source
stringlengths
1
2.05k
target
stringlengths
1
11.7k
Such accurate measurements are available for 1689. which we use to illustrate the method. and such data gaιο be obtainable for other clusters as well.
Such accurate measurements are available for A1689, which we use to illustrate the method, and such data should be obtainable for other clusters as well.
We note that this effective virial mass is defined from cdeprojecting 106 projected. mass assuming spherical svmumetrv. which is the closest. lensing observations can come to the standard theoretical definition. of the virial mass based. on a 3-D spherical average.
We note that this effective virial mass is defined from deprojecting the projected mass assuming spherical symmetry, which is the closest lensing observations can come to the standard theoretical definition of the virial mass based on a 3-D spherical average.
Lensing by a halo can be analvzed by calculating s= NNG.
Lensing by a halo can be analyzed by calculating $\kappa=\Sigma/\Sigma_{\rm cr}$ .
Dhe projected surface density is related to the three-imensional density p by an Abel integral transform.
The projected surface density is related to the three-dimensional density $\rho$ by an Abel integral transform.
This implies a relation between the integrated three-dimensional mass A(r) out to radius r and &8(£2) as a function of the projected radius 2: ο. }..0S)
This implies a relation between the integrated three-dimensional mass $M(r)$ out to radius $r$ and $\kappa(R)$ as a function of the projected radius $R$ : M(r) = dR , where f(x)= - .
The first term in equation (8)) is the total projected mass within a ring of projected radius r. and the second. term removes the contribution from mass elements ling at a 3-D radius greater than r.
The first term in equation \ref{eq:M}) ) is the total projected mass within a ring of projected radius $r$ , and the second term removes the contribution from mass elements lying at a 3-D radius greater than $r$ .
(1995): Girardi et al
; Girardi et al.
s analysis is based on digitized and calibrated survey plates from the COSMOS/UNST Southern Sky Object Catalog and is discrepant wilh Ixashikawa et al.
's analysis is based on digitized and calibrated survey plates from the COSMOS/UKST Southern Sky Object Catalog and is discrepant with Kashikawa et al.
Cluster mass-to-light determinations atf/f are scarce.
Cluster mass-to-light determinations at are scarce.
We thus transform to andV using the A—II colors above along wilh V—Lf~3 for earlv-tvpe galaxies and 1.37.
We thus transform to and using the $R-H$ colors above along with $V-H\sim3$ for early-type galaxies and $(V-H)_{\odot}=1.37$ .
The mass-to-light ratios within 1.5 7.! Mpe are then equivalent to: The caustic mass-to-light ratio is consistent wilh most previous measurements of cluster ALL valios: (vpical values using virial masses are M/Lj300hM.εἰ. (Carlbergetal. 2000).. with a range of 200—600M./L.j (Dressler1978).
The mass-to-light ratios within 1.5 $h^{-1}$ Mpc are then equivalent to: The caustic mass-to-light ratio is consistent with most previous measurements of cluster $M/L$ ratios; typical values using virial masses are $M/L_R\sim300\:h\:M_\odot/L_\odot$ \citep{car96,gir00}, with a range of $200-600\:h\:M_\odot/L_{\odot,R}$ \citep{dre78}.
. Our virial AL/L determination is consistent with the hieh end of this range.
Our virial $M/L$ determination is consistent with the high end of this range.
Estimates of M/L using X-ray masses tend to be somewhat lower.
Estimates of $M/L$ using X-ray masses tend to be somewhat lower.
Hradeckyetal.(2000). [id Αν.=154—468h solar units [or seven nearby Abell clusters and one group. andROSAT observations of eleven eroups and clusters vield AM/L,:=200—300hM./L. (David.Jones.&Forman1995).
\citet{hra00} find $M/L_V=154-468\:h$ solar units for seven nearby Abell clusters and one group, and observations of eleven groups and clusters yield $M/L_V=200-300\:h\:M_\odot/L_\odot$ \citep{dav95}.
