© The Authors 2024

1 Introduction

To understand the origins of our own Solar System and the distribution of the stars around us, it is important to have a broad understanding of the formation of stellar objects and their subsequent feedback on the molecular clouds where they were born. While models explaining the formation of individual low-mass stars are well established, the processes involved in high-mass star formation are more complicated due to their formation in clusters (Zinnecker & Yorke 2007; Krumholz et al. 2014; Motte et al. 2018). Usually, star-forming sites are located in close proximity to one another inside large-scale molecular clouds, and these sites are affected by the radiation of already existing massive O- and B-type stars.

Studying the effects of stellar feedback on star formation is crucial (e.g. Schneider et al. 2020). On one hand, the transferred momentum from stars can lead to a compression of the surrounding molecular gas, inducing star formation in dense regions. On the other hand, stellar radiation can contribute to the disruption of molecular clumps, preventing star formation.

The Lagoon Nebula (Messier 8, M8) is an interesting target for investigating the effects of stellar feedback, as it is an H II region associated with a star-forming (Kumar & Anandarao 2010) molecular cloud complex. It is located at a distance of 1325 pc (Damianí et al. 2019) in the Sagittarius-Carina arm. Its angular extent of about 40′ can be translated to a physical size of 15 pc. The position of M8 corresponds to galactic coordinates of approximately l = 6°, b = −1.3°, which locates the region slightly below the inner Galactic plane, not far from the direction to the Galactic centre. Based on Gaia proper motion data of associated cluster members, Damianí et al. (2019) conclude that the cloud complex crossed the Galactic plane around 4 Myr ago, which may have been the initial trigger for its recent star formation activity.

At optical wavelengths, M8 appears as a reddish emission nebula, with two particularly bright regions in the centre (see Fig. 1). One of them is the prominent H II region of the Lagoon Nebula, which is powered by the spectroscopic binary O star 9 Sagittarii (9 Sgr; Rauw et al. 2012) and the multiple stellar system Herschel 36 (Her 36, Arias et al. 2010) in the west. The other is the eastern illuminated region, associated with the open cluster NGC 6530 (Prisinzano et al. 2005), which is part of M8 and contains several massive stars. An overview of high-mass stars (of O and early B types) identified in this region is available in Wright et al. (2019, see their Table 1), and their positions are marked in Fig. 1 with coloured star symbols.

Special attention has been paid to the interstellar medium (ISM) in the M8 region since White et al. (1997) reported the Lagoon Nebula as the second brightest known CO-emitting source in our Galaxy. This study was followed up by Tothill et al. (2002), who identified 37 individual dense clumps in the inner region of the Lagoon Nebula based on dust continuum maps at 450 µm and 850 µm observed with the James Clerk Maxwell Telescope¹ (JCMT), towards which they observed the CO J = 2−1 line. As shown in Fig. 2, most of these clumps were loosely assigned to the west central region (WC), the east central ridge (EC), and two filaments in the southern part of the cloud (SE and SC).

Apart from the study conducted by Tothill et al. (2002), the molecular content of M8 and its environment has received little attention. However, this changed recently with Tiwari et al. (2018) studying the massive star-forming WC (M8-Main); they report that the WC clumps are located behind the optically visible H II region. Their spectroscopic observations revealed a photo-dissociation region (PDR) with a face-on geometry towards these sources, which is powered by Sgr 9 and Her 36 (see e.g. Fig. 15 of Tielens & Hollenbach 1985).

A further study by Tiwari et al. (2020) focused on the eastern part of the Lagoon Nebula (M8 East), which hosts clumps E, SE7, and SE8. This region is mainly illuminated by an embedded young stellar object (YSO), M8E-IR, which is likely to become a BO star (Linz et al. 2008). Tiwari et al. (2020) also observed an ionisation front moving in the south-east direction. These recent observations suggest that the remaining M8 clumps are also ideal targets for studying the effects of stellar feedback on the remnant gas in the region. Broad-bandwidth spectroscopic observations towards most molecular clumps in M8 are still missing. Through this work, we explore the complete sample of molecular clumps in M8 by combining the results of new spectroscopic observations at millimetre wavelengths with the information retrieved from archival infrared (IR) dust continuum images.

Section 2 gives an overview of the observations and the data reduction strategy. In Sect. 3, the dust continuum images are analysed and the physical properties of the clumps are derived by fitting spectral energy distributions (SEDs) to the flux densities of all clumps. Section 4 details the complete line survey of the 1.3 mm and 3 mm atmospheric windows towards the clumps in M8. The observed line emission is analysed in Sect. 5. The results derived from the analysis of the dust continuum and the line survey are discussed in Sect. 6. Finally, Sect. 7 summarises the results of this study.

Fig. 1 Image of the Lagoon Nebula observed using B, V, and R broadband filters. Blue and green stars respectively mark the positions of O- and early B-type stars associated with the M8 region, according to Table 1 of Wright et al. (2019). The dashed black box marks the molecular gas examined by Tothill et al. (2002). Data credits: Felipe Mac Auliffe (https://clusteroneobservatory.com).

2 Observations and data reduction

We used archival IR to submillimetre wavelength data of the dust continuum emission from M8 in addition to new spectroscopic data taken with the Atacama Pathfinder Experiment (APEX) 12 m submillimetre telescope and the 30 m telescope of the Institut de Radioastronomie Millimétrique (IRAM). The spec-troscopic observations were conducted in on-off mode on the 37 molecular clumps identified by Tothill et al. (2002, see our Appendix A). After inspecting the APEX spectra of all clumps, we decided to change the coordinates of WC3, SE8, and SC5 for the observations with the IRAM 30m telescope, in order to properly match the peak dust emission of the respective clumps. Therefore, the APEX observations of these three clumps were taken at the coordinates suggested by Tothill et al. (2002), while the observations taken with the IRAM 30m telescope are offset by up to 1′.

We chose a fixed reference position M8REF at the coordinates RA=18^h04^m40.0^s, Dec=−24°23′00.0″ (J2000), which is only slightly contaminated with line emission of ¹²CO and ¹³CO. In order to characterise the emission in M8REF, it was observed with a completely clean reference position at an offset of (+2000″, −2000″) from M8REF.

Inadvertently, first observations with APEX used the position switching mode with relative reference positions. These data were used to increase the sensitivity when the profiles of the thus observed lines agree with the data using the fixed reference position.

Fig. 2 SCUBA 850 µm image of the Lagoon Nebula adapted from Tothill et al. (2002). Markers show the position of the molecular clumps in the Lagoon Nebula. The blue and green rectangles respectively indicate the regions examined by Tiwari et al. (2018) and Tiwari et al. (2020).

2.1 APEX observations with nFLASH230

The Atacama Pathfinder Experiment is a 12 m diameter sub-millimetre telescope located on the Llano de Chajnantor in the Chilean High Andes at an altitude of 5107 m (Güsten et al. 2006). The data were taken under project M-0107.F-9530C-2021 (P.I. Karl M. Menten) during several runs between 2021 July and October with the new FaciLity APEX Submillimetre Heterodyne instrument (nFLASH²). The nFLASH receiver is a dual sideband (2SB) dual polarisation heterodyne receiver with two tunable frequency modules, of which we used the lower frequency nFLASH230 band for our observations. The centres of the upper and lower sidebands are separated by 16 GHz and each cover a 7.9 GHz bandwidth in two polarisations. Four observed setups cover a total bandwidth of 58.3 GHz in a frequency range between 210 GHz and 280 GHz (see Appendix A).

The receiver was connected to modules of the APEX fast Fourier transform spectrometer (FFTS), which is an evolved version of the instrument described by Klein et al. (2012) and records each sideband and polarisation with partially overlapping 4 GHz wide FFTS processor units. These units provide each a total of 2¹⁶ channels per 4 GHz bandwidth, resulting in a channel spacing of 61.04 kHz. At our lowest and highest frequencies of 213.1GHz and 279.7 GHz, this results in velocity resolutions of 0.09 km s⁻¹ and 0.07 km s⁻¹, respectively. For analysing the data, each two adjacent channels are averaged. While this reduces the velocity resolution to values between 0.18 km s⁻¹ and 0.14 km s⁻¹, the resulting resolution is sufficient to resolve well all observed spectral lines. At this velocity resolution, the spectra show a typical average root mean square (RMS) noise of 28 mK.

The system temperature during the observations typically ranged from 70 K to 250 K, with a few scans at system temperatures of up to 460 K. The conversion between antenna temperature $T_{\text{A}}^{*}$ and the main-beam brightness temperature T_MB is given by $T_{\text{MB}}=T_{\text{A}}^{*}\times {\eta }_{\text{FW}}/{\eta }_{\text{MB}}$, where η_MB is the main-beam efficiency and η_FW the forward coupling efficiency. Based on Jupiter continuum pointings during the observation period of this project³, we estimate an average conversion factor of η_FW/η_MB = 0.95/0.8. The heterodyne line intensity monitoring⁴ between 2021 July and October suggests a systematic calibration uncertainty of 5% for the nFLASH230 observations. A more conservative estimate of 10% is applied for the further analysis of the APEX data.

The full width at half maximum (FWHM) of the APEX beam, θ_B, at frequency ν (in GHz), in arcseconds is θ_B[″] = 7″.8 × (800/v[GHz]) (Güsten et al. 2006). It thus varies at the observed frequencies between 22.3″ and 29.3″ (respectively corresponding to 0.14 pc and 0.18 pc at the distance of M8).

2.2 IRAM 30 m telescope observations with EMIR 090

The IRAM 30 m telescope is located in the Spanish Sierra Nevada on Pico Veleta at an altitude of 2850 m⁵. The data were taken under project ID 141-21 (P.I. Friedrich Wyrowski) during several runs between 2022 March and June using the 3 mm band (‘Band 1’) of the heterodyne Eight Mixer Receiver (EMIR 090, Carter et al. 2012). Similar to nFLASH, EMIR is a 2SB two polarisation heterodyne receiver with a central sideband separation of 16 GHz and individual sideband bandwidths of 8 GHz. Using three setups, the observations cover a total bandwidth of 40.3 GHz between 70 GHz and 117 GHz. Additional on-off data of clump E taken by Tiwari et al. (2020) were used to increase the sensitivity and frequency coverage for this particular position. An overview of the frequency setups is shown in Appendix A.

EMIR was used in combination with the FFTS backend FTS200 that provides a total of 20737 frequency channels per 4 GHz wide sideband, resulting in a channel spacing of 192.89 kHz. This corresponds to 0.82 km s⁻¹ and 0.50 km s⁻¹ at our lowest and highest observed frequencies of 70.3 GHz and 116.5 GHz, respectively. While this resolution is sufficient for the broader bright lines, weak and narrow spectral lines are only covered by a few channels. At this resolution, the data have an average RMS noise level of 0.17 K.

The EMIR system temperatures during the observations varied for frequencies below 80 GHz between 130 K and 290 K, from 80 GHz to 105 GHz between 80 K and 200 K, and above 105 GHz between 150 K and 480 K. The conversion from antenna temperature $T_{\text{A}}^{*}$ to main-beam brightness temperature T_MB is given by $T_{\text{MB}}=T_{\text{A}}^{*}\times {\eta }_{\text{FW}}/{\eta }_{\text{MB}}$, with the main-beam efficiency η_FW and the forward coupling efficiency η_FW. Based on the average observed frequency of 93.4 GHz, we assumed the default conversion factor⁶ described by η_FW = 0.946 and η_MB = 0.797 for the calibration.

The FWHM, θ_B, of the 30m telescope beam is⁶ θ_B[″] = 2460/v[GHz], with ν being the observed frequency in GHz. Therefore, the beam widths vary between 35.0″ and 21.1″ (corresponding to 0.22 pc and 0.13 pc at the distance of M8) for frequencies between 70.3 GHz and 116.5 GHz, respectively.

2.3 Data reduction of spectroscopic data

The spectra taken with the APEX and the IRAM 30m telescope were reduced using the CLASS program of the GILDAS⁷ software package developed by IRAM. The spectra taken for each clump were combined and a first-order baseline was subtracted. This baseline was determined by averaged spectral channels located off, but in the vicinity, of the individual spectral lines.

In order to correct the CO and ¹³CO signal affected by a contaminated reference position, the spectrum observed at M8REF was re-added to the corresponding transitions. APEX observations obtained in frequency switching mode were compared to the on-off observations and combined if the residual between both spectra did not show emission with a significance above three times the baseline RMS.

The observations taken with the IRAM 30m telescope in 2022 June were affected by a technical defect that caused a frequency and sideband-dependent shift of the observed frequencies by approximately 4 MHz. This shift was corrected for all setups based on a comparison of the spectra at clump E taken before and in June. As this clump was observed at the start of each observing day, and also previously by Tiwari et al. (2020), it was possible to obtain a correction of the frequency scale for each band based on Gaussian fits to the strongest optically thin transitions.

