Datasets:

SouthernCrossAI
/

Arxiv

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 132k rows

Filename stringlengths 22 64	Paragraph stringlengths 8 5.57k
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	No other sources exhibit clustering as tight in the rest of the G08 sample; the next best examples are 4U 1323−62 and 4U 0836−429. 4U 1323−62 is quite well-known for it’s regular burst behaviour (e.g. G08), and in the RXTE sample, many of the bursts were long with τ = 28 ± 2 s and similar ﬂuences, of Eb = (100±10)×10−9 erg cm−2. However, several of the bursts were closely followed by weaker events only a few min- utes after the ﬁrst. This behaviour has been observed in a number of systems (e.g. Keek et al. 2010), and is thought to arise from delayed ignition of material left over from the initial burst. Such events are likely not suitable for a detailed test of the consistency of blackbody normal- isations between bursts. During a period of activity in 2003–4, 4U 0836−429 exhibited long (τ = 22±4 s) bursts at roughly similar ﬂuences, distributed approximately as a Gaussian with Eb = (270 ± 70) × 10−9 erg cm−2. De- spite good coverage of the outburst, the bursts were ap- parently not strictly periodic in nature, and there may also be contributions to the source ﬂux from a nearby HMXB pulsar which is within the PCA ﬁeld of view. Thus, we also exclude this source from our study.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	Where not explicitly stated, the data analysis proce- dures are as in G08. Time-resolved spectra in the range 2–60 keV covering the burst duration were extracted on intervals as short as 0.25 s during the burst rise and peak, with the bin size increasing gradually into the burst tail to maintain roughly the same signal-to-noise level. A spectrum taken from a 16-s interval prior to the burst was adopted as the background. We re-ﬁt the spectra over the energy range 2.5–20 keV using the revised PCA response matrices, v11.74 and adopted the recommended systematic error of 0.5%. The ﬁtting was undertaken us- ing XSpec version 12. In order to accommodate spectral bins with low count rates, we adopted Churazov weight- ing. No correction for instrumental deadtime was applied to the spectra.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	We modelled the eﬀects of interstellar absorption, us- ing a multiplicative model component (wabs in XSpec), with the column density nH frozen at 4 × 1021 cm−2 (for GS 1826−24, e.g. in ’t Zand et al. 1999) and 1.3 × 1022 cm−2 (for KS 1731−26, e.g. Cackett et al. 2006). In the original analysis carried out by G08, the neutral ab- sorption was determined separately for each burst, from the mean value obtained for spectral ﬁts carried out with the nH value free to vary. This has a negligible eﬀect on the ﬂuxes, but can introduce spurious burst-to-burst variations in the blackbody normalisation. We also com- puted averaged lightcurves of blackbody spectral param- eters for subsets of bursts from GS 1826−24, following the procedure adopted by Galloway et al. (2004).
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	over which a constant ﬁt to the blackbody normalisations was an acceptable ﬁt to below 3σ conﬁdence. We refer throughout to the best-ﬁt constant value as hKbbi, to distinguish from the individual Kbb values for each time- resolved spectrum.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	We found a best-ﬁt value of hKbbi = 103.3 ± 0.7 (km/d10 kpc)2 over the interval 3.0–65.25 s (relative to the start over the burst; Fig. 1). The reduced χ2/nDOF ≡ χ2 ν was 1.45, for 95 degrees of freedom. Re- markably, this consistency was maintained despite the ﬁtted blackbody temperature kTbb falling from 2.32 to 1.55 keV.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	In order to double-check our result, we carried out a similar procedure to determine the maximum extent over which the blackbody normalisation was consistent with a constant value, but instead of simply averaging the ﬁtted normalisations for each time bin, we performed a joint ﬁt in XSpec of the time-resolved spectra to an absorbed blackbody model, as used for the individual spectra. We froze the neutral column density at the same value as used for the individual spectral ﬁts, 4 × 1021 cm−2. We linked the blackbody normalisation for each of the indi- vidual spectra but allowed the blackbody temperatures to vary. With this treatment, the ﬁtted Kbb,joint was found over a shorter time interval, between 11.5–32.25 s following the burst start, and at a 3% higher value of 107.0 ± 1.1 (Table 1). Likely, the higher value is attained by omitting some of the spectra in the range 40–60 s, when there is a noticeable downward trend in the nor- malisation (Fig. 1). We note that the mean of the nor- malisation values over the reduced extent of the burst permitted by the joint ﬁts, returns a consistent best-ﬁt value, of 106.6 ± 1.1.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	Figure 1. Example burst observed by RXTE from GS 1826−24 on 2000 July 1 17:16:37 UT (#12 in G08), illustrating the con- stancy of the blackbody normalisation over a signiﬁcant extent of the burst tail. Shown is the burst ﬂux (top panel), blackbody (colour) temperature kTbb (middle panel), and blackbody normal- isation (in units of (km/d10kpc)2; bottom panel). The extent over which the constant-normalisation ﬁt is calculated is illustrated by the vertical dotted lines (top and middle panels), and the horizon- tal lines (lower panel), with the 1σ error range indicated (dashed lines). The reduced-χ2 = χ2 ν for the constant ﬁt is 1.45, for ν = 95 degrees of freedom.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	the average of the individual ﬁts hKbbi for the bursts from this sample to determine how the results from our sensitivity tests translated to a larger sample. Although in some cases the joint ﬁt durations were shorter than the constant ﬁt duration for the individual normalisa- tions the median duration was 94% of the average in- terval. Furthermore, we found no systematic diﬀerence between the normalisations determined by the two meth- ods, and the agreement was at better than 2% for 76% of the measurements (maximum deviation was 8.5%). This compares favourably to the typical 1σ statistical uncer- tainty on the measurements, of 1%. Thus, we conclude that the parameter averages hkbbi provide an unbiased measure of the blackbody normalisation in the tail of these bursts, and adopt that method for further analy- sis.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	Figure 2. Distribution of ﬁt statistic χ2 for blackbody ﬁts to time- resolved spectra from burst #12 from GS 1826−24, on 2000 July 1 17:16:37, over the time for which the blackbody normalisation was determined to be constant (3.0–65.25 s relative to the burst start). The expected distribution for statistically acceptable ﬁts with 22 degrees of freedom is overplotted (dotted line). The ﬁt χ2 show an excess of high values, as also indicated by the cumulative probability distributions (inset); a K-S test returns a value of 0.21, indicating disagreement at the 3.4σ level.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	after ≈ 60 s). Since the bursts reach approximately the same peak ﬂux, the deadtime correction is also approx- imately the same, and can be neglected for comparison purposes (we similarly neglect any issues related to the absolute PCA calibration). However, since the deadtime correction varies with time, it might also be expected to result in a slightly diﬀerent evolution of the normalisation throughout the burst. To test this possibility, we exam- ined the duration of the consant interval for the burst analysed in this section (#12 from G08), with and with- out the deatime correction. Although we measured an increase in the average normalisation with the deadtime correction, as expected, we found no change in the extent over which the normalisation was found to be constant. Thus, we neglect deadtime corrections for the remainder of the analysis in this paper.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	The hKbbi for the bursts from GS 1826−24 determined from the individual spectral ﬁt parameters (using the spectra with variable binsizes; see §3) is shown in Fig. 3. The typical span of the constant ﬁt was from < 3 s after the burst start (52% of the bursts) to approximately 60 s. For one burst the normalisation was found to be con- stant from 2–77 s after the start of the burst; the median duration was 48 s. Over the interval during which the blackbody normalisation was constant, the blackbody temperature typically decreased from 2.3 to 1.6 keV, while the ﬂux decreased from ≈ 3 × 10−8 erg cm−2 s−1 to ≈ 5 × 10−9 erg cm−2 s−1.