. We discuss a large.5 deep imagingDAS survey of a 5galaxy cluster in the near-infrared. covering5 a 1.35 deg? (9 h? Mpc?) region of A1644.
We discuss a large, deep imaging survey of a galaxy cluster in the near-infrared, covering a 1.35 $^2$ (9 $h^{-2}$ $^2$ ) region of A1644.
These data allow us to determine the IR LF atIl to roughly Mj,+3. deeper than most previous infrared. surveys of comparable size.
These data allow us to determine the IR LF at to roughly $M^*_H+3$, deeper than most previous infrared surveys of comparable size.
We also acquired spectra for 155 galaxies Gnceluding 141 cluster members) aud use (lese {ο compute AM/Ly for the cluster.
We also acquired spectra for 155 galaxies (including $\sim141$ cluster members) and use these to compute $M/L_H$ for the cluster.
The results are:
The results are:
detectecd.
detected.
To evaluate the siguilicauce of our survey and provide some guidance for future work. we have aliavzed iu detail a single Moue Carlo simulation.
To evaluate the significance of our survey and provide some guidance for future work, we have analyzed in detail a single Monte Carlo simulation.
We chose the Cumingetal.(2008). best fit values of a—1.3]| aud 3=—0.61. with the semimajor axis truncation radius set to 100 AU.
We chose the \citet{cumming} best fit values of $\alpha = -1.31$ and $\beta = -0.61$, with the semimajor axis truncation radius set to 100 AU.
Planets could rauge iu mass [rom 1 το 20Mj.
Planets could range in mass from 1 to 20.
Às« lescribed in Section 2 above. we uormalized the planet clistributioUs so that eac vstar had a probability of having a planet with semünajor axis between 0.3 :uxd 2 AU anc Linass between 1 ancl 13Δρ.
As described in Section \ref{sec:rv} above, we normalized the planet distributions so that each star had a probability of having a planet with semimajor axis between 0.3 and 2.5 AU and mass between 1 and 13.
The simulation consisted of 50.000 realizatious of our suvey with hese parameteδ,
The simulation consisted of 50,000 realizations of our survey with these parameters.
Ina L. 505.881 planets were simulated. of which 51.879 were detectect.
In all, 505,884 planets were simulated, of which 51,879 were detected.
In of the 20.QOQ realizatious. our survey fouxl zero planets. while of the time it found one. aud of the time it found two or mor e.
In of the 50,000 realizations, our survey found zero planets, while of the time it found one, and of the time it found two or more.
Tie. planet distribution we considered iu this simulation cannot be ruled out by our survey. since a null] result such as we actually obtained turus out not to be very improbable.
The planet distribution we considered in this simulation cannot be ruled out by our survey, since a null result such as we actually obtained turns out not to be very improbable.
The Luge uumber of survey realizatious in our sliuulation allows the caleulation of precise statistics for poteutially detectable planets.
The large number of survey realizations in our simulation allows the calculation of precise statistics for potentially detectable planets.
The mecliat mass of detected planets in our simulation was 11.36Mjap.. the median semimajor axis was 13.9 AU. the median angular sej)aration was 2.56 aresec. auc the meclian significance was 21.[o.
The median mass of detected planets in our simulation was 11.36, the median semimajor axis was 43.5 AU, the median angular separation was 2.86 arcsec, and the median significance was $\sigma$.
This last number is interesting because it suggests that. for our survey. auy real planet detected was likely to appear at high significance. obvious evel ola preliminary. quick-look" reduction of the data.
This last number is interesting because it suggests that, for our survey, any real planet detected was likely to appear at high significance, obvious even on a preliminary, `quick-look' reduction of the data.
This suggests that performing such reductions at the telescope should be a high priority. to allow iniuediate coufirmation aud followup if a candidate is seen.
This suggests that performing such reductions at the telescope should be a high priority, to allow immediate confirmation and followup if a candidate is seen.
Figure 1. preseuts as a histogram the significauce of all planets detected in this Monte Carlo simulation.
