``Cold HI Structures in the Perseus Arm:
Molecular Clouds in Formation?''

S. J. Gibson & A. R. Taylor

University of Calgary

1998, Bull. A.A.S., 30, 1341, #65.12

Press Release

A press release was made for this poster.

Electronic Poster Contents



HI self-absorption (HISA) against background HI 21cm emission reveals cold atomic gas structures in the interstellar medium. These often correlate spatially with molecular gas and dust (e.g., Knapp 1974; Peters & Bash 1987), though certainly not always; the exact physical relationship between HISA and H_2 is not well known. HISA features as a group are poorly understood, due to the high angular resolution required for accurate estimates of background spectra from adjacent sightlines, and to the lack of unbiased high-resolution searches for such objects.

We have begun a systematic study of HISA within a 73 x 9 degree region (Fig. 2) mapped by the Dominion Radio Astrophysical Observatory's Synthesis Telescope (DRAO-ST) for the ongoing Canadian Galactic Plane Survey (CGPS; Taylor et al. 1999; English et al. 1998). The ~1 arcminute DRAO-ST beam reveals considerable substructure within detected clouds (see Figs. 1, 3 & 5). Our investigation has uncovered a wealth of remarkable features in both the Local and Perseus Arms. Some have clear 12CO and dust counterparts, while many others do not.

Below, we investigate properties of sample CGPS HISA features and also consider aspects of the population as a whole. Their partial correlation with CO and dust, narrow linewidths, and intermediate appearance between diffuse HI and compact CO morphologies suggest the HISA features may represent HI in the act of molecular condensation. This possibility is augmented by a substantial abundance of HISA at the same velocities where molecular clouds are likely to form after encountering the Perseus Arm density wave shock.


We wish to obtain physical parameters for our HISA objects: spin temperature T_spin, optical depth tau_hisa, column and volume densities N_hisa and n_hisa, and mass M_hisa. This requires solving the radiative transfer equation. We consider the implicit 4-component formulation of Feldt (1993), with a HISA feature, warm, optically thin HI emission in front of and behind it, and a continuum background T_c:

[Radiative Transfer Equation]

Here T_on and T_off are continuum-subtracted ON and OFF brightnesses, and p is the fraction of T_off emission lying behind the HISA feature. Constraining p leaves two unknowns, tau_hisa and T_spin. Inclusion of other relations, such as that for Gaussian line column density, uniform gas density, and an ideal gas law then produce the transcendental equation

[Ugly Iterative Equation]

where L_|| is the HISA line-of-sight pathlength, and a canonical pressure such as P_therm / k = 4000 K cm^-3 is used (only thermal pressure is relevant).

In Table 1, we estimate properties for the objects in Figures 3 & 5. Both have T_c = 5 K from the general Galactic synchrotron background. Representative T_on & T_off values were taken from Figures 4 & 6. In more sophisticated future analyses, these will be measured separately for each HISA voxel, with T_off taken from a volume interpolation of non-HISA voxels surrounding the feature in space and velocity.

We assume the smallest high-contrast filaments in each object are roughly cylindrical (within a factor of 2) to obtain L_||; the pathlength will be that of a single filament if their filling factor inside the object is low, which typically appears to be the case. The ~2 kpc distance to the Perseus feature is based on the sightline velocity model of Roberts (1972), while the 200-500 pc distance for the Local feature is constrained merely to lie within the Local Arm, but outside the Local Bubble. p = 1 was used for both objects; if p < 1, T_spin, tau_hisa, and related values become lower limits. For the Perseus feature, M_hisa refers only to the central 10 x 45 arcminute component; the entire complex may be 10-100 times more massive.

Table 1: Sample Feature Properties

[GIF Table of Feature Properties]

Though our methods of estimating distances and background spectra are still crude at this stage, the values obtained serve as useful rough measures of the class of cold atomic structures revealed by our ongoing investigation.


The true power of a large survey is the ability to study ensembles of objects statistically. We are working to develop automated techniques for the identification of HISA features in velocity cubes and the measurement of their properties, which can then be used to examine group characteristics and correlations with other ISM constituents, e.g., CO emission. The complex shapes and sheer numbers of HISA features require the task of locating HISA voxels and their non-HISA neighbors (for background estimation) to be automated.