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	The hKbbi values were not signiﬁcantly correlated with either the duration over which the average was calcu- lated, nor the maximum time out to which the average was calculated. Interestingly, the earliest time to which the constant ﬁts could be extended (≈ 1–3 s after the burst start) was within the ≈ 5 s burst rise. For several of the bursts, the time at which the normalisation reaches the mean value corresponds approximately to a change in slope of the burst rise, seen in several of the bursts (e.g. Fig. 1). This feature is similar to the “bump” seen in model lightcurves (e.g. Heger et al. 2007). Al- though in the simulations this feature must arise due to some variation in the rate of increase of thermonuclear energy production (as the 1-D models cannot account for lateral propagation), the coincidence of the observed change in slope with the normalisation achieving its mean value suggests that this point marks where the eﬀects of spreading cease. Further analysis of the detailed shape of the burst rises may help to clarify this situation.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	Figure 3. Mean blackbody normalisation values for 67 thermonu- clear bursts observed by RXTE from GS 1826−24. The mean and standard deviation for the two observation epochs (1997–8 and 2000–7) are indicated. Note the marked discrepancy between the mean values for the two distributions; the K-S statistic indicate that they are discrepant at a signiﬁcance of 1.3 × 10−6, equivalent to 4.7σ.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	Figure 4. Comparison of averaged blackbody normalisation pro- ﬁles for bursts from GS 1826−24 measured in 1997–8 (#1–5 of G08) and 2000–7 (#9, 10, 11, 12, 13, 16, 17, 19, 20). The verti- cal dashed lines indicate the time of maximum ﬂux for each set of bursts. Note the agreement in the normalisation throughout the burst rise and maximum, and the increasing discrepancy from 10 s after the burst start. The inset shows the corresponding variation of the averaged burst ﬂux.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	sided Kolmogorov-Smirnov test conﬁrms that these dis- tributions are discrepant at the 1.3 × 10−6 signiﬁcance level (equivalent to 4.7σ). This variation was also noted by Galloway et al. (2004), who reported instead a cor- relation between the persistent ﬂux and the blackbody normalisation for a subset of the bursts analysed here.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	The observed variation is unlikely to arise from any instrumental eﬀect, as the PCA is precisely calibrated to maintain stable ﬂux measurements for calibration sources over the entire mission lifetime. As we discuss below, the discrepancy is also unlikely to result from vari- ations in the neutral column density nH as a function of epoch. Closer examination of the spectral variation in the two groups of bursts provides a possible explanation. In contrast with the example burst discussed in the pre- vious section, and other bursts observed in 2000–7, the constant ﬁt interval for the bursts observed in 1997–8 be- gan later than in the 2000–7 bursts; typically 10 s after the burst start for the 1997–8 bursts, compared to 3 s after the burst start for the other bursts. To put this another way, there was additional variation in the black- body normalisation during the burst rise for the 1997–8 bursts that prevented extension of the constant interval to the same point as in the later bursts. This is illus- trated in Fig. 4, which compares the blackbody normal- isation averaged over the 1997–8 bursts with that of the 2000–7 bursts. Remarkably, however, the behaviour of the normalisation in the two groups of bursts is virtually identical during the burst rise; the discrepancy sets in from 10 s after the burst start, with the normalisation of the 1997–8 bursts gradually increasing over another ≈ 10 s to a higher level.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	Figure 5. Example burst observed by RXTE from KS 1731−26 on 2000 September 29 14:08:35 UT (#24 in G08), illustrating the constancy of the blackbody normalisation over a signiﬁcant extent of the burst rise and tail. Panel descriptions are as for Fig. 1. The reduced-χ2 = χ2 ν for the constant ﬁt is 1.39, for ν = 79 degrees of freedom.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	signiﬁcantly shorter (τ = 24 s compared with ≈ 40 s for GS 1826−24). A shorter burst duration for KS 1731−26 suggests a smaller fraction of hydrogen in the burst fuel, although the shorter recurrence time also allows less hy- drogen to be consumed by steady burning prior to the burst.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	As with GS 1826−24, we measured the best-ﬁt normal- isation hKbbi from the time-resolved blackbody spectral ﬁts over the longest possible time interval without ex- ceeding the 3σ conﬁdence limit. With the shorter bursts from KS 1731−26, the constant Kbb ﬁts extended typ- ically over the range 2–30 s after the burst start. Over this time interval, the blackbody temperature dropped (typically) from 2.5 to 1.8 keV, with the ﬂux dropping from 3 × 10−8 erg cm−2 s−1 to 7 × 10−9 erg cm−2 s−1.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	The ﬁtted hKbbi values of the long bursts from KS 1731−26 exhibited signiﬁcant variability; a ﬁt with a constant model gave a χ2 = 112.7 for 14 degrees of freedom, indicating variability at the 8.5σ level. The mean value was 80 ± 2 (km/d10 kpc)2, with standard de- viation between the various measurements of 2.6%. The mean blackbody normalisation of the burst on 1999 Aug 26, measured between 1.25–36 s after the burst start at 79.7 ± 0.8 (km/d10 kpc)2, was fully consistent with the bursts observed in 2000 despite being observed a year earlier.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	Figure 6. Sensitivity of the blackbody normalisation in spectral ﬁts of RXTE data to the assumed column density nH . The ﬁts were calculated from a simulated spectrum with kT = 2.1 keV, including pre-burst persistent and instrumental background from burst #15 of KS 1731−26. The square symbols show the ﬁtted value of the blackbody normalisation (left-hand y-axis) as a func- tion of the assumed column density nH . The lower half of the plot shows the corresponding χ2 values; each ﬁt is statistically accept- able. The choice of the nH value can clearly introduce a systematic bias to the normalisation values, of the order a few per cent.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	we suggest that the diﬀerence arises from a systematic eﬀect due to the value adopted for the neutral column density. For the analysis presented here, we assumed nH = 1.3 × 1022 cm−2 (from multi-epoch Chandra and XMM-Newton spectra of the source during quiescence; Cackett et al. 2006) while G¨uver et al. (2011) adopted a value of 2.98 × 1022 cm−2, the mean of best-ﬁt values from ﬁts to the burst spectra measured by RXTE. For a simulated blackbody spectrum with known kT and nor- malisation, adopting a realistic model for the pre-burst emission as background, the ﬁtted value of the normal- isation depends linearly on the assumed nH (Fig. 6). That is, over (under) estimating the nH value assumed for the ﬁts will have the eﬀect of over (under) estimat- ing the Kbb. For the representative parameters chosen, the slope gives approximately 6.5 (km/d10kpc)2 for every additonal 1022 cm−2 in column that is adopted. The dif- ference between the adopted values for the two analyses could account for an oﬀset in the normalisations of up to 11 (km/d10kpc)2, suﬃcient to explain the discrepancy.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	The 1997–8 bursts from GS 1826−24, which exhib- ited longer recurrence times and reached slightly higher ﬂuxes, also exhibited larger mean normalisations hKbbi. Comparison of the normalisation time-series averaged over subsamples of bursts from the two epochs show that the normalisations were essentially identical during the burst rise and through the burst peak, with deviations becoming apparent between 10–20 s after the burst start. The reproducibility of the Kbb evolution throughout the burst rise and peak suggests that the system geometry, atmospheric composition and temperature (and hence fc) were essentially identical over that interval, and what- ever physical condition gave rise to the elevated Kbb in the 1997–8 bursts set in after the burst peak. We con- sider three possible mechanisms to give rise to the vari- ation.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	after the burst start. Such variations might be expected to lead to incorrect pre-burst emission subtraction later in the burst, although this is not observed consistently. Third, we consider the possibility that the fc value changes to produce the variation in the measured Kbb after the peak in the 1997–8 bursts from GS 1826−24. Modelling studies indicate that fc depends on ﬁxed parameters such as the neutron star surface gravity, but also the composition and eﬀective temperature of the scattering atmosphere (e.g. Madej et al. 2004; Suleimanov et al. 2011b). There is evidence in other sources that radius-expansion bursts can remove the outer, H-rich layers of the photosphere, leading to a change in the atmospheric composition during the burst (Galloway et al. 2006); no such eﬀects have been sug- gested from non-radius expansion bursts. Nevertheless, interpreting the maximum variation in hKbbi in the 1997–8 bursts, compared to the mean for the 2000–7 bursts, as a variation in fc implies a maximum vari- ation of 12% (8% in the mean). This result suggests that the assumption that fc is constant during bursts is not always true. While Suleimanov et al. (2011b) predict patterns of variation of fc with burst ﬂux, the epoch-to- epoch diﬀerences in the burst lightcurves in GS 1826−24 are relatively subtle, and do not appear suﬃcient to give rise to the inferred variation in fc.