source
stringlengths
1
2.05k
target
stringlengths
1
11.7k
Thus. by studyiug the observed ECGD anisotropy as a function of energy. it may be possible to diseutaugle the spectrum aud. amplitude
Thus, by studying the observed EGB anisotropy as a function of energy, it may be possible to disentangle the spectrum and amplitude
where II:gr and CO:gr represent 11 and CO on the grains.
where $:gr$ and $:gr$ represent H and CO on the grains.
The exceptions are and where we follow ?. in assuming that these ions form No rather than NII and ND when thev Ireezeout.
The exceptions are and where we follow \citet{roberts04} in assuming that these ions form $_2$ rather than NH and ND when they freezeout.
All species are assumed to [reezeout. except lor Ile which has a very low binding energy aud is therefore easily thermally clesorbed even αἱ very low temperatures.
All species are assumed to freezeout, except for He which has a very low binding energy and is therefore easily thermally desorbed even at very low temperatures.
Any that hits a exgrain is neutralized and returned immediately to the egas.
Any $^+$ that hits a grain is neutralized and returned immediately to the gas.
We include thermal desorption and desorption due (o cosmic rav heating of grains.
We include thermal desorption and desorption due to cosmic ray heating of grains.
Cosmic rav heating is able to maintain a low level of some volatile molecules such as CO and Ns in the cold midplane of our models.
Cosmic ray heating is able to maintain a low level of some volatile molecules such as CO and $_2$ in the cold midplane of our models.
These molecules can destroy ions such as IL, and its deuterated isotopes. aud produce ions such as | and in the midplane aud (thereby affect the ionization level in this region.
These molecules can destroy ions such as $_3^+$ and its deuterated isotopes, and produce ions such as $^+$ and $^+$ in the midplane and thereby affect the ionization level in this region.
We have therefore run models without cosmic ray heating to determine how the inclusion of this process will alfect molecular abundances ancl the ionization level in cold regions of the disk.
We have therefore run models without cosmic ray heating to determine how the inclusion of this process will affect molecular abundances and the ionization level in cold regions of the disk.
The rates for cosmic rav heating are taken from ?.. using updated binding energies for some species. notably. CO. [or which we use the value determined by 2..
The rates for cosmic ray heating are taken from \citet{hh93}, using updated binding energies for some species, notably CO, for which we use the value determined by \citet{oberg05}.
Table 7 gives the binding energies (775) used in our models.
Table \ref{tab:be} gives the binding energies $E_D$ ) used in our models.
Thermal desorption rates are calculated. [rom where 7,, is the grain temperature. and 7j is the frequency of oscillation between the absorbate and the surface given bv where n, is the surface density of sites ( 1.5 x LO’ 7) and m is the mass of the accreting species.
Thermal desorption rates are calculated from where $T_{gr}$ is the grain temperature, and $\nu_0$ is the frequency of oscillation between the absorbate and the surface given by where $n_s$ is the surface density of sites $\sim$ 1.5 $\times$ $^{15}$ $^{-2}$ ) and $m$ is the mass of the accreting species.
? recently observed a high column density of IDO in the disk around DM Tan.
\citet{cec05} recently observed a high column density of HDO in the disk around DM Tau.
They found NOIDO) ~ 1.6 x LOM ? in the outer disk. where the clensityv is ~ LO° * and the temperature is < 25 Ix. The emission comes from above the midplane and corresponds to a relatively high fractional abundance of ~ 3 x 7.
They found N(HDO) $\sim$ 1.6 $\times$ $^{13}$ $^{-2}$ in the outer disk, where the density is $\sim$ $^6$ $^{-3}$ and the temperature is $<$ 25 K. The emission comes from above the midplane and corresponds to a relatively high fractional abundance of $\sim$ 3 $\times$ $^{-9}$.
At these densities and temperatures. water is expected to be completely removed Irom the gas by accretion onto erains. where it will remain in the absence of a nonthermal desorption process.