The darker HISA features are quite easy to identify by eye but less so by computer, due to confusion imposed by noise and complex background HI emission structure. An initial method which traces at least the darkest HISA is to smooth each velocity channel in the cube spatially to improve S / N, and then convolve each sightline spectrum with a narrow (FWHM = 2 km/s) 1-D ``Mexican Hat'' wavelet function to pick out sharp local spectral minima. We have used this technique to flag the darkest HISA voxels in the velocity cube, interpolating non-HISA neighbors in the surrounding volume to estimate T_off. Resulting aggregate properties for each channel are shown in Figure 7.

The top panel shows the average T_off - T_on contrast. For some HISA voxels this can be 40 K or more, but for most it is only a few kelvins. Since the limit of detectability is 1-2 K, this suggests many more HISA features may exist which are too optically thin to see. The middle panel shows the average of T_off itself, with <T_B(HI)> overplot for comparison. Though <T_off> is a measure of average HISA neighbor brightness rather than average total HI brightness, a correlation between the two is apparent. Comparison of <T_B(HI)> with <T_off - T_on> above shows most detected features have a contrast of a few percent of background. The minimum <T_off> value is ~40 K (values of zero occur in channels where no HISA was found). Consideration of the radiative transfer equation suggests this may represent a lower limit to T_spin, as features become invisible when T_spin = p * T_off + T_c.  40 K is warmer than most HISA detected in previous searches of molecular clouds, but cooler than canonical temperatures of 80 K for ``cold'' neutral hydrogen. We use T_spin = 60 K to compute optical depths integrated over each channel in the bottom panel. This serves as an estimate of HISA mass vs. velocity. Since our present algorithm flags only the darkest HISA voxels, it uses many of their less-dark HISA neighbors to find T_off. This underestimation of background brightness makes all current contrast, optical depth, and mass measurements lower limits.


A key question about HISA concerns the nature of its physical environment. On small scales, we are interested in whether HISA occurs inside, around, or independent of molecular gas. Our investigation so far shows a mixture of these cases, perhaps indicating an evolutionary relationship. On large scales, the spatial distribution of HISA clouds gives clues relating their formation to global processes in the ISM. We wish to learn whether HISA is distributed in a homogeneous ``plum pudding'' fashion, or is more concentrated in specific regions.

Figure 7 shows a significant peak of integrated HISA opacity at a radial velocity of -41 km/s LSR. This peak coincides with others in mean CO and HI emission brightness, both of which in fact correspond to gas constituents of the Perseus Spiral Arm. A plum pudding distribution would also show HISA peaks coincident with background HI peaks, since the fraction of HISA to HI is constant. However the HISA peak at -41 km/s is much greater relative to the general HISA level than the HI emission peak is to other HI, arguing against a homogeneous distribution.

Instead, the peak in integrated opacity suggests a substantial abundance in HISA in the Perseus Arm, in full agreement with visual impressions of the amount of HISA in different channels of the velocity cube. While most velocities appear to have a certain low level of HISA, perhaps indicating a general faint homogeneous population, Perseus velocities show a major enhancement.

These velocities are the same predicted by Roberts (1972) to contain gas encountering the density wave shock of the Perseus Arm. Gas falling into the arm's gravitational potential slows abruptly upon colliding with gas already there, and picks up speed again on its way out, appearing to slow once more at larger distances, due to the perspective effects of a given sightline (Figures 8 & 9). The shock itself induces cloud compression, leading to the condensation of HI into H_2, and the eventual formation of new stars. We believe we may be seeing evidence of this initial phase transformation. The radiative transfer requirement of additional HI emission beyond the HISA is satisfied by the Roberts spiral shock geometry, which places the HI leaving the arm further along the sightline at the same velocity as gas in the shock. Though our measurements are currently only lower limits tracing the darkest of the HISA, we feel it is likely that a more thorough census will produce the same result.


