Interstellar Detection of 2-Cyanocyclopentadiene, C_5H_5CN, a Second Five-Membered Ring Toward TMC-1
Kin Long Kelvin Lee, P. Bryan Changala, Ryan A. Loomis, Andrew M. Burkhardt, Ci Xue, Martin A. Cordiner, Steven B. Charnley, Michael C. McCarthy, Brett A. McGuire
DDraft version February 22, 2021
Typeset using L A TEX twocolumn style in AASTeX62
Interstellar Detection of 2-Cyanocyclopentadiene, C H CN, a Second Five-Membered Ring Toward TMC-1
Kin Long Kelvin Lee, P. Bryan Changala, Ryan A. Loomis, Andrew M. Burkhardt, Ci Xue, Martin A. Cordiner,
5, 6
Steven B. Charnley, Michael C. McCarthy, and Brett A. McGuire
1, 3, 2 Department of Chemistry, Massachusetts Institute of Technology, Cambridge, MA 02139, USA Center for Astrophysics | Harvard & Smithsonian, Cambridge, MA 02138, USA National Radio Astronomy Observatory, Charlottesville, VA 22903, USA Department of Chemistry, University of Virginia, Charlottesville, VA 22904, USA Astrochemistry Laboratory and the Goddard Center for Astrobiology, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA Institute for Astrophysics and Computational Sciences, The Catholic University of America, Washington, DC 20064, USA
ABSTRACTUsing radio observations with the Green Bank Telescope, evidence has now been found for a secondfive-membered ring in the dense cloud Taurus Molecular Cloud-1 (TMC-1). Based on additional ob-servations of an ongoing, large-scale, high-sensitivity spectral line survey (GOTHAM) at centimeterwavelengths toward this source, we have used a combination of spectral stacking, Markov chain MonteCarlo (MCMC), and matched filtering techniques to detect 2-cyanocyclopentadiene, a low-lying isomerof 1-cyanocyclopentadiene, which was recently discovered there by the same methods. The new obser-vational data also yields a considerably improved detection significance for the more stable isomer andevidence for several individual transitions between 23–32 GHz. Through our MCMC analysis, we derivecospatial, total column densities of 8 . × and 1 . × cm − for 1- and 2-cyanocyclopentadienerespectively, corresponding to a ratio of ∼ Keywords:
Astrochemistry, ISM: molecules INTRODUCTIONThe recent astronomical detection of 1-cyano-1,3-cyclopentadiene (Fig. 1; hereafter 1-cyano-CPD) andother CN-functionalized ring molecules in the starlesscloud core, Taurus Molecular Cloud-1 (TMC-1), hasopened up an entirely new and unexplored area of aro-matic organic chemistry in space. In rapid succession,evidence has been found for both five-membered (Mc-Carthy et al. 2020b) and six-membered rings (McGuireet al. 2018), and even bicyclic ones (McGuire et al.2020a) in a primordial gas cloud which has long beenknown to be exceedingly rich in highly unsaturated car-bon chains, most notably cyanopolyynes and acetylenicfree radicals, among others. Intriguingly, the derivedabundances of these rings exceed—in some cases bymany orders of magnitude—those predicted from chem-ical models which well reproduce the abundance of a
Corresponding author: Kin Long Kelvin Lee, Brett A. [email protected], [email protected] wide assortment of chains regardless of length. Forthis reason, questions as to the relative importance ofbottom-up formation pathways versus inheritance fromprevious top-down routes that might survive to thedense cloud phase have been raised, but remain poorlyconstrained at present.Substitution of a H atom of cyclopentadiene, c -C H with a nitrile group yields three possible cyanocyclopen-tadienes. Quantum chemical calculations by McCarthyet al. (2020b), shown in Figure 1, predict 1-cyano-CPDis the most stable isomer, followed closely by 2-cyano-1,3-cyclopentadiene (2-cyano-CPD; 5 kJ/mol or 600 K)and then 5-cyano-1,3-cyclopentadiene, which lies farhigher in energy (by 26 kJ/mol or ∼ a r X i v : . [ a s t r o - ph . GA ] F e b 7 J Q F Y N [ J J S J W L ^ P / R T Q H ^ F S T ( 5 ) H ^ F S T ( 5 ) H ^ F S T ( 5 ) Figure 1.
