THE EXTENSION OF THE TRANSITION TEMPERATURE PLASMA INTO THE LOWER GALACTIC HALO

The Galaxy has a gaseous atmosphere consisting of different phases of the interstellar medium (ISM) including the cold neutral medium (CNM), the warm neutral medium (WNM), the warm ionized medium (WIM), and the hot ionized medium (HIM). The different phases exhibit different extensions away from the Galactic mid-plane into the halo. A study of the extensions provides information about the different phases, their different origins, and the way in which the phases interact with each other.

Absorption line spectroscopy toward objects with known distances provides an important method for measuring the extension of gas away from the Galactic plane. One can study the absorption to stars situated at different distances, d, and determine column densities, N(X), for a wide variety of interstellar species, X. With a large enough sample of measured values of N(X) to stars, it is possible to determine the distribution of the gas away from the Galactic plane by comparing observations of N(X) and its dependence on d, l, and b, to simple models of the expected distribution. It is most common to assume a plane-parallel model for the distribution with a simple exponential decline away from the plane with n(z) = n₀exp(−|z|/h) where n₀ is the mid-plane density, h is the exponential scale height, and z = d sin b is the distance away from the plane. In this case, the model column density integrated along the z-axis, N = ∫n(z)dz = n₀h[1−exp(−|z|/h)], can be compared to the projection of the observed column densities onto the z-axis, N(X)sin|b|. The simple model predicts N(X)sin|b| = n₀h[1−exp(−|z|/h)]. Plots of log[N(X)sin|b|] versus log|z| are expected to reveal an increase log[N(X)sin|b|] with log|z| until log|z| ∼ log h. For larger log|z|, log[N(X)sin|b|] should begin to approach log(n₀h). This simple exponential model assumes the gas has a smooth distribution. The irregularity in the actual gas distribution will produce a scatter of the data points about the simple exponential distribution model. The scatter produced by the gas irregularity will often be much larger than the scatter produced by the observational errors.

A problem with the absorption line method described above for determining the distribution of gas is that the sample of measurements must be relatively large in order to begin to deal adequately with the effects of the irregularities in the gas distribution and the effects of sample bias on the results. With only a few exceptions, such as Na i, Ca ii, and Ti ii, the absorption observations must generally be done in the ultraviolet region of the spectrum in order to record the relevant absorption lines of species that sample the different phases of the ISM. While results for H i (Diplas & Savage 1994), Ti ii (Lipman & Pettini 1995), Al iii (Savage et al. 1990), and Si iv, C iv, N v (Savage et al. 1997) have become available in recent years, the sample sizes for these earlier measurements have been relatively small, except for H i. With the successful operation of the Far Ultraviolet Spectroscopic Explorer (FUSE) satellite over the period 1999–2007, it has been possible to produce an absorption line data set tracing gas in the Galactic disk and halo for O vi toward a large number of Galactic (Bowen et al. 2008; Zsargó et al. 2003) and extragalactic objects (Savage et al. 2003). This provides the opportunity to inter-compare the extension of O vi away from the Galactic plane to the extension of other important interstellar species such as H i, Al iii, Si iv, C iv, and N v.

The species Si iv, C iv, N v, and O vi are particularly important for interstellar studies because they can be used to trace gas that cools relatively rapidly in the temperature range from (∼1–6) × 10⁵ K. This "transition temperature" gas differs from the more stable warm and hot gas phases with T ∼ 5000–10,000 K and T > 10⁶ K, respectively.¹ The transition temperature gas may arise in the conductive or turbulent interface regions between warm and hot gas or in cooling gas in a Galactic fountain or in old supernovae bubbles. The best tracer of transition temperature gas is O vi because it is unlikely that much O vi in the Milky Way disk and halo is created by photoionization in warm gas except in the vicinity of very hot white dwarfs with T > 50,000 K. This is because photons with E > 113 eV are required to convert O v into O vi. In contrast, Si iv and C iv be created with photons having E > 33 and >48 eV, respectively. Therefore, much of the Si iv and some of the C iv found in the ISM of the Milky Way could be produced by photoionization in warm gas. The ionizing radiation could be from hot Galactic stars, the intergalactic background radiation, or the radiation produced in warm–hot gas interfaces. N v, which requires E > 77 eV for its production, also is a good diagnostic of transition temperature collisionally ionized gas. However, the lower abundance of N compared to O makes N v more difficult to detect than O vi, particularly in complex stellar spectra.

Table 1 summarizes published values of the distributions of a number of interstellar species including H i, Ti ii, Al iii, Si iv, C iv, N v, O vi, and electrons derived from the N(X)sin|b| versus |z| method. The various columns in the table give the species, the number of Galactic and extragalactic sight lines used to derive h, σ_p(dex), a logarithmic measure of the patchiness or irregularity in the gas, n₀ ± σ, log(n₀h) ± σ, and h(kpc) ± σ. The sources of the measurements and comments are found in the footnotes to the table. The relatively small sample size of the measurements for Al iii, Si iv, and C iv from Savage et al. (1997) makes it unreliable to directly compare those scale height estimates with the much larger sample results for O vi, particularly given the finding that the distribution of O vi in the north Galactic polar region with b > 45° exhibits a 0.25 dex excess compared to the rest of the sky (Savage et al. 2003). Including or excluding this region when measuring h(O vi) affects the derived scale height considerably and presumably the same problem could occur for Si iv, C iv, and N v and other species. When trying to interrelate the Galactic distribution of different species, it is important to obtain measurements along the same lines of sight for the different species to minimize the possible impact of sample bias on the result.

All the results listed in Table 1 are based on ultraviolet (H i, Al iii, Si iv, C iv, and O vi) or optical (Ti ii) absorption line data with the exception of the results for the electron scale height which are derived from pulsar dispersion measures. Gaensler et al. (2008) have recently proposed that the electron galactic scale height is 1.89(+0.12, −0.25) kpc which is ∼2 times larger than values previously reported. This result is derived for 15 lines of sight to pulsars with |b| > 40°. However, the fit process does not allow for the irregular (patchy) nature of the WIM and the quoted value of the reduced χ² implies the fit is poor. In fact, the scale height derived by Gaensler et al. is based on an invalid fit to the pulsar observations as discussed in the Appendix. A revised fit for the 15 pulsars with |b| > 40° yields a smaller scale height, h = 1.41(+0.26, −0.21) kpc.

In anticipation of the major FUSE surveys of O vi, Savage et al. (2001a), determined column densities for Al iii, Si iv, and C iv toward 164 early-type stars in the Galactic disk and low halo from spectra obtained with the International Ultraviolet Explorer (IUE) satellite at a resolution of 25 km s⁻¹. O vi toward many of those stars were subsequently observed by FUSE at a resolution of 20 km s⁻¹ by Bowen et al. (2008). In this paper, we bring the two data sets together for the first time.

