considers the isotopic composition of molecules, thus enabling the calculation of not only the air mixture but also an arbitrary mixture with specified isotopic composition, information on which is in the HITRAN database. Moreover, the data of the high-temperature version HITEMP are considered. Practical significance. The combination of the ACE-FTS and HITRAN databases enables the calculation of a detailed profile of the spectral absorption coefficient of air for the specified month, temperature at the surface of the Earth, and geographic latitude and longitude, thus increasing the information content of the input data for the calculation of the brightness and transmission of the atmosphere at different heights. The data readout rate (parsing) and calculation speed of the developed program and the program HAPI implemented using Python are compared.

PRESSURE DEPENDENCES OF LINE PARAMETERS OF THE 𝝂𝟑 + 𝝂𝟒 R(7)F1 TRANSITION OF METHANE DILUTED IN NITROGEN USING THE HARDAND SOFT-COLLISION MODELS

In this work, the affection of the Dicke narrowing effect on the spectral shape of the 𝜈3 + 𝜈4 R(7)F1 transition of methane diluted in nitrogen was studied using the Hard- and the Soft- collision line-shape models. Room-temperature absorption spectra of CH4 broadened by N2 were measured at a wide range of pressure using a difference-frequency laser spectrometer. Measured spectra of the studied transition were then one-by-one fitted with the simple Voigt profile as well as the two line-shape models considered the Dicke narrowing effect at two different assumptions of velocity-changing collisions: hard- and soft- collisions. The obtained results confirmed that using these two models leads to better qualities of fit with respect to the Voigt profile. Pressure-dependences of the line-shape parameters were also observed for the considered transition. The results showed a good agreement with other studies.

Water vapor absorption line parameters in the 6760–7430 cm–1 region for application to CO2-rich planetary atmosphere

The absorption spectra of the mixtures of water vapor and carbon dioxide at five different partial pressures of both gasses have been recorded using the Bruker IFS 125 HR FTIR spectrometer at room temperature in the 6760–7430 cm–1 spectral region. The multispectrum fitting procedure has been applied to these spectra to recover the spectral line parameters. To obtain the spectral lines parameters the quadratic speed-dependent Voigt profile was used. The water vapor line broadening and shift coefficients for lines of three vibrational bands ν1+ν3, 2ν2+ν3, and 2ν1 were obtained.

Fundamental correction of self-absorption effect on the element concentration measurement in laser induced plasmas

For testing the fundamental correction effect of self-absorption (SA) on measurement of the element concentration in laser induced plasmas (LIPs), the LIPs spectra of soil are measured using a typical experiment set up. The SA coefficient (SAλ) of a spectral line is directly obtained from the fitting of the Voigt function with the experiment data of them, which are used to correct the intensity of the spectral line, electron temperature, electron density, and the element concentration of the LIPs. The nine elements (Al, Ca, Fe, K, Mg, Na, Li, Cu, Ti) concentrations in the soil are analyzed, which are closer to the results of chemical methods than that obtained without considering SA correction. Especially, SAλ of some lines (Mg I at 518.36 nm, Na I at 497.85 nm, Cu I at 324.75 nm) are equal to 1, which do not affect concentrations of corresponding elements, but the concentrations accuracies of them are still improved. This indicate the elements concentrations not only are corrected by the intensity of the spectral lines, but also are corrected by plasmas properties. So, the SA effect on element concentration can be fundamental corrected in this method.

Collisional broadening and shift coefficients for (n+1)s[3/2]2 → (n+1)p[1/2]1 argon and neon lines in parent and foreign rare gases

Results of simultaneous measurements of pressure shift β and broadening ξ coefficients for 912.3 nm Ar and 703.2 nm Ne spectral lines of the (n+1)s[3/2]2 → (n+1)p[1/2]1 transition are presented. This is a lasing transition in a recently proposed optically pumped rare gas laser. Absorption line shapes fitted to Voigt profiles permitted to assess gas temperature by determining Doppler width components. These coefficients for the Ar 912.3 nm line with He and Ne as the collisional partners were determined for the first time. The values of shift and broadening coefficients in units of 10−10 s-1 cm3, reduced to 300 K are: βAr-Ar = -2.11±0.05, βAr-Ne = -1.21±0.05, βAr-He = 0.37±0.03, βNe-Ne = -0.78±0.05, βNe-He = 0.0±0.05; ξAr-Ar = 2.8±0.1, ξAr-Ne = 1.68±0.05, ξAr-He = 3.7±0.1, ξNe-Ne = 1.94±0.05, ξNe-He = 3.18±0.05.

Harvesting the Ly α forest with convolutional neural networks

ABSTRACT We develop a machine learning based algorithm using a convolutional neural network (CNN) to identify low H i column density Ly α absorption systems (log NH i/cm−2 &amp;lt; 17) in the Ly α forest, and predict their physical properties, such as their H i column density (log NH i/cm−2), redshift (zH i), and Doppler width (bH i). Our CNN models are trained using simulated spectra (S/N ≃ 10), and we test their performance on high quality spectra of quasars at redshift z ∼ 2.5−2.9 observed with the High Resolution Echelle Spectrometer on the Keck I telescope. We find that ${\sim}78{{\ \rm per\ cent}}$ of the systems identified by our algorithm are listed in the manual Voigt profile fitting catalogue. We demonstrate that the performance of our CNN is stable and consistent for all simulated and observed spectra with S/N ≳ 10. Our model can therefore be consistently used to analyse the enormous number of both low and high S/N data available with current and future facilities. Our CNN provides state-of-the-art predictions within the range 12.5 ≤ log NH i/cm−2 &amp;lt; 15.5 with a mean absolute error of Δ(log NH i/cm−2) = 0.13, Δ(zH i) = 2.7 × 10−5, and Δ(bH i) = 4.1 km s−1. The CNN prediction costs &amp;lt; 3 min per model per spectrum with a size of 120 000 pixels using a laptop computer. We demonstrate that CNNs can significantly increase the efficiency of analysing Ly α forest spectra, and thereby greatly increase the statistics of Ly α absorbers.