betaFeedback

WorksPhysics and Astronomy

Estimating LuSEE-Night's Sensitivity to Cα Radio Recombination Lines

Abstract

Carbon radio recombination lines (CRRLs) arising from the cold neutral medium (CNM) of the interstellar medium are potential foregrounds in the 1-50 MHz band that LuSEE-Night will observe from the lunar far side. CRRLs are sensitive probes of the temperature, density, and radiation field of the CNM, and LuSEE-Night will provide sky-averaged measurements in a poorly explored frequency range. We model CRRLs across this band, accounting for doppler, collisional, and radiation broadening, and we estimate line strengths using two independent methods - fixing either the peak or the frequency-integrated optical depth - applied both uniformly across the sky and with region-dependent values informed by prior low-frequency observations. We simulate LuSEE-Night's expected response to these CRRLs and further convolve the beam-diluted spectra with LuSEE-Night's polyphase-filter-bank spectrometer response, including its zoomed-mode frequency resolution of ∼\sim390 Hz, to obtain the expected observed brightness temperature. Using these simulated spectra, we estimate the integration time required to detect individual CRRLs at SNR=5 across the six sky models considered. Required integration times range from as short as 30 hours (in the most optimistic model where the optical depth across the sky is ∼\sim10−3^{-3}) to 27.6 years (in the most conservative models where the optical depth across the sky is ∼\sim10−4^{-4}). Stacking the 361 simulated Cαα lines between 10 and 50 MHz reduces the integration time needed for an SNR=5 detection to ∼\sim11 minutes and ∼\sim151 hours respectively, indicating that LuSEE-Night has realistic chances of making a stacked detection of Cαα RRLs within a realistic mission lifetime, even under the most conservative optical-depth assumptions.

The LuSEE-Night collaboration

Introduction

Radio recombination lines (RRLs) of ionized hydrogen and carbon atoms in the galaxy are powerful diagnostics used to study the interstellar medium (ISM) of our Galaxy, providing constraints on parameters such as electron density, temperature, and emission measure (F. Salgado et al. [26]). RRLs have been observed from hydrogen (HRRLs), carbon (CRRLs), helium (HeRRLs), and heavier elements (XRRLs), spanning wavelengths from the submillimeter to decameter band (J. Oonk et al. 2019). Although the profiles of RRLs depend on the atomic species and the physical conditions of the gas from which they arise, they can be broadly classified as either spontaneous or stimulated. Spontaneous RRLs are associated with H ii regions and are observed above 1 GHz, while stimulated RRLs dominate lower frequencies and originate from diffuse, partially ionized gas where local thermodynamic equilibrium (LTE) conditions do not hold.

Previous observations of RRLs have targeted radio-bright sources, especially Cassiopeia A (Cas A) and Cygnus A (Cyg A). Toward Cas A, S. V. Stepkin et al. [29] detected carbon RRLs (CRRLs) near 26 MHz using the UTR-2 radio telescope, including α\alpha, β\beta, γ\gamma, and δ\delta transitions with principal quantum number n≈1009n \approx1009. Later, P. Salas et al. [24] observed RRLs toward this same source in the 33–78 MHz and 304–386 MHz ranges using LOFAR. P. Salas et al. [25] extended these measurements to even lower frequencies and detected individual Cα\alpha absorption lines, as low as 16 MHz and stacked detections between 11 and 16 MHz. Along with observations of the 158 μ\mum [C ii] fine-structure line, these measurements constrained the electron temperature of the line-of-sight clouds to Te=68T_e = 68–98 K and electron density to ne=0.02n_e = 0.02–0.035 cm−3^{-3}, indicating that low-frequency CRRLs arise from cold, low-density gas. J. B. R. Oonk et al. [22] used LOFAR to detect stacked Cα\alpha absorption along the line-of-sight of Cyg A between 33 and 57 MHz. They found Te∼110T_e \sim110 K and ne=0.06n_e = 0.06 cm−3^{-3}, implying that the lines came from the cold neutral medium (CNM) of the Orion arm of the Milky Way. L. Cros et al. [7] observed both sources with NenuFAR (the New Extension in Nançay Upgrading LOFAR telescope), between 10 and 85 MHz with higher spectral resolution and SNR than previous LOFAR detections. Toward Cas A, they detected hundreds of individual lines, while toward Cyg A, they detected RRLs by stacking groups of lines. From these observations, they inferred somewhat colder temperatures of Te=33T_e = 33–76 K and similarly low electron densities of ne=0.015n_e = 0.015–0.035 cm−3^{-3}. EDGES (Experiment to Detect the Global Epoch of Reionization Signature) measured beam-averaged spectra with 12.2 kHz spectral resolution across 50–87 MHz. A. K Vydula et al. [15] reported stacked Cα\alpha RRLs with peak amplitudes between −795±40-795 \pm40 mK toward the inner Galactic Plane (GP) and −33±11-33 \pm11 mK away from the GP, demonstrating that CRRLs become weaker as you move away from the GP. Their observations were most consistent with non-LTE models, as expected for low-frequency CRRLs.

The Lunar Surface Electromagnetics Experiment (LuSEE-Night) is a low-frequency radio astronomy mission that will observe the sky from the lunar farside. LuSEE-Night will observe at frequencies between 1–50 MHz, with a nominal frequency resolution of 25 kHz and the ability to spectrally zoom to a frequency resolution ×64\times64 to an effective resolution of 390.625 Hz on one chosen parent bin. Free from Earth’s ionospheric distortions and radio frequency interference, LuSEE-Night’s location on the far side of the Moon is ideal for observing low-frequency high-nn RRLs. The instrument consists of four antenna monopole elements arranged in a cross-dipole configuration. Recent work has shown that LuSEE-Night may be able to map the sub-50 MHz sky at a resolution of ∼5∘\sim5^\circ (H. Camacho et al. [4]). A preliminary description of the instrument can be found in S. D. Bale et al. [3], with a more complete instrument paper currently in preparation. In this paper, we model Cα\alpha recombination lines across LuSEE-Night’s frequency band 1. Because low-frequency CRRLs trace widespread cold, diffuse gas where carbon is ionized, we adopt physical parameters representative of CNM-like conditions. We use these models to simulate LuSEE-Night’s observations of CRRLs and estimate the integration times required to detect individual lines at SNR=5. These estimates motivate strategies for observation timescales and stacking line profiles.

Modeling RRLs

In this section, we discuss how we model RRLs across the LuSEE-Night frequency range of 1-50 MHz. In § 2.1, we discuss at which frequencies spectral lines are centered. In § 2.2, we discuss various line broadening mechanisms: Doppler broadening (2.2.1), collisional broadening (2.2.2), and radiation/pressure broadening (2.2.3). In § 2.3, we discuss how these mechanisms are combined into line profile that model RRLs that LuSEE-Night may observe. In § 2.4, we discuss how external radiation fields may alter observations of RRLs.