We have supplemented the IUE and FUSE disk/halo star effort with new measurements of Si iv and C iv absorption from the Space Telescope Imaging Spectrograph (STIS) archive of 25 extragalactic objects for which good quality O vi data exist from FUSE. These new observations are from B. P. Wakker et al. (2010, in preparation). Extragalactic observations are essential to derive the values of h for species where the largest values of |z| in the stellar sample are comparable to or smaller than h. The observations of the high ions and Al iii along with measures of N(H i) toward most of the stars by Diplas & Savage (1994) and toward the active galactic nuclei (AGNs) from Wakker et al. (2003) makes it possible for the first time with a large sample of measurements to directly and reliably compare the relative distributions of H i, Al iii, Si iv, C iv, and O vi away from the plane of the Milky Way. H i mostly traces the WNM with some contributions from diffuse clouds associated with the CNM. Al iii traces the WIM. Si iv and C iv trace a combination of photoionized gas and transition temperature gas. O vi mostly traces collisionally ionized transition temperature gas.

We have collected in Table 2 column densities for H i, Al iii, Si iv, C iv, and O vi for a large number of Galactic and extragalactic objects. The sources of these measurements for 109 stars in the Milky Way disk and halo are discussed in Section 2.1. The sources for these measurements for 25 AGNs and several stars in the Magellanic Clouds are discussed in Section 2.2. The basic requirement for an object appearing in Table 2 is that good measures of N(O vi) are available from the FUSE or the Copernicus satellites along with reliable measures or limits for Al iii, Si iv, or C iv from either IUE or Hubble Space Telescope (HST). Along extragalactic lines of sight, very few Al iii measurements exist. Therefore, it was necessary to include several directions with no measurements of O vi to study the behavior of Al iii at large |z|. The large size of our sample allows us to more reliably determine the extensions of the different species away from the galactic plane and to determine those extensions for the different ions for the same lines of sight. This reduces the effect of sample bias when inter-comparing the different scale height results.

2.1. Sight Lines in the Disk and Halo to Galactic Stars

2.2. Sight Lines Through the Galactic Halo to Extragalactic Objects

A major problem associated with the absorption method for the determination of the distribution of gas in the Galaxy is that measurements for a large number of lines of sight are required in order to average over the various sources of irregularity in the distribution of the gas. However, even when the object sample is large various sources of observational bias can influence the result. In this section, we discuss these issues.

In Figures 1(a) and (b), we display the distribution of the objects on the sky listed in Table 2. Figure 1(a) shows an Aitoff view of the whole sky with directions to stars shown as circles and directions to extragalactic objects shown as stars.

Zoom In Zoom Out Reset image size

Figure 1(b) shows a view of the object lines of sight from above the Galactic plane. In the left panels, the lines of sight to Galactic stars are shown as the solid lines and the lines of sight to the extragalactic objects are shown with dashed lines ending at a |z| distance of 8 kpc which corresponds to several times the high ion scale heights. In the right panels, the open circles display the projected positions of the survey stars.

In addition to directions through spiral arms and inter-arm regions, we also need to be aware of possible excess contributions to highly ionized gas from the Local Bubble (LB) surrounding the Sun to a distance of ∼100 pc, excess photoionized Al iii, Si iv, and C iv in pronounced H ii regions, excess shock heated Si iv, C iv, and O vi in young supernova remnants (SNRs), and excess O vi produced in the wind-shocked circumstellar regions around high-luminosity O and B stars.

Since most of the stars in our survey are at distances exceeding 1 kpc with median and average distances of 2.6 and 2.9 kpc, the effects of the contributions to the highly ionized gas absorption from the LB of hot gas is relatively small. The survey of O vi absorption toward white dwarfs of Savage & Lehner (2006) implies an average O vi column density of log N(O vi) = 13.05 at the ∼0.1 kpc boundary of the LB. The median and average values of log N(O vi) for all the stars listed in Table 2 are 14.15 and 14.11, respectively. Therefore, the LB contributes only ∼9% of the total O vi column density to the average star in the survey. Including or excluding the amount of O vi in the LB has only a marginal effect on the scale heights derived for O vi in this paper. The only detection of the Si iv and C iv ions in the LB is toward α Vir (HD 116658). The values of log N for Si iv, C iv, and O vi for this star listed in Table 2 imply the foreground LB contamination for Si iv and C iv will also hardly affect the derived scale heights for Si iv and C iv.

In order to investigate the effects of having the stellar or extragalactic lines of sight pass through or into pronounced H ii regions, we list in Column 7 of Table 2 the names of the prominent H ii regions associated with the survey stars. For the O stars, the associated H ii region names are taken from the O star catalog of Goy (1980). For the B stars, we performed our own independent search for the names of the associated H ii regions. Note that these names only include the prominent H ii regions found in the early Hα surveys of Sharpless (1957), Rodgers et al. (1960), and Gum (1955). Several lines of sight pass through the Vela SNR including the directions to HD74920 and HD 74711. When studying the general distribution of Al iii, Si iv, C iv, and O vi, we will not include the measurements for lines of sight contaminated by prominent regions of ionized gas associated with the background star.

Bowen et al. (2008) investigated the possible contributions of wind-shocked regions by looking for excess 0.25–1 keV X-ray emission in the regions surrounding the stars in their survey. Stars with R = 3 or 2 in Column 8 of Table 2 have relatively bright X-ray emission in their direction and may have an enhancement of O vi associated with a possible surrounding wind-blown bubble. R = 0 objects have weak surrounding X-ray emission and R = 1 objects represent an intermediate case. Note that many of the cases of objects with R = 2 and 3 lie in directions where there are prominent H ii regions associated with the star. For the extragalactic objects, we list R = 0 in all cases since extragalactic wind blown regions associated with these objects will produce redshifted absorption that is clearly distinguished from Galactic absorption.

Savage et al. (2003) found that the general distribution of Galactic thick disk O vi observed toward 100 extragalactic objects exhibits a 0.25 dex enhancement in the north Galactic polar region with b > 45°. Including or excluding this region of the sky when estimating the Galactic scale height of O vi has a substantial effect on the result (see Table 1). We need to be aware of this problem when determining the relative distributions of the highly ionized species.

The Galactic targets for the measurements reported in Table 2 are O and B stars that are relatively bright in the ultraviolet. Since ultraviolet extinction due to interstellar dust is large, the object sample is biased toward directions with much less than normal reddening per unit distance and with relatively small values of the average line of sight H i density. This is revealed by the average sample value in the mid-plane of n₀(H i) = 0.28 cm⁻³ (see Section 4) which is 3.9 times smaller than the overall Galactic H i mid-plane average of 1.1 cm⁻³ (Bohlin et al. 1978) in the vicinity of the Sun. Our disk sample, therefore, mostly avoids the denser interstellar regions containing cool and cold gas. The directions sampled with the measurements listed in Table 2 mostly trace gas in the lower density WNM rather than the CNM.

