Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.mrao.cam.ac.uk/~bn204/publications/2010/N6946.pdf Äàòà èçìåíåíèÿ: Fri Aug 10 20:18:05 2012 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 10:34:22 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ï ï ð ð ð ð ð ð ï ï ï ï ï ï ï ï

Mon. Not. R. Astron. Soc. 000, 000­000 (2009)

Printed 29 April 2010

A (MN L TEX style file v2.2)

Microwave observations of spinning dust emission in NGC 6946
AMI Consortium: Anna M. M. Scaife1 , Bojan Nikolic2,3 , David A. Green3 , Rainer Beck4 , Matthew L. Davies3 , Thomas M. O. Franzen3 , Keith J. B. Grainge2,3 , Michael P. Hobson3 , Natasha Hurley-Walker3 , Anthony N. Lasenby2,3 , Malak Olamaie3 , Guy G. Pooley3 , ´ Carmen Rodr´ iguez-Gonzalvez3 , Richard D. E. Saunders2,3 , Paul F. Scott3 , 3 , David J. Titterington3 , Elizabeth M. Waldram3 & Jonathan T. L. Zwart5 Timothy W. Shimwell 1
2 3 4 5

Dublin Institute for Advanced Studies, 31 Fitzwilliam Place, Dublin 2, Ireland Kavli Institute for Cosmology Cambridge, Madingley Road, Cambridge, CB3 0HA Astrophysics Group, Cavendish Laboratory, J J Thomson Avenue, Cambridge CB3 0HE MPIfR, Auf dem Hugel 69, 53121 Bonn, Germany ¨ Columbia Astrophysics Laboratory, Columbia University, 550 West 120th Street, New York 10027, USA

Accepted --; received --; in original form 29 April 2010

ABSTRACT

We report new cm-wave measurements at five frequencies between 15 and 18 GHz of the continuum emission from the reportedly anomalous "region 4" of the nearby galaxy NGC 6946. We find that the emission in this frequency range is significantly in excess of that measured at 8.5 GHz, but has a spectrum from 15­18 GHz consistent with optically thin free­free emission from a compact H I I region. In combination with previously published data we fit four emission models containing different continuum components using the Bayesian spectrum analysis package radiospec. These fits show that, in combination with data at other frequencies, a model with a spinning dust component is slightly preferred to those that possess better-established emission mechanisms. Key words: radiation mechanisms: general -- galaxies: individual(NGC 6946)

1 INTRODUCTION The complete characterization of microwave emission from spinning dust grains is a key question in both astrophysics and cosmology as it probes a region of the electromagnetic spectrum where a number of different astrophysical disciplines overlap: it is important for CMB observations in order to correctly characterise the contaminating foreground emission (Gold et al. 2010); for star and planetary formation it is important because it potentially probes a regime of grain sizes that is not otherwise easily observable (Rafikov 2000). Although a number of objects have now been found to exhibit anomalous microwave emission, attributed to spinning dust, it is still unclear what differentiates those objects from the many other seemingly similar targets that do not show the excess. In order to investigate this question a number of Galactic observations have been made towards known star formation regions (see e.g. AMI Consortium: Scaife et al. 2010 and references therein; Casassus et al. 2008; Tibbs et al. 2009). In addition Murphy et al. (2010; hereinafter M10) made the first extra-galactic search for anomalous
We request that any reference to this paper cites "AMI Consortium: Scaife et al. 2010". email: ascaife@cp.dias.ie c 2009 RAS

microwave emission within the star formation regions of the nearby galaxy NGC 6946 using the Caltech Continuum Back-end on the Green Bank Telescope. M10 found a significantly anomalous spectrum in only one of 10 star-forming regions: extra-nuclear region 4 (hereinafter NGC 6946-E4). The excess of emission was seen between 27­38 GHz relative to the continuum emission at 8.5 GHz measured using combined Effelsberg 100 m Telescope and VLA observations (Beck 2007). In this letter we present follow-up observations of NGC 6946E4 at frequencies from 15-18 GHz using the Arcminute Microkelvin Imager (AMI) Large Array (LA). In Section 2 we present the details of these observations, in Section 3 we present the results of the AMI-LA observations and a comparison with other radio data, and in Section 4 we discuss the implications of these results and form our conclusions. In what follows we use the convention: S , where S is flux density, is frequency and is the spectral index. All errors are quoted to 1 .