Center Frequencies

The frequency ν\nu of an RRL is determined by photon emitted due to an electronic energy change ΔE\Delta E from principal quantum states n+Δnn+\Delta n to nn is

νn+Δn→n=cRZ2(1n2−1(n+Δn)2),(1)\nu_{n+\Delta n\to n}=cRZ^2\left(\frac{1}{n^2}-\frac{1}{(n+\Delta n)^2}\right), \tag*{(1)}

where cc is the speed of light and Z=1Z=1 for an excited electron transitioning between energy levels in hydrogenic atoms. The Rydberg constant, RR is given by

R=(2π2mee4ch3)(MM+me),(2)R=\left(\frac{2\pi^2m_{\mathrm{e}}e^4}{ch^3}\right)\left(\frac{M}{M+m_{\mathrm{e}}}\right), \tag*{(2)}

where mem_{\mathrm{e}} is the mass of an electron, MM is the mass of the nucleus, e=1.602×10−19e=1.602\times10^{-19} C is the charge of the electron, and h=6.62607015×10−34h=6.62607015\times10^{-34} m2^2kg/s is Planck’s constant.

LuSEE-Night’s nominal operational frequency range is 1-50 MHz, thus we simulate Cα\alpha RRLs spanning from C509α\alpha at 49.745 MHz to C869α\alpha at 10.009 MHz. We note that the spacing between successive lines decreases with increasing nn (Δν∼300\Delta\nu\sim300 kHz at 50 MHz and ∼35\sim35 kHz at 10 MHz), with an average separation of 110.38 kHz.

Line Broadening

RRLs have an intrinsic natural linewidth due to the finite lifetimes of excited atomic states, with the linewidth inversely proportional to the level’s lifetime [13]. However, because these lifetimes are extremely long for high-nn transitions, the natural line widths are very narrow, of order ∼0.1\sim0.1 Hz, and can be considered negligible for the purposes of modeling [28, 13].

The observed line width, however is expected to be dominated by Doppler, radiation, and pressure broadening. High-nn lines, specifically those in the LuSEE-Night’s observation band, are significantly broadened by the Stark effect, which introduces Lorentzian wings to the line profile. To model these effects, we follow the framework of F. Salgado et al. [26] and adopt typical CNM parameters, which are representative of the gas traced by low-frequency CRRLs.

Doppler Broadening

Doppler broadening of RRLs is caused by line-of-sight motions of the emitting gas, producing a Gaussian line profile based on its thermal and turbulent components. The Doppler full width at half maximum (FWHM) is given by

ΔνD=ν0c[4ln⁡2(2kTeM+Vturb2)]1/2,(3)\Delta\nu_D=\frac{\nu_0}{c}\left[4\ln2\left(\frac{2kT_{\mathrm{e}}}{M}+V_{\mathrm{turb}}^2\right)\right]^{1/2}, \tag*{(3)}

where ν0\nu_0 is the line center frequency, TeT_{\mathrm{e}} is the temperature of the gas (∼100\sim100 K for the CNM) [17, 8], and MM is the mass of the ion [26]. The second term takes into account microturbulent effects, which can be approximated as ∣Vturb∣≈20\lvert V_{\mathrm{turb}}\rvert\approx20 km s−1^{-1} for observations with beams on degree scales [15, 13]. Together, thermal and turbulent Doppler broadening sets lower bounds of ΔνD∼1.1\Delta\nu_D\sim1.1 kHz at 10 MHz, increasing to ∼5.5\sim5.5 kHz at 50 MHz.

Because LuSEE-Night measures the whole sky, differences in radial velocity along different lines of sight, mainly from Galactic rotation, can smear the spectrum. Although this does not broaden the intrinsic line profile, the effective total Doppler width can be written as

ΔνG≈ν0c[4ln⁡2(2kTM+Vturb2)+VR2]1/2,(4)\Delta\nu_G\approx\frac{\nu_0}{c}\left[4\ln2\left(\frac{2kT}{M}+V_{\mathrm{turb}}^2\right)+V_R^2\right]^{1/2}, \tag*{(4)}

where VRV_R is the variation in radial velocities due to Galactic rotation [15]. Depending on the sky direction, this can broaden the lines by an additional ∼9\sim9 kHz for VR=60V_{\mathrm{R}} = 60 km s−1^{-1} toward the Galactic center. In this work, however, we do not include this additional broadening term, as its value will depend on how LuSEE-Night is calibrated. As a result, our model may underestimate the linewidth by a few kHz at the higher end of LuSEE-Night’s bandwidth, where Doppler broadening dominates.

Collisional Broadening

Collisional broadening, or pressure broadening, tends to dominate at high-nn energy levels. The Lorentzian FWHM due to collisional broadening is given by:

Δνcol=1πNe∑n′≠nCnn′,(5)\Delta\nu_{\mathrm{col}} = \frac{1}{\pi}N_{\mathrm{e}}\sum_{n'\ne n} C_{nn'}, \tag*{(5)}

where NeN_{\mathrm{e}} is the electron density and Cnn′C_{nn'} is the collision rate for electron-induced transitions from level nn to n′n' (F. Salgado et al. [26]). For levels with n>100n > 100, the collision rate is

∑n′≠nCnn′=10a(Te)nγc(Te),(6)\sum_{n'\ne n} C_{nn'} = 10^{a(T_{\mathrm{e}})}n^{\gamma_{\mathrm{c}}(T_{\mathrm{e}})}, \tag*{(6)}

where TeT_{\mathrm{e}} is the electron temperature. Values for a(Te)a(T_{\mathrm{e}}) and γc(Te)\gamma_{\mathrm{c}}(T_{\mathrm{e}}) depend on the electron temperature and are provided in Table 3 of F. Salgado et al. [26]. Collisional broadening is quite sensitive to the electron density NeN_{\mathrm{e}}, which in the CNM ranges between 0.01<Ne<0.10.01 < N_{\mathrm{e}} < 0.1 cm−3^{-3}. We assume an electron temperature of Te=100T_{\mathrm{e}} = 100 K.

Sky Modelτstack\tau_{\mathrm{stack}}
11.51 hrs
26.30 days
327.6 hrs
410.8 mins
52.88 hrs
61.68 hrs

Table 3. Integration Times Required to Detect Stacked RRLs with SNR = 5 for Different Sky Models. Times are given in minutes (mins), hours (hrs), or 24-hour Earth days (days) of integration.

