Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/QEDT/Papers/nlxg2.ps Äàòà èçìåíåíèÿ: Fri Jan 5 19:50:51 1996 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 23:18:14 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http astrokuban.info astrokuban

Mon. Not. R. Astron. Soc. 000, 000--000 (1993)
The Cambridge­Cambridge X­ray Serendipity Survey -- II.
Classification of X­ray Luminous Galaxies
B.J.Boyle, 1 R.G.McMahon, 2 B.J.Wilkes, 3 Martin Elvis 3
1. Royal Greenwich Observatory, Madingley Road, Cambridge, CB3 OEZ
2. Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge, CB3 0HA
3. Center for Astrophysics, 60 Garden St, Cambridge, MA 02138, USA
ABSTRACT
We present the results of an intermediate­resolution (1.5 š A) spectroscopic study of 17
X­ray luminous narrow emission­line galaxies previously identified in the Cambridge­
Cambridge ROSAT Serendipity Survey and the Einstein Extended Medium Sensitivity
Survey. Emission­line ratios reveal that the sample is composed of ten Seyfert and
seven starburst galaxies. Measured linewidths for the narrow Hff emission lines lie
in the range 170 \Gamma 460 km s \Gamma1 . Five of the objects show clear evidence for asym­
metry in the [OIII]–5007 emission­line profile. Broad Hff emission is detected in six
of the Seyfert galaxies, which range in type from Seyfert 1.5 to 2. Broad Hfi emis­
sion is only detected in one Seyfert galaxy. The mean full width at half maximum
for the broad lines in the Seyfert galaxies is FWHM = 3900 \Sigma 1750 km s \Gamma1 . Broad
(FWHM = 2200 \Sigma 600 km s \Gamma1 ) Hff emission is also detected in three of the starburst
galaxies, which could originate from stellar winds or supernovae remnants. The mean
Balmer decrement for the sample is Hff/Hfi = 3, consistent with little or no reddening
for the bulk of the sample. There is no evidence for any trend with X­ray luminosity in
the ratio of starburst galaxies to Seyfert galaxies. Based on our previous observations,
it is therefore likely that both classes of object comprise ¸ 10 per cent of the 2 keV
X­ray background.
Key words: X­rays: general -- galaxies: active -- quasars: general
1 INTRODUCTION
A number of recent spectroscopic surveys of soft (0.5--2 keV)
X­ray sources detected at faint fluxes with the ROSAT mis­
sion (Boyle et al. 1995, Georgantopoulos et al. 1995) have
all confirmed that, while QSOs comprise in excess of 50 per
cent of the total X­ray population down to these flux lev­
els, an increasingly large number of X­ray luminous, narrow
(FWHM ! 1000 km s \Gamma1 ) emission­line galaxies (NLXGs)
are identified as counterparts to faint X­rays sources with
fluxes S(0:5 \Gamma 2 keV) ! 10 \Gamma13 erg s \Gamma1 cm \Gamma2 . These galaxies
have X­ray luminosities in the range 10 42 \Gamma 10 43:5 erg s \Gamma1 ,
over 100 times more luminous than late­type galaxies (Fab­
biano 1989), whose low­resolution optical spectra they most
closely resemble. In a previous paper in this series (Boyle
et al. 1995; hereinafter Paper I), we have demonstrated
that, based on their space density and cosmological evo­
lution, these emission­line galaxies could comprise between
15--30 per cent of the soft 0.5--2 keV X­ray background. This
is entirely consistent with the upper limit of ¸ 50 per cent
for the contribution of QSOs to the 0.5--2 keV X­ray back­
ground, based on their luminosity function (Boyle et al.
1994), clustering properties (Georgantopoulos et al. 1993)
and X­ray spectra (Georgantopoulos et al. 1995).
Unfortunately, due to the poor quality of many of the
identification spectra, little is known about the precise na­
ture of this population. In particular, it is not clear whether
these emission­line galaxies are examples of starburst galax­
ies or `hidden' active galactic nuclei (e.g. Seyfert 2 galaxies),
both of which have previously been suggested as possible sig­
nificant contributors to the X­ray background (Griffiths &
Padovani 1990, Fabian & Barcons 1992) and are known to
exist in X­ray surveys (e.g. B¨oller et al. 1992), albeit at
much higher X­ray flux levels and lower space densities.
In order to understand the origin of this potentially
significant population of X­ray sources, we report in this
paper on a detailed intermediate­resolution spectroscopic
study of 17 NLXGs, 10 of which have been identified in the
Cambridge­Cambridge ROSAT Serendipity Survey (CRSS,
see Paper I) and a further 7 objects selected at random from
the Einstein Extended Medium Sensitivity Survey (EMSS,
Stocke et al. 1991) which we suspect are similar (Paper I).
This sample comprises all but two of the NLXG identified
in the CRSS (CRSS1514.4+5627 and CRSS1605.9+2554),
which were not observed due to lack of time.
