Near-infrared echelle spectroscopy of Class I protostars: molecular hydrogen emission-line (MHEL) regions revealed

Infrared echelle spectra are used to trace dynamic activity in the immediate vicinity of Class I outﬂow sources. The H 2 and Br g observations presented here trace different components of these emission-line regions; indeed, they are thought to trace the orthogonal processes of outﬂow and infall respectively. of within a hundred H 2 as ‘molecular hydrogen emission-line’ regions, or MHELs, and compare their properties to those of forbidden emission-line regions (FELs) observed in classical T Tauri and some Herbig AeBe stars. Like the FELs, both low- and high-velocity components (LVCs and HVCs) are observed in H 2 , with blueshifted velocities of the order of 5–20 and 50–150 km s 2 1 respectively. LVCs are more common than HVCs in MHEL regions, and like their FEL counterparts, the are spatially further offset from the exciting source in each case. The MHEL regions – which are in all cases preferentially blueshifted – are assumed to be associated with the base of each outﬂow. as well as towards the T Tauri star AS 353A. These lines are all broad and symmetric, the line peaks being blueshifted by , 30 km s 2 1 . The proﬁles are typical of the permitted hydrogen line proﬁles observed in many T Tauri stars, and probably derive from magnetospheric accretion ﬂows. We do not observe redshifted absorption features (inverse P-Cygni proﬁles) in any of the sources, however. Nor do we detect a dependence on linewidth with inclination angle of the system to the line of sight, as is predicted by such accretion models. No Br g is detected in the extended ﬂow lobes. Instead, the emission is conﬁned to the source and is spatially unresolved along each ﬂow axis.

Class II outflow sources is interpreted in terms of jets or outflows that are observed close to their driving sources (Hirth, Mundt & Solf 1994a, Hirth et al. 1994bHartigan et al. 1995;Corcoran & Ray 1998). These forbidden emission-line regions, or FELs, exhibit complex and/or multiple velocity components; the emission is also usually extended along the flow axis.
In this paper we focus on near-IR observations of more deeplyembedded, Class I outflow sources. From published low-resolution spectroscopic surveys, many of our targets are known to exhibit H I and H 2 lines in their spectra (Carr 1990;Greene & Lada 1996a;Reipurth & Aspin 1997). We discuss infrared echelle spectroscopy of nine outflow regions. The molecular hydrogen 1-0S(1) and H I Brg observations obtained trace very different excitation conditions; the H 2 traces dense, molecular gas of relatively low excitation ðn H2 $ 10 5 cm 23 ; T , 2000 KÞ while the Brg emission traces hot, hydrogen-recombination zones. In almost all of the regions observed, H 2 line emission is detected directly towards the outflow source, as well as in the more extended flow lobes. We associate the former with what we shall henceforth refer to as 'molecular hydrogen emission-line' (MHEL) regions. This molecular line emission region is generally confined to within a few arcsec (five to ten thousand au at the typical distances of our sources) of the outflow source. Brg emission is observed only towards a few of the outflow sources and is spatially unresolved along the outflow axes. We associate the H 2 emission with outflow activity and, akin to studies of T Tauri stars (e.g. Folha & Emerson 2000), the Brg with accretion.
The targets are discussed individually in Section 3. In Section 4 we consider the origins of the observed emission features in terms of published infall and outflow models, and look for correlations between H I and H 2 line luminosities, other outflow parameters and the ages and luminosities of the driving sources.

O B S E RVAT I O N S
High-resolution, near-IR echelle spectra were obtained at the UK Infrared Telescope (UKIRT) using the cooled grating spectrometer CGS 4 (see Table 1). The instrument utilizes a 256 Â 256 pixel InSb array and has a pixel scale of 0:41 Â 0:90 arcsec (0.41 arcsec in the dispersion direction); with a slit 2 pixels wide, the velocity resolution was ,15 km s 21 (although over-sampled spectra were obtained by physically shifting the array by 1/2 pixel, so that two detector positions were observed per resolution element). The instrumental profile in the dispersion direction, as measured from Gaussian fits to sky lines, was 17:7ð^2:5Þ km s 21 and 18:0ð^1:5Þ km s 21 in the H 2 and Brg spectra respectively. Spectra were obtained in both H 2 1-0S(1) (at l vac ¼ 2:1218334mm; Bragg, Smith & Brault 1982) and H I Brg (at l vac ¼ 2:166167mmÞ. For each target the slit orientation was set to that of the major jet or outflow axis (see Table 1), determined from images of the targets found in the literature. Object -sky -sky -object sequences were repeated a number of times on each source to build up signal-tonoise, the sky position being typically a few arcmin away from the source. Each spectral image was bias subtracted and flat-fielded (using a blackbody source in the calibration unit) before the data were combined to produce one 'reduced group' spectral image per target.
The data were wavelength calibrated using sky lines (for H 2 , four bright lines occupy the passband of the echelle at 2.122 mm; Davis et al. 2000) or a combination of sky and argon arc lines (both were needed to calibrate the Brg data): we used the OH calibrations of Oliva & Origlia (1992). The IRAF tasks used to accomplish this (IDENTIFY, RE-IDENTIFY, FITCOORDS and TRANSFORM) also corrects for distortion along the columns in each image (i.e. along arc or sky lines). The relative velocity calibration across each spectral image, measured from Gaussian fits to sky lines in 'velocity-calibrated and distortion-corrected' raw frames, is estimated to be accurate to #5 km s 21 . Instrument flexure over the duration of the observations could, however, introduce additional uncertainties in the absolute velocity calibration, i.e. by shifting the individual frames with respect to the wavelength reference used to calibrate the reduced group spectral image (Lumsden & Hoare 1999). By comparing the positions of sky lines in a number of raw frames we found that this effect was small; indeed, the narrowness of the H 2 emission features observed at some locations in the final, reduced data (as compared to the instrumental profile width, which is measured from just one frame), confirms this finding. Nevertheless, flexure could still result in an additional broadening of H 2 lines and a shift in the line centre by a few km s 21 . We therefore conclude that the overall velocity calibration is accurate to better than 10 km s 21 (the Brg calibration being slightly worse than the H 2 ), while perceived velocity shifts between adjacent spectra observed along the same slit will be considerably more accurate, within 1 km s 21 . A-and G-type bright standards were also observed at similar airmasses to each target to flux calibrate the extracted spectra and to correct for telluric absorption.