Radiation Broadening

Radiation broadening is a consequence of the interaction between atoms or ions and the surrounding radiation field. We define the RRL spectral line width contribution due to radiation broadening:

Δνrad=6.096×10−17T0n5.8,(7)\Delta\nu_{\mathrm{rad}} = 6.096 \times10^{-17} T_{0} n^{5.8}, \tag*{(7)}

where T0(ν)T_{0}(\nu) is the sky background brightness temperature at the reference frequency ν0=100\nu_{0} = 100 MHz. Equation 7, which is introduced in F. Salgado et al. [26], is derived by considering the case of a synchrotron radiation field characterized by a power law with a temperature T0T_{0} at a reference frequency ν0=100\nu_{0} = 100 MHz and a spectral index αpl=−2.6\alpha_{\mathrm{pl}} = -2.6. For our RRL models, we use a sky background brightness temperature of T0=2×103T_{0} = 2 \times10^{3} K towards the Galactic Plane and T0=103T_{0} = 10^{3} K away from the Galactic Plane, which is based on results from the LWA1 Low Frequency Sky Survey (J. Dowell et al. [9]). To calculate RRL spectral line width contributions due to radiation broadening, as in F. Salgado et al. [26], we also assume that the sky temperature follows a power law with a spectral index αpl=−2.6\alpha_{\mathrm{pl}} = -2.6 at 1-50 MHz. This assumption likely breaks down below ∼10\sim10 MHz, where free-free absorption is expected to dominate the sky spectrum (Dulk, G. A. et al. [10]; Manning, R. & Dulk, G. A. [18]).

Voigt Profile

We combine effects from the three described broadening mechanisms to construct a profile of the RRLs spectra in the LuSEE-Night bandwidth of 1-50 MHz. Collisional and radiation broadening effects introduce Lorentzian wings to spectral lines. The Lorentzian FWHM of an RRL spectral profile is the sum of the radiation and pressure broadening widths:

FWHMLorentzian=FWHMcoll+FWHMrad.(8)\mathrm{FWHM}_{\mathrm{Lorentzian}} = \mathrm{FWHM}_{\mathrm{coll}} + \mathrm{FWHM}_{\mathrm{rad}}. \tag*{(8)}

Doppler broadening introduces a Gaussian spread in the RRL profile. Because all three broadening mechanisms are expected to be present and observed, the RRL profile becomes a convolution of both Gaussian and Lorentzian profiles, which is known as a Voigt profile. The Voigt line profile of ϕ(ν)\phi(\nu) can be expressed using the complex error function:

ϕ(ν)=1σ2π⋅Re⁡[w(ν−ν0+iγσ2)],(9)\phi(\nu) = \frac{1}{\sigma\sqrt{2\pi}}\cdot\operatorname{Re}\left[w\left(\frac{\nu-\nu_{0}+i\gamma}{\sigma\sqrt{2}}\right)\right], \tag*{(9)}

where σ\sigma is the standard deviation of the Gaussian component, γ\gamma is the half-width at half-maximum of the Lorentzian component, and w(z)w(z) is the Faddeeva function (S. Abrarov & B. Quine [1]). Figure 1 summarizes the relative contributions of each of the three types of broadening mechanisms to the full line profile for a range of parameters. The Lorentzian components of the Voigt profile (radiation and collisional broadening) dominate the lower end of the LuSEE-Night frequency range, while Gaussian Doppler broadening dominates at higher frequencies.

Figure 1. Full width at half maximum (FWHM) contributions from Doppler, collisional, and radiation broadening as a function of frequency for varying parameters [top]. Combined FWHM of the CRRL Voigt profile, including contributions from Doppler, collisional, and radiation broadening used for estimating LuSEE-Night’s sensitivity [bottom].

Radiative Transfer of RRLs

In the presence of a strong background radiation field (as is true for low-frequency RRLs in the diffuse ISM), in the Rayleigh-Jeans regime (hν≪kTeh\nu\ll kT_{\mathrm{e}}), and when Δn/n≪1\Delta n/n \ll1, the ratio of the line intensity IνRRL(ν)I_{\nu}^{\mathrm{RRL}}(\nu) to the continuum intensity Iνcont(ν)I_{\nu}^{\mathrm{cont}}(\nu) in the diffuse optically thin (τν≪1\tau_{\nu} \ll1) ISM regime can be approximated as

IνRRLIνcont=−κνl(LTE)ϕ(ν)bnβn,n+ΔnL,(10)\frac{I_{\nu}^{\mathrm{RRL}}}{I_{\nu}^{\mathrm{cont}}} = -\kappa_{\nu}^{l}(\mathrm{LTE})\phi(\nu)b_{n}\beta_{n,n+\Delta n}L, \tag*{(10)}

where bnb_{n} is the departure coefficient from Local Thermodynamic Equilibrium (LTE), βn,n+Δn\beta_{n,n+\Delta n} is the correction factor for stimulated emission or absorption, κνl (LTE)\kappa_{\nu}^{l\,(\mathrm{LTE})} is the absorption coefficient, and ϕ(ν)\phi(\nu) is the Voigt line profile defined in Equation (9). This is equivalent to Equation (5) of F. Salgado et al. [26].

As outlined in F. Salgado et al. [26], the absorption coefficient can be calculated based on a specific cloud’s physical parameters, including electron temperature TeT_{\mathrm{e}}, electron density nen_{\mathrm{e}}, and line-of-sight path length LL. However, the CNM is quite inhomogeneous and there are a wide range of reported values for these parameters in the literature. Consequently, estimating the upper and lower limits of line strength would yield a very wide range of possible outcomes.

Instead, we opt for a more generalized and empirical approach by using the observed CRRL optical depths that already include non-LTE effects. This is a practical way to estimate the range of possible CRRL foreground strengths across LuSEE-Night’s band.

ground strengths across LuSEE-Night’s band. Therefore, rewriting Equation (10) in terms of brightness temperature of the background continuum (TcontT_{\mathrm{cont}}), we get

TRRL(ν)=−τνTcont(ν)ϕ(ν),τν≡κνlbnβn,n+ΔnL,(11)T_{\mathrm{RRL}}(\nu)=-\tau_{\nu}T_{\mathrm{cont}}(\nu)\phi(\nu), \qquad \tau_{\nu}\equiv\kappa_{\nu}^{l}b_n\beta_{n,n+\Delta n}L, \tag*{(11)}

where τν\tau_{\nu} is the optical depth with non-LTE effects included, TRRLT_{\mathrm{RRL}} is the brightness temperature of the RRL, and ϕ(ν)\phi(\nu) is the Voigt profile defined in Equation (9).

To estimate TcontT_{\mathrm{cont}} in Equation (11), we use the Ultralong Wavelength Sky Model with Absorption Effect (ULSA) maps (Y. Cong et al. [6]). We scale the ULSA maps by a factor of 0.2. We adopt this factor as a modeling assumption motivated by the background radiation models used in A. K Vydula et al. [15] and J. B. R. Oonk et al. (2019). This factor of 0.2 accounts for the fact that not all of the observed sky continuum actually illuminates the RRL-forming gas from behind. The gas sits at some distance along the line of sight, so a fraction of the total synchrotron and free-free emission we observe never interacts with the gas and therefore cannot drive stimulated emission or absorption. Only the emission originating behind the gas contributes to producing RRLs, and 0.2 is a rough estimate of that background fraction relative to the total observed continuum.

