Prof. W. Butler Burton
Scientific Editor, The Astrophysical Journal
Leiden University Observatory
National Radio Astronomy Observatory
  
RE: MS 64761

Dear Prof. Burton:

We have carefully considered the referee's comments, and I include
our responses below.  I have electronically submitted all relevant
files.  Again, EPS files with "BW" in their name refer to black & white
versions of the files to be used in the print version of the journal.

Thank you for your attention to our manuscript.

Best Regards,

Danny Dale
March ??, 2006

------------------------------------------------------------------------

We would like to thank the referee for the obvious care they took in
developing the report.  We find all of the referee's comments and 
suggestions to be reasonable, and we have incorporated nearly all of the
suggested modifications.  Below we explicitly respond to some of the 
referee's comments/suggestions.  Any suggestions not addressed below 
were completely incorporated into the new version of the manuscript.

>My main (but not major) criticism is concerned with the way the SiII 
>line is advertised as an alternative for weak high-ionization lines 
>(like OIV).  I do not find the Si version of the diagram in figure 3 
>very impressive. Compared to the version with OIV the "dynamic range" 
>in the Si version is much lower, i.e. the spread on the y-axis is a 
>factor of ~10 lower. The observed trend seems to me to be a trend 
>mostly in PAH and not so much in SiII/NeII. Quite a number of AGN have
>a SiII/NeII ratio in the starburst range, and quite a number of HII 
>regions and star forming regions are lying in the AGN range of that 
>line ratio, as the authors admit in paragraph 5. Also, the SiII line
>will be shifted out the IRS range for redshifts larger than ~0.1, 
>which limits its diagnostic usefulness (at least for IRS). 
>Nevertheless, in the absence of faint OIV lines the Si line can be a 
>valid and valuable alternative.  The SIII/SiII ratio (Figure 5) seems
>to be a better tool, in this respect. In topic 7) below I make a 
>suggestion how this perhaps could be used to improve the diagnostics.

We agree that most of the separation in the [SiII]/[NeII] vs PAH 
figure is due to differences in PAH equivalent width.  We also agree
that the dynamic range in the [OIV]/[NeII] ratio is larger than that
in the [SiII]/[NeII] (a factor of ~30 versus a factor of ~5).  We
address below the referee's useful suggestion of pursuing the
alternative ratio of [SIII]/[SiII] vs PAH.  

>"Interestingly, the Genzel et al. study found that 70-80% of their 
>ULIRG sample is mainly powered by starburst activity, yet optical 
>studies typically show a much smaller percentage of ULIRGs powered
>by star formation (e.g., Veilleux, Sanders, & Kim 1999)."
>
>This statement is not quite correct. Veilleux, Sanders & Kim (1999)
>confirm the starburst nature of most ULIRGs. Only above a luminosity
>of 10E12.3 they find an increasing fraction of AGN-dominated ULIRGs.
>In their paper, and in a paper by Lutz, Veilleux and Genzel (1999), 
>these authors themselves claim that optical and infrared 
>classifications agree quite well, if one assumes that most of the 
>infrared bright LINERs are actually starburst driven, i.e. are not low
>ionization AGN. The same claim is made by the cited Taniguchi et al. 
>(1999) paper. 
>
>The two respective sentences should be re-phrased a little bit, to 
>reflect this properly.

The introduction has been restructured to more accurately reflect
the results of these and other prior studies.

>The extraction area (the 3x5 grid in arcseconds) for Long High is 
>missing, right?  I assume the maps from the different modules do NOT
>all have the same orientation, or do they?  This should be clarified.
>Varying orientation could contribute to the scatter in the diagrams.
>The 23"x15" extraction aperture is used for all modules, not just 
>Short-High, right?

Yes, the last paragraph in this section explains that the same 23"x15"
aperture was used to extract spectra from Short-High, Long-High, and
Short-Low.  The referee is correct that the Short-High and Long-High
maps are oriented differently on the sky.  But the 45"x33" Long-High
map fully encompasses the smaller Short-High map, thus enabling a 
comparison over the ~same 23"x15" area.  [ 3x5 implies the 
observational grid, not the coverage in arcseconds ]  We have adopted
the referee's suggestion and dropped the confusing term 'Short-High' 
that prefaced 'extraction' in Section 3.2.

>Si (like Fe, Ca, Mg) is, at least in principle, susceptible to 
>depletion onto dust grains.  Does this play no role here?  One might 
>argue, that this paper is about empirical findings.  Sturm et al. 2002
>have also found that SiII can be used for this kind of diganostic 
>purposes (at least when normalized to Hydrogen recombination lines).
>Yet, Si depletion could, at least, be responsible for (some of) the 
>scatter in the diagrams.  If the authors have some thoughts on this 
>they should perhaps add a few sentences here.   
>
>Again, in addition to XDRs and interstellar density, dust depletion 
>could play a role to explain the Si diagnostic. 