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	it may be that the amount of neutral mate- rial close to the neutron star was reduced, so that the neutral column density nH decreased during the 1997–8 bursts, leading to an overestimation of Kbb (see e.g. Fig. 6). Simulations adopting the pre-burst emission from GS 1826−24 (as was done for KS 1731−26; see §3.2) in- dicate that in order to overestimate the hKbbi by the re- quired amount would necessitate decreasing the total nH value by 1.9 × 1022 cm−2. Since this value is almost ﬁve times the assumed line-of-sight value for GS 1826−24, we can rule out this mechanism.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	It is possible that the peak blackbody ﬂux (or tempera- ture) serves as the discriminant which results in elevated blackbody normalisations later in the burst. The 1997– 8 bursts reached maximum ﬂuxes about 8% higher on average than for the 2000–7 bursts. The discrepancy in the maximum temperature reached was proportionately smaller. The samples of bursts studied here were deliber- ately selected for the consistency of their lightcurves and regularity of their recurrence times. For other samples of bursts, which are typically much more heterogeneous, systematic errors in measurements of the blackbody ra- dius of order & 8% should be assumed, unless the varia- tion in fc can be modelled.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	The mean blackbody normalisations measured during the later (2000–7) bursts from GS 1826−24, and the reg- ular bursts from KS 1731−26, were consistent with a con- stant value over several tens of seconds. This constancy was observed independent of the speciﬁc method for de- termining the mean value, although the choice of method can introduce small biases. In particular, comparisons of joint ﬁts to the spectra with averages of the ﬁtted nor- malisations show that in general the former method ar- rives at shorter durations for the constant normalisation, likely because the spectra are not (en masse) statistically consistent with blackbodies. Trends in the blackbody normalisation late in the burst tail, coupled with these marginally deviant spectra, likely give rise to systematic errors of a few percent between the two methods. The choice of time binning strategy does not have as large an eﬀect.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	distortion factor fc as a result of the varying eﬀective temperature (as indicated by the decreasing blackbody temperature) is exactly balanced by some other varia- tion, e.g. a change in the emitting radius as the burst ﬂux decreases; or, that the color temperature correction fc is also constant throughout the interval in which Kbb is constant. The former explanation is rather contrived, particularly considering that in the burst tail the burning front is expected to have already spread to the entire sur- face area of the neutron star, and no further increase (or decrease) in burning area is expected. Thus, these mea- surements suggest that for the range of eﬀective temper- atures spanned in the burst tails, the color-temperature correction fc is roughly constant. This conclusion is dif- ﬁcult to reconcile with the predictions of atmosphere models (e.g. Suleimanov et al. 2011b), which indicate signiﬁcant variations in fc over most of the ﬂux range spanned during the burst. The distance for GS 1826−24 is thought to be ≈ 6 kpc (Heger et al. 2007), at which an Eddington-limited burst would be expected to reach F/FEdd = 3.7 × 10−8 (6.3 × 10−8) erg cm−2 s−1 for an H-rich (pure He) atmosphere (e.g. G08). This would imply that the range of ﬂux over which the burst nor- malisation is found to be constant is 0.16–0.75 (0.10– 0.46) F/FEdd. The greatest variation in the normali- sation predicted by Suleimanov et al. (2011a) is outside these ranges, although signiﬁcant variations would yet be expected (particularly for the solar metallicity mod- els; see also Z11).
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	The ≈ 3–5% fractional variation in hKbbi observed from GS 1826−24 and KS 1731−26 within each obser- vational epoch was smaller than that seen in two radius- expansion bursts observed from EXO 1745−248, of 25% ( ¨Ozel et al. 2009). This observation perhaps suggests an explanation of the variation. EXO 1745−248 exhib- ited during it’s 2000 outburst an initial period of strong variability, reminiscent of dipping behaviour observed in high-inclination systems (e.g. Galloway et al. 2008). No such variability has ever been observed from GS 1826−24 or KS 1731−26, suggesting that perhaps the inclination is lower in those systems than in EXO 1745−248. If system inclination is the main factor in determining the variation in apparent blackbody normalisation, a possi- ble explanation is the reprocessing of some fraction of the burst ﬂux oﬀ an accretion disk whose projected area varies with time. In 4U 1728−34, the timescale inferred for the variation was several tens of days (Galloway et al. 2003), much longer than the expected orbital period of this system (e.g. Galloway et al. 2010). However, in GS 1826−24 and KS 1731−26, the variation is on a much shorter timescale, comparable to the recurrence time of the bursts themselves (hours). For comparison, the dis- crepant bursts from EXO 1745−248 were separated by 8.5 d. We note that such an explanation fails to ac- count for the factor of ≈ 2 diﬀerence in normalisation in the bursts from 4U 1724−307 (Suleimanov et al. 2011a), which does not show dips and thus is unlikely to be at high inclination. However, those bursts also exhibited markedly diﬀerent timescales, indicative of a varying ac- cumulated fuel reservoir at ignition; we suggest instead that diﬀerent physical conditions (temperature, compo- sition) gave rise to the diﬀerence in the measured black- body normalisation.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	for GS 1826−24 and KS 1731−26 suggests that there may be an orbital component of the variation. The speciﬁc value of the blackbody normalisation depends upon the assumed neutral column density, as illustrated in Fig. 6. Thus, an orbital variation in the line-of-sight column den- sity, perhaps arising from cool clouds of material above the point of contact of the accretion stream with the disk, will manifest as a variation in the blackbody normalisa- tion on the same timescale. In order to test this hy- pothesis, we calculated a Lomb-normalised periodogram on the blackbody normalisation measurements as a func- tion of time. Recall the orbital period in both sources is unknown, but for GS 1826−24 is thought to be around 2 hr; based on the typical orbital periods for other burst sources, we searched a frequency range of 0.5–24 hr. For GS 1826−24, we divided the measurements from the two epochs through by the appropriate mean, and found a peak Lomb power of 10.48; for KS 1731−26, we found a peak Lomb power of 5.29. Neither of these detections is signiﬁcant. For KS 1731−26, the known sensitivity of the normalisation measurement to discrepancies between the assumed and true value of nH indicates that a vari- ation of 0.3 × 1022 cm−2 in nH could account for the variation in the measured blackbody normalisation. For GS 1826−24, the measurements are slightly more sensi- tive to discrepancies in nH , so that a variation in nH of 0.45 × 1022 cm−2 could account for the variation in the blackbody normalisation within each epoch. If orbital or longer-timescale variations in nH are driving the varia- tion in the blackbody normalisation, it may be possible to verify through time-resolved high-spectral resolution measurements of absorption edges in the X-ray spectra.
Processed_On_the_consistency_of_neutron-star_radius_measurem.txt	We thank the anonymous referee, who made several substantial comments which improved the paper. DKG is a member of an International Team in Space Science on type-I X-ray bursts sponsored by the International Space Science Institute (ISSI) in Bern, Switzerland, and we thank ISSI for hospitality during part of this work. This research has made use of data obtained through the High Energy Astrophysics Science Archive Research Center Online Service, provided by the NASA/Goddard Space Flight Center. DKG is the recipient of an Australian Re- search Council Future Fellowship (project FT0991598).