We estimate values for τν\tau_{\nu} in Equation (11) using two methods. The first method uses representative order-of-magnitude values from previous observations to fix the peak optical depth of each line profile. We use this approach to build general lower and upper bounding cases, as well as a more realistic region-dependent model. However, the few observations available below 50 MHz suggest that peak optical depths decrease at lower frequencies as the line widths increase, resulting in a roughly constant integrated optical depth (P. Salas et al. [25]). Our second method therefore fixes the integrated optical depth instead. This approach allows us to examine the smaller line-to-line variations in peak optical depth expected as the line width changes with frequency. Figure 2 shows the resulting Cα\alpha line profiles across 10–50 MHz for both methods, computed with the broadening parameters for regions away from the Galactic plane, before accounting for the effects of beam dilution and the spectrometer response.

Figure 2. Simulated Cα RRL brightness temperature spectra, ΔTRRL\Delta T_{\mathrm{RRL}}, as a function of frequency for the two optical-depth methods described in § 2.4, computed using broadening parameters representative of sky regions away from the Galactic plane (T0=1000T_0 = 1000 K; § 2.2.3). Top: Method 1 (fixed peak optical depth), showing the full Voigt line profile of every simulated Cα line across the 10–50 MHz band for τν=10−3\tau_\nu= 10^{-3}, 5×10−45 \times10^{-4}, and 10−410^{-4}. Bottom: Method 2 (fixed integrated optical depth), showing the Cα RRL profiles for τν=20\tau_\nu= 20, 10, and 5 Hz.

Method 1: Fixed Peak Optical Depth

The first method of estimating the optical depth of RRL observations fixes the peak value of τνϕ(ν)\tau_{\nu}\phi(\nu). τν\tau_{\nu} of Equation (11) is the peak optical depth, and ϕ(ν)\phi(\nu) is the Voigt line profile (Equation (9)) scaled such that it is unity at the line center.

We use representative values of peak optical depths τν\tau_\nu of 10−410^{-4}–10−310^{-3}, which cover most of the low-frequency CRRL detections in the literature, including W. C. Erickson et al. [11], N. G. Kantharia et al. [16], Roshi, D. Anish et al. [23], and J. B. R. Oonk et al. [22]. Results from EDGES are especially useful for our model because, like LuSEE-Night, the instrument averages over large regions of the sky. We therefore use the EDGES-based values in our fiducial region-dependent model, where the optical depth τν\tau_\nu varies with sky direction and the RRL absorption is strongest toward the Galactic center and weaker away from the Galactic plane. We also include two uniform cases, with τν=10−4\tau_\nu= 10^{-4} and 10−310^{-3} applied in all sky directions, as general lower and upper bounds. The top panel of Figure 2 shows the profiles of every simulated Cα line for peak optical depths of 10−310^{-3}, 5×10−45 \times10^{-4}, and 10−410^{-4}.

Method 2: Fixed Integrated Optical Depth

The second method of estimating the optical depth of RRL observations fixes the value of the integral ∫τνϕ(ν)dν\int\tau_\nu\phi(\nu)d\nu. The Voigt profile ϕ(ν)\phi(\nu) is normalized such that ∫ϕ(ν)dν=1\int\phi(\nu)d\nu= 1. We use integrated optical-depth values based on the trend reported in Figure 8 of P. Salas et al. [25]. The LOFAR data used in P. Salas et al. [25]

Figure 8. Simulated C654α\alpha line before (left) and after (right) convolution with the LuSEE-Night spectrometer response. The line is centered at 23.467 MHz with a FWHM of 3.78 kHz, the narrowest of all simulated CRRLs in the 10–50 MHz range in the region away from the Galactic Plane. Applying the spectrometer response (right) preserves the main line feature but introduces two smaller peaks near 23.44 MHz and 23.49 MHz, arising from aliasing between frequency channels (Figure 7b).

et al. (2017b) show that the integrated optical depth under non-LTE conditions increases gradually toward lower frequencies, from values of ∼14\sim14 Hz near 50 MHz Cα frequency to ∼20\sim20 Hz below ∼30\sim30 MHz Cα frequency, where it then becomes roughly constant. Although these values are measured toward Cas A, they are still useful for our purposes because integrated optical depth is measured relative to the continuum, so the bright background source does not itself make τν\tau_\nu larger. While the values are still specific to the absorbing gas along the line of sight to Cas A, their agreement with the undiluted integrated optical depth value reported by EDGES of ∼9\sim9 Hz at an effective frequency of 66.1 MHz (A. K. Vydula et al. [15]) measured when the Galactic center was overhead suggests that they provide a reasonable scale for this comparison model. The bottom panel of Figure 2 shows the corresponding profiles for integrated optical depths of 20, 10, and 5 Hz.

Simulation Pipeline

In this section, we discuss how we simulate LuSEE-Night’s observations of RRLs. In § 3.1, we discuss how we simulate the LuSEE-Night beam pattern with Ansys HFSS. In § 3.2, we discuss how we model the spatial distribution of optical depth τν\tau_\nu in the galaxy. In § 3.3, we calculate the effect of beam-dilution on RRL observations. In § 3.4, we convolve the beam-diluted RRLs with LuSEE’s spectrometer response. In § 3.5, we discuss how we estimate the integration time required for LuSEE-Night to detect individual RRLs.

Beam Model

We use Ansys HFSS’s Finite Element Method (FEM) solver (ANSYS, Inc. 2023) to simulate the far-field gain, impedance, and electromagnetic fields of the LuSEE-Night antenna-lander system at frequencies between 1-50 MHz. A model of the LuSEE-Night antenna and lander system in HFSS is shown in Figure 3. The model includes 4 antenna monopole elements labeled North, South, East, and West, which are arranged in a cross-dipole configuration. Additionally, this model assumes perfectly matched layer (PML) boundary conditions for five out of the six outer boundary faces (the four sides and the top). PML boundary conditions simulate an infinitely extending 3D space by absorbing outgoing waves at the edges of the finite simulation volume and preventing artificial reflections.

Model of LuSEE-Night instrument and lander in Ansys HFSS

Figure 3. Model of LuSEE-Night instrument and lander in Ansys HFSS used to simulate antenna far-field beam pattern. The instrument includes four 3m BeCu stacer antenna monopole elements arranged in a cross-dipole configuration.

The sixth boundary face of the Ansys HFSS model, at the bottom of the model volume, represents the lunar surface and is assigned a layered-impedance boundary condition. The lunar subsurface is modeled as consisting of eight layers. The depths and dielectric constants of each layer are taken from Figure 7 of R. E. Grimm [14]. The dielectric loss tangent of each layer is 0.01 and the relative permeability of each layer is 1.0. Modeling the lunar subsurface as a multi-layered stratified medium is consistent with previous work, including W. Fa & Y.-Q. Jin (2007), P. Chen et al. [5], and R. E. Grimm [14]. However, this eight-layer model of the lunar subsurface was derived using data from the Apollo 17 Surface Electrical Properties (SEP) experiment G. Simmons et al. [27], which was deployed on the lunar nearside. The properties of the lunar subsurface at the LuSEE-Night landing site on the lunar farside are currently unknown.

