Main

Understanding the origin of life remains one of the most profound and enduring challenges in science. Over many years, numerous theories have been advanced to address this question (for example, ref. 1). Among these, the hypothesis that life’s building blocks were synthesized in interstellar space and later delivered to the primitive Earth via comets and meteorites has garnered substantial attention2,3. This model is supported by the detection of a rich inventory of prebiotic organic molecules in comets (for example, refs. 4,5), meteorites (for example, ref. 6) and asteroids (for example, refs. 7,8).

The role of sulfur-bearing molecules is pivotal in the general discussion of prebiotic chemistry. Sulfur (S) is indispensable to many biological processes and is considered an essential element for life on Earth (for example, ref. 9). Yet, there exists a notable discrepancy between the S compounds observed in interstellar space and those found in meteoritic material. In the gas phase of the interstellar medium (ISM), only a limited array of S species—predominantly small molecules with up to nine atoms (CH3SCH3 (ref. 10) and C2H5SH (refs. 11,12))—have been detected. Furthermore, whereas the observed sulfur abundance in the diffuse ISM is consistent with the cosmic value, the overall abundance of gas-phase S species in dense molecular clouds is significantly lower (for example, one order of magnitude lower13) than expected compared with the cosmic S/H abundance ratios. This S depletion in the ISM indicates that a substantial fraction of sulfur may be sequestered in forms that are not readily detectable, such as refractory compounds or as constituents of dust grains, leaving the dominant carriers of sulfur in interstellar space unidentified.

By contrast, meteoritic analyses have consistently uncovered a diverse suite of larger S-bearing molecules, most of which contain more than nine atoms, for example, benzothiophene C8H6S (refs. 6,14,15). The absence of these larger, potentially prebiotically relevant S species in the ISM raises critical questions about the chemical evolution that bridges interstellar matter and the chemical composition of cometesimals and planetesimals. It is possible that our current observational capabilities are insufficient to detect these elusive large compounds in the ISM or that substantial chemical transformations occur during the transfer from interstellar clouds to planetary surfaces, such as the early Earth.

To date, more than 340 interstellar molecules have been detected in the ISM and circumstellar envelopes16,17. The record of these discoveries is characterized not merely by a gradual increase in numbers but also by the episodic emergence of entirely new molecular groups. The most recent breakthrough occurred following the detection of benzonitrile by McGuire et al.18 in 2018 within Taurus Molecular Cloud 1. Subsequently, several cyclic hydrocarbon cyanides have been identified, with 1- and 2-cyanonaphthalene (c-C10H7CN) being recognized as the first polycyclic hydrocarbons19. Following this, tetracyclic hydrocarbon cyanides, such as 1-, 2- and 4-cyanopyrene (c-C16H9CN)20,21 were found in 2024. At present, the largest interstellar molecule, apart from fullerenes, is the seven-ring polycyclic aromatic hydrocarbon (PAH) cyanocoronene (c-C24H11CN)22. It is important to note that the cyclic hydrocarbons detected include not only cyanide species but also pure hydrocarbons, for example, ethynylbenzene (c-C6H5CCH)23. Among these, indene (c-C9H8), a bicyclic hydrocarbon, is the largest species identified to date24. In total, 17 cyclic hydrocarbon species, including benzene25, have been observed. In a natural progression of these findings, Yang et al.26 proposed that S-bearing PAHs, particularly those incorporating S heterocycles, might serve as a reservoir for the missing S. This proposal is supported by the detection of compounds such as thiophenol (c-C6H5SH)6, diphenyl disulfide, dibenzothiophene, thianthrene14 and thiophene along with related species15 in meteorites. Nevertheless, although S-bearing species account for approximately 15% of the interstellar molecules detected, S-bearing cyclic hydrocarbons have yet to be identified in interstellar space except for the claimed detection of the small species c-C3H2S (ref. 27), possibly due to their missing spectral experimental characterization.