2.4 Archival continuum data

To derive the physical properties of the M8 clumps, we fitted their SEDs as in Urquhart et al. (2018) to compare the M8 clumps to the sources identified through the APEX Telescope Large Area Survey of the Galaxy (ATLASGAL). For this, we used archival data from Spitzer, Herschel, APEX, the Midcourse Space Experiment (MSX; Price et al. 2001), and the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010). While WISE and MSX provide all-sky surveys, dedicated Galactic plane surveys have been performed with Spitzer, Herschel, and APEX. We used data from the GLIMPSE (Churchwell et al. 2009), MIPSGAL (Carey et al. 2009), Hi-GAL (Molinari et al. 2010), and ATLASGAL (Schuller et al. 2009) surveys.

Since M8 is located at a galactic latitude of b = −1.3°, the Hi-GAL maps do not fully cover the nebula. Due to this, these surveys were supplemented with data from the AKARI (Doi et al. 2015) all-sky survey. In addition, we also used the 850 µm data of the Lagoon Nebula taken with the Submillimetre Common-User Bolometer Array (SCUBA; Holland et al. 1999) of the JCMT by Tothill et al. (2002).

The flux densities of each clump were extracted analogously to Urquhart et al. (2018) using several tools of the astropy (Astropy Collaboration 2022) and Photutils (Bradley et al. 2022) packages for Python. For this, the flux density of each clump was extracted in an aperture of two or three times the clump size derived by Tothill et al. (2002), depending on the proximity of neighbouring clumps. This flux density was corrected for background flux based on the median flux density in an annulus around the respective clump. The RMS of the Gaussian noise was calculated for each band based on emission inside a defined mask of all annuli around the clumps, which excludes the clump emission.

The AKARI FWHM beam sizes of 63.4″ at 65 µm (N60), 77.8″ at 90 µm (WIDE-S), and 88.3″ for the 140 µm and 160 µm bands (WIDE-L and N160) (Takita et al. 2015) are not sufficient to resolve the M8 clumps that have FWHM sizes smaller than 40″. Due to this, the AKARI flux density was extracted at the exact position of the M8 clumps (see Table A.1). This allowed the determination of the flux density within the corresponding AKARI beam, which covers most of the respective emission for the M8 clumps. In cases where multiple clumps are contained inside the extracted AKARI beam, the individual contribution of each clump was estimated based on the area fraction of each of the respective clumps inside the beam and the corresponding 350 µm Hi-GAL flux densities of the clumps. Due to the low resolution, we estimated the AKARI flux densities to have a measurement uncertainty of 50% in addition to the uncertainty introduced by the RMS noise. In order to verify that the use of AKARI instead of Hi-GAL for wavelengths between 65 µm and 160 µm leads to results that are comparable to the ATLASGAL sample of clumps (Urquhart et al. 2018), we tested the modified method on a sample of clumps in the NGC 6334 cloud complex, which is similar in distance as M8. This comparison is presented in Appendix B, where we find almost identical luminosities and only minor deviations in the derived masses, which can likely be attributed to the different methods in source size computation, instead of the usage of PACS rather than AKARI data.

3 Dust continuum emission at M8

Figure 3 shows the MIPSGAL 24 µm image overlaid with the contours of the ATLASGAL 870 µm emission. In contrast to the optical image (see Fig. 1), the dust continuum emission at 24 µm not only shows an emission peak at M8-Main (at clump HG, extending to WC1-6) but also a peak of similar brightness at clump E in the massive star-forming region M8 East. The position of HG is, within a few arcseconds, coincident with that of the O7.5V star Her 36, one of main the ionisation sources. Additional fainter, point-like sources can be seen in the vicinity of the clumps WC7, SE2, SE3, SE7, SE8, and SC1. These IR-bright clumps may contain intermediate- to high-mass YSOs, of which the IR radiation penetrates the surrounding colder dust (König et al. 2017). Further 24 µm emission is located in the vicinity of the EC region, slightly offset from the M8 clumps and extending to the central region of the nebula. As this emission does not coincide with the 870 µm emission from the clumps, it likely originates from a diffuse foreground gas layer. Individual pointlike 24 µm sources in this region coincide with the positions of stars from the open cluster NGC 6530 (see Fig. 3).

As shown in Fig. 4, the 8 µm emission extends in a bubblelike structure around the main condensations of the Lagoon Nebula (see Deharveng et al. 2010 for other examples). Further inspections of the individual clumps additionally reveal that the emission is strongest on the edges of the clumps traced by the 870 µm emission (see Fig. 5). The 8 µm band includes fluorescent emission from polycyclic aromatic hydrocarbons (PAHs; Draine & Li 2007), which is pumped by far-UV photons radiated by the present O-type stars. This is a clear indicator of the feedback from the stars on the surrounding remnant gas. In particular, the emission peaking on the clump edges indicates the presence of PDRs on clump surfaces across the nebula.

To examine the impact of stellar feedback on the physical properties of molecular clumps in the Lagoon Nebula, they were compared to the ATLASGAL sample of clumps examined by Urquhart et al. (2018). Analogously to these authors’ analysis, we reconstructed the cold dust SED of the clumps using a modified blackbody model: $$F_{\lambda }\left(T_{\text{d}},{\tau }_{\lambda ,\text{ref}}\right)={\Omega }_{\text{d}}\cdot B_{\lambda }\left(T_{\text{d}}\right)\cdot \left[1-\exp \left(-{\tau }_{\lambda ,\text{ref}}{\left(\frac{870\mu \text{m}}{\lambda }\right)}^{\beta }\right)\right].$$(1)

In this model, the dust blackbody emission, B_λ(T_d), at the dust temperature T_d is modified by a factor that is composed of the opacity ${\tau }_{{\lambda }_{\text{ref}}}$ at the reference wavelength λ_ref = 870 µm and a wavelength-dependent power law. This reference wavelength was chosen such that it matches the value used by Urquhart et al. (2018). The spectral index β was set to a fixed value of 1.75, which corresponds to the mean value across the dust models of Ossenkopf & Henning (1994) for star-forming regions. The clump size Ω_d is given by $\pi R_{\text{d}}^2$with the clump radius R_d, which we set to half of the FWHM source size of the respective clumps.

We used a two-component model to describe the full SED of clumps, presuming internal heating based on the examination of the 24 µm emission. The second component resembles a black body spectrum of a hotter embedded object: $$F_{\lambda }\left(T_{\text{d}},{\tau }_{\lambda ,\text{ ref }},T_{\text{h}},{\Omega }_{\text{h}}\right)=F_{\lambda }\left(T_{\text{d}},{\tau }_{\lambda ,\text{ ref }}\right)+{\Omega }_{\text{h}}\cdot B_{\lambda }\left(T_{\text{h}}\right).$$(2)

In addition to the parameters used in the single-component model, the size of the compact blackbody component Ω_h and its temperature T_h were determined.

For the cold component fit, only data points with longer than 65 µm were used. Due to the presence of diffuse warm gas in the vicinity of M8-Main and the EC region, mid-IR flux densities extracted at the associated clump positions are likely to be unrelated to the actual clump emission and were therefore also excluded for the fitting.

As mentioned above, the 8 µm band is dominated by the emission of PAHs in the outer layers of the clumps. Flux detected at this wavelength was used as upper limit when fitting the hot SED component, to avoid overestimating the continuum flux originating from the embedded object. Since the 65 µm band might contain additional emission from very small grains (Compiègne et al. 2010), the flux density at this wavelength was also used as an upper limit, in order to avoid an overestimation of the dust temperature.

Mass M and H₂ column density N(H₂) of the clumps were calculated according to the Eqs. (1) and (2) of Schuller et al. (2009). The bolometric luminosity L of the clumps was derived by integrating the flux density of the reconstructed models between 1 µm and 1000 µm and assuming isotropically radiating sources. The derived quantities and the corresponding SED plots can be found in Appendix C. Figure 6 illustrates the derived distribution of dust temperatures and clump masses. In particular, the central clumps C1–2 and clumps surrounding the central condensations in M8-Main show increased temperatures and comparably small clump masses. As discussed further in Sect. 6, these lower masses and increased dust temperatures are found for the whole sample of M8 clumps when comparing them to the ATLASGAL sample of clumps in the inner galaxy.

Fig. 3 Dust continuum image of M8 at 24 µm from MIPSGAL. The ATLAS-GAL 870 µm emission is shown as contours. Blue and green stars show the positions of O- and early B-type stars in this region (Wright et al. 2019), respectively. The image is saturated near the H II region M8-Main.

Fig. 4 8 µm dust continuum image obtained from the GLIMPSE survey. Blue and green stars respectively mark the positions of present O- and B-type stars (Wright et al. 2019). Contours of the ATLASGAL 870 µm emission are shown in grey.

Fig. 5 8 µm dust continuum image of the southern clumps in the Lagoon Nebula obtained from the GLIMPSE survey. Contours of the ATLASGAL 870 µm emission are shown in grey.

Fig. 6 Dust temperatures (upper panel) and masses (lower panel) of clumps in the Lagoon Nebula. The circle size corresponds to the aperture size used for flux extraction in the analysis. For the temperatures, the smaller inner circles visualise the derived temperatures of the hot components. The grey-scale background image shows the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2).

4 Line survey of M8 clumps

The rotational molecular line transitions in the M8 clumps were identified by matching rest frequencies from the Jet Propulsion Laboratory Line Catalog (JPLC⁸; Pickett et al. 1998) and the Cologne Database for Molecular Spectroscopy (CDMS⁹; Müller et al. 2001, 2005; Endres et al. 2016) to the observed spectra. In the first step, we visually inspected the spectrum of E, as we expected the lines to be brightest towards this massive star-forming region and because E has the longest integrated observing time of the clumps in the sample. Each significant line detected towards E was matched to a transition in the JPLC and CDMS with upper-level energy E_up < 200 K. If transitions of multiple species have rest frequencies close to observed transitions, the identification favours species commonly present in the ISM and transitions with low E_up and high Einstein A coefficients A_ij.

For the remaining M8 clumps, we executed a CLASS script to examine the respective spectra for line emission. The script flagged emission inside a 10 km s⁻¹ interval around the respective clump velocities (see Sect. 5.1) for all rest frequencies of transitions identified at E. All of the spectra were then visually inspected to identify any lines that are not present at E and to rule out false detections due to emission from unrelated nearby transitions. This visual inspection led to the detection of N₂D⁺, which is not seen at E.

A transition is considered to be detected if it has at least three adjacent channels with intensity higher than three times the baseline RMS noise for velocity resolutions of 0.4 km s⁻¹ (APEX) or 0.7 km s⁻¹ (IRAM 30m). Line candidates that show at least one channel above three times the RMS noise were confirmed based on their appearance in both polarisations and the presence of other transitions of the same species in the corresponding clump.

Across all clumps in the nebula, it was possible to identify a total of 346 transitions of 70 molecular species, including iso-topologues. Table 1 provides an overview of all the detected species in the M8 region.

All species (except N₂D⁺) and most of their emission lines were observed towards clump E, which hosts the embedded YSO M8E-IR. The large chemical richness of this object is explained by it being an early-stage massive star-forming region with an associated PDR. In addition, it was possible to detect also fainter transitions in the 3 mm band towards this clump, as the increased observing time at E reduces the RMS noise of the combined spectrum to 10 mK, as compared to 17 mK on average for the other clumps.

Figure 7 shows that almost all M8 clumps have a complex chemistry that varies across the cloud with an average of 30 species detected at each position. While the category of O-bearing species contains mostly molecules with bright lines detected in the entire cloud, the clumps show interesting differences in the number of (and ratio between) detections of N-bearing, S-bearing, and deuterated species. For instance, the number of nitrogen- and sulphur-bearing species detected in the SE region is overall larger than in the SC clumps. In contrast, we observe each region to have a few individual clumps with a large number of deuterated species, without a pronounced trend of a whole filament with a higher deuteration fraction.

The chemistry seen in the individual clumps is discussed further in Sects. 5.4 and 6.3. A detailed description of the individual detected transitions in each clump can be found in Appendix D, which lists the line parameters of each detected transition in Table D.1 as well as the derived line properties in Table D.2.

4.1 Line properties

In order to further analyse the chemical properties of the clumps in M8, Gaussian profiles were fitted to all detected transitions. In general, the fits are performed using the GAUSS method of the MINIMIZE function of CLASS, which returns the integrated intensity, I, the FWHM line width, ∆v, and the line of sight (LOS) velocity with respect to the local standard of rest (LSR) velocity, v_LSR, of a given line. Partially blended transitions and blended velocity components of the same transition were fitted simultaneously. If well resolved, a maximum of two Gaussian components of the same transition were fitted as possible additional components were almost exclusively detected for transitions of CO isotopes.