Geometric structures of the three low-lying cyanocyclopentadiene isomers, 1-, 2-, and 5-cyanocyclopentadiene, c -C H CN and G3//B3LYP (Baboulet al. 1999) energetics at 0 K. The dipole moment projec-tions are µ a = 4.15 D and µ b = 0.27 D for 1-cyano-CPD, and µ a = 4.36 D and µ b = 0.77 D for 2-cyano-CPD (Sakaizumiet al. 1987). 5-cyano-CPD has not yet been observed, butpossesses similarly favorable dipole moments ( µ a = 3.91 D, µ b = 1.47 D calculated at the ω B97X-D/cc-pVQZ level oftheory). ing very sharp lines ( ∼ < . − , which ismore than sufficient to conduct a rigorous search in thecoldest, most quiescent molecular clouds. Both isomersare highly polar, with measured dipole moments alongtheir a -inertial axes in excess of 4 D [4.15(15) D and4.36(25) D, respectively; (Sakaizumi et al. 1987)], andcomparable to that of benzonitrile ( µ a = 4 . a bands (26–40 GHz), a number of the authors hererecently reported the astronomical detection of 1-cyano-CPD using spectral stacking and matched filtering tech-niques (McCarthy et al. 2020b). From these observa-tions, which represent ∼
30% of the project goal, anupper limit for 2-cyano-CPD relative to 1-cyano-CPDwas estimated to be roughly 1/3. Because the abun-dance ratio in the laboratory ranges from 1/2 to 1/4(Sakaizumi et al. 1987; McCarthy et al. 2020a), it wasunclear if 1-cyano-CPD is formed selectively in TMC-1or if the apparent absence of 2-cyano-CPD is simply aquestion of sensitivity, given its somewhat lower stabil-ity and consequently lower abundance. With the second data release (DR2) and additional laboratory measure-ments, this ambiguity has now been resolved, and indoing so a common formation for this isomeric pair isimplicated. OBSERVATIONS AND DATA ANALYSISThe observations in DR2 fold in new observationsmade between February 2018 and June 2020 on theRobert C. Byrd 100-m Green Bank Telescope in GreenBank, West Virginia under project codes GBT18A-333,GBT18B-007, and GBT19A-047. Although the fre-quency range of the present dataset is only slightlywider than that covered in DR1 (McGuire et al. 2020b),it is considerably more sensitive at some frequencies(McGuire et al. 2020a). The spectral coverage of DR2extends from 7.906 to 33.527 GHz (25.6 GHz band-width) with continuous coverage between 22–33.5 GHz,at a uniform frequency resolution of 1.4 kHz (0.05–0.01 km/s in velocity) and an RMS noise of ∼ T A ∗ ) across the spectrum.As before, the target was the cyanopolyyne peak ofTMC-1 at (J2000) α = 04 h m s δ = +25 ◦ (cid:48) (cid:48)(cid:48) .The calibrator source for pointing and focus observa-tions was J0530+1331; focus and pointing offsets wereperformed at the beginning of each observing session,and subsequently every 1 to 2 h, depending on theweather; typical pointing convergence was (cid:46) (cid:48)(cid:48) . Obser-vations were performed using position switching (ON-OFF), in which the target and the off position were ob-served in a sequential manner, each for 2 min. The offposition was chosen to be 1 ◦ off target and was confirmedto be clear of emission. Additionally data from projectGBT17A-164 and GBT17A-434 have also been folded inthe DR2. The observing strategy for this archival datais outlined in McGuire et al. (2018), but it is very simi-lar to that used here. To ensure uniformity and consis-tency with the present data set, the archival data werere-calibrated and re-reduced. Uncertainty due to fluxcalibration is expected to be ∼ > σ ) is omitted so as to avoid any interlop-ers: this corresponds to 275 transitions for 1-cyano-CPDand 326 for 2-cyano-CPD, without interlopers detected.A signal-to-noise weighted average of the spectra wasthen performed based on the expected intensity of theline and the RMS noise of the observations. Only tran-sitions of the target species that have a predicted flux ≥
5% of the strongest line are considered in our analysis.This procedure is built into the molsim package (Lee &McGuire 2020), which performs the spectral simulationand wraps the affine-invariant MCMC implementationof emcee (Foreman-Mackey et al. 2013) and posterioranalysis routines from