R E S U LT S
Echelle spectra of each outflow region were obtained in both H 2 1-0S(1) and H I Brg emission (Table 1); the slit position angles (p.a.) were set to that of each large-scale outflow, as indicated by previous published observations of each flow. H 2 positionvelocity (P-V) diagrams are shown in Fig. 1. These show not only line emission coincident with the outflow sources, but also complex velocity structure in many of the extended outflow lobes and Herbig -Haro (HH) objects.
The Brg observations are shown in Fig. 2. In all regions where Brg was detected, the emission was confined to the outflow source itself. Gaussian fits to cuts made perpendicular to the stellar continua in each Brg P-V plot show that the emission is unresolved along the slit/flow axis, to within ,2 arcsec of the outflow source. Consequently, we present H I spectra extracted from each spectral image rather than P -V plots. For comparison, the equivalent H 2 spectra are also shown.
Below we discuss the results pertaining to each source separately.

SVS 13 (HH 7-11)
The HH 7-11 outflow in NGC 1333 ðd , 220 pcÞ has been observed at near-IR wavelengths by a number of groups (e.g. Garden, Russell & Burton 1990;Carr 1993;Gredel 1996;Everett 1997;Chrysostomou et al. 2000;Davis et al. 2000). The HH features in the south-eastern, blueshifted lobe of this bipolar flow are distributed along the walls of a cavity, the leading edge of which is capped by the HH 7 bow shock. HH 7-11 is also associated with a high-velocity CO outflow (Bachiller & Cernicharo 1990). Although the identification of SVS 13 as the powering source of the flow was recently questioned as a result of the location of other, nearby radio sources (Rodríguez, Anglada & Curiel 1997) this young stellar object (YSO) is known to have undergone an outburst in recent years (Eislöffel et al. 1991;Aspin & Sandell 1994) and, from the data presented in Figs 1 and 2, is clearly associated with complex line emission. Recent highresolution, interferometric CO maps of the HH 7-11 region also indicate that SVS 13 is the powering source of the outflow (Bachiller et al. 2000).
The P-V plot in Fig. 1(a) illustrates the velocity structure of the H 2 emission along the HH 7-11 outflow axis. SVS 13 is situated at the bottom of this plot, with H 2 emission detected out as far as HH 7 (the bright emission at the top of the P-V diagram). The H 2 emission along the flow axis has been discussed by Davis et al. (2000) and previously by Carr (1993); here we consider only the emission observed directly towards the outflow source.
H 2 and Brg spectra are presented in Fig. 2. The spectra each represent the average of three rows centred on SVS 13. Both lines are blueshifted with respect to the systemic local standard of rest (LSR) velocity of 18 km s 21 (Bachiller & Cernicharo 1990). The main H 2 peak is at 220ð^5Þ km s 21 ; the Brg profile peaks at around 225ð^15Þ km s 21 . The Brg profile is clearly much wider than the H 2 profile (as is the case in all YSOs observed here); the full width half-maximum (FWHM) of the former measures 180ð^10Þ km s 21 . Yet the H 2 profile is more complex, exhibiting a secondary blueshifted peak at 289ð^5Þ km s 21 as well as extensive, non-Gaussian blue and redshifted line wings. The red wing is fainter than the blue wing/peak, presumably because of extinction by a dense equatorial disc that obscures our view of the innermost regions of the receeding jet lobe. The broad, blue H 2 wing extends out to at least 2110 km s 21 . The two peaks in the H 2 profile, at 220 and 290 km s 21 , are probably the molecular counterparts of the low-and high-velocity jet components identified in optical spectra of classical T Tauri stars and some Herbig Ae/Be stars (Hirth et al. 1994a,b;Corcoran & Ray 1998).

B5-IRS 1 (HH 366)
Barnard 5 is situated at the eastern end of the Perseus cloud complex ðd , 350 pcÞ. The embedded YSO B5-IRS 1 drives an extensive east -west molecular outflow (Goldsmith, Langer & Wilson 1986;Fuller et al. 1991;Yu, Billawalla & Bally 1999) that excites two groups of HH objects, HH 366E and 366W, situated some 30 arcmin (,1 parsec) to the east-north-east (ENE) and westsouth-west (WSW) of IRS 1 (Bally, Devine & Alten 1996). Interferometric millimetre-wave observations trace a toroidal dust/ gas disc structure as well as the inner regions of a bipolar CO outflow cavity (which has a notably broad opening angle of almost 908; Langer, Velusamy & Xie 1996). Both flow lobes are also traced by extensive H 2 emission; within 30 arcsec of the source Yu et al. (1999) observe emission in both lobes, with a bright H 2 peak coincident with the optical knot HH 366 E5 in the eastern, blueshifted flow.
In our echelle observations H 2 emission is detected towards the outflow source itself and towards HH 366 E5 in the eastern flow lobe (Fig. 1b). The latter is strongly blueshifted with a radial velocity of ,265 km s 21 . A very faint patch of redshifted emission (at V LSR , 150 km s 21 Þ, found at roughly the same distance from the source as E5, though in the western lobe, was detected at the 2-3s level in our data. Here the CGS 4 slit passed just north of an H 2 knot WSW of IRS 1. This knot is evident in the H 2 data of Yu et al. (1999); indeed, their data show spatial and kinematic symmetry in the B5-IRS1 bipolar flow close to the source.
No Brg emission was detected towards B5-IRS 1 (Fig. 2). The H 2 profile observed towards the source is somewhat asymmetrical, exhibiting a weak blueshifted wing that extends out to at least 240 km s 21 . The line profile is centred at ,4ð^5Þ km s 21 , close to the systemic velocity of ,10 km s 21 (Yu et al. 1999), though it extends over almost 100 km s 21 full width zero intensity (FWZI).

IRAS 0423912436 (HH 300)
The low-luminosity Class I protostar IRAS 0423912436 has a rich near-IR spectrum (Greene & Lada 1996b). The source, situated in the B18 cloud in Taurus ðd , 140 pcÞ, probably excites a group of extensive optical HH bow shocks (HH 300: Reipurth, Bally & Devine 1997) found about 30 arcmin to the south-west, as well as a more compact conical HH feature just 30 arcsec to the north-east of the source. Lucas & Roche (1998) present near-IR imaging and polarimetric observations of the source and associated conical reflection nebula, which opens in a direction towards the northeastern HH knot 300D.
Although H 2 emission was only detected close to the source, the emission peak (superimposed on the stellar continuum in Fig. 1c) is extended along the southwestern flow lobe (traced faintly in Fig. 1c at negative slit offsets). The H 2 profile in Fig. 2 is considerably narrower than the Brg profile observed towards IRAS 0423912436, the FWZI of the latter measuring ,470 km s 21 . Both lines peak within a few km s 21 of the systemic velocity of ,8 km s 21 , however.