2

OBSERVATIONS

The AMI-LA consists of eight 13 m antennas and is sited at Lord's Bridge, Cambridge (AMI Consortium: Zwart et al. 2008). The telescope observes in the band 13.9­18.2 GHz with cryogenically


2

Scaife et al.
those of the AMI-LA 15.0 GHz data (channel 4) in order to compare consistent angular scales. The resulting uv data set was then mapped and cleaned in the same way as the AMI-LA data. Since channel 4 of the AMI-LA recovers a large amount of the extended structure around NGC 6946-E4 the flux density for this source from the sampled 8.5 GHz map is very similar to the unsampled value, see Table 1. Flux densities were extracted from both the AMI-LA and sampled Effelsberg 8.5 GHz maps using the FI T FL U X software (Green 2007; AMI Consortium: Scaife et al. 2009). This method calculates flux densities by removing a tilted plane fitted to the local background and integrating the remaining flux. We do this by drawing a polygon around the source and fitting a tilted plane to the pixels around the edge of the polygon. Where an edge of the polygon crosses a region confused by another source the background is subjective and we omit this edge from the fitting. An example of where this might be appropriate is shown in Fig. 1 (b). Since the extracted flux density is dependent to some degree on the background emission we repeat this process using five irregular polygons, varying each slightly in shape. The final flux density is the average of that extracted from these five polygons. Errors on the flux densities were calculated as =
2 2 (0.05S )2 + rms, + fit , where rms, is the rms noise outside the primary beam on each channel map and fit is the standard deviation of the fluxes measured in the five polygonal apertures. The errors are dominated by fit , which is large due to the complicated background emission in this crowded field. Since the uv coverage of the AMI-LA varies across the frequency channels we quantified the amount of flux lost in each channel relative to channel 4 by sampling the total power 8.5 GHz map to match the uv coverage in each of the AMI-LA frequency channels. This showed that flux loss was negligible in channels 4­6 and notable ( 5%) in channels 7 & 8 only. We corrected for this loss in the measured AMI-LA flux densities with corrected values shown in parentheses in Table 1. In the same manner we checked for corrections to the flux densities of NGC 6946-E8. In this case the corrections were negligible (< 2%) and no corrections were made. We found the 15­18 GHz data for NGC 6946-E8 to be consistent with the spectrum for this object presented in M10. The flux density of NGC 6946-E4 across the AMI-LA band is in excess of the 8.5 GHz flux density by approximately 10 . The collected observations of both region E4 and E8 were analysed using an updated version of the radiospec package1 (Nikolic 2009). This software tool calculates the Bayesian posterior distribution of the parameters of a model for the radio spectrum and, furthermore, the Bayesian evidence. The implementation of these calculations is based on the nested sampling algorithm by Skilling (2004). For the analysis of the data in this paper, we used several different models which each consist of a number of components, each with physically parameterized properties. Two components that are present in all models are a synchrotron component, which we parametrise in terms of the supernova rate within the beam, and an un-absorbed free­free component, which we parametrise in terms of the star-formation rate within the beam. The conversions from these physical parameters to radio luminosities are made according to the formulae given by Condon (1992). In order to explain the excess of emission at cm wavelengths, it is necessary to introduce another component that contributes to the