Processed_The_Nature_of_Low-Density_Star_Formation.txt	How do stars manage to form within low-density, HI-dominated gas? Such environments provide a laboratory for studying star formation with physical conditions distinct from starbursts and the metal-rich disks of spiral galaxies where most effort has been invested. Here we outline fundamental open questions about the nature of star formation at low-density. We describe the wide-field, high-resolution UV-optical-IR-radio observations of stars, star clusters, and gas clouds in nearby galaxies needed in the 2020’s to provide definitive answers, essential for development of a complete theory of star formation.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	Star formation (SF) is a principle driver of galaxy evolution. It occurs under an enormous range of conditions, as metallicity, gas richness, the interstellar radiation field and phase balance vary dramatically across cosmic space and time. Decades of observations across the electromagnetic spectrum have enabled the detailed characterization of SF in the environments of starbursts and the metal-rich disks of spiral galaxies.0 We know far less about how stars manage to form within low-density, HI dominated gas.1 Such environments are important because they characterize star forming dwarfs, the most common type of gas-rich galaxy and building blocks of larger systems, as well as the outskirts of more massive galaxies, the sites of cold gas reservoirs for inner disks, where galaxy growth occurs today, and where disks must end and meet the intergalactic medium. Presumably, it is within such environments that the first stars formed. Addressing this shortcoming is crucial for understanding key parameters that influence the SF and chemical enrichment that occurs in the vast majority of neutral HI gas in the universe out to high redshift, otherwise accessible only as absorption line systems such as DLAs.3,4 Fundamentally, it is essential for development of complete theory of star formation.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	GALEX2 was a revolutionary probe of recent star formation in this low surface brightness (LSB) regime. Its wide (1.2°) field-of-view enabled discovery of “XUV disks” with UV-bright SF extending to several times beyond the optical disk (Fig. 1).5,6,7 GALEX provided the UV data for studies that revealed more SF in low-density environments than previously inferred from Ha observations.8,9,10,11 This renewed interest in understanding the physics that sets the conversion efficiency of gas to stars in such environments 12,13,14 (top-right Fig. 1). It motivated development of a new generation of stellar population synthesis models required for inferring the physical properties of populations where the massive end of the stellar initial mass function is not fully sampled.15,16 It leads to questions about the fraction of stars formed in bound clusters (G) as a function of density17,18 (Fig. 1) and how variations in G might mirror changing cloud populations.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	formation is in principle possible with the VLA, but the time requirements are prohibitive.29 Statistically sound investigations probing the full range of ΣSFR with stars and gas, in particular controlling for galactic environmental parameters such as metallicity, and mitigating against degeneracies between mass function shape and SF history, would require years of time with HST and the JVLA. Here, we outline critical problems that can be solved with new capabilities in the 2020’s to achieve a more complete understanding of SF in all of its myriad forms.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	When SF is struggling to occur, how much SF activity is locked into long-lived clusters? Most stars are born in clustered environments (including associations), but their subsequent fate is far less clear. The cluster formation efficiency, G = SFRcluster / SFRtotal (fraction of SF occurring in bound clusters30) at low ΣSFR is vigorously debated with evidence of reduced efficiency18 (Fig. 1) but counterclaims of nearly constant G .31 Additional constraint on the theoretical models17 (predicting a downturn vs. Σgas hence ΣSFR ) is now vital. The outer disks of galaxies are an ideal place to make the G measurement, but a census of sufficient integrated SFR is necessary, implying a wide area. Total <SFR> of ~ few x 10-2 "⨀yr-1 over a 300-500 Myr duration is required per each independent sample of an environmental regime, otherwise the lowest G values (G < 0.1) become indistinguishable due to low expected cluster number counts. Note that it is essential to exclude the youngest clusters, possibly dissolving until ~10 Myr, on the basis of UV-optical SED fitting. A G- ΣSFR study conducted at the extreme limit of SF activity, in an environment less hostile to dynamical cluster disruption, would also gauge the importance of internal vs. external cluster destruction mechanisms with clarity impossible in the complex inner disk and inter-arm zones.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	Are there upper IMF variations in LSB star forming environments? If so, what drives them? The stellar initial mass function (IMF) is a fundamental parameter that encodes the complex physics of SF, and is crucial for interpreting most observations in extragalactic astronomy. Since the idea of the “original” mass function was introduced by Salpeter in 1955, considerable effort has been invested to verify its form, and historically most studies have concluded that the IMF is invariant32. However, studies of LSB environments have suggested a possible deficiency of high mass stars. Forward-modeling of the luminosity function of main sequence stars33,34,35 show trends in the IMF slope (α) above 1"⨀, or potentially severe limits on the maximum stellar mass (Mu), perhaps linked to the low pressures, gas densities, ΣSFR and/or metallicity 9,36,37 of outer disks. Indirect constraints from integrated light studies (e.g., M/L ratio, deficiencies of Hα emission)38,39,10,40,41,9 frequently support this picture, but there are exceptions42. Efforts in the next decade must address the significant limitations of current studies in order to provide conclusive results. IMF experiments must be conducted with large samples of massive stars from different galaxies to (1) probe a range of LSB environments (e.g. varied pressure, metallicity, local dynamics) and provide insight into the drivers of possible variations, (2) overcome Poisson uncertainties, (3) address degeneracies between IMF shape and star formation history (e.g., by averaging over many independent regions, and using relatively IMF-insensitive core He burning stars, best distinguished in blue/UV-optical CMDs, to directly derive the SFH in the last few 100 Myr) 43. In the Local Volume (< 10 Mpc), HST-resolution spectroscopy in the optical or UV will be essential to break α-Mu degeneracy, relying on relative line strength of different ionization stages (He I / He II, Si III / Si IV) and P-Cygni (C IV, N V) profiles to assign spectral type.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	Fig. 1: How do the physics of star formation change as a function of density? Wide-field, high-resolution imaging is required to study SF in its primary quanta (stars, star clusters, gas clouds) over the complete range of densities probed by nearby galaxies, with new work needed on the low-density, HI-dominated regime. (Left-center) M83 (d=4.5 Mpc) and its archetypal extended UV disk. GALEX FUV/NUV is shown in (blue/green) with HI gas (red). A magenta rectangle shows the CASTOR FOV, dashed red squares show the WFIRST FOV and small white rectangles mark existing outer disk HST data with blue imaging. (Left, top+bottom) M83 HST imaging at typical and low-density, respectively. (Right-top) The K-S relation between SFR surface density ΣSFR and gas surface density, from Bigiel+201045. (Right-bottom) The relation between the fraction of stars formed in bound clusters and ΣSFR, from Adamo 201718. (In all panels) Regions highlighted by red circles correspond to the well-studied central regions of galaxies while blue circles correspond to the low densities characteristic of galaxy outskirts, typical dwarf galaxies, tidal dwarfs, and LSB galaxies (including the sub-population of HI-bearing ultra-diffuse galaxies, UDGs51, called HUDS52).
Processed_The_Nature_of_Low-Density_Star_Formation.txt	Fig. 2: CMDs of NGC2403 from GHOSTS53 the difficulty a statistical sample of stars and clusters at low-density. The HST field probing low-density shows only a few upper main sequence (uMS) stars and a modest core He-burning (HeB) population, but is otherwise dominated by old stars. Few stellar clusters are found in this field. The inner disk has a well-populated uMS and many clusters. To accumulate a similar statistical sample in the LSB regime would be very costly using HST.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	on this topic have relied on surface photometry in the UV, Ha and/or IR and are prone to systematic errors. A direct reassessment is needed of the resolved, low-density K-S relation54,55,56 via counting of young massive star candidates (plus incorporation of any young clusters) occupying NHI bins. Relative motion of stars and gas is a complication, but can be controlled for by averaging over scales consistent with stellar lifetimes and typical velocity dispersions. Perhaps more importantly, this study would also yield a statistically meaningful description of the “SF products” [e.g. max/median cluster mass, max stellar mass] and small-scale clustering properties thereof, plus gas-based metrics (below), as a function of density and ΣSFR . Such studies will provide direct constraints on the origin of scatter in the faint end of the K-S relation.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	constraints on many questions can come by correlating average physical quantities (e.g., column65 or line width distribution functions, cloud mass functions66) against young population indicators.