L 1551-IRS5
The proximity of this source to the earth ðd , 140 pcÞ, the welldefined bipolar CO outflow and the array of optical and near-IR shock features associated with it have attracted considerable attention from observers (e.g. Mundt & Fried 1983;Stocke et al. 1988;Moriarty-Schieven & Snell 1988;Davis et al. 1995). The south-western, blueshifted flow lobe consists of a compact collimated flow and an extensive wind-swept cavity. IRS 5 is in fact a binary system that appears to drive two jets (Fridlund & Liseau 1998;Hartigan et al. 2000), a brighter (at optical wavelengths) jet that terminates at a bow shock some 10-12 arcsec from the source, and a fainter, though more extended, jet that is traced twice as far from the source. The wind-swept cavity, on the other hand, extends over 10 arcmin (0.5 pc).
In the H 2 P-V plot in Fig. 1(d) we see line emission coincident with IRS 5 as well as in both the blue and redshifted jet lobes. In the south-western blue lobe (negative offsets in Fig. 1d) two velocity components are observed, a stationary component that is detected along the full length of the bright optical jet, and a steadilyincreasing velocity component that reaches about half-way along the jet. The latter is suggestive of steady acceleration along the flow, the gas reaching a radial velocity of almost 260 km s 21 . This is in contrast to optical kinematic studies which indicate decreasing jet velocities with distance from the source (Hartigan et al. 2000). In Ha, for example, the velocity of the peak of the observed emission profiles decreases from ,2290 km s 21 at the base of the bright optical jet to ,2120 km s 21 at the end of the optical jet, about 20 arcsec from the driving source (Stocke et al. 1988). This suggests a different origin for the optical and H 2 emissions; the latter could be associated with the second, slower and much fainter optical jet that runs parallel with the brighter, much faster optical jet. The Fabry -Perot observations of Hartigan et al. (2000) indicate radial velocities along the slower optical jet of about 260 km s 21 , similar to the velocities measured in H 2 .  , 6, 9, 12, 15, 20, 30, 50 and 100Â the standard deviation to the mean background signal in each image. y-axis offsets are in arcsec along the slit. The exciting source in each system is centred at 0-arcsec offset. The vertical dotted lines mark the systemic LSR velocity in each region. Arrows in each plot show the slit orientation on the sky. Thus for SVS 13 -increasing y-axis offsets are to the south-east; for B5-IRS1 -increasing offsets are to the east; for IRAS 0423912436 -increasing offsets are to the north-east. For L 1551-IRS5increasing y-axis offsets are to the north-east; for HH 34-IRS -increasing offsets are to the south. For HH 72-IRS -increasing y-axis offsets are to the east. For GGD 27 -increasing y-axis offsets are to the south; for AS 353A -increasing offsets are to the east; for HH 379-IRS -increasing offsets are to the east. Two continuum sources were observed simultaneously in the GGD 27 and AS 353A regions: in the main text the brighter stars are denoted by GGD 27(1) and AS 353A(1), the fainter stars by GGD 27(2) and AS 353A(2).
The H 2 emission along the IRS5 flow terminates in a doublepeaked profile that is spatially coincident with the optical jet bow shock (labelled in Fig. 1d and discussed further in Section 4.5).
No Brg emission was detected towards IRS5; however, the blueshifted, high-velocity jet component is traced back towards the source, this being evident as a 'bump' on the blueshifted side of the otherwise Gaussian H 2 profile in Fig. 2. The profile peaks at

HH 34-IRS
The archetypal HH 34 optical jet and bow shock in L 1641 in Orion ðd , 450 pcÞ forms part of a parsec-scale 'superjet' that includes HH 33, 40 and 85 to the north, and HH 86-88 in the south (Devine et al. 1997a). Overall, the flow has a total length of at least 3 pc. HH 34 itself has been the subject of intense scrutiny (e.g. Reipurth et al. 1986; Bürke, Mundt & Ray 1988;Eislöffel & Mundt 1992;Heathcote & Reipurth 1992), yet the jet has only recently been detected in H 2 , the emission being relatively weak (Stanke, McCaughrean & Zinnecker 1998). HH 34 is also associated with a weak CO outflow (Chernin & Masson 1995).
Here we detect distinct H 2 emission features towards HH 34-IRS and along the southern, blueshifted optical jet (Fig. 1e). The H 2 in the jet coincides with optical knots C, I-J and L (Bürhke et al. 1988; Ray et al. 1996), the distances of these H 2 counterparts from the near-IR source position being 7.7, ,20 and 27.9 arcsec respectively. The H 2 emission found between knots I and J is probably associated with the extended bow wings of knot J. Indeed, in this portion of the HH 34 jet knots I and J are the most laterally extended and most resemble bow shocks (see e.g. the HST images of Ray et al. 1996). Also, the H 2 observed near knot L actually peaks just ahead of the optical emission and may likewise be associated with the flanks of a bow shock further downwind. Deep, near-IR imaging is required to investigate these associations further.
The HH 34 jet is known to be of low excitation (Reipurth et al. 1986). The central portion of the jet -around knot I -is brightest in [S II] emission, the optical knot brightnesses fading nearer the source and further downwind. Correspondingly, the gas excitation in the jet knots drops in this central region, as traced by the ½S ii=Ha ratio along the flow (Bührke et al. 1988). The detection of H 2 emission in this portion of the jet is therefore not too surprising.
The H 2 emission along the jet is all blueshifted out to radial velocities of ,100 km s 21 . These velocities are roughly equal to those measured at optical wavelengths (though note that Bürhke et al. (1988) and Heathcote & Reipurth (1992) report somewhat Towards HH 34-IRS, both H 2 and Brg emission is detected. As with the other YSOs observed, the profile of the latter is much wider than that of the former. Both lines are blueshifted with respect to the systemic velocity of ,8 km s 21 (Chernin & Masson 1995); Gaussian fits to the profiles yield peak velocities of 11ð^5Þ and 224ð^12Þ km s 21 for H 2 and Brg respectively.