cooled NRAO indium-phosphide front-end amplifiers. The overall system temperature is approximately 25 K. Amplification, equalization, path compensation and automatic gain control are applied to the IF signal. The back-end has an analogue lag correlator with 16 independent correlations formed at double the Nyquist rate for each baseline using path delays spaced by 26 mm. From these, eight complex visibilities are formed in 0.75 GHz bandwidth channels. In what follows, the three lowest frequency channels (1, 2 & 3; 13.9­14.6 GHz) are not used due to interference from geostationary satellites. s NGC 6946-E4 (J 20h 34m 19. 17 + 60 10 08. 7) was observed by the AMI-LA in a single 12 hour synthesis. Data reduction was performed using the local software tool R E D U C E. This applies both automatic and manual flags for interference, shadowing and hardware errors, phase and amplitude calibrations and then Fourier transforms the correlator data to synthesize frequency channels before output to disk in uv FITS format suitable for imaging in A I P S. Flux calibration was performed using short observations of 3C48 at the beginning of the run and 3C286 at the end of the run. We assumed I+Q flux densities for these sources in the AMI-LA channels consistent with the frequency dependent model of Baars et al. (1977), 1.64 and 3.48 Jy, respectively, at 15 GHz. As Baars et al. measure I and AMI-LA measures I+Q, these flux densities include corrections for the polarisation of the calibrator sources derived by interpolating from VLA 5, 8 and 22 GHz observations. A correction is also made for the changing intervening airmass over the observation. From cross-calibration of 3C48 and 3C286, we find the flux calibration is accurate to better than 5 per cent. The phase was calibrated using interleaved observations of J 2031+5455, selected from the Jodrell Bank VLA Survey (JVAS; Patnaik et al. 1992). After calibration, the phase is generally stable to 5 for channels 4­7, and 10 for channel 8. The FWHM of the primary beam of the AMI-LA is 6 at 16 GHz. Reduced data were imaged using the A I P S data package. C L E A N deconvolution was performed using the task I M AG R . C L E A N deconvolution maps were made from both the combined channel set and for individual channels.

3 RESULTS 3.1 AMI Large Array Data The combined channel map from the AMI-LA observation of NGC 6946-E4 is shown in Fig. 1. E4 is located at the phase s centre with extra-nuclear region 8 (J 20h 34m 32. 52 + 60 10 22. 0; NGC 6946-E8) to the east and the galactic nucleus of NGC 6946 s (J 20h 34m 52. 34 + 60 09 14. 2) further to the east, outside the primary beam FWHM. Flux densities were extracted for both NGC 6946-E4 and NGC 6946-E8. Since NGC 6946-E8 also lies within the FWHM of the primary beam we used the recovered flux densities for this source as a check on the absolute calibration for this field. The flux densities measured for NGC 6946-E4 and NGC 6946-E8 are listed in Table 1.

3.2 NCC6946-E4: Comparison with other radio data We compared the AMI-LA data with that at 8.5 GHz from the Effelsberg 100 m and VLA telescopes (Beck 2007). These data in their original form, see Fig. 1(b), constitutes a total power measurement of the region at a resolution of 15 . We Fourier transformed these combined data and sampled them at uv positions identical to

The complete code and data used for this spectrum analysis are available for public download under GPL license: http://www.mrao.cam.ac.uk/ ~bn204/galevol/speca/sdgals.html. c 2009 RAS, MNRAS 000, 000­000

1


Spinning dust emission in NGC 6946

3

(a)

(b)

Figure 1. NGC 6946-E4. (a) Greyscale and contours at -3, 3, 6, 12 etc are from the AMI-LA where rms = 25 µ Jy beam-1 , greyscale units are mJy beam-1 . Data are not corrected for the primary beam response. The AMI-LA synthesized beam, 29.2 â 25.2 , is shown as a filled ellipse in the bottom left corner. (b) Contours are from the AMI-LA and greyscale at 8.5 GHz is from Effelsberg-VLA measurements, greyscale units are µ Jy beam-1 (Beck 2007). The positions of NGC 6946-E4 (at the phase centre) and NGC 6946-E8 (to the east) are marked with crosses, and a line is drawn to indicate a non-fitted polygon edge (see text for details). The AMI-LA primary beam is shown as a black circle in both images.