Thiophenol is one of the most fundamental of the S-bearing cyclic hydrocarbons and has been detected in meteoritic material. However, its dipole moment is relatively small (a- and b-dipole moments: μa = 0.83 D and μb= 0.75 D; Extended Data Table 1), which might be a main limiting factor for its possible identification in the interstellar gas through millimetre-wave astronomical observations. Also, the rotation of the SH group relative to the phenyl ring causes a splitting of its rotational lines and, hence, an overall spread of its emitted intensity (and, therefore, molecular population) across a higher number of rotational levels. By contrast, 2,5- and 2,4-cyclohexadien-1-thione (two isomeric forms of c-C6H6S, hereafter 2,5-CT and 2,4-CT, respectively; a-type dipole moments: 4.73 D and 3.87 D; Extended Data Table 1), which are structural isomers of thiophenol, do not exhibit splitting due to spin–rotation interactions, internal rotation or hyperfine effects. These rigid molecules display a pure rotational spectrum characteristic of an asymmetric top—analogous to the behaviour observed by 2,5- and 2,4-cyclohexadien-1-one (c-C6H6O)28—and, as discussed later, they have substantially larger dipole moments. Consequently, the rotational lines of these two species might be brighter than those of their most stable isomeric form, rendering them favourable for interstellar detection in molecular clouds. Similar cyclic hydrocarbons incorporating a doubly hydrogenated carbon (–CH2–) within their ring structures, have already been detected in interstellar gas (for example, cyclopentadiene29, ethynyl cyclopentadiene29, indene24, 1-cyanocyclopentadiene30 and 2-cyanocyclopentadiene31), stressing the potential interest of these thiophenol isomers for interstellar detection.

We report here the discovery of 2,5-CT, the largest S-bearing molecule detected so far, towards the Galactic Centre molecular cloud G+0.693-0.027 (hereafter, G+0.693). Situated within the Sgr B2 complex, this source is one of the principal reservoirs of complex organic molecules (having six or more atoms in total) in our Galaxy and has been the site of over 20 first interstellar detections (see, for example, refs. 32,33,34,35). We selected G+0.693 for our search for 2,5-CT due to its exceptional richness in S-bearing species10,12,36,37,38 and because the aromatic ring benzonitrile has also been recently detected (Rivilla, V. M., private communication). This chemical diversity is probably attributable to the enhanced sputtering of icy grain mantles induced by large-scale shocks from cloud–cloud collisions, which release a significant fraction of the S budget expected to be locked in interstellar ices39.

Results and discussion

Laboratory identification

The experimental search for 2,5-CT and 2,4-CT was guided by ab initio quantum chemical calculations that derived the main spectroscopic parameters from geometry optimization and harmonic force field analysis (section ‘Theoretical calculations’ and Extended Data Fig. 1). The rotational transitions of 2,5-CT and 2,4-CT in the 8–40-GHz range were observed using a combination of a chirped-pulse Fourier transform microwave spectrometer and a pulsed-discharge supersonic jet (section ‘Laboratory measurements’ and Extended Data Figs. 2 and 3 for details). Molecules were generated by a pulsed discharge in the throat of a 10-Hz supersonic jet using a vapour pressure of thiophenol at room temperature of 25 °C. The experimental settings were optimized by monitoring the production of 2,5- and 2,4-cyclohexadien-1-one generated from the discharge of anisole (c-C6H5OCH3).

In the 8–40-GHz region, 92 and 75 rotational lines of 2,5-CT and 2,4-CT, respectively, were detected, covering quantum numbers up to J = 15 and Ka = 7, as shown in Supplementary Data 1 and 2 (ref. 40). Their rest frequencies, determined with a precision of about 5 kHz, were fitted to an effective Watson-type Hamiltonian in S reduction, which included the rotational and centrifugal distortion constants listed in Table 1. The three rotational constants A0, B0 and C0 and the centrifugal distortion constants DJ, DJK, d1 and d2 were determined by the fit, whereas the centrifugal distortion constant DK was kept fixed to the values obtained by the CAM-B3LYP/cc-pCVTZ calculations41,42. The fit reproduces the experimental frequencies with a root-mean-square (RMS) deviation of 2.8 kHz and 3.3 kHz for 2,5-CT and 2,4-CT, respectively. Based on this analysis, rest frequencies for transitions used in the analysis of the astronomical spectra are predicted with uncertainties sufficiently better than 10 kHz. For the frequency ranges of the spectroscopic survey carried out towards the G+0.693 cloud, this uncertainty corresponds to uncertainties in velocity of less than 0.1 km s−1. This is small compared with the typical linewidths of the molecular emission measured towards this cloud (of ~20 km s−1; ref. 39).

