Dear Sir/Madam,

Please find below our response to the issues you raized in your last referee report. 
Our response is structured in the following way: first we list your initial comments, 
then our first response, next your second remarks and last our current response. 
We hope that we have answered all of your questions/remarks to your satisfaction.

Regards,

Jeroen Bouwman

********************************************************************************

2. The results of the error analysis (p.9) look plausible, but the method is
unorthodox. It would help readers orient themselves if the authors would comment
briefly on how such error estimates would compare to anything more conventional
(e.g. the parameter range that increases chi-squared/degree of freedom by unity
compared to the minimum): how conservative or aggressive is this uncertainty
estimate likely to be? Or turn them into confidence intervals, which they almost
are. 

Authors reply: 
The method we use is the most robust and flexible way to get an accurate error
estimate. The error estimate is, to our best knowledge and in the tests we
performed, the same as when using a delta-chi2=1 method and represents a one
sigma error. It is, in fact, the preferable method in most cases and also the
first method suggested by Numerical Recipes. The reason that it is not so
frequently employed is not for reasons of accuracy or robustness, but for the
simple reason that this method requires one to make a large number of fits,
which is usually computationally very demanding. In our case making a single fit
is extremely fast (it is a linear least squares fit, which can be done in a
fraction of a second) making this method feasible. To help the reader we have
added the reference to chapter 15.6 of Numerical Recipes for more details on the
applied method, in the last paragraph of section 2.3.

Referee's response:
Fair enough. Please remember to specify the edition of NR in your reference.
It's not in that section in my copy. 

Authors second reply:

The reference to correct version of NR is: "Numerical Recipes in Fortran , The art of Scientific Computing", second edition, ed. William H. Press, Saul A. Teukolsky, William T. Vetterling, Brian P. Flannery., Cambridge University Press, 1992
The chapter is called "Modeling of data" (Chaper 15) and the relevant subsection is 15.6, titled
"Confidence limits on estimated model parameters"


********************************************************************************

3. Also worth a stray mention: PAH features show up in these spectra far more
frequently than they do in other samples of Class II objects (e.g.
Kessler-Silacci et al 2005, Furlan et al 2006). This seems likely to be a result
of the selection by FEPS of stars of earlier type (G - early K) than is typical
of T Tauri stars (K-M). 

Authors’ reply: 
This is indeed an interesting point. Certainly what the 6.2 and 11.3 micron PAH
feature is concerned, there seems to be a correlation with spectral type and
strength of the PAH bands.
This is for instance the conclusion of a recent C2D Spitzer legacy study by
Geers et al 2006, which found no PAH emission for sources later than G8. In that
sense, our K and M0 systems should not have had any PAH emission. However, in
the Geers et al study they did not do an extensive analysis of the 8 micron
region. As Peeters et al already noticed, the sources showing a PAH band at 8.2
micron have much weaker 6.2 and 11.3 micron bands. These sources could have been
missed in a study focussing on these latter 2 bands.
Also, the difficulty with the 8 micron region is that is coincides with the
strong silicate band. The relatively weak PAH band appears as a weak shoulder on
the blue wing of the 10 micron silicate feature and could be easily overlooked.
Just looking at figure 1, we would not have claimed to see any PAH emission
apart from those seen in the spectrum of the HD143006 system. One can only make
a firm conclusion concerning the presence of PAH emission bands after carefully
modelling the silicate emission bands. 
The claim, therefore, that systems with lower mass than those studied by us do
not have any PAH emission should only be made after a carefull fitting of the
silicate emission band. We have added a few sentences to section 4.3 to reflect
the above discussion, addressing the referee's comments.

Referee's response:
I like the new additions to section 4.3. Note, however, that there's still some
repair necessary to remove the effects of hasty editing: for instance, the stray
phrase between section 4.3 and the acknowledgements, and the fifth sentence,
which is ungrammatical and incomplete enough that I'm not sure how to interpret
it. 

Authors' second reply:

We have edited section 4.3. We have re-written the mentioned sentence and removed
the stray phrase. The questionable sentence now read:

"Characterizing the PAH population is also important for understanding the 
observed gas temperatures in disks, as the stochastically heated molecules strongly influence the temperature through photo-electric heating. Since the gas temperature determines the pressure scaleheight of the disk, the presence of PAH molecules can, therefore, have a substantial 
influence on the disk geometry. Five out of the seven TTS systems of our sample show a band at 8.2~$\mu$m we identify with emission from PAH molecules. This is the first time this band is observed in low-mass pre-main-sequence systems."


********************************************************************************


5. They have only seven objects, of uncertain age and point of origin, with
which to work. The statistical significance of correlations, of whatever
magnitude, is questionable in such a small sample, and the authors have
presented no quantitative estimates of the confidence that can be placed in the
correlations they find. There are several other Spitzer observing teams
currently publishing similar analyses on much larger samples. This places the
authors at risk of claiming trends (or lacks thereof) that are not reproduced in
the larger studies, and thereby tarring the other, solid results. And the FEPS
survey was aimed at solar analogue stars of intermediate age (>3 Myr), in the
field as well as in associations, and it is thus somewhat accidental that they
find even seven objects with optically-thick dusty disks; the seven objects
cannot be taken as typical of the disks around T Tauri stars, which are mostly
younger, of later spectral type, and strongly concentrated in clust
 ers. (The lone M star on the list truly was a gatecrasher in this party, as
explained in Appendix B, p. 26.)