In Section 2 we report on the observation and analysis
of the NLXG spectra. Based on the results obtained from
these spectra, we discuss the properties and classification of
the NLXG in Section 3, including the implications for the

2 B.J. Boyle et al.
Table 1 Log of Observations
Name RA (J2000) Dec z L(0.5­2keV) V \Lambda Exposure Dichroic Date
h m s ffi 0 00 (\Theta10 43 erg s \Gamma1 ) (seconds)
CRSS NLXG
CRSS0009.0+2041 00 09 01.7 20 41 36 0.189 0.34 19.0 3600 -- 1993 Sept 20/21
CRSS0030.2+2611 00 30 17.4 26 11 38 0.077 0.17 16.0 900 6100 1994 June 10/11
CRSS0030.7+2616 y 00 30 47.9 26 16 50 0.246 0.59 18.4 3600 -- 1993 Sept 20/21
CRSS1406.7+2838 14 06 47.9 28 38 53 0.331 1.52 21.5 1800 7500 1994 June 10/11
CRSS1412.5+4355 14 12 31.6 43 55 36 0.094 0.89 16.2 1800 6100 1994 June 9/10
CRSS1413.3+4405 14 13 19.9 44 05 34 0.136 0.45 17.2 1800 6100 1994 June 9/10
CRSS1415.0+4402 14 15 00.1 44 02 08 0.136 0.25 17.6 1800 6100 1994 June 9/10
CRSS1429.0+0120 14 29 04.7 01 20 17 0.102 0.20 16.6 1800 6100 1994 June 9/10
CRSS1605.6+2554 16 05 39.9 25 43 10 0.278 1.65 18.9 1800 7500 1994 June 10/11
CRSS1705.3+6049 17 05 18.3 60 49 54 0.572 4.72 21.0 1800 6100 1994 June 10/11
EMSS NLXG
MS1252.4\Gamma0457 z 12 54 58.1 \Gamma05 13 22 0.158 1.47 19.1 1800 6100 1994 June 10/11
MS1334.6+0351 13 37 09.8 03 35 55 0.136 1.36 17.7 900 6100 1994 June 9/10
MS1412.8+1320 14 15 16.9 13 06 01 0.139 1.23 18.7 900 6100 1994 June 9/10
MS1414.8\Gamma1247 14 17 34.0 \Gamma13 01 06 0.198 7.42 19.3 1800 6100 1994 June 10/11
MS1555.1+4522 15 56 40.6 45 13 38 0.181 3.80 18.0 1800 6100 1994 June 9/10
MS1614.1+3239 16 16 01.9 32 32 25 0.118 0.96 17.0 1200 6100 1994 June 9/10
MS2044.1+7532 20 43 18.3 75 43 39 0.183 2.08 20.1 900 6100 1994 June 9/10
\Lambda O mag for CRSS NLXG
y Declination incorrect in Paper I
z Wrong source identified as z = 0:158 AGN in Stocke et al. (1991)
composition for the soft X­ray background. We present our
conclusions in Section 4.
2 DATA
2.1 Observations
We obtained intermediate­resolution spectra of 17 emission­
line galaxies previously identified in the CRSS and EMSS
using the ISIS double arm spectrograph at the WHT on
the nights of 1994 June 10--11. We operated ISIS with the
Tektronix CCD on the blue arm and the EEV CCD on the
red arm. We used 600 lines mm \Gamma1 gratings in both arms,
giving an instrumental resolution of 1:5 š A (0:67 š A pix \Gamma1 ).
For each galaxy, we observed the redshifted Hfi/[OIII]–5007
and Hff/[NII]––6717,6731 regions in the blue and red arms
respectively. Throughout the run, conditions were good
(1 arcsec seeing) and we observed all objects with a 1.5 arcsec
slit. The redshifts of the program objects allowed us to ob­
serve the Hfi/[OIII]–5007 and Hff/[NII]––6717,6731 regions
in all galaxies with only one change of grating position and
dichroic. Details of the observations, including exposure
times for each program object, are given in table 1. A fur­
ther two emission­line galaxies were observed with ISIS as
part of the WHT service observation program on the night
of 1994 September 20. The observations were made with
the ISIS red arm, EEV detector and the 1200 lines mm \Gamma1
grating, giving an overall resolution of 0.7 š A (0:33 š A pix \Gamma1 )
in the Hff/[NII]––6717,6731 region. X­ray luminosities in
the 0.5--2 keV band and optical magnitudes for all NLXG are
also listed in table 1. The V magnitudes listed for the EMSS
QSOs are taken from Stocke et al. (1991). The Palomar O
plate magnitides for the CRSS were obtained from the APM
Northern Sky Survey (Irwin, McMahon and Maddox 1994).
The optical magnitudes correspond to the total galaxy mag­
nitude and cannot therefore be used as an accurate mea­
sure of the nuclear magnitudes for these objects. To derive
the X­ray luminosities we assumed H 0 = 50 km s \Gamma1 Mpc \Gamma1 ,
q 0 = 0:5 and an X­ray spectral slope ff X = 1 (f X / š \Gammaff X ).