ArviZ (Kumar et al. 2019).Given that the 1-cyano-CPD isomer was characterizedin our earlier work, we used the posterior distributionsobtained in McCarthy et al. (2020b) as priors for both1-cyano-CPD and 2-cyano-CPD MCMC simulationshere—effectively refining the previous model with thenew experimental and observational data. The modelspace comprises 14 parameters corresponding to fourknown velocity components within TMC-1 (Dobashiet al. 2018, 2019), each having an independent sourcesize [SS], column density ([ N T ], and radial velocity[ v lsr ], while a common excitation temperature [ T ex ] andlinewidth [ dv ] are assumed. The parameters are used tosimulate the expected flux in a forward model, takinginto account beam dilution and optical depth effects,with the MCMC sampling guided by the observed spec-tra. Finally, to determine an overall significance of adetection, the model spectra are stacked using identi-cal weights, and that stacked model is then used as amatched filter that is cross-correlated with the stackedobservations. The resulting response spectrum providesa lower limit on the statistical significance to the detec-tions. LABORATORY MEASUREMENTSThe astronomical detection of 1-cyano-CPD wasbased on laboratory measurements made using a cavity-enhanced Fourier transform microwave spectrometer inwhich this molecule and many others (McCarthy et al.2020a,b) were produced in discharge of benzene andmolecular nitrogen. To improve the accuracy of thespectroscopic constants of 1-cyano-CPD and measure acomparable number of transitions for 2-cyano-CPD, adischarge of dicyclopentadiene (the Diels-Alder dimerof CPD) and acetonitrile was used here. This pre-cursor combination yielded nearly a ten-fold increasein line intensity for both isomers relative to our ear-lier work. For 1-cyano-CPD, the number of hyperfinecomponents in the laboratory data set approximatelydoubled, while a three-fold increase was achieved for2-cyano-CPD. In doing so, the frequency range of themeasurements increased commensurately: from 30 to36 GHz for 1-cyano-CPD, and from 19 GHz to 33 GHz for 2-cyano-CPD. The best-fit spectroscopic constantsfor both species—the most complete summary of therotational data yet—are summarized in Table 1.As part of these measurements, improved SNR forboth isomers enabled transitions between higher J and K a levels to be observed. Inclusion of these weaker tran-sitions proportionally increases the rotational partitionfunctions, and alters somewhat the column densities de-rived for both species. At 300 K, the new values for Q rot are approximately 1.4 times larger than in McCarthyet al. (2020b), resulting in a substantially lower columndensity, and one outside the 2 σ uncertainties reportedin our previous analysis. RESULTS AND DISCUSSIONShown in Fig. 2 are stacked spectra and the impulseresponses for the two cyanocyclopentadiene isomers thathave been derived from the DR2 data. Additional re-sults of the MCMC results can be found in the Ap-pendix. Relative to the DR1 results, these additional ob-servations improve the detection significance of 1-cyano-CPD by a factor of 1.8 (from 5.8 σ to 10.7 σ ), and moreimportantly, provide compelling evidence for 2-cyano-CPD. As the visual representations of the statisticalanalysis illustrate, we can now conclude with good confi-dence that both isomers are present in TMC-1. Further-more, albeit at low signal-to-noise ratio (SNR), evidencefor several individual rotational lines of 1-cyano-CPD asshown in Fig 3; these correspond to the strongest tran-sitions at ∼ . σ un-certainty) has been determined for the two isomers; wenote, however that this value treats the two models asstatistically independent, and a proper estimate wouldrequire explicit treatment of the column density covari-ances between the two molecules.The astronomical detection of 2-cyano-CPD with amarginally similar abundance to 1-cyano-CPD stronglysuggests both isomers are formed via a common path-way. The simplest and most direct route involves the Table 1.