Authors’ reply: 
How typical should we expect our results to be compared to other studies of
mid-IR spectra obtained with Spitzer for T Tauri stars? As mentioned in the
introduction, five of our targets were identified from an unbiased sample of
FEPS targets spanning a range in age from 3-30 Myr (Silverstone et al. 2006).
These stars represent older actively accreting classical T Tauri stars with
remnant primordial gas rich disks. To that sample, we have added two targets:
one with an optically-thick primordial disk chosen for the sample on the basis
of its circumstellar properties (HD 143006) and another young star with a disk
which was observed by accident and was not originally part of the FEPS sample
(see appendix B). None of our targets are particularly unusual in terms of their
stellar or circumstellar properties compared to larger samples of T Tauri stars.
Because they are all slightly older, the accretion rates of these objects tend
to be low (< 10^-8 Msun/yr; Pascucci et al. 2007) and 
presumably they lack remnant infalling envelopes. Perhaps for these reasons,
the correlations we observe were discernable from such a small sample of seven.
Though larger samples should be statistically more significant, the problem with
a large, and most importantly young sample of TTauri systems, is that one might
have a sample which is by far not as uniform as the sample we study here. Our
sample represents those systems in which the disk survives for almost the entire
PMS evolution, and are the last ones to be dissipated. In a much younger sample,
one has mix of systems of which the disk will be dissipated at an early stage
and systems similar as studied in this paper. This difference in evolutionary
time scales might also be reflected in differences in the dust composition and
processing. By mixing these systems as one does in a young sample, any
correlation might be lost. 

Referee's response:
It's even easier to see a correlation if you have only two objects instead of
seven. I wouldn't say that this is a virtue of small sample sizes! And perhaps
the correlations would have been even larger if two of the objects had been
chosen from the Small Magellanic Cloud instead of the Greater Solar
Neighborhood! 

A correlation between two parameters, however large its value, is meaningless
unless the sample is composed in a fashion that controls other important
parameters in a well-defined way, and unless one can demonstrate that a
correlation of that value cannot have been drawn from a random sample. 

In your reply to the next point, you demonstrate admirably that the correlations
you observe are statistically significant. But the fact remains that these seven
objects are a heterogeneous sample: they have different origins, and have been
subjected to different environments through their lives, in addition to lying on
the high-age and high-stellar-mass end of the range for classical T Tauri
systems. There may be systematic differences between parts of the sample that
are due to the differences in origin and environment, for example, instead of
processing within the disks. No claim in the paper is so strident that I would
consider it dangerous to publish, so I will not *insist* on any modifications on
this account; nevertheless, and in view of the heterogeneity of this small
sample, I would still *urge* you consider tempering or further qualifying the
conclusions that are based on trends within the sample, in case the potential
systematic differences get illuminated some day by observations of larger, 
homogeneous samples of objects in this age range and younger. 

Authors Second reply:

We offcourse realize that it is a small sample, only 7 objects is just a start and we
fully agree that a larger sample is needed to make final conclusions. We would only like
to make the remark that sample size alone will not nesseserely give a statistically more 
meaningfull result if the sample is extremely inhomogeneous, consisting of different types of 
objects following different evolutionary paths. Concerning the homogeneity of our sample, the systems might be from different regions, but their environment can't be all that different. They are all from low-mass star forming environments (nothing like Orion) and they can't have had strong interactions with other stars else they would have lost their disk.

Having said this, to determine how the disk evolution depends on stellar age, mass, binarity, 
environment etc. one needs much larger samples, spanning a sufficiently large range of parameters. 
This is obviously not the case with our sample, and indeed, as the referee remarks, 
systems with a lower mass might behave quite different. Also, the systems of our sample have retained their disks longer than average, which could imply that other, more shortlived disks, behave differently. It is our feeling that this is the most stringent limitation of our study. 

To make sure the reader is aware of these limitations and to properly address the concerns of the referee, we have added the following sentences to the first paragraph of the discussion:

"Note that our sample only spans a limited range in stellar parameters and consists of older, 
long surviving disks. Our conclusions are applicable to similar systems, but 
possible effects of stellar properties or age on the evolution
of circumstellar disks can not be address directly by this study. For this our results will
have to be compared to the disk properties of a larger sample, spanning a wide range in stellar properties and evolutionary stages."
 

********************************************************************************

7. Page 22: FU Ori is not a very apt comparison here, as the silicate emission
feature in this object may arise in the envelope rather than the disk (Green et
al 2006, Hartmann 1998, Adams et al 1987).