To convert the EMSS 0.3--3.5 keV band luminosities to the
0.5--2.5 keV band we divided by a factor of 1.8, the conver­
sion factor between the Einstein and ROSAT bands for a
spectral index ff X = 1 (see Boyle et al. 1995).
2.2 Data Reduction and Analysis
The data were reduced using standard routines in the
IRAF reduction package on the Cambridge SPARC clus­
ter. Optimally­extracted galaxy spectra were wavelength­
calibrated using copper­argon arc spectra taken throughout
each night during the observing run. The spectra were flux­
calibrated using spectrophotometric standards taken during
evening and morning twilight on each night. We present the
reduced spectra for each galaxy in figure 1. The large gap
in each spectrum corresponds to the gap in the wavelength
coverage between the blue and red arms of ISIS.
Before measuring the emission lines, we first shifted
each spectrum to the rest­frame using the redshift given in
Table 1 and then divided each spectrum by a second­order
polynomial fit to its own continuum, after excluding the
regions \Sigma150 š A around the Hfi/[OIII] and Hff/[NII] lines.
This division created a flat continuum (advantageous in the
line­fitting routine described below) without removing any
weak, broad features which may be present in the Hff or Hfi
emission.

CRSS II 3
Figure 1 Intermediate­resolution (1.5 š A) spectra for all 17 narrow emission­line galaxies observed in this program.
We then used the SPECFIT routine (written by Dr Ger­
ard Kriss) in IRAF to measure emission line ratios, rest
equivalent widths (W ) and full width at half maximum
intensity (FWHM) of the prominent emission lines in the
spectra of the galaxies. This routine uses a Marquand ü 2 ­
minimisation routine to fit a user­specified number of func­
tions (power­law or linear continuum, gaussian or logarith­
mic line profiles) to the input spectrum. In this case, the
simplifying step of continuum division allowed us to fix the
continuum at a constant value of 1. The fitting process
yields typical Poission errors of 15 per cent and 10 per cent
in the measurement of the equivalent widths and FWHM
respectively.
For each of the 13 spectra in which [OIII]–5007
was observed at a high signal­to­noise ratio (excluding
CRSS1406.7+2838 and CRSS1705.3+6049, see Fig. 1), we
first used the SPECFIT routine to fit both gaussian and loga­
rithmic profiles to this emission line, in order to establish the
correct profile shape to fit to the narrow emission lines. We
chose the [OIII]–5007 line for this purpose because it is the
strongest narrow line observed in each spectrum which does
not have any weak, broad (FWHM ? 1000 km s \Gamma1 ) compo­
nent and is not blended with any other lines. We found that,
in every case, the ü 2 value for the gaussian profile fit was less
than that for the logarithmic profile fit. In 10 cases, the F ­
ratio test (based on the ratio of the ü 2 statistics, see Mood

4 B.J. Boyle et al.
Figure 1 contd.
& Graybill 1963) implied that the gaussian fit was preferred
at the 99 per cent confidence level over the logarithmic fit.
Based on this test, we subsequently used gaussian profiles
throughout the fitting procedure.
Several of the [OIII]–5007 emission lines appeared to
exhibit a significant blue asymmetry. To quantify this ob­
servation, we performed another F ­ratio test, this time on
the ü 2 values obtained from fitting a single and double gaus­
sian to the [OIII]–5007 line. In the latter case, the second
gaussian component was blueshifted with respect to the rest
wavelength of the line. We also investigated a double gaus­
sian with a redshifted component. We stress that the use of
this double gaussian fit is not intended to reflect any physical
significance for the origin of any asymmetry. It simply pro­
vides us with a simple way to establish the significance of the
asymmetry and obtain a more accurate measurement of the
equivalent width, while retaining a consistent measurement
process with all the other emission lines.
In five cases (¸ 35 per cent of the sample) we found
a significant blue asymmetry, with the extra gaussian
component improving the preferred fit at the 99 per cent
confidence level. No redshifted components were de­
tected at the same level of significance. The NLXG
in which the blueshifted components were observed are
CRSS1429.0+0120, MS1252.4\Gamma0457, MS1414.8\Gamma1247,
MS1555.1+4522 and MS2044.1+7532. Examples of these

CRSS II 5
Table 2 Rest­frame emission line properties of NLXG sample
[OIII] Hfi/[OIII] Hff/[NII]
Name W Hfi W [OIII] a(20%) y FWHM W [OI] W Hff FWHM W [NII] W [SII] ID
( š A) ( š A) (km s \Gamma1 ) ( š A) ( š A) (kms \Gamma1 ) ( š A) ( š A) ( š A)
CRSS NLXG
CRSS0009.0+2041 0.6 28.2 289 15.8 4.0 3.3 Narrow
8.1 358 8.6 Blueshifted Hff/[NII]?