HH 72-IRS
HH 72 ðd , 1500 pcÞ in L 1660 is associated with an east -west orientated bipolar molecular outflow (Schwartz, Gee & Huang 1988;Reipurth & Graham 1988). The optical HH knots are only associated with the eastern end of the flow, where it exits the dense core that harbours the powering source. Near-IR observations reveal additional shock features along the flow axis, as well as the likely driving source, HH 72-IRS (Davis et al. 1997). HH 72-IRS is situated about 25 arcsec to the east of the catalogue position of IRAS 07180-2356. However, the CO 2-1 maps of Schwartz et al. (1988), combined with unpublished higher-resolution CO observations obtained by one of us (CJD) and the near-IR observations discussed below, strongly indicate that the infrared source HH 72-IRS is the true outflow source.
Complex, high-velocity H 2 line emission is observed along much of the eastern, blueshifted flow lobe in HH 72. Once again, double-peaked velocity profiles are observed, this time towards the HH 72B bow shock (labelled in Fig. 1f; see also the H 2 image of Davis et al. 1997). Consistently high H 2 velocities are observed along much of the flow between the IRS source and HH 72B; shock velocities in excess of 150 km s 21 are thus inferred, not just for the leading bow shock, but for most of the H 2 features along the flow axis. There is also a faint patch of marginally detected redshifted emission in the counterflow that again hints at a symmetrical bipolar flow centred on HH 72-IRS. Our east-west orientated CGS 4 slit passed just to the north of the H 2 emission observed in the western flow lobe (Davis et al. 1997).
No Brg emission was detected in this system. However, the H 2 profile towards the source is somewhat unique in being highly structured (Fig. 2). At least three velocity peaks are observed, at 113ð^8Þ, 240ð^5Þ and 2129ð^5Þ km s 21 ; these sit on a blue wing that extends to a velocity of , 2165ð^5Þ km s 21 (the L 1660/HH 72 systemic velocity is 120 km s 21 ; Schwartz et al. 1988). In most of the other YSOs observed, the H 2 emission coincident with the outflow source is confined to low radial velocities, even when high velocities are observed further out in the flow lobes (see the plots in Fig. 2). However, in the HH 72-IRS flow H 2 at very high velocities is observed directly towards the source as well as in the extended flow. Of the other sources observed, only SVS 13/HH 7-11 exhibits a similarly complex, high-velocity H 2 profile coincident with the source.

GGD 27 (HH 80/81)
GGD 27 is a well-studied high-mass star forming region situated in L 291 in Sagitarius ðd , 1:7 kpcÞ. Near-and mid-IR observations reveal a number of embedded sources associated with IRAS Figure 2. Brg and H 2 spectra observed towards the driving source of each Herbig-Haro flow. Each spectrum represents the average of three adjacent rows extracted from the spectral images. The x-axis are in LSR velocity (km s 21 ) and are plotted to the same scale; the y-axis are calibrated in Jy, although some of the spectra have been scaled (as noted in the figure). Also, an offset has been added to each spectrum so that the data could be plotted together in the same figure. The instrumental profiles -which have not been removed from the spectra -are also shown. 18162-2048 (Aspin et al. 1991. These IR sources lie at the geometric centre of a massive CO outflow (Yamashita et al. 1989). Luminous HH objects, HH 80-81, are also observed (Heathcote, Reipurth & Raga 1998) some 4-5 arcmin south of the source region. The flow is probably driven by a source undetected at near-IR wavelengths; designated GGD 27-ILL by Aspin et al. (1991), this mid-IR source is offset by approximately 22, 21.5 arcsec from the bright near-IR peak IRS 2. GGD 27-ILL is also, to within a few arcsec, coincident with a radio continuum source which drives a highly-collimated bipolar radio jet (Martí et al. 1993); this jet is orientated along the CO flow axis and points towards the distant HH objects 80/81. For both H 2 and Brg observations of this region, the slit position was set so that it bisected the two bright near-infrared sources, IRS 1 and IRS 2. Here, for simplicity, we refer to the brighter source as GGD 27 (1) and the fainter, more northerly source, as GGD 27 (2); this latter peak likely represents nebulosity associated with the outflow driving source, GGD 27-ILL.
H 2 emission was detected along much of the slit (Fig. 1g). Faint, diffuse emission is observed near the near-IR reflection nebula to the north of the two stars GGD 27 (1) and (2) (offset ,240 arcsec in Fig. 1g), while much brighter knots of emission are detected towards GGD 27 (2) and in between the two stars. In all cases, the line emission peaks at precisely the systemic velocity of 12.5 km s 21 . The H 2 peak situated midway between GGD 27 (1) and (2) is almost certainly associated with the collimated flow, being only a few arcseconds south of the embedded mid-IR/radio source GGD 27-ILL.
H 2 and Brg observations of the two infrared stars are shown in Fig. 2. It is interesting to note that towards the brighter source (1) only Brg emission is observed, whereas towards the fainter star (2) only H 2 emission is detected. Although the H 2 profile towards the latter peaks at the systemic velocity, the Brg profile towards the former is clearly blueshifted, peaking at 260ð^20Þ km s 21 ; it also exhibits a distinct redshifted wing.
As with GGD 27, two infrared stars were observed in the AS 353A region (Fig. 1h). AS 353A itself, the source of the bipolar HH 32 outflow, is here referred to as AS 353A (1), while the fainter source situated ,15 arcsec to the east of AS 353A (1) is referred to as star (2). H 2 emission associated with the HH 32 bow shock has been discussed by Davis et al. (1996). We will therefore only consider the line emission associated with the two stars.
We again see contrasting line emission properties in the two stars observed (Fig. 2); Brg is observed only in the brighter T Tauri star AS 353A (1), while H 2 is only detected in the fainter star (2). In both cases the line profiles are marginally blueshifted, to LSR radial velocities of 224ð^5Þ km s 21 [Brg-AS 353A (1) (Edwards & Snell 1982). Both profiles are also very symmetrical (Gaussian).

HH 379-IRS
HH 379 is situated in Cygnus near the molecular cloud 093.5-04.3 (in the catalogue of Dobashi et al. 1994) at a distance of ,0.9 kpc. HH 379 may be associated with IRAS 2143214719. The IRAS position is, however, offset by almost 1 arcmin from a compact optical nebula that probably marks the true outflow source location (Devine, Reipurth & Bally 1997b); source confusion and/or poor IRAS coordinates may account for this offset.
We assumed that the conical nebula is associated with the HH energy source and therefore positioned our spectrograph slit through the nebula orientated along the axis that links the nebula with the HH object. No continuum emission was detected from the outflow source in our high-resolution echelle data. However, faint H 2 line emission was observed towards the source position (Fig. 1i). The H 2 peak in the P-V diagram does appear to be elongated at a position angle that implies a blueshifted flow component towards the west of the source and a redshifted component to the east, although further kinematical studies of the source and more extended HH 379 outflow system are needed to confirm this. The central H 2 peak, plotted in Fig. 2, is, nevertheless, centred precisely at the systemic velocity of the parent cloud (,3.5 km s 21 , Dobashi et al. 1994).