Table 1 Molecular parameters for 2,5-CT and 2,4-CT derived from the least square fits
Full size table

Detection of 2,5-CT in G+0.693

We analysed an unbiased, ultrasensitive broadband spectral survey of G+0.693 carried out with the IRAM 30-m and Yebes 40-m radio telescopes (for details of the observations, see Methods). The observed data were compared with simulated spectra of 2,5-CT generated with the Spectral Line Identification and Modeling (SLIM) tool within the MADCUBA software package43, under the assumption of constant excitation temperature, here referred to as local thermodynamic equilibrium (LTE). We note that the intermediate H2 volume densities (104–105 cm−3) in G+0.693 (refs. 39,44) result in the subthermal excitation of molecular emission, thus yielding excitation temperatures (Tex = 5–20 K, which is lower than the kinetic temperature Tk = 50–150 K; ref. 39). Unlike massive hot cores or low-mass hot corinos—where numerous rotational transitions, including those from vibrationally excited states, are observed—only low-energy rotational transitions in the ground vibrational state are detectable in G+0.693, significantly reducing the levels of line blending and confusion due to the excitation temperatures. Consequently, with the current sensitivity, we anticipate the detection of a few tens of transitions of 2,5-CT at these low excitation temperatures.

After assessing the emission of more than 140 molecules previously identified towards G+0.693, we detected many a-type transitions of 2,5-CT with an integrated signal-to-noise ratio >5 covering from the upper rotational levels Jup = 12 to 19, including several pairs of transitions belonging to two nearly complete progressions of (J + 1)0,J+1 ← J0,J and (J + 1)1,J+1 ← J1,J transitions (Fig. 1a), with the exception of the 121,12–111,11 transition, which lies out of the covered frequency range, and the 171,17–161,16 and 170,17–160,16 transitions, which seem to be heavily blended with unidentified lines. These pairs of lines progressively converge as the frequency increases, eventually coalescing into a doubly degenerate line for Jup = 19. Overall, we found 22 unblended or slightly blended features, the latter being contaminated by less than 25% (Fig. 1; spectroscopic information is listed in Extended Data Table 2, including an analysis of the contamination). These were subsequently used in the LTE fitting and to derive the physical parameters of 2,5-CT (detailed information about the LTE fitting using the MADCUBA-SLIM tool and the definition of an unblended line is provided in the Methods section). We stress that no missing lines were observed within the whole dataset, and the remaining lines were either heavily blended or too weak to be observed (transitions at 2 mm and 3 mm that did not rise above the noise). Our results are in agreement with the observed spectra.

Fig. 1: Lines of 2,5-CT detected towards the Galactic Centre molecular cloud G+0.693.
Fig. 1: Lines of 2,5-CT detected towards the Galactic Centre molecular cloud G+0.693.
Full size image

a, Pairs of Ka = 0 and 1 transitions that progressively converge with increasing frequency, ultimately coalescing into a doubly degenerate line. b, Ka > 1 transitions of 2,5-CT observed in the astronomical data that were also used to derive the LTE physical parameters of the molecule (see text; listed in Extended Data Table 2). The quantum numbers for each transition are shown in the upper part of each panel. The red lines depict the result of the best LTE fit to the 2,5-CT rotational transitions. The blue lines are the emission from all the molecules identified to date in the survey, including 2,5-CT, overlaid with the observed spectra (grey histograms and light grey shaded area). The three-dimensional structure of 2,5-CT is also shown (carbon atoms in grey, S atom in yellow and hydrogen atoms in white).