Cao, Y., Terebey, S., Prince, T. A., & Beichman, C. A., 1997, Ap. J. Supp., 111, 387
Dame, T.M., Ungerechts, H., Cohen, R.S., de Geus, E., Grenier, I.A., May, J.,
     Murphy, D.C., Nyman, L.A., & Thaddeus, P., 1987, Ap. J., 322, 706
English, J., et al., 1998, Pub. Ast. Soc. Aust., 15, 56
Feldt, C., 1993, A. & A., 276, 531
Heyer, M. H., Brunt, C., Snell, R. L., Howe, J. E., Schoerb, F. P., & Carpenter,
     J. M., 1998, Ap. J. Supp., 115, 241
Knapp, G. R., 1974, A. J., 79, 527
Laustsen, S., Madsen, C., & West R, 1987, Exploring the Southern Sky,
     Springer-Verlag: Berlin
Peters, W. L., & Bash, F. N., 1987, Ap. J., 317, 646
Roger, W. W., 1972, Ap. J., 173, 259
Taylor, A. R., et al., 1999, A. J., in preparation


Click on the GIF images below to obtain a larger version. Some have plain, unmarked versions available.

Figure 1: Large Field View

Center: 12CO J=1-0 emission at the same velocity, with a display ceiling of 3 K. Data from Heyer et al. (1998). Right: IRAS HIRES 60µm dust emission, from ~30 MJy/sr. Data from Cao et al. (1997).

Figure 2: Milky Way Context

Optical Milky Way, showing the 73 x 9 degree CGPS coverage (yellow) with the area of Figure 1 marked. B1950 coordinates (green) and constellations (red) are also shown. Image from a mosaic of scanned photographs by Laustsen, Madsen, & West (1987), courtesy of the NASA Astrophysical Data Facility.

Figures 3-6: Perseus & Local Example HISA Features & Spectra

Figure 3 (Upper Left): Close-up of a HISA complex in the Perseus Arm at -41 km/s, with brightness ranging from 50-135 K. Yellow CO contours (Heyer et al.) are at 1, 3, & 5 K, and red 60 µm contours (Cao et al.) are at 15, 17, & 19 MJy/sr. ON (white) and OFF (black) spectrum boxes used in Fig. 4 are shown.

Figure 4 (Upper Right): ON (solid) and OFF (dashed) HI and CO representative spectra for the Perseus HISA feature, extracted from boxes marked in Fig. 5.

Figure 5 (Lower Right): A much fainter feature in the Local Arm at $+1\kms$, with brightness from 55-80 K. Yellow CO contours (Dame et al. 1987) are at 0.33, 0.67, 1.0, and 1.33 K. No correspondence in dust emission was found. ON (white) and OFF (black) spectrum boxes used in Fig. 6 are shown.

Figure 6 (Lower Left): ON (solid) and OFF (dashed) HI and CO representative spectra for the Local HISA feature, extracted from boxes marked in Fig. 5. The CO detection is marginal but suggestive, being limited in S/N by low spatial resolution.

Plain (unmarked) image
Plain (unmarked) image

Figure 7: HISA vs. Velocity

Top: HISA brightness temperature contrast averaged over each channel. Middle: Average channel T_off brightness (zero where no HISA detected), compared to mean overall HI brightness in each channel (blue dashed). Bottom: HISA optical depth integrated over each channel, assuming T_spin = 60 K. Since our algorithm only detects the darkest HISA voxels at present, this measurement is only a lower limit, as is the integrated HISA mass estimate on the right axis. Average CO channel brightness (red dotted) is shown for comparison.

Figure 8: HISA Location in Spiral Arms

This plan view of the Milky Way shows the area covered by the Canadian Galactic Plane Survey (green), and the sightline of the objects in the images shown above (black). Most of the HISA has radial velocities typical of the Perseus Arm.

Figure 9: HISA Geometry in Spiral Shock Model

This adaptation of a figure from Roberts (1972) shows the predicted velocities of gas along this sightline for his spiral shock model. The arm shock itself is shown in red. Material flows into the arm from the left, decelerates in the shock, speeds up again leaving the shock, and finally appears to decelerate further due to a perspective effect. A likely location for the HISA is just downstream of the shock (green), with HI emission on the further side of the velocity hill providing background illumination.


Return to the Neutral Hydrogen Self-Absorption page.