Table 1. Flux densities. AMI-LA Channel Number 6 7 (mJy) (mJy) [5] [6] 3.00 ± 0.28 1.40 ± 0.23 3.15 ± 0.43 (3.31 ± 0.43) 1.51 ± 0.34

Source

8.5 GHz (mJy) [2] 1.94 ± 0.29 2.11 ± 0.21

4 (mJy) [3] 3.17 ± 0.36 1.90 ± 0.27

5 (mJy) [4]

8 (mJy) [7] 2.78 ± 0.34 (3.09 ± 0.35) 1.42 ± 0.19

NGC 6946-E4 NGC 6946-E8

3.16 ± 0.33 1.51 ± 0.27

Column [1] contains the source name, [2] the uv-sampled 8.5 GHz flux density for that source, [3]-[7] contain the AMI-LA flux densities for each source from AMI-LA channels 4­8, with flux loss corrected values in parentheses..

emission. As outlined above we have considered two options: emission due to spinning dust, and absorbed free­free emission from a compact H I I region. For the spinning dust emission component we used the warm ionized medium model described by Draine & Lazarian (1998)2 . The only degree of freedom in this model is the overall amplitude, which we parametrised in terms of the total mass of gas carrying the spinning dust. Our model for the absorbed free­free emission from H I I is again parametrised in terms of the star-formation rate in the region, but it now also has the filling factor, i.e. the fraction of the area covered by the telescope beam that the H I I region subtends, as a free parameter. This area is used to compute the electron opacity as a function of frequency and to correct the unabsorbed free­free model for the effects of absorption.

error on this estimate is large and we cannot rule out other mechanisms. Since the spectral index between the Effelsberg-VLA mea16 surement at 8.5 GHz and the AMI band is rising (8.5 = 0.67 ± 0.08, see Fig. 2), we need to consider the possibility that region E4 contains one or more compact H I I regions with their opacity reaching unity at approximately 12 GHz. Such an opacity would require an emission measure of 5 â 108 pc cm-6 , assuming Te = 104 K, and would be appropriate for a compact H I I region. We therefore examine two alternative hypotheses for the emission from region E4. The first is that the emission is due to the usual diffuse synchrotron and free­free mechanisms associated with starformation, with an additional high-opacity free­free component (Hypothesis 1; H1). The second hypothesis is that there is a spinning dust component rather than high-opacity free­free (H2). A summary of how well these two hypotheses fit the observed data is shown in the form of fan-diagrams in Fig. 2. As can be seen from this figure, neither of the hypotheses can be ruled out, although the spinning-dust appears to somewhat better match the data. This is also confirmed by a simple comparison of the models: assuming flat priors and no a priori difference between the models, the logarithmic Bayes factor is 3.7 ± 0.3 in favour of the spinning dust model. From the Jeffreys' scale of evidence (Jeffreys 1961; Kass & Raftery 1995; see e.g. Efron & Gous 2001 for further discussion of this scale) this would indicate a weak positive preference for

4 DISCUSSION AND CONCLUSIONS Considered on their own, the AMI-LA data (after correction for flux loss) have a spectral index of AMI = -0.11 ± 0.77. Although this value is consistent with optically thin free­free emission, the

2

ftp://ftp.astro.princeton.edu/draine/dust/ spin/emit4.jnu.wim a c 2009 RAS, MNRAS 000, 000­000


4

Scaife et al.
Synchrotron + free­free + absorbed free­free
0.01 3 · 103 2.5 · 103 0.005 2 · 103 F (Jy) 1.5 · 103 0.002 1 · 103 5 · 103 0.001 1 2 5 10 (GHz) 20 50 100
5 5

Synchrotron + free­free + spinning dust
0.01

8 · 103
5

6

0.005 6 · 103 F (Jy)
6

5

4 · 103 0.002 2 · 103

6

5

6

4

0

0.001 1 2 5 10 (GHz) 20 50 100

0

Figure 2. The observed radio spectrum of region E4 of NGC 6946 (points and error bars) with the fan-diagram of two model fits to these data: on the left is the model with a highly absorbed free­free emission region (H1) and on the right is the model with spinning dust emission (H2). Low frequency data are taken from M10, scaled to the uv-sampled flux density at 8.5 GHz, with the exception of points between 15 and 18 GHz which are from the AMI-LA (this work). The assumed error is 10% unless stated otherwise in Table 1. The colour scale indicates the evidence contribution as a function of frequency and flux density, for details see Nikolic 2009. Synchrotron + free­free
0.01 4 · 103 0.005
9