Brg emission from accretion
Profiles of permitted hydrogen line emission profiles, observed towards a large number of classical T Tauri stars, have been interpreted in terms of both outflow and accretion (Calvet, Hartmann & Hewett 1992;Hartmann et al. 1994;Edwards et al. 1994;Muzerolle et al. 1998a). Reipurth, Pedrosa & Lago (1996) proposed a classification scheme for the observed line shapes, which they applied to their survey of Ha line profile observations of 43 T Tauri stars. They found that while 25 per cent of the sources studied exhibited single-peaked, symmetric (or 'Type I') profiles, 54 per cent had blueshifted absorption features (P Cygni profiles) and 21 per cent had redshifted absorption features (inverse P Cygni profiles, or IPC). By comparison, a similar study by Folha & Emerson (2000) made in near-IR Pab and Brg indicates that most of these profiles were symmetric; 73 per cent of the 50 T Tauri stars observed in Brg exhibited symmetric Type I profiles, while 20 per cent had redshifted absorption features (IPC); in Pab 54 per cent were Type I and 34 per cent had IPC profiles. These differences, particularly between the optical and near-IR lines, may be caused by opacity effects.
The absorption features that define P Cygni and IPC profiles are thought to be signposts of outflow and accretion processes respectively. Redshifted absorption probably derives from accretion as the infalling gas absorbs emission from the hot accretion shock; blueshifted absorption likely arises from outflow, the cooler outflowing gas again absorbing emission from the protostar.
In their study of T Tauri stars, Folha & Emerson (2000) measured Brg linewidths (FWHM) that were typically 200 km s 21 , although lines with FWHM ranging from ,110 to ,300 km s 21 were observed. Towards the Class I sources observed here (AS353A and SVS 13 have been observed before by Najita, Carr & Tokunaga 1996), the Brg profiles have similar widths. The line peaks are typically blueshifted by a few tens of km s 21 , again like the Brg profiles observed by Folha & Emerson (2000). Models that include only outflow predict redshifted line peaks (e.g. Calvet et al. 1992) and in many cases broad, blueshifted absorption features. Neither is observed in our small sample of Class I sources, nor in the observations of Class II sources by Folha & Emerson (2000).
Instead, the blueshifted peaks and broad, symmetric line profiles observed compare (qualitatively at least) more favourably with magnetospheric accretion models (Hartmann et al. 1994;Muzerolle et al. 1998a). These models do, however, often produce redshifted absorption features as well as lines that are typically much narrower than are observed here and elsewhere (Najita et al. 1996;Folha & Emerson 2000). The linewidth and depth of the redshifted absorption is strongly dependent on inclination angle. A weak or low-temperature accretion shock might also explain an absence of redshifted absorption. Linewidths are predicted to increase, and the red absorption become more prominent, at large inclination angles (measured with respect to the line of sight; Muzerolle et al. 1998a). For the few sources observed here, for which Brg was detected, the flow inclination angle is reasonably well known (listed later in Table 3). Yet we do not observe any evidence of changing Brg line shapes with flow inclination angle; if we take AS 353A, SVS 13 and HH 34 as examples of 'pole-on' (small inclination angle), intermediate, and 'disc-on' systems respectively, we find that the linewidths are roughly the same (the HH 34-IRS profile being about 20 per cent wider), and we see no evidence of redshifted absorption, even in the Brg profile observed towards HH 34-IRS (although admittedly this line is relatively weak). As was noted by Folha & Emerson (2000), redshifted absorption is more typically observed in hydrogen Balmer lines than Brackett lines [for example, the Ha -Hd Balmer lines observed towards AS 353A by Alencar & Basri (2000) do have strong red absorption; the Brg line observed here (and by Najita et al. 1996) does not]. The lower-energy Balmer lines may thus be considered as a better probe of accretion processes, even though extinction affects are more of a problem. When considering model predictions for linewidths, one should also note that the accretion models of Hartmann, Muzerolle and co-workers do not include rotation, outflow or turbulence, process that will almost certainly broaden the theoretical line profiles and, in the case of turbulence, serve to produce more symmetric profiles (Edwards et al. 1994).
From low-resolution spectroscopic observations of 30 T Tauri stars, Muzerolle, Hartmann & Calvet (1998b) also find a very convincing correlation between Brg line luminosity and accretion luminosity, the latter being derived independently from the infrared-excess continuum emission. They find that the line strength decreases with decreasing accretion rate. The Brg luminosities measured for the five sources observed in this paper (see Table 4 below) are typical of those observed for T Tauri stars. If we assume that the Brg line luminosity versus accretion luminosity relationship applies also to Class I sources, then following Muzerolle et al. (1998b) we predict accretion luminosities in the range 0:2-8 L ( (the limiting sources in this range being IRAS 0423912436 and SVS 13). There is, however, considerable uncertainty in the measured line luminosity because of the large extinctions towards the embedded, Class I sources (listed in Table 3 below). If we assume a 'typical' extinction of A v , 20, then the true line luminosity would increase by a factor of 10 A2:2/ 2:5 , 6:3, where the extinction at 2.2 mm, A 2:2 , A v / 10. The accretion luminosity would then be in the range ,2-80 L ( for the low-mass Class I sources observed [and ,3 Â 10 4 L ( for the distant, high mass source GGD27-IRS (1)]. This very crude analysis suggests that the accretion luminosity is roughly comparable to the bolometric luminosity of each source (see Table 4).

H 2 emission from outflow
There are a number of distinct differences between the H 2 and Brg observations presented in this paper. The Brg profiles in Fig. 2 are much wider than their H 2 counterparts, although the H 2 profiles are, in a few cases at least, more structured (less Gaussian). Also, in all cases the Brg emission is confined to the outflow source, being unresolved along the slit/outflow direction, whereas the H 2 emission is extended along the flow axis, even towards sources where H 2 is detected only towards the source position (e.g. HH 379-IRS). To better illustrate these extended MHEL regions we plot in Fig. 3 contour diagrams of five outflow source regions. The continuum emission associated with each YSO has been fitted and removed, row-by-row via linear least-squares fits to the emission on either side of the H 2 line emission peaks, so that only the emission associated with the MHEL regions remains. Multiple velocity components are revealed in the MHEL regions associated with four of the five YSOs considered.
The continuum-subtracted P -V diagram for SVS 13 is dominated by two velocity peaks, a low-velocity component (LVC) at V LSR , 220 km s 21 and a high-velocity component (HVC) at V LSR , 290 km s 21 . Both components have faint extensions (with positive offsets in Fig. 3; labelled 1 and 2) along the blueshifted HH 7-11 outflow lobe; the LVC also has two shorter, though brighter, extensions in the 'counter-flow direction' (negative offsets in Fig. 3; labelled 3 and 4), one at near-stationary velocities and the other approaching 250 km s 21 . All four features could be from the same wide-angled, blueshifted HH 7-11 flow lobe, features 3 and 4 deriving from the near-side of the flow cavity that is seen projected 'behind' the SVS 13 source. The outflow clearly has a wide opening angle on large scales -the HH objects 7 -11 themselves represent shock nebulae along the walls of a conical cavity seen in deep optical images (e.g. Davis et al. 1995)and the flow is inclined at a reasonably oblique inclination angle of ,408 to the line of sight (Carr 1990).
The IRAS 0423912436 MHEL region comprises only a single LVC, although this component extends into both the blueshifted (decreasing offsets in Fig. 3) and redshifted outflow lobes. The blueshifted extension comprises two 'jet' components (labelled in Fig. 3), a near-stationary cusp and a higher-velocity extension that is indicative of acceleration. The same pattern of emission features is observed in the LVC associated with the L 1551-IRS5 MHEL region; again both blueshifted (decreasing offsets in Fig. 3) and redshifted extensions are detected. Like IRAS 0423912436, the blue MHEL lobe in L 1551-IRS5 comprises a near-stationary component and an increasing-velocity feature; both jet components in L 1551-IRS5 extend out towards the known jet bow-shock (also labelled in Fig. 3). The near-stationary and 'accelerating' MHEL components in the blue outflow lobes in both IRAS 0423912436 and L 1551-IRS5 could be associated with two separate jets (as already mentioned, optical imaging of L 1551-IRS 5 indicates the possible presence of two jet flows; Fridlund & Liseau 1998) or they could represent emission from entrained gas in the walls of a single jet; the two MHEL extensions would then represent emission from the near and far sides of a 'hollow' H 2 component in each flow, the former being inclined towards the observer and the latter almost parallel with the plane of the sky.
Subtraction of the stellar continuum emission associated with

