betaFeedback

WorksPhysics and Astronomy

Observational diagnostics of external photoevaporation at millimetre wavelengths

Abstract

External photoevaporation is increasingly recognised as an important influence on protoplanetary disk evolution. However, there are no established molecular line diagnostics that can reliably distinguish external irradiation from variations in intrinsic stellar and disk properties. We use a large grid of thermochemical models to predict the fluxes of 48 transitions from eight species observable with ALMA, spanning a wide range of system properties and external FUV fields of 100 to 100,000 G0. At fixed stellar and disk properties, all transitions generally brighten with increasing FUV field strength. HCO+ and CO fluxes increase monotonically in virtually all disks, whereas lower-excitation lines, particularly those of C2H and N2H+, respond weakly and non-monotonically. However, variations in stellar mass and disk size introduce substantial degeneracies, limiting the usefulness of individual line fluxes as tracers of external irradiation. We therefore assess all pairwise line flux ratios, which can significantly reduce these degeneracies, evaluating their correlation with the external field, the magnitude of their change across the FUV range, and predicted ALMA detectability. We identify 135 promising ratios, including several combining strong FUV sensitivity and high detectability with efficient single-tuning ALMA observations. The strongest diagnostics either contrast high- and low-excitation lines of the same or co-spatial species (e.g. CO J=8-7/J=1-0), probing different regions of the photoevaporative wind, or pair an FUV-sensitive wind tracer with comparatively insensitive emission arising closer to the disk (e.g. HCO+ J=3-2/C2H N=3-2). Our results provide a framework for identifying externally photoevaporating systems from spatially unresolved emission and constraining their incident FUV field strengths.

Introduction

Nearly everything we know about the chemistry of planet-forming disks has been learned from systems evolving in isolation. Observational studies have focused on individual benchmark sources such as TW Hya and disk populations in low-mass star-forming regions such as Taurus and Lupus, which are nearby, relatively bright, and generally show little evidence of environmental influence (e.g. [100] (Ansdell et al. 2016; Bergin et al. 2016; Guilloteau et al. 2016; Isella et al. 2016; Miotello et al. 2017; Öberg et al. 2021)). Such quiescent environments, however, are the exception. Most stars instead form in massive clusters containing OB stars (Lada & Lada 2003), where disks are exposed to far-ultraviolet (FUV) radiation fields thousands of times stronger than the interstellar background (e.g. Fatuzzo & Adams 2008; Winter & Haworth 2022; Allen et al. 2025). If that irradiation leaves an imprint on disk chemistry, and by extension on the planets that disks produce, then the sample on which our understanding is built is poorly placed to reveal it.

The physical consequences of external irradiation are now well established. Disks in strongly irradiated environments are smaller (Eisner et al. 2018; Boyden & Eisner 2020; Portilla-Revelo et al. 2025), less massive ([125, 124] (Mann et al. 2014; Ansdell et al. 2017; Maucó et al. 2025)), and shorter lived than their isolated counterparts ([36] (Bally et al. 1998; Haworth et al. 2018; Concha-Ramírez et al. 2019; Nicholson et al. 2019; Winter et al. 2020)), with many displaying the classical ‘proplyd’ cometary morphology indicative of ongoing mass loss (O’Dell et al. 1993; O’Dell & Wen 1994; Bally et al. 2000).

How external irradiation shapes disk chemistry is far less clear, and the relatively small number of observations to date present a mixed picture. Several irradiated systems appear chemically similar to isolated disks, both in the inner disk probed by JWST (Ramírez-Tannus et al. 2023) and in the outer disk probed by ALMA (Díaz-Berríos et al. 2024). Others carry signatures that are difficult to explain without environmental effects. In the outer disk, these include a lack of CO depletion, in contrast to the significant depletion routinely inferred in isolated systems (Boyden & Eisner 2023), and bright, extended [CI] emission tracing warm atomic gas (Goicoechea et al. 2024). JWST has additionally revealed photochemically produced H3+_{3}^{+}, CH3+_{3}^{+} and CH+^{+} in the irradiated disk surface (Berné et al. 2023; Schroetter et al. 2025), and a CO2_{2}-enriched inner disk around a Herbig star (Frediani et al. 2025). These results suggest that external irradiation can alter disk chemistry, although the strength and spatial extent of its influence may vary between systems.

Chemical models generally predict that the radiation environment

© 2026 The Authors leaves a measurable imprint on the composition of the outer disk (Walsh et al. 2013; Keyte & Haworth 2025; Keyte & Ran 2025; Gross & Cleeves 2025; Calahan et al. 2025), although the effect scales with field strength and may only become clearly detectable in the most extreme environments. The inner disk is expected to be less sensitive, since its radiation field is dominated by the host star (Ndugu et al. 2024), and models of the terrestrial planet-forming region predict mid-infrared tracers that are largely insensitive to the external field (Frediani et al. 2026). Reconciling these observational and theoretical results therefore requires robust diagnostics capable of identifying which systems are affected, where the effects occur, and how strongly their chemistry has been altered.

A key difficulty is that molecular line fluxes depend not only on the external FUV field, but also on a wide range of intrinsic stellar and disk properties. Stellar mass, disk mass and size, dust-to-gas ratio, dust opacity, and turbulence all shape the disk density and temperature structure and regulate the attenuation of radiation, and can therefore influence line emission as strongly as, or more strongly than, external irradiation. The influence of individual disk and stellar properties on molecular abundances and line fluxes in quiescent systems has been explored in great detail (e.g. Aresu et al. 2011; Vasyunin et al. 2011; Cleeves et al. 2013; Walsh et al. 2015; Woitke et al. 2016; Kamp et al. 2017; Agúndez et al. 2018; Wakelam et al. 2019; Gavino et al. 2021), where it is well established that no single line maps cleanly to any individual physical parameter. External irradiation introduces an additional source of degeneracy. The problem is further complicated by the fact that irradiated systems typically lie at much further distances than the bright, well-studied isolated disks, so their emission is only marginally resolved, if at all. A bright line measured in an irradiated disk might therefore be attributed to a strong external FUV field, or could equally arise from the disk being larger than assumed and so presenting a greater emitting area. Comparing individual line fluxes between an irradiated and isolated system cannot, on its own, separate these possibilities.

Line flux ratios offer a way to deal with these degeneracies. If two transitions depend similarly on a given stellar/disk parameter, then that shared dependence partially cancels when one is divided by the other, leaving the response to the external field. Such diagnostics have long been used in studies of photodissociation regions (PDRs), where line ratios constrain the incident FUV field and gas density (e.g. Fuente et al. 1993; Kaufman et al. 1999; Boger & Sternberg 2005; Wolfire et al. 2022), and in isolated disks, where ratios such as CN/HCN have been proposed as probes of photochemistry (e.g. Chapillon et al. 2012; Guzmán et al. 2015; Visser et al. 2018; Bergner et al. 2021). For externally irradiated systems, comparable diagnostics have been developed only at optical wavelengths, which predominantly trace the ionised flow (Péake et al. 2025). No equivalent framework exists at millimetre wavelengths, despite this being the regime in which the bulk molecular content of disks is typically observed (e.g. Ansdell et al. 2016; Long et al. 2017, 2019; Öberg et al. 2021; Booth et al. 2024; Zhang et al. 2025).

Constructing such a framework requires models that capture the full structure of the irradiated system, not just the disk. Above a threshold external field strength, FUV photons drive a thermal wind from the disk surface, so that the system must be treated as a coupled disk–wind structure rather than a disk in isolation (e.g. Johnstone et al. 1998; Adams et al. 2004; Facchini et al. 2016). The wind has its own density, temperature, and opacity structure, with dust properties that differ from the disk, since only small grains are entrained (Facchini et al. 2016; Owen & Altaf 2021; Paine et al. 2025). Depending on system properties, emission from the wind can contribute a significant fraction of the total line flux, and the wind can also alter the composition of the disk itself through the way in which it reprocesses the external FUV field inward (Keyte & Haworth 2025). A model that does not include the wind may therefore reproduce an observed flux, but wrongly attribute it to the disk, and infer the wrong composition as a result. Any diagnostic intended for irradiated systems must therefore be calibrated against models which include both the disk and wind components.

Until recently, this was not practical. Radiation-hydrodynamic simulations resolve the disk-wind structure accurately, but cost weeks-to-months to run per model, which restricts them to a handful of representative cases (Haworth & Clarke 2019; Ballabio et al. 2023), ruling out the broad parameter survey required to identify robust diagnostics. To overcome this limitation, Keyte & Haworth (2026) developed the PUFFIN framework, which parametrically approximates disk-wind density structures at much lower computational cost. Coupling PUFFIN to established thermochemical codes makes grids of thousands of irradiated disk models feasible, and allows a systematic search of the parameter space for diagnostics.

We carry out such a search here, computing a large grid of models with DALI (Bruderer et al. 2012; Bruderer 2013) that span a range of stellar masses, disk sizes and masses, and external FUV field strengths. From these we synthesise integrated fluxes for a broad set of transitions across eight species accessible to ALMA. We then quantify the response of each transition to the external FUV field, separating this from the effects of the other model parameters, and rank all pairwise flux ratios by their FUV sensitivity, dynamic range, and detectability. The resulting diagnostics are designed to identify systems in which the photoevaporative wind contributes substantially to the observed flux, and to constrain the strength of the incident field.

This paper is organised as follows. § 2 describes the model grid and synthetic observations. § 3 presents the response of the individual line fluxes to the external FUV field and the impact of degeneracies, and finally assesses the robustness of the line flux ratios. § 4 examines the chemical origins of the observed trends and provides guidance on applying the diagnostics with ALMA. We summarise our conclusions in § 5.

Modelling

We construct a grid of thermochemical models spanning a broad range of stellar masses, disk properties, and external FUV field strengths (Table 1). Each model comprises a two-dimensional density structure generated with the PUFFIN parametric framework (Keyte & Haworth 2026), which serves as input to thermochemical calculations performed with the DALI code (Bruderer et al. 2012; Bruderer 2013). The physical setup, stellar parameters, dust treatment, and chemical network follow the approach of Keyte & Haworth (2026), which we summarise below. We refer the reader to that work for a complete description of the modelling procedure.

Physical structure

For each combination of input parameters, we generate a two-dimensional gas density structure using the PUFFIN Python package. PUFFIN is a parametric framework for computing the density structures of protoplanetary disks undergoing external photoevaporation. Given a set of physical inputs and a mass-loss rate, it produces density structures that approximate the results of full radiation-hydrodynamic simulations, but can be evaluated in minutes rather than the weeks-to-months typically required. This efficiency makes it well-suited to the systematic exploration of large parameter spaces.

ParameterDescriptionValues
M∗M_\astStellar mass [M⊙M_\odot]0.3, 0.6, 1.0, 2.0, 3.0
rdr_{\mathrm{d}}Disk outer radius [au]20, 40, 60, 80, 100, 150
Σ1au\Sigma_{\mathrm{1au}}Surface density at 1 au [g cm−2^{-2}]10110^1, 5×1015 \times10^1, 10210^2, 5×1025 \times10^2, 10310^3, 5×1035 \times10^3, 10410^4
FFUVF_{\mathrm{FUV}}External FUV field [G0G_0]10210^2, 5×1025 \times10^2, 10310^3, 5×1035 \times10^3, 10410^4, 5×1045 \times10^4, 10510^5

Table 1. Parameter grid used to construct the thermochemical model grid. All combinations of the listed values are computed, yielding 1470 models in total.

The density structure comprises three components: a disk in vertical hydrostatic equilibrium (accounting for external FUV heating of the disk surface), a spherically diverging photoevaporative wind, consistent with the outflow geometry adopted in one-dimensional models of external photoevaporation (Facchini et al. 2016; Haworth et al. 2018), and a smooth transition region that bridges the two, approximating the disk-wind structure found in two-dimensional radiation-hydrodynamic simulations (Haworth & Clarke 2019). The wind density profile depends on the mass-loss rate, which is interpolated from the FRIED grid of hydrodynamic simulations (Haworth et al. 2018, 2023) and multiplied by a factor of two to account for the systematic underprediction of mass-loss rates by one-dimensional models relative to two-dimensional simulations found by Haworth & Clarke (2019). The density structure therefore depends on FFUVF_{\mathrm{FUV}} through both the mass-loss rate and the heating of the disk surface, and is computed once with puffin and held fixed during the thermochemical calculation.

The dust distribution follows the approach of Haworth & Clarke (2019), in which the photoevaporative wind has a smaller maximum grain size and a lower dust-to-gas mass ratio than the disk because only sufficiently small grains can be entrained in the outflow (Facchini et al. 2016; Owen & Altaf 2021; Paine et al. 2025). We adopt a dust-to-gas mass ratio of Δd/g=0.01\Delta_{\mathrm{d/g}} = 0.01 in the disk, consistent with the canonical interstellar value (Bohlin et al. 1978), while the wind is dust-depleted with Δd/g=3×10−4\Delta_{\mathrm{d/g}} = 3 \times10^{-4} following Haworth & Clarke (2019). Two grain populations are included in the disk: small grains (0.005–1 μ\mum), which share the gas scale height, and large settled grains (0.005–1000 μ\mum), whose scale height is reduced by a factor of χ=0.2\chi= 0.2 following the standard dali formulation based on the prescription of D’Alessio et al. (2006). The large grain mass fraction is f=0.9f = 0.9. The wind contains only the small grain population, following a standard ISM size distribution (Mathis et al. 1977).