Figure \ref{fig:sighist} presents as a histogram the significance of all planets detected in this Monte Carlo simulation.
We suspected that there would be a detection bias toward very eccentric planets. because these would speud most of their orbits uear apastrou. where they would be easier to detect.
We suspected that there would be a detection bias toward very eccentric planets, because these would spend most of their orbits near apastron, where they would be easier to detect.
This bias did not appear at auy measurable level in our simulation.
This bias did not appear at any measurable level in our simulation.
However. there was a weak but clear bias toward planets iu low-inclination orbits. which. of course. spend more of their time at large separatious from their stars than do planets with nearly. edege-ou orbits.
However, there was a weak but clear bias toward planets in low-inclination orbits, which, of course, spend more of their time at large separations from their stars than do planets with nearly edge-on orbits.
A concern with any planet imagine survey is how strougly the results hinge ou the best (i.e. nearest aud youngest€) few stars.
A concern with any planet imaging survey is how strongly the results hinge on the best (i.e. nearest and youngest) few stars.
A survey of 51 st:us may have far less statistical power than the number would im D “the best two or three stars |ad most of the probabilty of hosting detectable planets.
A survey of 54 stars may have far less statistical power than the number would imply if the best two or three stars had most of the probabilty of hosting detectable planets.
Table 3 elvesoO the percentage of planets cdeected around each star in our sample based ou our detailed Mone Carlo simulation.
Table \ref{tab:sdp} gives the percentage of planets detected around each star in our sample based on our detailed Monte Carlo simulation.
Due to poor data quality. binary orbit coustraints. or other issues. a few stars had zero probability of detected. planets giveu the distribution used here.
Due to poor data quality, binary orbit constraints, or other issues, a few stars had zero probability of detected planets given the distribution used here.
Iu eeneral. however. the likelihood of hostiug detectable plauets is fairly well clistributect.
In general, however, the likelihood of hosting detectable planets is fairly well distributed.
moving down along the diagonal brauch (not shown).
moving down along the diagonal branch (not shown).
The overall iuteusitv on this brauch is higher in the rise of an outburst (100.300. counts 1 + for Aql N-1 in Fiewre 2)) than in the decay (100. counts PpCU 3 or Aql N-l: see also Figure 3)).
The overall intensity on this branch is higher in the rise of an outburst (100–300 counts $^{-1}$ $^{-1}$ for Aql X-1 in Figure \ref{hid}) ) than in the decay $<$ 100 counts $^{-1}$ $^{-1}$ for Aql X-1; see also Figure \ref{cctime}) ).
Since the soft color increases With intensity along the top brauch (Figure 2)). he value of soft color at which the diagonal aud top xauches connect also could be smaller iu the rise of au outburst than in the decay. as may be seen by comparing he soft colors near the dates idicated by dashed lines in Fieure 3..
Since the soft color increases with intensity along the top branch (Figure \ref{hid}) ), the value of soft color at which the diagonal and top branches connect also could be smaller in the rise of an outburst than in the decay, as may be seen by comparing the soft colors near the dates indicated by dashed lines in Figure \ref{cctime}.
Similar hwsteresis has been observed frou ransieut LAINBs coutaiuiug black holes (Alivamotoctal. 1995).
Similar hysteresis has been observed from transient LMXBs containing black holes \citep{miy95}.
. The bottom portion of the color-color track from the atoll sources (hard color of 0.3) as traced frou left to vieht as the count rate creases once again by a factor of 10 (from το1500 counts 1 lin Figure 2)).
The bottom portion of the color-color track from the atoll sources (hard color of 0.3) is traced from left to right as the count rate increases once again by a factor of 10 (from 70–1500 counts $^{-1}$ $^{-1}$ in Figure \ref{hid}) ).
This auch is traced on time scales of davs to weeks in the ransicnt sources, while Ser N-1 has remained iu this soft state throughout the nnulssion.