Examples of other factors that could influence the results include the presence or absence of IVCs along the line of sight and the possible overlap in velocity space of high-ionization absorption associated with HVCs and the gas in the thick Galactic disk. Savage et al. (2003) studied the possible impact of IVCs on the amount of O vi in the Galactic thick disk and found no evidence for an enhancement or deficiency of O vi in directions with IVCs compared to other directions. Separating high-ionization absorption associated with HVCs and the Galactic thick disk gas is often not easy because the HVC absorption can overlap the absorption by thick disk gas (Wakker et al. 2003; Sembach et al. 2003). Therefore, there may be extragalactic directions where the gas absorption we have identified as thick disk gas is contaminated by the lower velocity gas associated with an HVC. An example is the direction to Complex C where O vi HVC absorption often extends from −100 km s⁻¹ to −200 km s⁻¹ (Wakker et al. 2003). The thick disk absorption toward Complex C typically has been assumed to lie between ∼ −90 and +100 km s⁻¹ based on the appearance of the absorption in the velocity range of overlap. However, Complex C could have a lower velocity tail of absorption extending into the −90 to 0 km s⁻¹ velocity range. Since Complex C is situated ∼4–11 kpc away from the Galactic plane (Wakker et al. 2007), it could be interacting with and slowed down by the high-z extension of thick disk gas.

It is possible that there is an excess of transition temperature gas in the Galactic center. Therefore, lines of sight passing over or near the Galactic center might have more highly ionized gas than elsewhere as a consequence of a wind driven by the central regions of the Galaxy (Keeney et al. 2006).

A major goal of this study is to evaluate similarities or differences in the distribution of H i, Al iii, Si iv, C iv, and O vi away from the Galactic plane. Although we have assembled a relatively large database from the perspective of ultraviolet absorption line standards, the sample is small given the complexity of the range of phenomena occurring in the Galaxy that might influence the distribution of the gas (see Section 3). We will therefore need to be careful in deciding what observations should be included or excluded from the analysis in order to remove the possible bias introduced by various types of ISM phenomena.

In Figure 2, we display log[N(X)sin|b|] versus log|z| for H i, Al iii, Si iv, C iv, and O vi for all the observations listed in Table 2. Detections are denoted with the open or filled circles. Upper or lower limits are denoted with the downward or upward pointing open or filled triangles. Open symbols are for objects associated with pronounced nebular H ii regions or SNRs or strong X-ray 0.25 to 1 keV X-ray emission (R = 2 or 3 in Column 8 of Table 2). The open symbols are therefore for directions where H ii region photoionization may enhance the column densities of the transition temperature ions observed in this survey and where the X-ray emission may be revealing wind-shocked regions around the observed stars. The extragalactic measurements plotted on the right-hand side of the individual panels are ordered according to Galactic latitude with b = −90° to the left and b = 90° to the right. Extragalactic objects with b > 45° and the star vZ 1128 with b = 7869 and z = 10 kpc are also plotted with open circles because they lie in the north Galactic polar direction where the projected, log(N sin|b|), column density of O vi exhibits a 0.25 dex excess.

Zoom In Zoom Out Reset image size

Most of the observed scatter in the panels of Figure 2 is produced by the irregularity in the ISM rather than observational errors. In each panel of Figure 2, we show the behavior of log{n₀h[1−exp(−|z|/h)]} plotted against log|z| describing a smooth exponential gas distribution with the mid-plane densities listed in Table 3 and scale heights of 0.1, 0.3, 1, 3, and 10 kpc. A visual inspection of the different panels reveals the increase of implied scale heights through the sequence H i to Al iii to the higher ions.

Since we are interested in the distribution of the transition temperature ions in the general ISM, we should avoid particular directions if there is evidence for an excess of the transition temperature ions associated with phenomena local to the star. A simple visual inspection of the various panels of Figure 2 reveals that the open symbols do lie higher than the filled symbols in the plots of log[N(X)sin|b|] versus log|z| for Al iii, Si iv, C iv, and O vi. Toward the Galactic stars the effect is strongest for Al iii and Si iv. Since these two ions can be created by photoionization with photon energies greater than 18.8 and 33.5 eV, respectively, it is likely we are seeing an excess of Al iii and Si iv in the photoionized gas of the H ii regions associated with the survey stars.

After a consideration of the various sources of bias, we conclude the estimate of the Galactic scale heights of the transition temperature gas in the general ISM should omit lines of sight passing through the following regions: (1) prominent H ii regions; (2) SNRs; (3) objects with surrounding X-ray bright emission (R = 2 or 3 in Table 2); (4) extragalactic directions toward the north Galactic pole with b > 45° along with the measurement for vZ 1128. We will refer to the remaining sample of objects as the "restricted sample."

The visual inspection of Figure 2 reveals that the measurements for objects in the restricted sample and in the full sample exhibit a large dispersion about the prediction of a smooth simple exponential distribution for the gas. An improved gas distribution model is one where the model is allowed to have a degree of irregularity or patchiness in the gas distribution but with an exponential decline in the number of clouds or structures sampled with distance away from the Galactic plane. Therefore, when comparing the observations to the smooth exponential model, we include in the fitting process the patchiness parameter, σ_P, to describe the irregular distribution of the gas in the improved model. The patchiness parameter is adjusted to make the quality of the fit reasonable. This technique follows the method first introduced by Savage et al. (1990) and subsequently adopted in many of more recent derivations of scale heights from absorption line observations (see Table 1).

Values of the patchiness parameter, log(n₀h), mid-plane density, scale height, and their errors derived from the observations are listed in Table 3 for the restricted sample. The best fit of the observed values of N(X)sin|b| to the exponential scale height model, N(X)sin|b| = n₀(X)[1 − exp(−|z|/h(X)], was performed through a χ² minimization procedure applied in log[N(X)sin|b|]–log|z| space. For the stellar lines of sight, the logarithmic error for each observed value of log[N(X)sin|b|] is taken to be the quadrature addition of the logarithmic column density error, σ_N, the logarithmic distance error, σ_d, and the logarithmic patchiness parameter, σ_P. Therefore, σ = (σ_N²+σ_d²+σ_P²)^1/2. During the minimization, σ_P is varied so that the value of the reduced χ², χ_ν² = χ²/(N−2) = 1.0. Here, N refers to the number of data points and 2 to the number of fitting parameters. By varying σ_P to obtain χ_ν² = 1.0, we achieve an acceptable fit for our exponential scale height model with an irregular (patchy) cloud distribution, where σ_P is a measure of the patchiness. If the patchiness parameter is not included in the χ² minimization process, the resulting values of χ_ν² (minimum) are very large and the resulting best-fit values of h and log(n₀h) are highly dependent on the most accurate column density measurements. When most of the scatter about the smooth exponential model is due to the irregularity of the ISM, it is incorrect to give too much weight in the fitting process to the most accurate measurements since the scatter in the measurements reveal the basic inadequacy of a simple smooth model.