The computational grid is initially constructed with 1000 logarithmically spaced cells in both the radial and vertical directions, extending from r=0.1r = 0.1 au to rout=8rdr_{\mathrm{out}} = 8 r_{\mathrm{d}} and z=0.01z = 0.01 au to zout=8rdz_{\mathrm{out}} = 8 r_{\mathrm{d}}. This resolution is necessary to obtain accurate solutions from the hydrostatic equilibrium solver within PUFFIN and to capture the full spatial extent of the photoevaporative wind. The structure is subsequently regridded onto a coarser 200 ×\times 200 grid to ensure that the thermochemical calculations are computationally feasible, while retaining sufficient resolution to resolve chemical abundances on scales of a few au or smaller. The resulting density structures are passed to DALI, which computes the UV radiation field and dust temperature via Monte Carlo radiative transfer before iteratively solving for the gas temperature, molecular excitation, and time-dependent chemistry.

Stellar parameters

For each stellar mass, we generate a blackbody spectrum with effective temperature and luminosity determined from the empirical scaling relations of Eker et al. (2018). X-ray emission is modelled as a thermal spectrum with temperature TX=7×107T_{\mathrm{X}} = 7 \times10^7 K over the 1–100 keV energy range, adopting a fixed X-ray luminosity of LX=7.94×1028L_{\mathrm{X}} = 7.94 \times10^{28} erg s−1^{-1} across all stellar masses. Although X-ray luminosity is observed to scale with stellar mass (Flaischlen et al. 2021), holding LXL_{\mathrm{X}} constant isolates the effects of external FUV irradiation on disk chemistry and simplifies the interpretation of trends across the parameter grid. We also neglect accretion-driven UV emission, which can dominate the FUV output of low-mass stars (e.g. France et al. 2014). This additional FUV may warm the inner disk, particularly around the lowest-mass stars, whose photospheric FUV emission is weakest, although its impact is typically confined to the inner few au (e.g. Calahan et al. 2025). A treatment of accretion is beyond the scope of this work, and, as with LXL_{\mathrm{X}}, omitting accretion luminosity as a grid parameter allows us to more easily isolate the effects of the external FUV field.

Chemical network and initial conditions

The chemical evolution is calculated using a network based on a subset of UMIST 06 (Woodall et al. 2007), as described in Visser et al. (2018). This network comprises 134 species connected through 1855 reactions, encompassing gas-phase chemistry, freeze-out and thermal desorption, photodissociation and photoionisation, X-ray and cosmic-ray-induced processes, polycyclic aromatic hydrocarbon (PAH) charge exchange, and reactions involving vibrationally excited H2_2. Non-thermal desorption is included for CO, CO2_2, H2_2O, CH4_4, NH3_3, and N2_2, while grain surface chemistry is restricted to hydrogenation of simple species. A complete description of these processes is provided in Bruderer et al. (2012). We adopt initial molecular abundances based on Ballering et al. (2021), representing a typical protoplanetary disk with Solar-like C/O = 0.5 and composition inherited from the parent molecular cloud (see Table B1). The chemistry is evolved for 1 Myr.

SpeciesAbundance (X/H)
H2_25.00×10−15.00 \times10^{-1}
He9.00×10−29.00 \times10^{-2}
H2_2O1.00×10−41.00 \times10^{-4}
CO1.00×10−41.00 \times10^{-4}
N2_23.51×10−53.51 \times10^{-5}
C+^+1.00×10−51.00 \times10^{-5}
NH3_34.80×10−64.80 \times10^{-6}
HCN2.00×10−82.00 \times10^{-8}
H3+_3^+1.00×10−81.00 \times10^{-8}
S+^+9.00×10−99.00 \times10^{-9}
HCO+^+9.00×10−99.00 \times10^{-9}
C2_2H8.00×10−98.00 \times10^{-9}
Mg+^+1.00×10−111.00 \times10^{-11}
Si+^+1.00×10−111.00 \times10^{-11}
Fe+^+1.00×10−111.00 \times10^{-11}

Table B1. Initial abundances for all chemical models presented in § 2.3, adapted from Ballering et al. (2021). These values represent a typical protoplanetary disk with Solar-like C/O=0.5C/O=0.5 and composition inherited from the parent molecular cloud.

Parameter grid

Our model grid spans stellar masses from M∗=0.3M_\ast= 0.3 to 3.0M⊙3.0 M_\odot, disk radii from rd=20r_{\mathrm{d}} = 20 to 150 au, surface density normalisations from Σ1au=101\Sigma_{\mathrm{1au}} = 10^1 to 10410^4 g cm−2^{-2}, and external FUV field strengths from FFUV=102F_{\mathrm{FUV}} = 10^2 to 10510^5 G0G_0 (where G0=1.6×10−3G_0 = 1.6 \times10^{-3} erg cm−2^{-2} s−1^{-1} is the average interstellar radiation field in the solar neighbourhood; Habing 1968). The range of FFUVF_{\mathrm{FUV}} is set by the validity of PUFFIN, whose density structures have been calibrated against radiation-hydrodynamic simulations only up to 10510^5 G0G_0 (Keyte & Haworth 2026). Structures generated at stronger fields cannot be assumed to be reliable and we do not extrapolate our results beyond this limit. The ranges of the remaining parameters were chosen to span the stellar masses and compact disk sizes and masses representative of the externally irradiated disk populations accessible to current observations (e.g. Mann et al. 2014; Eisner et al. 2018; Boyden & Eisner 2020). The full set of parameters is listed in Table 1. The resulting grid contains 1470 models in total.

Synthetic observations

For each model in the grid, we extract disk-integrated line fluxes for a set of species and transitions observable with ALMA at (sub-)millimetre wavelengths. We focus on eight species that are typically abundant in protoplanetary disks and readily detectable in this regime: CO, HCO+^{+}, CN, HCN, C2_2H, N2_2H+^{+}, H2_2CO, and atomic C.

For each species, we select transitions that collectively span the frequency range of ALMA Bands 3-10 (84–950 GHz), limiting the selection to at most two transitions per ALMA band per species and sampling a range of upper energy levels. In total, this yields 48 transitions that probe the broad range of temperatures and chemical conditions expected under external FUV fields of varying strength. The full list of transitions is provided in Table A1. For brevity, throughout this paper CN and C2_2H transitions are labelled by their rotational quantum number NN alone, omitting the fine- and hyperfine-structure designations, and H2_2CO transitions by JJ alone, omitting the KaK_{\rm a} and KcK_{\rm c} sublevels. Full transition designations are given in Table A1.

MoleculeTransitionFreq (GHz)EuE_{\mathrm{u}} (K)AulA_{\mathrm{ul}} (s−1^{-1})gug_{\mathrm{u}}
[CI]3P1−3P0{}^{3}P_{1} - {}^{3}P_{0}492.16123.67.88×10−87.88 \times10^{-8}3
[CI]3P2−3P1{}^{3}P_{2} - {}^{3}P_{1}809.34262.52.65×10−72.65 \times10^{-7}5
C2_2HN=1−0N = 1 - 0, J=3/2−1/2J = 3/2 - 1/2, F=2−1F = 2 - 187.3174.21.53×10−61.53 \times10^{-6}5
C2_2HN=2−1N = 2 - 1, J=5/2−3/2J = 5/2 - 3/2, F=3−2F = 3 - 2174.66312.61.47×10−51.47 \times10^{-5}7
C2_2HN=3−2N = 3 - 2, J=5/2−3/2J = 5/2 - 3/2, F=3−2F = 3 - 2262.06525.24.89×10−54.89 \times10^{-5}7
C2_2HN=4−3N = 4 - 3, J=9/2−7/2J = 9/2 - 7/2, F=5−4F = 5 - 4349.33841.91.31×10−41.31 \times10^{-4}11
C2_2HN=5−4N = 5 - 4, J=11/2−9/2J = 11/2 - 9/2, F=6−5F = 6 - 5436.66162.92.61×10−42.61 \times10^{-4}13
C2_2HN=7−6N = 7 - 6, J=15/2−13/2J = 15/2 - 13/2, F=8−7F = 8 - 7611.267117.47.36×10−47.36 \times10^{-4}17
COJ=1−0J = 1 - 0115.2715.57.15×10−87.15 \times10^{-8}3
COJ=2−1J = 2 - 1230.53816.66.86×10−76.86 \times10^{-7}5
COJ=3−2J = 3 - 2345.79633.22.48×10−62.48 \times10^{-6}7
COJ=4−3J = 4 - 3461.04155.36.08×10−66.08 \times10^{-6}9
COJ=6−5J = 6 - 5691.473116.22.12×10−52.12 \times10^{-5}13
COJ=7−6J = 7 - 6806.652154.93.39×10−53.39 \times10^{-5}15
COJ=8−7J = 8 - 7921.800199.15.09×10−55.09 \times10^{-5}17
CNN=1−0N = 1 - 0, J=3/2−1/2J = 3/2 - 1/2113.4955.51.19×10−51.19 \times10^{-5}4
CNN=2−1N = 2 - 1, J=5/2−3/2J = 5/2 - 3/2226.87616.31.14×10−41.14 \times10^{-4}6
CNN=3−2N = 3 - 2, J=5/2−3/2J = 5/2 - 3/2340.03132.63.85×10−43.85 \times10^{-4}6
CNN=4−3N = 4 - 3, J=9/2−7/2J = 9/2 - 7/2453.60754.41.02×10−31.02 \times10^{-3}10
CNN=6−5N = 6 - 5, J=11/2−9/2J = 11/2 - 9/2680.047114.23.50×10−33.50 \times10^{-3}12
CNN=7−6N = 7 - 6, J=13/2−11/2J = 13/2 - 11/2793.336152.35.64×10−35.64 \times10^{-3}14
o-H2_2COJKa,Kc=21,1−11,0J_{K_{\mathrm{a}},K_{\mathrm{c}}} = 2_{1,1} - 1_{1,0}150.49822.66.47×10−56.47 \times10^{-5}5
o-H2_2COJKa,Kc=31,2−21,1J_{K_{\mathrm{a}},K_{\mathrm{c}}} = 3_{1,2} - 2_{1,1}225.69833.42.77×10−42.77 \times10^{-4}7
o-H2_2COJKa,Kc=53,2−43,1J_{K_{\mathrm{a}},K_{\mathrm{c}}} = 5_{3,2} - 4_{3,1}364.289158.48.90×10−48.90 \times10^{-4}11
o-H2_2COJKa,Kc=71,7−61,6J_{K_{\mathrm{a}},K_{\mathrm{c}}} = 7_{1,7} - 6_{1,6}491.968106.33.44×10−33.44 \times10^{-3}15
o-H2_2COJKa,Kc=91,8−81,7J_{K_{\mathrm{a}},K_{\mathrm{c}}} = 9_{1,8} - 8_{1,7}674.810174.09.09×10−39.09 \times10^{-3}19
o-H2_2COJKa,Kc=111,10−101,9J_{K_{\mathrm{a}},K_{\mathrm{c}}} = 11_{1,10} - 10_{1,9}823.083249.41.67×10−21.67 \times10^{-2}23
HCNJ=1−0J = 1 - 088.6324.32.41×10−52.41 \times10^{-5}3
HCNJ=2−1J = 2 - 1177.26112.82.31×10−42.31 \times10^{-4}5
HCNJ=3−2J = 3 - 2265.88625.58.36×10−48.36 \times10^{-4}7
HCNJ=4−3J = 4 - 3354.50542.52.05×10−32.05 \times10^{-3}9
HCNJ=5−4J = 5 - 4443.11663.84.10×10−34.10 \times10^{-3}11
HCNJ=7−6J = 7 - 6620.304119.11.16×10−21.16 \times10^{-2}15
HCNJ=8−7J = 8 - 7708.877153.11.74×10−21.74 \times10^{-2}17
HCNJ=10−9J = 10 - 9885.971233.93.43×10−23.43 \times10^{-2}21
HCO+^{+}J=1−0J = 1 - 089.1894.34.25×10−54.25 \times10^{-5}3
HCO+^{+}J=2−1J = 2 - 1178.37512.84.08×10−44.08 \times10^{-4}5
HCO+^{+}J=3−2J = 3 - 2267.55825.71.48×10−31.48 \times10^{-3}7
HCO+^{+}J=4−3J = 4 - 3356.73442.83.63×10−33.63 \times10^{-3}9
HCO+^{+}J=5−4J = 5 - 4445.90364.27.24×10−37.24 \times10^{-3}11
HCO+^{+}J=7−6J = 7 - 6624.209119.82.04×10−22.04 \times10^{-2}15
HCO+^{+}J=8−7J = 8 - 7713.342154.13.07×10−23.07 \times10^{-2}17
HCO+^{+}J=10−9J = 10 - 9891.558235.46.07×10−26.07 \times10^{-2}21
N2_2H+^{+}J=1−0J = 1 - 0, F1=1−1F_1 = 1 - 1, F=0−1F = 0 - 193.1724.53.90×10−53.90 \times10^{-5}1
N2_2H+^{+}J=2−1J = 2 - 1, F1=3−2F_1 = 3 - 2, F=4−3F = 4 - 3186.34513.43.74×10−43.74 \times10^{-4}9
N2_2H+^{+}J=3−2J = 3 - 2, F1=4−3F_1 = 4 - 3, F=5−4F = 5 - 4279.51226.81.35×10−31.35 \times10^{-3}11
N2_2H+^{+}J=4−3J = 4 - 3, F1=5−4F_1 = 5 - 4, F=6−5F = 6 - 5372.67344.73.32×10−33.32 \times10^{-3}13
N2_2H+^{+}J=5−4J = 5 - 4, F1=6−5F_1 = 6 - 5, F=7−6F = 7 - 6465.82567.16.64×10−36.64 \times10^{-3}15

Table A1. Molecular and atomic transitions used in this work.