This branch is traced on time scales of days to weeks in the transient sources, while Ser X-1 has remained in this soft state throughout the mission.
The color-color diagrams of Z sources are known to be raced smoothlv. without jumping between branches. but he color changes are not accompanied by large variations in the N-rav intensity (right panels of Figure 2)).
The color-color diagrams of Z sources are known to be traced smoothly, without jumping between branches, but the color changes are not accompanied by large variations in the X-ray intensity (right panels of Figure \ref{hid}) ).
The top xxrtion of the color-color track from GN 1 Chard color of 0. [in Figure 1)) is traced as the count rate mereases w (Figure 2)).
The top portion of the color-color track from GX $-$ 1 (hard color of 0.4 in Figure \ref{cconly}) ) is traced as the count rate increases by (Figure \ref{hid}) ).
Ou the diagonal portion of the color track of 6X5. 1. the count rate returns to its lowest values.
On the diagonal portion of the color-color track of GX $-$ 1, the count rate returns to its lowest values.
In atoll sources. the inteusitv is nearly constant.
In atoll sources, the intensity is nearly constant.
Ou the bottom of the color-color track. the count rate of GN L steadily increases vet again by.
On the bottom of the color-color track, the count rate of GX $-$ 1 steadily increases yet again by.
. The intensity from other Z sources changes by up to a factor of three ou this track ([email protected]).
The intensity from other Z sources changes by up to a factor of three on this track \citep[e.g.,][for GX 17$+$2]{hom01}.
. The hDarduessuteusitv diaerani of the atoll source Aql N-1iu Figure 2. also reveals important sub-structure ou short time scales that is conunon to all of the atoll sources in Table Ἐν,
The hardness-intensity diagram of the atoll source Aql X-1 in Figure \ref{hid} also reveals important sub-structure on short time scales that is common to all of the atoll sources in Table \ref{stats}.
This structure is most evideut in ιο plot of soft color agaiust count rate. where narrow. xmwallel tracks are traced by observations within single lays.
This structure is most evident in the plot of soft color against count rate, where narrow, parallel tracks are traced by observations within single days.
At low intensities (less than 200 counts !À 1). 1ο soft color varies as the count rate remains ucarly constant. while at the highest inteusitics. the soft color and count rate both increase significantly together,
At low intensities (less than 200 counts $^{-1}$ $^{-1}$ ), the soft color varies as the count rate remains nearly constant, while at the highest intensities, the soft color and count rate both increase significantly together.
Ou 1e color-color diagram in Figure 1.. these daily tracks are ecnorally aligned perpendicular to motion along the Z. aud ws broaden the color-color tracks of the atoll sources.
On the color-color diagram in Figure \ref{cconly}, these daily tracks are generally aligned perpendicular to motion along the Z, and thus broaden the color-color tracks of the atoll sources.
Iu contrast. the parallel tracks from the Z sources are forme. when the cutive Z shifts (vauderIlis.2000).
In contrast, the parallel tracks from the Z sources are formed when the entire Z shifts \citep{vdk00}.
. We have cemonstrated that the color-color diagrams of both Z and atoll sources have similar three-branched shapes (Figures 1)). but that the rauge of X-ray intensity and the time scale over which the diagrams are traced are one to two orders of magnitude larger iu the atoll sources (Figures 2. and 3)).
We have demonstrated that the color-color diagrams of both Z and atoll sources have similar three-branched shapes (Figures \ref{cconly}) ), but that the range of X-ray intensity and the time scale over which the diagrams are traced are one to two orders of magnitude larger in the atoll sources (Figures \ref{hid} and \ref{cctime}) ).
We note that simular conclusions have been reported independently by Ciüerliüski&Done (20023..
We note that similar conclusions have been reported independently by \citet{gd02}.
There are also significant spectral differences between the two classes of source (Schulzctal.1989:Christian&Swauk1997:Barretoetal. 2000).
There are also significant spectral differences between the two classes of source \citep{sht89,cs97,bar00}.