The spectra partially resolve the hyperfine structure (HFS) due to the non-zero nuclear spin of ¹⁴N for DCN, HCN, H¹³CN, N₂H⁺, N₂D⁺, and NH₂D. In addition, a splitting of the transitions of C³³S and C¹⁷O is observed as a consequence of the non-zero nuclear spin of ³³S and ¹⁷O. Corresponding transitions show a non-Gaussian line profile and were therefore fitted using the HFS method of MINIMIZE, which additionally returns the optical depths of the transitions. This function was also used to fit the well-resolved hyperfine components of lines from NS, NO, HCO, c-C₃H, CN, ¹³CN, C₂H, C₂D, and C¹³CH.

Observed line intensities of the CN 226 GHz transitions do not match the relative intensities expected based on the HFS calculations provided by the CDMS. A similar behaviour is noted by Kim et al. (2020) for the HFS transitions of CN at 113 GHz and HCO at 87 GHz, which they attribute to optical depth effects and non-local thermodynamic equilibrium (LTE) excitation. As the corresponding transitions of CN and HCO are only weakly affected in the M8 clumps, we used the respective HFS fits for the column density estimation in Sect. 5.4. In contrast, we refrained from computing column densities for the CN 226 GHz transitions.

A complete overview of the fitted line profiles for detected transitions is provided in Table D.2. For emission lines with multiple velocity components that are blended due to small differences in their observed frequencies, we give the integrated intensities over the full line profiles. The results of this survey are analysed and discussed in Sects. 5 and 6.

Fig. 7 Distribution of species across the M8 clumps. Different colours differentiate between S-, O-, and N-bearing species, deuterated species, carbon chains, complex organic molecules (COMs), and other species, as given in Table 1.

4.2 Radio recombination lines

In addition to the molecular rotational transitions, radio recombination lines (RRLs) of hydrogen, helium, and carbon are detected in the vicinity of M8-Main. In particular, the frequency setups observed with the IRAM 30m telescope cover HNα, He Nα, and C Nα recombination lines with N=39−42 and N=44. Towards HG, which coincides with the main ionisation source Her 36, we additionally detect multiple HNβ and HNγ transitions with N=48−56 and N=54−63, respectively. Using APEX, we also detect the H 29−30α and H 36−38β lines at HG.

For the analysis of the RRLs, we focused on the HNα, He Nα, and C Nα transitions with N=39−42 and N=44. The covered transitions with different N were averaged along the velocity axis to create individual combined spectra of the respective RRL towards each clump. Gaussians were fitted to each combined spectrum, in order to derive the line properties for the detected RRLs. The results of these fits are given in Appendix E, where we also give the line parameters of the H Nβ and H Nγ lines at HG. Figure 8 provides an overview of clumps with H 39−44α emission and the respective line intensities.

While the H Nα and He Nα RRLs observed towards HG have broad spectral line profiles and central velocities close to the systemic LSR velocity (between 2 km s⁻¹ and 4 km s⁻¹ ), the CNα lines are observed to be relatively narrow and shifted to velocities of 11.1 km s⁻¹ (see Fig. 9). These findings are in agreement with the structure of M8-Main derived by Tiwari et al. (2018), who place the dense molecular gas in the background of the H II region at velocities between 10 km s⁻¹ and 13 km s⁻¹. The main fraction of ionised carbon hereby originates from the warm gas of the PDR between the dense molecular clump and the H II region, as indicated by the narrow line widths and velocity shift of the CNα transitions. This is confirmed by the observations of the 157 µm CII transition at HG from Tiwari et al. (2018), which is shown as the blue spectrum in Fig. 9. While the v_LSR of the more strongly emitting CII component agrees with the velocities of the C Nα RRLs, the weaker component at approximately 6 km s⁻¹ is likely associated with the hotter foreground layer. This foreground gas contributes to most of the H Nα and He Nα emission and is expanding towards us with relative velocities between 2 km s⁻¹ and 6 km s⁻¹.

In contrast to HG, the remaining clumps in the Lagoon Nebula either only show the H 39-44α lines or no RRL emission at all (see Fig. 8). Unsurprisingly, the brightest recombination lines are observed in the vicinity of HG, where the clumps, in projection, are closest to the O-type stars (Her 36 and Sgr 9). The velocities in this region hereby largely follow the trend observed for the CII emission in Fig. 6 of Tiwari et al. (2018). Their channel maps indicate an enhanced ionisation of the molecular gas in WC4, which could possibly explain the low degree of chemical variety observed at this position (see Fig. 7). Additional weak H 39−44α emission is detected towards the EC clumps. The weak line strengths and the measured velocities of v_LSR = 0 km s⁻¹ imply that the emission does not originate from the associated clumps, which we observe at velocities of order v_LSR = 16 km s⁻¹. Instead, we might observe a less dense foreground gas layer, which is affected by the radiation of the nearby massive stars.

Fig. 8 Intensity and LOS velocities of the H 39−44α transitions observed at the positions of the clumps in the Lagoon Nebula. The fillings of the coloured dots indicate the intensities and the borders describe the velocities. Blue and green stars respectively mark the positions of present O- and B-type stars. The grey-scale background image shows the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2).

Fig. 9 CII 157 µm transition (blue) and stacked spectra of the HeNα and CNα RRLs with N=39−42 and 44 (black) towards clump HG (Her 36). The velocity scale of the RRLs is computed based on the rest frequency of the CNα lines. The Gaussians fitted to the RRLs are displayed in red. The temperature scale of the 158 µm CII fine structure transition is divided by 2000 to compare it with the RRLs.

4.3 Methanol maser emission

Methanol masers are commonly associated with star formation. As described by Menten (1991), these objects can be classified into two distinct classes. The Class II methanol masers are associated with the presence of high-mass protostars (Urquhart et al. 2015), where they are presumably pumped by the radiation of the surrounding warm dust (Sobolev et al. 1997). In contrast, Class I methanol masers are collisionally pumped (Lees 1973), due to which they are associated with the shocked material of protostellar outflows (Cyganowski et al. 2009).

Leurini et al. (2016) provide an overview of all known Class I maser transitions, some of which are also detected in the clumps of the Lagoon Nebula. Of these, the transitions with the highest detection rate for clumps in M8 are the lines at 84.5 GHz, 95.2 GHz, and 218.4 GHz. The detection of these transitions alone does not automatically imply the presence of maser emission, as the corresponding transitions could also be thermally excited. In order to probe if this is the case, the line properties derived in Sect. 4.1 were used to examine the line widths of these transitions.

As methanol masers amplify the emission from the respective transitions, typical line profiles are narrow and do not necessarily possess Gaussian shapes. Figure 10 shows the line profiles of the maser transitions observed at E and SE7, which are characteristic of the line profiles observed at the remaining clumps. The methanol 84.5 GHz and 95.2 GHz transitions at E have FWHM line widths of 1.04 km s⁻¹ and 1.17 km s⁻¹, respectively, less than half the median line width of 2.45 km s⁻¹ for non-masing methanol transitions at this source. The 218.4 GHz transition at SE7 has a width of 1.06 km s⁻¹, which is about 0.8 times the median width of 1.37 km s⁻¹ for non-masing methanol transitions at SE7.

M8 East is a known host to Class I methanol masers (see, for example, Kogan & Slysh 1998; Sarma & Momjian 2009), which is in agreement with it showing the brightest 95.2 GHz transition of the sample. The narrow line profiles indicate maser emission for the two transitions at 84.5 GHz and 95.2 GHz, originating from the vicinity of M8 East, the massive star-forming region containing clump E. Similar narrow line widths between 0.5 and 0.8 times the median line width of non-masing methanol lines are detected for the 95.2 GHz transition at HG, WC1, and SE7, while only SE2 shows potential maser emission at 84.5 GHz.

In contrast to the bright transitions in the 3 mm atmospheric band, the potential maser transitions observed at 218.4 GHz are very faint (see Fig. 10). Given that only a few sources have been reported to detect 218.4 GHz maser emission (Hunter et al. 2014; Leurini et al. 2016), it is not surprising that the emission observed at M8 is also not very strong. Nevertheless, we potentially observe 218.4 GHz masers at the positions of HG, EC4, EC5, SE1, SE7, and SC8, where the line widths of the 218.4 GHz transition are about 0.8 times the respective median methanol line widths of non-masing transitions.

The line widths provide a reliable indicator of the presence of methanol masers in M8. A confirmation of the maser activities in the clumps would require high angular resolution interfero-metric observations, supported by a detailed radiative transfer modelling of maser and thermally excited methanol transitions, which would go beyond the scope of this work. Furthermore, interferometry alone of the M8 clumps could confirm maser action.

Fig. 10 Typical line profiles of the discussed maser transitions. Upper panel: methanol maser emission of the 84.5 GHz and 95.2 GHz transitions at clump E, respectively shown in blue and red. Lower panel: maser emission of the 218.4 GHz transition at the position of SE7.

5 Analysis

5.1 Systemic clump velocities

A first overview of the velocity structure of the Lagoon Nebula was obtained by Tothill et al. (2002) based on observations of ¹³CO and C¹⁸O. While they noted the presence of double-peaked line profiles, they only give the central line velocity for the stronger component at each clump. To gain a more detailed view of the velocities in the Lagoon Nebula, the analysis of CO line profiles was repeated based on our new APEX data of the J = 2−1 transitions at each clump.

While the profiles of ¹²CO and ¹³CO lines show wings and optical depth effects, the transitions of the less abundant isotopes C¹⁸O and C¹⁷O are mostly optically thin. Due to this, we used the C¹⁸O and C¹⁷O J = 2−1 transition data to derive the LOS velocities v_LSR of the clumps (see Sect. 4.1). For positions at which both transitions are detected, the weighted average of both shifts was computed. For clumps with multiple velocity components observed in these optically thin transitions, we chose the peak velocity for the strongest two components. Table 2 gives an overview of the derived v_LSR at each clump position. As WC3, SE8, and SC5 have been observed at deviating coordinates with both telescopes, the velocities at these clump positions observed with the IRAM 30m telescope were estimated based on the J = 1−0 transitions of the same species.

As can be seen in the upper panel of Fig. 11, the clumps in M8 show velocity gradients along the filaments. This suggests that the clumps in the respective cloud parts are likely to be kinematically related. In contrast, the systemic velocities of the individual filaments differ across the nebula. With respect to the southern clumps, the WC clumps in M8-Main show blue-shifted emission, while the EC clumps of the central ridge show significantly high redshifted velocities. This relatively large-scale velocity gradient in the western half of M8 is also apparent in the position-velocity (PV) diagram shown in the lower panel of Fig. 11 and might be caused by the radiation or mechanical feedback of the massive stars on the remnant gas. The SE clumps do not seem to follow this trend, as they branch out to lower velocities in the PV diagram. The cloud-scale kinematics in M8 is discussed further in Sect. 6.2.

Fig. 11 LOS velocities of clumps in the Lagoon Nebula. Upper panel: positions of the individual clumps (circles), with the grey-scale image showing the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2). The circle size corresponds to the beam size of the on-off observations with APEX and the IRAM 30m telescope. Bi-coloured circles visualise distinct velocity components observed in the same beam. Lower panel: LOS velocity of the M8 clumps as a function of RA separation from the cloud centre at RA=18h04m 06s. Multiple velocity components inside one beam are shown as individual points.

5.2 Kinetic temperatures and H₂ volume densities from para-formaldehyde

The dust temperatures of the clumps in M8 were derived in Sect. 3 using the SEDs obtained from the dust continuum images. In order to complement the derived temperatures with values from the line emission, we estimate the rotational temperatures of para-formaldehyde (p-H₂CO), acetonitrile (CH₃CN), and methyl acetylene (CH₃C₂H) in the following two sections.

Formaldehyde in particular has been shown in the past to be a good thermometer for dense molecular clumps (Mangum & Wootten 1993). As this molecule is a slightly asymmetric rotor (described by $J_{K_a,K_c}$), each respective energy level is split further into multiple levels with different K values as a consequence of different projections of the rotational axis on the symmetry axis of the molecule. While line ratios involving transitions with different angular momentum quantum numbers, J, are sensitive to density deviations, transitions with the same J can be used to obtain reliable temperature estimates of the gas (Mangum & Wootten 1993).