LuSEE-Night has a nominal frequency resolution of 25 kHz, as well as the ability to spectrally zoom in to a frequency resolution of 390.625 Hz. We simulate the far-field gain, impedance, and electromagnetic fields of LuSEE-Night from 1-50 MHz in 0.5 MHz steps. Then, to estimate the far-field beam at these spectrally zoomed-in frequencies, we linearly interpolate this simulated far-field beam pattern to frequencies where the Cα\alpha RRLs are centered.

Sky Models

We model the galactic foregrounds for LuSEE-Night by injecting the modified versions of ULSA sky maps (Y. Cong et al. [6]) with our models of RRLs, described in § 2. We consider 6 different sky models to test with our simulation pipeline. Each sky model implements different models of spectral and spatial distribution of the optical depth τν\tau_\nu. Table 1 summarizes how

Sky ModelMethod of Calculating τνϕ(ν)\tau_\nu\phi(\nu)Region Dependency of τν\tau_\nu
1Fixed Peak ValueUniform Sky, Upper Bound
2Fixed Peak ValueUniform Sky, Lower Bound
3Fixed Peak ValueRegion Dependent
4Fixed Integrated ValueUniform Sky, Upper Bound
5Fixed Integrated ValueUniform Sky, Lower Bound
6Fixed Integrated ValueRegion Dependent

Table 1. Sky Models for RRL Sensitivity Estimations

the spectral and spatial distribution of optical depth τν\tau_\nu was calculated for each of the six sky models.

We also consider two cases of how the optical depth τν\tau_\nu varies spatially: (i) a uniform case and (ii) a region-dependent case. For the uniform case, we assume that the optical depth τν\tau_\nu used to simulate RRLs is the same everywhere. We run the simulation pipeline with the uniform sky map using a lower- and upper-bound value for τν\tau_\nu. When fixing the peak optical depth, we use a maximum value of τν=10−3\tau_\nu= 10^{-3} (Sky Model 1) and a minimum value of τν=10−4\tau_\nu= 10^{-4} (Sky Model 2). When fixing the integrated optical depth, we use a maximum value of 20 Hz (Sky Model 4) and a minimum value of 5 Hz (Sky Model 5).

For the region-dependent case, we divide the sky map into three regions: the galactic center, the galactic plane, and away from galactic plane. We inject these three regions of the ULSA sky map with RRLs simulated with different optical depth τν\tau_\nu values. Figure 4 plots the region-dependent sky map (Sky Models 3 and 6 of Table 1) and how the optical depth values used for modeling RRLs with these sky maps vary spatially. When fixing the peak optical depth, we assign the galactic center a peak optical depth of τν=10−3\tau_\nu= 10^{-3}, the galactic plane a peak optical depth of τν=5×10−4\tau_\nu= 5 \times10^{-4}, and the region away from the galactic plane a peak optical depth of τν=10−4\tau_\nu= 10^{-4} (Sky Model 3). When fixing the integrated optical depth, we assign the galactic center an integrated optical depth of 20 Hz, the galactic plane an integrated optical depth of 10 Hz, and the region away from the galactic plane an integrated optical depth of 5 Hz (Sky Model 6).

Spatial distribution of the optical depth used in the region-dependent RRL simulations

Figure 4. Spatial distribution of the optical depth used in the region-dependent RRL simulations (Sky Models 3 and 6 of Table 1). The ULSA sky map at 25 MHz is also shown for reference. The Galactic center (red dashed box) is assigned the largest optical depth, the Galactic plane (within the white lines) is assigned an intermediate value, and regions away from the galactic plane are assigned the smallest value. This approximates the expected variation in low-frequency RRL absorption across the sky.

In Figure 5, we show the simulated Cα RRL brightness temperatures for Sky Model 3 for each of the three different sky regions. We sample the RRL brightness temperature for both of LuSEE’s frequency resolution options: a nominal 25 kHz resolution and a spectrally zoomed resolution of ∼390\sim390 Hz. LuSEE-Night’s nominal 25 kHz resolution can not fully resolve simulated RRL brightness temperature profiles between 10–50 MHz. Therefore, we only consider spectral zoom frequency resolutions of 390.625 Hz.

Simulated Cα RRL brightness temperature for Sky Model 3 at two frequency resolutions

Figure 5. Simulated Cα RRL brightness temperature for Sky Model 3 for each of the three sky regions (Galactic center, Galactic plane, and away from the Galactic plane), before beam dilution. Spectra are shown at the nominal 25 kHz resolution (top) and at the spectrally zoomed resolution of ∼390\sim390 Hz (bottom). In the top panel, each channel is modeled as an idealized boxcar response spanning ±12.5\pm12.5 kHz about the channel center, so the plotted value is the line profile averaged over the channel. The nominal 25 kHz resolution cannot fully resolve the RRL line profiles, so the line depths are diluted relative to the zoomed-mode spectra (bottom). In the bottom panel the channel width is much smaller than the line widths, so the profiles are sampled directly.

Beam Dilution

We simulate the effects of beam dilution on the spectral brightness temperature that LuSEE-Night will observe on the lunar far side using luseepy24, a set of python utilities developed for performing various LuSEE-Night related calculations. luseepy utilizes the lunarsky25 package to accurately determine the orientation of the Moon, and thus the LuSEE-Night instrument, with respect to the celestial sphere at any given time. We compute the temperature spectrum as a convolution of the antenna beam pattern and the Galactic foregrounds (see e.g. [19])

Tobs(ν,t)=∫skyTsky(ν,t,θ,ϕ)B(θ,ϕ,ν) dΩ∫B(θ,ϕ,ν) dΩ(12)T_{\mathrm{obs}}(\nu,t)=\frac{\int_{\mathrm{sky}}T_{\mathrm{sky}}(\nu,t,\theta,\phi)B(\theta,\phi,\nu)\,\mathrm{d}\Omega}{\int B(\theta,\phi,\nu)\,\mathrm{d}\Omega} \tag*{(12)}

where B(θ,ϕ,ν)B(\theta,\phi,\nu) refers to the simulated far-field LuSEE-Night beam described in § 3.1 and TskyT_{\mathrm{sky}} refers to the sky models described in § 3.2. We run the simulation at Cα RRL center frequencies to calculate the effects of beam dilution on the observed RRL amplitude.

Equation 12 is implemented in LUSEEPY in harmonic space. We compute TobsT_{\mathrm{obs}} for lmax⁡=32l_{\max}=32 over one lunar day and night (∼4\sim4 weeks), assuming that the temperature of lunar ground is 0 K. lmax⁡=32l_{\max}=32 corresponds to an angular scale of ∼5−6∘\sim5-6^\circ, comparable to the ∼5∘\sim5^\circ resolution at which LuSEE-Night is expected to map the sky ([4]). Computing TobsT_{\mathrm{obs}} over one full lunar day and night (∼4\sim4 weeks) ensures that the full range of beam orientations relative to the sky is sampled. Assuming the lunar ground is 0 K is equivalent to assuming that the effect on observations of a realistic moon brightness temperature has been perfectly calibrated out.