. The spectra of Z sources are always soft. aud cau be described by the sum of a cool (1 keV) black body aud Comptonized enission fom warm (5 keV) optically thick electrous (6.8.Claistian&Swank1997:DiSalvoetal.2000).
The spectra of Z sources are always soft, and can be described by the sum of a cool (1 keV) black body and Comptonized emission from warm (5 keV) optically thick electrons \citep[e.g.,][]{cs97, dis00}.
. Spectral changes along the Z are quite subtle (e.g...Schulzetal.1989)..
Spectral changes along the Z are quite subtle \citep[e.g.,][]{sht89}. .
In coutrast. changes in the energy spectra of atoll sources are dramatic.
In contrast, changes in the energy spectra of atoll sources are dramatic.
While soft energy spectra resenibliug those of Z sources are characteristic of atoll sources at the bottom of their color-color diagrams (e.g.Oosterbroekeal.2001.forSerN-1).. a hard state with a Pz1.8 power euergv spectrum between 2100 keV occurs at the top of this diaerani when the sources are faint (0.8.Barreteal.2000.forGS1826.238)..
While soft energy spectra resembling those of Z sources are characteristic of atoll sources at the bottom of their color-color diagrams \citep[e.g.,][for Ser~X-1]{oos01}, a hard state with a $\Gamma \approx 1.8$ power-law energy spectrum between 2–100 keV occurs at the top of this diagram when the sources are faint \citep[e.g.,][for GS~1826$-$238]{bar00}.
Such a hard spectrin is no observed from Z sources. probably because they are no observed at low Iuuinosities (Table 1. aud Figure 2)).
Such a hard spectrum is not observed from Z sources, probably because they are not observed at low luminosities (Table \ref{stats} and Figure \ref{hid}) ).
Recent work has suggested that the timing properties of Z and atoll sources also may have iuterestig similarities.
Recent work has suggested that the timing properties of Z and atoll sources also may have interesting similarities.
It has lone been suggested that the broad-band noise iu power spectra of Z aud atoll sources can probably be deseribed by simular components (vanderIls1995).. and detailed studies of ddata have shown that the frequencies of QPOs and baud-liuuted noise exhibit similar correlations iu both types of sources (Wijuands&vanderElis1999:Psaltisetal. 1999).
It has long been suggested that the broad-band noise in power spectra of Z and atoll sources can probably be described by similar components \citep{vdk95}, and detailed studies of data have shown that the frequencies of QPOs and band-limited noise exhibit similar correlations in both types of sources \citep{wk99,pbk99}.
. ILoxcever. whether the timing properties are correlated with the brauches of the color-color diagrams in a simular miauner in cach class of source remains to be investigated.
However, whether the timing properties are correlated with the branches of the color-color diagrams in a similar manner in each class of source remains to be investigated.
The timing properties of neutron star LAINBs eventually may allow us to understand what drives the spectral variability.
The timing properties of neutron star LMXBs eventually may allow us to understand what drives the spectral variability.
Several studies have found that the kz OPO frequencies ave svelb-correlated. with position ou the color-color diagram in both Z (οι,Tomanet and atoll (e...vauStraateuetal.2001) sources,
Several studies have found that the kHz QPO frequencies are well-correlated with position on the color-color diagram in both Z \citep[e.g.,][]{hom01} and atoll \citep[e.g.,][]{vst01b} sources.
Examine Figure 2.. the short-term correlations between the colors and PCA count rate are extremely simular to the "parallel tracks” observed im comparisous of the frequencies of kilohertz quasi-periodic oscillations Oz OPO) with the PCA count rate (vanderlis2000. 2001).
Examining Figure \ref{hid}, the short-term correlations between the colors and PCA count rate are extremely similar to the “parallel tracks” observed in comparisons of the frequencies of kilohertz quasi-periodic oscillations (kHz QPO) with the PCA count rate \citep{vdk00,vdk01}.