Authors’ reply: 
As the referee rightly remarks, the emission of FUORI objects, given the young
nature of these objects, could be strongly influenced or dominated by material
in an envelope surrounding the system. The Green et al study and also a resent
study by Quanz et al show that there is a large variation between the FUORI
objects, where the emission of some of them is clearly envelope dominated.
However, in the case of FU Ori itself, this is probably not the case as also
Green et al state in section 4.3 : "We therefore reinforce the conclusion of
KH91 that the SED of FU Ori can be explained by a flared disk without
significant envelope contributions in the IRS wavelength range." The Quanz et al
2006 study of FU Ori, using spatially resolved interferometric mid-IR
observations obtained with VISIR also suggests that the mid-IR emission of FU
Ori originates from a disk rather than an envelope. We therefore think that the
comparison we make with our TTS sample is valid, showing that also in the disks 
of the youngest Class II objects a significant change in grain size
compared to the ISM can occur. 

Referee’s response:
OK, so the vote is 3-1 (only Adams dissenting) against the silicate-feature
emission in FU Ori arising in an envelope; the comparison to FU Ori itself may
therefore be useful. But in the text you emphasize that FU Ori is the
prototypical case of its class,” which will surely tempt the reader to conclude
that all FUOrs satisfy the comparison. Perhaps it would be better to make that
sentence read “Further, recent studies of FU Orionis objects show no sign of
crystalline silicates, independent of whether the silicate emission originates
in disks or envelopes (Quanz et al 2006, Green et al 2006).”

Authors second reply:

We have changed the sentence as suggested by the referee.


********************************************************************************

10. The authors generate continuum spectral indices by performing synthetic
photometry on the spectra, with filters at wavelengths 13, 24 and 33 microns (p.
12, middle paragraph). Unlike 13 microns, the longer two wavelengths would seem
from the spectra to have significant silicate-dust emission: crystalline and
amorphous at 24, and the longest wavelength forsterite feature at [33] microns. 
This is supposed to be about the continuum; why not choose wavelengths with only
continuum emission? 

(Note typo fixed: 33, not 36.)

Authors’ reply: 
As one can see from figure 1, the amorphous silicate band at around 10 micron
can dominate the continuum, with the strongest band a factor 3.6 stronger than
the underlying local continuum. This is offcourse the reason why we choose to
determine a synthetic photometry point at around 13 microns, to prevent the 10
micron silicate band influencing our results.
We choose 24 micron as to have a direct comparison to the MIPS 24 micron
measurements. The 33 micron band represent the wavelength the longest
wavelengths (so probing the coldest dust) with the IRS instrument at which one
could obtain good photometry without being influenced by the excess noise at
wavelengths longer than 35 micron. 
The 24 and 33 micron bands are also far enough apart that their flux ratio is a
meaningfull measurement of the SED slope. A further advantage of having to
photometry points within the LL1 slit, is that the flux ratio is independent of
absolute flux calibration 
Indeed, as the referee notices, the bands at around 24 and 33 microns can be
influenced by crystalline silicate bands. However, the typical band strengths
are around 10% over continuum, so the influence is far smaller than the
amorphous silicate band at 10 micron would have. Also, both bands are expected
to be similarly influenced by forsterite emission (see for instance also figure
11) so that any influence of forsterite emission to the flux ratio of F33/F24
can be expected to cancel out. We have checked other flux ratios, i.e.
F30/F13, which yields identical results as to those plotted in figure 6. We
prefer to keep the current synthetic photometry points and central wavelengths
to allow for a direct comparison to other FEPS studies which also use these
wavelengths. 

Referee’s response:
It's not just crystalline silicates at 24 microns; the amorphous silicate
feature  still contributes at that wavelength. It is true that 24 and 33 microns
are far enough apart that they would provide a useful SED-slope measurement if
they represented continuum emission. It is also true that the silicate features
have tolerably small contrast in these objects. But they don’t always, in other
groups of disks; other observers, who will have to deal with larger
crystalline-silicate features, will prefer to compare their results to real
*continuum* SED slopes derived from your data. And that’s not easy to get from
your figures.

Since you have already checked things with a band at 30 microns (which *is* all
continuum), and since you have internal-consistency reasons for preserving the
bands you started with, I **strongly** suggest that you add a 30-micron
synthetic filter to the discussion, add a F_30/F_13 frame to Figure 6 (with
perhaps F_33/F_13 plotted as well), and use this to demonstrate, in the text,
that the results involving the crystalline-free 30 micron filter, and the 
"contaminated" 33 micron filter, are similar. You have spectra here, not 
photometry; why endow your analysis with the uncertainties that would attend
photometry, if you don't have to?

Authors second reply:

We agree with the referee that the use of a photometry band near the 18 micron 
amorphous silicate feature, next to the contamination by crystalline silicate
emission bands, could e problematic. We have, therefore, replaced the top panel of
Figure 6 with a new figure showing the correlation between the F30/F13 flux ratio 
and the 10 micron silicate band strength. The 2 panels of Figure 6, showing the 
F30/F13 and F70/F13 flux ratio's probe a broad enough wavelength range and thus 
temperatures, to link overall disk structure to the measured grain size of the
amorphous silicates. The chosen wavelengths now only probe the continuum,
preventing any confusion with the crystalline bands. We have changed the 
text where Figure 6 is discussed (section 3.2) accordingly.  


********************************************************************************