At these densities and temperatures, water is expected to be completely removed from the gas by accretion onto grains, where it will remain in the absence of a non–thermal desorption process.
Cosmic rav heating is not efficient enough to remove such a strongly bound molecule as water.
Cosmic ray heating is not efficient enough to remove such a strongly bound molecule as water.
7? suggested that photodesorption. arising from the action of the interstellar farUV. field could be efficient enough to retain water vapor in the eas phase in the cold outer disk. and so account for the
\citet{dom05} suggested that photodesorption, arising from the action of the interstellar far–UV field could be efficient enough to retain water vapor in the gas phase in the cold outer disk, and so account for the
over near 2)2—1.2 and obtains a similar shallow slope (dashed line in Figure 2).
over near $\beta \approx -1.2$ and obtains a similar shallow slope (dashed line in Figure 2).
For LBCs in our sample with jm1.2. the clumpy model predicts larger values of IRN than the IRA-.> relation of MIIC'99.
For LBGs in our sample with $\beta > -1.2$, the clumpy model predicts larger values of IRX than the $\beta$ relation of MHC99.
ILwius defined the UV reddenius relation expected for the dust models. we again use the same technique to predict the N-vav huuinosities of the LBC sample.
Having defined the UV reddening relation expected for the dust models, we again use the same technique to predict the X-ray luminosities of the LBG sample.
The homogeneous model slehthy πάς predicts the mean N-ray luminosity (L1<10H Creswoe 1) compared to the observed value.
The homogeneous model slightly under predicts the mean X-ray luminosity $1.1 \times 10^{41}\,$ $\,$ $^{-1}$ ) compared to the observed value.
Within strict error τς Guoclel and observed) the homogeneous model agrees with the observation.
Within strict error limits (model and observed) the homogeneous model agrees with the observation.
The chuupy model slightly over predicts it (LT«10H Cres Ly but the observed mean is well within he lo eror.
The clumpy model slightly over predicts it $4.7 \times 10^{41}\,$ $\,$ $^{-1}$ ), but the observed mean is well within the $1\sigma$ error.
The histogram in Panel C (Panel D) of Figure 3 represeuts the homogeneous (cluupy) model.
The histogram in Panel C (Panel D) of Figure 3 represents the homogeneous (clumpy) model.
To within the uncertainties iu our technique. we are unable o differentiate between the empiicallv derived relation of MIIC99 aud either of the SMC-dust/shell geometry nodels.
To within the uncertainties in our technique, we are unable to differentiate between the empirically derived relation of MHC99 and either of the SMC-dust/shell geometry models.
Witt Cordon claim that the clumpy model reproduces he effective starburst attenuation curve of Calzetti (1997: Calzetti et al.
Witt Gordon claim that the clumpy model reproduces the effective starburst attenuation curve of Calzetti (1997; Calzetti et al.
2000).
2000).
We have calculated the UV reddening relation for the Calzetti attenuation curve (dotted line of Figure 2) using the same method aud find it very similar to the elunipy model over the narrow range of j| measured for the LBGs.
We have calculated the UV reddening relation for the Calzetti attenuation curve (dotted line of Figure 2) using the same method and find it very similar to the clumpy model over the narrow range of $\beta$ measured for the LBGs.
When we use the Calzetti effective attenuation to predict N-rav cussion. we fud the mean luuinosity L5s10H cress ly Gs essentially the same as that from the ‘hunpy model aud also in agreement with the observed value (Figure 3: Panel E).
When we use the Calzetti effective attenuation to predict X-ray emission, we find the mean luminosity $4.5 \times 10^{41}\,$ $\,$ $^{-1}$ ) is essentially the same as that from the clumpy model and also in agreement with the observed value (Figure 3: Panel E).
We should note the effect our choice of intrinsic starburst imodel has ou the computed UV reddening relations.
We should note the effect our choice of intrinsic starburst model has on the computed UV reddening relations.