. The parallel tracks are observed from kz QPOs both in stucies of individual sources (Méndezetal.1999)... and in comparing the frequencies of QPOs from Z aud atoll sources that span a wide ranee of luuinosity etal. 2000).
The parallel tracks are observed from kHz QPOs both in studies of individual sources \citep{men99}, and in comparing the frequencies of QPOs from Z and atoll sources that span a wide range of luminosity \citep{for00}.
. Based upon these parallel tracks. vauderIlis(2001) has sugeested that the kIIz QPO frequencies are determined by a feedback process that ix seusitive to deviations in the accretion rate about its value averaged over a few davs.
Based upon these parallel tracks, \citet{vdk01} has suggested that the kHz QPO frequencies are determined by a feedback process that is sensitive to deviations in the accretion rate about its value averaged over a few days.
Further work is needed to determine whether the same feedback mechauisui can operate to produce spectral variations in both Z aud atoll sources. ou widely different time scales and over ranges in Μποςτν that differ bv factors of 100.
Further work is needed to determine whether the same feedback mechanism can operate to produce spectral variations in both Z and atoll sources, on widely different time scales and over ranges in luminosity that differ by factors of 100.
We are erateful for questions and coments frou Michiel van der Klis and Exik EKuulkers that helped in clavitvine these results.
We are grateful for questions and comments from Michiel van der Klis and Erik Kuulkers that helped in clarifying these results.
This work was supported bv NASA. under contract NAS 5-30612aud eraut. NAC 5- aud has made use of data obtained from the Wiel Euergv Astroplivsics Science Archive Research Ceuter (IEASARC). provided by NASA’s Goddard Space Flight Center
This work was supported by NASA, under contract NAS 5-30612and grant NAG 5-9184, and has made use of data obtained from the High Energy Astrophysics Science Archive Research Center (HEASARC), provided by NASA's Goddard Space Flight Center
The rate at which star formation occurs is still an area of ongoing debate.
The rate at which star formation occurs is still an area of ongoing debate.
The two competing paradigms of rapid star formation on dynamical timescales (eg22). and slow star-formation on timescales of several Myr (e.g2).. are difficult to choose between on observational grounds.
The two competing paradigms of rapid star formation on dynamical timescales \citep[e.g][]{elmegreen00,hartmann01}, and slow star-formation on timescales of several Myr \citep[e.g][]{shu87}, are difficult to choose between on observational grounds.
One argument in favour of slow star-formation (SSF) has been the observation of apparent age spreads in young clusters and associations (e.g2222)..
One argument in favour of slow star-formation (SSF) has been the observation of apparent age spreads in young clusters and associations \citep[e.g][]{herbst82,sung98,pozzo03,dolan01}.
These apparent age spreads manifest themselves as a spread in the location of young stars within a Hertzsprung-Russell (H-R) or a colour-magnitude diagram (CMD): at a given luminosity stars with a range of effective temperatures are observed. implying a range of radii or a spread in ages.
These apparent age spreads manifest themselves as a spread in the location of young stars within a Hertzsprung-Russell (H-R) or a colour-magnitude diagram (CMD); at a given luminosity stars with a range of effective temperatures are observed, implying a range of radii or a spread in ages.
? have used ages obtained from the H-R diagram to argue that star formation takes place over ~10MMyr and accelerates towards the present day.
\cite{palla00} have used ages obtained from the H-R diagram to argue that star formation takes place over $\sim$ Myr and accelerates towards the present day.
However. the reality of these apparent age spreads has been questioned by various authors.
However, the reality of these apparent age spreads has been questioned by various authors.
9 suggested a number of factors which could produce a spread in the CMD or H-R diagram without requiring a genuine spread in ages.
\cite{hartmann01} suggested a number of factors which could produce a spread in the CMD or H-R diagram without requiring a genuine spread in ages.
These included photometric variability. inadequate correction for variable extinction and the presence of unresolved binaries.
These included photometric variability, inadequate correction for variable extinction and the presence of unresolved binaries.