The abundances of ortho- and para-formaldehyde are not equal, which is why transitions of the same symmetry state have to be considered when deriving temperatures and densities. As a consequence, we limited this analysis to the p-H₂CO transitions 3_0,3−2_0,2, 3_2,2−2_2,1, 3_2,1 −2_2,0, and 1_0,1 −0_0,0. While the J = 3−2 transitions enable the derivation of rotational temperature T and p-H₂CO column density N, adding the J = 1−0 transition allowed us to estimate the H₂ volume density $n_{{\text{H}}_2}$ in the M8 clumps.

In order to derive the physical properties of the M8 clumps, we followed the approach introduced by Christensen et al. (in prep.) by utilising the Python wrapper pyradex for the non-LTE radiative transfer code RADEX (van der Tak et al. 2007) in combination with the emcee (Yang et al. 2017) package for Python, which implements a Markov chain Monte Carlo (MCMC) algorithm. To obtain the line parameters of p-H₂CO with pyradex, we assumed a background temperature of T_bkg = 2.7315 K and used the collisional rate coefficients calculated by Wiesenfeld & Faure (2013). We fixed the line width in the computation to the weighted average line width of the p-H₂CO transitions at the individual clumps. This line width should have minimal variations from line to line as all transitions should be probing the same gas. The line properties in M8 derived in Sect. 4.1 were then fitted to obtain the physical parameters of volume density, kinetic gas temperature and p-H₂CO column density. The starting position for the MCMC was obtained by scipy.curve_fit, after which the MCMC algorithm explored the parameter space within $n_{{\text{H}}_2}=10-{10}^7{\text{cm}}^{-3}$, T = 10−300 K and N = 10¹⁰−10¹⁷ cm⁻². For each clump, 1100 steps were taken where the first 100 were discarded, and the last 1000 steps were converging on the best fit physical conditions. This process is detailed in Christensen et al. (in prep.). Derived temperatures and column densities as well as the posterior probability distributions visualising the explored parameter space are shown in Appendix F.

Formaldehyde is ubiquitous in the ISM and its transitions studied by us probe gas layers with temperatures < 100 K (Mangum & Wootten 1993). Hence, it is not surprising that it has a high detection rate among the clumps in M8. As can be seen in Fig. 12, the temperatures derived from para-H₂CO data are on average 20 K higher than the dust temperatures. To probe the presence of a linear correlation between dust and rotational temperatures, we calculated the corresponding Pearson correlation coefficient R and P values with the python module scipy.stats (Virtanen et al. 2020). Given the Pearson coefficient R of 0.72, the dust temperatures and p-H₂CO rotational temperatures are linearly correlated in the M8 sample. This indicates that the analysed formaldehyde transitions probe the clump envelope.

As the upper-level energy of both the H₂CO 3_2,2−2_2,1 and 3_2,1 −2_2,0 transitions is 68 K above the para ground state, only the warmer gas contributes to the H₂CO temperature measurement. In particular, the WC4 clump shows relatively high temperatures compared to the surrounding clumps. As discussed in Sect. 4.2, Tiwari et al. (2018) detect an influence of the ionised gas on the position of WC4, which might also attribute to the heating of the clump.

Fig. 12 Temperatures derived from the rotational transitions of paraformaldehyde, acetonitrile, and methyl acetylene as a function of dust temperatures at the clumps in M8. The legend in the upper-left corner provides the Pearson correlation coefficient R and the P values for the respective samples.

5.3 Rotational temperatures of acetonitrile and methyl acetylene

Both acetonitrile and methyl acetylene are symmetric top molecules (described by J_K) that show multiple transitions with the same angular momentum quantum number J and with different values for the projected angular momentum K at similar frequencies. Thus, both species have been found to provide good temperature estimates for astrophysical environments (e.g. Askne et al. 1984; Bisschop et al. 2007). A study of galactic molecular clumps by Giannetti et al. (2017) reveals that both these species trace warm gas in the clumps, while higher-energy CH₃CN transitions additionally trace even warmer embedded hot cores. Based on their findings, all the transitions we detected towards the M8 clumps are arising from the extended warm component.

We used rotation diagrams with a simple least-squares fit to derive temperatures and column densities of CH₃CN and CH₃C₂H assuming LTE. This was done for the M8 clumps towards which we detect at least three spectral lines. A detailed explanation of the method, the derived physical parameters, and the respective rotation diagrams are shown in Appendix F. Figure 13 visualises the rotational temperatures of acetonitrile and methyl acetylene alongside the kinetic temperatures from para-formaldehyde.

The analysed transitions of CH₃ C₂H and CH₃ CN are excited either internally by star formation or externally by the feedback from the surrounding massive stars. In contrast to the p-H₂CO temperatures, the temperatures derived with CH₃C₂H and CH₃CN do not correlate with the dust temperatures (see Fig. 12), which may suggest that we primarily observe internal heating. Overall, we find similar to slightly higher rotational temperatures for acetonitrile as compared to methyl acetylene. In particular, we measure comparably high acetonitrile temperatures for the clumps EC1, EC3, and SC8, which do not have a temperature estimate based on methyl acetylene. Finding higher acetonitrile temperatures is compatible with the results of Giannetti et al. (2017), who conclude that the gas layers traced by CH₃C₂H extend further to the outer parts of the clump core than the regions traced by CH₃CN. Interestingly, the opposite is seen for the HG and WC1 clumps in the vicinity of M8-Main, where methyl acetylene traces higher temperatures. A possible explanation may be an influence of the nearby H II region, which externally heats the outer layers of these clumps, where methyl acetylene is more common than acetonitrile. Overall, the temperatures derived from CH₃C₂H and CH₃CN are within the expected values for the warm gas component surrounding clump cores derived by Giannetti et al. (2017).

Despite being bright in the 24 µm dust continuum (see Fig. 3), no CH₃CN emission is detected towards the WC7 core. This missing line emission hints at the absence of a hot core at this position, which implies that the observed mid-IR emission at this position is unrelated to the clump. This is further confirmed by the methanol emission at WC7, which we only detect in the lowest energy transitions at 96.7 GHz. These transitions are typically very strong and therefore also detected in most IR-dark clouds (e.g. Leurini et al. 2007). The non-detection of higher-energy methanol transitions towards this clump suggests a low excitation of this species and therefore the absence of a hot core at WC7.

5.4 Column densities

Using the line intensities, I, derived in Sect. 4.1, column densities of the detected species can be computed. Assuming optically thin emission and LTE, the column density, N, can be described as a function of I, the clump temperature T, and the background temperature Tbg according to $$\begin{array}{l}{N}_{\text{total }}^{\text{thin }}\left(T,I,T_{\text{bg}}\right)=\frac{Q(T)}{g_{\text{u}}}\cdot \frac{\exp \left(\frac{E_{\text{u}}}{k_{\text{B}}T}\right)}{\exp \left(\frac{hv}{k_{\text{B}}T}\right)-1}\\ \cdot \frac{1}{J_{v}T-J_v\left(T_{\text{bg}}\right)}\cdot \frac{8\pi v^3}{c^3{A}_{\text{ul}}}\frac{I}{f},\end{array}$$(3)

by introducing the Rayleigh-Jeans equivalent temperature J_v(T)=hv/[k_B(exp(hv/(k_BT)) − 1)] (Mangum & Shirley 2015). Additional parameters are the Planck constant, h, the Boltzmann constant, k_B, the speed of light, c, and the transition-specific frequency, v, upper-level energy, E_u, upper-level degeneracy, g_u, and spontaneous Einstein A coefficient, A_ul. In order to account for the clump sizes θ_S, the measured intensities were corrected by the beam filling factor $f={\theta }_{\text{S}}^2/\left({\theta }_{\text{S}}^2+{\theta }_{\text{B}}^2\right)$ with the FWHM beam width, θ_B. The corresponding clump sizes for this were adopted from Tothill et al. (2002). Analogous to Kim et al. (2020), we approximated the kinetic temperature, T, in Eq. (3) with the dust temperatures, T_dust, derived in Sect. 3, as they are available for the full sample of M8 clumps. This approximation can lead to an underestimation of the derived column densities, as the dust temperature of a respective clump tends to be lower than its actual kinetic temperature, especially for species that trace the warmer gas component, such as H₂CO, CH₃CN, and CH₃C₂H. A more precise approximation for the kinetic temperatures of the clumps are the excitation temperatures derived from rotational transitions of certain molecular species (see Sect. 5.2). Nevertheless, we used the dust temperatures here since they are available for all M8 clumps, while deriving rotational temperatures was only possible for a subsample of them.

The assumptions mentioned above do not hold in many cases, for example where emission is optically thick like in CO, HCN, or CS. As the optical depth τ of transitions with fitted HFS is known (see Table D.2), the derived column densities for the corresponding species were corrected according to $$N_{\text{total }}=N_{\text{total }}^{\text{thin }}\cdot \frac{\tau }{1-e^{-\tau }}.$$(4)

For other optically thick transitions, the derived column densities, , act as lower limits to the actual column density of the species.

If multiple transitions of a species are detected in a molecular clump, the median of the column densities derived from each line was computed. For species that remain undetected in some clumps, we estimated upper limits to their column densities by using the RMS at lines of these species detected in other clumps of M8. For all possible transitions of the species, we independently calculated an upper limit of the column density based on the median line width at the respective clump and a peak intensity equal to three times the spectrum RMS close to the non-detected transition. The lowest limit obtained with this method was then considered to be the upper limit of the column density.

All derived column densities and upper limits are presented in Table G.1. Figure 14 shows an overview of all species that have been detected in at least ten clumps of M8.

In addition to the most common tracers of dense clumps in the ISM, we also detect PDR tracers such as HCO, c-C₃ H₂, CN, and C₂H (see Kim et al. 2020 and the references therein), in a large fraction of clumps in M8. This is consistent with the widespread 8 µm emission detected on the surfaces of the M8 clumps (see Sect. 3). About half of the clumps also show the presence of shocks, as indicated by the detection of SiO (e.g. Bergin & Tafalla 2007; Schilke et al. 1997; Bachiller et al. 1991). In contrast, some clumps also show the presence of cold and dense gas tracers such as NH₂D, N₂D⁺, and DNC. A more detailed comparison of the chemical conditions in the M8 clumps is given in Sects. 6.3 and 6.4.

Fig. 13 Temperatures in clumps in the Lagoon Nebula derived from different molecular species. Coloured circles show the kinetic temperatures from para-formaldehyde (annuli) and the rotational temperatures of acetoni-trile (left inner circle) and methyl acetylene (right inner circle). The size of the inner circles corresponds to the approximate beam size of the observations. The grey-scale background image shows the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2).

Fig. 14 Column densities of species detected in at least ten M8 clumps. The colour scale gives the relative abundance of a respective species to the column density of C18O at the same position. White cells indicate the non-detection of a species in the associated clump.

6 Discussion

6.1 Dust continuum clump properties

The physical properties of the M8 clumps were derived in Sect. 3. In this section, they are compared to the overall population of clumps in the inner Galactic plane examined by Urquhart et al. (2018). Their sample of clumps is based on the ATLAS-GAL survey of the Galactic plane and therefore contains a large variety of different sources.

ATLASGAL has the highest source densities at distances between 2 and 4 kpc, further away than the M8 clumps, which have a distance of 1.3 kpc. This imposes constraints on the spatial resolution and sensitivity, implying that the typical ATLASGAL source will have a larger physical size and is more massive compared to the clumps in M8. At these larger distances, it is even possible that the corresponding ATLASGAL sources contain sub-structure similar to multiple M8 clumps. In order to compare similar objects, we therefore also compared the clumps of M8 with a distance-limited sample of the ATLASGAL clumps, containing only sources that are at distances smaller than 1.5 kpc.

Furthermore, the ATLASGAL survey is likely to be incomplete for clumps below a few hundred M_⊙. For instance, out of the 37 molecular clumps studied in this work, only 14 are retrieved in the ATLASGAL survey. While the comparably bright clumps such as SC1-3 might have been rejected due to their extended elongated shapes, other adjacent clumps like EC4 and 5 are recognised as a single clump. In contrast, the 870 µm peak emission from the weaker clumps SW1 and C1-3 is below the detection threshold of about 0.4 Jy beam⁻¹, which was applied for the source extraction of the ATLASGAL compact source catalogue (Contreras et al. 2013). To account for these differences in source selection, we compared the distance-limited ATLASGAL sample with both the full sample of M8 clumps and a subsample of M8 clumps that do have an ATLASGAL counterpart.

We performed two-sample Kolmogorov-Smirnov (KS) tests between physical properties of clumps in the ATLASGAL and the M8 samples using the Python module scipy.stats. These tests estimate the p-value, which indicates the likelihood that the properties of both samples are distributed equally. A low p-value indicates that both samples originate from different distributions, while values above p > 0.0013 suggest that the samples have the same underlying distribution (Urquhart et al. 2018). Cumulative distribution plots of the clumps’ radii, masses, luminosities, L/M ratios, and dust temperatures are shown alongside the corresponding p-values in Figs. 15 and 16.