Running the LUSEEPY simulation at a frequency resolution of ∼390\sim390 Hz is computationally quite expensive. Because the amplitude and the width of RRLs can vary between the three different sky regions: the galactic center, the galactic plane, and away from the galactic plane, a brute force approach would require running the simulation with a ∼390\sim390 Hz frequency resolution. To reduce the computing time required to simulate the beam-diluted RRLs, we take advantage of the fact that the beam pattern and the background continuum vary slowly across a single RRL. The simulated beam is already interpolated from 0.5 MHz steps, while the Cα\alpha line widths are a few to ∼16\sim16 kHz (Figure 1), so beam dilution approximately

acts as a single multiplicative factor on each line rather than changing its shape.

We therefore do not run LuSEEpy on the full ∼105\sim10^{5}-channel grid at 390.625 Hz resolution. Instead, we run the LuSEEpy simulation only at the 361 Cα\alpha line-center frequencies (C509α\alpha–C869α\alpha), separately for each of the three sky regions (Galactic center, Galactic plane, and away from the Galactic plane). In each run, the sky map contains only RRLs with the optical depth assigned to that region (§ 3.2). This yields the beam-diluted line-center amplitude, ΔTRRL(νi,t)\Delta T_{\mathrm{RRL}}(\nu_i,t), for each line, region, and time step (Figure 6a,Figure 6b).

Four-panel demonstration of beam dilution calculation

Figure 6. A step-by-step demonstration of how the expected beam dilution of RRLs is calculated, using Sky Model 6 as an example. First, we identify the brightness temperature of each of the 361 simulated RRLs at the frequencies where RRLs are centered, which are marked in (a). We then run the LuSEEpy simulation only for the 361 frequencies where RRLs are centered for each of the three different sky regions. The LuSEEpy output at one time step is plotted in (b). We then re-sample the RRL brightness temperature at the spectrometer resolution of ∼390\sim390 Hz with the fixed widths calculated in § 2.2. The re-sampled RRLs are plotted in (c). Finally, we add the RRL brightness temperature spectra for each of the three different sky regions together to get a simulated LuSEE-observation of beam-diluted RRLs, which is shown in (d).

We then resample each line at the spectrometer resolution of 390.625 Hz. To do this, we take the unit-peak Voigt profile with the width computed in § 2.2 for that line frequency and region (Figure 1, bottom). We scale the Voigt profile to the beam-diluted line-center amplitude and sample it on the zoomed-mode frequency grid (Figure 6c). The three regional spectra are summed to produce the final simulated LuSEE-Night observation of beam-diluted RRLs (Figure 6d). This approach reduces the number of beam-convolution calculations by roughly a factor of ∼300\sim300 (∼105\sim10^{5} channels →\rightarrow 361 frequencies), at the cost of assuming that the beam and the sky continuum are constant across each individual RRL profile.

Spectrometer Response Model

LuSEE-Night has a spectral engine that runs an ADC with a sampling rate of 102.4 MS/s, corresponding to a Nyquist frequency of 51.2 MHz. It employs a 4-tap 4096 point polyphase filter bank (PFB) spectrometer, resulting in 2048 spectral channels nominally spaced by 25 kHz. The PFB samples can pass through another 64-point FFT engine, resulting in zoomed-in bins with 64×64\times improved spectral resolution of 390.625 Hz.

We plot the spectrometer response of a parent PFB bin and sub-bins in Figure 7a. This spectrometer response characterizes the response of the 64 zoom sub-bins of a parent PFB bin (N=0,…,63N=0,\ldots,63) to a monochromatic source at a frequency offset δ\delta from the parent-bin center. Each sub-bin response function is tabulated on a uniform 10 Hz grid over ∣δ∣≤50\lvert\delta\rvert\le50 kHz.

Figure 7. Spectrometer response as a function of frequency offset from the channel center. (a) The parent-bin response (dashed black line) is the weighted sum of the response across the window within roughly ±20\pm20 kHz of the channel center frequency. Each solid color line represents a sub-bin, of 390 Hz width. Note that the amplitude of the sub-bin weights are scaled by 1/64 so they can be compared with the parent bin response. (b) The spectrometer response weighting matrix for one parent bin, showing the normalized weight contributed by each source frequency νj\nu_j to each observed channel νi\nu_i. The dominant diagonal band reflects the response of the sub-bins shown in (a). The faint off-diagonal streaks near νi−νj≈±24\nu_i-\nu_j \approx\pm24 kHz indicate aliasing between channels.

The simulated RRL spectra are defined on the zoom-mode grid with spacing Δν=390.625\Delta\nu=390.625 Hz, so the spectrometer response is resampled onto a grid that is commensurate with it. We linearly interpolate each sub-bin response functions to a frequency resolution of 195.3125 Hz. Each resampled sub-bin response is then normalized so that every sub-bin response has unit total weight. This re-sampled spectrometer response is plotted in Figure 7b.

We convolve the beam-diluted brightness temperature from LuSEEpy with the spectrometer response function shown in Figure 7(b):

Tobs(νi,t)=∑jTRRL(νi,t)⋅K(νi,νj)(13)T_{\mathrm{obs}}(\nu_i,t)=\sum_j T_{\mathrm{RRL}}(\nu_i,t)\cdot K(\nu_i,\nu_j) \tag*{(13)}

where K(ν)K(\nu) is the spectrometer response, TRRL(ν,t)T_{\mathrm{RRL}}(\nu,t) is the beam-diluted brightness temperature of the RRLs, and Tobs(ν,t)T_{\mathrm{obs}}(\nu,t) is the simulated observed antenna brightness temperature.

This aliasing can in principle be corrected for by combining zoom response from neighboring bins or forward-modelling the data, but we ignore this possibility in this paper.

The full spectrometer response is given by

K(νi,νj)=W(νi−νjΔν,N(νj)),(14)K(\nu_i,\nu_j)=W\left(\frac{\nu_i-\nu_j}{\Delta\nu},N(\nu_j)\right), \tag*{(14)}

where νi\nu_i is the source frequency, νj\nu_j is the observed frequency, Δν\Delta\nu is the spectral zoom frequency resolution of 390.625 Hz, and NN is the sub-bin channel that corresponds to the observed frequency. W(νi,νj)W(\nu_i,\nu_j) is plotted in Figure 7(b). The spectrometer weighting function aliases measurements in frequency channels ±10\pm10 kHz from the center of each parent bin.

In Figure 8, we plot the beam-diluted C654α\alpha line before (left) and after (right) applying the spectrometer response. The C654α\alpha line is centered at 23.467 MHz and has the smallest width of all the simulated RRLs in the 10-50 MHz frequency range in the region away from the Galactic Plane, with a FWHM of 3.78 kHz. Applying the spectrometer response preserves the main features of the C654α\alpha line. However, aliasing in the spectrometer response manifests as smaller peaks in spectra, centered at approximately 23.44 MHz and 23.49 MHz.