CRSS0030.2+2611 3.8 3.1 0.15 238 0.8 25.6 197 13.6 3.2 3.5 Narrow
8.0 2366 Broad Hff (weak)
CRSS0030.7+2629 1.0 12.5 421 16.0 Narrow
CRSS1406.7+2838 8.0 4.9 157 -- 78.7 237 11.0 6.9 4.9 Narrow
CRSS1412.5+4355 6.4 25.5 0.13 479 1.5 33.1 209 20.9 6.6 5.9 Narrow
20.4 3007 Broad Hff (weak)
CRSS1413.3+4405 1.7 9.1 \Gamma0.08 491 1.4 9.6 338 7.0 1.1 2.6 Narrow
22.6 2353 Broad Hff
CRSS1415.0+4402 1.9 7.1 0.13 269 1.2 9.3 178 7.2 2.8 2.7 Narrow
49.0 7668 Broad Hff
CRSS1429.0+0120 4.2 5.7 0.31 508 0.5 21.9 347 11.3 2.1 1.6 Narrow
1.3 2.5 434 8.4 499 4.2 Blueshifted Hfi/[OIII]
Blueshifted Hff/[NII]
CRSS1605.6+2554 8.4 4.5 \Gamma0.14 356 -- 49.3 459 19.5 -- -- Narrow
CRSS1705.3+6049 -- 99.4 311 Narrow
EMSS NLXG
MS1252.4\Gamma0457 4.1 23.1 0.26 445 2.5 31.5 317 25.6 3.2 3.9 Narrow \Lambda
20.4 686 Blueshifted [OIII]
MS1334.6+0351 2.3 16.0 0.15 315 0.6 13.3 280 15.7 2.7 2.4 Narrow
13.1 3669 48.1 2768 Broad Hff/Hfi
MS1412.8+1320 7.3 9.9 \Gamma0.35 314 0.8 28.1 163 10.2 4.2 5.2 Narrow
20.3 2613 80.9 1889 Broad Hff/Hfi
MS1414.8\Gamma1247 8.9 45.2 0.27 441 2.3 53.7 373 13.1 5.8 5.6 Narrow
10.0 550 Blueshifted [OIII]
25.9 2250 89.5 1745 Broad Hff/Hfi
MS1555.1+4522 7.9 48.6 0.24 392 5.3 33.9 366 33.0 10.4 8.5 Narrow
15.7 642 Blueshifted [OIII]
MS1614.1+3239 1.3 16.7 0.20 423 1.4 3.4 363 8.9 3.0 1.8 Narrow
49.4 3725 Broad Hff
MS2044.1+7532 11.2 24.5 0.30 385 0.7 16.7 363 13.1 3.0 5.1 Narrow
13.1 816 Blueshifted [OIII]
90.3 4142 Broad Hff
\Lambda Hff/N[II] on atmospheric B band.
y Whittle (1985) asymmetry parameter, corresponding to the relative wavelength shift between the 10 percentile areas in the blue and
red wings of the emission­line profile from the line centre.
asymmetries with the additional fitted components can be
seen in figure 2, where we have plotted expanded spectra of
the regions surrounding the major emission lines for repre­
sentative sample of 4 NLXG observed in this survey. The
strengths of the additional components range from 25 per
cent to 88 per cent of principal line, with velocity shifts be­
tween the two fitted components ranging from 210 km s \Gamma1
to 653 km s \Gamma1 (see table 3). We have confirmed that the
summed equivalent width of both fitted components in these
asymmetric lines is also good estimate (accurate to within
15 per cent) of the total equivalent width measured without
line­fitting.
Similarly asymmetric [OIII]–5007 profiles have also
been seen in active galaxies by Whittle (1985). Whittle iden­
tified blue asymmetries in most of the AGN/HII regions he
studied. If we use the same asymmetry parameter as defined
by Whittle (a(20%)), we find that the five NLXG (out of 13
with measurable [OIII]–5007) in this analysis which require
blueshifted components all have an asymmetry parameter
a(20%) ? 0:2. In Whittle's sample approximately 40 per
cent of the AGN had asymmetries with a(20%) ? 0:2, and
so the results found here would appear to be consistent with
Whittle's observations. Based on the Kolmogorov­Smirnoff
(KS) test statistic, we find that the overall distribution of
the measured [OIII]–5007 line asymmetries (irrespective of
their significance level) for the NLXG sample is consistent
at the 95 per cent confidence level with that observed by
Whittle (1985). Measured a(20%) values for all [OIII]–5007
lines are given in table 2. The origin of these asymmetries
is still a matter of some debate, but they are most likely to

6 B.J. Boyle et al.
Figure 2 Expanded spectra of the regions around the prominent emission lines in a respresentative sample of 4 NLXG observed in
our sample. The spectra have been corrected to the rest frame and been divided through by a low­order polynomial fit to the continuum.
The accepted fit is denoted by the short dashed line. The contribution of the individual emission lines is shown by the longer dashed
lines. (a) CRSS1429:0+ 0120: a starburst galaxy with asymmetric [OIII]/Hfi and [NII]/Hff.
originate from wind­driven nuclear outflows in which dust
preferentially obscures the emission from the far (red) side
(Whittle 1985, Veilleux 1991).