The clump masses and physical radii are compared in Fig. 15. Effective radii for the ATLASGAL sources are provided by Urquhart et al. (2018, see their Table 5), who derive the values by multiplying the geometric mean of the deconvolved semimajor and semi-minor axes by 2.4 (Contreras et al. 2013). The radii of the M8 clumps were calculated analogously based on the distance of 1 .3 kpc and the clump sizes derived by Tothill et al. (2002, see their Table 1).

Both masses and radii are significantly lower in the M8 clumps compared to the full ATLASGAL sample of clumps. While for the clumps in M8, the median radius and mass is about 0.16pc and 10 M_⊙, respectively, the ATLASGAL clumps tend to be larger and more massive with a median radius of 0.37 pc and a median mass of about 500 M_⊙. As argued above, this discrepancy is less pronounced when comparing with the distance-limited sample that has median radii of 0.12pc and masses of 50 M_⊙.

The sample of M8 clumps with ATLASGAL counterparts has a median mass of 25 M_⊙, which is only a factor of two smaller than the median mass of the distance-limited ATLAS-GAL clumps. However, the p-value of p = 7.7 × 10⁻⁴, obtained when comparing these samples, suggests that the underlying distribution of masses still differs. A possible physical mechanism that could cause these smaller clump masses is the fragmentation of the filaments by the radiation pressure of the nearby O stars in M8. Advocation of this mechanism, however, is in contrast with a recent study by Mazumdar et al. (2021), who examined the clump properties in the star-forming complex G305. They find increased clump masses as a result of the collect and collapse feedback mechanism. We believe this contradiction is due to the differences in the ages and star formation histories of M8 and G305. While the ages of the stellar clusters in the G305 complex vary between 1 Myr and 3 Myr (Davies et al. 2012), the initial trigger of the star formation in M8 is estimated to have occurred around 4Myr ago (Damianí et al. 2019). As a consequence, the M8 cloud might be allowing for a stronger dispersal of the clumps, while in the G305 complex the fragmentation has only started recently, maybe as a consequence of a previously triggered star formation phase. Furthermore, the distance of G305 of about 4 kpc suggests that it is not possible to resolve the sub-structures in that complex in the same detail as in M8.

With a median value of 200 L_⊙, luminosities in the M8 sample are only lower by a factor of four compared to the typical luminosities of about 800 L⊙ observed for the full sample of clumps in the inner Galactic plane. Additionally, the M8 luminosities are slightly higher when compared to nearby ATLASGAL clumps, which have a median luminosity of 83 L_⊙ (see the upper panel of Fig. 16). As shown in the middle panel of Fig. 16, the combination of small measured masses and comparable luminosities leads to increased L/M ratios with a median of 17 L_⊙ $M_{\odot }^{-1}$ for the M8 clumps as compared to about 2 L_⊙ $M_{\odot }^{-1}$ for the ATLASGAL samples.

Moreover, the derived dust temperatures of the M8 clumps, shown in the bottom panel of Fig. 16, are higher by approximately 5 K with respect to the ATLASGAL sources. The full and close-by samples of ATLASGAL clumps have median dust temperatures of 18.6 K and 19.2 K. In contrast, the median dust temperature of M8 clumps is 23.6 K, indicating that the temperatures are approximately 25% higher. The evolutionary sequence for dense clumps introduced by König et al. (2017) and refined by Urquhart et al. (2018) predicts these conditions of increased temperatures and large L/M ratios only for clumps associated with evolved, massive YSOs and H II regions. While such sources exist in the major star-forming regions (M8-Main and M8 East) of the Lagoon Nebula, arguably not every observed clump is host to a massive stellar object. For instance, some of the warmest clumps are the central clumps C1-2, which due to their low masses below 3 M_⊙ are not capable of harbouring a massive object.

In addition to evolved star-forming regions, a likely cause for the higher temperatures and luminosities is the presence of an externally heating source. As shown in Sect. 3, almost all clumps in the Lagoon Nebula are being exposed to the radiation from the nearby massive stars, which is revealed by the presence of numerous PDRs traced by 8 µm PAH emission, which is the fluorescent result of UV radiation from the nearby O stars that heats the outer regions of the clumps. This external heating increases the luminosities of low-mass M8 clumps, which intrinsically are likely to have low luminosities characteristic for clumps of the inner Galactic plane.

When only considering the M8 sample of clumps that have an ATLASGAL counterpart, this temperature offset seemingly vanishes at higher temperatures. A possible reason for this is that the higher temperature clumps also correspond to the less massive clumps offset from the filament that were rejected in the ATLASGAL survey. We therefore conclude that especially these less massive clumps are affected by the external heating.

Fig. 15 Cumulative distributions of the clump radii derived by Tothill et al. (2002, upper panel) and masses (lower panel) in M8 and for the full and distance-limited sample of ATLASGAL clumps. The M8 clump masses shown in orange colour are corrected based on the results of Appendix B. The dotted orange distribution only contains M8 clumps that have a nearby counterpart in the ATLASGAL sample. The p-values of KS tests between selected samples are provided as grey text.

Fig. 16 Cumulative distributions of the clump luminosities (upper panel), the L/M ratios of the respective clumps (middle panel), and clump dust temperatures (lower panel) in M8 and for sources of the full and distance-limited ATLASGAL sample. The M8 distributions shown in the orange colour are corrected based on the results of Appendix B. The dotted orange distribution only contains M8 clumps that have a nearby counterpart in the ATLASGAL sample. The p-values of KS tests between selected samples are provided as grey text.

Fig. 17 Line profiles of the C18 O (2−1) transition for clumps located in the EC ridge.

6.2 Kinematics in the Lagoon Nebula

The velocities of the individual LOS components towards the M8 clumps were examined in Sect. 5.1 based on the position of the peak intensities of the optically thin C¹⁸ O and C¹⁷ O spectral lines. The v_LSR overview in Fig. 11 shows velocity gradients of the clumps along the filaments. For example, the LOS velocity gradually changes from 13.5 km s⁻¹ to 12.1 km s⁻¹ for the SE1-SE5 clumps. The only exception is the other relatively faint redshifted component of SE2 that likely originates from a more diffuse gas component.

The distinct velocity components observed in lines from many species towards M8-Main were examined in detail by Tiwari et al. (2018), who observe the dense gas components to be seen with higher systemic velocities above 10 km s⁻¹, while the blue-shifted gas is accelerated towards us and is powered by the radiation of the nearby O-type stars. The two velocity components found in our observations are consistent with that interpretation. We note that we only analysed one bright component of the WC1 clumps, while the CO line profiles suggest the presence of multiple weak components in this region.

Another striking example of velocity and intensity gradients is observed towards the EC ridge whose C¹⁸O (2−1) line profiles are shown in Fig. 17. While the component at 12 km s⁻¹ shows approximately the same LOS velocity for all EC1-EC3 clumps, the peak of the redshifted component at about 15 km s⁻¹ shifts by 1.3 km s⁻¹ along the ridge. It is also interesting to examine the intensities of the respective components, as the brightness of the 12 km s⁻¹ component increases from south to north, while the 15 km s⁻¹ emission gets fainter along this direction. The EC4 and EC5 clumps follow the velocity shift in the northern direction, with bright emission centred at approximately 17 km s⁻¹. In contrast to the three southern clumps, these clumps show no emission component at 12 km s⁻¹.

As both EC velocity components are seen in high-density tracers such as HNC, it is probable that both components correspond to dense gas layers. Nevertheless, it is difficult to judge if the two components actually interact, as these findings are purely based on selected on-off observations, while a complete mapping of the region is still missing. The overall redshifted emission of the EC region may be explained by a location of the clumps behind the massive stars of the open cluster NGC 6530. It is possible that the radiation from the associated massive stars transfers momentum to the dense molecular gas, which would explain the observed high redshift velocity component with receding clump motions.

In contrast to the EC region, the velocity components observed in the SC2 and SC3 clumps are less separated and approximately show the same intensities for both clumps. The similarity in the spectral line profiles for both clumps is caused by an overlap of the observed beams. In contrast, the close proximity of the two line components shows that the two gas layers have a similar LOS velocity. This suggests similar motions of both gas layers, indicating that they are exposed to the same external forces.

On larger scales, the velocity distribution in the ISM around M8 shows two distinct northern regions, with the blueshifted components between 7 km s⁻¹ and 12 km s⁻¹ in the fore- and background of M8-Main, and the redshifted EC ridge in the background of the NGC 6530 cluster with velocities between 12 km s⁻¹ and 17 km s⁻¹. In contrast, the LOS velocities of the southern clumps are more clustered around a systemic velocity of 10 to 12 km s⁻¹. This small variation in velocity for the southern clumps only reflects their motion along the LOS. It is therefore possible that the southern filament instead moves perpendicular to the LOS, which may be due to an acceleration in southern direction caused by the massive stars located in the north.

6.3 Chemistry of the M8 clumps

The line survey towards all the clumps in M8 was described in Sect. 4, based on which we estimated the column densities of all the detected species in Sects. 5.2−5.4. The detection and column density distributions of the observed species across all the clumps are summarised in Table 1 and Figs. 7 and 14, respectively.

In general, a large variety of molecular species is detected towards all the clumps, including complex organic molecules (COMs) and deuterated species. M8 East in particular shows an overall higher number of detected species, which can attributed to its hosting of an embedded massive star-forming region (Tothill et al. 2008) and a PDR on the associated clump surface (Tiwari et al. 2020). The impact of these physical conditions extends to the nearby clumps SE7 and SE8, which also show a large number of observed species. In general, the southeastern filament is chemically the richest, which is likely caused by the presence of PDRs on the northern clump surfaces (see Fig. 4) that alter their chemistry. The high number of detected species in the SE1 clump hereby may be due to its location at the western end of the filament. As large parts of its surface are exposed to the incoming radiation from the O- and B-type stars, a large fraction of the observed beam might be occupied with the PDR towards this clump. As PDR tracers are detected towards all of the clumps in M8, they are discussed in detail in Sect. 6.4.

In other regions of M8, the individual clumps EC4, EC5, and SC8 also show an enhanced number of species, possibly hinting at a rich chemistry. As discussed further in Sect. 6.5, the EC4 and SC8 clumps are found to be forming low-mass stars. This can also be seen in their chemical composition, which contains cold and dense gas tracers such as N₂D⁺ and DCO⁺ in addition to the shock tracer SiO, possibly tracing an outflow from a low-mass protostar. In contrast, we detect neither SiO nor N₂D⁺ in EC5, indicating that the condensation may not yet have cooled enough to initiate star formation.

Towards the C1-3 and WC4 clumps, we only detect a few species that are mostly limited to those with the highest detection rates in the M8 cloud. As mentioned previously in Sect. 4.2, the position of WC4 is strongly affected by the ionising radiation of the nearby massive stars, possibly halting more complex chemical mechanisms in this clump. Comparing them to the remaining M8 clumps, all WC3-7 clumps show a decrease in chemical complexity, which is likely also caused by the radiation of the M8-Main H II region. While the C1-3 clumps are located at a larger distance from the massive stars, their small masses (see Table C.1) enable the ionising radiation to penetrate the clump surfaces, reducing the chemical complexity of the clumps. Moreover, it is possible that the incoming radiation could lead to a complete disintegration of the clumps, due to their small masses.

A cold and dense gas tracer, the N₂D⁺ emission line, is detected in a total of ten clumps, which are distributed over various regions across the nebula. Bright emission of N₂D⁺ is detected in the SE6-8 clumps, which are neighbouring the M8 East star-forming region. This further strengthens the argument of Tiwari et al. (2020), who suggest that the compression of gas introduced by the ionisation front north of M8 East may trigger star formation in this region. Interestingly, we do not detect N₂D⁺ directly at the E clump. As E is associated with an IR-bright source, the absence of N₂D⁺ could indicate a more evolved protostellar object in the clump core (Emprechtinger et al. 2009).

The shocked regions traced by SiO generally coincide well with the positions towards which we detect star formation (see Sect. 6.5). In addition, this species is also observed in almost all clumps of the SE filament. This indicates that the ionisation front that may be triggering star formation in M8 East also extends to the SE clumps located in the west.

6.4 PDR tracers in the Lagoon Nebula

PDR tracers such as HCO and c-C₃H₂, CN, and C₂H are detected towards all the clumps in M8. As shown in Fig. 18, clumps associated with the known PDRs at M8-Main and M8 East show higher HCO column densities as compared to the surrounding clumps. In addition, the southern clumps SC8 and SC9 have similar column densities when compared to the known PDR regions M8-Main and M8 East, while the column densities observed towards the SC1-SC4 clumps are significantly higher. Comparing the location of these clumps with the 8 µm emission in Fig. 4, it can be seen that these SC clumps are associated with bright PAH emission coming from a structure that looks like an ionisation front, which is receding away from the HG region. The C2 clump can also be associated with this structure, which may explain the high HCO column densities in this overall lower density clump.