Disk-integrated fluxes are extracted using the ray-tracing module in dali, assuming a face-on disk (i=0∘i = 0^\circ) at a distance of 400 pc, representative of the Orion star-forming region, which hosts many of the best-studied externally irradiated systems (e.g. O’Dell & Wen 1994; Bally et al. 2000; Mann et al. 2014). We focus on spatially unresolved, disk-integrated emission because our aim is to identify diagnostics that can be applied independently of angular resolution. This is particularly important because externally irradiated disks are also studied in massive star-forming regions at distances of ∼1−3\sim1 - 3 kpc, such as NGC 2244, Cygnus OB2, and the Carina Nebula, where spatially resolving individual disks is substantially more challenging than in the ONC (e.g. Balog et al. 2007; Mesa-Delgado et al. 2016; Champion et al. 2017; Guarcello et al. 2023). Disk-integrated measurements also offer the practical advantage of not requiring a specific high-resolution array configuration, allowing observations to be conducted across a wider range of ALMA configurations and observing periods.

A small number of line fluxes (∼1.6%\sim1.6\%) return unphysical values in excess of 101010^{10} Jy km s−1^{-1}. These arise in models combining a high stellar mass with a dense, high-surface-density disk. In these cases, the deep gravitational potential limits the mass removed by external photoevaporation, so the wind is very tenuous, while the high disk surface density sets up a steep density contrast between the disk surface and the wind. Under these conditions the level populations of some transitions invert within the wind, giving a negative absorption coefficient. Since dali solves the radiative transfer assuming this coefficient is positive, and includes no saturation mechanism, the intensities can be amplified without limit where the populations invert. The affected fluxes therefore reflect a failure of the solver in a regime it does not accurately model. To mitigate this, we exclude from our analysis any individual line flux exceeding 10510^{5} Jy km s−1^{-1}, together with any non-finite values. This threshold lies well above the brightest physical fluxes in the grid (∼104\sim10^{4} Jy km s−1^{-1}) and far below the affected values, so our results are insensitive to its precise choice.

Results

Response of individual line fluxes to model parameters

We begin by examining the disk-integrated line fluxes computed across the full model grid for all transitions listed in Table A1. Figure 1 shows the flux distributions as a function of their rest frequencies across the ALMA wavelength regime, coloured by species. For each transition, the thick bar spans the interquartile range of the flux distribution across all models, while the thin line extends to the minimum and maximum values.

Disk-integrated line fluxes for all models in the grid

Figure 1. Disk-integrated line fluxes for all models in our grid, assuming a face-on geometry and a source distance of 400 pc. Each marker represents a single transition, coloured by species and plotted at the corresponding rest frequency across ALMA Bands 3-10. For each transition, the thick bar spans the interquartile range across the model grid and the thin line covers the full range. A small horizontal jitter is applied for clarity. The shaded grey region indicates fluxes below an approximate 3σ3\sigma ALMA detection threshold, estimated for a 10-hour integration, assuming all emission is unresolved within a single synthesised beam.

The line fluxes span approximately 13 orders of magnitude across the full model grid, reflecting the wide range of stellar and disk properties explored. For any individual transition, the total range typically covers a few orders of magnitude. However, the interquartile ranges are much narrower, generally spanning only ∼1\sim1–2 orders of magnitude, indicating that the most extreme flux values arise from only a small subset of models.

To assess the observability of each transition, we estimate a representative ALMA 3σ3\sigma detection threshold, shown by the shaded grey region in Figure 1. We assume a total integration time of 10 hours, broadly representative of a standard ALMA spectral-line programme, and a line width of 10 km s−1^{-1}, with the emission distributed uniformly across the line profile. The source is assumed to be spatially unresolved, such that all of the flux is contained within a single synthesised beam. We also adopt the default atmospheric water-vapour level corresponding to the third octile. Using these assumptions, we calculated the detection threshold with the ALMA Sensitivity Calculator2 at intervals of 100 GHz and smoothly interpolated between these values across the full frequency range. The resulting threshold increases from approximately 7 mJy km s−1^{-1} at 100 GHz to 104 mJy km s−1^{-1} at 900 GHz. Although intended only as an approximate guide, this threshold provides a useful first-order measure of detectability and is used in later sections to identify the most promising molecular line-flux ratio diagnostics.

Sensitivity of line fluxes to the external FUV field

Our primary aim is to identify observational signatures of external photoevaporation, so we first assess how the disk-integrated line fluxes of individual species respond to the external FUV field strength FFUVF_{\mathrm{FUV}}. To quantify this, we use the Spearman rank correlation coefficient, which measures the strength of monotonic association between two variables on a scale from −1-1 (perfectly decreasing) through 0 (no trend) to +1+1 (perfectly increasing), and is computed by replacing each variable with its rank before evaluating the Pearson correlation of the resulting rank sequences. We compute this statistic in two complementary ways, which answer different questions. The first, the grid-wide correlation ρgrid\rho_{\mathrm{grid}}, is evaluated between log⁡(FFUV)\log(F_{\mathrm{FUV}}) and the logarithm of the disk-integrated flux across all models in the grid, in which M∗M_{\ast}, rdr_{\mathrm{d}} and Σ1au\Sigma_{\mathrm{1au}} vary alongside the external field. It asks whether a line flux tracks FFUVF_{\mathrm{FUV}} across a population of disks with diverse, and generally unknown, properties, as in an observational survey. The second, the fixed-structure correlation ρfixed\rho_{\mathrm{fixed}}, asks whether a line responds to the external field when nothing else changes. For each combination of (M∗,rd,Σ1au)(M_{\ast}, r_{\mathrm{d}}, \Sigma_{\mathrm{1au}}) the corresponding models differ only in FFUVF_{\mathrm{FUV}}, and we compute the Spearman coefficient separately within each such set. We summarise each transition by the median of the resulting distribution and its central quartile range. A median ρfixed\rho_{\mathrm{fixed}} of unity therefore means that, in a typical disk, the line flux increases at every step in FFUVF_{\mathrm{FUV}}. Where ρfixed\rho_{\mathrm{fixed}} is high but ρgrid\rho_{\mathrm{grid}} is low, the line responds consistently to the external field, but this response is masked by variations in the stellar and disk properties. The difference between the two therefore quantifies the degeneracy that motivates the use of line flux ratios in § 3.4. Where both are low, the line is intrinsically insensitive to the external field. We emphasise that both statistics measure the consistency of a monotonic response rather than the magnitude of the underlying flux change.

Figure 2 shows both statistics for every transition, grouped by species. Open circles denote ρgrid\rho_{\mathrm{grid}}, filled circles the median ρfixed\rho_{\mathrm{fixed}}, and coloured bars the corresponding central quartile range. For every transition, the median ρfixed\rho_{\mathrm{fixed}} exceeds ρgrid\rho_{\mathrm{grid}}. This is expected, since the stellar and disk parameters vary independently of FFUVF_{\mathrm{FUV}} in our grid, and therefore dilute rather than strengthen a consistent response to the field. The size of this gap is nonetheless informative, since it reflects how strongly variations in stellar and disk properties mask the intrinsic FUV response. HCO+^{+} and CO exhibit the strongest correlations, with median ρfixed=1\rho_{\mathrm{fixed}} = 1 and negligible central quartile spread across all transitions. Their grid-wide correlations are considerably lower, with ρgrid\rho_{\mathrm{grid}} increasing with excitation from ∼0.55\sim0.55 to ∼0.8\sim0.8 for HCO+^{+} and from ∼0.3\sim0.3 to ∼0.5\sim0.5 for CO, reflecting the dilution of the perfectly monotonic response to external irradiation by structure-to-structure flux variations. HCN approaches similarly strong behaviour at moderate and high excitation, with its median ρfixed\rho_{\mathrm{fixed}} rising from

Monotonic sensitivity of molecular line fluxes to the external FUV field, showing Spearman rank correlations for molecular transitions grouped by species

Figure 2. Monotonic sensitivity of molecular line fluxes to the external FUV field, quantified by the Spearman rank correlation ρ\rho between line flux and FUV field strength FFUVF_{\mathrm{FUV}}. Each row is a single transition, grouped by species (colour). For each line, a higher Spearman ρ\rho value indicates a stronger correlation between the FUV field and the line flux. Open circles show the grid-wide correlation (ρgrid\rho_{\mathrm{grid}}), evaluated across all models, in which M∗M_{*}, rdr_{\mathrm{d}}, Σ1au\Sigma_{\mathrm{1au}} and FFUVF_{\mathrm{FUV}} all vary. Filled circles show the median fixed-structure correlation (ρfixed\rho_{\mathrm{fixed}}), computed separately within each set of models that share the same (M∗M_{*}, rdr_{\mathrm{d}}, Σ1au\Sigma_{\mathrm{1au}}) and differ only in FFUVF_{\mathrm{FUV}}. Coloured bars span the central quartile range of that distribution, indicating how consistently a line responds to FUV across different disk structures. Note that hyperfine transition labels have been shortened for clarity, see Table A1 for full transition designations.

∼0.65\sim0.65 for J=1J = 1--00 to ≈1\approx1 for J=5J = 5--44, and remaining above ∼0.95\sim0.95 through J=10J = 10--99 with narrow central quartile ranges. CN shows a weaker but similarly excitation-dependent trend, with the median ρfixed\rho_{\mathrm{fixed}} increasing from ∼0.15\sim0.15 for N=1N = 1--00 to ∼0.85\sim0.85 for N=7N = 7--66, accompanied by a modest central quartile spread. C2_2H and N2_2H+^{+} show qualitatively similar increases with excitation, but with weaker overall correlations. Their median ρfixed\rho_{\mathrm{fixed}} reaches only ∼0.7\sim0.7--0.80.8 for the highest-excitation transitions, while at low excitation their central quartile ranges extend below zero, indicating that for an appreciable fraction of disk structures these line fluxes are insensitive to, or even anti-correlated with, the external field.

[CI] and H2_2CO exhibit the strongest divergence between the two statistics. Both [CI] transitions have ρgrid∼0.2\rho_{\mathrm{grid}} \sim0.2 but a median ρfixed∼0.8\rho_{\mathrm{fixed}} \sim0.8, although their central quartile ranges are very broad, spanning ∼0.6\sim0.6 to unity. So, while atomic carbon typically responds strongly to the external field, the strength of that response depends sensitively on the underlying disk structure. H2_2CO exhibits the most extreme contrast, with low ρgrid\rho_{\mathrm{grid}} values across all transitions (≲0.3\lesssim0.3) but a uniformly high median ρfixed\rho_{\mathrm{fixed}} (∼0.8\sim0.8--0.90.9). The H2_2CO flux therefore rises consistently with FFUVF_{\mathrm{FUV}} for any given disk structure, but this response is masked in the overall model grid by a strong dependence on the stellar and disk parameters, which we quantify in the following section. More generally, both statistics tend to increase with excitation energy within each species, such that higher-EuE_{\mathrm{u}} transitions correlate more strongly with the external field.

Overall, the line responses vary substantially between species and transitions, both in their strength and consistency across disk structures. We discuss the chemical and physical origins of these trends in § 4.1.

Degeneracies with stellar and disk parameters

The gap between ρgrid\rho_{\mathrm{grid}} and ρfixed\rho_{\mathrm{fixed}} varies substantially between transitions, indicating that variations in the stellar or disk properties obscure the intrinsic response to the external field to differing degrees. We now examine which of the remaining grid parameters are responsible. We compute the grid-wide Spearman correlations between the disk-integrated fluxes and the stellar mass (ρM∗\rho_{M_{\ast}}), disk radius (ρrd\rho_{r_{\mathrm{d}}}), and surface density normalisation (ρΣ1au\rho_{\Sigma_{\mathrm{1au}}}), evaluated across all models in the same manner as ρgrid\rho_{\mathrm{grid}}. Figure 3 presents the results as three panels, each showing the correlation with a given parameter against ρgrid\rho_{\mathrm{grid}} for every transition, coloured by species.

Spearman rank correlation coefficients between disk-integrated line fluxes and stellar and disk parameters

Figure 3. Spearman rank correlation coefficients between the disk-integrated line fluxes and the stellar and disk parameters, for all 48 molecular transitions in the model grid. Each point represents a single transition, coloured by species. From left to right, panels show the correlation of each line flux with the stellar mass (M∗M_\ast), disk outer radius (rdr_{\mathrm{d}}), and surface density at 1 au (Σ1au\Sigma_{\mathrm{1au}}), plotted against its correlation with the external FUV field strength, ρgrid\rho_{\mathrm{grid}} (§ 3.2). All coefficients are evaluated across the full model grid, in which all four parameters vary simultaneously, and they are therefore directly comparable to ρgrid\rho_{\mathrm{grid}} in Figure 2 (open circles), rather than ρfixed\rho_{\mathrm{fixed}}, which is computed at fixed disk structure. Dashed lines indicate ρ=0\rho= 0 in each dimension. Transitions towards the upper right of each panel are positively correlated with both the external FUV field and the given stellar or disk parameter.

Most transitions display a negative correlation with stellar mass (left panel), with ρM∗\rho_{M_{\ast}} typically ranging from −0.5-0.5 to 0. The anti-correlation is strongest for the lowest-EuE_{\mathrm{u}} lines and weakens with increasing EuE_{\mathrm{u}}. This is because more luminous stars heat a larger fraction of the disk, depopulating the low-EuE_{\mathrm{u}} levels that trace colder gas. High-EuE_{\mathrm{u}} transitions, on the other hand, are more efficiently excited in warmer gas, which offsets the loss of cold emitting material and weakens their anti-correlation with M∗M_{\ast}. The main exceptions are H2_2CO, whose transitions show weak to moderate positive correlations (ρM∗∼0.2\rho_{M_{\ast}} \sim0.2--0.70.7), and HCN, for which the high-JJ transitions are weakly positively correlated with stellar mass while the low-JJ transitions remain weakly anti-correlated.

Strong positive correlations are found between disk-integrated fluxes and rdr_{\mathrm{d}} for nearly all species and transitions (centre panel of Figure 3). This is expected simply because a larger disk provides a greater emitting area. However, because changing rdr_{\mathrm{d}} also modifies the temperature and chemical structure in a non-linear fashion, the degree of correlation is species- and transition-dependent, with ρrd\rho_{r_{\mathrm{d}}} varying from ∼0.4\sim0.4 to ∼0.85\sim0.85. As with the stellar mass, the correla-

tion is strongest for low-EuE_{\mathrm{u}} transitions that trace the extended outer disk, and decreases with increasing EuE_{\mathrm{u}}.