The best-fitting LTE model for 2,5-CT (shown in red in Fig. 1) yields an excitation temperature Tex = 14.3 ± 3.4 K, a radial velocity vLSR = 71.7 ± 0.9 km s−1, a linewidth with a full-width at half-maximum (FWHM) of 20.0 km s−1 and a molecular column density N = (5.6 ± 0.3) × 1012 cm−2, which translates into a fractional abundance with respect to H2 of (4.1 ± 0.7) × 10−11, using N(H2) = 1.35 × 1023 cm−2 as derived by ref. 45. A complementary population diagram analysis has also been performed46, which gave physical parameters that are in good agreement with the SLIM-AUTOFIT analysis: N = (5.6 ± 1.4) × 1012 cm−2 and Tex = 12.5 ± 1.5 K (Methods section and Extended Data Fig. 4). The partition functions used are listed in Extended Data Table 3.

Astrophysical implications

The detection of 2,5-CT opens a new window into the chemistry of large S-bearing cyclic molecules in the ISM, and it provides the first basis for elucidating their abundance and formation. The most straightforward comparison is between 2,5-CT and its structural isomers, 2,4-CT and thiophenol, which are not clearly detected in the current astronomical data (Methods section). Based on the derived upper limits for both molecules (N(2,4-CT) ≤ 3.2 × 1012 cm−2 and N(thiophenol) ≤ 8 × 1013 cm−2), we expect that 2,5-CT is more than twice as abundant as 2,4-CT (which is close to the factor of ~2 estimated in our laboratory from the relative intensities of rotational lines) whereas N(thiophenol)/N(2,5-CT) < 14. Of the two structural isomers, 2,4-CT and 2,5-CT, it would be expected from the energy level diagram (Extended Data Fig. 1) that the low-energy one would be more abundant, in agreement with the detection, but the relatively low dipole moment of thiophenol compared with those of both 2,5-CT and 2,4-CT (Extended Data Table 1) prevents us from unveiling conclusively whether only 2,5-CT is selectively produced in the ISM or whether 2,4-CT and thiophenol are also present but remain undetected due to sensitivity limitations. Besides the emergence of 2,5-CT as the only structural isomer identified to date for the C6H6S family, our findings now confirm the existence of large (more than ten atoms) sulfur-containing cyclic species in the ISM. Although 2,5-CT itself accounts for only a small fraction of the S budget detected towards G+0.693 so far (~0.05%; Sanz-Novo, private communication), its discovery may be just the tip of the iceberg of a yet unexplored chemistry. This scenario might closely mirror that of benzonitrile, whose initial detection in the ISM by McGuire et al.18 preceded the discovery of numerous cyano-substituted PAHs, and is in line with the rich inventory of S-bearing rings in meteorites, with over 80 species detected (including thiophenol, dibenzothiophene and thianthrene6,14. By analogy with the nearly flat abundance trend of cyano-bearing rings found in Taurus Molecular Cloud 120,22, the cumulative contribution of the C6H6S isomers in G+0.693 could approach ~1.5% of the total S reservoir, hinting that S-bearing cyclic hydrocarbons and S-bearing PAHs might not represent a relevant sink of sulfur in the ISM. In this context, if a small portion of sulfur is locked up in S-bearing PAHs and related S-containing cyclic species, we anticipate that the James Webb Space Telescope is capable of detecting several infrared features26, even though some of the prominent bands (for example, the 10-μm C–S-band) may be obscured by the 9.7-μm silicate absorption band. Moreover, upcoming radioastronomical facilities, such as the Square Kilometre Array or the Atacama Large-Aperture Submillimeter/millimeter Telescope, will probably find a rich reservoir of large cyclic S-bearing species, including potentially prebiotic molecules.