Processed_The_Nature_of_Low-Density_Star_Formation.txt	This path forward requires innovative capabilities: (1) UV-optical imaging with ~HST resolution and sensitivity but with extremely wide FOV (Table 1) to inventory stars and clusters at low ΣSFR. (2) LUVOIR or HabEx for highly-multiplexed UV-optical spectral typing of O star candidates within ~10 Mpc. (3) cm-wave radio interferometry with 10× JVLA HI sensitivity to characterize dense HI gas clouds on small scales. (4) Efficient FIR spectroscopic mapping to capture dust properties and ISM cooling. (5) Deployment of heterodyne arrays on existing mm-wave telescopes.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	We present the analysis of MOA-2007-BLG-197Lb, the ﬁrst brown dwarf companion to a Sun-like star detected through gravitational microlensing. The event was alerted and followed-up photometrically by a network of telescopes from the PLANET, MOA, and µFUN collaborations, and observed at high angular resolution using the NaCo instrument at the VLT. From the modelling of the microlensing light curve, we derived basic parameters such as, the binary lens separation in Einstein radius units (s (cid:39) 1.13), the mass ratio q = (4.732 ± 0.020) × 10−2 and the Einstein radius crossing time (tE (cid:39) 82 d). Because of this long time scale, we took annual parallax and orbital motion of the lens in the models into account, as well as ﬁnite source eﬀects that were clearly detected during the source caustic exit. To recover the lens system’s physical parameters, we combined the resulting light curve best-ﬁt parameters with (J, H, Ks) magnitudes obtained with VLT NaCo and calibrated using IRSF and 2MASS data. From this analysis, we derived a lens total mass of 0.86 ± 0.04 M(cid:12) and a lens distance of DL = 4.2 ± 0.3 kpc. We ﬁnd that the companion of MOA-2007-BLG-197L is a brown dwarf of 41 ± 2 MJ observed at a projected separation of a⊥ = 4.3 ± 0.1 AU, and orbits a 0.82 ± 0.04 M(cid:12) G-K dwarf star. We then placed the companion of MOA-2007-BLG-197L in a mass-period diagram consisting of all brown dwarf companions detected so far through diﬀerent techniques, including microlensing, transit, radial velocity, and direct imaging (most of these objects orbit solar-type stars). To study the statistical properties of this population, we performed a two-dimensional, non-parametric probability density distribution ﬁt to the data, which draws a structured brown dwarf landscape. We conﬁrm the existence of a region that is strongly depleted in objects at short periods and intermediate masses (P (cid:46) 30 d, M ∼ 30 − 60 MJ), but also ﬁnd an accumulation of objects around P ∼ 500 d and M ∼ 20 MJ, as well as another depletion region at long orbital periods (P (cid:38) 500 d) and high masses (M (cid:38) 50 MJ). While these data provide important clues on the diﬀerent physical mechanisms of formation (or destruction) that shape the brown dwarf desert, more data are needed to establish their relative importance, in particular as a function of host star mass. Future microlensing surveys should soon provide more detections, in particular for red dwarf hosts, thus uniquely complementing the solar-type host sample.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	Gravitational microlensing is a powerful technique for detecting extrasolar planets (Mao & Paczynski 1991), and it holds great promise for detecting populations of brown dwarf companions to stars. Compared to other detection techniques, microlensing provides unique information on the population of exoplanets, be- cause it allows the detection of very low-mass planets (down to the mass of the Earth) at long orbital distances from their host stars (typically 0.5 to 10 AU). It is also the only technique that allows discovery of exoplanets and brown dwarfs at distances from the Earth greater than a few kiloparsecs, up to the Galactic bulge, which would have been hard to detect with other methods. Exoplanets are found to be frequent by all detection tech- niques (e.g., Cassan et al. 2012; Bonﬁls et al. 2013; Mayor et al. 2011; Sumi et al. 2011; Gould et al. 2010b), and recent statistical microlensing studies even imply that there are, on average, one or more bound planets per Milky Way star (Cassan et al. 2012).
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	Conversely, brown dwarfs appear to be intrinsically rare, to the point that shortly after the ﬁrst exoplanet detections, it led to the idea of a “brown dwarf desert” (Marcy & Butler 2000) bridg- ing the two well-deﬁned regions of binary stars and planetary systems. While in the past, brown dwarfs were deﬁned as ob- jects of mass within the deuterium- and hydrogen-burning limits (13 − 74 MJ, Burrows et al. 2001), it appears today that diﬀer- ent formation scenarios can build objects with similar masses but with diﬀerent natures (super-massive planets, or low-mass brown dwarfs). An object formed via core accretion and reach- ing 13 MJ would, for example, be able to start deuterium burn- ing, as would an object of same mass formed by gravitational collapse of a cloud or in a protoplanetary disk (Mollière & Mor- dasini 2012).
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	Despite their low occurrence, a number of brown dwarf com- panions to stars have been discovered by diﬀerent methods: ra- dial velocity and transit (e.g., Moutou et al. 2013; Díaz et al. 2013; Sahlmann et al. 2011; Johnson et al. 2011; Deleuil et al.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	include isolated brown dwarfs, brown dwarfs hosting planets, and brown dwarf com- panions to stars. The ﬁrst isolated brown dwarf detected through gravitational microlensing (a 59 ± 4 MJ brown dwarf located at 525 ± 40 pc in the thick disk of the Milky Way) was re- ported by Gould et al. (2009) in microlensing event OGLE-2007- BLG-224. Two brown dwarfs with planetary-mass companions were discovered in events OGLE-2009-BLG-151/MOA-2009- BLG-232 and OGLE-2011-BLG-0420 by Choi et al. (2013). In both cases, the planets were super Jupiters (7.9 ± 0.3 MJ and 9.9 ± 0.5 MJ, respectively) with the hosts being low-mass brown dwarfs (19 ± 1 MJ and 26 ± 1 MJ, respectively), with very tight orbits (below 0.4 AU). Similarly, Han et al. (2013) report a 23 ± 2 MJ ﬁeld brown dwarf hosting a 1.9 ± 0.2 MJ planet in a tight system after the analysis of the event OGLE-2012-BLG- 0358.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	The ﬁrst published microlensing detection of a brown dwarf companion to a star is OGLE-2008-BLG-510/MOA-2008-BLG- 369, which was ﬁrst reported by Bozza et al. (2012) as an am- biguous case between a binary-lens and a binary-source event. The data were reanalysed by Shin et al. (2012a), who con- cluded that the binary-lens model involving a massive brown dwarf orbiting an M dwarf was preferred. Shin et al. (2012b) conducted a database search for brown dwarf companions by focusing on microlensing events that exhibit low mass ratios. Among seven good candidates with well-determined masses (combination of Einstein radius and parallax measurements), they found two events that involve brown dwarfs: OGLE-2011- BLG-0172/MOA-2011-BLG-104 with mass 21 ± 10 MJ around an M dwarf, and MOA-2011-BLG-149, a 20±2 MJ brown dwarf also orbiting an M dwarf. Similarly, Bachelet et al. (2012) re- ported the detection of another ∼ 52 MJ brown dwarf orbiting an M dwarf in MOA-2009-BLG-411, although the lens mass could not be determined exactly, and was estimated through sta- tistical realisations of Galactic models. In microlensing event MOA-2010-BLG-073, Street et al. (2013) ﬁnd that the lens was composed of a 11.0 ± 2.0 MJ companion (hence near the planet/brown boundary) orbiting an M dwarf of 0.16 ± 0.03 M(cid:12). Jung et al. (2015) report the detection of a star at the limit of the brown dwarf regime hosting a companion at the planet/brown dwarf boundary (13 ± 2 MJ). More recently, Park et al. (2015) have reported the discovery of a binary system composed of a 33.5 ± 4.2 MJ brown dwarf orbiting a late-type M dwarf in mi- crolensing event OGLE-2013-BLG-0578.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	Here we report the ﬁrst microlensing discovery of a brown dwarf orbiting a Sun-like star. This new brown dwarf has a mass of 42 MJ and it was observed at a projected separation of 4.3 AU from its G-K dwarf host star.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	probability density estimation tools. In sec. 6 we summarise our results, and underline the importance of future microlensing ob- servations to characterise the populations of objects in the mass region between planets and stars.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	On June 19 (THJD (cid:39) 4272), PLANET observers noticed that a few MOA data points were slightly above the standard single-lens theoretical curve. As this usually happens quite often in microlensing data, no alert was released. Indeed SAAO soon revealed that the alleged deviation was a false alert, but follow- up observations were continued. On July 5 (THJD (cid:39) 4287), even though the full moon aﬀected the quality of the observa- tions, a public alert was issued after Danish 1.54m data were found to be above the single lens curve by more than 0.2 mag for more than ﬁve consecutive days. Intensive follow-up obser- vations from SAAO conﬁrmed the rise in brightness, announcing a caustic crossing. While Perth was overclouded, Canopus then took over, and was the only telescope to densely cover the caus- tic exit, which took place during the night of July 4 in Australia (THJD (cid:39) 4287.0-4287.4, see inset of Fig. 1).
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	Notes. (a) Number after data cleaning. (b) Error bar rescaling factor (Sect. 2.2). (c) MOA broad R/I ﬁlter.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	Fig. 1. In the upper panel, the light curve of MOA-2007-BLG-197 and the best-ﬁt model (solid line) are plotted with a zoom on the caustic exit on the right-hand side. On the left-hand side, the structure of the resonant caustic is drawn in red, as well as the trajectory of the source, in black (axes are in Einstein radius units). The source is too small to be distinguished. The four intervals E1−4 indicate time intervals where possible source caustic crossing (caustic entry) have been investigated. In the lower panel, the residuals of the best-ﬁt model are shown.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	The ﬁnal light curve data amounts to a total of 802 data points. They are summerised in Table 1. As seen in the table, all telescopes use a similar I-band ﬁlter, apart from the MOA 1.8m telescope which is equipped with a broad R/I ﬁlter (re- ferred to as RM). Additionally, a few V-band images were taken by PLANET and µFUN to produce colour-magnitude diagrams (see Sect. 4.1).