Integration Time Estimation

We estimate the integration time required to observe the CRRLs with a given signal-to-noise ratio (SNR) using the radiometer equation ([30]):

SNR=TRRLTsysΔντ(15)\mathrm{SNR}=\frac{T_{\mathrm{RRL}}}{T_{\mathrm{sys}}}\sqrt{\Delta\nu\tau} \tag*{(15)}

where TRRLT_{\mathrm{RRL}} is the brightness temperature of the RRLs after beam-dilution effects and the spectrometer response are accounted for, TsysT_{\mathrm{sys}} is the spatially-averaged and beam-diluted brightness temperature of the ULSA sky maps, Δν\Delta\nu is the LuSEE-Night frequency bandwidth of 390.625 Hz, and τ\tau is the integration time.

We assume that the sky continuum TBGT_{\mathrm{BG}} is the dominant noise source, so that the system temperature is Tsys≈TBGT_{\mathrm{sys}} \approx T_{\mathrm{BG}}. In general Tsys≈TBG+TrecT_{\mathrm{sys}} \approx T_{\mathrm{BG}} + T_{\mathrm{rec}}, where TrecT_{\mathrm{rec}} is the receiver noise, and the required integration time increases by a factor of (1+TrecTBG)2\left(1 + \frac{T_{\mathrm{rec}}}{T_{\mathrm{BG}}}\right)^2. We do not include TrecT_{\mathrm{rec}} in this work. We expect this assumption to matter least at the low-frequency end of the band, where the sky is brightest, and most near 50 MHz, where the sky is faintest.

Results

We calculate the integration times τ\tau required to detect individual RRLs with SNR=5 for all 6 sky models described in Table 1. We note the minimum and maximum calculated integration times for each sky model in Table 2. Note that a “day” denotes a 24-hour Earth day (86,400 s) of cumulative integration time, not a lunar day (∼29.5\sim29.5 Earth days), and integration times refer to total effective observing time, which may be accumulated over multiple lunar nights.

Sky ModelMinimum τ\tauMaximum τ\tau
113.4 days101 days
21,340 days10,070 days
3253 days1610 days
430.5 hrs22.0 days
520.3 days353 days
611.6 days229 days

Table 2. Integration Times Required to Detect Individual RRLs with SNR = 5 for Different Sky Models. Times are given in hours (hrs) or in 24-hour Earth days (days) of integration.

Sky Models 2 and 5 represent the most pessimistic scenarios, with the lowest optical depths. Sky Models 1 and 4 represent the most optimistic scenarios, with the highest optical depths. Across all sky models, RRLs at high nn and low frequencies require much longer integration times because the background brightness temperature of the galaxy is much higher. RRLs at lower nn and higher frequencies are more likely to be observed with LuSEE-Night, as they require shorter integration times to make a detection with the same SNR.

Stacking Line Profiles

We extract a window around each frequency where Cα RRLs are expected to peak, centered on the channel that contains the line center. Each time-averaged window spans 25 kHz and contains 128 frequency channels. This results in a total of 360 windows, which are averaged together to yield the stacked line profile.

Figure 9 shows the resulting stacked Cα profile for each of the six sky models described in Table 1. The depth and width of each stacked profile reflect the combined effects of the assumed optical depth, beam dilution, and spectrometer response for that sky model. Sky Model 4, which uses the fixed integrated optical depth method with uniform upper-bound values, produces the deepest stacked profile, consistent with its correspondingly short integration times in Table 2. Sky Model 2, the uniform lower-bound case, produces the shallowest profile and requires the longest integration times. The spread among the six stacked profiles in Figure 9 illustrates how strongly the assumed optical-depth model affects the detectability of the stacked signal.

Figure 9. Stacked Cα RRL line profiles for each of the six sky models described in Table 1. Each profile is produced by extracting a 25 kHz window (128 frequency channels) around every simulated Cα line between 10 and 50 MHz and averaging the resulting 360 windows together. The line depth reflects the combination of optical depth, beam dilution, and spectrometer response assumed for each sky model. Sky Model 4 (red), which uses the fixed integrated optical depth method with uniform upper-bound optical depth values, produces the deepest stacked profile, consistent with its shortest required integration times in Table 2. Sky Model 2 (orange), the uniform lower-bound case, produces the shallowest profile, consistent with its correspondingly longer required integration times. The spread among the six curves illustrates how sensitive the predicted detectability of stacked CRRLs is to the assumed optical-depth model.

Stacking line profiles reduces the integration time required to make a detection as

τstack=(∑iNτi−1)−1(16)\tau_{\mathrm{stack}} = \left(\sum_{i}^{N} \tau_i^{-1}\right)^{-1} \tag*{(16)}

where NN is the number of stacked observation windows, τi\tau_i is the integration time required to observe a single Cα\alpha RRL with a given SNR without stacking, and τstack\tau_{\mathrm{stack}} is the integration time required to observe the stacked profile with the same given SNR.

For each sky model, we use Equation (16) to estimate the integration time required for a detection of the predicted Cα\alpha signal across all 361 line centers with SNR=5, assuming independent, sky-dominated radiometer noise (Table 3). The integration times required to make a detection are much smaller than those in Table 2, and the likelihood of a detection within a reasonable mission lifetime is much higher.

Conclusions and Future Work

Our work suggests that LuSEE-Night will make a stacked detection of CRRLs within the mission lifetime under reasonable astrophysical and instrument modeling, even under conservative astrophysical assumptions. We estimate that a stacked detection would require an integration time between hours to days.

We have modeled the expected brightness temperature of Cα\alpha RRLs across LuSEE-Night’s 1–50 MHz observing band, incorporating Doppler, collisional, and radiation broadening (§ 2), beam dilution from a simulated LuSEE-Night far-field antenna pattern (§ 3.1), and the instrument’s zoomed-mode spectrometer response (§ 3.4). Using six sky models that span a range of plausible optical-depth assumptions—fixed-peak and fixed-integrated methods, applied both uniformly and with region-dependent variation informed by prior low-frequency observations (Table 1)—we estimated the integration time required to detect individual Cα\alpha lines at SNR = 5 (Table 2).

Required integration times vary widely across the six sky models, from as short as 30 hours with the most optimistic optical depth assumptions (Sky Model 4) to as long as 27.6 years with the most conservative optical depth assumptions (Sky Model 2). Within any single sky model, integration times also vary substantially with frequency. Higher-frequency, lower-nn lines generally require far less integration time than lower-frequency, higher-nn lines. This is driven primarily by the steep rise in Galactic synchrotron background temperature toward lower frequencies. Stacking simulated Cα\alpha lines across the band, following the approach outlined in § 4.1 and shown in Figure 9, substantially deepens the detectable line profile and offers a promising strategy for reducing the required observing time relative to targeting individual lines.

These results suggest that the most readily detectable Cα\alpha RRLs with LuSEE-Night will be the higher-frequency, lower-nn transitions, while detection of the weakest, lowest-frequency lines individually is unlikely to be feasible within a realistic mission lifetime without stacking or additional sensitivity improvements. Stacking the simulated Cα\alpha lines reduces the integration time required for an SNR = 5 detection to between ∼11 minutes and ∼6.3 days across all six sky models (Table 3), indicating that LuSEE-Night is very likely to make a stacked detection of Cα\alpha RRLs within a reasonable mission lifetime, even under the most conservative optical-depth assumptions. Note that the real LuSEE-Night spectrometer can perform zoom over a single parent channel at a time and therefore real stacking would need to involve an optimal combination of zoomed and un-zoomed spectra. We leave this for future work.