Only a weak positive correlation exists between the disk-integrated fluxes and the surface density normalisation Σ1au\Sigma_{\mathrm{1au}} (right panel of Figure 3). For all species and transitions, ρΣ1au\rho_{\Sigma_{\mathrm{1au}}} clusters around ∼0.15\sim0.15, indicating that increasing the disk surface density, and hence disk mass, generally produces only a negligible to modest change in the emergent flux. This does not imply that the disk itself makes no contribution, but rather that the emitting gas is not set by the total disk mass. Because the incident FUV radiation is attenuated rapidly at the disk surface, the irradiated layers responsible for much of the emission are comparatively unaffected by changes in Σ1au\Sigma_{\mathrm{1au}}, and the same is true of optically thick lines, which trace the temperature and area of the τ=1\tau= 1 surface rather than the column beneath it.

The disk can nevertheless contribute substantially to the emission, particularly for low-EuE_{\mathrm{u}} transitions at weak external fields, where the emission arises predominantly from the disk surface. As FFUVF_{\mathrm{FUV}} increases, a growing fraction of the emission arises from the warm photoevaporative wind, whose temperature and excitation structure are regulated primarily by the external field rather than the underlying disk. We examine this in greater detail in § 4.1.

The degeneracies discussed here demonstrate that absolute (sub-)millimetre line fluxes alone are unlikely to provide a clean diagnostic of the external FUV field. The disk radius and stellar mass, in particular, introduce correlations of comparable strength to FFUVF_{\mathrm{FUV}} for many transitions. This motivates the use of line flux ratios as diagnostics, where the dependence on parameters that affect two lines similarly can be reduced or cancelled. We explore this approach in the following section. If two transitions respond similarly to a given parameter, their shared dependence partially cancels in the ratio. However, cancellation is exact only when both lines scale by the same factor. Where their responses differ, a residual sensitivity survives and the raw Spearman correlation of the ratio with FFUVF_{\mathrm{FUV}} may still partly reflect covariance with other parameters.

Line flux ratios as diagnostics of external FUV field strength

Several species exhibit strong Spearman correlations between their disk-integrated line fluxes and the external FUV field strength, but these correlations can be significantly degenerate with stellar mass and disk radius. Line flux ratios offer a natural way to reduce these degeneracies.

To assess the dependence of line ratios on the external FUV field, we calculated all possible combinations of the 48 transitions in our grid, yielding 2256 ratios (including reciprocal pairs but excluding self-ratios). We characterise each ratio using three quantities. The first is the fixed-structure correlation, ρfixed\rho_{\mathrm{fixed}}, which, as for individual transitions, measures the rank-based relationship between the ratio and FFUVF_{\mathrm{FUV}} after accounting for variations in stellar and disk properties. We calculate this using a partial Spearman correlation controlling for M∗M_\ast, rdr_{\mathrm{d}}, and Σ1au\Sigma_{\mathrm{1au}} (e.g. (Kendall 1942)). However, a high ρfixed\rho_{\mathrm{fixed}} does not by itself ensure that a ratio is useful observationally. A ratio may follow FFUVF_{\mathrm{FUV}} monotonically but vary only weakly across several orders of magnitude in irradiation, making it difficult to distinguish different FUV environments given realistic observational uncertainties. We therefore also compute a matched enhancement factor, ηFUV\eta_{\mathrm{FUV}}, which measures the total change in a ratio across the full range of external fields considered. For each combination of M∗M_\ast, rdr_{\mathrm{d}} and Σ1au\Sigma_{\mathrm{1au}}, we divide the value of the ratio at 105 G010^{5}\,G_{0} by its value at 102 G010^{2}\,G_{0}, and take ηFUV\eta_{\mathrm{FUV}} as the median of this quantity across all combinations. A value of ηFUV=1\eta_{\mathrm{FUV}} = 1 therefore indicates no net change, while ηFUV=10\eta_{\mathrm{FUV}} = 10 and 100 indicate that the ratio increases by one and two orders of magnitude, respectively, across the three decades in FFUVF_{\mathrm{FUV}}. Because it depends only on the end points, ηFUV\eta_{\mathrm{FUV}} does not by itself guarantee a monotonic response, which is instead captured by ρfixed\rho_{\mathrm{fixed}}. Finally, to assess observability, we record the detection fraction, fdetf_{\mathrm{det}}, defined as the fraction of models in which both lines exceed the 3σ3\sigma threshold of § 3.1. Together, these metrics test whether a ratio changes consistently and measurably with FUV irradiation while both lines remain detectable.

Because reciprocal ratios carry identical information with the sign of ρfixed\rho_{\mathrm{fixed}} reversed, we consider each pair only once in our analysis below. Across the full set, ∣ρfixed∣|\rho_{\mathrm{fixed}}| ranges from 0.003 ([Cl] 3P2^{3}P_{2} 3P1^{3}P_{1}/HCN J=10−9J = 10 - 9) to 0.936 ([CI] 3P1^{3}P_{1}–3P0^{3}P_{0}/[CI] 3P2^{3}P_{2}–3P1^{3}P_{1}), and ηFUV\eta_{\mathrm{FUV}} reaches values of up to ∼125\sim125, confirming that a substantial number of ratios combine a strong, structure-independent correlation with a large dynamic range. To isolate the most practical diagnostics we apply three cuts: ∣ρfixed∣≥0.8\lvert\rho_{\mathrm{fixed}}\rvert\ge0.8, ensuring at least a strong correlation with FFUVF_{\mathrm{FUV}}; ηFUV≥5\eta_{\mathrm{FUV}} \ge5, ensuring the ratio varies by a factor large enough to be measured against realistic uncertainties (see § 4.2); and fdet≥0.5f_{\mathrm{det}} \ge0.5, ensuring both lines are detectable across at least half the model grid. These criteria yield a shortlist of 135 ratios that are well-suited to identifying and quantifying the impact of external photoevaporation, and are listed in Table C1 ordered by ρfixed\rho_{\mathrm{fixed}}.

Line flux ratioALMA bandsρfixed\rho_{\mathrm{fixed}}ηFUV\eta_{\mathrm{FUV}}fdetf_{\mathrm{det}}
[CI] 3P2^3P_2-3P1^3P_1 / CN N=1–0B10 / B30.918.60.80
HCO+^+ J=7–6 / CO J=1–0B9 / B30.9111.51.00
CO J=7–6 / CO J=1–0B10 / B30.915.61.00
CO J=8–7 / CO J=1–0B10 / B30.917.21.00
CO J=8–7 / HCN J=2–1B10 / B50.9112.60.99
HCO+^+ J=4–3 / CN N=1–0B7 / B30.9046.00.80
HCO+^+ J=3–2 / CN N=1–0B6 / B30.9044.30.80
HCO+^+ J=5–4 / CO J=1–0B8 / B30.9010.11.00
HCO+^+ J=2–1 / CN N=1–0B5 / B30.9042.20.80
HCO+^+ J=5–4 / CO J=2–1B8 / B60.906.11.00
[CI] 3P1^3P_1-3P0^3P_0 / CN N=1–0B8 / B30.906.30.80
CO J=8–7 / HCN J=3–2B10 / B60.908.71.00
HCO+^+ J=1–0 / CN N=1–0B3 / B30.8930.20.60
HCO+^+ J=8–7 / CO J=1–0B9 / B30.8912.31.00
HCO+^+ J=5–4 / CN N=1–0B8 / B30.8946.50.80
HCO+^+ J=4–3 / CN N=2–1B7 / B60.8926.10.98
HCO+^+ J=3–2 / CN N=2–1B6 / B60.8925.50.98
CO J=8–7 / C2_2H N=3–2B10 / B60.8923.20.58
CO J=7–6 / HCN J=2–1B10 / B50.8910.10.99
HCO+^+ J=3–2 / C2_2H N=3–2B6 / B60.8925.40.58
HCO+^+ J=2–1 / C2_2H N=3–2B5 / B60.8923.40.58
HCO+^+ J=2–1 / CN N=2–1B5 / B60.8924.60.91
HCO+^+ J=5–4 / CN N=2–1B8 / B60.8827.40.98
HCO+^+ J=4–3 / C2_2H N=3–2B7 / B60.8825.70.58
CO J=7–6 / C2_2H N=3–2B10 / B60.8818.30.58
HCO+^+ J=7–6 / CO J=2–1B9 / B60.886.41.00
CO J=8–7 / CN N=1–0B10 / B30.8843.90.80
HCO+^+ J=4–3 / CO J=1–0B7 / B30.889.61.00
HCO+^+ J=1–0 / CN N=2–1B3 / B60.8717.90.63
HCO+^+ J=4–3 / CO J=2–1B7 / B60.875.81.00
CO J=7–6 / HCN J=3–2B10 / B60.876.81.00
HCO+^+ J=1–0 / C2_2H N=3–2B3 / B60.8718.70.53
CO J=7–6 / CN N=1–0B10 / B30.8732.60.80
CO J=8–7 / CN N=2–1B10 / B60.8724.90.98
CO J=6–5 / HCN J=2–1B9 / B50.878.20.99
HCO+^+ J=7–6 / CN N=1–0B9 / B30.8759.30.80
HCO+^+ J=3–2 / C2_2H N=4–3B6 / B70.8716.80.65
CO J=6–5 / C2_2H N=3–2B9 / B60.8614.50.58
HCO+^+ J=5–4 / C2_2H N=3–2B8 / B60.8626.40.58
HCO+^+ J=2–1 / C2_2H N=4–3B5 / B70.8616.00.65
CO J=6–5 / CN N=1–0B9 / B30.8625.10.80
CO J=8–7 / HCN J=1–0B10 / B30.8622.50.87
HCO+^+ J=4–3 / CN N=3–2B7 / B70.865.50.99
CO J=8–7 / HCN J=4–3B10 / B70.867.01.00
CO J=7–6 / CN N=2–1B10 / B60.8619.00.98
HCO+^+ J=3–2 / CN N=3–2B6 / B70.8613.00.99
HCO+^+ J=7–6 / HCN J=3–2B9 / B60.8612.41.00
[CI] 3P2^3P_2-3P1^3P_1 / CN N=2–1B10 / B60.865.50.88

Table C1.

Every ratio in the shortlist correlates positively with FFUVF_{\mathrm{FUV}}. In each case a strongly responding line is placed over a more weakly responding one, so that the ratio grows as the external field increases. The numerators are drawn exclusively from the species identified in § 3.2 as the most FUV-responsive, including HCO+^{+} across its full rotational ladder, high-JJ CO, and the [CI] lines. The denominators broadly separate into two categories that trade correlation strength against dynamic range. In the first, the denominator is a low-excitation line tracing essentially the same gas as the numerator, either a low-JJ line of the same molecule or of a co-spatial species (e.g. CO J=8−7J = 8 - 7/J=1−0J = 1 - 0, CO J=8−7J = 8 - 7/HCN J=2−1J = 2 - 1, HCO+^{+} J=7−6J = 7 - 6/CO J=1−0J = 1 - 0). Because both lines share the same dependence on M∗M_{*}, rdr_d and Σ1au\Sigma_{\mathrm{1au}}, that dependence cancels almost exactly, leaving as a signal the excitation contrast produced by the warm photoevaporative wind. These ratios attain the highest ρfixed\rho_{\mathrm{fixed}} (up to ∼0.91\sim0.91) and are detectable in nearly every model (fdet≳0.8f_{\mathrm{det}} \gtrsim0.8), but their enhancement factors are modest (rFUV∼5−17r_{\mathrm{FUV}} \sim5 - 17) because the denominator also responds to the FUV field. In the second category, the denominator is a comparatively FUV-insensitive line (see § 4.2). Here the change is set almost entirely by the numerator, producing the largest enhancement factors in the shortlist (up to ηFUV=89\eta_{\mathrm{FUV}} = 89, e.g. HCO+^{+} J=10−9J = 10 - 9/CN N=1−0N = 1 - 0), at the cost of a lower detection fraction (fdet∼0.6−0.8f_{\mathrm{det}} \sim0.6 - 0.8) owing to the faintness of the denominator. The ratios with the greatest dynamic range are therefore less reliably detected, while those detectable in every disk offer more limited contrast.

Figure 4 illustrates both categories using all ratios formed from CO, HCO+^{+}, and C2_2H, with colour denoting ρfixed\rho_{\mathrm{fixed}} and circle size the enhancement factor ηFUV\eta_{\mathrm{FUV}}. The first category appears in the CO high-JJ/low-JJ blocks and HCO+^{+}/CO blocks, which are strongly correlated with FFUVF_{\mathrm{FUV}}, but have modest enhancement factors. The second category appears in the HCO+^{+}/C2_2H block, where the weakly responding C2_2H denominator produces the largest enhancement factors in the figure. In contrast, ratios between different HCO+^{+} transitions show little dependence on FFUVF_{\mathrm{FUV}} since all respond to the external field in a similar way. The complete matrix of 2256 ratios is presented in Appendix E.

Diagnostic power of molecular line flux ratios as tracers of external photoevaporation

Figure 4. Diagnostic power of molecular line flux ratios as tracers of external photoevaporation, shown for a representative subset of molecular species (CO, HCO+^{+}, C2_2H). These species are chosen to illustrate the two classes of diagnostic identified in § 3.4: excitation-sensitive ratios (e.g. CO high-JJ/low-JJ, HCO+^{+}/CO) and chemistry-sensitive ratios with a denominator that responds weakly to the external FUV field (e.g. HCO+^{+}/C2_2H). Each circle represents the ratio of the transition on the yy-axis (numerator) to that on the xx-axis (denominator), with colour denoting ρfixed\rho_{\mathrm{fixed}} and circle size the enhancement factor ηFUV\eta_{\mathrm{FUV}}. Blue indicates that the ratio increases with FFUVF_{\mathrm{FUV}} and red that it decreases. Wedges along the axes indicate increasing EuE_u within each species (see Table A1). The full matrix for all 8 species is shown in Appendix E.