Choosing a higher metallicity (16. Z..) stellar population will produce a redder iutrinsic (unextincted) spectral slope CAJj~ 0.1)but would have little effect on IRN values for 3>2.
Choosing a higher metallicity (i.e. $Z_{\odot}$ ) stellar population will produce a redder intrinsic (unextincted) spectral slope $\Delta\beta_0 \sim 0.1$ ) but would have little effect on IRX values for $\beta > -2$.
Differences in either burst age or the IME upper mass lint will effect both ου aud IRN in the seuse that older bursts or smaller upper mass πες will redden y aud lower IRN.
Differences in either burst age or the IMF upper mass limit will effect both $\beta_0$ and IRX in the sense that older bursts or smaller upper mass limits will redden $\beta_0$ and lower IRX.
This would then result ina sanaller predicted N-rav flux for the LBC sample.
This would then result in a smaller predicted X-ray flux for the LBG sample.
We can use the machinery we have developed to test the extreme lypothesis that LBCs suffer no £u-UV extinction.
We can use the machinery we have developed to test the extreme hypothesis that LBGs suffer no far-UV extinction.
[f the LBC sample is actively forming stars and harbors little or uo dust. we can reasonably assuue that the far-UV flux is unattenuated auc dominates the bolometrie bhDuiuositv.
If the LBG sample is actively forming stars and harbors little or no dust, we can reasonably assume that the far-UV flux is unattenuated and dominates the bolometric luminosity.
The reddened UV colors mist then be interpreted ax a simple effect of burst age.
The reddened UV colors must then be interpreted as a simple effect of burst age.
Although continuous star formation modes could not casily account for the rauge of J measured in the LBG sauple. imstantaneous burst modes with ages betwoeeou 302fna100 Mr can naturally redden enoush to explain the observed UV properties (Leitherer et al.
Although continuous star formation modes could not easily account for the range of $\beta$ measured in the LBG sample, instantaneous burst modes with ages between $30 > t_{\rm burst} > 100$ Myr can naturally redden enough to explain the observed UV properties (Leitherer et al.
1999).
1999).
ILlowever. uuder this scenario. the mean ταν luuinosity is uuder predicted by. a factor of 6 (L6«10/9 eres. 1).
However, under this scenario, the mean X-ray luminosity is under predicted by a factor of 6 $4.6 \times 10^{40}\,$ $\,$ $^{-1}$ ).
Given the large observational aud model uucertaiuties the difference is ~26 (Figure 3: Paucl FE).
Given the large observational and model uncertainties the difference is $\sim2\sigma$ (Figure 3: Panel F).
Evidence is beginning to emerge that 28 keV Χανς are a good star formation rate indicator (Ranalli ct al.
Evidence is beginning to emerge that 2–8 keV X-rays are a good star formation rate indicator (Ranalli et al.
2002).
2002).
Iu local starbursts. the 28 keV N-rav enission is believed to be produced. primarily bv hiehauass N-rav binaries (IDNEND: Persic Rephacli 2002).
In local starbursts, the 2–8 keV X-ray emission is believed to be produced primarily by high-mass X-ray binaries (HMXB; Persic Rephaeli 2002).
This suggests that the Brandt et al. (
This suggests that the Brandt et al. (
2001) stacking technique applied to the IIDE-N LBC sample may be detecting binary stars at ~~3.
2001) stacking technique applied to the HDF-N LBG sample may be detecting binary stars at $z \sim 3$.
Low luminosity AGN aud the class of ultraluninous N-rav sources (ULNs or INOs: Colbert AMiushotzisy "ur which may be the beamed emission frou IINNDs (ie. Roberts et al.
Low luminosity AGN and the class of ultraluminous X-ray sources (ULXs or IXOs; Colbert Mushotzky 1999), which may be the beamed emission from HMXBs (i.e., Roberts et al.