IRAS 04239
Offset (arcsec) the HH 34-IRS MHEL region reveals only a single, relatively featureless LVC (Fig. 3) that is blueshifted by about 10 km s 21 from the systemic velocity. In stark contrast, the HH 72-IRS MHEL region exhibits very complex velocity structure. This is probably caused in part by the greater distance to this intermediatemass outflow source. As a consequence, the HH 72-IRS MHEL will be less well resolved from the rest of the bipolar, molecular outflow than in the other sources in Fig. 3. We do nevertheless identify two LVC components, at low-and intermediateblueshifted velocities, and a HVC at V LSR , 2130 km s 21 . It seems unlikely that molecular hydrogen would survive in the high-temperature shocks associated with the magnetospheric accretion processes discussed in Section 4.1. The fact that the H 2 emission regions are extended suggests that they originate in a wind rather than an accretion zone. Indeed, the MHEL regions are in many respects similar to the forbidden emission-line (FEL) regions observed in many classical T Tauri stars (Hamann 1994;Hartigan et al. 1995;Hirth et al. 1997) and some Herbig Ae/Be stars (Corcoran & Ray 1998). Spectroscopy of these optical emission-line regions reveal in many cases multiple velocity components, much like those seen here in H 2 towards the Class I sources. Typically, both LVC and HVC, with velocities in the ranges 25 to 220 km s 21 and 250 to 2150 km s 21 , are observed in the FELs; similar components are observed in at least two of our Class I YSOs. The optical FELs are preferentially blueshifted, again like the MHEL regions. The optical LVC is also often spatially more compact than the HVC; the same trend is observed in two of our MHEL regions (SVS 13 and HH 72-IRS), and in a third (L1551-IRS5) the extent and offset of the LVC increases with velocity. To illustrate this last point more clearly we plot the offset of the LVC and HVC components from the YSO continuum centroid in Fig. 4.
Gaussian fits to the MHEL emission peaks and continuum strips either side of the H 2 emission regions were made using the P-V plots in Fig. 3. These fits yield the spatial offsets of the MHEL components from the continuum position (note that there was no measurable offset of the LVC in HH 34-IRS from the continuum position); a second-order polynomial fit to the continuum points is drawn in each plot. The scatter of the continuum position measurements about this 'stellar centroid' fit is of course dependent on the brightness of the observed continuum. Nevertheless, in all four cases in Fig. 4, clear offsets of the MHEL emission from the driving YSOs are apparent. The separate LVCs and HVCs in the SVS 13 and HH 72-IRS MHEL regions are represented by individual clusters of measurements: note in particular how the HVCs in these two sources are situated further from their sources than the LVCs, and in L 1551-IRS 5 and IRAS 0423912436 how the MHEL is offset further from the source at increasing blueshifted velocities.