The referee raises an interesting point.  Heavy elements such as Si, Mg,
and Fe may be returned to the gas phase by dust destruction (e.g., 
sputtering) in regions subject to strong shocks caused by stellar winds,
starbursts, and AGN activity.  So according to this reasoning perhaps 
(gas phase) Si lines are stronger in active galaxies.  Words to this
effect have been added to the manuscript to include this possibility.

>I am surprised about the low value for the 100%AGN anchor in the 
>Dale et al. study: OIV/NeII=0.3.  This is almost a factor 10 below
>the value used in Sturm et al. 2002, which was, according to their
>figure 7, about 2.5.  How can that be if a significant fraction of the
>Seyferts in this study is actually taken from Sturm et al. (2002)?
>Sturm used only those AGN which were free of starburst admixture in 
>their spectra for the average 100%AGN.  Could it be that the low value
>of 0.3 results from some of the additional Seyferts having some 
>contribution from star forming regions in the extraction aperture?
>Or did the authors include LINERs for the determination of the average
>AGN value?  If so, I would strongly suggest to exclude them.  This 
>should be checked and explained. 

The 100% AGN anchor point from our study was determined using only the
Seyfert data points.  It is a factor of ~3 lower than that shown in 
Figure 5 of Genzel et al. (OIV/NeII~1, which is not inconsistent
with the average, on a log scale, for the Sturm et al distribution).  
We have added a note to the text that this difference may arise from
small number statistics.  We note here that archival 6.2um PAH data 
are only available for 9 of the 29 sources studied in Sturm et al.;
the Sturm et al. data comprise only a fraction of the archival data
displayed in our Figure 3.

>Because of the dependance on the metallicity the NeIII/NeII ratio is 
>not an unambiguous tracer of the excitation type (AGN vs. star 
>formation).  Figure 5 to me looks like the SIII/SiII ratio alone is
>already a good separator (like, e.g., OIV/NeII, although the trend is
>not as pronounced).  The dynamical range is larger than for NeII/SiII.
>So, Figure 3 (bottom) and 5 both have weaknesses and strengths.  Why
>not take the strengths of both diagrams and combine it into a better 
>diagram, i.e. SiII/SIII (y-axis) vs. PAH (x-axis)?  Using SiII/SIII~5
>and ~0.25 (to be checked) as anchors for 100% and 0%AGN one could 
>replace the y-axis in figure 3 (bottom panel), i.e. SiII/NeII, with 
>SiII/SIII, which would presumably improve that diagram, wouldn't it?
>I do not know, though,  why, in principle, SIII should be the better
>line than NeII, when compared to SiII.  But this seems to be an 
>empirical finding of this study.  The only disadvantage of this new
>diagram would be that SIII34 can be weaker than NeII (compare Verma 
>et al. 2003, underabundance of S), and that it lies in the Long-High
>module of IRS, which is much noisier than Short-High (but one could use
>SIII 18 instead).  Can the authors check this, please, and comment on
>this?

We have adopted this fine idea and added a third panel to Figure 3, one
that includes the SiII/SIII ratio.  We prefer to keep the SiII/NeII
panel for the reasons outlined by the referee.

>Since the authors mention their optical spectroscopy, which they use
>for an optical classification of the SINGS objects and for an 
>evaluation of their diagnostic diagrams, it would be good if these 
>types could be given in Table 1, as is done in Table 3 for the archival
>targets (I know, Table 1 is already quite wide). 

Some of our classifications still derive from the literature, so we are
in the process of obtaining a bit more optical long-slit spectroscopy 
for the SINGS project.  The final set of data will (hopefully) be 
obtained by the end of May 2006.  Our intention is to soon afterwards
publish a paper that more completely presents the metallicities and 
nuclear classifications for the SINGS project.

>In the HII region spectrum there seems to be a discrepancy between the
>low and high resolution spectra in the overlap region (10-14 micron).
>Can the authors explain this?  Is it consistent with the 25% flux 
>calibration uncertainty they claim in section 3.1?  Or is it only an
>apparent discrepancy?  Or due to different aperture orientations?

The variable offset between the low resolution and high resolution data
portrayed in Figure 1 is largely due to the lack of subtracting the sky
for the high resolution data.  Each target lies in a different portion 
of the sky, with a different sky level.  We have reworded the caption to
better explain the discrepancy.

>I find it very difficult to distinguish between star-forming objects 
>and AGN in the figure. I wonder if the figure could be improved. E.g.
>individual error bars could be replaced by a "typical" (average) error
>bar in one corner of the diagram. It could also help to use more colors
>(at least for the electronic version of the paper), and to increase the
>symbol size a bit.

We have increased the color usage and symbol size in this figure.  We
prefer to retain the individual error bars, as they help to show that 
the data on the low side of the envelope are not inconsistent with the
low density limit.