2002) or a new class of intermediate mass (107 10! AL.) black holes. may also contribute to the 2S keV flux.
2002) or a new class of intermediate mass $10^2$ $10^4$ $_{\odot}$ ) black holes, may also contribute to the 2–8 keV flux.
Furthermore. 28 keV. N-vavs suffer little intrinsic absorption and can escape regions where the far-UY tracers ofstar formation may be heavily extincted by a dusty interstellar mediuu.
Furthermore, 2–8 keV X-rays suffer little intrinsic absorption and can escape regions where the far-UV tracers of star formation may be heavily extincted by a dusty interstellar medium.
For these reasons. the 28 keV N-rav huninosity strongly correlates with the bolometric huninosity of star fornüng galaxies.
For these reasons, the 2–8 keV X-ray luminosity strongly correlates with the bolometric luminosity of star forming galaxies.
This mcaus that 28 keV A-vavs can serve as a proxy for the far-IR thermal dust emission at hiel-:.
This means that 2–8 keV X-rays can serve as a proxy for the far-IR thermal dust emission at $z$.
We have derived the bolometric to 28 keV Nav correlation for a local souple of normal and starburst ealaxies and have used it. iu combination with several UV reddening schemes. to predict the mean 28 keV ταν lununesity for a sample of 21 spectroscopically confirmed high redshift Lyiman-break galaxies.
We have derived the bolometric to 2–8 keV X-ray correlation for a local sample of normal and starburst galaxies and have used it, in combination with several UV reddening schemes, to predict the mean 2–8 keV X-ray luminosity for a sample of 24 spectroscopically confirmed high redshift Lyman-break galaxies.
Tis simple analysis demonstrates that LBCs can not have far-IR to &u-UV flux ratios similar to those found for nearby ULICs. nor are they likely to he unatteuuated by dust.
This simple analysis demonstrates that LBGs can not have far-IR to far-UV flux ratios similar to those found for nearby ULIGs, nor are they likely to be unattenuated by dust.
Of the extinction methods considered. we find that the starburst reddening relation of MIIC99 is the most accurate predictor of the mean X-rav Duuinositv for the sample.
Of the extinction methods considered, we find that the $\beta$ starburst reddening relation of MHC99 is the most accurate predictor of the mean X-ray luminosity for the sample.
The very similar reddening relations derived frou Witt Cordon (2000) extinction models of low metallicity dust in a shell ecometiv aud the Calzetti et al. (
The very similar reddening relations derived from Witt Gordon (2000) extinction models of low metallicity dust in a shell geometry and the Calzetti et al. (
2000) effective starburst attenuation curve are also consistent with the observed X-ray enission.
2000) effective starburst attenuation curve are also consistent with the observed X-ray emission.
These results provide additional evidence that LDCis can be considered as scaled-up local starbursts.
These results provide additional evidence that LBGs can be considered as scaled-up local starbursts.
Equally iuportanut. it sugeests that Πο may be a reasonable tool for estimating the UV extinction of high redshift LDCs.
Equally important, it suggests that $\beta$ may be a reasonable tool for estimating the UV extinction of high redshift LBGs.
If this is the case. all 21 LBCs iu this sample have «που<3.1 Mae with a mean Qaedian) of 1.1 (1.5) Mag iurplviug a mean dadust correction factor of ~I.
If this is the case, all 24 LBGs in this sample have $A_{1600} < 3.1$ Mag with a mean (median) of 1.4 (1.5) Mag implying a mean dust correction factor of $\sim4$.
This moderate level of UV. extinction is consistent with the results of Papovich et al. (
This moderate level of UV extinction is consistent with the results of Papovich et al. (
2001) who find typical ccorrection factors of 3Ll from an analysis of the UV-optical spectral energv distributions for a sample of 33 IIDF-N LBGs which includes 23 from our sample.
2001) who find typical correction factors of 3–4.4 from an analysis of the UV-optical spectral energy distributions for a sample of 33 HDF-N LBGs which includes 23 from our sample.