Figure 5 extends this comparison, showing the joint distribution of ρfixed\rho_{\mathrm{fixed}} and ηFUV\eta_{\mathrm{FUV}} alongside the detectability of each pair. The figure highlights the trade-off between correlation strength, dynamic range, and detectability, as well as the fraction of ratios that satisfy our selection criteria. Ratios over low-excitation CO attain the highest ρfixed\rho_{\mathrm{fixed}} but only modest enhancement factors, while those over CN and C2_2H reach the largest dynamic ranges at the cost of detectability.

Diagnostic performance of all 2256 line flux ratios

Figure 5. Diagnostic performance of all 2256 line flux ratios, shown as the partial Spearman correlation with the external FUV field strength, ρfixed\rho_{\mathrm{fixed}}, against the median FUV enhancement factor, ηFUV\eta_{\mathrm{FUV}}. Points are coloured by the species forming the denominator of the ratio, and grey crosses mark ratios that fall below our detectability threshold (fdet<0.5f_{\mathrm{det}} < 0.5). Dotted lines mark the cuts at ρfixed=0.8\rho_{\mathrm{fixed}} = 0.8 and ηFUV=5\eta_{\mathrm{FUV}} = 5, and the shaded region indicates where both are satisfied, containing the 135 shortlisted ratios of Table C1.

We note that satisfying all three criteria does not necessarily make a ratio the optimal diagnostic for a particular target. The ranking rewards a strong, structure-independent correlation but is largely blind to practical observability beyond the coarse fdetf_{\mathrm{det}} cut. Many of the highest-ranked ratios pair widely separated frequencies, often involving Band 9 or Band 10, which demand excellent atmospheric conditions and are therefore considerably more expensive to observe than e.g. a same-band pair such as HCO+^{+} J=3−2J = 3 - 2/C2_2H N=3−2N = 3 - 2. The detection fraction is similarly a relatively crude filter. A ratio detectable in only a minority of systems may remain an excellent diagnostic within that subset, and even a non-detection of one line yields a useful limit on the ratio. The most appropriate diagnostic therefore depends on the target and the specific science goal. We turn to these practical considerations, and to guidance on applying the ratios to real observations, in § 4.2.

Discussion

Chemical origins of the line flux trends

We have shown that most line fluxes increase with FFUVF_{\mathrm{FUV}} when the stellar and disk properties are held fixed, but that the strength and consistency of this response varies substantially between species and transitions. To explore the origin of these differences, we first consider how the abundance of each species, and the emitting region of its highest-excitation transition, change with FFUVF_{\mathrm{FUV}} for a representative model (M∗=0.6 M⊙M_{*} = 0.6\,M_{\odot}, Σ1au=1000 g cm−2\Sigma_{\mathrm{1au}} = 1000\,\mathrm{g\,cm^{-2}}, rd=40 aur_d = 40\,\mathrm{au}; Figure 6). At the weakest field (102 G010^{2}\,G_{0}), the emission of every species arises from a thin layer close to the disk surface. As the field strengthens, [CI], C2_2H, CO, CN, HCO+^{+} and HCN become increasingly abundant in the photoevaporative wind, and their emitting regions shift outwards to follow. In contrast, H2_2CO and N2_2H+^+ remain in low abundance in the wind, and their emission stays confined to the disk–wind interface. We now examine the chemistry underlying these changes for each species in turn.

Abundance maps for the eight chemical species considered in this work

Figure 6. Abundance maps for the eight chemical species considered in this work (rows) for a representative model (M∗=0.6 M⊙M_\ast= 0.6\,M_\odot, Σ1au=1000 g cm−2\Sigma_{1\mathrm{au}} = 1000\,\mathrm{g\,cm}^{-2}, rd=40 aur_{\mathrm{d}} = 40\,\mathrm{au}) at external FUV field strengths of 10210^{2}–10510^{5} G0G_{0} (columns). Black contours enclose the central 25% (solid), 50% (dashed) and 75% (dotted) of the emission of the highest-EuE_{\mathrm{u}} transition of each species (Table A1), computed for a face-on disk. The white dotted line marks the disk–wind boundary, defined as the gas density contour equal to the midplane density where τ=1\tau= 1. With increasing FFUVF_{\mathrm{FUV}}, the emission of [CI], C2_2H, CO, CN, HCO+^{+} and HCN shifts from the disk surface into the photoevaporative wind, whereas H2_2CO and N2_2H+^{+} remain confined to the disk–wind interface.

HCO+^+ displays the most consistent intrinsic response to external irradiation of any species in our grid. All transitions have median ρfixed=1\rho_{\mathrm{fixed}} = 1, with negligible variation between disk structures, showing that their fluxes increase monotonically with FFUVF_{\mathrm{FUV}} when the stellar and disk properties are held fixed. The lower ρgrid\rho_{\mathrm{grid}} values reflect variations induced by these other parameters rather than an inconsistent response to the external irradiation field. Although the peak HCO+^+ abundance remains approximately constant (HCO+^+/H ∼10−8\sim10^{-8}), the high abundance layer moves outwards and expands into the photoevaporative wind when FFUVF_{\mathrm{FUV}} is increased. Within this layer, HCO+^+ is predominantly produced through FUV-driven chemistry. C+^+ reacts with OH to form CO+^+, which is subsequently converted to HCO+^+ either directly or via HOC+^+. A secondary minor formation pathway is the reaction of CH3+_3^+ with atomic O. The outer boundary of the high-abundance layer is set by rapid dissociative recombination with electrons, while the inner boundary is controlled by attenuation of the FUV field, which reduces the C+^+ abundance and suppresses the main formation pathway. Stronger irradiation shifts and broadens this region into the wind, increasing the volume of warm HCO+^+-emitting gas. Deeper in the disk, where the external FUV is attenuated, HCO+^+ instead forms predominantly through proton transfer from H3+_3^+ to CO, and its abundance is regulated by ionisation from stellar X-rays and cosmic rays (e.g. Cleeves et al. 2014, 2015). These are held fixed across our grid, but variations in either would alter this disk component, and could therefore affect HCO+^+-based ratios at weak external fields, where the disk contribution is largest.

CO behaves similarly to HCO+^+ in the wind, maintaining a high peak abundance (CO/H ∼10−4\sim10^{-4}) at all field strengths and filling a progressively larger volume as FFUVF_{\mathrm{FUV}} increases. Every CO transition has median ρfixed=1\rho_{\mathrm{fixed}} = 1 with negligible variation between disk structures, so the fluxes increase monotonically with the external field whenever the stellar and disk properties are held fixed. The grid-wide correlations are considerably lower (ρgrid∼0.3\rho_{\mathrm{grid}} \sim0.3–0.5, rising with EuE_u) and fall below the corresponding HCO+^+ values across the full ladder. This stronger dilution arises because CO remains abundant throughout the disk as well as the wind, so the low-JJ lines receive a large contribution from gas whose emission is regulated by the disk radius and stellar mass rather than by the external field. Higher-JJ transitions preferentially trace the warm wind and are correspondingly less diluted. Uniquely among the species considered here, increasing FFUVF_{\mathrm{FUV}} also enhances the gas-phase CO abundance within the disk itself. This occurs because CO has a relatively low sublimation temperature (∼20\sim20 K). Although the external FUV is efficiently attenuated at the disk surface even at the highest field strengths, reprocessing of FUV by the photoevaporative wind into infrared radiation can raise dust temperatures within the disk by a few kelvin, sufficient to desorb a substantial fraction of the CO ice reservoir (Keyte & Haworth 2025). Nevertheless, the dominant effect of increasing FFUVF_{\mathrm{FUV}} on the total CO column is still the expansion of the warm wind. Both the desorption of CO ice and the additional emission from the wind may contribute to the weak CO depletion inferred in compact ONC disks by Boyden & Eisner (2023), although testing this would require source-specific models that include the photoevaporative wind.

CN and HCN both respond positively to increasing FFUVF_{\mathrm{FUV}}, but the strength and consistency of this response depend on rotational excitation. For HCN, the median ρfixed\rho_{\mathrm{fixed}} rises from ∼0.65\sim0.65 for J=1J = 1–0 to unity for J=5J = 5–4 and remains above ∼0.95\sim0.95 at higher JJ, with little variation between disk structures. CN shows a much stronger dependence on excitation, with ρfixed\rho_{\mathrm{fixed}} increasing from ∼0.15\sim0.15 for N=1N = 1–0 to ∼0.85\sim0.85 for N=7N = 7–6. As for the other species, the grid-wide correlations are lower, reflecting the additional scatter introduced by variations in the stellar and disk properties. These trends are linked to the coupled chemistry of the two species. CN forms predominantly through the photodissociation of HCN and is destroyed primarily through CN + H2_2 →\rightarrow HCN + H, which is also a major HCN formation pathway alongside the dissociative recombination reaction HCNH+^+ + e−^- →\rightarrow HCN + H. CN is therefore most abundant in the exposed gas where HCN is photodissociated, and extends farther into the wind, whereas HCN is concentrated in a layer coincident with, or lying beneath, the CN-rich region. As FFUVF_{\mathrm{FUV}} increases, the high-abundance layers of both species expand into the warm wind. The higher-excitation transitions preferentially trace this warm gas and so brighten consistently with FFUVF_{\mathrm{FUV}}. The lower-excitation lines are more easily excited in cooler gas and receive a larger contribution from the disk, particularly at weak fields, which dilutes their response to the wind and produces the rise in ρfixed\rho_{\mathrm{fixed}} with EuE_u seen for both species.

Atomic C has previously received particular attention as a tracer of the photoevaporative flow interior to the ionisation front, through both the sub-millimetre [C I] fine-structure lines and the optical [C I] 8727 Å line (Haworth & Owen 2020; Haworth et al. 2022; Aru et al. 2024). In our models, both [C I] transitions show a large gap between the two statistics, with ρgrid∼0.2\rho_{\mathrm{grid}} \sim0.2 but a median ρfixed∼0.8\rho_{\mathrm{fixed}} \sim0.8. Their broad centrality ranges further indicate that, although the [C I] flux generally increases with FFUVF_{\mathrm{FUV}} at fixed disk structure, the strength and consistency of this response depend sensitively on the underlying stellar and disk properties. Neutral carbon occupies a relatively narrow layer between the CO-rich molecular gas and the C+^+-dominated outer wind, bounded on the inside by the transition to CO and on the outside by photoionisation of C to C+^+. As FFUVF_{\mathrm{FUV}} increases, this layer shifts outwards and extends further into the photoevaporative

wind, increasing the volume of warm gas in which atomic carbon is abundant. Since both [CI] transitions have upper-level energies well below the typical gas temperature in this layer (Eu≈24E_{\mathrm{u}} \approx24 and 62 K), this additional gas is readily excited and contributes to the emission. The position and thickness of the layer are set by the balance between CO photodissociation and C photoionisation, which in turn depends on the density structure of the wind and hence on how rapidly the incident FUV radiation is attenuated. Because the ability of the wind to shield its inner regions varies substantially across the grid, so do the location and volume of the atomic carbon layer. This helps explain why [CI] correlates strongly with FFUVF_{\mathrm{FUV}} within each disk structure, but only weakly when all models are combined.

H2_2CO exhibits the largest gap between ρgrid\rho_{\mathrm{grid}} and ρfixed\rho_{\mathrm{fixed}} of any species in our grid. Its ρgrid\rho_{\mathrm{grid}} values are among the lowest we measure (≲0.3\lesssim0.3 for all transitions), yet its median ρfixed\rho_{\mathrm{fixed}} is uniformly high (∼0.8\sim0.8–0.90.9), so the flux rises consistently with FFUVF_{\mathrm{FUV}} at fixed disk structure while this response is almost entirely masked across the full grid. This reflects the fact that H2_2CO is the only species whose fluxes are all positively correlated with stellar mass (ρM∗∼0.2\rho_{M_*} \sim0.2–0.70.7; § 3.3), so that variations in M∗M_* imprint scatter comparable in magnitude to the FUV response itself. H2_2CO is confined to a thin layer on the disk surface at 10210^2 G0_0 and spreads into the wind at higher FFUVF_{\mathrm{FUV}}, but remains relatively diffuse and in low abundance. In our models it is produced predominantly through gas-phase neutral–radical reactions in the warm irradiated gas such as CH3_3 + O →\rightarrow H2_2CO + H (rather than by grain-surface hydrogenation of CO, which typically dominates in colder disk regions). Its destruction is controlled by proton transfer to the rising HCO+^+ abundance (HCO+^+ + H2_2CO →\rightarrow H3_3CO+^+ + CO), which counteracts the outward expansion of the H2_2CO layer and limits the growth of the emitting volume as FFUVF_{\mathrm{FUV}} increases.

C2_2H shows a weaker and more structure-dependent response to FFUVF_{\mathrm{FUV}} than the species discussed above, particularly at low excitation. Its median ρfixed\rho_{\mathrm{fixed}} rises from ∼0.2\sim0.2 for N=1−0N = 1 - 0 to ∼0.7\sim0.7 for the highest-NN transitions, while the central quartile ranges of the low-NN lines extend below zero. Thus, although higher-excitation C2_2H emission generally strengthens with increasing FFUVF_{\mathrm{FUV}}, the low-excitation flux can be insensitive to, or even decrease with, the external field for some disk structures. Its ρgrid\rho_{\mathrm{grid}} values are lower still, increasing from approximately zero at N=1−0N = 1 - 0 to ∼0.4\sim0.4 at N=7−6N = 7 - 6, showing that variations in the underlying stellar and disk properties further dilute this intrinsic trend. Its abundance morphology closely resembles that of atomic C, forming a thin shell near the disk surface that shifts outwards into the photoevaporative wind as FFUVF_{\mathrm{FUV}} increases. In our models, C2_2H forms primarily through the dissociative recombination of C2_2H2+_2^+ and C2_2H3+_3^+. The production of these parent ions is promoted by FUV-driven ionisation and photochemistry within the hydrocarbon network. By contrast, C2_2H destruction is dominated almost entirely by photodissociation. The C2_2H-rich shell is therefore bounded on the inside by the availability of the hydrocarbon ions required for its formation, and on its outside by rapid photodestruction. As FFUVF_{\mathrm{FUV}} increases, the shell moves outwards and reaches its greatest spatial extent at FFUV=104F_{\mathrm{FUV}} = 10^4 G0_0. At FFUV=105F_{\mathrm{FUV}} = 10^5 G0_0, it contracts slightly as the increased photodissociation rate begins to outpace its formation pathways.