The detection of 2,5-CT can be rationalized in terms of its large dipole moment (a-type dipole moment μa = 4.73 D; Extended Data Table 1). This result establishes this organosulfur species as a promising observational link between the rich S inventory found in the minor bodies of the Solar System (asteroids, comets and meteorites), which includes a wide array of cyclic S-bearing compounds, ranging from thiophenol6 and thiophene15 to the more complex diphenyl disulfide, dibenzothiophene and thianthrene14, and the known S budget in the ISM, which has been limited so far to the detection of molecules with up to nine atoms10,11,12. With 13 atoms, 2,5-CT now ranks as the largest S-bearing molecule detected so far in the ISM, marking an important step forwards in molecular size and complexity within interstellar sulfur chemistry. Previously, the largest S-bearing interstellar species had up to nine atoms (for example, ethyl mercaptan, CH3CH2SH, and its isomer dimethyl sulfide, CH3SCH3 (refs. 10,12)). In this context, 2,5-CT is the largest S-bearing complex organic molecule detected so far and also the most complex S-bearing cyclic species. Thus, it provides a novel view on cyclic interstellar chemistry, which now extends beyond pure hydrocarbons and their cyano (–CN) and ethynyl (–CCH) derivatives. Additionally, our findings highlight the need for caution when analysing mass spectrometric measurements of cometary, meteoritic and asteroid material targeting thiophenol, as the mass peak could be contaminated by 2,5-CT, given that both molecules share the same molecular mass (110.02 u). Consequently, although 2,5-CT has not, to our knowledge, been searched for in extraterrestrial material from these minor bodies, it may still be present but unidentified.

To date, the potential formation routes for 2,5-CT and related S rings remain largely unexplored, both experimentally and theoretically. Consequently, we can only hypothesize its possible formation routes by either studying chemically related species or by drawing analogies with bottom-up pathways proposed for related cyclic species. Given the uncertain efficiency of gas-phase pathways, such as those invoked for benzene formation through ion–molecule reactions47, which are considered to be the bottleneck in the growth of larger PAHs, a potential dust-grain origin appears particularly promising in G+0.693. Laboratory simulations have shown that cosmic-ray irradiation of low-temperature acetylene (C2H2) ices efficiently produces benzene48, indicating that an analogous chemistry involving small S-bearing carbon chains (for example, C2S and C3S, and also the detection in the ISM of up to the five-carbon member, C5S; ref. 49) and C2H2 on icy grains could lead to 2,5-CT. Although this hypothesis still needs to be tested in the laboratory, it is supported by two key factors that shape the chemistry of G+0.693, where linear S chains are also abundant (for example, N(CCS) = 1.5 × 1014 cm−2): (1) Its elevated cosmic-ray ionization rate (10−14–10−15 s−1), estimated through chemical modelling involving cations such as PO+ and HOCS+ (ref. 37 and references therein), which favours radical formation and recombination on grain mantles34. (2) The occurrence of large-scale low-velocity shocks associated with a cloud–cloud collision scenario50, which enhance the sputtering of icy grain mantles and could facilitate the desorption of molecules such as 2,5-CT. An analogous connection has already been suggested between benzonitrile, which is also detected in G+0.693 (Rivilla, V. M., private communication), and the cyanopolyyne family51, but for 2,5-CT, the inclusion of ring defects (a (–CH2–) moiety that disrupts the electron delocalization and, thus, the aromaticity within the ring) needs to be addressed. Alternatively, benzene could be directly released from the grains through shocks and subsequently react through radical–neutral reactions52,53, which are considered to be some of the main formation routes for diverse PAHs, such as c-C6H5CN (ref. 18), c-C10H7CN (ref. 19), c-C5H5CN (ref. 30) and c-C16H9CN (refs. 20,21). However, apart from a recent study on the production of c-C3H2S via c-C3H2 + SH (ref. 27), there are no theoretical or experimental data that support an analogous formation route starting from c-C6H6 and yielding 2,5-CT or any of its isomers (for example, thiophenol).