Similar UV extinctions are also deduced for larecr LBG samples by Steidel et al. (
Similar UV extinctions are also deduced for larger LBG samples by Steidel et al. (
1999) and MIICO99.
1999) and MHC99.
Our results are the first to use low extinction X-rav enüssion to test aud confini these clans.
Our results are the first to use low extinction X-ray emission to test and confirm these claims.
Although it is tempting to use the results developed here to predict the N-rav fluxes of individual LDCs. we caution the reader that our results are statistical.
Although it is tempting to use the results developed here to predict the X-ray fluxes of individual LBGs, we caution the reader that our results are statistical.
The accuracy of the estimated N-vav cuussion for any suele LBC in this sample is only ~0.18 dex.
The accuracy of the estimated X-ray emission for any single LBG in this sample is only $\sim0.48$ dex.
With this iu nüud. it is interesting to note that when the UV reddening relations of MIIC99. Witt Cordon (clampy model}. aud Calzetti are applied. to this technique. more than of the 2ὃ keV flux originates from the half of the sample (12) with
With this in mind, it is interesting to note that when the UV reddening relations of MHC99, Witt Gordon (clumpy model), and Calzetti are applied to this technique, more than of the 2–8 keV flux originates from the half of the sample (12) with
bancs 4.5-6.0 keV. and 3.0-4.5 keV and the Lard colour (LIC) as the ratio of count rates in energy. bands 9.7-16.0. keV and 6.0-9.7 keV. Average Crab colours obtained are ος = 1.986+0.002 and LIC = 0.5992+ 0.0007.
bands 4.5-6.0 keV and 3.0-4.5 keV and the Hard colour (HC) as the ratio of count rates in energy bands 9.7-16.0 keV and 6.0-9.7 keV. Average Crab colours obtained are SC = $1.986 \pm 0.002 $ and HC = $0.5992 \pm 0.0007$ .
Figure 2. is the CD obtained by plotting normalised Aql X-1 HC versus SC.
Figure \ref{fig:cd} is the CD obtained by plotting normalised Aql X-1 HC versus SC.
All ποιακο. αι N-1 colours having error greater than of the observed. colour were ignored.
All normalised Aql X-1 colours having error greater than of the observed colour were ignored.
This amounted in rejecting 1H out of 307 points of the CD.
This amounted in rejecting 14 out of 307 points of the CD.
From Figure 2. we conclude that during the second. third and fourth observations. Aql N-1 was in the ELS (compareFigure2. of his paper with Figure 2 We will refer to these three observations as ELL. E12. ELS respectively.
From Figure \ref{fig:cd} we conclude that during the second, third and fourth observations, Aql X-1 was in the EIS \citep[compare Figure \ref{fig:cd} of this paper with Figure 2 We will refer to these three observations as EI1, EI2, EI3 respectively.
I is noted that E12 observation ias the best statistics in the ELS and also EI2 is brighter han ELI indicating irregular decline in the lishteurve (Table 1).
It is noted that EI2 observation has the best statistics in the EIS and also EI2 is brighter than EI1 indicating irregular decline in the lightcurve (Table \ref{tab:table1}) ).
There are no simultaneous PCA observations during the irst observation to ascertain the spectral state of he source.
There are no simultaneous PCA observations during the first observation to ascertain the spectral state of the source.
Hence we use a cdillerent method to find the normatisecl Aql Χο ancl HC during this observation.
Hence we use a different method to find the normalised Aql X-1 SC and HC during this observation.
The best fit spectral model that can reproduce the observed spectrum is used to simulate a PCA spectrum.
The best fit spectral model that can reproduce the observed spectrum is used to simulate a PCA spectrum.
The command of NSPISC package is used for this simulation.
The command of XSPEC package is used for this simulation.
Average count rates in the four energy. bands required. to calculate the SC and LIC are obtained by this simulated PCA spectrum.