Kim et al. (2020) examined PDR tracer abundances for a sample of massive clumps in the inner Galactic plane and find a relation of decreasing abundances with increasing H₂ column densities. This relation can be interpreted to mean that these species are more abundant in the less shielded outer parts of the PDRs, which are associated with stronger UV emission from external sources.

To investigate whether this anti-correlation is also seen in the clumps of M8, Fig. 19 shows the PDR tracer abundances (X(HCO) =N(HCO)/N(H) and X(c-C₃ H₂)=N(c-C₃H₂)/N(H₂)) as a function of the Hydrogen column densities (N(H₂)) determined in Sect. 3. As explained in Appendix B, the method used to calculate H₂ column densities overestimates the corresponding values by approximately a factor of 10^0.3 for source sizes around 30″. Due to this, Fig. 19 also shows data points computed with H₂ column densities that were corrected by this factor. The top axis of this figure displays the visual extinction A_v, as derived according to the conversion A_v = N(H₂)/1.88 × 10²¹ cm⁻² (Bohlin et al. 1978; Frerking et al. 1982).

It can be seen that the observed c-C₃ H₂ and HCO abundances in the M8 clumps are in agreement with the trend observed by Kim et al. (2020). Additionally, this trend can also be seen in the clumps with relatively low observed column densities in M8. This is an extension of the study by Kim et al. (2020), who purely rely on massive clumps. The HCO column densities observed towards the clumps in M8 are systematically higher than the values derived by Kim et al. (2020). While we used all transitions of the HCO hyperfine structure line for deriving the column density, Kim et al. (2020) only consider the weakest component.

In addition to HCO and c-C₃ H₂, the estimated abundances for CN and C₂H also follow the same trend as the clumps observed by Kim et al. (2020). This further confirms the suggestion that PDR species are located at the outer edges of the dense molecular clumps.

Fig. 18 HCO column densities at the M8 clumps. The circle markers correspond to the positions of the individual clumps, with the grey-scale image showing the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2). The circle size corresponds to the beam size of the on-off observations with APEX and the IRAM 30m telescope. Two-coloured circles visualise distinct velocity components observed in the same beam.

Fig. 19 Abundance of HCO (left panel) and c-C3 H2 (right panel) as a function of H2 column density. The Pearson correlation coefficient R and the P value for the M8 clumps are given in the upper-right corner.

6.5 Star formation in M8

This study of the Lagoon Nebula illustrates several signatures of star formation in the clumps based on the analysis of the dust continuum emission and the observed chemical species. In order to provide an overview of the star-forming clumps in M8, the most reliable probes are introduced in this section and the results are compared in Fig. 20.

The large abundance of ¹²CO in the ISM makes the line wings of its transition commonly used tracers to identify pro-tostellar outflows (e.g. Duarte-Cabral et al. 2013; Kahle et al. 2023). In M8, the spectral line profiles of ¹²CO are too complex for such an analysis due to the presence of many different velocity components along the LOS. As a consequence, we instead used formylium (HCO⁺) as a probe for outflow activities, since the line profiles of this species are also very sensitive to motions of the associated gas layers (e.g. Wyrowski et al. 2016). Excess emission in the line wings of the HCO⁺ therefore indicates the presence of a protostellar outflow in the clump and that of a protostellar object driving it.

For examining the M8 clumps, spectral line profiles of the HCO⁺ (1−0) and (3−2) transitions from each clump were inspected for this excess emission by comparing the spectra of the line wings with fitted Gaussian profiles for the line emission. As an example, excess emission in the line profile at SE3 is shown in Fig. 21. Among the clumps where we detect excess emission, about half show a one-sided excess at redshifted velocities like SE3, while the other half show excess emission in both line wings. The kinematic complexity of the region makes it difficult to pinpoint the origin of this excess. While it is possible that excess emission is hidden by the presence of a second velocity component, weak emission from additional gas layers might be mistaken for excess emission originating from molecular outflows.

Examining the mid-IR dust continuum towards M8 in Sect. 3 resulted in the identification of IR-bright sources that comprise all the clumps in the M8-Main region and several fainter point-like sources. The associated clumps might contain a bright internal heating source whose IR emission is strong enough to escape the molecular clump. As a consequence, the IR-bright clumps might be sites of intermediate to high-mass star formation. All IR-bright clumps are included in the diagram shown in Fig. 20, although it is not clear if the emission observed towards the clumps in the M8-Main region originates from the clumps themselves or from the surrounding H II region.

CH₃CN has often been used to probe hot molecular cores (e.g. Bisschop et al. 2007). While we find temperatures in the clumps of M8 that are lower than in typical hot cores (which are about 100 K), CH₃ CN does seem to probe warm parts of the clumps that do not correlate with their general dust envelope (see Sect. 5.2). Therefore, we used the CH₃CN emission here as a probe for additional heating within the clumps.

In contrast to the mid-IR emission, the millimetre line emission of CH₃CN escapes the molecular clumps even if the contained protostar is not very bright. The detection of this species therefore allows us to additionally trace ongoing low-mass star formation in the M8 clumps.

Based on these criteria, at least eight of the M8 clumps show signs of intermediate to high-mass star formation. In addition to the known sites of high-mass star formation in HG and E, the clumps WC1, SE2, SE3, SE7, SE8, and SC1 are likely to contain a protostellar object. WC1 and SC1 correspond to some of the closest clumps to the O star Her 36, which drives the H II region at M8-Main. Due to this, the star formation in these objects might have been triggered by the compression introduced by the radiation of this star. The SE7 and SE8 clumps are located in the region observed by Tiwari et al. (2020), who find signs of triggered star formation across M8 East. The independent observation of star formation in this region presented in this study validates their findings.

In addition to sites of intermediate to high-mass star formation, the clumps EC3, EC4, SE1, SC8, and SC9 are likely to contain low-mass protostellar objects. This follows from the detection of CH₃CN towards these clumps together with signs of outflows, while the absence of IR emission hints at very weakly emitting embedded objects. While the presence of excess emission could not be completely justified towards EC1 due to its complex velocity structure (see Fig. 17), the CH₃CN emission towards this clump indicates the presence of a hot core.

The clumps WC2, WC3, WC4, and C1 show signs of outflows and mid-IR emission. Nevertheless, the lack of CH₃CN emission implies that the corresponding clumps do not harbour a hot core. Consequently, the observations do not confirm star formation in these clumps. It is likely that the associated IR emission corresponds to the foreground H II region, while lines that suggest an origin in outflows rather trace emission from unrelated gas layers.

As a tracer species for shocked gas, SiO is also commonly used to identify outflows of protostars (e.g. Bachiller et al. 1991; Hirano et al. 2001). We detect it in most of the possibly star-forming clumps discussed above, except for EC1, EC3, SC1, and SC9. The absence of SiO may indicate that the observed excess emission in the line wings of HCO⁺ is not related to outflow activity at these clumps. However, this does not rule out weaker embedded outflows, as the detection of CH₃CN implies that these clumps host warm cores. In addition to the star-forming clumps, SiO is also detected at WC7, EC2, SC2, SC6, SE4, and SE5. As we do not expect the presence of protostellar outflows at these clumps, their gas is likely shocked by external factors.

Recent star formation in the Lagoon Nebula was previously examined by Arias et al. (2007) and Kumar & Anandarao (2010) based on optical and near-IR observations. Both studies found a variety of pre-main sequence objects in the Lagoon Nebula and the associated cluster NGC 6530. Figure 22 shows an overview of the YSOs identified by Kumar & Anandarao (2010) in the Lagoon Nebula. We have highlighted the clumps (in yellow) where star formation is observed in our study. It can be seen that most of the clumps that show signs of active star formation are located inside regions with the highest number of YSOs. While the EC clumps are associated with nearby IRAC class 0/I sources, the remaining clumps are primarily found in the vicinity of class II sources. As the IRAC class 0/I and II sources are associated with class 0/I and II protostars (Billot et al. 2010), this implies that the protostars contained in the EC clumps may be less evolved than objects in the remaining M8 clumps.

Fig. 20 Venn diagram visualising the detected signs of star formation in the M8 clumps. Clumps shown in the overlapping regions show multiple tracers of star formation.

Fig. 21 Line profiles of the HCO+ (1−0) and (3−2) transitions at SE3 with the corresponding Gaussian fits displayed in red. Excess emission can be seen around 15 km s−1.

Fig. 22 Spitzer IRAC3 5.8 µm image of the Lagoon Nebula adapted from Kumar & Anandarao (2010). Red circles and blue triangles mark the positions of IRAC class 0/I and class II sources, respectively. Stars show the position of M8 clumps as indicated in Table 1 of Tothill et al. (2002). Clumps associated with star formation based on the analysis presented here are shown in yellow. White contours show YSO densities of 5 YSO pc−2 and 10 YSO pc−2.

7 Summary

We have presented the first spectroscopic observations towards 37 dense molecular clumps in the Lagoon Nebula since their identification by Tothill et al. (2002). Using the heterodyne receivers nFLASH230 and EMIR at APEX and the IRAM 30m telescope, we conducted pointed on-off observations in the complete frequency ranges from 210 GHz to 280 GHz and from 70 GHz to 117 GHz.

We identified a total of 346 transitions from 70 different molecular species towards the dense clumps, confirming the chemical complexity of the nebula. For every spectral line, we determined its parameters, which were then used to estimate temperatures and column densities towards every clump, applying the optically thin approximation and assuming LTE.

Combining the insights from the spectral line analysis with archival dust continuum maps, we investigated the morphology and kinematics of the region. We observe velocity gradients along the filaments, with many clumps showing multiple emission components, which may originate from kinematically related gas layers. On larger scales, we find a velocity gradient between the central EC clumps and the western WC clumps, which may be attributed to the momentum transfer from the massive stars into the surrounding gas. The intermediate LOS velocities of the clumps in the southern filaments indicate they are moving to the south.

Heating from the O- and B-type stars in M8 is responsible for radiative feedback, due to which extended PDRs are observed throughout the nebula. Apart from the widespread Spitzer 8 µm emission, we detected PDR tracers such as HCO, c-C₃H₂, CN, and C₂H towards all the clumps. Furthermore, the ionising radiation in the vicinity of M8-Main is traced by the detection of hydrogen, helium, and carbon RRLs.

We examined the clumps in M8 for signs of low- and highmass star formation, based on the presence of possible outflow signatures in the line profiles of HCO⁺, the detection of mid-IR emission, and the detection of a warm core by observing CH₃CN emission. We find that 38% of the M8 clumps are likely to host a protostellar object.

Using archival dust continuum maps from Spitzer, MSX, WISE, Herschel, AKARI, APEX, and the JCMT, we modelled SEDs for the clumps in M8 at wavelengths between 8 µm and 870 µm. Based on these SEDs, we estimated the masses, luminosities, dust temperatures, H₂ column densities, and upper limits on the virial parameter for all the clumps. We compared these clump properties to those in the ATLASGAL survey of clumps in the inner Galactic plane, finding that the M8 clumps show higher dust temperatures and L/M ratios, which are likely caused by external heating of the clumps by the surrounding O-and B-type stars.

The clumps in M8 are found to be slightly less massive than the clumps of the ATLASGAL sample at comparable distances. This suggests a possible fragmentation of the remnant gas in M8 by the radiation pressure on the massive stars. But since the ATLASGAL sample of clumps is probably incomplete for lower masses, this scenario is not clearly proven.

Acknowledgements

The authors would like to thank Nick Tothill for kindly providing the dust continuum maps of the Lagoon Nebula at 450 µm and 850 µm wavelengths. We also thank James Urquhart for an early reading of the manuscript and for valuable suggestions. Furthermore, we thank the anonymous referee, whose many useful comments have considerably improved this work. This publication is based on data acquired with the Atacama Pathfinder Experiment (APEX) under programme ID [M-0107.F-9530C-2021]. APEX is a collaboration between the Max-Planck-Institut für Radioastronomie, the European Southern Observatory, and the Onsala Space Observatory. In addition, this work is based on observations carried out under project number 141-21 with the IRAM 30m telescope. IRAM is supported by INSU/CNRS (France), MPG (Germany) and IGN (Spain). This work was partially funded by the Collaborative Research Council 956 “Conditions and impact of star formation” funded by the Deutsche Forschungsgemeinschaft (DFG).