In summary, the study of interstellar chemistry of large cyclic species (>12 atoms) has been bound so far to pure cyclic hydrocarbons or cyano (–CN) and ethynyl (–CCH) derivatives, as their derivatization provides a sizeable dipole moment to the parental, typically nonpolar hydrocarbon (for example, benzene, naphthalene or pyrene), thus enabling their radioastronomical identification. The interstellar detection of 2,5-CT presented here, which is based on new high-resolution rotational data, demonstrates that interstellar cyclic chemistry extends beyond the aforementioned families to encompass S-bearing compounds. These findings open the window to a yet uncharted S chemistry, which, although it might not account for the missing sulfur in dense interstellar environments, does contribute considerably to our understanding of the origin of sulfur-containing molecules in meteorites and comets and provides insight into sulfur reservoirs in young solar-type systems.

Methods

Theoretical calculations

All the quantum chemical calculations were carried out with ORCA54 using CAM-B3LYP41. The correlation-consistent polarized core-valence quintuple-zeta basis set (cc-pCV5Z)42 was used for molecular geometry optimizations and energy calculations of thiophenol, 2,5-CT and 2,4-CT. Effective rotational constants and centrifugal distortion constants for 2,5-CT and 2,4-CT were evaluated theoretically at cc-pCVTZ through harmonic force field calculations. The calculated energies and dipole moments are listed in Extended Data Table 1; the rotational and centrifugal distortion constants are shown in Table 1 along with their experimental values. The accuracy of the dipole moments is expected to be around 8.5% based on comparisons of the values calculated by this method with the observed dipole moments for H2S, dimethylsulfide, thiophene and 3-methylthiophene.

Laboratory measurements

High-resolution rotational spectra of 2,5-CT and 2,4-CT were observed using a high-resolution broadband microwave spectrometer in combination with a pulsed-discharge supersonic jet, as shown in Extended Data Figs. 2 and 3.

Pulsed-discharge supersonic jet

Molecules of 2,5-CT and 2,4-CT were generated via pulsed discharge in a 10-Hz supersonic jet (CASJet55) using thiophenol (Thermo Fisher Scientific, without further purification) maintained at its vapour pressure at room temperature (25 °C) and diluted in neon as a buffer gas with a flow of 50–55 sccm. The jet was produced with a pulse valve (Parker) controlled by a pulse driver (IOTA ONE, Parker). By employing a backing pressure of ~1 kTorr, the molecular beam was cooled to a rotational temperature of approximately 5 K, as estimated from the relative line intensities of thiophenol. Under these conditions, the pressure in the vacuum chamber was maintained at approximately 0.5 × 10−4 Torr to 1.0 × 10−4 Torr using a combination of a diffusion pump, a root blower and a rotary pump.

The discharge nozzle, mounted directly after the pulse valve, comprised electrodes made of small copper discs with a thickness of 3 mm, each featuring a central aperture through which the molecular beam flowed—4.3 mm for the downstream electrode and 2.3 mm for the upstream electrode. The downstream electrode was grounded and served as the anode, whereas the upstream electrode was negatively charged and functioned as the cathode. These electrodes were separated by a Teflon insulator cylinder (13.4 mm in length with a 3.5 mm central aperture) and connected in series with a 50-kΩ ballast resistor. A voltage of 1,000 V was applied across the assembly, resulting in a current of a few milliamperes. Under these conditions, a plasma discharge was generated between the electrodes immediately before the supersonic expansion in the vacuum chamber.

The chirped-pulse Fourier transform microwave spectrometer

The chirped-pulse Fourier transform spectrometer is a high-resolution broadband instrument covering frequencies in the range 8–40 GHz. An intense pulse of length 2 μs, chirped in frequency, produced a macroscopic polarization of the molecular sample. The subsequent free induction decay was recorded in the time domain using a heterodyne receiver. A two-channel arbitrary waveform generator (Keysight, M8190A) generated the chirped pulse, which was frequency upconverted using an IQ modulator and a tunable signal generator (Agilent Technologies, E8257D) to cover the frequencies of interest (8–18 GHz, 18–26 GHz and 26–40 GHz). A solid state amplifier (8–18 GHz: Microsemi C0618-43-T680; 18–26 GHz: Microsemi C1826-36-T964 and 26–40 GHz: Eravant SBP-2734033530-KFKF-S1-HR) amplified the signal before it was emitted via a quad ridge horn antenna, the other port of which was used to detect the molecular signal. A rooftop mirror in the chamber was used to rotate the polarization by 90°. The receiver consisted of a low-noise amplifier protected from the intense excitation pulse by a fast pin diode switch, followed by an identical IQ modulator using the same local oscillator signal to downconvert the free induction decay. The intermediate frequency (IF) signal was then fed into a 5-GHz low pass filter, which was followed by another amplifier, before being digitized using an Acqiris U9510A digitizer card. The measurements were repeated and averaged in the time domain to improve the signal-to-noise ratio. During the measurements, the phase of the chirped pulse was cycled (0°, 90°, 180°, 270°), which allowed sideband separation and the suppression of spurious harmonics.