For each spectrum, separate fits were then carried out
for the following combination of lines over the wavelength in­
tervals indicated: Hfi/[OIII]––4959; 5007 (4800 š A -- 5070 š A);
[OI]–6300 (6280 š A -- 6320 š A); [NII]–6549/Hff/[NII]–6584
(6440 š A -- 6690 š A); [SII] ––6717; 6734 (6697 š A -- 6754 š A). For
each narrow emission line we tried two fits; a single narrow
(FWHM ! 1000 km s \Gamma1 ) gaussian, and two narrow gaus­
sians with an additional blueshifted component. For the
Balmer lines we also tried a fit which comprised a nar­
row plus broad (FWHM ? 1000 km s \Gamma1 ) gaussian profile.
To improve the robustness of the fits, we made every at­
tempt to minimize the number of free parameters in each
fit. As discussed above, the continuum­divided spectra
first allowed us to fix the continuum at a constant value
of 1. We also fixed the [NII]–6734:–6717 emission line ra­

CRSS II 7
Figure 2(b) CRSS0030:2 + 2611: a starburst galaxy with weak broad Hff.
tio at 3.01:1 and the [OIII]–5007:–4959 emission line ra­
tio at 2.88:1. Similarly, we fixed the relative rest wave­
length ratios of the [NII], [SII] and [OIII] emission line pairs
at 6549:6584, 6717:6734 and 4959:5007 respectively. Fi­
nally, for each group of narrow lines fitted simultaneously
(i.e. Hfi/[OIII]––4959; 5007, [NII]–6549/Hff/[NII]–6584
and [SII] ––6717; 6734), we adopted a single value for the
FWHM of the narrow gaussian component fitted to all lines.
For each set of fits (narrow only, narrow plus blueshifted
component, narrow plus broad) we again used the F ­ratio
test to discriminate between them, accepting the more com­
plex fit (i.e. including the blueshifted or broad components)
where it was preferred to the narrow­line­only fit at greater
than 99 per cent confidence level.
3 RESULTS
Based on the fitting procedure outlined in the previous
section, the measured rest­frame emission line equivalent
widths and FWHM for each AGN are listed in table 2.
For each object the data is presented in two lines; the first
provides the information for the narrow lines and the sec­
ond lists the measured parameters for the second compo­
nent (blueshifted line or broad emission line), if present.
The identification of each component (narrow, blueshifted
or broad) is given at the end of each line. Expanded spectra

8 B.J. Boyle et al.
Figure 2(c) EMSS1555:1+ 4522: a Seyfert 2 galaxy with asymmetric [OIII].
of the emission line regions in 4 reprensentative NLXG are
plotted in figure 2. In each spectrum the total fit is shown
by the short dashed line with the individual emission lines
components represented by long dashed lines. The principal
narrow emission line ratios, together with the velocity shift
(\Deltav) of any blueshifted [OIII]–5007 lines observed, are given
in table 3. The emission line ratios quoted in table 3 (and
used throughout the following discussion) are ratios of the
total narrow emission­line equivalent widths (i.e. including
both components of any `double gaussian') and not intensity
ratios. For the Balmer line ratios, the subscripts in table 3
refer to the ratio of the narrow (1) or broad (2) components.
Similarly, the subscripts on the [OIII]–5007 line ratios in
table 3 correspond to the rest (1) and blueshifted (2) com­
ponents respectively. For a power­law continuum fš / š \Gammaff
it can be shown straightforwardly (see Boyle 1990) that the
ratio of equivalent widths W 1 and W 2 , measured at – 1 and
– 2 respectively correspond to a ratio of intensities I 1 and
I 2 :
I 1
I 2
= W 1
W 2
i – 2
– 1
j 2\Gammaff
With the exception of the Hff/Hfi line ratio (discussed sepa­
rately below), the correction from equivalent width ratio to
intensity ratio is negligible for all line ratios quoted in table
3 (e.g. ! 4 per cent in the [OIII]/Hfi ratio) for most realistic

CRSS II 9
Figure 2(d) CRSS1413:3 + 4405: a Seyfert 1.5 galaxy.
values of the continuum slopes, ff ¸ 0:5.
In table 3 we also list the classification assigned to each
NLXG on the basis of its position in the [NII]–6584/Hff
v (see figure 3) using the scheme of Baldwin, Philips
and Terlevich (1981). From figure 3, we can see that
there is a good separation between the objects with HII­
like spectra (i.e. starburst galaxies) and AGN­like spec­
tra (Seyferts 1.5­2). The values of the other emission
line ratios e.g. Hff/[OI]–6300 and Hff/[SII]–6717 + 6734
in each NLXG are also consistent with the classification
based on this diagram. For two NLXGs we have no ob­
servations of the Hfi/[OIII]–5007 region. The large value
of the Hff/[OI]–6300 and Hff/[SII]–6717 + 6734 ratios in
CRSS0009.0+2041 mean that this object is likely to have an
HII­like spectrum, whereas the much lower Hff/[OI]–6300
ratio in CRSS0030.7+2629 implies an AGN­like spectrum
(see Filippenko & Terlevich 1992). Although only the
[OIII]–5007 line is reliably detected in CRSS1705.3+6049
the upper limit of the equivalent width of the much weaker
Hfi suggests that this object is also likely to be an AGN,
although this classification is still rather uncertain. For
each object identified as an AGN, we further classified the
object as a Seyfert 1.5, 1.8, 1.9 or 2, based on the rela­
tive strengths of the broad and narrow Hff components us­
ing the approximate relation given by Netzer (1990): 1 +
(Narrow/Total) 0:4 .