Appendix A Observational parameters: Coordinates and setups

The on-off observations of the Lagoon Nebula were conducted towards 37 clumps identified by Tothill et al. (2002). Coordinates of the observed positions are given in Table A.1. Further inspection of the dust continuum emission and the APEX spectra revealed that the clumps WC3, SE8, and SC5 are at an offset from the observed positions given in Tothill et al. (2002). Thus, the coordinates for these clumps were adjusted for the new observations taken with the IRAM 30m telescope. The frequency setups used with each telescope are shown in Fig. A.1.

Fig. A.1 Visualisation of the frequency coverage. The spectra originate from the observations of E, which are divided by two (upper panel) or three (lower panel) and shifted on the temperature scale for illustrative purposes. Upper panel: The four setups used in the APEX observations. Lower panel: Setups (in colour) of the new observations conducted with the IRAM 30m telescope. Grey spectra show the additional frequency bands covered by the on-off observations of Tiwari et al. (2020), which are used to supplement our data at the position of clump E.

Appendix B Testing AKARI data as a substitute for Hi-GAL-PACS data in SEDs

In order to derive SEDs for the clumps in M8, data from the AKARI satellite was used to estimate flux densities between 65 µm and 160 µm. In comparison to data from the Hi-GAL survey, the AKARI images have a lower resolution, with FWHM beam widths of 63.4″ at 65 µm (N60), 77.8″ at 90 µm (WIDE-S), and 88.3″ for the 140 µm and 160 µm bands (WIDE-L and N160) (Takita et al. 2015). As a consequence, the M8 clumps with FWHM sizes lower than 38.3 are not spatially resolved. Due to this, the flux extraction method used here for the AKARI images differs from the method applied by Urquhart et al. (2018). Instead of extracting the AKARI flux in an aperture with a radius based on the source size, the flux density within an AKARI beam was extracted at the exact position of the M8 clump (see Table A.1), which accounts for most of a clump’s emission.

In order to compare the M8 clumps to the sample of ATLAS-GAL clumps located in the inner Galactic plane examined by Urquhart et al. (2018), it is necessary to identify possible differences in the results of the two flux retrieval methods. For this, the SEDs of a sample of clumps in the NGC 6334 region were computed with the AKARI images as a replacement for the Hi-GAL PACS data. The clumps in this region provide ideal targets for testing the usability of AKARI data for our cause, as NGC 6334 is located at a distance of 1.34 pc (Urquhart et al. 2018), which is similar to the distance of the Lagoon Nebula. This ensures that the method is tested on objects with similar properties as the M8 clumps.

From the clumps covered by both AKARI and Hi-GAL, a selection was made considering only bright sources that are well separated from other objects in the region. While AKARI conducted an all-sky survey that covers 99% of the sky, the data have a large image defect covering the brightest sources of NGC 6334, limiting the sample size of sources suitable for this test to 16.

The flux densities of the NGC 6334 clumps were extracted analogously to the M8 clumps, as explained in Sect. 2.4. Since there is no available SCUBA data from the JCMT that covers the whole NGC 6334 region, the 870 µm flux density was derived purely based on the ATLASGAL data. Because the 870 µm data of the JCMT and ATLASGAL are in good agreement for the M8 clumps, the SEDs should be unaffected by this change.

Fig. B.1 Upper panel: Comparison of the flux density extracted inside a 40.9″ beam for the AKARI and PACS images of selected clumps in NGC 6334. Lower panel: Luminosities of these clumps as derived from fitting SEDs with the AKARI data as compared to the luminosities derived by Urquhart et al. (2018) for the same clumps. The fitted linear relation and the Pearson correlation coefficient R and the P value are given in the upper-left corner.

Firstly, we verified the calibration of the AKARI data in order to exclude the possibility of a systematic under- or overes-timation of the derived flux densities. This is done by extracting the flux densities measured by AKARI-FIS and Hi-GAL-PACS instruments according to the method explained in Sect. 2.4. As shown in the upper panel of Fig. B.1, the flux densities extracted from the Hi-GAL datasets at 70 µm and 160 µm are in good agreement with the flux densities derived from AKARI images. The slightly lower flux density values of AKARI in the 70 µm band are likely due to the different wavelengths covered by the respective receivers. While the AKARI band is centred on 65 µm, the PACS band is centred on 70 µm. This difference is also seen when comparing the spectral response functions of both instruments shown in Fig. 1 of Shirahata et al. (2009) and Fig. 6 of Poglitsch et al. (2010).

To test the influence of the different extraction methods applied for the AKARI bands, the SED fitting procedure was applied to all sample sources, and the physical parameters were derived analogously to Urquhart et al. (2018) using the equations given by Schuller et al. (2009). The lower panel of Fig. B.1 shows a comparison between the computed luminosities and the values derived by Urquhart et al. (2018) for the same sources. The luminosities derived using both methods are in good agreement for the examined range between 10 and 10⁴ M_⊙ .A linear fit leads to a slope slightly below 1, indicating that lower luminosities might be overestimated, while higher luminosities could be underestimated. This can be explained by the constant extraction size of one AKARI beam size, which for larger and brighter sources may not cover the full source emission. Thus, when dealing with much larger sources, it might be beneficial to extract the AKARI emission with a defined aperture analogous to the other bands. In contrast, accurate luminosity values for weak sources cannot be assured if multiple sources are located within the AKARI beam.

As shown in the upper panel of Fig. B.2, the applied method introduces a slight overestimation of the clumps masses, which is similar across the considered clump masses between 3 and 300 M_⊙. Nevertheless, there is a good correlation between the derived masses, and the values computed by Urquhart et al. (2018). On the contrary, The derived H₂ column densities do not show a clear correlation to the column densities derived by Urquhart et al. (2018, see the lower panel of our Fig. B.2). While the column density is overestimated for intermediate to larger clumps, the column density is underestimated for smaller clumps. Both quantities are beam-specific quantities, and might differ due to the different methods used to derive the source sizes. While we used the source FWHM based on the ATLAS-GAL measurements as size estimate, Urquhart et al. (2018) based their size estimate on the extent of the 350 µm emission inside the extraction aperture.

Finally, the dust temperatures between 15 K and 30 K derived by both methods are consistent with one another, independent of the respective source sizes (see Fig. B.3). This is expected, as the cold gas component is usually broadly extended, such that it fills the beam in all considered cases.

This analysis demonstrates that AKARI has the potential to be used for deriving SEDs of clumps away from the Galactic plane, which were not covered by the PACS instrument of Her-schel. We note that the lower resolutions of the AKARI images, as compared to Hi-GAL, limits the use of the data in crowded regions, as the large beam may capture emission from neighbouring sources. While it is possible to correct for the influence of additional sources inside the AKARI beam as described in Sect. 2.4, this method may become inaccurate for regions with a higher number of sources than in M8. We note that for large sources it will be beneficial to extract the AKARI flux density inside an aperture that is based on the source size, as applied for the data of other surveys. This would prevent emission in the outer areas of these extended sources from being excluded.

Fig. B.2 Comparison of the computed masses (upper panel) and column densities (lower panel) for selected clumps in NGC 6334 to the values derived by Urquhart et al. (2018) for the same clumps. The Pearson correlation coefficient, R, and the P value are given in the lower-right and upper-left corner, respectively, where also the fitted linear relations are indicated.

Fig. B.3 Comparison of the computed dust temperatures for selected clumps in NGC 6334 to the values derived by Urquhart et al. (2018) for the same clumps. The fitted linear relation and Pearson correlation coefficient R and the P value are given in the upper-left corner.

Appendix C Physical properties of the M8 clumps

Sect. 3 describes the procedure we used for modelling the SEDs of the M8 clumps. This appendix provides an overview of the physical properties obtained from the modelling in Table C.1. The fitted SEDs for all clumps in M8 are shown in Fig. C.1-C.5.

Fig. C.1 Fit to the SEDs of HG, E, WC1, SC9, and SW1. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered, as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

Fig. C.2 Fits to the SEDs of the WC2–9 clumps. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

Fig. C.3 Fits to the SED of the SE1-8 clumps. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

Fig. C.4 Fits to the SED of the EC1-5 clumps and the clumps C1-3. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

Fig. C.5 Fits to the SED of the SC1-8 clumps. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

Appendix D Identified transitions and line parameters

This appendix provides a full overview of molecular rotational transitions observed in the Lagoon Nebula. As an example of the analysed data, spectrumextracts ofSE1 are shown in Fig. D.1.

Table D.1 lists the name of each species, the quantum numbers or identifiers of the transition, the associated frequency in MH_z, the upper-level energy, E_up, in K, the Einstein A coefficient, A_ij, in s⁻¹ and the degeneracy g_up according to the information provided by the databases (last column).

The second column describes the transition according to the values stated by the respective database (usually J or J_F for linear molecules, J_K for symmetric tops and $J_{K_a,K_c}$ for asymmetric tops). The third subscript number for NH₂D describes the symmetry state, where the anti-symmetric state is labelled 1 and the symmetric state is labelled 0. An exception is methanol, for which the symmetry state and, as subscript, k for the E-type and K for the A-type species are given. Lines from NO, NS, CN, C₂H, and C₂D are described as N_J,F, where the additional subscript of NO and NS states the parity. HCO and c-C₃H lines are described as $N_{K_a,K_c,J,F}$ in order to fully identify the transitions. The descriptions of ¹³CN and C¹³CH lines additionally contain the intermediate quantum numbers as $N_{F_1,F_2,F}$ and N_JF₁,F, respectively.

Table D.2 gives a detailed overview of the derived line properties for detected emission from rotational transitions towards all M8 clumps. The full versions of Tables D.1 and D.2 are available at the CDS.

Fig. D.1 Spectrum extracts of the data obtained with the IRAM 30m telescope (upper panel) and APEX (lower panel) at SE1.

Appendix E Radio recombination lines detected in M8

Table E.1 provides an overview of the computed Gaussian profiles for the combined spectra of the detected RRLs.

Appendix F Temperatures and column densities of acetonitrile and methyl acetylene and formaldehyde

As described in Sects. 5.2 and 5.3, we derived the kinetic temperatures, H₂ volume densities and column densities of formaldehyde with an MCMC method, and the properties of acetonitrile and methyl acetylene using rotation diagrams. This appendix gives a short overview of the concepts used in rotation diagrams and provides the resulting physical quantities in Table F.1 alongside the corresponding rotation diagrams in Figs. F.1-F.15 and MCMC posterior density distributions in Figs. F.16-F.38.

When multiple transitions of the same species are detected, it is possible to derive rotation temperatures T_rot and column densities N_total of the respective species. In the frame of an energy level system, the Boltzmann distribution can be written as $$\frac{N_{\text{u}}}{N_{\text{total }}}=\frac{g_{\text{u}}}{Q\left(T_{\text{rot }}\right)}\exp \left(-\frac{E_{\text{u}}}{k_{\text{B}}{T}_{\text{rot }}}\right),$$(F.1)

with the Boltzmann constant, k_B, the column density of a respective upper energy level, N_u, and the degeneracy, g_u, and energy, E_u, of the same level. Q(T_rot) is the rotational partition function that describes the sum over all rotational energy levels in a molecule for a given rotation temperature (Mangum & Shirley 2015). N_u can be computed in the optically thin limit from the measured intensity I of a transition between the upper energy level (u) and the lower energy level (l) according to (Goldsmith & Langer 1999) $$N_{\text{u}}^{\text{thin }}=\frac{8\pi k_{\text{B}}{v}^2}{h{c}^3{A}_{\text{ul}}}I\text{, }$$(F.2)

with the speed of light, c, the Planck constant, h, the frequency, ν, of a respective transition and the corresponding spontaneous emission Einstein (A) coefficient, A_ul.

By rearranging Eq. F.1 as $$\ln \left(\frac{N_{\text{u}}}{g_{\text{u}}}\right)=-\frac{1}{k_{\text{B}}{T}_{\text{rot }}}{E}_{\text{u}}+\ln \left(\frac{N_{\text{total }}}{Q\left(T_{\text{rot }}\right)}\right),$$(F.3)

it can be seen that the logarithm of the column density of an upper energy level N_u is proportional to the energy of the respective level. This relation can be used to perform a linear fit on the known values of ln(N_u/g_u) and E_u, in order to obtain the rotation temperature T_rot from the slope and the column density from the ordinate interception.

For this analysis, transitions of the same acetonitrile or methyl acetylene multiplets were fit simultaneously with multiple Gaussian components of the same width and velocity shift to account for line blending. According to Equation F.2, the values of N_u and therefore of ln(N_u/g_u) depend on the velocity-integrated line intensity, I. Their uncertainty was therefore derived by propagating the uncertainty of I, which is composed of the fit uncertainty for the line profiles with the MINIMIZE function of CLASS, and an additional 10% uncertainty to account for the calibration uncertainty of the data.