The origin of the MHEL regions
A number of theories have been presented to explain the FEL regions observed in TTSs. Here we compare some of these models with our H 2 observations. Models which interpret the HVC and LVC emission as being from the near and far sides of a hollow flow, observed in projection along the line of sight, seem to be ruled out. Observations clearly show that FEL line ratios in the HVCs and LVCs of the same flows differ, indicating contrasting excitation conditions in the two flow components (Hammann 1994; Hartigan et al. 1995); this is in contrast to what one would expect from the two sides of a single, hollow flow. Indeed, line ratios suggest higher densities though lower excitation in LVCs as compared to the HVCs in FEL regions. This is consistent with the LVCs being observed more often in H 2 than the HVCs in our (limited) sample of Class I MHEL regions. Of course, measurements of the H 2 excitation in the HVCs and LVCs observed in the Class I flows would be of considerable interest in similarly ruling out such a model for the MHEL regions. Ouyed & Pudritz (1994) model FELs with a conical shock in a disc wind; they reproduce [O II] and [S II] line profiles in the hot, post-shock gas just above the disc plane. The shock serves to recollimate the magnetohydrodynamic disc wind. The model does produce complex and double peaked profiles similar to the FEL observations. The post-shock emission zone is also offset from the outflow source, though only by about 10 11 cm, and again in the axially-symmetric case one would expect to see the same excitation conditions in the low-and high-velocity post-shock components, if these correspond to the LVCs and HVCs observed. Thus, such a model could at best only explain one flow component.
A model that assumes two independent flow components is preferred for the FELs (e.g. Hirth et al. 1997), and we have (as yet) no reason to believe that such a model would not also apply to the MHEL regions. Kwan & Tademaru (1988) and later Kwan (1997) suggest that while the HVC derives from a collimated, highvelocity jet, the LVC could be produced in a warm disc corona or a slow disc wind. Kwan & Tademaru (1995) model [O I] profiles for an LVC produced in a disc corona; the velocity profile extends for a few km s 21 out to ,100 km s 21 , with the peak only slightly blueshifted from the stellar rest velocity. Their model corona, with densities of the order of 10 7 cm 23 and temperatures of about 10 4 K (needed to reproduce the observed [O I] line ratios) would be conducive to H 2 excitation. However, the corona probably only extends to a height of about 10 11 -10 12 cm above the disc surface, whereas the observed LVC MHEL emission regions in SVS 13 and L 1551-IRS 5 are offset by ,10 15 cm from their sources (Fig. 4), with the more distant HH 72-IRS MHEL emission being offset by about 10 16 cm. Thus, such coronal H 2 emission would have to be incorporated into the flow to distance it from the exciting source in each case.
The line profile widths displayed in Fig. 2 could be broadened by either Keplerian rotation in or near the accretion disc surface (Hartigan et al. 1995), or simply through lateral expansion of the outflowing gas (Hamann 1994). If the former applies, then the FWZI of each H 2 profile should reflect the maximum velocity attained in the inner regions of the accretion disc (illustrated in Fig. 5). Given the inclination angle of the flow to the line of sight, u, one may then estimate the maximum Keplarian velocity associated with the H 2 and Brg emission regions, as V kep ¼ DV FWZI / ð2:0 sin uÞ; u is the flow inclination angle to the line of sight and DV FWZI ¼ V max ðredÞ 2 V max ðblueÞ. Estimates for each source are listed in Table 2. The narrow H 2 profiles obviously lead to a much lower velocity than the Brg observations. As the square of V kep is inversely related to the radial distance of the emitting 'ring', it is not surprising that the higher-excitation Brg emission would derive from much closer in to the star. Indeed, for a 1-M ( star, a typical broadened Brg profile with V kep , 250 km s 21 could originate from a ring of radius ,2 Â 10 11 cm; the H 2 would then derive from further out in the disc plane, at a radius of ,9 Â 10 12 cm for a keplerian velocity of 40 km s 21 (Kwan & Tademaru 1995).
As we have established that the MHEL regions are associated with the outflow, the H 2 profiles could instead be broadenend by expansion of the flow lobe. Taking into account the inclination angle of the flow axis with respect to the line of sight, u, the flow opening angle, a ¼ 2 arctanðDV FWZI cos u/ 2V peak Þ, where V peak is the velocity of the HVC or LVC emission peak. For 'typical' values of DV FWZI , 70 km s 21 and u , 458, the full opening angle of an LVC with V peak , 20 km s 21 would be roughly 1008; by comparison, an HVC with a similar FWZI though a V peak , 70 km s 21 would have an opening angle of the order of 408. These values are likely to be overestimated, because turbulence and thermal motions will contribute to the line broadening, even in the case of a magnetic C-shock where the excited H 2 molecules predominantly radiate from a region just ahead of the shock front. The observations, particularly of SVS 13 where both LVC and HVC components are clearly detected in H 2 , do in any case show that the HVC is probably more highly collimated than the LVC, as is the case with optical FEL regions.
From our flux-calibrated spectra (Fig. 2) we may also estimate the mass of hot H 2 gas associated with each MHEL LVC, and subsequently the mass outflow rate specific to the H 2 flow component observed. The H 2 and Brg spectra plotted represent the average of three rows (centred on the source) extracted from each spectral image, and therefore cover an area of 0:8 Â 2:7 arcsec across each source, the major axis being aligned with the flow axis. The column of H 2 molecules populating the v ¼ 1, J ¼ 3 ro-vibrational level is directly related to the observed flux (assuming optically thin emission) via the equation N 1;3 ¼ 4pI 120Sð1Þ / ðhn 120Sð1Þ A 120Sð1Þ Þ; where I 120Sð1Þ is measured in W m 22 sr 21 , and n 120S(1) and A 120S(1) are the frequency of the transition and the Einstein A coefficient respectively. With the H 2 molecules in local thermodynamic equilibrium, the total column of H 2 molecules is then given by N H2 ¼ N 1;3 ZðTÞ=ðg 1;3 exp½2hn 120Sð1Þ / kTÞ; where the partition function, Z(T), is related to the gas excitation temperature, T, by ZðTÞ ¼ 0:024T/ ð1 2 exp½26000/ TÞ (Smith & Mac Low 1997). For the 1 2 0Sð1Þ transition observed, A 120Sð1Þ ¼ 3:47 Â 10 27 s 21 and the statistical weight, g 1;3 ¼ g s ð2J 1 1Þ ¼ 21 (for an ortho-para-H 2 ratio of 3). H 2 column density estimates for the sources where emission was detected are listed in Table 3. We assume a temperature of 2000 K in the hot, post-shock gas; this is equivalent to the excitation temperatures typically observed in HH objects (e.g. Gredel 1996). The flux measurements are also corrected for extinction; although in some cases the A v estimates used may be slightly greater than the true extinction to the MHEL regions. Observed masses are given in Table 3; these are likely to be accurate to within a factor of a few, the greatest source of error being from the distance to the source used (accurate to perhaps 30 per cent) and our assumptions concerning LTE and the H 2 level populations. Moreover, our 0.8-arcsec wide CGS 4 slit may not include all of the H 2 emission region.
To calculate the mass outflow rate in each region we take the peak H 2 velocity observed in Fig. 2 and correct this for the flow inclination angle (if known). In a few cases, the H 2 profile peaks at approximately the systemic velocity. For the purpose of this analysis, we then use an arbitrary velocity of 5 km s 21 . We arrive at the values given in Table 3. Not surprisingly, the higher massloss rates are attributed to the more massive (and more distant) sources, GGD 27-IRS and HH 72-IRS. The quoted mass loss rates compare favourably with typical values for Class I YSOs (,10 27 -10 28 M ( yr 21 Þ, derived for example from CO observations of molecular flows from low-mass sources (Bontemps et al. 1996). However, one should remember that we do not know the relationship between the observed hot H 2 component and the rest of the flow.

Line emission versus source and outflow parameters
Is there a correlation between the presence or absence of an MHEL or Brg emission region and the source age, luminosity, or any of the known outflow parameters? If, for example, the Brg emission derives from an accretion zone in the inner disc and the H 2 emission derives from the base of a large-scale molecular outflow, one would expect extinction and the inclination angle of the source to affect the observed H 2 and/or Brg line strengths. Moreover, one might assume that high-mass YSOs would produce more luminous MHEL or Brg emission regions, or higher-velocity components. We have investigated these possibilities in some detail and list in Tables 3 and 4 various parameters for each outflow region.
All HH sources observed possess a rising spectral energy distribution through the far-IR bands observed by IRAS and thus are considered Class I or borderline Class I/II YSOs. Source confusion is likely to affect IRAS flux measurements, however, particularly in busy star forming regions like SVS 13 and GGD 27, where multiple mid-IR and/or radio sources have been detected within a few arcseconds of the outflow origin. Higher-resolution near/mid-infrared photometry is used by Greene & Lada (1996a) to classify the YSOs in their extensive spectroscopic survey; they designate Class I to sources with spectral slope indices a . 0:3, and Class II to those with a , 20:3; sources with intermediate indices are referred to as 'flat spectrum' sources. Greene & Lada also note that Class I YSOs are distinct from Class II sources in having no absorption features, although Luhman & Rieke (1999) have since detected weak absorption in 25 per cent of the Class I and flat-spectrum sources in their survey of the r Oph cloud core.
In Table 4 we therefore list each source as I/II if it has a flat near-IR spectral energy distribution (SVS 13, HH 34-IRS and HH 379-IRS;Reipurth & Aspin 1997;Greene & Lada 1996a;Carr 1989) or distinct absorption lines (L 1551-IRS 5 exhibits strong CO absorption; Reipurth & Aspin 1997). Younger, more deeply embedded protostars are known to drive more powerful outflows and, consequently, are assumed to possess higher accretion rates (e.g. Bontemps et al. 1996). We find, however, no obvious correlation between source classification (that is Class I, I/II or Class II) and the presence, absence or indeed line flux in either Brg or H 2 . Observations of a larger sample of outflow sources, together with independent near-and mid-IR photometry of these sources would be useful to search for such a correlation. Another indicator of source age is the ratio of bolometric-tosubmillimetre luminosities, a lower ratio being associated with the more deeply embedded YSOs; we therefore also list this ratio in Table 4, as taken from the survey of Reipurth et al. (1993). As before, source confusion associated with the large (sub) millimetre photometry beams may introduce uncertainties in this classification scheme. This may explain why we again see no real correlation between this age indicator and the H 2 and/or Brg luminosities.
We lastly mention that we also find no relationship between H 2 or Brg line luminosity and flow inclination angle or the extinction towards each source. The only possible correlation we do find is between the Brg/H 2 line ratio and the source luminosity (Fig. 6), although this is at best a very weak correlation (based largely on lower and upper limits) that again requires further observations of a wider sample of outflow sources. The higher ambient gas densities and the inclusion of more of the extended molecular flow in the Outflow inclination angle with respect to the line of sight. f Approximate H 2 flow velocity (accurate only to within 10 km s 21 ), assumed to be the peak H 2 velocity observed (see Fig. 2), corrected for the flow inclination angle (if known). g H 2 mass outflow rate.