Astronomical observations

We analysed a new unbiased, ultradeep molecular line survey conducted towards the Galactic Centre molecular cloud G+0.693 using the Yebes 40-m (Guadalajara, Spain) and the IRAM 30-m (Granada, Spain) radio telescopes. This broadband survey spanned a frequency range of ~91 GHz and had higher sensitivity compared with the data used in previous works (for example, ref. 12). The observations were carried out in position switching mode towards the equatorial coordinates of G+0.693 (right ascension α = 17 h 47 min 22 s and declination δ = −28° 21′ 27″) using an off position shifted by Δα = −885″ and Δδ = 290″.

Yebes 40-m radio telescope

New Yebes 40-m observing runs (Project 21A014; PI V. M. Rivilla) were performed between March 2021 and March 2022. We used the ultra-broadband Nanocosmos Q-band (7 mm) high electron mobility transistor (HEMT) receiver, which enables broadband observations across the whole Q-band (18.5 GHz between 31.07 GHz and 50.42 GHz) in two linear polarizations56. The 16 fast Fourier transform spectrometers provided a raw channel width of 38 kHz. We used two distinct spectral set-ups, centred at 41.4 GHz and 42.3 GHz, to identify possible spurious lines. The detailed procedure employed for the data reduction, combination and averaging of both the new Yebes 40-m and the IRAM 30-m data is presented in ref. 34. Subsequently, the spectra were imported into MADCUBA43 and smoothed to a frequency resolution of 256 kHz (velocity resolutions of 1.5–2.5 km s−1 in the range observed). An extraordinary sensitivity has been reached, with r.m.s. noise levels ranging between 0.25 mK and 0.9 mK across the whole Q-band at this spectral resolution in units of the antenna temperature scale (\({T}_{{\rm{A}}}^{* }\)), as the molecular emission towards G+0.693 is extended over the beam57. The half-power beam-width of the telescope ranged between ~35″ and ~55″ (at 50 GHz and 31 GHz, respectively).

IRAM 30-m radio telescope

The new IRAM 30-m observations (Project 123-22; PI Jiménez-Serra) were carried out between 1 and 18 February 2023. We employed the multiband millimetre-wave eight mixer receiver and various frequency set-ups to cover three frequency windows, 83.2–115.41 GHz, 132.28–140.39 GHz and 142–173.81 GHz. Each frequency set-up was shifted in frequency to identify possible contamination of spurious lines coming from the image band. We achieved an initial spectral resolution of 195 kHz using a fast Fourier transform spectrometer (FTS200), even though we finally smoothed the spectra within MADCUBA to 615 kHz, which translates to velocity resolutions of 1.0–2.2 km s−1 in the observed frequency range. The half-power beam-width of the telescope varied between 14″ and 29″ across the frequency range covered. We note that for those frequency ranges that are not covered within these new data, we used the previous IRAM 30-m survey (further details are given elsewhere; for example, ref. 12). Overall, we obtained final noise levels between 0.5 mK and 2.5 mK at 3 mm and between 1.0 mK and 1.6 mK at 2 mm per channel.