Fig. F.1 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at HG. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.2 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at WCl. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.3 Rotation diagram of CH3C2H at WC2. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.4 Rotation diagram of CH3CN at EC1. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.5 Rotation diagram of CH3CN at EC3. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.6 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at EC4. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.7 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at E. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.8 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE1. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.9 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE3. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.10 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE7. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.11 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE8. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.12 Rotation diagram of CH3CN at SC1. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.13 Rotation diagrams of CH3C2H at SC2. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.14 Rotation diagram of CH3CN at SC8. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.15 Rotation diagram of CH3CN at SC9. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.

Fig. F.16 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at HG.

Fig. F.17 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at WCl.

Fig. F.18 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at WC4.

Fig. F.19 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at EC1.

Fig. F.20 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC2.

Fig. F.21 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC3.

Fig. F.22 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC4.

Fig. F.23 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC5.

Fig. F.24 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at E.

Fig. F.25 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE1.

Fig. F.26 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE2.

Fig. F.27 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at SE3.

Fig. F.28 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE4.

Fig. F.29 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE5.

Fig. F.30 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE6.

Fig. F.31 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE7.

Fig. F.32 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE8.

Fig. F.33 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at SC1

Fig. F.34 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC2.

Fig. F.35 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at SC3.

Fig. F.36 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC4.

Fig. F.37 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC6.

Fig. F.38 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC8.

Fig. F.39 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC9.

Appendix G Column densities of detected species at all clumps in M8

Table G.1 provides column density values or upper limits for all species detected in M8 for all clumps, as described in Sect. 5.4. The full version of this table is available at the CDS.

References

All Tables

All Figures

	Fig. 1 Image of the Lagoon Nebula observed using B, V, and R broadband filters. Blue and green stars respectively mark the positions of O- and early B-type stars associated with the M8 region, according to Table 1 of Wright et al. (2019). The dashed black box marks the molecular gas examined by Tothill et al. (2002). Data credits: Felipe Mac Auliffe (https://clusteroneobservatory.com).
In the text

	Fig. 2 SCUBA 850 µm image of the Lagoon Nebula adapted from Tothill et al. (2002). Markers show the position of the molecular clumps in the Lagoon Nebula. The blue and green rectangles respectively indicate the regions examined by Tiwari et al. (2018) and Tiwari et al. (2020).
In the text

	Fig. 3 Dust continuum image of M8 at 24 µm from MIPSGAL. The ATLAS-GAL 870 µm emission is shown as contours. Blue and green stars show the positions of O- and early B-type stars in this region (Wright et al. 2019), respectively. The image is saturated near the H II region M8-Main.
In the text

	Fig. 4 8 µm dust continuum image obtained from the GLIMPSE survey. Blue and green stars respectively mark the positions of present O- and B-type stars (Wright et al. 2019). Contours of the ATLASGAL 870 µm emission are shown in grey.
In the text

	Fig. 5 8 µm dust continuum image of the southern clumps in the Lagoon Nebula obtained from the GLIMPSE survey. Contours of the ATLASGAL 870 µm emission are shown in grey.
In the text

	Fig. 6 Dust temperatures (upper panel) and masses (lower panel) of clumps in the Lagoon Nebula. The circle size corresponds to the aperture size used for flux extraction in the analysis. For the temperatures, the smaller inner circles visualise the derived temperatures of the hot components. The grey-scale background image shows the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2).
In the text

	Fig. 7 Distribution of species across the M8 clumps. Different colours differentiate between S-, O-, and N-bearing species, deuterated species, carbon chains, complex organic molecules (COMs), and other species, as given in Table 1.
In the text

Fig. 8 Intensity and LOS velocities of the H 39−44α transitions observed at the positions of the clumps in the Lagoon Nebula. The fillings of the coloured dots indicate the intensities and the borders describe the velocities. Blue and green stars respectively mark the positions of present O- and B-type stars. The grey-scale background image shows the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2).

In the text

	Fig. 9 CII 157 µm transition (blue) and stacked spectra of the HeNα and CNα RRLs with N=39−42 and 44 (black) towards clump HG (Her 36). The velocity scale of the RRLs is computed based on the rest frequency of the CNα lines. The Gaussians fitted to the RRLs are displayed in red. The temperature scale of the 158 µm CII fine structure transition is divided by 2000 to compare it with the RRLs.
In the text

	Fig. 10 Typical line profiles of the discussed maser transitions. Upper panel: methanol maser emission of the 84.5 GHz and 95.2 GHz transitions at clump E, respectively shown in blue and red. Lower panel: maser emission of the 218.4 GHz transition at the position of SE7.
In the text

Fig. 11 LOS velocities of clumps in the Lagoon Nebula. Upper panel: positions of the individual clumps (circles), with the grey-scale image showing the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2). The circle size corresponds to the beam size of the on-off observations with APEX and the IRAM 30m telescope. Bi-coloured circles visualise distinct velocity components observed in the same beam. Lower panel: LOS velocity of the M8 clumps as a function of RA separation from the cloud centre at RA=18h04m 06s. Multiple velocity components inside one beam are shown as individual points.

In the text

	Fig. 12 Temperatures derived from the rotational transitions of paraformaldehyde, acetonitrile, and methyl acetylene as a function of dust temperatures at the clumps in M8. The legend in the upper-left corner provides the Pearson correlation coefficient R and the P values for the respective samples.
In the text

Fig. 13 Temperatures in clumps in the Lagoon Nebula derived from different molecular species. Coloured circles show the kinetic temperatures from para-formaldehyde (annuli) and the rotational temperatures of acetoni-trile (left inner circle) and methyl acetylene (right inner circle). The size of the inner circles corresponds to the approximate beam size of the observations. The grey-scale background image shows the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2).

In the text

	Fig. 14 Column densities of species detected in at least ten M8 clumps. The colour scale gives the relative abundance of a respective species to the column density of C18O at the same position. White cells indicate the non-detection of a species in the associated clump.
In the text

Fig. 15 Cumulative distributions of the clump radii derived by Tothill et al. (2002, upper panel) and masses (lower panel) in M8 and for the full and distance-limited sample of ATLASGAL clumps. The M8 clump masses shown in orange colour are corrected based on the results of Appendix B. The dotted orange distribution only contains M8 clumps that have a nearby counterpart in the ATLASGAL sample. The p-values of KS tests between selected samples are provided as grey text.

In the text

Fig. 16 Cumulative distributions of the clump luminosities (upper panel), the L/M ratios of the respective clumps (middle panel), and clump dust temperatures (lower panel) in M8 and for sources of the full and distance-limited ATLASGAL sample. The M8 distributions shown in the orange colour are corrected based on the results of Appendix B. The dotted orange distribution only contains M8 clumps that have a nearby counterpart in the ATLASGAL sample. The p-values of KS tests between selected samples are provided as grey text.

In the text

	Fig. 17 Line profiles of the C18 O (2−1) transition for clumps located in the EC ridge.
In the text

Fig. 18 HCO column densities at the M8 clumps. The circle markers correspond to the positions of the individual clumps, with the grey-scale image showing the JCMT SCUBA 870 µm dust continuum flux (see Fig. 2). The circle size corresponds to the beam size of the on-off observations with APEX and the IRAM 30m telescope. Two-coloured circles visualise distinct velocity components observed in the same beam.

In the text

	Fig. 19 Abundance of HCO (left panel) and c-C3 H2 (right panel) as a function of H2 column density. The Pearson correlation coefficient R and the P value for the M8 clumps are given in the upper-right corner.
In the text

	Fig. 20 Venn diagram visualising the detected signs of star formation in the M8 clumps. Clumps shown in the overlapping regions show multiple tracers of star formation.
In the text

	Fig. 21 Line profiles of the HCO+ (1−0) and (3−2) transitions at SE3 with the corresponding Gaussian fits displayed in red. Excess emission can be seen around 15 km s−1.
In the text

Fig. 22 Spitzer IRAC3 5.8 µm image of the Lagoon Nebula adapted from Kumar & Anandarao (2010). Red circles and blue triangles mark the positions of IRAC class 0/I and class II sources, respectively. Stars show the position of M8 clumps as indicated in Table 1 of Tothill et al. (2002). Clumps associated with star formation based on the analysis presented here are shown in yellow. White contours show YSO densities of 5 YSO pc−2 and 10 YSO pc−2.

In the text

Fig. A.1 Visualisation of the frequency coverage. The spectra originate from the observations of E, which are divided by two (upper panel) or three (lower panel) and shifted on the temperature scale for illustrative purposes. Upper panel: The four setups used in the APEX observations. Lower panel: Setups (in colour) of the new observations conducted with the IRAM 30m telescope. Grey spectra show the additional frequency bands covered by the on-off observations of Tiwari et al. (2020), which are used to supplement our data at the position of clump E.

In the text

Fig. B.1 Upper panel: Comparison of the flux density extracted inside a 40.9″ beam for the AKARI and PACS images of selected clumps in NGC 6334. Lower panel: Luminosities of these clumps as derived from fitting SEDs with the AKARI data as compared to the luminosities derived by Urquhart et al. (2018) for the same clumps. The fitted linear relation and the Pearson correlation coefficient R and the P value are given in the upper-left corner.

In the text

	Fig. B.2 Comparison of the computed masses (upper panel) and column densities (lower panel) for selected clumps in NGC 6334 to the values derived by Urquhart et al. (2018) for the same clumps. The Pearson correlation coefficient, R, and the P value are given in the lower-right and upper-left corner, respectively, where also the fitted linear relations are indicated.
In the text

	Fig. B.3 Comparison of the computed dust temperatures for selected clumps in NGC 6334 to the values derived by Urquhart et al. (2018) for the same clumps. The fitted linear relation and Pearson correlation coefficient R and the P value are given in the upper-left corner.
In the text

Fig. C.1 Fit to the SEDs of HG, E, WC1, SC9, and SW1. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered, as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

In the text

Fig. C.2 Fits to the SEDs of the WC2–9 clumps. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

In the text

Fig. C.3 Fits to the SED of the SE1-8 clumps. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

In the text

Fig. C.4 Fits to the SED of the EC1-5 clumps and the clumps C1-3. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

In the text

Fig. C.5 Fits to the SED of the SC1-8 clumps. Single-component fits only consider flux at wavelengths longer than 65 µm to avoid a contribution from unrelated diffuse warm gas to the clumps’ SEDs. SCUBA 450 µm emission was not considered as we found it to systematically underestimate the flux of all clumps. Flux densities at 8 µm and 65 µm are considered as upper limits due to the possible contributions of PAHs and very small grains.

In the text

	Fig. D.1 Spectrum extracts of the data obtained with the IRAM 30m telescope (upper panel) and APEX (lower panel) at SE1.
In the text

	Fig. F.1 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at HG. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.2 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at WCl. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.3 Rotation diagram of CH3C2H at WC2. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.4 Rotation diagram of CH3CN at EC1. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.5 Rotation diagram of CH3CN at EC3. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.6 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at EC4. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.7 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at E. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.8 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE1. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.9 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE3. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.10 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE7. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.11 Rotation diagrams of CH3C2H (upper panel) and CH3CN (lower panel) at SE8. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.12 Rotation diagram of CH3CN at SC1. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.13 Rotation diagrams of CH3C2H at SC2. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.14 Rotation diagram of CH3CN at SC8. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.15 Rotation diagram of CH3CN at SC9. Black data points are inferred from the line intensities, the red line shows the least squares linear fit.
In the text

	Fig. F.16 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at HG.
In the text

	Fig. F.17 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at WCl.
In the text

	Fig. F.18 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at WC4.
In the text

	Fig. F.19 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at EC1.
In the text

	Fig. F.20 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC2.
In the text

	Fig. F.21 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC3.
In the text

	Fig. F.22 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC4.
In the text

	Fig. F.23 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at EC5.
In the text

	Fig. F.24 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at E.
In the text

	Fig. F.25 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE1.
In the text

	Fig. F.26 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE2.
In the text

	Fig. F.27 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at SE3.
In the text

	Fig. F.28 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE4.
In the text

	Fig. F.29 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE5.
In the text

	Fig. F.30 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE6.
In the text

	Fig. F.31 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE7.
In the text

	Fig. F.32 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SE8.
In the text

	Fig. F.33 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at SC1
In the text

	Fig. F.34 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC2.
In the text

	Fig. F.35 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin(NH2CO/dv)at SC3.
In the text

	Fig. F.36 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC4.
In the text

	Fig. F.37 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC6.
In the text

	Fig. F.38 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC8.
In the text

	Fig. F.39 Posterior probability distributions of the H2 volume density $\left(n_{H_2}\right)$, kinetic temperature (Tkin), and H2CO column density per velocity bin (NH2CO/dv) at SC9.
In the text