Echelle spectroscopy of molecular bow shocks
Although the focus of this paper has been on line emission from the vicinity of outflow sources (the MHEL regions), we have at the same time observed H 2 bow shocks in the more extended flow lobes. Double-peaked profiles are observed in three notable bows; in HH 7 (offset ,70 arcsec in Fig. 1a), in the L 1551-IRS5 jet bow about 10-12arcsec to the southwest of IRS5 (negative offsets in Fig. 1d), and in the main HH/H 2 bow HH 72B situated about 25 arcsec to the east of HH 72-IRS (positive offsets in Fig. 1f). In all three sources the bows are blueshifted with respect to the driving source.
The double-peaked velocity profiles observed in each source are expected of hollow, shell-like bow shocks (e.g. Carr 1993; Davis & Smith 1996a,b). In all three cases the low-velocity component is brighter than the high-velocity peak; this is consistent with a highvelocity 'bullet' interpretation, rather than a 'shocked stationary clump' model (e.g. Hartigan et al. 1987).
In HH 7 the velocity separation between the low-and highvelocity peaks is of the order of 70 km s 21 : in the HH 72 and IRS 5 bow shocks, the peak-to-peak separation is considerably larger, approaching 150 km s 21 . The full extent of these H 2 profiles therefore indicates a shock velocity of the order of 100-150 km s 21 (Hartigan et al. 1987); the emission is also testament to the survival of H 2 in the hot, post-shock gas. Molecular hydrogen is dissociated in planar 'jump' shocks with velocities exceeding 30-50 km s 21 (Smith 1994) provided the gas density is reasonably high (this is expected given the strong/detectable line emission). However, the curved geometry of a bow, when combined with jet variability (in direction and velocity) can result in peak-to-peak velocity separations of this order for medium-to high-velocity (,100-200 km s 21 Þ bows orientated out of the plane of the sky (Downes 1996;Völker et al. 1999;Tedds, Brand & Burton 1999). Note from Table 3 that the observed orientation of each flow is approximately 458, the possible exception being HH 72 for which the flow orientation is not well known. If magnetic fields play a role, resulting in a 'continuous' or C-type shock, the low-velocity emission would then be excited in the slow or near-stationary gas just ahead of the shock front. Some or all of the higher-velocity component could also derive from the jet or Mach disc (i.e. the jet shock that separates the shock working surface from the jet itself). The application of existing numerical models (see e.g. Volker et al. 1999;Downes & Ray 1999) to give position-velocity diagrams like those discussed here (in particular for different viewing angles) would be of considerable interest.
Collectively, these echelle observations of the H 2 emission in the extended outflows offer further proof of the propensity of molecular bow shocks in outflows from embedded YSOs. Shock speeds associated with these bows are high, exceeding in many cases the H 2 dissociation speed limit [see also the high H 2 proper motions measured in HH 7-11 (and other sources) by Chrysostomou et al. 2000]. Flow variability and shock geometry are probably the dominant mechanisms that allow the H 2 molecules to survive at high speeds. The velocity-separation between peaks in the double-peaked profiles observed, measuring 80-150 km s 21 , nevertheless remain a challenge for those modelling molecular bow shocks in YSO outflows.

S U M M A R Y A N D C O N C L U S I O N S
High-resolution echelle spectra of embedded, Class I outflow sources reveal H I and H 2 line emission regions coincident with many of the young stellar objects (YSOs) observed. Brg emission is detected towards ,45 per cent of our target list; H 2 emission is observed towards ,80 per cent of the YSOs. The Brg emission is only observed coincident with the protostars, the emission being spatially unresolved along the slit/flow axis. Conversely, the H 2 towards each YSO is marginally extended along the flow axis (over a few hundred to a few thousand AU). Spatially and kinematically independent H 2 features are also observed in the more distant molecular outflow lobes and known HH objects in most flows. Towards YSOs where both H 2 and Brg emission lines have been observed, the latter are always considerably broader than the former. Brg profiles typically measure .200 km s 21 FWZI; the width of H 2 profiles are of the order of 70-200 km s 21 FWZI, although the H 2 profiles do in some cases comprise of multiple velocity components.
The H 2 and Brg emissions probably trace the orthogonal processes of outflow and infall. The characteristics of the Brg profiles are similar to those observed in the more evolved T Tauri stars, so we attribute this emission to magnetospheric accretion processes. Conversely, the H 2 emission regions are comparable to the forbidden emission line (FEL) regions observed in classical T Tauri and some Herbig Ae/Be stars. We refer to the observed H 2 regions -defined as the emission zones within a few arcsec of each outflow source -as molecular hydrogen emission-line, or 'MHEL', regions. Like their optical FEL counterparts, the MHEL regions are preferentially blueshifted: the H 2 emission is generally offset from the stellar centroid by a few tenths of an arcsecond (a few hundred au in each case). In a few targets two or more velocity components are observed; the high velocity H 2 is spatially offset further from the source than the lower-velocity H 2 , as is also the case in some FEL regions. Lastly, like FELs, MHEL regions seem to be a feature of both low and high-mass protostars, since MHELs are observed towards the distant, luminous outflow sources in GGD27 and HH72, as well as the low-luminosity, and considerably closer YSOs IRAS 0423912436, B5-IRS1, L 1551-IRS 5 and HH 34-IRS.