10 B.J. Boyle et al.
Table 3 Emission line ratios of NLXG sample
Name Hfi
[OIII]
Hff
[OI]
Hff
[NII]
Hff
[SII]
Hff1
Hfi1
[SII] 6717
[SII] 6734
Hff2
Hff1
Hfi2
Hfi1
Hff2
Hfi2
[OIII]2
[OIII]1 \Deltav Classification
(km/s)
CRSS NLXG
CRSS0009.0+2041 60.5 1.49 4.97 1.21 0.29 HII
CRSS0030.2+2611 1.22 32.0 1.88 3.82 6.73 0.91 0.31 HII
CRSS0030.7+2616 12.5 0.78 Sy 2
CRSS1406.7+2838 1.63 ? 20 7.15 6.90 9.83 1.53 HII
CRSS1412.5+4355 0.25 22.1 1.58 2.65 5.17 1.12 0.62 Sy 1.8
CRSS1413.3+4405 0.19 6.9 1.37 2.59 5.64 0.42 2.35 Sy 1.5
CRSS1415.0+4402 0.26 7.8 1.29 1.69 4.89 1.04 5.27 Sy 1.5
CRSS1429.0+0120 0.67 60.6 1.95 8.19 5.51 1.31 0.38 0.31 6.46 0.44 653 HII
CRSS1605.6+2554 1.86 ? 50 2.32 ? 25 5.87 HII
CRSS1705.3+6049 ! 0:1 Sy 2?
EMSS NLXG
MS1252.4\Gamma0457 0.09 12.6 1.23 4.44 7.68 0.82 0.88 383 Sy 2
MS1334.6+0351 0.14 22.1 0.85 2.61 5.78 1.12 3.61 5.70 3.67 Sy 1.5
MS1412.8+1320 0.73 35.1 2.75 2.99 3.85 0.81 2.88 2.78 3.99 HII
MS1414.8\Gamma1247 0.20 23.3 4.10 4.71 6.03 1.04 1.67 2.90 3.45 0.24 210 HII
MS1555.1+4522 0.12 6.4 1.03 1.79 4.29 1.22 0.32 425 Sy 2
MS1614.1+3239 0.07 2.4 0.38 0.71 2.61 1.67 Sy 1.5
MS2044.1+7532 0.30 14.4 1.27 2.06 1.49 0.59 5.41 0.53 491 Sy 1.5
Based on the observed emission line ratios in the 17
NLXGs observed, we have identified 7 starburst galaxies
and 10 AGN (including 5 Sy 1.5, 4 Sy 1.8­2 galaxies and
one uncertain classification). We found no LINERS (see
Heckman 1980) in the NLXG sample. These results are
broadly in agreement with the results of Fruscione, Griffiths
and MacKenty (1993), who found similar numbers of star­
burst galaxies and Seyfert galaxies amongst a similar sample
of EMSS `ambiguous' sources. Using the KS test, we were
able to determine that there is no significant difference in
the relative numbers of starburst galaxies/AGN found in
the CRSS and EMSS samples.
The AGN and the starbursts cannot be distinguished
in the present small sample by X­ray luminosity, redshift,
optical magnitude, presence of a broad component, or line
asymmetry. There is also no evidence for any X­ray lumi­
nosity dependence in the ratio of starburst galaxies to AGN.
The X­ray to optical ratio (ff OX ) might be able to discrimi­
nate, since nearby starbursts are relatively X­ray faint (Fab­
biano 1989), but the large galaxy contribution to the optical
magnitudes prevents us measuring this ratio in a meaningful
fashion. High resolution imaging is needed.
The relative numbers of starbursts and AGN in the
sample is the same as that reported in Paper I, although
the numbers of objects with broad components is now bet­
ter understood due to improved analysis of the higher res­
olution spectra. The FWHM of the narrow Hff lines lies
in the range 170 ! FWHM ! 460 km s \Gamma1 , with no sig­
nificant difference between the distribution of FWHM for
the starburst galaxies and AGN samples. Nine NLXG
were found to exhibit broad Hff components, including
6 of the 9 objects classified as AGN on the basis of
their emission line ratios. Three of the starburst galax­
ies (CRSS0030.2+2611, MS1412.8+1320, MS1414.8\Gamma1247)
also exhibit broad components, with FWHM ranging from
1700 km s \Gamma1 to 2400 km s \Gamma1 . Broad Hff profiles (up to
FWHM = 3500 km s \Gamma1 ) have previously been detected in
HII regions in starburst galaxies (e.g. NGC2363, Roy et
al. 1992, Gonzalez­Delgado et al. 1994), although their
origin is uncertain (stellar winds, supernovae remnants, su­
perbubbles). Moreover, we cannot rule out the possibilty
that the broad emission is due to a `mini­QSO' embedded
in the starburst galaxy. The equivalent widths of the broad
Hff components observed in this sample of starburst galax­
ies are also roughly consistent with the range observed in
NGC2363 (few š A ­ 40 š A).