N2_2H+^+ shows the weakest response of any species considered here. Its median ρfixed\rho_{\mathrm{fixed}} rises from ∼0.15\sim0.15 at J=1−1J = 1 - 1 to ∼0.8\sim0.8 at J=5−4J = 5 - 4, but the central quartile ranges of the low-JJ lines extend well below zero, indicating that for an appreciable fraction of disk structures the flux is insensitive to, or even anti-correlated with, the external field. Its ρgrid\rho_{\mathrm{grid}} values are close to zero across the entire ladder (≲0.2\lesssim0.2). This behaviour follows from the tight coupling of N2_2H+^+ to CO. It forms through H+^+ + N2_2 →\rightarrow N2_2H+^+ + H2_2 and is destroyed rapidly by CO + N2_2H+^+ →\rightarrow HCO+^+ + N2_2, so it survives only where CO is photodissociated while N2_2 and H3+_3^+ remain available. This confines the species to a narrow layer near the disk–wind interface, which expands only modestly with increasing FFUVF_{\mathrm{FUV}}, while within the disk the FUV-driven desorption of CO ice described above further suppresses the reservoir. The emission is typically the faintest of any species considered here, with median fluxes that fall below our detectability threshold across all lines, although the brightest models exceed it for J=2−1J = 2 - 1 and above (Figure 1). This contrasts with the growing number of N2_2H+^+ detections in isolated disks. However, these detections come predominantly from systems with extensive cold outer regions, where CO is frozen out beyond the snowline and no longer destroys N2_2H+^+ (e.g. Qi et al. 2013, 2015; van ’t Hoff et al. 2017; Qi et al. 2019; Cataldi et al. 2021). Such conditions are largely absent from our models. The disks in our grid are mostly compact and so contain little cold gas beyond the CO snowline, while the wind-driven heating described above desorbs much of the CO ice that remains. The CO-depleted reservoir in which N2_2H+^+ is normally abundant is therefore small or absent, and the N2_2H+^+ abundance is correspondingly suppressed. We note, however, that N2_2H+^+ is also among the species most sensitive to the adopted chemical network, as discussed in § 4.3.

Practical selection of line-ratio diagnostics for ALMA

The ratios listed in Table C1 respond strongly and consistently to the external FUV field across our model grid, but no single ratio is optimal for every target. These ratios have two main observational applications. In surveys of unresolved disks, elevated wind-sensitive ratios can identify systems whose molecular emission is strongly affected by external irradiation. For systems already known or suspected to be externally irradiated, comparison with the model grid can constrain FFUVF_{\mathrm{FUV}}, with the precision depending on how well the stellar and disk properties are known. The practical choice of ratio must additionally account for whether the two transitions can be observed in the same spectral setup, how observational and model uncertainties affect the inferred FFUVF_{\mathrm{FUV}}, and whether either line is likely to be contaminated by emission from the surrounding environment.

The shortlisted ratios broadly fall into two classes. The first comprises excitation-sensitive ratios in which a high-EuE_{\mathrm{u}} transition is divided by a low-EuE_{\mathrm{u}} transition of the same molecule or of a species with a similar spatial distribution. Examples include CO J=8−7J = 8 - 7 / CO J=1−0J = 1 - 0 and HCO+^+ J=7−6J = 7 - 6 / CO J=1−0J = 1 - 0. At low and moderate FFUVF_{\mathrm{FUV}}, the low-excitation line receives a substantial contribution from the cooler disk, whereas the high-excitation line preferentially traces the warm photoevaporative wind. At higher FFUVF_{\mathrm{FUV}}, both lines arise largely in the wind, but the high-excitation line is sub-thermally excited at typical wind densities and therefore brightens more steeply as the wind becomes denser and warmer. Their similar dependence on disk size, stellar mass, and surface density produces high ρfixed\rho_{\mathrm{fixed}} and, in many cases, detection fractions close to unity. Their main disadvantage is that the large excitation contrast generally corresponds to a large frequency separation, often requiring observations in widely separated receiver bands.

The second class exploits differences in chemical and spatial response rather than excitation alone. The clearest example is HCO+^+ J=3−2J = 3 - 2 / C2_2H N=3−2N = 3 - 2. As the external field increases, HCO+^+ develops a substantially more extended wind contribution, whereas the low-excitation C2_2H emission responds much more weakly to FFUVF_{\mathrm{FUV}} (Figure 6). The same principle applies to ratios in which H2_2CO or a low-frequency CN transition forms the denominator. These ratios can provide large dynamic ranges without requiring a large separation in excitation energy, although their detection fractions are often lower because the more weakly responding denominator is comparatively faint.

For diagnostics with broadly similar predictive power, the possibility of observing both transitions in a single setup can determine the preferred choice. Of the 135 ratios in Table C1, 13 combine transitions within the same ALMA receiver band. Whether such a pair can be measured simultaneously is determined by the band-dependent intermediate frequency range and the placement of the spectral windows. Using the current ALMA frequency coverage, five of the shortlisted ratios can be accommodated within a single tuning: HCO+^{+}/HCN for the matched J=2J = 2–1, 3–2, 4–3, 5–4, and 7–6 transitions (corresponding to Bands 5, 6, 7, 8 and 9). These ratios have ρfixed≃0.81\rho_{\mathrm{fixed}} \simeq0.81–0.84, ηFUV≃9.3\eta_{\mathrm{FUV}} \simeq9.3–15.8, and fdet≃0.91f_{\mathrm{det}} \simeq0.91–1.00. They therefore provide somewhat less dynamic range than the strongest cross-band diagnostics, but offer a strong combination of detectability and observational efficiency.

Observing both lines in a single tuning also improves the relative calibration of the ratio, since uncertainties on the absolute flux and atmospheric transmission are largely common to both transitions. Ratios constructed from separate observations retain the calibration uncertainty associated with each band and may also differ in weather, elevation, synthesised beam, and uvuv coverage. Adopting typical flux uncertainties of 5% in Bands 1–5, 10% in Bands 6–8, and 20% in Bands 9–10 (Cortes et al. 2026), and treating the two measurements as independent, gives calibration uncertainty contributions of approximately 11% for a B3/B6 ratio, 21% for a B3/B10 ratio, and up to 28% when both observations use Bands 9–10.

The effect of this uncertainty on the inferred external field depends on the dynamic range of the diagnostic. Ratios with large enhancement factors are relatively robust to observational uncertainty, whereas ratios with more limited dynamic ranges provide weaker constraints. For example, a 20% uncertainty in the measured ratio corresponds to an uncertainty of roughly a factor of 1.5 in FFUVF_{\mathrm{FUV}} for the strongest diagnostics, but can increase to a factor of 2–3 for ratios with smaller enhancement factors (ηFUV=3\eta_{\mathrm{FUV}} = 3–5). These values are only approximate, because the relation between the ratio and FFUVF_{\mathrm{FUV}} is not necessarily uniform across the grid, and residual dependence on M∗M_{*}, rdr_{\mathrm{d}}, and Σ1au\Sigma_{\mathrm{1au}} introduces additional scatter. The choice of line ratio should therefore account for how well the relevant disk and stellar properties of the target are constrained.

Ratios involving Bands 9 and 10 are subject to additional complications. The low precipitable water vapour (PWV) and stable atmospheric phase required at these frequencies occur during only a limited fraction of observing time, with their availability depending on season, time of day, target elevation, and array configuration (Cortes et al. 2026). However, this cost is partly offset by the density of useful transitions at high frequency. For example, the CN N=7N = 7–6, CO J=7J = 7–6 and [C I] 3P2{}^{3}P_{2}–3P1{}^{3}P_{1} transitions span only 16 GHz and can be captured across the two sidebands of a single Band 10 tuning. Where a high-frequency observation is undertaken, multiple transitions can be included within a single spectral setup at no additional cost.

Since nearly all of the shortlisted ratios already require two tunings, those tunings can be chosen to deliver several diagnostics rather than one. The numerators in Table C1 are drawn from a small number of FUV-responsive species, whereas the denominators are predominantly low-excitation lines of CO, CN, HCN and C2_2H. A single tuning can therefore collect several numerators, a second several denominators, and every cross-pairing between them becomes available for the cost of two observations. For example, pairing Bands 7 and 3 allows CO J=3J = 3–2 and HCO+^{+} J=4J = 4–3 to be placed in opposite sidebands of a single Band 7 tuning, while CN N=1N = 1–0 and CO J=1J = 1–0 share a single sideband in Band 3, yielding three shortlisted ratios. Similar pairings are possible across other bands. The number of lines that can be captured at once is ultimately limited by the sideband separation, but this restriction can be lifted using spectral scan mode, which covers a continuous frequency range with several tunings rather than a single LO setting (Cortes et al. 2026). Band 7 is the most useful for our purposes, containing five relevant transitions within 17 GHz. A scan across this region captures all five, yielding two additional shortlisted ratios that would otherwise require separate tunings.

A final consideration relates to the wider environments in which externally irradiated disks are typically found. For example, many benchmark systems lie within the Orion Nebula Cluster (O’Dell & Wen 1994; Bally et al. 2000; Mann et al. 2014), where spatially extended molecular emission can contaminate disk spectra or produce absorption against a bright cloud background (Bally et al. 2015; Boyden & Eisner 2020). Comparable complications are likely in other irradiated regions, such as NGC 2024 and NGC 1977, where disks are likewise found in dense molecular environments or associated with externally driven ionised flows (e.g. Graf et al. 2012; Kim et al. 2016; Haworth et al. 2021). Low-JJ CO transitions are especially vulnerable because the ambient cloud is bright at similar excitation temperatures and velocities (Boyden & Eisner 2023). Free-free continuum from the ionised photoevaporative envelope can also contribute substantially at long millimetre wavelengths, reducing molecular-line contrast and complicating continuum subtraction (Ballering et al. 2023; Boyden & Eisner 2024). Chemistry-based ratios that avoid low-JJ CO are therefore preferable in strongly nebulous regions.

We note that the criteria adopted in § 3.4 exclude a further 28 ratios that satisfy the correlation and enhancement thresholds, but have fdet<0.5f_{\mathrm{det}} < 0.5. All contain C2_2H in the denominator, paired with CO, HCO+^{+}, or [C I] in the numerator. These ratios remain strongly correlated with the external field, with ρfixed=0.80\rho_{\mathrm{fixed}} = 0.80–0.92, and span substantial dynamic ranges of ηFUV=8.8\eta_{\mathrm{FUV}} = 8.8–79.0, but are limited by the faintness of the low-excitation C2_2H lines. Ratios using HCO+^{+} N=1N = 1–0 provide the largest contrasts, including HCO+^{+} J=10J = 10–9 / C2_2H N=1N = 1–0 with ηFUV=79.0\eta_{\mathrm{FUV}} = 79.0, but both transitions are detectable in only ∼8% of the model grid. Using C2_2H N=2N = 2–1 instead raises the detection fraction to ≃0.42–0.43, while retaining enhancement factors of 8.8–54.6. These ratios may therefore remain valuable where sufficiently deep observations are feasible. Conversely, a secure detection of the wind-sensitive numerator together with a non-detection of C2_2H provides a lower limit on the ratio and may therefore place a useful lower bound on FFUVF_{\mathrm{FUV}}.

Finally, we note that the co-observability classification adopted here reflects ALMA’s current capabilities. The upcoming Advanced Technology ALMA Correlator, delivered as part of the Wideband Sensitivity Upgrade, will increase the instantaneous bandwidth to ≥\geq 16 GHz per sideband (Carpenter et al. 2023). Many of the ratios that currently require two independent tunings will become accessible within a single spectral setup, decreasing their observational cost, and the balance of recommendations presented here should be revisited accordingly.

Caveats and limitations

Our models omit external EUV radiation, and therefore do not include an ionisation front. In reality, the ionisation front truncates the photoevaporative flow, and the PDR becomes more compact as the external field increases. Since most of the emission in our strongly irradiated models comes from the wind, this would reduce the emitting volume at high FFUVF_{\mathrm{FUV}} and potentially weaken the trends we identify. The effect should be small over much of our parameter space, where the ionisation front lies far from the disk, but becomes significant for systems in which it sits close to the disk surface (e.g. Johnstone et al. (1998)). Our predicted fluxes at the highest field strengths should therefore be treated as upper limits. How much this matters depends on the irradiating star, since the EUV-to-FUV ratio of the incident field rises with effective temperature. Our diagnostics are therefore best suited to regions dominated by cooler massive stars, and complement the optical line ratios of Peake et al. (2025), which trace the ionised outflow beyond the ionisation front and work best in the extreme environments where our molecular emitting region is most reduced.

The geometry of our models is also simplified. We assume an isotropic external radiation field and a spherically diverging wind, whereas real proplyds are irradiated from a particular direction and develop the familiar cometary morphology (e.g. O’Dell & Wen (1994)). Each model also represents a static snapshot of the disk–wind structure. In reality, external photoevaporation progressively truncates the disk, which in turn alters the wind it supplies and hence the wind-dominated emission. Our grid captures this only through the range of disk radii sampled (§ 3.3), rather than by following the evolution self-consistently. Additionally, all synthetic fluxes presented in the main analysis assume a face-on disk. We assess the sensitivity of our results to disk inclination in Appendix D, and find that most of the shortlisted ratios are only weakly affected, though those pairing HCO+\mathrm{HCO}^{+} with CO, HCN, or (in a few cases) higher-NN CN lose part of their dynamic range when viewed edge-on.