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	that one data set dominates over the others at the modelling stage. A relatively robust method to prevent these drawbacks is to rescale the error bars, based on the best model ﬁtting the data. For each data set, the (classical) χ2 is set up to the number of de- grees of freedom by adjusting a rescaling factor f in the formula σ(cid:48)2 = f 2σ2 + σ2 0, where σ(cid:48) and σ are respectively the rescaled and initial error bars on the magnitudes, and σ0 = 4 × 10−4 a constant accounting for the data most highly magniﬁed. The f factors are given in Table 1 for each data set.
Processed_MOA-2007-BLG-197:_Exploring_the_brown_dwarf_desert.txt	tics (AO) images in the near-infrared bands J, H and Ks using the NaCo instrument, mounted on 8.2m ESO VLT Yepun tele- scope (Fig. 2). The source star was then still magniﬁed by a fac- tor of about A (cid:39) 1.30. A year later, on the night 2008 August 3/4 (THJD (cid:39) 4682.1), we carried out second epoch observa- tions3 while the event was back to baseline magnitude. In princi- ple, when two epochs are obtained at diﬀerent magniﬁcations, it should be possible to directly disentangle the lens ﬂux from the source ﬂux. In our case, however, the combination of a relatively high blending factor and low magniﬁcation did not support this direct measurement. Therefore, in the global analysis we used only the ﬁrst epoch images. We reduced and calibrated the NaCo images following the general method outlined in Kubas et al. (2012) and brieﬂy described below.
Processed_H2O_Maser_Polarization_of_the_Water_Fountains_IRAS.txt	ABSTRACT We present the morphology and linear polarization of the 22-GHz H2O masers in the high-velocity outﬂow of two post-AGB sources, d46 (IRAS 15445−5449) and b292 (IRAS 18043−2116). The observations were performed using The Australia Telescope Compact Ar- ray. Different levels of saturated maser emission have been detected for both sources. We also present the mid-infrared image of d46 overlaid with the distribution of the maser features that we have observed in the red-shifted lobe of the bipolar structure. The relative position of the observed masers and a previous radio continuum observation suggests that the continnum is produced along the blue-shifted lobe of the jet. It is likely due to synchrontron radiation, implying the presence of a strong magnetic ﬁeld in the jet. The fractional polarization levels measured for the maser features of d46 indicate that the polarization vectors are tracing the poloidal component of the magnetic ﬁeld in the emitting region. For the H2O masers of b292 we have measured low levels of fractional linear polarization. The linear polarization in the H2O maser region of this source likely indicates a dominant toroidal or poloidal magnetic ﬁeld component. Since circular polarization was not detected it is not possible to determine the magnetic ﬁeld strength. However, we present a 3-σ evaluation of the upper limit intensity of the magnetic ﬁeld in the maser emitting regions of both observed sources.
Processed_H2O_Maser_Polarization_of_the_Water_Fountains_IRAS.txt	Key words: masers – stars: AGB and post-AGB – stars: late-type – stars: magnetic ﬁelds – polarization – stars: circumstellar matter.
Processed_H2O_Maser_Polarization_of_the_Water_Fountains_IRAS.txt	jet of W43A, the archetypal water fountain. They found that the jets of W43A are magnetically collimated. Therefore, the detection of polarized maser emission from water fountains is useful to deter- mine the role of the magnetic ﬁelds on the onset of wind asymme- tries during the evolution from AGB stars to aspherical PNe. Here we report the detection of linear polarization of 22-GHz H2O maser emission from two water fountains d46 and b292 (IRAS 15445−5449, IRAS 18043−2116).
Processed_Scale-free_networks_in_complex_systems.txt	In the past few years, several studies have explored the topology of interactions in diﬀerent complex systems. Areas of investigation span from biology to engineering, physics and the social sciences. Although having diﬀerent microscopic dynamics, the results demonstrate that most systems under consideration tend to self-organize into structures that share common features. In particular, the networks of interaction are characterized by a power law distribution, P (k) ∼ k−α, in the number of connections per node, k, over several orders of magnitude. Networks that fulﬁll this propriety of scale-invariance are referred to as “scale-free”. In the present work we explore the implication of scale-free topologies in the antiferromagnetic (AF) Ising model and in a stochastic model of opinion formation. In the ﬁrst case we show that the implicit disorder and frustration lead to a spin- glass phase transition not observed for the AF Ising model on standard lattices. We further illustrate that the opinion formation model produces a coherent, turbulent-like dynamics for a certain range of parameters. The inﬂuence, of random or targeted exclusion of nodes is studied.
Processed_Scale-free_networks_in_complex_systems.txt	In the following section we brieﬂy describe the algorithm used for generating the SFN, that is the Barab´asi- Albert model with tunable clustering. We then study the implications of a SFN topology in two diﬀerent complex systems. The ﬁrst one, Sec. 3, in the antiferromagnetic (AF) Ising model while the second, Sec. 4, is a stochastic model for opinion formation.
Processed_Scale-free_networks_in_complex_systems.txt	The ubiquity of SFNs in nature has inspired physicists to investigate the dynamics of standard models in the new case where the interactions between elements are described by complex interactions. These include the study of various magnetic models such as the Ising model. An intriguing issue concerns how the unusual topology acts to inﬂuence the cooperative behaviour of the spins. Studies of the ferromagnetic (FM) Ising model on a SFN, using several theoretical techniques26–29 including the Monte Carlo (MC) method,29 have found the robustness of ferromagnetic ordering against thermal ﬂuctuations for the degree distribution exponent α ≤ 3.
Processed_Scale-free_networks_in_complex_systems.txt	The robustness feature is naturally expected as SFNs have large connectivities. This is analogous to the FM Ising model on a regular lattice above the lower critical spatial dimension, dl = 2. There the ordered phase is very robust against thermal ﬂuctuations. However, for the antiferromagnetic (AF) case with a SFN, the situation is diﬀerent.
Processed_Scale-free_networks_in_complex_systems.txt	Two factors come to play a central role in the dynamics of the AF-SFN model; namely the competition induced by the AF interaction in the elementary triangles of the network and the randomness related to the non-regular connections. The abundance of elementary triangles in the network leads to frustration, as, for example, only two of the three spins can be anti-aligned. More generally, frustration refers to the inability of the system to remain in a single lowest energy state (ground state). These ingredients lead the AF SFN to belong to a class of randomly frustrated systems commonly referred to as spin glasses (SGs).
Processed_Scale-free_networks_in_complex_systems.txt	We consider the AF Ising model on a Barab´asi-Albert network with a tunable clustering coeﬃcient, as described in Sec. 2. We illustrate that the AF model undergoes a SG transition. Such a transition is not observed on a regular triangular lattice where, for the AF Ising model, the spins are fully frustrated.
Processed_Scale-free_networks_in_complex_systems.txt	Here the summation is performed over the connected spins si and sj occupying sites i and j, respectively. The coupling interaction Jij = J < 0 is AF. As previously mentioned, each vertex with the local cluster coeﬃcient Ci > 0 together with its neighbours, compose elementary triangles. Due to the AF interactions the local system is frustrated.
Processed_Scale-free_networks_in_complex_systems.txt	As a random system, each realization of a network of size N will diﬀer in the “structure” of connectivities. Therefore, in order to have reliable statistics, we average over many realizations of the SF network for each speciﬁed size. In general, one takes into account more realizations for small system sizes and less for large system sizes as the latter tend to self-average. The system sizes that we simulate are N = 1024, 2048, 4096, and 8192. Since the self-averaging of physical quantities for larger system sizes are interfered by the increase of ground state degeneracy, we did not take less realizations. Instead all physical quantities of interest for each system size are averaged over 1000 network realizations.
Processed_Scale-free_networks_in_complex_systems.txt	Another peculiarity of SF networks regards the existence of a broad distribution of “hubs”, that is nodes with a large number of connections, k. The energy diﬀerence in a spin ﬂip actually depends on the number of ki j=1 sj. Thus in the AF case for the ith spin with ki connections, connections of the spin itself, ∆Ei = −2si the hubs are more likely to “freeze” into a particular conﬁguration compared to the nodes with just few links. This fact resembles the spin glass behaviour of particular alloys where some elements freeze into a particular orientation at a higher temperature than others.