LTE analysis of 2,5-CT with MADCUBA

Once the rotational spectroscopic data of 2,5-CT had been imported into the MADCUBA software package43, we used the SLIM tool (version dated 15 June 2024) within MADCUBA to analyse the astronomical data under the assumption of LTE. We generated the LTE simulated spectra with SLIM and then conducted a nonlinear least-squares LTE fit of the brightest transitions of 2,5-CT that were either unblended or exhibited a slight blending (shown in Fig. 1 and listed in Extended Data Table 2) to the observed spectra using the AUTOFIT tool within SLIM43. The selection of these lines followed the criteria established in previous works10,36 to identify unblended lines while considering potential contamination from a yet unidentified line. To evaluate the level of line blending, we analysed the region surrounding the 2,5-CT lines within a velocity range of ±FWHM/2, where FWHM represents the observed linewidth. The level of contamination was assessed by subtracting the LTE fit for 2,5-CT from the observed spectrum and calculating the residual area. The contributions to the residuals for all the selected lines are shown in Extended Data Table 2. A line was classified as unblended if the residual area contributed 25% or less to the total. Additionally, if a known molecule that lies within this range contributed less than 25% of the total integrated intensity, the line was considered to be slightly blended. The SLIM-AUTOFIT method enables us to derive the following physical parameters: molecular column density (N), excitation temperature (Tex), radial velocity (vLSR) and FWHM. Only the latter parameter was fixed in the fit to a value of 20 km s−1 to achieve convergence, which is in agreement with the characteristic FWHM measured for the molecular transitions in G+0.693 (FWHM ≈ 15–20 km s−1; see for example, ref. 39). We, thus, derived the following parameters: N = (5.6 ± 0.3) × 1012 cm−2, Tex = 14.3 ± 3.4 K and vLSR = 71.7 ± 0.9 km s−1.

Rotational diagram analysis

As an alternative to the SLIM-AUTOFIT analysis, we can profit from applying the rotational diagram method46 to derive the physical parameters of 2,5-CT. We used the reduced subset of clean and slightly blended transitions listed in Extended Data Table 2, excluding the transitions for which the contamination by an unidentified line accounted for >25% of the overall area, We obtained results that are in good agreement with both the column density and excitation temperature obtained using SLIM-AUTOFIT: N = (5.6 ± 1.4) × 1012 cm−2 and Tex = 12.5 ± 1.5 K. The results of the rotational diagram method are shown in Extended Data Fig. 4.

Non-detection of 2,4-CT and thiophenol

We implemented in MADCUBA-SLIM the newly measured rotational data for 2,4-CT and searched for it in the survey towards G+0.693. 2,4-CT was not clearly detected, despite the emergence of several weak and unblended transitions within the noise. However, unlike 2,5-CT, it lacks clear enough spectroscopic features in the astronomical data for a conclusive detection. Therefore, we used the LTE parameters obtained for 2,5-CT to derive the upper limit for its molecular abundance. We searched for the brightest predicted spectral features of 2,4-CT that seem to be completely unblended with emission from other molecules previously identified in the astronomical data. Specifically, we used the 130,13–120,12 transition (at ~33.677 GHz), which is one of the frequency regions of the survey with the highest sensitivity (r.m.s. ≈ 0.5 mK), which enabled us to place stringent constraints on the abundance of 2,4-CT towards G+0.693. We, thus, derived a 3σ upper limit for its column density (σ is the r.m.s. noise of the spectra) of N ≤ 3.2 × 1012 cm−2, which yields an upper limit for the molecular abundance with respect to molecular hydrogen of 2.7 × 10−11 and does not produce any overly bright features at other frequencies. Based on the derived upper limit, 2,4-CT is a factor of ~2 less abundant than 2,5-CT.

We also searched for thiophenol using available laboratory rotational data58, without yielding a detection. Therefore, we derived the 3σ upper limit for its molecular abundance using the 151,14–141,13 transition (at ~40.906 GHz), which is the brightest transition and a fully unblended transition predicted by the LTE model. We obtained N ≤ 8 × 1013 cm−2 by adopting the physical parameters found for 2,5-CT. In terms of the molecular abundance with respect to H2, the above value translates into an upper limit of 6 × 10−10. Thus, we found that thiophenol is ≤14 times more abundant than 2,5-CT, in line with the energy ordering of the three structural isomers (Extended Data Fig. 1).