Spectroscopic constants of 1-cyano-CPD and 2-cyano-CPD. The fits were performed with the Watson A-reducedHamiltonian including quartic centrifugal distortion. All values are given in MHz with 1 σ uncertainties in parentheses. Valuesbracketed with [ ] were held fixed.3 Parameter 1-cyano-CPD 2-cyano-CPDThis work McCarthy et al. (2020b) This work McCarthy et al. (2020b) A . . . . B . . . . C . . . . J × . . . . JK × . . . . K × [0 . a [0 . ] [0 . a [0 . ] δ J × . . . . ] δ K × . . . . ] χ aa (N) − . − . − . − . χ bb (N) 2 . . . . N lines b
154 68 110 38(
J, K a ) max (11 ,
3) (9 ,
3) (10 ,
2) (5 , a Calculated at the ω B97X-D/cc-pVQZ level of theory. b The number of hyperfine-resolved transitions included in the fit. 8 3 7 H ^ F S T ( 5 ) ; J Q T H N Y ^ P R X 8 3 7 ; J Q T H N Y ^ P R X H ^ F S T ( 5 ) Figure 2.
Velocity stacked and matched filter spectra of 1-cyano-CPD and 2-cyano-CPD. (
Left ) The stacked spectra from theGOTHAM DR2 data are display in blue, overlaid with the expected line profile in red from our MCMC fit to the data. Thesignal-to-noise ratio is on a per-channel basis. (
Right ) Impulse response functions of the stacked spectra of same moleculesusing the simulated line profiles as matched filters. The intensity scales are the signal-to-noise ratios of the response functionswhen centered at a given velocity. The values denoted in each figure indicate the peak of the impulse response functions whichprovide a minimum significance for the detection. reaction of cyclopentadiene with CN radical. Although apparently not examined experimentally, theoretical cal- T A 0
2, 5
2, 4
1, 6
1, 5 v LSR P R X T A 0
0, 8
0, 7 v LSR P R X
0, 9
0, 8
Figure 3.
Four rotational transitions of 1-cyano-CPD observed towards TMC-1 using the 100-m GBT telescope. The overlap-ping red trace and shaded regions represent the simulated flux based on mean parameters derived from the MCMC posterior.The corresponding asymmetric top ( J K a ,K c ) quantum numbers for each transition are shown above each feature. The dashedline indicates the nominal 5.8 km/s source velocity. culations (McCarthy et al. 2020b) at the G3//B3LYPlevel of theory predict both 1-cyano-CPD and 2-cyano-CPD are formed exothermically and barrierlessly, irre-spective of the small difference in their stabilities. Thisreaction is analogous to benzonitrile formation frombenzene and CN, which has been extensively studiedby multiple experimental (Balucani et al. 2000; Tre-vitt et al. 2010; Lee et al. 2019; Cooke et al. 2020)and theoretical techniques (Woon 2006; Lee et al. 2019),and is known to occur rapidly at low temperature, im-plying the possibility of in situ formation of cyanidederivatives of hydrocarbons under the cold, dark con-ditions of TMC-1. By analogy, the same logic impliesthe presence of cyclopentadiene in TMC-1, although thesmall permanent dipole moment [ µ b = 0 .
420 D; (Laurie1956; Damiani et al. 1976)] of this hydrocarbon ring willmake its direct detection very challenging in the radioband. It is also worth noting the abundance of the twocyanocyclopentadienes taken together close to that ofbenzonitrile [1 . × cm − ; Burkhardt et al. (2020)].Since both the five and six-membered rings likely formvia reaction with a common precursor, the CN radical,and the cyanation reactions are highly exothermic andbarrierless, this ratio may reflect the nascent hydrocar-bon abundances in TMC-1. Although there is no obvi- ous a priori expectation of the benzene/cyclopentadieneabundance ratio, the former is aromatic while the latteris not, and thus might be expected to more abundantbased on thermodynamic stability.In a more general sense, the present work further high-lights that carbon chemistry of considerable richness andcomplexity lies just below the noise floor of previousspectral line surveys toward this