Synchrotron + free­free + spinning dust
0.01 2.5 · 103
9

3 · 103

9

0.005

2 · 103

9

F (Jy)

F (Jy)

1.5 · 103

9

2 · 10 0.002 1 · 10

39

0.002
39

1 · 103

9

5 · 103

8

0.001 1 2 5 10 (GHz) 20 50 100

0

0.001 1 2 5 10 (GHz) 20 50 100

0

Figure 3. The observed radio spectrum and fan-diagrams for region E8 of NGC 6946, with data and errors as in Fig. 2. The model on the left only consists of a synchrotron and un-absorbed free­free components (H1) while the model on the right also has a spinning dust component (H2). Low frequency data are taken from M10, scaled to the uv-sampled flux density at 8.5 GHz, with the exception of points between 15 and 18 GHz which are from the AMI-LA (this work). The assumed error is 10% unless stated otherwise in Table 1. Colour scale as above. Region E4
0.4 0.3
0.04 0.06

Region E8

0.2 0.1 0 5.5 6 6.5 7 7.5 8 8.5 log10 (MH /M )

f

f 0.02 0 5.5 6 6.5 7 7.5 8 8.5 log10 (MH /M )

Figure 4. Marginalised distribution of inferred gas mass that bears the spinning dust: the left panel shows extra-nuclear region 4 and the right panel extranuclear region 8.

c 2009 RAS, MNRAS 000, 000­000


Spinning dust emission in NGC 6946
Table 2. Derived model parameters and errors Model SNe rate log10 (yr-1 ) [2] (H1) (H2) (H1) (H2) (-5, -1) -3.84 ± 0.11 -3.98 ± 0.16 -3.65 ± 0.04 -3.64 ± 0.04 sync [3] (-1.0, -0.5) -0.74 ± 0.14 -0.71 ± 0.15 -0.66 ± 0.13 -0.65 ± 0.12 SF Rabs log10 (yr-1 ) [4] (-3, -1) -0.93 ± 0.07 ­ ­ ­ SF Runabs log10 (yr-1 ) [5] (-3, 0) -1.80 ± 0.48 -1.40 ± 0.36 -1.70 ± 0.41 -1.75 ± 0.41 Mgas log10 (M ) [6] (5, 9) ­ 8.12 ± 0.09 ­ 6.19 ± 0.66 f log10 (str) [7] (-6, -3) -4.97 ± 0.18 ­ ­ ­