Processed_Scale-free_networks_in_complex_systems.txt	The calculation of the thermal averages of the physical quantities of interest is performed using the replica exchange MC method,32 appropriate for systems such as spin-glass. For a given network conﬁguration, replicas having an associated inverse temperature, β, are created. In using this method, we deﬁne a “local” MC (LMC) update as a MC update for each spin of each replica, either consecutively through all elements of the network or randomly. Given that we can group the inverse temperatures in even and odd pairs, (βm, βm+1), after each LMC update we alternate attempts to switch conﬁgurations from one temperature to the next. According to this procedure, we deﬁne a Monte Carlo step (MCS) as a LMC plus a half (m odd or even) exchange trial.
Processed_Scale-free_networks_in_complex_systems.txt	For each network realization we run 3 × 105 MCSs after a transient period of 103 LMC updates. We take a total of 6 × 104 measures for the thermal averages. The simulation is run down to low temperatures in a search for the possible existence of a phase transition. All the thermal averages obtained are then averaged over the whole ensemble of networks. In the following, we indicate h...i as the thermal average and [...]av as the ensemble average. The statistical errors in the plots, where reported, are calculated via the bootstrap method.
Processed_Scale-free_networks_in_complex_systems.txt	i X where the superscripts α and β denote two copies of the same conﬁguration of connectivity at the same temper- ature.
Processed_Scale-free_networks_in_complex_systems.txt	In particular, for the Ising system, due to the Z2 symmetry, it is important to evaluate the absolute value of the order parameter, \|q\| = , to overcome the implication of the Z2 symmetry of the Hamiltonian. That is, if the system is at thermal equilibrium and if we take quite long MCS then the usual q should average out and give an approximately zero value. The existence of a spin glass phase is indicated by the convergence of \|q\| to a ﬁnite value as we increase the network size and, at the same time, a convergence of \|q\| to zero at high temperatures. In the latter case the system is in the paramagnetic phase.
Processed_Scale-free_networks_in_complex_systems.txt	Figure 1. Left: Temperature dependence of the overlap parameter, q, for diﬀerent system sizes N . The increasing value of q at low temperatures indicates a SG phase. For a given network size, 1000 realizations of the SFN are averaged over. Right: The distribution of q at various temperatures for diﬀerent system sizes, including (a) N = 1024, (b) N = 2048, (c) N = 4096 and (d) N = 8192. The temperatures are provided in units of J/kB, where kB is the Boltzmann constant.
Processed_Scale-free_networks_in_complex_systems.txt	The temperature dependence of \|q\|, resulting from the simulations, is shown in Fig. 1 (Left). The existence of a SG phase is indicated by the ﬁnite value of \|q\| in the low temperature region, and the approach of \|q\| to zero at higher temperatures associated with paramagnetic phase. For high temperatures and large networks, \|q\| is approaching zero in accord with the thermodynamic limit where \|q\| = 0.
Processed_Scale-free_networks_in_complex_systems.txt	The existence of these two diﬀerent phases can also be observed from the distributions of q, as shown in Fig. 1 (Right). For higher temperatures we observe simple Brownian ﬂuctuations of the values of q, leading to a singly peaked Gaussian distribution characteristic of a paramagnetic state. By decreasing the temperature, the distribution starts to spread out, reﬂecting the increasing number of metastable disordered states reﬂecting the presence of substantial frustration. At lower temperatures the distribution develops double peaks associated with the Edward-Anderson parameter representative of the SG phase. The transition between these two phases is roughly estimated at T ∼ 4. We note that the shape of the observed distribution is diﬀerent from that of the conventional Ising system where the double peaks approach delta-like double peaks reﬂecting a simple doubly degenerate ground state.
Processed_Scale-free_networks_in_complex_systems.txt	where hq2i and hq4i are respectively the second and the fourth cumulant moment of q and 0 ≤ gL ≤ 1. At high temperature, when the thermal ﬂuctuation overcomes all cooperative interaction, the system is expected to exist in the paramagnetic phase where there is no spatial nor temporal autocorrelation. As a result, the distribution of q should be Gaussian centered at q = 0. In this case the ratio of the cumulants, hq4i/hq2i2 → 3, resulting in gL → 0. At low temperatures, the cooperative interaction becomes dominant and the ratio of the cumulants approaches unity so that gL = 1.
Processed_Scale-free_networks_in_complex_systems.txt	Figure 2. Top: Scaling behaviour of the Binder cumulant, gL, for diﬀerent system sizes. Each system size is averaged over 1000 realizations of the network conﬁguration. Bottom: Scaling plot of the data illustrated above (Top), ﬁt to Eq. 4.
Processed_Scale-free_networks_in_complex_systems.txt	is T ∼ 4.0. For temperatures above T ∼ 4.0 the Binder parameter, while remaining always above zero, does indeed order in an opposite manner indicative of a genuine crossing of the curves and in accord with a genuine spin glass transition at ﬁnite temperature.
Processed_Scale-free_networks_in_complex_systems.txt	with ν > 0. At Tc the Binder cumulant does not depend on L. For the SFN, the system size scales logarithmically with the number of nodes N ,23 and therefore we take L = log(N ). The parameters Tc and ν are determined by constraining the temperature dependence of the Binder parameter for each network size to lie on a single curve. The curves following the scaling bahaviour of Eq. (4) are shown in Fig. 2 (Bottom). ¿From this ﬁt we estimate the critical temperature Tc ∼ 4.0(1) and the exponent of the SG correlation length ν ∼ 1.10(2). It is important to underline that this kind of behaviour is not observed for an AF system on a regular triangular lattice.
Processed_Scale-free_networks_in_complex_systems.txt	We turn now our attention to the role played by the SFN on a model of stochastic opinion formation. In this case, once the scale-free network has been built, we randomly assign the spin values, ±1, to every node. These values correspond to a Boolean kind of opinion while the bonds of the networks represent the interactions between agents.
Processed_Scale-free_networks_in_complex_systems.txt	Figure 3. Left: Time series of the average opinion, r, for diﬀerent values of the group interaction strength parameter a: (i) a = 0.8, (ii) a = 1.5, (iii) a = 1.8 and (iv) a = 2.3. The parameters used for the simulations are: N = 104 nodes, clustering probability θ = 0.9, initial nodes and links per new node m0 = m = 5 and we take the upper bound of the distribution of personal response strengths equal to the group interaction strength, that is κ = a. The results involve 10 realizations of the scale free network each displayed for 5000 time steps. For values of a greater than 1 a turbulent-like state, characterized by large ﬂuctuations, starts to appear in the process of opinion formation. Right: PDFs of the time series relative to the time series in (Left). The shapes of the distributions converge to a Gaussian for small values of the group interaction strength a = κ. A Gaussian distribution is also plotted for comparison. All the PDFs in this paper are obtained over 50 realizations of the SF network. In order to compare the ﬂuctuations at diﬀerent scales, the time series in the plot have been normalized according to r(t) → r(t)−¯r , where ¯r and σ denote the average and the standard deviation over the period considered respectively.
Processed_Scale-free_networks_in_complex_systems.txt	The ﬁrst term on the right-hand side of Eq. (5) represents the time dependent interaction strengths between the node i and his/her ˜Ni information sources, which are the ﬁrst neighbours in the network. The second term N j=1 σj(t), instead reﬂects the personal reaction to the system feedback, that is the average opinion, r(t) = 1/N resulting from the previous time step. The terms ξ(t) and ηi(t) are random variables uniformly distributed in the interval (-1,1) with no correlation in time nor in the network. They represent the conviction, at time t, with which agent i responds to his/her group (common for all the agents) and the global opinion of the network respectively. The strength term, a, is constant and common for the whole network, while hi is speciﬁcally chosen for every individual from a uniform distribution in (0,κ) and are both constant in the dynamics of the system. By varying the parameter κ we can give more or less weight to the role of feedback in the model. The strength coeﬃcients a and hi in the local ﬁeld, Ii, characterizing the attributes of the agents, play a key role in the dynamics of the model. They represent the relative importance that each agent of the network gives, respectively, to his/her group and to the variation of the average opinion itself. While a is a parameter associated with the network, hi is speciﬁcally chosen for each individual at the beginning of each simulation.
Processed_Scale-free_networks_in_complex_systems.txt	At ﬁrst we investigate the importance of the group strength a by ﬁxing κ = a. In this case the dynamical behaviour is similar to that found in the stock market context in Refs..34–36 For a ˜<1 the resulting time series of average opinion is largely uncorrelated Gaussian noise with no particularly interesting features, as illustrated in Fig. 3(i) (Left).