well-known primordialmolecular cloud. Despite discoveries of monocyclic andeven bicyclic rings in rapid succession, this work hasraised many more questions than it has helped answer.Nevertheless, it serves to illustrate that there is poten-tially much more to be learned about a well-studied anduntil recently a seemingly well-understood source. Inthis context, it would be disappointing if still other func-tionalized rings and potentially their precursors are noteventually found with sustained effort. DATA ACCESS & CODEData used for the MCMC analysis can be found in theDataVerse entry (GOTHAM 2020). The code used toperform the analysis is part of the molsim open-sourcepackage; an archival version of the code can be accessedat Lee & McGuire (2020).M.C.M, K.L.K.L., and P.B.C. acknowledge financialsupport from NSF grants AST-1908576, AST-1615847,and NASA grant 80NSSC18K0396. A.M.B. acknowl-edges support from the Smithsonian Institution as aSubmillimeter Array (SMA) Fellow. S.B.C. and M.A.C.were supported by the NASA Astrobiology Institutethrough the Goddard Center for Astrobiology. C.X. isa Grote Reber Fellow, and support for this work wasprovided by the NSF through the Grote Reber Fellow- ship Program administered by Associated Universities,Inc./National Radio Astronomy Observatory and theVirginia Space Grant Consortium. The National Ra-dio Astronomy Observatory is a facility of the NationalScience Foundation operated under cooperative agree-ment by Associated Universities, Inc. The Green BankObservatory is a facility of the National Science Foun-dation operated under cooperative agreement by Asso-ciated Universities, Inc.REFERENCES
Baboul, A. G., Curtiss, L. A., Redfern, P. C., &Raghavachari, K. 1999, The Journal of Chemical Physics,110, 7650, doi: 10.1063/1.478676Balucani, N., Asvany, O., Huang, L. C. L., et al. 2000, TheAstrophysical Journal, 545, 892Burkhardt, A. M., Loomis, R. A., Shingledecker, C. N.,et al. 2020, Nature Astronomy, 0, 0Cooke, I. R., Gupta, D., Messinger, J. P., & Sims, I. R.2020, Astrophysical Journal Letters, 891, L41Damiani, D., Ferretti, L., & Gallinella, E. 1976, Chem.Phys. Lett., 37, 265,doi: https://doi.org/10.1016/0009-2614(76)80212-6Dobashi, K., Shimoikura, T., Nakamura, F., et al. 2018,The Astrophysical Journal, 864, 82Dobashi, K., Shimoikura, T., Ochiai, T., et al. 2019, TheAstrophysical Journal, 879, 88,doi: 10.3847/1538-4357/ab25f0Ford, R. G., & Seitzman, H. A. 1978, J. Mol. Spectrosc., 69,326, doi: https://doi.org/10.1016/0022-2852(78)90068-1Foreman-Mackey, D., Hogg, D. W., Lang, D., & Goodman,J. 2013, Publications of the Astronomical Society of thePacific, 125, 306, doi: 10.1086/670067GOTHAM. 2020, Spectral Stacking Data for Phase 2Science Release of GOTHAM, V3, Harvard Dataverse,doi: 10.7910/DVN/K9HRCKKumar, R., Carroll, C., Hartikainen, A., & Martin, O. 2019,Journal of Open Source Software, 4, 1143,doi: 10.21105/joss.01143Laurie, V. W. 1956, The Journal of Chemical Physics, 24,635, doi: 10.1063/1.1742585Lee, K. L. K., & McGuire, B. A. 2020, molsim, Zenodo.https://zenodo.org/record/4122749 Lee, K. L. K., McGuire, B. A., & McCarthy, M. C. 2019,Phys. Chem. Chem. Phys., 21, 2946,doi: 10.1039/C8CP06070CLoomis, R. A., Burkhardt, A. M., Shingledecker, C. N.,et al. 2020, Nature Astronomy, In press.McCarthy, M. C., Lee, K. L. K., Carroll, P. B., et al. 2020a,The Journal of Physical Chemistry A, 124, 5170,doi: 10.1021/acs.jpca.0c02919McCarthy, M. C., Lee, K. L. K., Loomis, R. A., et al.2020b, Nature Astronomy, In press.,doi: 10.1038/s41550-020-01213-yMcGuire, B. A., Burkhardt, A. M., Kalenskii, S., et al.2018, Science, 359, 202, doi: 10.1126/science.aao4890McGuire, B. A., Loomis, R. A., Burkhardt, A. M., et al.2020a, Science, 0, 0McGuire, B. A., Burkhardt, A. M., Loomis, R. A., et al.2020b, The Astrophysical Journal, 900, L10,doi: 10.3847/2041-8213/aba632—. 