Since then. ? have shown that photometric variability cannot explain the observed spreads: nor ean carefully. controlling. for the effects of binaritv. variable reddening and contaminating light from the accretion dise explain the observed spreads in the Orion Nebular Cluster (2) or LH 95 in the Large Magellanic Cloud (2)..
Since then, \cite{burningham05} have shown that photometric variability cannot explain the observed spreads; nor can carefully controlling for the effects of binarity, variable reddening and contaminating light from the accretion disc explain the observed spreads in the Orion Nebular Cluster \citep{dario10a} or LH 95 in the Large Magellanic Cloud \citep{dario10b}.
In addition ? used a novel geometrical technique to show that the apparent age spreads in the H-R diagram of the ONC are associatec with a genuine spread in stellar radii.
In addition \cite{jeffries07} used a novel geometrical technique to show that the apparent age spreads in the H-R diagram of the ONC are associated with a genuine spread in stellar radii.
These recent results have added support to the idea that the luminosity spreads observed in young clusters and associations are a genuine phenomenon. anc not an observational artifact.
These recent results have added support to the idea that the luminosity spreads observed in young clusters and associations are a genuine phenomenon, and not an observational artifact.
However. as accretion can affect the evolution of the central star. inducing a spread of luminosities and radii in a population which is co-eval. such spreads may not be attributable to a spread in ages.
However, as accretion can affect the evolution of the central star, inducing a spread of luminosities and radii in a population which is co-eval, such spreads may not be attributable to a spread in ages.
Most studies agree that curren accretion rates typically observed in pre-main-sequence objects
Most studies agree that current accretion rates typically observed in pre-main-sequence objects
the internal magnetic field eventually leads to sudden relaxation of the twist in the magnetospheric field, releasing the energy necessary to power the flare.
the internal magnetic field eventually leads to sudden relaxation of the twist in the magnetospheric field, releasing the energy necessary to power the flare.
The initial configuration and subsequent evolution of magnetic fields in highly-magnetized neutron stars is a complicated problem.
The initial configuration and subsequent evolution of magnetic fields in highly-magnetized neutron stars is a complicated problem.
The field evolves continuously due to the effects of ohmic decay, ambipolar diffusion, and Hall drift.
The field evolves continuously due to the effects of ohmic decay, ambipolar diffusion, and Hall drift.
Recently, ? studied the evolution of magnetic fields in neutron star crusts, emphasizing the importance of Hall drift.
Recently, \citet{ponsgeppert} studied the evolution of magnetic fields in neutron star crusts, emphasizing the importance of Hall drift.
Their results indicate that Hall drift of crustal fields can create small-scale magnetic field structures, and that those structures can drift to regions of higher resistivity.
Their results indicate that Hall drift of crustal fields can create small-scale magnetic field structures, and that those structures can drift to regions of higher resistivity.
The simulations of ? were restricted to magnetic fields in the inner crust.
The simulations of \cite{ponsgeppert} were restricted to magnetic fields in the inner crust.
In this paper, we focus on the outer crust, and show that large currents can lead to a thermo-resistive instability, affecting the thermal evolution of the star.
In this paper, we focus on the outer crust, and show that large currents can lead to a thermo-resistive instability, affecting the thermal evolution of the star.
As the instability evolves, large portions of the crust may melt, allowing the magnetic field to evolve on a timescale much faster than the average ohmic decay and Hall timescales.
As the instability evolves, large portions of the crust may melt, allowing the magnetic field to evolve on a timescale much faster than the average ohmic decay and Hall timescales.
The enhanced magnetic evolution resulting from instability may be related to flare activity in magnetars.
The enhanced magnetic evolution resulting from instability may be related to flare activity in magnetars.
'This paper is organized as follows.
This paper is organized as follows.
In section 2, we describe the relevant physics of the thermo-resistive instability in neutron star crusts.
In section 2, we describe the relevant physics of the thermo-resistive instability in neutron star crusts.