Processed_Scale-free_networks_in_complex_systems.txt	As soon as we exceed the value of a ≈ 1 a turbulent-like regime sets in, characterized by large intermittent ﬂuctuations, Fig. 3(ii → iv) (Left). These large ﬂuctuations, or coherent events, can be interpreted in terms of a multiplicative stochastic process with a weak additive noise background.34, 37 For a > 2.7 we observe that the bursts of the time series begin to saturate the bounds −1 ≤ r ≤ 1.
Processed_Scale-free_networks_in_complex_systems.txt	In Fig. 3 (Right) we plot the probability distribution functions (PDFs) associated with the time series of Fig. 3 (Left). The large ﬂuctuations, for a greater than ≈ 1, are reﬂected in the fat tails of the relative PDFs. Decreasing the value of a, and so the number of coherent events, the PDF converges to a Gaussian distribution generated by a random Poisson process.
Processed_Scale-free_networks_in_complex_systems.txt	In order to test the relevance of the network structure on the process of opinion formation, the previous simulations have been repeated, with a large number of nodes, N , and κ = a, for diﬀerent values of the clustering parameter, θ, and the node-edge parameter, m. While varying θ, does not lead to any substantial diﬀerence in the dynamics of the model, the increase of the average number of links per node, ¯k = 2m, has a dramatic large scale eﬀect in the turbulent-like phase, which deviations from a Gaussian regime increase dramatically: synchronizations are more likely to occur for large m. This behaviour is intrinsically related to the model of Eq. (5). In fact, the turbulent-like regime is a consequence of the random ﬂuctuations of the interaction strengths between agents around a bifurcation value separating the ordered and disordered phase.
Processed_Scale-free_networks_in_complex_systems.txt	It is also worth pointing out that an increase of ¯k is related to a decrease in the average path length between nodes; that is, the network “shrinks” and becomes more compact. In relation to our previous discussion, the more compact the network is the more the dynamics of our system approaches to the mean ﬁeld approximation. It becomes easier for the agents to synchronize. This characteristic of compactness, referred to as the small world eﬀect,19, 23, 24 is actually very common in both real and artiﬁcial networks.
Processed_Scale-free_networks_in_complex_systems.txt	These results conﬁrm that the critical topological characteristic leading to herding behaviour in the framework of stochastic opinion formation is the presence of mean ﬁeld eﬀects enhanced by small-world structure. The more information (links) that an agent has, the more likely it is for him/her to have an opinion in accord with other agents.
Processed_Scale-free_networks_in_complex_systems.txt	We now extend our model in order to include the concept of indecision. In practice a certain agent i, at a time step t, may take neither of the two possible decisions, σi = ±1, but remain in a neutral state. Keeping faith to the spirit of the model, we address this problem introducing an indecision probability, ǫ: that is the probability to ﬁnd, at each time step, a certain agent undecided. This is equivalent to introducing time dependent failures in the structure of the network by setting σ = 0.
Processed_Scale-free_networks_in_complex_systems.txt	Focusing on the turbulent-like regime, the shape of the PDF in the opinion ﬂuctuations changes according to diﬀerent concentrations of undecided persons. The results of the simulations, in Fig. 4 (Left), show how the dynamics of the model move from an intermittent state for ǫ = 0 toward a noise state for ǫ ≈ 0.6. The convergence to a Gaussian distribution is obtained only for quite high concentrations of undecided agents at about 60%. The robustness of the turbulent-like behaviour is related to the intrinsic robustness of SF networks against random failures.20–22 In fact, because there is a large absolute number of poorly connected nodes, related to the power law shape of P (k), the probability of setting one of them to inactive is much higher compared to the “hubs” that are relatively rare.
Processed_Scale-free_networks_in_complex_systems.txt	We can claim that, in large social networks governed by stochastic reactions in their elements, large ﬂuc- tuations in the average opinion can appear even in the case in which a large part of the network is actually “inactive” provided that the structure is scale free and the indecision is randomly distributed. The existence of large hubs provides for the survival of extended sub-networks in which synchronization can give rise to coherent events. The structure of the network itself supplies the random indecision.
Processed_Scale-free_networks_in_complex_systems.txt	Now we address the question of how the dynamics may change if we do not choose randomly the inactive nodes but we target the nodes having the most links. What we do in practice is to sort the nodes according to their number of links and then deactivate the nodes having the largest number of links in decreasing order. Fig. 4 (Right) illustrates how the fragmentation process is much faster and the noise regime is reached already when only the 10% of the hubs are deactivated. As emphasized in Ref.,20–22 the hubs have a great importance in the structural properties of SF networks and speciﬁcally targeting these nodes can lead to sudden isolation of a large fraction of the nodes of the network.
Processed_Scale-free_networks_in_complex_systems.txt	Figure 4. Left: Transition from coherent bahaviour, indecision probability ǫ = 0, to noise using a random selection for the inactive agents. For ǫ ≈ 0.6 we reach a noise-like behaviour. The parameters used in the simulation are: N = 104 nodes, θ = 0.9 for the clustering probability, m = m0 = 5 for the links of each new node, a = 1.8 and κ = a for the group and global opinion response respectively. Right: In this simulation we progressively turn oﬀ the largest hubs in the network. Once we have turned oﬀ about the 10% of agents, N = 104, the coherence in opinion formation disappears.
Processed_Scale-free_networks_in_complex_systems.txt	Motivated by recent empirical ﬁndings for complex systems interaction topologies, we have explored the implica- tions of a Barab´asi-Albert SFN topology in two diﬀerent models of complex systems. In particular we have found that the random frustration introduced by this topology of interactions induces a transition from a paramagnetic state to a spin glass state for an AF Ising model at a ﬁnite temperature. The critical spin glass phase transition temperature is estimated to be Tc ∼ 4.0(1). Such behaviour is not observed for the AF Ising model on regular lattices.
Processed_Scale-free_networks_in_complex_systems.txt	The SFN topology also has important consequences in our model for opinion formation. In this case, we discovered that the “hubs” of the network are more likely to synchronize due to mean ﬁeld eﬀects. Conversely, these eﬀects are not strong enough to synchronize the poorly connected nodes. Moreover, introducing inactive agents and spreading them randomly on the network, does not spoil the turbulent-like state, even for high concentrations of “gaps” up to approximately 60% of agents. This is a consequence of the implicit robustness of SF networks against random failures. If instead of selecting randomly the undecided individuals, we aim directly to the “hubs” of the network then the situation changes. In this case the network is disaggregate, composed of very small sub-networks and isolated nodes. Synchronization cannot signiﬁcantly eﬀect the resulting global opinion and the time series approximates Gaussian noise.
Processed_Scale-free_networks_in_complex_systems.txt	Part of the computation of this work has been done using super computer facilities of the South Australian Partnership for Advanced Computing (SAPAC).
Processed_Geometrically_Graded_h-p_Quadrature_Applied_to_the.txt	Boundary integral methods for the solution of boundary value PDEs are an alternative to ‘interior’ methods, such as ﬁnite diﬀerence and ﬁnite element methods. They are attractive on domains with corners, particularly when the solution has singularities at these corners. In these cases, interior methods can become excessively expensive, as they require a ﬁnely discretised 2D mesh in the vicinity of corners, whilst boundary integral methods typically require a mesh discretised in only one dimension, that of arc length.
Processed_Geometrically_Graded_h-p_Quadrature_Applied_to_the.txt	Consider the Dirichlet problem. Traditional boundary integral methods ap- plied to problems with corner singularities involve a (real) boundary integral equation with a kernel containing a logarithmic singularity. This is both te- dious to code and computationally ineﬃcient. The CBIEM is diﬀerent in that it involves a complex boundary integral equation with a smooth kernel. The boundary integral equation is approximated using a collocation technique, and the interior solution is then approximated using a discretisation of Cauchy’s integral formula, combined with singularity subtraction.
Processed_Geometrically_Graded_h-p_Quadrature_Applied_to_the.txt	A high order quadrature rule is required for the solution of the integral equation. Typical corner singularities are of square root type, and a ‘geometri- cally graded h-p’ composite quadrature rule is used. This yields eﬃcient, high order solution of the integral equation, and thence the Dirichlet problem.
Processed_Geometrically_Graded_h-p_Quadrature_Applied_to_the.txt	This report describes a research project carried out from March to October 1992, at the Department of Mathematics, The University of Queensland, Australia. It was carried out under the supervision of Dr Graeme A. Chandler, and was accredited as a #30 project, coded MN882.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.