There are also a number of uncertainties associated with the chemical modelling itself. Our models are static, so the chemistry is solved at fixed position and does not follow material as it is advected outwards through the wind. Where the flow timescale is shorter than the relevant chemical timescale, the resulting wind abundances may differ from those predicted here. For example, one-dimensional models that include advection show that molecular ices can survive much farther into the flow than static models predict (Keyte & Ran (2025)). More generally, thermochemical codes differ from one another in a number of ways, including their chemical networks, heating and cooling prescriptions, treatment of molecular excitation, and can produce noticeably different line fluxes for the same physical structure. Absolute fluxes are therefore the least reliable of our predictions. Ratios are less affected, since these systematic differences act in a similar way on both lines and partially cancel, in the same way that they cancel the dependence on the stellar and disk properties.

This cancellation does not, however, extend to uncertainties that affect only one species, which arise from e.g. our choice of chemical network. Here, we have used a reduced subset of UMIST06 (Woodall et al. (2007)), which includes the principal formation and destruction pathways of the species considered here but omits larger molecules and some reaction channels. Comparisons between chemical networks show that the (sub-)millimetre fluxes of species like HCO+\mathrm{HCO}^{+}, HCN, and N2H+\mathrm{N}_{2}\mathrm{H}^{+} in particular can depend on the network adopted, with N2H+\mathrm{N}_{2}\mathrm{H}^{+} among the most sensitive species to the underlying chemical assumptions (e.g. Kamp et al. (2017); Wakelam et al. (2019)). C2\mathrm{C}_{2}H is also likely to be affected, since our network lacks some of the larger hydrocarbons that can act as both carbon reservoirs and photodissociative sources of C2\mathrm{C}_{2}H, and its abundance is additionally sensitive to our fixed C/O = 0.5 (e.g. Bergin et al. (2016); Bosman et al. (2021)). N2H+\mathrm{N}_{2}\mathrm{H}^{+} does not appear in our shortlist, but HCO+\mathrm{HCO}^{+}, HCN and C2\mathrm{C}_{2}H do, so the absolute values and detection fractions of the affected ratios should be treated with caution. Our diagnostics, however, are based on relative changes with FFUVF_{\mathrm{FUV}} computed with the same net-

work across the grid, which are expected to be more robust than absolute fluxes.

Finally, the criteria used to construct the shortlist in § 3.4 are deliberately simple. They identify ratios that respond strongly and consistently to the external field across our grid, but they take no account of the properties of any particular target, nor of what a given observation is intended to measure. As discussed in § 4.2, the most appropriate diagnostic depends on the source and the science goal. Table C1 should therefore be treated as a starting point for selecting lines, rather than a definitive ranking.

Conclusions

We have used a large grid of thermochemical models, tailored to the chemistry of externally irradiated protoplanetary disk + wind systems, to identify robust observational diagnostics of external photoevaporation at (sub-)millimetre wavelengths. The grid spans a broad range of stellar masses, disk radii, surface densities, and external FUV field strengths. For each model we synthesised disk-integrated fluxes for 48 transitions of eight species (CO, HCO+\mathrm{HCO}^{+}, CN, HCN, C2\mathrm{C}_{2}H, N2H+\mathrm{N}_{2}\mathrm{H}^{+}, H2\mathrm{H}_{2}CO, and atomic C), sampling a wide range of excitation energies accessible to ALMA. Our main conclusions are:

(i) Disk-integrated line fluxes vary significantly across the grid, spanning roughly 13 orders of magnitude between the brightest and faintest values. This variation is driven both by the external FUV field and by the other disk and stellar parameters. For any individual transition the full range typically covers a few orders of magnitude, while the interquartile range spans only ∼1\sim1–2 orders of magnitude, indicating that the most extreme fluxes arise from a small subset of models.

(ii) We quantify the sensitivity of each transition to the external FUV field using the Spearman rank correlation coefficient, computed in two ways. The grid-wide correlation ρgrid\rho_{\mathrm{grid}}, evaluated across all models, measures how reliably a line traces the external field in a population of disks with unknown properties. The fixed-structure correlation ρfixed\rho_{\mathrm{fixed}}, evaluated separately for each combination of stellar and disk parameters, measures how the line responds to the external field alone. As expected, ρfixed\rho_{\mathrm{fixed}} exceeds ρgrid\rho_{\mathrm{grid}} for every transition, but the size of this gap varies substantially, showing that the underlying FUV response of some species is strongly masked by variations in the other parameters. Both statistics tend to increase with excitation energy, so higher-EuE_{\mathrm{u}} transitions are generally the more sensitive tracers.

(iii) HCO+\mathrm{HCO}^{+} and CO are the most robust single-line tracers of the external field, with ρfixed≃1\rho_{\mathrm{fixed}} \simeq1 and little variation across their rotational ladders. HCN and high-NN CN transitions show similarly strong behaviour at moderate-to-high excitation. Atomic carbon ([CI]) and H2\mathrm{H}_{2}CO also respond strongly to the external field when the disk parameters are held fixed, as indicated by their high ρfixed\rho_{\mathrm{fixed}} values. However, their low ρgrid\rho_{\mathrm{grid}} values show that this response is readily obscured by variations in other disk properties, making them reliable tracers only when the disk structure is independently constrained. N2H+\mathrm{N}_{2}\mathrm{H}^{+} and C2\mathrm{C}_{2}H are the least sensitive species, particularly at low excitation, where their line fluxes are largely insensitive to, or may even decrease with, increasing external FUV field strength.

(iv) Of the secondary parameters, stellar mass and disk radius exert the strongest influence on the absolute line fluxes. Most transitions correlate negatively with stellar mass (ρM∗∼0\rho_{M_*} \sim0 to −0.5-0.5), most strongly at low EuE_{\mathrm{u}}, because more luminous stars warm a larger fraction of the disk and depopulate the low-EuE_{\mathrm{u}} levels that trace cold gas. Disk radius correlates positively with flux (ρrd\rho_{r_{\mathrm{d}}} up to ∼0.85\sim0.85), simply because larger disks present a greater emitting area. Surface density exerts only a weak positive influence (ρΣ1au∼0.15\rho_{\Sigma1\mathrm{au}} \sim0.15). Because disk radius and stellar mass introduce correlations of comparable strength to the FUV field itself, absolute line fluxes alone do not provide a clean diagnostic of external photoevaporation.

(v) We computed all 2256 pairwise flux ratios and characterised each by three quantities: the fixed-structure correlation with the FUV field (ρfixed\rho_{\mathrm{fixed}}), an enhancement factor ηFUV\eta_{\mathrm{FUV}} measuring how strongly the ratio changes between the weakest and strongest fields, and a detection fraction fdetf_{\mathrm{det}} measuring how often both lines are detectable. Applying cuts on all three yields a shortlist of 135 diagnostic ratios (Table C1), all correlating positively with the external FUV field. In each, a strongly responding line (typically HCO+^+, high-JJ CO, or [CI]) sits over a more weakly responding one. These separate into two classes: those whose numerator and denominator trace the same gas and vary due to excitation conditions, and those whose numerator and denominator trace gas of varying spatial extent due to underlying differences in the chemistry.

(vi) Selecting the best ratio for a given target requires practical considerations beyond this ranking. Almost all shortlisted ratios pair transitions in different ALMA bands, so calibration uncertainties do not fully cancel and ratios with large ηFUV\eta_{\mathrm{FUV}} are more robust. Bands 9 and 10 demand excellent atmospheric conditions but contain many useful transitions that can be captured in a single setup. More generally, two tunings can be chosen to deliver several diagnostics at once. For example, a Band 7 tuning covering CO J=3−2J = 3 - 2 and HCO+^+ J=4−3J = 4 - 3, combined with a Band 3 tuning covering CN N=1−0N = 1 - 0 and CO J=1−0J = 1 - 0, yields three shortlisted ratios. The ALMA Wideband Sensitivity Upgrade, which will at least double the instantaneous bandwidth, will bring many more ratios within a single setup, including the line-rich Band 7 window.

Our results show that no single line flux cleanly separates the effects of external irradiation from variations in intrinsic disk properties. Carefully chosen flux ratios, however, trace the increasing contribution of the warm photoevaporative wind across a wide range of disk properties. They can therefore be used to identify unresolved disks whose molecular emission is strongly shaped by external irradiation and, when combined with constraints on the stellar and disk properties, to place limits on FFUVF_{\mathrm{FUV}}. The shortlist presented here offers a practical starting point for applying this approach with ALMA in irradiated star-forming regions, particularly where individual disks cannot be spatially resolved. It also follows that, in strongly irradiated regions, unresolved molecular emission cannot be assumed to trace the disk alone, and models that neglect the wind risk drawing the wrong conclusions about the underlying disk.

Acknowledgements

LK is funded by UKRI guaranteed funding for a Horizon Europe ERC consolidator grant (EP/Y024710/1). TJH acknowledges UKRI guaranteed funding for a Horizon Europe ERC consolidator grant (EP/Y024710/1) and a Royal Society Dorothy Hodgkin Fellowship. TP is funded as part of the Dorothy Hodgkin Fellowship of TJH. AR has received funding from the Royal Society through a University Research Fellowship grant number URFR1\241791.

Data Availability

This work makes use of the open-source Python package PUFFIN, available at https://github.com/lukekeyte/PUFFIN/. The thermochemical models were computed with DALI, which is not publicly available, but the model outputs underlying this article are available from the corresponding author upon reasonable request.

Synthetic atomic and molecular transitions

DALI model parameters

Flux ratio shortlist

Table A1. Molecular and atomic transitions used in this work. Spectroscopic and collisional rate data were obtained from the Leiden Atomic and Molecular Database (LAMDA; Schöier et al. 2005), which compiles data from the following sources for the species considered here: Thomson & Dalby (1968); Allison & Dalgarno (1971); Launay & Roueff (1977); Johnson et al. (1987); Roueff & Le Bourlot (1990); Staemmler & Flower (1991); Schroder et al. (1991); Klisch et al. (1995); Jankowski & Szalewicz (2005); Müller et al. [90]; Daniel et al. (2005); Faure et al. (2007); Dumouchel et al. (2010); Yang et al. (2010); Lique et al. (2010); Spielfiedel et al. (2012); Wiesenfeld & Faure (2013); Denis-Alpizar et al. (2020).

Table C1: Ranked shortlist of line flux ratio diagnostics of the external FUV field. ρfixed\rho_{\mathrm{fixed}} is the partial Spearman correlation with FFUVF_{\mathrm{FUV}}, controlling for M∗M_\ast, rdr_{\mathrm{d}}, and Σ1au\Sigma_{1\mathrm{au}}. A ratio and its reciprocal contain identical information, with the sign of ρfixed\rho_{\mathrm{fixed}} reversed, so we only retain ratios for which ρfixed\rho_{\mathrm{fixed}} is positive. ηFUV\eta_{\mathrm{FUV}} is the median change from 10210^2 to 10510^5 G0G_0 at fixed structure. fdetf_{\mathrm{det}} is the fraction of models in which both lines exceed the 3σ3\sigma detection threshold defined in Section 3.1. Note that hyperfine transition labels have been shortened for clarity. Full transition details are presented in Table A1.