Average count rates in the four energy bands required to calculate the SC and HC are obtained by this simulated PCA spectrum.
Aql N-1 8C and LIC thus obtained are normalised by the average Crab SC and LC respectively.
Aql X-1 SC and HC thus obtained are normalised by the average Crab SC and HC respectively.
In figure 2 the point marked by an open plus sign with error bars represents this normalised Aql N-1 5€ and HC.
In figure \ref{fig:cd} the point marked by an open plus sign with error bars represents this normalised Aql X-1 SC and HC.
Thus we conclude that during the first observation. Aql X-1 was in the DS.
Thus we conclude that during the first observation, Aql X-1 was in the BS.
Leneclorth this observation. will be referred. to as the DS observation.
Henceforth this observation will be referred to as the BS observation.
The PCA spectra. from. simultaneous observations (Table 1)) ave used for comparison with the spectra.
The PCA spectra from simultaneous observations (Table \ref{tab:table1}) ) are used for comparison with the spectra.
We have used the Std2 spectra and response files provided in the standard. products.
We have used the Std2 spectra and response files provided in the standard products.
Energy. channels were appropriately regrouped before the final spectral fitting.
Energy channels were appropriately regrouped before the final spectral fitting.
Lor cach front-illuminated NIS detector. (XISO and X183) we extracted the source spectra using a 260 arc-sec circular extraction region centered on the source.
For each front-illuminated XIS detector (XIS0 and XIS3) we extracted the source spectra using a 260 arc-sec circular extraction region centered on the source.
The energy. scale is reprocessed using the task.
The energy scale is reprocessed using the task.
Dackground spectra are extracted using appropriate circular regions outside the source region.
Background spectra are extracted using appropriate circular regions outside the source region.
C'orresponding response files are. generated using the and tools.
Corresponding response files are generated using the and tools.
The spectra from respective ALSO and. NIS3 detectors are then added together using the tool.
The spectra from respective XIS0 and XIS3 detectors are then added together using the tool.
During high count rates. as is the case for the BS observation. there is a possibility of pile-up in the AIS detectors.
During high count rates, as is the case for the BS observation, there is a possibility of pile-up in the XIS detectors.
To check [or pile-up we extracted the source spectrum. from an annular region with inner radius of 40 arc-sec and an outer radius of 260 arc-sec.
To check for pile-up we extracted the source spectrum from an annular region with inner radius of 40 arc-sec and an outer radius of 260 arc-sec.
Corresponding response files were also generated.
Corresponding response files were also generated.
Figure 3. shows the ratio of NISO background subtracted spectrum. extracted. [rom the full circular region. ancl from the annular region.
Figure \ref{fig:compare-xis0} shows the ratio of XIS0 background subtracted spectrum extracted from the full circular region and from the annular region.
The constant ratio indicates that both these spectra are similar and cilfer only in the observed. count. rates.
The constant ratio indicates that both these spectra are similar and differ only in the observed count rates.
Similar result is seen for NIS3 detector also.
Similar result is seen for XIS3 detector also.
This confirms that the Bs observation is not allected by. pile-up.
This confirms that the BS observation is not affected by pile-up.
The data processing version used for the LIND cleaned products is 2.1.6.15.
The data processing version used for the HXD cleaned products is 2.1.6.15.
Hence the PIN spectra are extracted using the cleaned event files Following the standard analysis threads on the The PIN non-X-rav background. is extracted. [rom the observation specificmodel provided by the instrument team.
Hence the PIN spectra are extracted using the cleaned event files following the standard analysis threads on the The PIN non-X-ray background is extracted from the observation specificmodel provided by the instrument team.
(Batveinetal.2009).
\citep{bat09} \citep{ver09}.
(Verasetal.2009).. 2009).. (saneetal.2009).. (Dollinger
\citep{bar07a} \citep{lau09,mou09}, \citep{kan08,kan09a}, \citep{kan09b}.