Several extensions of this work would refine these estimates. First, this analysis focuses exclusively on Cα\alpha RRLs. Extending the model to Hα\alpha, Cβ\beta, and higher-order transitions, as well as their possible blending, would provide a more complete picture of the RRL foreground and additional candidates for stacking. Second, our lunar subsurface model is derived from Apollo 17 nearside measurements (R. E. Grimm [14]). As in-situ characterization of the LuSEE-Night landing site becomes available, updating the beam simulation accordingly will improve the fidelity of the beam-dilution estimates in § 3.1. Third, we have not assessed the impact of CRRLs on global Dark Ages 21-cm measurements. Because Cα\alpha lines occur at known frequencies and the optical depth and line width vary across the sky, the beam-weighted CRRL contribution may add a spectral structure that a smooth foreground model would not absorb. Quantifying the effects of RRLs on the detectability of the Dark Ages 21-cm monopole is left to future work.

Acknowledgments

This research was supported by the U.S. Department of Energy, Office of Science under the LuSEE-Night Science program. The LuSEE-Night program is funded under NASA Contract 80MSFC22CA018 and DOE Office of Science Contract DE-SC0012704. This work was also supported in part through HPC computational resources provided by NERSC (ERCAP0033114). Stuart D. Bale acknowledges the support of the Royal Society Wolfson Visiting Fellowship program. David Rapetti acknowledges support by NASA APRA grant award 80NSSC23K0013. L.V.E.K. acknowledges support from the European Research Council (ERC) under the European Union’s Horizon 2020 Research and Innovation Programme, grant agreement No. 884760 (CoDEX).

Facilities: NERSC

References

  1. [1]Abrarov, S., & Quine, B. 2015, Journal of Mathematics Research, 7, 163, doi: 10.5539/jmr.v7n2p163
  2. [3]Bale, S. D., Bassett, N., Burns, J. O., et al. 2023, https://arxiv.org/abs/2301.10345
  3. [4]Camacho, H., Rotermund, K. M., Slosar, A., et al. 2026, The Open Journal of Astrophysics, 9, doi: 10.33332/001c.158626DOI
  4. [5]Chen, P., Huang, P., He, Y., et al. 2013, Radio Science, 48, 265, doi: https://doi.org/10.1002/rds.20027
  5. [6]Cong, Y., Yue, B., Xu, Y., et al. 2021, ApJ, 914, 128, doi: 10.3847/1538-4357/abf55cDOI
  6. [7]Cros, L., Gusdorf, A., Salomé, P., et al. 2025, Astronomy & Astrophysics, 701, A148
  7. [8]Dickey, J. M., & Lockman, F. J. 1990, Annu. Rev. Astron. Astrophys., 28, 215
  8. [9]Dowell, J., Taylor, G. B., Schinzel, F. K., Kassim, N. E., & Stovall, K. 2017, Monthly Notices of the Royal Astronomical Society, 469, 4537, doi: 10.1093/mnras/stx1136
  9. [10]Dulk, G. A., Erickson, W. C., Manning, R., & Bougeret, J.-L. 2001, A&A, 365, 294, doi: 10.1051/0004-6361:20000006DOI
  10. [11]Erickson, W. C., McConnell, D., & Anantharamaiah, K. R. 1995, Astrophys. J., 454, 125, doi: 10.1086/176471DOI
  11. [13]Gordon, M. A., & Sorochenko, R. L. 2009, Astrophysics and Space Science Library, Vol. 282, Radio Recombination Lines: Their Physics and Astronomical Applications (New York: Springer), doi: 10.1007/978-0-387-09691-9
  12. [14]Grimm, R. E. 2018, Icarus, 314, 389, doi: https://doi.org/10.1016/j.icarus.2018.06.007
  13. [15]K Vydula, A., Bowman, J., Lewis, D., et al. 2023, The Astronomical Journal, 167, 2, doi: 10.3847/1538-3881/ad08ba
  14. [16]Kantharia, N. G., Anantharamaiah, K. R., & Payne, H. E. 1998, Astrophys. J., 506, 758, doi: 10.1086/306266DOI
  15. [17]Kulkarni, S. R., & Heiles, C. 1987, in Interstellar Processes, ed. D. J. Hollenbach & H. A. Thronson (Dordrecht: Springer Netherlands), 87–122
  16. [18]Manning, R., & Dulk, G. A. 2001, A&A, 372, 663, doi: 10.1051/0004-6361:20010516DOI
  17. [19]Monsalve, R. A., Bye, C. H., Sievers, J. L., et al. 2024, The Astrophysical Journal, 961, 56, doi: 10.3847/1538-4357/ad0f1b
  18. [22]Oonk, J. B. R., van Weeren, R. J., Salgado, F., et al. 2014, Monthly Notices of the Royal Astronomical Society, 437, 3506, doi: 10.1093/mnras/stt2158
  19. [23]Roshi, D. Anish, Kantharia, N. G., & Anantharamaiah, K. R. 2002, A&A, 391, 1097, doi: 10.1051/0004-6361:20020899DOI
  20. [24]Salas, P., Oonk, J. B. R., van Weeren, R. J., et al. 2017a, Monthly Notices of the Royal Astronomical Society, 467, 2274, doi: 10.1093/mnras/stx239
  21. [25]Salas, P., Oonk, J. B. R., van Weeren, R. J., et al. 2017b, Monthly Notices of the Royal Astronomical Society, 467, 2274, doi: 10.1093/mnras/stx239
  22. [26]Salgado, F., Morabito, L. K., Oonk, J. B. R., et al. 2017, Astrophysical Journal, 837, 142, doi: 10.3847/1538-4357/aa59da
  23. [27]Simmons, G., Strangway, D., Annan, P., & Baker, R. 1973, Apollo 17: Preliminary Science Report, 330
  24. [28]Sobel’man, I. I., Vainshtein, L. A., & Yukov, E. A. 1995, Springer Series on Atomic, Optical, and Plasma Physics, Vol. 15, Excitation of Atoms and Broadening of Spectral Lines (Springer), doi: 10.1007/978-3-642-57825-0DOI
  25. [29]Stepkin, S. V., Kovalenko, A. A., Kantorina, N. G., & Udaya Shankar, N. 2007, Monthly Notices of the Royal Astronomical Society, 374, 852, doi: 10.1111/j.1365-2966.2006.11190.xDOI
  26. [30]Thompson, A. R., Moran, J. M., & Swenson, George W., J. 2017, Interferometry and Synthesis in Radio Astronomy, 3rd edn., Astronomy and Astrophysics Library (Cham: Springer), doi: 10.1007/978-3-319-44431-4DOI

Paper details

Contents