2020c, The Astrophysical Journal Letters, 900, L10,doi: 10.3847/2041-8213/aba632Sakaizumi, T., Kikuchi, H., Ohashi, O., & Yamaguchi, I.1987, Bull. Chem. Soc. Jpn., 60, 3903,doi: 10.1246/bcsj.60.3903Trevitt, A. J., Goulay, F., Taatjes, C. A., Osborn, D. L., &Leone, S. R. 2010, J. Phys. Chem. A, 114, 1749Wentrup, C., & Crow, W. 1970, Tetrahedron, 26, 4375,doi: https://doi.org/10.1016/S0040-4020(01)93084-4Wohlfart, K., Schnell, M., Grabow, J.-U., & K¨upper, J.2008, Journal of Molecular Spectroscopy, 247, 119 ,doi: https://doi.org/10.1016/j.jms.2007.10.006Woon, D. E. 2006, Chem. Phys., 331, 67
APPENDIX A. MCMC RESULTSThe following summarizes the posteriors obtained for 1-cyano-CPD and 2-cyano-CPD from the MCMC modeling.The data comprises corner plots and statistics that visualizes and quantifies the sampling results. Corner plots areinterpreted in two ways: the off-diagonal contours show the covariance between model parameters, while the diagonaltraces are empirical cumulative distribution function (ECDF) plots. The latter indicates the integrated marginalizedlikelihood of a given parameter—in other words, the parameter space comprised by the posterior.A.1.
Figure A1 shows the posterior traces for the MCMC modeling analysis of 1-cyano-CPD. These simulations indicatedetections in all four components, with the source sizes very poorly constrained and covariant with the column densities.Table A1 provides summary statistics of the marginalized posteriors.
Table A1. v lsr Size N † T T ex ∆ V (km s − ) ( (cid:48)(cid:48) ) (10 cm − ) (K) (km s − )C1 5 . +0 . − . +67 − . +0 . − . . +0 . − . . +0 . − . C2 5 . +0 . − . +30 − . +1 . − . C3 5 . +0 . − . +5 − . +1 . − . C4 6 . +0 . − . +153 − . +0 . − . N T (Total) †† . +0 . − . × cm − Note – The quoted uncertainties represent 95% highest posterior density. † Column density values are highly covariant with the derived source sizes. †† Total column density derivedfrom combining posterior column densities of each component. The uncertainty corresponds to the 95%highest joint posterior density. 8 8 8 8 8 8 8 8 ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X N c o l × N c o l × N c o l × N c o l × T e x ( K ) 8 8 d v P R X 8 8 8 8 8 8 ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X N col ×10 N col ×10 N col ×10 N col ×10 T ex ( K ) dv P R X Figure A1.
Corner plot for 1-cyano-CPD The diagonal traces correspond to ECDF plots, and off-diagonal plots show the kerneldensity covariance between model parameters. In the former, lines represent the 25th, 50th, and 75th percentiles respectively.The length scale for the kernel density plots is chosen with Scott’s rule.
A.2.
Figure A2 shows the posterior traces for the MCMC modeling analysis of 2-cyano-CPD. These simulations suggestdetections in components 8 8 8 8 8 8 8 8 ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X N T × N T × N T × N T × T e x 0 8 8 d v P R X 8 8 8 8 8 8 ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X ; 1 8 7 P R X N T ×10 N T ×10 N T ×10 N T ×10 T ex 0 dv P R X Figure A2.
Corner plot for 2-cyano-CPD The diagonal traces correspond to ECDF plots, and off-diagonal plots show the kerneldensity covariance between model parameters. In the former, lines represent the 25th, 50th, and 75th percentiles respectively.The length scale for the kernel density plots is chosen with Scott’s rule. Table A2. v lsr Size N † T T ex ∆ V (km s − ) ( (cid:48)(cid:48) ) (10 cm − ) (K) (km s − )C1 5 . +0 . − . +71 − . +0 . − . . +0 . − . . +0 . − . C2 5 . +0 . − . +31 − . +0 . − . C3 5 . +0 . − . +140 − . +0 . − . C4 6 . +0 . − . +140 − . +0 . − . N T (Total) †† . +0 . − . × cm − Note – The quoted uncertainties represent the 95% highest posterior density. † Column density values are highly covariant with the derived source sizes. ††††