The broad Hff emission lines in the AGN sample have
a mean value of 3900 \Sigma 1900 km s \Gamma1 . Broad Hfi was only
conclusively detected in 1 AGN. This has important conse­
quences for the classification of such objects from spectra
with limited wavelength coverage, particularly in the case
when the region around Hff is not observed. For NLXGs
with z ? 0:25, this will frequently be the case.
The Hff/Hfi equivalent width ratios for the sample
range from 1.5 to 9.8, with a mean of 5.4. The mean val­
ues for the Seyfert and starburst galaxy samples are 4.6
(range 1.5 to 7.7) and 6.3 (range 3.8 to 9.8) respectively.
In order to derive Hff/Hfi intensity ratios, we have multi­
plied these equivalent width ratios by 0.64, i.e. assuming
a spectral index ff = 0:5 (see above). Note that this fac­
tor is relatively insensitive to spectral index, only changing
from 0.58 to 0.78 even over the wide range in spectral in­
dices, 0:2 ! ff ! 1:2 observed by Francis et al. (1992).
The mean Seyfert galaxy Hff/Hfi intensity ratio derived in
this manner is 2.9, consistent with the value predicted from
photoionisation models (Netzer 1990). For the starburst
galaxies, the mean value is 4.0, although it reduces to 3.6 if
the anomalously high Hff/Hfi ratio measured from the low
signal­to­noise spectrum of CRSS1406.7+2838 is excluded.
This value is slightly higher than the predicted range in the
Hff/Hfi intensity ratios for HII regions: 2:8 ! Hff=Hfi ! 3:0
(Aller 1974). However, given the typical uncertainties in

CRSS II 11
Figure 3 [NII]–6584/Hff -- [OIII]–5007/Hfi emission­line ratio diagram for the NLXGs observed in this paper. CRSS objects are
indicated by the filled circles and EMSS objects by the open circles. The division between AGN­like and HII­like spectra (dotted line)
is based on the criterion of Baldwin, Philips & Terlevich (1981).
the derivation of these narrow­line ratios (dominated by the
¸ 15 per cent uncertainity in the equivalent width mea­
surement of each individual line), this is not a significant
discrepancy. The lack of significant reddening from the
observed Hff/Hfi ratios is also consistent with the results
from the X­ray spectral analysis (Ciliegi et al. 1995) in
which none of the NLXG X­ray spectra (with the exception
of CRSS1412.5+4355) show any evidence for any intrinsic
absorption due to neutral hydrogen in excess of the galac­
tic value (although few NLXG have sufficient X­ray counts
to permit a detailed spectral fit). The Seyfert 1.8 galaxy
CRSS1412.5+4355 has an intrinsic neutral hydrogen X­ray
column density NH = 2:5 \Sigma 1:0 \Theta 10 20 cm \Gamma2 , correspond­
ing (for galactic gas­to­dust ratios) to a visual extinction
of A V = 0:14 mag (Zombeck 1990). This small amount of
extinction is consistent with the mild amount of redden­
ing implied by the intensity ratio Hff/Hfi = 3:3 derived for
CRSS1412.5+4355. Thus, while individual NLXG may ex­
hibit some reddening, it would appear that significant ob­
scuration is not a general feature of either the starburst or
Seyfert population in this sample.

12 B.J. Boyle et al.
As demonstrated in Paper I, the NLXG sample com­
prises between 15--35 per cent of the soft (2 keV) X­ray back­
ground. With an approximate ratio of 10:7 AGN:starbursts
identified in this paper, this suggests that the approxi­
mate contributions of the two populations also lie in the
approximate ratio 10:7 per cent. However, the numbers
of NLXGs identified are still small, and we can not rule
out equal contributions from both classes of object. Given
the composition of the NLXG sample, it is not surpris­
ing that the rate of cosmological evolution derived in Pa­
per I, LX / (1 + z) 2:6\Sigma1 , is so similar to that of QSOs
(L X / (1 + z) 3:0\Sigma0:2 ). Unified models of AGN (in which
the appearance of an object as a Seyfert 1 or 2 is merely
dependent on viewing angle) naturally imply that Seyfert
2s (or similar types) must evolve at the same rate as Seyfert
1s/QSOs. In addition, it is also known that starburst galax­
ies also undergo a rate of cosmological evolution in infra­red
luminosity L IR / (1 + z) 3:0\Sigma1:0 (Saunders et al. 1990)
which is consistent with that of QSOs in the optical and
X­ray regimes (see Boyle 1993).