In the fitting procedure, we combine in quadrature the column density (y-axis) errors of the fit with the distance (x-axis) errors. As discussed by Bowen et al. (2008), this is an acceptable way to include the distance errors in the minimization process for measurements made at heights |z| < h. For the extragalactic lines of sight, we do not include the distance error since for |z| ≫ h the error in distance does not influence the fit. Our procedure for including the distance error is not strictly valid for H i and Al iii where a number of stars have |z| > h. However, we do not believe this will have much of an impact on the derived value of the scale height because the value of σ_P required to obtain χ_ν² = 1.0 is considerably larger than the average values of σ_N and σ_d. This implies that the variation of the fit about the smooth exponential model is dominated by the irregular distribution of the gas and not by column density or distance errors.

The ±1σ errors derived for the parameters listed in Table 3 were obtained by determining the χ² (minimum) +2.3 contour in h–log(n₀h) space. Here, 2.3 is the increase in χ² corresponding to the 68% confidence or ±1σ contour for a two-parameter fit. For detailed discussions see Lampton et al. (1976). Note that in some of the earlier applications of this method the 68% contour was incorrectly determined from χ² (minimum) +1.0. This expression is only correct when the number of fit parameters is 1. The best fit results from Table 3 are shown in Figure 2 with the heavy lines. The ±1σ errors in the scale height are displayed with the dashed lines.

For the restricted sample, the transition temperature high ions (O vi, C iv, and Si iv) are found to have scale heights ranging from 2.6 to 3.6 kpc with errors of 0.5 to 1.2 kpc. The scale height of O vi, h = 2.6 ± 0.5 kpc appears to be marginally smaller than the scale heights for Si iv and C iv, h = 3.2(+1.2, −0.6) kpc, and 3.6(+1.0, −0.8) kpc, respectively. The scale heights for the transition temperature ions are ∼3 times larger than for Al iii, a tracer of the WIM, with h = 0.90(+0.62, −0.33) and ∼12 times larger than for H i, mostly a tracer of the WNM with h = 0.24 ± 0.06 kpc. Note that the comparison of the relative sizes of these scale heights is more reliable than the absolute values of the scale heights. Sample bias issues will affect the absolute values of the scale heights. However, the bias problems cancel to first order when evaluating the relative values of the scale heights because the samples are approximately the same for all the scale heights derived for the restricted sample.

We considered the effects of including measurements of upper and lower limits on the estimates of the scale heights. This was done by assuming the limit is a measurement with a relatively large column density error of σ_N = 0.2 dex. The inclusion of the limits did not affect the derived value of the fit parameters.

The scale heights for the transition temperature ions in this paper can be compared to the earlier results listed in Table 1. Because of the larger sample size and better control over sample bias issues, the new scale heights are better suited for a comparison of any scale height differences from one species to the next. In the case of O vi, the value derived in this paper of 2.6 ± 0.5 kpc is somewhat smaller than h = 3.2 ± 0.8 kpc derived by Bowen et al. (2008) using measurements from the FUSE O vi disk survey combined with the FUSE O vi halo survey of Savage et al. (2003) for extragalactic objects in the southern Milky Way with b < −20°. Most of the difference between these two measures of h(O vi) arises from the smaller value estimated for the mid-plane density of O vi for the Bowen et al. survey compared to this survey (n₀ = 1.33 × 10⁻⁸ versus 1.64 × 10⁻⁸ cm⁻³). The Bowen et al. survey includes many more stars at d < 1 kpc than the survey of this paper. When considering disk stars at these smaller distances, Bowen et al. included a correction for O vi absorption in the LB. Evidently, the mid-plane density of O vi is somewhat larger when averaging over the larger distances toward the disk stars used in our survey sample.

The values of the patchiness parameter listed for the various ions in Table 3 provide important information about the degree of irregularity of the various gas phases sampled in the ISM. The five species sampled have patchiness parameters ranging from 0.172 (H i) to 0.273 (C iv) dex. Therefore, the WNM, WIM, and transition temperature gas sampled by these measurements are irregularly distributed with degrees of patchiness ranging from 49% to 87%. The patchiness parameter measures a complex combination of irregularities caused by intercepting absorbing structures having different size, shape, properties, and line of sight distributions. Note that when deriving the patchiness parameter through the scale height χ² analysis method we assumed the level of patchiness is the same for lines of sight to stars in the disk and for the extragalactic lines of sight. In the following section, we will see that there is more irregularity in the behavior of ion ratios along disk lines of sight than along extragalactic lines of sight.

The ∼0.25 dex excess of log[N(O vi)sin|b|] toward the north Galactic pole in directions with b > 45° found by Savage et al. (2003) is also seen toward the extragalactic objects in the Si iv and C iv ions (see panels in Figure 2 and Table 4). In Table 4, the difference between 〈log[N(X)sin|b|]〉 for the samples with b < 45 ° versus b > 45° is 0.21, 0.19, and 0.23 dex for O vi, C iv, and O vi, respectively. For our sample, all three of these high ions have a similar average excess of ∼0.21 dex for b > 45°. Evidently, the processes responsible for the origin of the excess are not significantly affecting the relative behavior of the three high ions. We note that the excess high-ionization absorption seen toward the nine AGNs at b > 45° is also seen toward the globular cluster star vZ 1128 at z = 10 kpc in the direction l = 425 and b = 787. For vZ 1128 the values of log[N(X)sin|b|] for the high ions are as large as the largest measured for any extragalactic object. This implies that the excess high-ionization absorption for directions with b > 45° occurs at z < 10 kpc.

The north/south Galactic asymmetry in the distribution of interstellar gas as seen from the solar position in the Galaxy will influence scale height derivations for many species. In the Appendix, we discuss how the asymmetry has likely affected the electron scale height reported by Gaensler et al. (2008). In that discussion, we note that the asymmetry is also seen in FUSE measures of S iii and Fe iii which mostly trace gas in the WIM.

The different models for the origins of the transition temperature plasma in the Galactic disk and halo make different predictions for the ionic ratios of Si iv, C iv, and O vi. In Figure 3, we display the values of log[N(C iv)/N(O vi)], log[N(Si iv)/N(O vi)], and log[N(Si iv)/N(C iv)] as a function of log|z| and log d. The restricted sample is plotted with the filled circles. Stars in prominent H ii regions, SNRs, and with strong associated surrounding X-ray emission, and extragalactic objects with b > 45° and vZ 1128 are plotted as open circles. Lower and upper limits and data points for measurements with large error bars (σ_N ≥ 0.26 dex) are not plotted to reduce the confusion. Table 4 gives medians, averages and standard deviations for the logarithmic ionic ratios for various samples of objects as defined in the footnotes to the table.

Zoom In Zoom Out Reset image size

A visual inspection of the various panels in Figure 3 shows log[N(C iv)/N(O vi)], log[N(Si iv)/N(O vi)], and log[N(Si iv)/N(C iv)] are often larger for objects with suspected H ii region or stellar wind contamination (open circles) versus directions where the line of sight does not pass through these contaminated environments. The behavior justifies the omission of those directions when determining the scale heights in Section 4. The column densities of Si iv and C iv and sometimes O vi are clearly enhanced for directions passing through prominent H ii regions.

Examples of lines of sight with very large ratios of the Si iv to C iv column densities include HD 101131, HD 101205, HD 101298, HD 101413, and HD 101436 with 〈log[N(Si iv)/N(C iv)]〉 = −0.02 dex which is 0.57 dex larger than for the restricted sample average. All five of these stars are situated in the pronounced H ii region Gum 42 with the exciting cluster IC 2944. Deep Hα imaging of this region of the sky around (l, b) = (295°, −17) is discussed by Georgelin et al. (2000). The lines of sight to HD 101190 and HDE 308803 also pass toward Gum 42. However, the C iv line is saturated for HD101190 and the Si iv line is saturated for HDE 308803. The limit log[N(Si iv)/N(C iv)] ≥ −0.17 for HDE 308803 is consistent with the results for the five stars discussed above.

Other lines of sight with large Si iv to C iv column density ratios are for HD 152623 and HD 152723 which are located in Gum 56 and HD 153426 in Gum 57a. These three stars have 〈log[N(Si iv)/N(C iv)]〉 = −0.07 dex which is 0.52 dex larger than for the restricted sample average.

Stars in the Carina nebular complex including HD 93129a, HD 93205, HD 93206, HD 93222, HD 93146, HD 93250, CPD –59 2603, HDE 303308 also have enhanced ratios of Si iv to C iv. The Si iv lines of four of these stars are so strong that they are saturated and therefore only permit the derivation of a lower limit to N(Si iv). The average lower limit to log[N(Si iv)/N(C iv)] for the four stars is ≥−0.20. The other four stars with secure measures of N(Si iv) and N(C iv) have 〈log[N(Si iv)/N(C iv)]〉 = −0.26. The stars in the Carina region have Si iv to C iv column density ratios enhanced by an average of >0.39 and 0.33 dex compared to the restricted sample average.

Starlight photoionization of Si iv and to some extent C iv by the O stars in these H ii regions is likely responsible for the excess Si iv and C iv and the large ratio of Si iv and C iv. The O stars also have strong winds that can create excess O vi and C iv in the surrounding wind-shocked regions.

For the restricted sample, there is a considerable dispersion in the measured values of the column density ratios particularly for N(C iv)/N(O vi) and N(Si iv)/N(O vi). The dispersion significantly decreases when moving from lines of sight to stars in the Galactic disk and low halo with log|z| and log d < 1 to the extragalactic lines of sight with log|z| and log d > 1. This is well illustrated for the values listed in Table 4 for the median, average, and standard deviation (stdev) of the ionic ratios when moving from the results for Galactic stars using the restricted sample to the results for the extragalactic restricted sample. In particular, 〈log[N(Si iv)/N(C iv)]〉 = −0.55 ± 0.21(stdev) in the Galactic star restricted sample and −0.60 ± 0.11(stdev) in the extragalactic restricted sample with b < 45°. The dispersion as measured by the stdev decreases from 0.21 to 0.11 dex. 〈log N[(C iv)/N(O vi)]〉 = −0.29 ± 0.35(stdev) in the Galactic star restricted sample and −0.15 ± 0.17 in the extragalactic restricted sample. In this case, the stdev decreases from 0.35 to 0.17 dex.

The logarithmic dispersion in the ion ratios as measured by the standard deviation is approximately two times larger for log N[(C iv)/N(O vi)] and log[N(Si iv)/N(O vi)] than for log[N(Si iv)/N(C iv)] for both the Galactic restricted sample (stdev = 0.35 and 0.38 versus 0.21 dex) and for the extragalactic restricted sample (stdev = 0.17 and 0.19 versus 0.11 dex). Si iv and C iv are much better correlated with each other than C iv versus O vi or Si iv versus O vi. The relatively good correlation between Si iv and C iv has been known for many years (see Pettini & West 1982; Savage et al. 1994) suggesting the production and destruction of these two species are controlled by regulating processes in the general ISM of the disk and halo. However, it is also clear that strong increases occur in the Si iv to C iv ratio in dense H ii regions where an excess of Si iv is created by starlight photoionization.

The average values of log N(Si iv)/N(C iv), log[N(C iv)/N(O vi)], and log[N(Si iv)/N(O vi)] listed in Table 4 for the restricted galactic and restricted extragalactic samples change for the sample of sight lines to stars in the Galactic thick disk to the extragalactic sample by −0.05, 0.14, and 0.09 dex, respectively. These changes are consistent with the expectations for species with the somewhat different scale heights of 2.6 ± 0.5, 3.2(+1.2, −0.6), and 3.6 (+1.0, −0.8) kpc, for O vi, Si vi, and C iv, respectively (see Table 3). Therefore, the differences in the scale heights for Si iv, C iv, and O vi listed in Table 3 are consistent with the z distribution of the ionic ratios of the three species. This gives added support to the reality of the scale height differences found among the transition temperature ions.

Si iv, C iv, and O vi extend to roughly similar distances (h ∼ 3 kpc) away from the Galactic plane. The high ion extension is ∼3 times the warm ionized gas extension as traced by Al iii. The high ion extension is ∼12 times the extension of the WNM as traced by H i. Since the extensions of these species have been measured for the same lines of sight, we can be reasonably confident in the validity of the relative behavior of the extensions from one species to the next. The new scale heights from this paper are displayed in Figure 4 with the filled circles. The open circle for the uncertain N v scale height from Savage et al. (1997) is based on a small sample of extragalactic lines of sight.

Zoom In Zoom Out Reset image size

For an isothermal atmosphere at temperature T in hydrostatic equilibrium, the Galactic scale height is given by h = kT/〈m〉g(|z|), where 〈m〉 is the average mass per particle and g(|z|) is the gravitational acceleration toward the disk. For |z| in the range of 1–10 kpc in the solar neighborhood g(|z|) ∼ 10⁻⁸ cm s⁻² (Kalberla & Dedes (2008), see their Section 6.1 and Figure 13). For an isothermal Galactic thick disk in hydrostatic equilibrium, a scale height of 3 kpc would require T ∼ 0.8 × 10⁶ K for a fully ionized plasma with solar abundances with 〈m〉 = 0.73m_H. The required temperature is 4–8 times larger than the temperature where O vi and C iv peak in abundance in collisional ionization equilibrium. The existence of the transition temperature ions in the Galactic thick disk therefore requires nonthermal support mechanisms (turbulence, cosmic ray pressure, magnetic pressure) or the presence of a flow process (the Galactic fountain) driven by gas with T > 0.8 × 10⁶ K.

It is surprising that Si iv and C iv may have slightly larger scale heights than O vi since O vi is mostly associated with production in collisionally ionized gas, while Si iv and C iv are associated with both photoionized and collisionally ionized gas. In a simple Galactic fountain flow where the plasma cools and recombines during the return flow to the disk, the scale height of the higher temperature ions (O vi) should exceed that for the lower temperature ions (C iv and Si iv).

A hybrid model describing different relative extensions of Si iv, C iv, and O vi into the halo was suggested by Ito & Ikeuchi (1988) who proposed a Galactic fountain flow of cooling hot (10⁶ K) gas to create collisionally ionized Si iv, C iv, and O vi in the disk and low halo in combination with photoionization by the UV extragalactic background and Galactic EUV sources to produce an enhanced production of Si iv and C iv at high |z|. Another hybrid model capable of explaining the somewhat more extended |z| distribution of Si iv and C iv compared to O vi is that of Shull & Slavin (1994) who combine the collective effects of the production of hot gas in cooling SN bubbles at low |z| with the turbulent mixing layers created by Galactic chimneys at large |z|. However, because of the large number of adjustable parameters in such hybrid models, it is difficult of devise tests to determine which models best describe the dominating processes operating in the transition temperature gas. Hydrodynamical simulations have begun to provide important insights about the effects of supernovae explosions and hot stellar winds on gas in the Galactic disk and low halo (de Avillez & Breitschwerdt 2005). However, the modeling of the behavior of the transition temperature plasma is difficult to implement because nonequilibrium processes strongly influence the ionization particularly for Si iv and C iv.

The large decrease in the dispersion of the high ion ratios from sight lines emphasizing disk gas to sight lines emphasizing halo gas is probably the result of moving from highly disturbed regions of the Galactic disk where the primary energy injection and ionization processes are occurring through supernova explosions, stellar winds, and O star photons to the more quiescent halo regions where direct localized energy injection processes are less common. Gaseous structures at large distances away from the Galactic plane may be larger and longer-lived than structures in the disk. This might also explain the decrease in dispersion. The possible existence of large and very long-lived supernova remnants in the halo as modeled by Shelton (1998, 2006) might produce halo gas with relatively uniform high ion ratios.

The much better correlation between the column densities of Si iv and C iv than between C iv and O vi for the gas in the restricted sample of the general ISM is interesting. As we move from Si iv to C iv to O vi, it becomes more and more difficult to create the ion by starlight photoionization. Therefore, we would expect starlight photoionization to strongly affect the behavior of Si iv along a given line of sight. Indeed lines of sight extending through pronounced H ii regions in the disk have large Si iv to C iv ratios as expected. If the Si iv in the general ISM is also created by starlight photoionization, one might also expect to see large variations in the Si iv to C iv ratio in that medium as the ratio of the ionizing photon density to the gas density changes from place to place along a line of sight. The absence of large variations suggests that the ionization of much of the Si iv in the general ISM may not be due to starlight photoionization. A high-resolution (3.5 km s⁻¹) and good signal-to-noise ratio (S/N, 22–38) study of the high ionization line absorption along the 4 kpc line of sight to HD 167756 in the inner Galaxy by Savage et al. (1994) revealed the presence of Si iv and C iv absorbing components with properties well explained by their production in warm–hot gas conductive interfaces where the C iv occurs in cooling collisionally ionized interface gas and the Si iv is produced by the He⁺ ionizing radiation emitted by the interface. In those components, log[N(Si iv)/N(C iv)] = −0.54 close to the value −0.55 dex found for the general ISM of the disk in this paper. In this case, the physics of the interface processes provides the formation and destruction regulation that yields a standard value for the column density ratio.

Extending such high-resolution studies to other lines of sight in the Galactic disk will be important in order to determine if the behavior of the highly ionized gas toward HD 166756 is common to that found for other lines of sight through the Galactic disk. In Figure 5, we combine the results for ∼3.5 km s⁻¹ resolution observations of the Si iv and C iv absorption components studied in detail along five lines of sight including HD 116852 (Fox et al. 2003), HD 119608 (Sembach et al. 1997), HD 167756 (Savage et al. 1994), HD 177989 (Savage et al. 2001a), and HD 215733 (Fitzpatrick & Spitzer 1997). A total of 20 components containing C iv are detected. For each of the 17 components also detected in Si iv, we plot N(C iv) versus log[N(Si iv)/N(C iv)] in Figure 5. The five lines of sight which have a combined extent of 20.5 kpc mostly sample gas in the Galactic disk and low halo, although in two cases the paths extend through special environments. The sight line to HD 177989 samples gas in the Scutum supershell at v ∼ 45 km s⁻¹. The sight line to HD 119608 samples gas in the supershell over the Sco–Oph association which is associated with Galactic radio loop IV and the North Polar Spur. The three absorption components associated with these supershells are marked SSS and NPS in Figure 5. These supershell components have small values of log[N(Si iv)/N(C iv)] = −0.70, −0.92, and −0.99. In the remaining 14 components which trace less disturbed gas in the general ISM of the disk and low halo, the values of log[N(Si iv)/N(C iv)] range from −1.63 to +0.15. The four larger column density components in this group have log[N(Si iv)/N(C iv)] between −0.75 and −0.63. Nine of the 10 low column density components with N(C iv) < 2 × 10¹³ cm⁻² have log[N(Si iv)/N(C iv)] ranging from −0.62 to +0.15. The weak component with N(C iv) = 1.05 × 10¹³ cm⁻² and log[N(Si iv)/N(C iv)] = −1.63 ± 0.40 from Savage et al. (2001a) has C iv and Si iv well aligned in velocity but the very different b values (b(Si iv) = 4.3 ± 2.8 and b(C iv) = 15.9 ± 1.8 km s⁻¹) imply the two species do not co-exist in the same absorbing medium.

Zoom In Zoom Out Reset image size

There is a large spread in the Si iv to C iv ratio in the weak components. The weak components with values of log[N(Si iv)/N(C iv)] = 0 ± 0.2 likely arise in photoionized gas as confirmed by the narrow C iv line widths (Fox et al. 2003). In Figure 5, the four components with narrow C iv line profiles implying gas with T < 5 × 10⁴ K are plotted as the filled squares. See Fox et al. (2003) for additional discussions.

Our brief examination of the results for five lines of sight implies that even though Si iv and C iv seem relatively well regulated when viewed at 25 km s⁻¹ resolution, the situation looks more complex at 3.5 km s⁻¹ resolution. While there may be a regulation process affecting the stronger components, a wide range of ionic ratios are found in the weak components. A more extensive high-resolution study of Si iv and C iv absorption would help to determine the relative frequency of occurrence of strong components with log[N(Si iv)/N(C iv)] near −0.6 dex versus the weaker components exhibiting a large spread in the values of log[N(Si iv)/N(C iv)].

Toward the north Galactic polar region in directions with b > 45°, there is a 0.20 dex excess in the amounts of O vi, C iv, and Si iv. This excess likely occurs in the thick disk of the Galaxy with z < 10 kpc since a similar excess is found along the line of sight to the star vZ 1128 in the Globular cluster M3 at z = 10 kpc. The excess is possibly associated with a high-z extension of the local superbubble traced by radio loops I and IV and powered by the Sco–Cen OB association. Other possibilities include an excess produced by a low-velocity tail of absorption associated with HVC Complex C or an excess that traces an outflow of transition temperature plasma from the center of the Galaxy. Origins occurring beyond the Galactic thick disk appear ruled out by the vZ 1128 measurements.

We have assembled absorption line column densities of H i, Al iii, Si iv, C iv, and O vi tracing gas in the Galactic disk and halo toward 109 Galactic stars, 25 AGNs, and 5 Magellanic Cloud stars. New column densities for C iv and Si iv are presented for the AGNs. The measurements are drawn from observations with the IUE, FUSE, and HST observatories. The observations are used to determine the relative extensions away from the Galactic plane of a tracer of the WNM (H i), the WIM (Al iii), and tracers of warm photoionized and transition temperature gas (Si iv, C iv, and O vi). Our sample size is much larger than for previous studies of the relative extensions of these species into the halo. The large sample size for absorption line measurements allows an evaluation of the effects of various types of galactic phenomena that can seriously bias the results for small samples. The results are as follows.

• 1.  
  The transition temperature ions traced by Si iv, C iv, and O vi exhibit similar extensions away from the Galactic plane with exponential scale heights h ∼ 3 kpc. The scale height for O vi (h = 2.6 ± 0.5 kpc) appears to be marginally smaller than the Si iv and C iv scale heights of h = 3.2(+1.2, −0.6) and h = 3.6 (+1.0, −0.8) kpc, respectively.
• 2.  
  The transition temperature plasma is ∼3 times more extended than the WIM with h = 0.90(+0.62, −0.33) kpc as traced by Al iii.
• 3.  
  The transition temperature plasma is ∼12 times more extended than the WNM with h ∼ 0.24 ± 0.06 kpc as traced by H i.
• 4.  
  The marginally smaller scale height of O vi compared to Si iv and C iv is consistent with the predictions of several hybrid halo gas models that incorporate collisional ionization in cooling transition temperature gas with photoionization by the extragalactic background or the turbulent mixing of warm and hot gas in Galactic supershells.
• 5.  
  All of the species studied exhibit an irregular (patchy) distribution about a simple uniform plane-parallel exponential model for the extension of the gas away from the Galactic plane. The level of the irregularity as measured by the patchiness parameter (σ_P) ranges from 0.172 (H i) to 0.273 (C iv) dex. The measured dispersion of the observations about the smooth exponential model is dominated by the irregular distribution of the gas and not by column density or distance errors.
• 6.  
  There is a factor of ∼2 decrease in the logarithmic dispersion of the ionic column density ratios for the transition temperature plasma in moving from lines of sight in the Galactic disk to lines of sight through the entire halo.
• 7.  
  The dispersions for log[N(C iv)/N(O vi)] and log[N(Si iv)/N(O vi)] in the disk and halo are ∼2 times that for log[N(Si iv)/N(C iv)]. Si iv and C iv are relatively tightly correlated with each other with 〈log[N(Si iv)/N(C iv)]〉 = −0.55 ± 0.21 (stdev) and −0.60 ± 0.11 (stdev) in the disk and halo, respectively.
• 8.  
  The assembled set of observations reveals lines of sight along which the high ions have properties deviating significantly from the average results discussed above. These include lines of sight through prominent H ii regions where there is a strong enhancement in the amount of Si iv compared to C iv.
• 9.  
  Toward the north Galactic polar region in directions with b > 45°, there is a 0.20–0.25 dex excess in the amounts of O vi, C iv, and Si iv. This excess likely occurs in the thick disk of the Galaxy with z < 10 kpc since a similar excess is found along the line of sight to the star vZ 1128 in the globular cluster M3 at z = 10 kpc.
• 10.  
  In the Appendix, we discuss a revision to the Galactic electron scale height recently proposed by Gaensler et al. (2008) based on electron dispersion measures. The revised result is strongly influenced by the north/south Galactic asymmetry found for many interstellar species. More dispersion measures are needed for pulsars with b < −45° to determine the actual size of the electron asymmetry and its effect on the inferred value of the electron scale height.

The results reported here used observations from a number of NASA observatories and instruments including the Copernicus satellite, IUE, FUSE, HST-GHRS, and HST-STIS. We appreciate the dedication and hard work required to plan, build, and successfully operate these facilities that span nearly four decades of space astronomy. We thank the referee for providing a number of suggestions that helped to improve the manuscript. The Si iv and C iv measurements obtained toward the extragalactic objects were mostly obtained with STIS aboard the NASA/ESA HST which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-25655. The O vi results were obtained by the NASA-CNES FUSE mission operated by Johns Hopkins University. Financial support to US participants has been provided by NASA contract NAS5-32985.

Gaensler et al. (2008) have reported that the electron Galactic scale height based on the dispersion measures of pulsars is a factor of ∼2 larger than previously determined. The Gaensler et al. analysis techniques illustrate a number of problems associated with obtaining reliable estimates of the |z| distribution of interstellar species when the gas distribution is irregular. Gaensler et al. collected electron dispersion measures, DM (cm⁻³ pc), for pulsars with reliable distances estimated either from parallaxes or from the pulsar locations in globular clusters. For globular clusters with more than one pulsar, they adopted the mean dispersion measure for the cluster and considered the mean as a single measurement in the scale height analysis. Their final sample included 51 lines of sight into the Galactic disk and halo. They constructed plots of log(DMsin|b|) versus log|z| and considered various ways of fitting the observations in order to reduce the effects of the contamination of the lines of sight by pronounced Galactic H ii regions in the disk. They decided to limit the fit to the 15 pulsars at relatively high latitude with |b| > 40° when determining their preferred scale height estimate. Since the dispersion measure errors were very small compared to the distance errors, Gaensler et al. performed the fit using |z| as the dependent variable. They obtained h = 1.89 (+0.12, –0.25) kpc which is approximately a factor of 2 larger than previous estimates. See Table 1 for the other fit parameters. There are several problems with the Gaensler et al. fit to the pulsar dispersion measures. Five of their six data points with |z| > 1 kpc lie an average of ∼0.08 dex (∼20%) below the fitted curve (see the blue dashed curve and blue data points in Figure 1 of Gaensler et al. 2008). The sixth data point is on the fitted curve. This suggests a systematic problem associated with the fit process. By using |z| as the dependent variable, fits to the observations cannot be made if any of the data points at large |z| violate the condition DMsin|b| (observation) > DMsin|b| (model). If that condition is violated, the difference between the observation and the model treating |z| as a dependent variable, |z| (observed) – |z| (model), is undefined (i.e., forbidden). This constraint creates a strong bias in the fitted curve causing the derived scale height to be too large. In addition, the reported fit χ_ν² = 5.2 is large implying the fit is of low quality even though an additional 10% error was added to the errors on |z| for each measurement. The large value of χ_ν² is most likely the result of the patchy nature of the ISM which should be built into the model or into the fitting procedure. The Gaensler et al. scale height fit is invalid.

We discussed our concerns about the new electron scale height with Dr. Bryan Gaensler. He provided us with the values of DM and distance for the 15 pulsars with |b| > 40°. We used our software for estimating the electron scale height by considering DMsin|b| to be the dependent variable and followed the fitting procedures used for the absorption line measurements presented in this paper. The logarithmic errors for DM, distance, and the patchiness parameter were combined in quadrature to produce a total error including the patchiness σ (total) = [σ_DM²+σ_d²+σ_P]^1/2. The distance errors for pulsars in the globular clusters were not included in the total error because at their large z distances, DMsin|b| achieves its maximum value. Again σ_P was adjusted in order to achieve an acceptable fit with χ_ν² = 1.0. We obtain for σ_P = 0.06 dex, h = 1.41 (+0.26, −0.21) kpc and log(DMsin|b|) = 1.34 ± 0.04, which corresponds to log[N(e)sin|b|] = 19.83 ± 0.04. The other fit parameters are listed in Table 1. The fit method closely parallels the approach we have taken in this paper. We believe this second estimate of the electron scale height is more reliable than that given in Gaensler et al. because it addresses the basic fact that the medium being studied is inhomogeneous even for relatively high latitude directions. By using DMsin|b| as the dependent variable, the fit process does not forbid fits when DMsin|b| (observed) exceeds DMsin|b| (model) at large |z|.

The revised electron scale height of 1.41 (+0.26, −0.21) kpc from the pulsar dispersion measures is in reasonable agreement with h = 0.90 (+0.62, −0.33) kpc derived by us for Al iii which should also trace the WIM. The Al iii scale height fit required σ_P = 0.256 dex to achieve χ_ν² = 1.0 which is considerably larger than the value 0.07 dex required for the pulsar dispersion measures.

The revised electron scale height of h = 1.41 (+0.26, −0.21) kpc may contain substantial systematic errors because of the small size of the data sample and the biased sampling of the high latitude sky. Ten of the 15 measurements at |b| > 40° lie in the north Galactic sky. Five measurements have b > 70°. Five measurements lie in the southern Galactic sky with b between −40° and −60°. The high latitude north and south Galactic sky is asymmetrical in the distribution of many species including H i, Si iv, C iv, and O vi. In fact, for the analysis of Si iv, C iv, and O vi in this paper, the northern Galactic sky with |b| > 45° was excluded from the scale height derivation because all three species exhibited an excess of ∼0.2 dex or 46% in the values of log N(ion)sin|b| toward extragalactic objects compared with the rest of the sample. In the case of H i, the asymmetry is in the opposite sense. Nine AGNs with b > 45° have 〈log[N(H i)sin|b|]〉 = 20.16 ± 0.19(stdev) while nine with b < −45° have 〈log[N(H i)sin|b|]〉 = 20.25 ± 0.15(stdev). The north/south asymmetry in H i is 0.09 dex or 23% when averaged over nine sight lines in the north and in the south. Possible origins of the high ionization column density asymmetries are discussed in Savage et al. (2003).

A north/south Galactic asymmetry in the WIM is confirmed from a study of Fe iii and S iii absorption with FUSE toward AGNs (B. P. Wakker et al. 2010, in preparation). Fe iii is produced and destroyed in photoionized gas with photon energies between 16.19 and 30.65 eV. For S iii, the range is 23.34 and 34.79 eV. While the Fe iii observations are affected by the presence of Fe in dust, sulfur is mostly found in the gas phase (Savage & Sembach 1996). The existence of dust grain cores in the WHIM has been revealed via studies of S iii and Al iii (Howk & Savage 1999). While some Fe iii and S iii may be produced in collisionally ionized gas, photoionization is likely the primary production mechanism for these ions.

The column density of Fe iii in the Galactic thick disk was obtained from FUSE observations of the unsaturated line at 1122.208 Å with an oscillator strength f = 0.0788. The S iii column density is from FUSE observations of the unsaturated 1012.502 Å line with f = 0.0355. These f values are from Morton (2003). For both ions, the column densities are from the AOD method and refer to gas absorbing between ∼±100 km s⁻¹. The objects are drawn from the AGN sample listed in Table 2. We list in Table 5 the relevant averages of log[N(X)sin|b|] for lines of sight to AGNs in directions in the north Galactic sky with b > 40° and in the south with b < −40°. The table lists the object number and values of 〈log[N(X)sin|b|]〉 ± stdev for Fe iii and S iii. Although there is a considerable dispersion in the results as revealed in the standard deviations which range from 0.16 to 0.20 dex, both ions reveal an excess in the logarithmic average for b > 40° compared to b < −40°. For Fe iii, the excess is 0.20 dex or 58%. For S iii, the excess is 0.15 dex or 41%. The larger excess for Fe iii may be related to differences in the level of dust processing in the north versus south Galactic polar regions.

Given the known north/south Galactic asymmetries, it would be highly desirable to obtain dispersion measures toward a sample of distant pulsars that covers the south Galactic sky at |b| < −40°. Given the strong asymmetry in Fe iii and S iii which likely trace photoionized gas in the WIM, the electron dispersion measures are expected to show an excess of ∼0.15–0.2 dex in the north compared to the south. The Gaensler et al. (2008) sample contains four globular clusters with b > 40° and 2 with b < −40°. The four globular clusters with b > 40° have 〈DMsin|b|〉 = 22.7 pc cm⁻³, while the two with b < −40° have 〈DMsin|b|〉 = 17.8 pc cm⁻³. Although the sample is small, the existing observations imply a 27.5% or 0.11 dex excess of DMsin|b| in the north Galactic polar region. The derivation of the scale height from the pulsar data should allow for the presence of the north/south asymmetry. If the electrons behave like the transition temperature ions, the origin of the asymmetry is an excess in the north Galactic sky for b > 40° compared to the rest of the sky with −90° < b < 40°.