5

Prior NGC NGC NGC NGC

6946-E4 6946-E4 6946-E8 6946-E8

the spinning dust model above the free­free model. The maximum likelihood parameters for these models are listed in Table 2. For comparison we have also carried out a similar analysis on the region E8, shown in Fig 3. In this case the two hypotheses are a simple diffuse synchrotron plus free­free model (H1) versus the same model with an additional spinning dust component (H2). In this case the logarithm of the Bayesian evidence ratio is 0.5 ± 0.3 in favour of the simpler model without the spinning dust. A ratio of this size indicates no perceptible difference between the two models. In region E4 where a spinning dust model is the preferred hypothesis, we can marginalise the posterior distribution of the model parameters to obtain an estimate of the gas mass containing the spinning dust, shown as a histogram in the left panel of Fig. 4 and in numerical form in Table 2, i.e., 108.1±0.1 M . In region E8 spinning dust is not the preferred hypothesis but proceeding with this hypothesis anyway, we find an upper limit for the mass of the gas bearing the spinning dust, which is around 107.5 M . We draw a conclusion that if the conditions in E4 and E8 are similar, then the mass of any gas bearing spinning dust in E8 must be at least a factor of five smaller than in E4. Region E4 is located on the dense rim of a "remarkable" H I hole (Boomsma et al. 2008) within NGC 6946. Such an association may be relevent to the differentiation of this star formation region from the eight others found to exhibit no anomalous emission by M10. The hole is remarkable for a number of reasons, notably the almost unbroken symmetry of its dense H I rim, unusual in so large a hole, and the small scale high velocity gas complexes observed in connection with it. As mentioned above, the spinning dust model is preferred by the evidence calculation for NGC 6946-E4, but not at a very high level. Definitive confirmation of the nature of the emission requires measurements at frequencies above 50 GHz where the spinning dust and compact H I I region models have significantly different behaviour. For example, Fig. 2 shows that at 100 GHz the difference between these two models should be at least a factor of two in brightness. In the sub-mm there are data available from SCUBA at 850 µ m for this region by Di Francesco et al. 2008, which might be used to constrain the mass of NGC 6946-E4 and place constraints on the frequency at which the optical depth reaches unity. If a compact H I I region is present, with = 1 at > 8.5 GHz it should have a correspondingly high dust mass. From analysis of the SCUBA data we obtained a flux estimate for region E4 of S850 = 11 ± 14 mJy beam-1 . However, the errors on this estimate are too high to allow any useful constraint on the properties of the thermal dust emission or to calculate a reliable dust mass estimate. From the existing data the possibility of a spinning dust component in this region cannot be ruled out, but the evidence is not yet
c 2009 RAS, MNRAS 000, 000­000

definitive. Further observations of this object at frequencies covering the higher frequency minimum between spinning dust emission and thermal dust emission ( 90 GHz) would be most useful.

5

ACKNOWLEDGEMENTS

We thank the staff of the Lord's Bridge Observatory for their invaluable assistance in the commissioning and operation of the Arcminute Microkelvin Imager. The AMI-LA is supported by Cambridge University and the STFC. AS thanks Thijs van der Hulst for drawing attention to the H I complexes in NGC 6946. CRG, TS, TF, MO and MLD acknowledge the support of PPARC/STFC studentships.

REFERENCES AMI Consortium: Scaife A. M. M., et al., 2010, MNRAS, L16 AMI Consortium: Scaife A. M. M., et al., 2009, MNRAS, 400, 1394 AMI Consortium: Zwart, J. T. L., et al. 2008, MNRAS, 391, 1545 Beck, R. 2007, A&A, 470, 539 Boomsma R., Oosterloo T. A., Fraternali F., van der Hulst J. M., Sancisi R., 2008, A&A, 490, 555 Casassus S., et al., 2008, MNRAS, 391, 1075 Condon J. J., 1992, ARA&A, 30, 575 Di Francesco, J., Johnstone, D., Kirk, H., MacKenzie, T., & Ledwosinska, E. 2008, ApJS, 175, 277 Draine, B. T., & Lazarian, A. 1998, ApJ, 508, 157 Efron, B., & Gous, A., 2001, IMS Lecture Notes - Monograph Series, vol. 38 Gold, B., et al. 2010, arXiv:1001.4555 Green, D. A. 2007, Bulletin of the Astronomical Society of India, 35, 77 Jeffreys, H., 1961, OUP, Theory of Probability, 3rd ed. Kass, R. E., & Raftery, A. E., 1995, J. Amer. Statist. Asso. 90, 773 Murphy, E. J., et al. 2010, ApJ, 709, L108 Nikolic, B. 2009, arXiv:0912.2317 Patnaik, A. R., Browne, I. W. A., Wilkinson, P. N., & Wrobel, J. M. 1992, MNRAS, 254, 655 Rafikov, R. R. 2006, ApJ, 646, 288 Skilling, J., 2004, Nested Sampling, Bayesian inference and maximum entropy methods in science and engineering, eds. R. Fischer, R. Preuss, and U. von Toussaint, No. 735 Tibbs, C. T., et al. 2010, MNRAS, 402, 1969 Visser, A. E., Richer, J. S., & Chandler, C. J. 2001, MNRAS, 323, 257