4 CONCLUSIONS
We have obtained intermediate­resolution spectra of 17
NLXG identified from the CRSS and EMSS samples. Based
on their emission line ratios, we estimate that the sample
contains 7 starburst galaxies and 10 Seyfert galaxies. Six of
the Seyfert galaxies show evidence for broad Hff emission,
although only one conclusively exhibits broad Hfi emission.
The Seyfert types range from Seyfert 1.5 to 2. In addition,
3 of the starburst galaxies exhibit evidence for weak broad
(¸ 2000 km s \Gamma1 ) Hff emission. Thus, the NLXG sample as
originally identified in Paper I appears to be a heteroge­
neous mix of Seyfert and starbursts. Only the line ratios
distinguish the two classes. In all other characteristics they
share similar properties (z, optical magnitude, LX , and line
asymmetry distributions). If the two classes are powered
by different processes this is surprising. Further discrimi­
nants need to be searched for in larger samples. Both classes
contribute approximately equally to the 2 keV X­ray back­
ground at a level of between 7 and 17 per cent.
ACKNOWLEDGEMENTS
RGM acknowledges the receipt of a Royal Society Univer­
sity Research Fellowship. We are indebted to Dr Mike Irwin
for observing the two spectra obtained during spectroscopic
service time on the William Herschel Telescope. We also
would also like to thank Dr Gerard Kriss for providing help
with the SPECFIT routine. BJB acknowledges the support
and hospitality of the Smithsonian Astrophysical Observa­
tory. The X­ray data was obtained from the Leicester and
Goddard ROSAT archives. This work was partially sup­
ported by NASA grants NAGW­2201 (LTSA) and NAS5­
30934 (RSDC). The optical spectra were obtained at the
William Herschel Telescope at the Observatory of the Roque
de los Muchachos operated by the Royal Greenwich Obser­
vatory.
REFERENCES
Aller L.H. 1974, in Physics of Thermal Gaseous Nebulae,
Reidel, Dordrecht, p.77
Baldwin J. A., Philips M. M., Terlevich R. J., 1981, PASP, 93,
5
B¨oller Th., Meurs E.J.A., Brinkman W., Fink H., Zimmermann
U., Adorf H.­M., 1992, A&A, 261, 57
Boyle B. J., 1990, MNRAS, 243, 231
Boyle B. J. 1993, in Shull, J.M., Thronson, H.A., eds, The
Environment and Evolution of Galaxies, Kluwer, Dordrecht,
p.433
Boyle B. J., Shanks T., Georgantopoulos I. G., Stewart
G. C., Griffiths R. E., 1994, MNRAS, 271, 639
Boyle B. J., McMahon R. G., Wilkes B. J., Elvis M., 1995,
MNRAS, 272, 462 (Paper I)
Ciliegi P., Elvis M., Wilkes B. J., Boyle B. J., McMahon
R. G. 1995, MNRAS, submitted
Fabbiano G., 1989, Ann. Rev. A&A., 27, 87
Fabian A. C., Barcons X., 1992, ARAA, 30, 329
Filippenko A. V., Terlevich R.J., 1992, ApJ, 397, L79
Francis P. J., Hewett P. C., Foltz C. B., Chaffee F. H., 1992,
ApJ, 398, 476
Fruscione A., Griffiths R. E., Mackenty J. W, 1993, in
Crincarini,D. et al. eds. Observational Cosmology, ASP
Conf. Ser. 51, Astron. Soc. Pacif., San Francisco, p296
Georgantopoulos I. G., Stewart G. C., Shanks T., Griffiths
R. E., Boyle B. J., 1993, MNRAS, 262, 619
Georgantopoulos I. G., Stewart G. C., Shanks T., Griffiths
R. E., Boyle B. J., 1995, MNRAS, in press
Gonzalez­Delgado M. et al., 1994, ApJ, 437, 239
Griffiths R. E., Padovani P., 1991, ApJ, 360, 483
Heckman T. M. 1980, A&A, 87, 152
Irwin M. J., McMahon R.G., Maddox S.J., 1994, Spectrum, 2,
14
Mood R., Graybill R., 1963, Introduction to the theory of
statisitics, (2nd edition), p231
Netzer H., 1990, in Active Galactic Nuclei, Saas­Fee Advanced
Course 20, Springer, Berlin, p57
Roche N., Shanks T., Georgantopoulos I., Stewart G. C.,
Boyle B.J., Griffiths R. E., 1995, MNRAS, 275, L15
Roy J­R., Boulesteix J., Joncas G.,Grundseth B., 1992, ApJ
386, 498
Saunders W., Rowan­Robinson, M., Lawrence A., Efstathiou
G., Kaiser N., Ellis R. S., Frenk C. S. 1990, MNRAS, 242,
318
Stocke J. T., Morris S. L., Gioia I. M., Maccacaro T., Schild
R., Wolter A., Fleming T. A., Henry J. P., 1991, ApJS, 76,
813
Whittle M. 1985, MNRAS, 213, 1
Veilleux S., 1991, ApJS, 75, 383
Zombeck M. 1990, in Handbook of Space Astronomy and
Astrophysics, Cambridge University Press, p.103
This paper has been produced using the Blackwell
Scientific Publications T E X macros.