Line flux ratioALMA bandsρfixed\rho_{\mathrm{fixed}}ηFUV\eta_{\mathrm{FUV}}fdetf_{\mathrm{det}}
CO J=8J=8–77 / C2_2H N=4N=4–33B10 / B70.8613.50.65
HCO+^+ J=5J=5–44 / CN N=3N=3–22B8 / B70.8613.90.99
HCO+^+ J=4J=4–33 / C2_2H N=4N=4–33B7 / B70.8517.10.65
HCO+^+ J=8J=8–77 / HCN J=3J=3–22B9 / B60.8513.41.00
HCO+^+ J=7J=7–66 / CN N=2N=2–11B9 / B60.8532.70.98
HCO+^+ J=7J=7–66 / HCN J=2J=2–11B9 / B50.8517.00.99
CO J=8J=8–77 / CN N=3N=3–22B10 / B70.8511.20.99
HCO+^+ J=8J=8–77 / CN N=1N=1–00B9 / B30.8567.60.80
HCO+^+ J=2J=2–11 / CN N=3N=3–22B5 / B70.8512.60.91
CO J=6J=6–55 / CN N=2N=2–11B9 / B60.8514.50.98
HCO+^+ J=8J=8–77 / HCN J=2J=2–11B9 / B50.8518.20.99
HCO+^+ J=4J=4–33 / CN N=4N=4–33B7 / B80.8512.01.00
HCO+^+ J=7J=7–66 / HCN J=4J=4–33B9 / B70.8510.31.00
CO J=8J=8–77 / CN N=4N=4–33B10 / B80.8510.81.00
HCO+^+ J=5J=5–44 / HCN J=2J=2–11B8 / B50.8516.60.99
HCO+^+ J=8J=8–77 / HCN J=4J=4–33B9 / B70.8511.01.00
HCO+^+ J=5J=5–44 / CN N=4N=4–33B8 / B80.8512.11.00
CO J=7J=7–66 / HCN J=1J=1–00B10 / B30.8517.80.87
CO J=6J=6–55 / HCN J=3J=3–22B9 / B60.855.41.00
HCO+^+ J=5J=5–44 / HCN J=3J=3–22B8 / B60.8512.81.00
HCO+^+ J=3J=3–22 / CO J=1J=1–00B6 / B30.849.31.00
HCO+^+ J=4J=4–33 / HCN J=3J=3–22B7 / B60.8412.01.00
HCO+^+ J=3J=3–22 / CN N=4N=4–33B6 / B80.8411.91.00
HCO+^+ J=5J=5–44 / HCN J=4J=4–33B8 / B70.8410.11.00
HCO+^+ J=4J=4–33 / HCN J=2J=2–11B7 / B50.8417.40.99
CO J=4J=4–33 / CN N=1N=1–00B8 / B30.8416.00.80
HCO+^+ J=8J=8–77 / CO J=2J=2–11B9 / B60.847.31.00
HCO+^+ J=8J=8–77 / HCN J=5J=5–44B9 / B80.8410.30.99
HCO+^+ J=1J=1–00 / C2_2H N=4N=4–33B3 / B70.8412.20.58
HCO+^+ J=4J=4–33 / HCN J=4J=4–33B7 / B70.849.51.00
HCO+^+ J=5J=5–44 / C2_2H N=4N=4–33B8 / B70.8418.20.65
HCO+^+ J=8J=8–77 / CN N=2N=2–11B9 / B60.8437.80.98
HCO+^+ J=3J=3–22 / HCN J=2J=2–11B6 / B50.8417.50.99
HCO+^+ J=3J=3–22 / HCN J=3J=3–22B6 / B60.8411.51.00
HCO+^+ J=5J=5–44 / HCN J=5J=5–44B8 / B80.839.30.99
CO J=7J=7–66 / C2_2H N=4N=4–33B10 / B70.8310.80.65
HCO+^+ J=7J=7–66 / HCN J=5J=5–44B9 / B80.8310.50.99
CO J=7J=7–66 / CN N=3N=3–22B10 / B70.838.80.99
HCO+^+ J=7J=7–66 / C2_2H N=3N=3–22B9 / B60.8332.30.58
HCO+^+ J=10J=10–99 / CO J=1J=1–00B10 / B30.8317.30.96
HCO+^+ J=10J=10–99 / HCN J=3J=3–22B10 / B60.8316.90.96
CO J=7J=7–66 / HCN J=4J=4–33B10 / B70.835.41.00
HCO+^+ J=7J=7–66 / CN N=3N=3–22B9 / B70.8316.50.99
HCO+^+ J=3J=3–22 / CO J=2J=2–11B6 / B60.835.81.00
HCO+^+ J=2J=2–11 / CN N=4N=4–33B5 / B80.8311.50.91
HCO+^+ J=7J=7–66 / CN N=4N=4–33B9 / B80.8314.91.00
CO J=6J=6–55 / HCN J=1J=1–00B9 / B30.8314.50.87
HCO+^+ J=4J=4–33 / HCN J=5J=5–44B7 / B80.838.70.99
CO J=7J=7–66 / CN N=4N=4–33B10 / B80.838.21.00
HCO+^+ J=10J=10–99 / HCN J=5J=5–44B10 / B80.8311.20.96
HCO+^+ J=10J=10–99 / HCN J=4J=4–33B10 / B70.8312.90.96
HCO+^+ J=2J=2–11 / HCN J=2J=2–11B5 / B50.8315.80.91
HCO+^+ J=3J=3–22 / HCN J=4J=4–33B6 / B70.838.71.00
HCO+^+ J=1J=1–00 / CN N=3N=3–22B3 / B70.8310.10.63
HCO+^+ J=8J=8–77 / HCN J=1J=1–00B9 / B30.8232.20.87
Table 1 (PDF p. 19).

Table 1.

Line flux ratioALMA bandsρfixed\rho_{\mathrm{fixed}}ηFUV\eta_{\mathrm{FUV}}fdetf_{\mathrm{det}}
CO J=3–2 / CN N=1–0B7 / B30.8212.20.80
HCO+^{+} J=7–6 / HCN J=1–0B9 / B30.8231.20.87
HCO+^{+} J=10–9 / HCN J=2–1B10 / B50.8224.10.96
HCO+^{+} J=2–1 / HCN J=3–2B5 / B60.8210.50.91
CO J=4–3 / CN N=2–1B8 / B60.829.30.98
HCO+^{+} J=5–4 / HCN J=1–0B8 / B30.8228.10.87
HCO+^{+} J=10–9 / HCN J=7–6B10 / B90.8211.40.96
HCO+^{+} J=8–7 / CN N=4–3B9 / B80.8215.31.00
HCO+^{+} J=10–9 / HCN J=1–0B10 / B30.8242.50.86
HCO+^{+} J=8–7 / CN N=3–2B9 / B70.8118.70.99
HCO+^{+} J=4–3 / HCN J=1–0B7 / B30.8128.10.87
HCO+^{+} J=8–7 / C2_2H N=3–2B9 / B60.8136.90.58
HCO+^{+} J=3–2 / C2_2H N=5–4B6 / B80.819.90.62
CO J=8–7 / HCN J=5–4B10 / B80.816.20.99
HCO+^{+} J=3–2 / HCN J=1–0B6 / B30.8129.20.87
CO J=6–5 / CN N=3–2B9 / B70.817.20.99
HCO+^{+} J=7–6 / C2_2H N=4–3B9 / B70.8121.30.65
HCO+^{+} J=3–2 / HCN J=5–4B6 / B80.818.00.99
CO J=4–3 / C2_2H N=3–2B8 / B60.818.90.58
HCO+^{+} J=10–9 / CN N=1–0B10 / B30.8189.00.79
HCO+^{+} J=7–6 / HCN J=7–6B9 / B90.8110.80.98
CO J=4–3 / HCN J=2–1B8 / B50.815.30.99
HCO+^{+} J=2–1 / C2_2H N=5–4B5 / B80.819.30.62
HCO+^{+} J=8–7 / HCN J=7–6B9 / B90.8111.40.98
HCO+^{+} J=2–1 / HCN J=1–0B5 / B30.8129.60.84
HCO+^{+} J=2–1 / CO J=1–0B5 / B30.818.60.91
HCO+^{+} J=2–1 / HCN J=4–3B5 / B70.817.90.91
HCO+^{+} J=4–3 / C2_2H N=5–4B7 / B80.8110.20.62
CO J=6–5 / C2_2H N=4–3B9 / B70.808.70.65
CO J=6–5 / CN N=4–3B9 / B80.806.51.00
HCO+^{+} J=1–0 / CN N=4–3B3 / B80.809.10.63
HCO+^{+} J=10–9 / HCN J=8–7B10 / B90.8012.20.95

Table 2.

Sensitivity to disk inclination

All fluxes presented in this work assume a face-on geometry (i=0∘i = 0^\circ). Recomputing the full grid at multiple inclinations is computationally prohibitive, so we instead assess the sensitivity of our results using the representative model of Figure 6 (M∗=0.6 M⊙M_\ast= 0.6\,M_\odot, Σ1au=1000 g cm−2\Sigma_{1\mathrm{au}} = 1000\ \mathrm{g\ cm}^{-2}, rd=40 aur_{\mathrm{d}} = 40\ \mathrm{au}), with all fluxes recomputed edge-on (i=90∘i = 90^\circ) at each of FFUV=102F_{\mathrm{FUV}} = 10^2, 10310^3, 10410^4 and 105 G010^5\ G_0.

For each transition we define an inclination correction factor CX=Fi=90∘/Fi=0∘C_X = F_{i=90^\circ}/F_{i=0^\circ}, where Fi=90∘F_{i=90^\circ} is the disk-integrated flux at i=90∘i = 90^\circ and Fi=0∘F_{i=0^\circ} is the disk-integrated flux at i=0∘i = 0^\circ. Figure D1 (panel a) shows CXC_X as a function of FFUVF_{\mathrm{FUV}}. The behaviour varies according to whether emission from a given species predominantly traces the disk or the photoevaporative wind. At 102 G010^2\ G_0 the wind is negligible and the emission from all species arises from thin layers near the disk surface. In this scenario, the column density along the line of sight is much larger when the disk is edge-on compared to face-on. For the edge-on disk, median fluxes therefore increase by factors of 4.9, 5.8 and 4.0 for HCO+^+, C2_2H and CN respectively, while optically thick CO instead decreases to CCO=0.63C_{\mathrm{CO}} = 0.63 as the projected emitting area is reduced. As FFUVF_{\mathrm{FUV}} increases, these species begin to fill a large, spherical wind, and the correction converges towards unity, reaching CX=1.0C_X = 1.0–1.41.4 for CO, HCN, [CI], CN, HCO+^+ and C2_2H at 104 G010^4\ G_0 and above. The exceptions are H2_2CO and N2_2H+^+, which retain CX∼7C_X \sim7–1313 and 1010–2121 respectively at all external FUV field strengths. Neither expands into the wind, remaining confined to a narrow layer close to the disk–wind interface (as described in § 4.1).

Effect of disk inclination on the line fluxes and flux ratio diagnostics

Figure D1. Effect of disk inclination on the line fluxes and flux ratio diagnostics, based on the representative model of Figure 6 recomputed edge-on. Left: Inclination correction factor CX=Fi=90∘/Fi=0∘C_X = F_{i=90^\circ}/F_{i=0^\circ} against external FUV field strength. Faint lines show individual transitions and heavy lines the median for each species. The dashed line marks CX=1C_X = 1, where the flux is independent of viewing geometry. Right: FUV enhancement factor ηFUV\eta_{\mathrm{FUV}} at i=0∘i = 0^\circ compared with i=90∘i = 90^\circ, for the 135 ratios shortlisted in Table C1. Colour denotes the denominator species and shape the numerator. The dashed line is the 1:1 relation and the shaded region marks ηFUV<5\eta_{\mathrm{FUV}} < 5, below which a ratio no longer satisfies the selection criteria of § 3.4.

Because the correction is largest at weaker FUV fields and negligible at strong ones, its principal effect on the diagnostics is to compress or extend the dynamic range ηFUV\eta_{\mathrm{FUV}}, rather than to remove the FUV dependence itself. Figure D1 (panel b) compares ηFUV\eta_{\mathrm{FUV}} at the two inclinations for the ratios shortlisted in Table C1. Of the 135 ratios, 93 retain ηFUV≥5\eta_{\mathrm{FUV}} \ge5 when viewed edge-on. Just as ratios were used in § 3.4 to cancel the dependence on the stellar and disk properties, ratios whose numerator and denominator respond similarly to inclination also leave ηFUV\eta_{\mathrm{FUV}} essentially unchanged. This is the case for same-species excitation ratios, HCO+^+/C2_2H, and HCO+^+/CN N=1N = 1–00. Of the 42 ratios that fall below the threshold, 39 pair HCO+^+ with CO or HCN, for which the edge-on brightening of HCO+^+ at low FFUVF_{\mathrm{FUV}} raises the weak-field end of the ratio and reduces ηFUV\eta_{\mathrm{FUV}} by factors of ∼5\sim5–77. This includes the matched-JJ HCO+^+/HCN pairs identified in § 4.2. The remaining three pair low-JJ HCO+^+ with CN N=3N = 3–22 or 44–33, where the reduction is smaller but their face-on enhancement factors are already modest (ηFUV≈9\eta_{\mathrm{FUV}} \approx9–1212). Conversely, ratios of high-JJ CO over CN or C2_2H gain dynamic range when edge-on, since their numerator and denominator respond in opposite directions. This is not necessarily an advantage, however, since it reflects a strong sensitivity to inclination, making them poor choices where the geometry is unknown. For a target with no constraint on its inclination at all, the corrections in Figure D1 (panel a) imply a systematic uncertainty of ∼1\sim1 dex on the inferred field strength. In practice this is rarely the situation, since inclination is routinely measured from continuum imaging, and can therefore be applied as a known correction rather than left as a degeneracy.

Two caveats apply to this. First, this test uses a single disk structure. Second, the photoevaporative wind in PUFFIN is spherically diverging by construction, and a spherical emitting volume looks the same from every viewing angle, so the convergence of CXC_X towards unity at high FFUVF_{\mathrm{FUV}} is partly imposed by the model. Real flows are asymmetric and, at high FFUVF_{\mathrm{FUV}}, truncated by an ionisation front. Their emission would depend on the star–disk-observer orientation rather than on inclination alone. Our corrections are therefore minimum estimates, and are most reliable at 10210^2–103 G010^3\ G_0, where the emission arises from the disk itself.

Full ratio matrix

Figure E1 shows the diagnostic performance of all 2256 line flux ratios considered in this work, extending the subset shown in Figure 4 to all eight species.

This paper has been typeset from a TeX/LaTeX file prepared by the author.

Full matrix of the 2256 line flux ratios formed from the 48 transitions considered in this work

Figure E1. Full matrix of the 2256 line flux ratios formed from the 48 transitions considered in this work, grouped by species. Each circle represents the ratio of the transition on the yy-axis (numerator) to that on the xx-axis (denominator). Self-ratios along the diagonal are omitted. Colour denotes ρfixed\rho_{\mathrm{fixed}}, the correlation of each ratio with the external FUV field when the stellar and disk properties are held fixed, with blue indicating ratios that increase with FFUVF_{\mathrm{FUV}} and red those that decrease. Circle size indicates the enhancement factor ηFUV\eta_{\mathrm{FUV}}. Reciprocal ratios appear as mirror images across the diagonal, with the same circle size and the opposite sign of ρfixed\rho_{\mathrm{fixed}}. Wedges along each axis indicate increasing EuE_u within each species (see Table A1 for full details of the individual transitions).

References

  1. [36]Coleman G. A. L., Haworth T. J., 2022, MNRAS, 514, 2315
  2. [90]Müller H. S. P., Schlöder F., Stutzki J., Winnewisser G., 2005, Journal of Molecular Structure, 742, 215
  3. [100]Qi C., et al., 2013, Science, 341, 630
  4. [124]van Terwisga S. E., Hacar A., 2023, A&A, 673, L2
  5. [125]van Terwisga S. E., et al., 2020, A&A, 640, A27

Paper details

vlnc:W_7754dd8n
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents