Observations of Outflowing UV Absorbers in NGC 4051 with the Cosmic Origins Spectrograph

Abstract

We present new Hubble Space Telescope (HST)/Cosmic Origins Spectrograph
observations of the Narrow-Line Seyfert 1 galaxy NGC 4051. These data were
obtained as part of a coordinated observing program including X-ray
observations with the Chandra/High Energy Transmission Grating (HETG)
Spectrometer and Suzaku. We detected nine kinematic components of UV
absorption, which were previously identified using the HST/Space Telescope
Imaging Spectrograph. None of the absorption components showed evidence for
changes in column density or profile within the \sim 10 yr between the STIS and
COS observations, which we interpret as evidence of 1) saturation, for the
stronger components, or 2) very low densities, i.e., n_H < 1 cm^-3, for the
weaker components. After applying a +200 km s^-1 offset to the HETG spectrum,
we found that the radial velocities of the UV absorbers lay within the O VII
profile. Based on photoionization models, we suggest that, while UV components
2, 5 and 7 produce significant O VII absorption, the bulk of the X-ray
absorption detected in the HETG analysis occurs in more highly ionized gas.
Moreover, the mass loss rate is dominated by high ionization gas which lacks a
significant UV footprint.

Full text

Page 1

Observations of Outflowing UV Absorbers in NGC 4051 with the
Cosmic Origins Spectrograph1
S.B. Kraemer2, D.M. Crenshaw3, J.P. Dunn4, T.J. Turner5, A.P. Lobban6, L. Miller7, J.N.
Reeves6, T.C. Fischer3and V. Braito8
ABSTRACT
We present new Hubble Space Telescope (HST)/Cosmic Origins Spectrograph
observations of the Narrow-Line Seyfert 1 galaxy NGC 4051. These data were
obtained as part of a coordinated observing program including X-ray observa-
tions with the Chandra/High Energy Transmission Grating (HETG) Spectrome-
ter and Suzaku. We detected nine kinematic components of UV absorption, which
were previously identified using the HST/Space Telescope Imaging Spectrograph.
None of the absorption components showed evidence for changes in column den-
sity or profile within the ∼ 10 yr between the STIS and COS observations, which
we interpret as evidence of 1) saturation, for the stronger components, or 2) very
low densities, i.e., nH< 1 cm−3, for the weaker components. After applying a
+200 km s−1offset to the HETG spectrum, we found that the radial velocities of
the UV absorbers lay within the O VII profile. Based on photoionization models,
1Based on observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space
Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy,
Inc. under NASA contract NAS 5-26555. These observations are associated with proposal 11834.
2Institute for Astrophysics and Computational Sciences, Department of Physics, The Catholic University
of America, Washington, DC 20064, USA; steven.b.kraemer@nasa.gov
3Department of Physics and Astronomy, Georgia State University, Astronomy Offices, One Park Place
South SE, Suite 700, Atlanta, GA 30303, USA
4Department of Chemistry and Physics, Augusta State University, 2500 Walton Way, Augusta, Ga. 30904,
USA
5Department of Physics, University of Maryland Baltimore County, Baltimore, MD 21250, USA
6Astrophysics Group, School of Physical and Geographical Sciences, Keele University, Keele, Staffordshire
ST5 5BG, UK
7Department of Physics, University of Oxford, Denys Wilkinson Building, Keble Road, Oxford OX1 3RH,
UK
8Department of Physics and Astronomy, University of Leicester, University Rd., Leicester, LE1 7RH,
U.K.

Page 2

– 2 –
we suggest that, while UV components 2, 5 and 7 produce significant O VII ab-
sorption, the bulk of the X-ray absorption detected in the HETG analysis occurs
in more highly ionized gas. Moreover, the mass loss rate is dominated by high
ionization gas which lacks a significant UV footprint.
Subject headings: galaxies: active – galaxies: Seyfert – galaxies
1.Introduction
According to the standard paradigm, Active Galactic Nuclei (AGN) are powered by
accretion of matter onto a supermassive black hole. Outflowing winds may arise from an
accretion disk surrounding the black hole (e.g., Rees 1987) or at larger distances. As evi-
dence for such winds, more than 50% of Seyfert 1 galaxies, relatively local (z < 0.1), modest
luminosity (Lbol< 1045ergs−1) AGN, show intrinsic X-ray and UV absorption (Crenshaw,
Kraemer, & George 2003, and references therein), suggesting that the absorbers have global
covering factors > 0.5. Blue-shifted absorption lines in their UV (Crenshaw et al. 1999),
and X-ray (Kaastra et al. 2000; Kaspi et al. 2000) spectra reveal significant outflow veloci-
ties (up to −4000 km s−1; Dunn et al. 2007). The inferred mass-loss rates are comparable
to the accretion rates needed to produce the observed luminosities of AGN. Hence, mass
outflows are a critical component in the structure, energetics, and evolution of AGN. Vari-
ous acceleration mechanisms have been proposed for these outflows, in particular radiative
driving (e.g., Murray et al. 1995), thermal winds (Begelman, McKee, & Shields 1983), and
magneto-hydrodynamic flows (Blandford & Payne 1982).
Overall, there appears to be a one-to-one correspondence between X-ray and UV ab-
sorption in Seyfert galaxies (Crenshaw et al. 2003; but, see Dunn et al. 2008). Nevertheless,
the nature of the physical connection between the sources of UV and X-ray absorption may
vary among these objects and, indeed, among different kinematic components in individual
objects. While it has been suggested that at least some of the UV absorption arises in the
same gas as the X-ray absorbers (e.g., Mathur, Elvis, & Wilkes 1999), it is apparent that
there is a range of physical conditions within the absorbing gas (e.g., Kraemer et al. 2002;
2003; 2005; Gabel et al. 2005). In fact, there is strong evidence that the X-ray absorbers
in individual objects span a range in ionization (e.g., Steenbrugge et al. 2005; Turner et al.
2011). While there are cases in which the UV absorption lines are simply the “footprint”
of the X-ray absorbers (e.g., Kraemer et al. 2005), an interesting possibility is that they
are condensations in a more highly ionized medium (Kriss et al. 1996, 2000; Krolik & Kriss
2001). For example, UV and X-ray absorbers may be detected at the same radial velocities,
but there is evidence, such as lower line-of-sight covering factors for the UV absorbers (Krae-

Page 3

– 3 –
mer et al. 2006) or constraints based on photo-ionization modeling, that they are distinct
physical components. If these components are indeed co-located, it raises questions about
the stability of lower-ionization condensations, which may not be in pressure equilibrium
with the surrounding medium (e.g. Gabel et al. 2005), and the overall dynamics of the out-
flows, because the integrated cross-section for absorption of radiation is significantly greater
for lower ionization gas (eg., Arav, Li & Begelman 1994).
NGC 4051 is a nearby (distance = 15.2 Mpc, Russel 2004), narrow-line Seyfert 1 (NLSy1)
galaxy (see Osterbrock & Pogge 1985), as evidenced by the narrowness of its Hβ emission
line profile, with a full width at half maximum (FWHM) ≈ 1070 km s−1(Peterson et al.
2004). Via optical reverberation mapping, Denney et al. (2009) determined the mass of
the central black hole to be 1.73+0.55
smaller black holes than broad-line Seyfert 1s of similar luminosity, hence have higher relative
mass-accretion rates and are radiating at or near the Eddington limit (e.g., Mathur 2000).
Interestingly, while there is evidence that NGC 4051 is substantially sub-Eddington (see
Wang & Netzer 2003), hence may not be a typical NLSy1 in that sense, it does exhibit the
extreme X-ray variability characteristic of the class (e.g., Turner et al. 1999; Leighly 1999).
−0.52× 106M⊙. NLSy1s are thought to possess relatively
While observations such as those with ROSAT (Komossa & Fink 1997) and ASCA
(Guainazzi et al. 1996) originally revealed the presence of intrinsic X-ray absorption in NGC
4051, it has been the subject of several more recent studies at higher spectral resolution,
both in the X-ray and UV. Collinge et al. (2001) observed NGC 4051 on 2000 April 24 – 25
with the Chandra High Energy Transmission Grating (HETG) Spectrometer, with a total
exposure time of ∼ 81.5 ksec, and on 2000 March 24 – 25 with the Hubble Space Telescope
(HST)/Space Telescope Imaging Spectrograph (STIS). They found that the UV absorption
showed 9 distinct kinematic components, all seen in C IV, N V, and several in Si IV, Si III,
Si II, and C II, while the X-ray absorption consisted of two systems. They suggested that
lower velocity X-ray system, with radial velocity vr∼ −600 km s−1, may be connected to
the UV absorbers. In Far Ultraviolet Spectroscopic Explorer (FUSE) spectra, obtained on
2002 March 29, 2003 January 18, and 2003 March 19, Kaspi et al. (2004) detected O VI,
H I Lyman, and C III absorption consistent with the components detected by Collinge
et al.In their analysis of a 100 ksec XMM-Newton observation from 2001 May 16-17,
Krongold et al. (2007) found evidence for two X-ray absorption components. Based on time
variability and photoionization modeling, they argued that both systems were within 3.5
lt-days of the central source. Steenbrugge et al. (2009; hereafter S2009) analyzed ∼ 200 ksec
of Chandra/Low Energy Transmission Grating (LETG) spectrometer data, obtained 2001
December 31 – 2002 January 1 and 2003 July 23 – 24. They modeled the X-ray absorption
with four components, three spanning the range of −200 km s−1≤ vr≤ −610 km s−1and one
at ∼ −4760 km s−1, which could possibly be identified with the high velocity/high ionization

Page 4

– 4 –
absorber suggested by Collinge et al. (although that system was at vr∼ −2340 km s−1).
As with Krongold et al., Steenbrugge et al. constrained the radial distances of the X-ray
components based on time variability (we will discuss this in more detail in Section 5.).
We observed NGC 4051 for 300 ksec with Chandra/HETG, from 2008 November 6 –
30, and for 350 ksec with Suzaku, on 2008 November 23. The analysis of these data are
discussed in detail in Miller et al. (2010), Turner et al. (2010) and Lobban et al. (2011;
hereafter L2011). To summarize the results from L2011, based on photoionization modeling
and spectral analysis, we found evidence for 5 zones of absorption, spanning a range of −180
km s−1≤ vr≤ −710 km s−1and a highly ionized zone at vr∼ −5800 km s−1, similar to
the results from S2009. Additionally, the continuum-fitting required another zone, with a
covering factor of 30%. In this paper, we present our analysis of HST COS observations of
NGC 4051. In addition to the characterization of the UV absorption, via spectral analysis
and photoionization modeling, we discuss the relation between the UV absorbers and the
components of X-ray absorption, the latter as modeled by L2011 and S2009.
2. Observations and Analysis
We obtained HST COS observations of the nucleus of NGC 4051 on 2009 December 11
(UT) as part of a multiwavelength effort. Due to scheduling constraints, the UV observations
were not simultaneous with our Suzaku and Chandra X-ray observations. We obtained optical
spectra of NGC 4051 through a 2′′wide slit in photometric conditions on 2009 December 17
and 18, close in time to the COS observations, with the Perkins 1.8-m telescope and DeVeny
Spectrograph at Lowell Observatory. The optical spectra span the range 4000 – 7000˚ A at a
spectral resolution of ∼3.3˚ A. We also retrieved FUSE archival spectra from 2002 March 29,
2003 January 18, and 2003 March 19 (Kaspi et al. 2004; Dunn et al. 2008), obtained through
the 30′′× 30′′aperture, and averaged these together to improve the signal-to-noise ratios
(SNRs). Finally, we made use of the HST STIS echelle E140M observation of of NGC 4051
by Collinge et al. (2001), obtained through a 0′′.2 × 0′′.2 aperture.
We observed NGC 4051 with COS over 4 HST orbits with the G130M and G160M
gratings through the Primary Science Aperture, which is 2′′.5 in diameter. COS far-UV
observations with these gratings are imaged onto two side-by-side detectors, and thus each
observation consists of two spectra separated by a small wavelength gap (Dixon et al. 2010;
Green et al. 2012). We therefore obtained the spectra at different wavelength offsets to
ensure full coverage from 1137 – 1772˚ A. We give the wavelength coverages and exposure
times of the individual spectra in Table 1.

Page 5

– 5 –
We extracted the one-dimensional COS spectra from the STScI “x1d” files and interpo-
lated them onto a common wavelength grid. The absolute fluxes of the individual spectra in
the regions of overlap differ by <5%, and we therefore combined them without any scaling
in flux. We flagged regions at the ends of each spectrum that were not useful (and not
flux-calibrated in the pipeline processing) as well as regions containing obvious artifacts (in
particular, shadows cast by ion repeller grid wires), and then averaged fluxes in good regions
on a pixel-by-pixel basis. Individual pixels in the final averaged spectrum contain contribu-
tions from one to four original spectra; most are from two spectra. The resolving power of
the combined COS spectrum is λ/∆λ ≈ 16,000, which translates to a velocity resolution of
19 km s−1(∼8 pixels) FWHM. Comparison with the STIS echelle spectra of NGC 4051 in-
dicated a slight constant offset in the central wavelengths of the Galactic lines; we corrected
for the offset by adding +0.07˚ A to the COS spectra to allow for direct comparison with the
STIS observations.
We show the full COS spectrum and identify the principal UV emission lines in Figure
1. The SNRs per resolution element range from 5 to 10 in the continuum regions and
up to 50 at the peaks of the emission lines, about twice those in the STIS spectrum (for
the same wavelength bins) even though the COS spectrum was obtained when NGC 4051
was in a low flux state. In Figure 2, we show the UV continuum light curve for NGC 4051
spanning 32 yr, derived primarily from International Ultraviolet Explorer (IUE) observations
(Dunn et al. 2006). The STIS and COS observations occurred when NGC 4051 was in
“average” (F[1365˚ A] = 1.38(±0.16)×10−14erg s−1cm−2˚ A−1) and relatively low (F[1365˚ A]
= 0.83(±0.05)×10−14erg s−1cm−2˚ A−1) continuum states, respectively, although it is clear
from the previous IUE observations that large amplitude variations could have occurred in
the ∼10 yr time interval between these latest observations.
As noted by Collinge et al. (2001), the STIS UV spectrum of NGC 4051 contains a
complex set of absorption lines from our Galaxy, the host galaxy of NGC 4051, and clouds
of gas that are intrinsic to NGC 4051 and outflowing with respect to its nucleus. We list
their absorption components, radial velocities relative to the emission-line redshift, and full-
width at half-minima (FWHM; we use the same acronym for absorption and emission lines)
in Table 2. In Figures 3 and 4, we show expanded versions of the COS and STIS spectra
around absorption lines spanning a wide range in ionization (N V λλ1238.821, 1242.804; C IV
λλ 1548.202, 1550.774;Si IV λλ1393.755, 1402.770, Si III λ1206.500; C II λ1334.532; Si II
λ1260.422). We also show the radial velocities of the kinematic components of absorption (for
the strongest member of the doublet) in Figure 3. In order to be consistent with the radial
velocities of Collinge et al., we use their systemic redshift of z = 0.002295 for NGC 4051
based on its optical emission lines (de Vaucouleurs et al. 1991). Using the H I 21-cm redshift
(z = 0.002418, de Vaucouleurs et al. 1991) would shift the radial velocities of the absorption

Page 6

– 6 –
lines by an additional −37 km s−1. The higher SNRs of the COS spectrum are apparent
in Figures 3 and 4, and the lower spectral resolution of COS (19 km s−1) compared to that
of the STIS E140M grating (7 km s−1) is also clearly evident, for example in the depth of
Component 1 in N V.
We find the same absorption components (G, 1– 9) in the COS spectrum that were
identified by Collinge et al. (2001) in the STIS spectrum. We also confirm that Component
8, most clearly seen in N V, is shifted to more negative radial velocities in lower ionization
lines (Collinge et al. suggest that this is a new low-ionization component and identify it as
Component 10). We did not find any UV absorption line corresponding to the −2340 km
s−1absorption system identified in X-ray observations (Collinge et al. 2001), in agreement
with the lack of detection of this system in the STIS (Collinge et al.) and FUSE (Kaspi et
al. 2004) spectra.
In Figure 4, the Galactic components (labeled “G”) in the strong low-ionization lines
are completely saturated and black (i.e., consistent with zero flux in their cores), but these
same lines in the COS spectra have a slight pedestal at ∼6 × 10−16erg s−1cm−2˚ A−1(about
7% of the continuum flux). The latter is likely due to instrumental scattering by the broad
wings of the COS line-spread function (Dixon et al. 2010), and we take this extra emission
into account when analyzing the absorption components.
3.Observational Results
We first identify those absorption components that are likely associated with the inter-
stellar medium (ISM) in our Galaxy or the host galaxy of NGC 4051, based primarily on
their radial velocities. Taking a closer look at Figures 3 and 4, we confirm that the “G”
component identified by Collinge et al. (2001) arises in our Galaxy’s ISM; the absorption
lines are broad, completely saturated at low ionization, and at a heliocentric radial velocity
of only −39 km s−1. Strong Galactic lines at this velocity can also be seen in the FUSE
spectrum of NGC 4051 (Kaspi et al. 2004); these lines, and in particular the Galactic H2
absorption, are studied in detail in Wakker (2006). Component 1 is present in N V, C IV,
Si IV and possibly Si III, and is at a heliocentric radial velocity of only +41 km s−1; it is
likely a high-ionization component of gas in our Galaxy (see Bowen et al. 2008; Fox et al.
2006), although an origin in NGC 4051 cannot be ruled out. Component 8 shows broad,
saturated absorption in C IV, Si IV, Si III, C II, and Si II at a central radial velocity of −48
to −80 km s−1with respect to NGC 4051 (Collinge et al. 2001); it is likely due to gas in
the disk or halo of NGC 4051. Component 9 shows moderate absorption in these lines as
well, at a radial velocity of +30 km s−1. This component and two others seen only in Lyα at

Page 7

– 7 –
+110 km s−1and +260 km s−1(Collinge et al. 2001) are likely associated with high-velocity
clouds in the host galaxy of NGC 4051, similar to those seen in our Galaxy (Fox et al. 2006;
Shull et al. 2009). The remaining components are likely intrinsic to and outflowing from the
active nucleus in NGC 4051.
The troughs of the low-ionization lines of Component 8 in Figure 4 are heavily saturated
and go down to zero flux in the STIS data, but show excess emission in the COS data. The
excess flux cannot be due to instrumentally scattered light, because it does not appear in
the troughs of the nearby Galactic lines, and there is no extra line emission near the lines
of Component 8, at least in the regions of Si III and Si II λ1260. The Component 8 lines
bottom out at ∼1.2 ×10−15erg s−1cm−2˚ A−1in the COS data, rather than ∼6 ×10−16erg
s−1cm−2˚ A−1for the Galactic lines in the same areas of the spectrum. Thus, in addition
to the instrumental scattered light, the COS spectrum show a slight excess UV continuum
at the level of ∼6 ×10−16erg s−1cm−2˚ A−1in its 2′′.5 aperture compared to the 0′′.2 × 0′′.2
aperture of STIS.
Intrinsic absorption components 2 – 7 are present in N V and C IV, but none are clearly
evident in Si IV or lower ionization lines, indicating that the gas in these components is
relatively highly ionized. Components 2, 5, and 7 show the strongest lines, but Component
7 is strongly blended with Component 8 in C IV. The remaining components (3, 4, and 6)
are relatively weak and difficult to separate from the major components due to blending. We
therefore concentrate on Components 2, 5, and 7, because they are strong, less affected by
blending, and likely dominate the mass outflow in the UV. We also investigate the nature of
components 8 and 9 in more detail.
Convolving the STIS spectrum with the COS line-spread function (LSF, Dixon et al.
2010) and scaling it down to match the COS fluxes, we find no clear evidence for intrinsic
changes in the absorption components over the ∼10 yr interval between observations, as
demonstrated for C IV in Figure 5.The apparent difference in depths of some of the
absorption components is likely due to excess emission in the COS spectra, as discussed
in more detail below. This is somewhat unusual, because most Seyfert 1 galaxies show
strong absorption variability, including the appearance and disappearance of absorption
components, on time scales of years (e.g., NGC 3516, Kraemer et al. 2002; NGC 3783,
Gabel et al. 2005; NGC 4151, Kraemer et al. 2006; NGC 7469, Scott et al. 2005; Mrk 279,
Scott et al. 2009; NGC 5548, Crenshaw et al. 2009; and references therein). However, this
may not be too surprising, because many of the absorption lines in NGC 4051 are completely
saturated (Collinge et al. 2001; this paper), and would not show detectable variations unless
the ionic column densities decreased dramatically or a significant fraction of a background
emission source (e.g., the broad line region) was covered or uncovered (e.g., see Crenshaw et

Page 8

– 8 –
al. 2004) over the 10-year interval between STIS and COS observations. Thus, we cannot
rule out ionic column density changes in the saturated lines between the STIS and COS
observations.
To further investigate the nature of the uncovered emission in the COS spectrum and
its effect on measurements of the ionic column densities, we generated a detailed model of
the background emission spectrum. We fitted a spline to the continuum across the entire
spectrum using regions unaffected by emission or absorption features. To generate template
profiles for the emission lines, we used the He II λ1640 line, which is not affected by absorp-
tion. Inflections in the He II profile, shown in Figure 6, suggest the presence of 3 components:
broad, intermediate, and narrow. We fitted the broad component with a cubic spline and
the other two components with Gaussians. The resulting widths of the emission-line com-
ponents are 260, 1090, and 4500 km s−1(FWHM). We associate the first and last of these
with the narrow-line region (NLR) and broad-line region (BLR), and suggest that the middle
component arises in an “intermediate-line region” (ILR), which we have also identified in
low-flux UV spectra of NGC 4151 (Crenshaw & Kraemer 2007) and NGC 5548 (Crenshaw
et al. 2009). However, we note that our “ILR” likely corresponds to the “BLR” detected
in optical spectra, which show a “broad” component of Hβ with FWHM = 1070 km s−1
(Peterson et al. 2004).
To model the emission-line profiles of the high-ionization lines, we reproduced the He II
templates at the expected positions of the lines, retaining the same velocity widths, and
scaled them in intensity to obtain the best fits to the observed profiles..
doublet ratios (e.g., N V λ1238.8/λ1242.8) to vary between 1 and 2 (see Crenshaw et al.
2009) made little difference to the overall fits, so we fixed them to a value of 1. We adopted
the minimum narrow-line fluxes needed to fit the observed profiles. Figure 7 shows that our
procedure yields an excellent fit to the observed profiles in the COS spectra, and provides
an accurate deconvolution of the NLR, ILR, and BLR contributions to the emission-line
profiles.
Allowing the
As noted above, absorption components 2, 5, and 7 are saturated at zero flux in their
cores in the STIS spectra. Their non-zero fluxes in the COS spectra must be due to uncovered
emission. To model the uncovered emission, we use our derived values for the instrumentally-
scattered light and the excess emission from the host galaxy, plus contributions from the NLR
fluxes derived above, as we have done in the past (Kraemer et al. 2002; Crenshaw et al.
2009). Thus, a reasonable model for the uncovered emission is a combination of the excess
continuum flux at a level of 1.2 ×10−15erg s−1cm−2˚ A−1(half due to instrumental light,
half to extra UV flux in the COS aperture) plus the NLR emission. As shown in Figure
6, the model provides provides a reasonably good fit to the uncovered emission, in that it

Page 9

– 9 –
skims just beneath the unblended saturated lines associated with Components 2, 5, and 7.
Furthermore, it demonstrates that Component 5 and 7 must be mostly inside the NLR,
whereas Component 8 absorbs most, if not all, of the NLR emission (we cannot constrain
Component 2 in this manner because it does not overlap the NLR emission in velocity space).
On the other hand, Components 2, 5 and 7 absorb most, if not all, of the continuum, BLR,
and ILR emission, and therefore lie outside of these regions.
We measured the ionic column densities of Components 2, 5, 7, 8, and 9 by subtracting
the uncovered emission components from the spectra and dividing by the covered emission
components (covered continuum, BLR, and ILR fluxes) to obtain normalized profiles. We
then converted these to optical depth as a function of wavelength, and integrated across
the optical depth profiles to obtain column densities (Crenshaw, Kraemer, & George 2003).
Uncertainties in the measured column densities come from propagation of photon noise and
different reasonable fits to the underlying emission. Measurements of unsaturated lines in
both COS and STIS spectra resulted in differences smaller than the quoted uncertainties,
reinforcing our claim that these absorption features did not vary. Components 2, 5, and 7
are completely saturated in N V and C IV, and we can only provide lower limits to their
ionic column densities. We used the higher SNR COS spectra, and assumed the least amount
of uncovered emission, given the uncertainties, to determine these limits. To obtain upper
limits to the Si IV columns, we added model absorption-line profiles to the spectra and
decreased their equivalent widths until they were not detectable above the noise.
We used the FUSE spectra of NGC 4051 to obtain further constraints on the ionic
column densities. The COS and FUSE observations are not concurrent, and our underlying
assumption is that the limits obtained from the COS observations apply to the epochs of
FUSE observations as well. By matching the COS and average FUSE spectra in the region
of wavelength overlap, we find the continuum flux at 1365˚ A is 1.06×10−14erg s−1cm−2˚ A−1
projected from the FUSE spectra, intermediate between the STIS and COS observations,
providing further justification for the above assumption. As noted by Kaspi et al. (2004),
the O VI absorption components are blended together into big troughs, similar to Lyα, and
no useful limits can be obtained. We agree with Kaspi et al. that N III λ989.790 absorption
is undetectable, primarily due to contamination by geocoronal emission and Galactic H2
absorption, and that C III λ977.020 absorption is present in Component 5 and possibly
present in Component 2 (Component 7, if present, is blended with 8). We see no evidence
for P V λλ1117.98, 1128.01 absorption, which would have indicated very high columns of
high-ionization gas, nor C III∗λ1175 absorption in either the FUSE or COS spectra. As
noted by Kaspi et al., H I Lyman absorption lines are seen extending down to the Lyman
edge. We identified H I absorption to Ly6 λ930.748 in Component 2 and Ly9 λ920.963
in Component 5, whereas Lyman absorption in Component 7 is always blended with that

Page 10

– 10 –
in Component 8. The longer wavelength lines of Lyβ and Lyγ are strongly affected by
geocoronal emission and likely saturated in any event, so we used the shorter wavelength
lines to determine the H I columns for Components 2 and 5. The C III absorption profiles
for Components 2 and 5 (shown in Kaspi et al. 2004) appear not to be saturated, and we
measured their columns directly by integrating their optical depths across the profiles after
subtracting off the uncovered emission.
We present the ionic column densities or limits for Components 2, 5, 7, 8, and 9 in Table
3. Using a different technique for the FUSE analysis, Kaspi et al. (2004) found that the H I
column density in Component 5 is 1.0+0.6
but the error bars still overlap. They do not give column densities for any of the other lines.
We do not detect O I in any absorption component; its ionic column density is therefore <
1.0 × 1014cm−2.
−0.5× 1016cm−2. Our value is about half of theirs,
To gauge the reddening in the line of sight to the nucleus of NGC 4051, we measured
the total fluxes of the He II λ1640 emission line in the COS spectrum and the He II λ4686
emission line in the concurrent optical spectra. The lines are due to recombination and the
intrinsic H II λ1640/λ4686 ratio should be ∼7.25, for an electron density ne= 106cm−3and
temperature T = 5×104K (Seaton 1978), which we chose to match the combined BLR and
NLR emission. To ensure an accurate absolute flux, we scaled our averaged optical spectrum
slightly (by a factor of 1.15) so that our [O III] λ5007 flux matched the well-established value
of Peterson et al. (2000). The resulting observed He II λ1640/λ4686 ratio is then (14 ×
10−14erg s−1cm−2)/(4.0 × 10−14erg s−1cm−2) = 3.5. This value yields E(B−V) = 0.19
(only 0.01 from our Galaxy) for a Galactic reddening curve (Savage & Mathis 1979) and NH
= 1.0 × 1021cm−2for a standard Galactic dust-to-gas ratio (Shull & van Steenberg 1985).
Kaspi et al. (2004) present an observed spectral energy distribution (SED) for NGC 4051
which has a UV spectral index of α = −2.0 (where Fν = Kνα) between 1160 and 3000
˚ A. Correction for extinction based on the above reddening yields α = −1.3 and F(1365˚ A)
= 6.4 × 10−14erg s−1cm−2˚ A−1for the STIS observation used in Kaspi et al.’s SED.
We searched for and did not find any H2absorption in the FUSE spectrum near the
redshift of NGC 4051. Using the methods of Dunn et al. (2007), we find that N(H2) <
5.0 × 1014cm−2for any rovibrational level. Thus, the reddening is not due to dust in cold
molecular gas.

Page 11

– 11 –
4.Photoionization Models
The photoionization models used for this study were generated using the photoioniza-
tion code Cloudy, version 08.00 (last described by Ferland et al. 1998). We assumed an
open, or “slab”, geometry. As per convention, the models are parametrized in terms of the
dimensionless ionization parameter, U = Q/(4πr2cnH), where r is the radial distance of the
absorber, nHis hydrogen number density, in units of cm−3and Q =?∞
number of ionizing photons s−1emitted by a source of luminosity Lν, and the total hydrogen
column density, NH(in units of cm−2), where NH= NHI+ NHII.
13.6eV(Lν/hν) dν, or the
4.1.Model Inputs
In our previous paper (L2011), we characterized the intrinsic X-ray continuum in the
Suzaku XIS + HXD (0.5 – 10 keV and 20 – 50 keV, respectively) as a power-law with a
photon index Γ ≈ 2.5. This is consistent with constraints on the EUV-soft X-ray SED
obtained via photoionization models and the ratios of [Ne V] 14.32 µm, 24.32 µm, [Ne III]
15.56 µm, and [O IV] 25.89 µm observed in Spitzer IRS spectra (Mel´ endez et al. 2011).
Noting that the UV observations found NGC 4051 in a relatively low-flux state, we used the
lowest of the 3 Suzaku fluxes, 3.4 × 10−12ergs cm−2s−1at 2 keV, corrected for absorption,
and extrapolated down to 1365˚ A, assuming a spectral index α = 1.5. This predicts a flux of
7.4 × 10−14ergs cm−2s−1˚ A−1, which is close to the dereddened flux from the STIS spectrum
used by Kaspi et al. (2004). Based on this, and the fit to the UV continuum (see Section
3), we have parametrized the SED in the form of a broken power law, such that Lν∝ ναas
follows: α = −1.3 for energies < 9.8 eV (1365˚ A), α = −1.5 over the range 9.8 eV ≤ hν < 50
keV1. We included a low energy cut-off at 0.1 eV and a high energy cutoff at 100 keV. Using
this SED and the unabsorbed flux at 2 keV, we determined that Q ≈ 6.4×1052photons s−1.
For these models, we have assumed roughly solar elemental abundances (e.g., Asplund et al.
2009) as follows (in logarithm, relative to H, by number): He: −1.00, C: −3.57, N: −4.17,
O: −3.31, Ne: −4.07, Na; −5.76, Mg: −4.40, Al: −5.55, Si: −4.49, P: −6.59, S: −4.88, Ar:
−5.60, Ca: −5.66, Fe: −4.50, and Ni: −5.78.
In these spectra, we were only able to derive column densities for a small number of ions
in the various kinematic components (see Section 3 and Table 3). Therefore, our modeling
method is to fit these derived columns, within the measurement uncertainties, exceed lower
1In L2011, the photoionization models were parametrized in terms of ξ = (?13.6keV
our assumed SED, U ≈ 0.04 × ξ.
13.6eV
Lν dν)/(nHr2); for

Page 12

– 12 –
limits for the saturated lines, and under-predict upper limits for those ions which were
not detected. To achieve this, we used the “optimize” command in Cloudy, adjusting the
commandable errors as needed to improve the fit (see Cloudy Manual, HAZY; Ferland 1996).
However, uncertainties in both elemental abundances and atomic data (e.g., dielectronic
recombination rates) can affect the predictions of ionic column densities, particularly for
poorly populated ionic states. The final model parameters for the strongest absorption
components and the predicted ionic column densities are listed in Tables 4 and 5, respectively.
4.2.Model Results
As discussed in Section 2, components 2 – 7 are associated with mass outflow, while
components 8 and 9 are more likely formed in the ISM or halo of the host galaxy. Of
the mass-outflow components detected in the STIS and COS spectra, the strongest are
components 2, 5, and 7. For components 2 and 5, we have measured column densities for
H I and C III (see Table 3), lower limits for N V and C IV, and an upper limit for Si IV.
Given these limited constraints, models spanning a range of U and NHcan fit the measured
values. For example, using the measured columns for H I and C III and their respective
uncertainties, and the lower limit for C IV, acceptable models parameters for component
5 are logU = −0.45−0.2
s−1and FWHM = 133 km s−1, which are quite close to the values determined by S2009 for
their zone 4, i.e., vr= 270 km s−1and FWHM = 170 km s−1. Furthermore, the S2009 zone
4 model parameters, logU = −0.64 and logNH= 20.4, are within the range acceptable for
component 5, hence we have opted to use these values. The predictions for this model are a
good fit for component 5, albeit with a slight overprediction of the C III column density. To
summarize, while our final model parameters could be off by 0.2dex in U and 0.4dex in NH,
requiring consistency with the X-ray results can, at least, direct us to a preferred location
in U-NH-space (we discuss the relationship between the UV and X-ray absorption in detail
in Section 5).
+0.1and logNH= 20.74−0.40
+0.25. However, for component 5, vr= 268 km
For component 2, there are no clear X-ray constraints and our best-fit model overpredicts
the H I column density and slightly underpredicts the lower limits for N V and C IV.
Na¨ ıvely, one might assume that this is an indication of super-solar N/H and C/H ratios.
Although that cannot be ruled out, increasing the carbon abundance results in a factor of
? 2 overprediction of C III. The physical parameters of component 7 are even less well-
constrained by the observations, because only lower limits for C IV and N V and an upper
limit for Si IV can be determined, and other ions, including H I, are blended with component
8. However, we are able to fit the available constraints using model parameters adapted from

Page 13

– 13 –
zone 2 (L2) in L2011.
As shown in Figures 3 and 4, while there is N V and C IV absorption spanning the
velocities of components 7 and 8, the lower ionization lines, e.g., C II, Si III and Si II, appear
to be associated primarily with component 8. This is consistent with our parametrization
for component 7, since our model predicts column densities for these ions Nion< 1012cm−2.
In Figure 3, the C IV line is saturated for component 8, N V is weak, and Si IV is strong,
but not saturated. On the other hand, C II, Si III, and, possibly, Si II appear saturated,
and at a more positive radial velocity than the high-ionization lines, which is why there were
identified as a separate component (10) in Collinge et al. (2001). This suggests that there
are two physical sub-components near this velocity. One other constraint is that there is no
detectable C II∗, which we find requires that NCII∗< 1.3 × 1013cm−2. Using the ratio of
the upper limit to the column density for C II∗to the lower limit for C II (see Table 4),
we obtain an electron density ne≈ 1 cm−3, for an electron temperature Te= 8 × 103K
(e.g., Srianand & Petitjean 2000); the density must be somewhat lower than this value, since
it was estimated using the lower limit for NCII,. For the lower ionization model, 8L, with
nH= 0.36cm−3, logU = −3.5 and logNH= 18.8, we were able to match the NCII∗constraint
and obtain acceptable predictions for the column densities or lower limits for C II, Si III,
Si II, Fe II. For the higher ionization component, 8H, we were able to match NNV, within
the uncertainties, and obtained a reasonable fit for NSiIVwith logU = −2.4 and logNH= 19,
which requires that nH= 0.03 cm−3, if the two sub-components are co-located. Based on
the model parameters and our estimate of Q, component 8 is > 12.5 kpc from the AGN,
hence, possibly in the halo of the host galaxy. Note, this is a larger radial distance than that
found for the similar low-density, low-ionization components, A and C, in NGC 4151, which
were determined to be at radial distances of 681 pc and 2.15 kpc, respectively (Kraemer
et al. 2001). Finally, component 9 shows only weak C IV and lower-ionization ions (Table
3). We obtained a reasonable fit, save the over-prediction of Si III, which could be due to
uncertainties in third-row element dielectronic recombination rates (e.g., Ali et al. 1991).
As with component 8, the absence of CII∗suggests a low density. Assuming the same nHas
for 8L, component 9 would be 19.3 kpc from the AGN.
Our models for components 2, 5, and 7 predict NOVI> 1016cm−2(see Table 5). To
test whether the predictions are consistent with the data, we plotted the normalized flux
for O VI as a function of velocity, using the measured values of FWHM and vr for each
of the three components. As shown in Figure 8, the combined profile is characterized by a
deep, square trough, spanning velocities −700 km s−1? v ? 0 km s−1, in good agreement
with the FUSE O VI λ1032 profile presented in Kaspi et al. (2004; see their Figure 3),
except for the presence of some uncovered emission in the latter. However, while our UV
model predictions are consistent with the observed O IV absorption, we cannot rule out

Page 14

– 14 –
the presence of additional absorption over this range in velocity, as may be present in more
highly ionized gas (i.e., the X-ray absorbers).
None of the component models predict NH > 2.5 ×1020cm−2. Coupled with their
relatively high values of U, one consequence of the low column densities is that none of the
three strongest components is sufficiently optically thick to the ionization radiation to justify
including the effects of screening by intervening absorbers, e.g. as is the case in NGC 3516
(Kraemer et al. 2002; Turner et al. 2005) and NGC 4151 (Kraemer et al. 2001; 2005; 2006).
Also, while, as noted in Section 3, the continuum and BLR emission is reddened, consistent
with an intervening column of NH= 1.0 × 1021cm−2of dusty gas, none of the models for the
UV absorber components has a sufficient column density to account for the reddening. In
fact, the sum of the column densities of components 2, 5, 7, 8, and 9 is only 5.4 × 1020cm−2.
Furthermore, the presence of Fe II in component 8, along with the likelihood, confirmed by
our photoionization model, that most of the iron is in the form of Fe III, hence not detectable
in these spectra, indicates that there is very little depltion of iron onto dust grains. Hence,
there cannot be a significant amount of dust in this component. One possibility is that the
dust is distributed within components 2, 5, and 7, and that additional dust exists in the
higher ionization gas. However, there are no tight constraints from the modeling of the UV
or X-ray absorption (e.g. L2011) that can provide a definitive answer, hence we must leave
the question as to origin of the reddening open.
5.Physical Properties of the Absorbers
5.1. Connection with the X-ray absorbers
As shown in Table 5, the models for UV components 2, 5, and 7 predict NOVII> several ×
1016cm−2. Hence, these components may be detectable in the HETG spectra. On the other
hand, the UV component models all predict NOVIII< 1016cm−2, which suggests the presence
of more highly ionized gas without a strong UV “footprint”. In order to compare the X-ray
and UV absorption in detail, we first compared the kinematic profiles of strong absorption
lines, specifically O VII and O VIII with N V and C IV. Initially we found no overlap of
velocities, because the X-ray absorption lines are offset to more negative velocities than the
UV. However, this velocity offset is likely due to the calibration of the Chandra gratings (see
the Chandra Proposer Guide: http://cxc.harvard.edu/proposer/POG/html/HETG.html).
For example, in L2011, we had used only the MEG in the 0.5-1.0 keV regime. The accuracy of
the wavelength scale for MEG is ±0.011˚ A. Hence, for O VII, this yields absolute uncertainty
in velocity of ±150 km s−1, while for Ne IX it would be ± 235 km s−1. For Si XIII, we
used the MEG and HEG combined. For the HEG, the accuracy of the wavelength scale

Page 15

– 15 –
is ±0.006˚ A , which results in an uncertainty for Si XIII of ± 250 km s−1. This justifies
some flexibility in comparing the X-ray and UV line kinematics. Indeed, after we applied a
+ 200 km s−1shift to the X-ray spectra, we found visual correlation between the UV and
X-ray absorption, as shown in Figure 9. The trough of the O VII profile encompasses UV
components 2 through 8, and, while the deepest part of O VIII profile covers UV components
1 though 5, there is significant O VIII absorption blueward of component 1.
In L2011, we determined there were 6 separate zones of absorption. The highest velocity
zone, at vr∼ -0.02c, in which H and He-like Fe lines form, is too highly ionized to produce
detectable UV lines and, in fact, no UV absorption is detected at this velocity. And, as we
have found in previous studies (e.g. Turner et al. 2005), no lines are detected from the high-
ionization, high-column, partial-covering zone. For the remaining zones, we list the model
parameters, with U given for the equivalent ξ, and possible associated UV components, in
Table 6.
The two lower ionization X-ray absorbers, zones L1 and L2, should produce strong UV
signatures. With our velocity correction to the Chandra spectra, L1 and L2 should be near
the systemic velocity, which would associate them, kinematically, with UV components 7,
8, and 9. As we noted above, a model using U and NHcorresponding to zone 2 provides a
good match to UV component 7, which suggests that these indeed are the same component.
Given that, one would expect that L1 is associated with UV components 8 and 9. However,
the column density determined by L2011 for L1 is ∼ 30 times larger than that of our models
8L, 8H and 9 combined. Furthermore, using the parameters for L1 (logU = −2.25, logNH=
20.4), a Cloudy-generated model predicts heavily saturated Si IV, with log NSiIV> 15, of
which there is no evidence in the STIS and COS spectra. Therefore, while we find significant
absorption near systemic, in agreement with the HETG analysis, the column density derived
from the UV analysis is much smaller than that derived from the X-ray. The source of the
discrepancy may be the uncertainty in the strength of the ∼0.1 keV black-body component
used in the parametrization of the HETG data. If the black-body were weaker than assumed,
the continuum fitting would require less low-ionization absorption. For example, if we leave
the black-body in the spectrum, but allow its normalization to adjust, we find a 90% lower
limit on the column of zone L1 of log NH > 19.47 which would be in better agreement
with UV components 8 and 9. The other zones (L2-L4) are unaffected, as they are well
constrained by the strong absorption lines in the spectrum. However, although the exact
form of this emission is uncertain, removing the black-body component completely leads to
an unacceptable fit to the data.
L2011’s zones L3a and L3b are significantly more ionized than the modeled UV com-
ponents (see Table 6), hence they do not predict any significant UV absorption. However,