A STATE CHANGE IN THE MISSING LINK BINARY PULSAR SYSTEM PSR J1023+0038

The pulsar "recycling" scenario (Alpar et al. 1982; Radhakrishnan & Srinivasan 1981) proposes that the rapid rotation rates of millisecond pulsars (MSPs¹⁰) originate from a period of mass accretion from a binary companion in a low-mass X-ray binary (LMXB). The idea that such systems could also result in the complete ablation of the companion star and produce isolated MSPs—like the first MSP discovered, PSR B1937+21 (Backer et al. 1982)—seemed to be confirmed by the discovery of the first eclipsing "black widow" pulsar PSR B1957+20 (Fruchter et al. 1990). The eclipses of the radio pulsations in this system showed that mass was being ablated from the very low mass (M_c ≪ 0.1 M_☉) companion (Ryba & Taylor 1991), however the mass loss rate appeared to be insufficient to ever completely destroy the companion. The LMXB and ablation phases were expected to be short-lived and so the subsequent discovery of another black widow, PSR J2051−0827 (Stappers et al. 1996), seemed to challenge the connection between these systems and the isolated MSPs. Nonetheless, clear evidence that MSPs are indeed spun-up in LMXBs came with the discovery of coherent X-ray pulsations at a period of 2.5 ms from the LMXB SAX J1808.4−3658 during an X-ray outburst (Wijnands & van der Klis 1998). To date, there are some 15 of these accreting X-ray millisecond pulsars (AMXPs) known (Patruno & Watts 2012).

With the rapid rate of MSP discoveries in the last few years (Ransom 2013, see his Figure 1), the diversity of MSP binary companions has continued to grow beyond the standard low-mass white dwarf or ultra-low-mass black widow companions that were initially found. This variety has shown that the pulsar recycling process produces a rich array of systems, some of which are exceedingly exotic (e.g., Bailes et al. 2011; Ransom et al. 2014). Furthermore, some of these new systems are excellent case studies for better understanding the accretion-induced spin-up of neutron stars.

Recently, a completely different class of eclipsing binary MSP systems, dubbed "redbacks," with more massive (M_c ≳ 0.1 M_☉) and likely non-degenerate companions have been discovered in the Galactic field (Roberts 2011 and references therein; such systems are also known in globular clusters (GCs) and were discovered earlier, e.g., D'Amico et al. 2001). Targeted searches for radio pulsars in unassociated Fermi γ-ray sources have been particularly fruitful, producing the majority of these discoveries (e.g., Ray et al. 2012; Hessels et al. 2011). Such searches have also found many more examples of the previously known black widow systems. It has also been shown that during the accretion stage the companion star can be ablated down to the mass of a planet (Bailes et al. 2011; van Haaften et al. 2012) while at the same time (barely) surviving complete destruction via ablation. This again suggests that in some cases the companion does not survive and an isolated MSP is left behind.

With the wide variety of MSP systems now known, it is necessary to better understand the various potential accretion and ablation processes in LMXBs and eclipsing binary MSPs. While most MSP systems are clearly well past any accretion episode, the redbacks and black widows appear to provide the closest available link to the LMXBs. There are now two exceptional systems, the Galactic field binary pulsar J1023+0038 (hereafter J1023; Archibald et al. 2009) and PSR J1824−2452I/IGR J18245−2452 (hereafter M28I; Papitto et al. 2013), located in the GC M28. Both are providing us with a detailed look at the transition between these two phases. In fact, the behavior of these two systems in the last decade has made it abundantly clear that the transition between an LMXB and an MSP is not a sudden or unidirectional transformation.

J1023 was discovered in 2007 as a 1.7 ms radio pulsar in a 4.8 hr, circular-orbit eclipsing binary system. It was soon recognized to be the same source as FIRST J102347.6+003841, which had originally been identified as a magnetic cataclysmic variable (Bond et al. 2002) but was subsequently identified as possibly being an LMXB with an accretion disk during 2001 (Thorstensen & Armstrong 2005). Later optical and X-ray observations indicated that the source no longer possessed an accretion disk (Woudt et al. 2004; Homer et al. 2006; Archibald et al. 2009, 2010; Bogdanov et al. 2011). It was therefore identified as the first object to have been seen to transition from an LMXB-like state to an MSP. Archibald et al. (2013) also show that since its discovery as a pulsar this "original Galactic field redback" has exhibited radio behavior typical of this class of eclipsing binary MSPs. During this phase, X-rays are produced by the system (Archibald et al. 2010; Bogdanov et al. 2011) but they originate from a combination of pulsed magnetospheric emission and an intra-binary shock between the companion and MSP winds.

M28I links LMXBs and MSPs in a complementary, and even more direct, way because it has been seen to transition from a radio emitting MSP to an AMXP and back again: i.e., it has shown accretion-powered pulsations at the same rotational period as the previously known radio pulsar (Riggio et al. 2013; Papitto et al. 2013). Moreover, it was suggested that this object has been seen to swing between these states several times over the past decade.

Here we report a sudden change of state in J1023, starting in 2013 June. This change was heralded by the cessation of detectable pulsed radio emission from the MSP and coincides with a dramatic, five-fold increase in the γ-ray flux from the system. Along with the reported changes in the X-ray behavior (Patruno et al. 2014) and the emergence of double-peaked optical spectral lines (Halpern et al. 2013), this points to an accretion disk having re-formed in the system but with a still-active pulsar mechanism also present.

2.1. Radio Observations

2.2. Gamma-Ray Observations

3.1. Radio Results

A summary of the radio observations made with the LT and WSRT since 2009 is presented in Figure 1, where circles represent clear detections of pulsed emission and all other symbols indicate non-detections. This plot shows clearly the effect of the eclipses around orbital phase 0.25, when the pulsar is behind its companion. The eclipse duration approaches 50% at low radio frequencies (red symbols); see Archibald et al. (2013) for a detailed discussion. Variability away from eclipse in the L-band (1380 MHz and 1500 MHz) observations (black symbols) is caused by strong scintillation in the interstellar medium, shown to have a bandwidth of about 100 MHz (Archibald et al. 2013) at this frequency, and as expected for a pulsar with a dispersion measure of only 14.3 pc cm⁻³. The 350 MHz observations have a sufficiently large fractional bandwidth that they are unaffected by scintillation, but the eclipses are longer and there are also short-duration, up to a few minutes long, eclipses at other orbital phases. A part from this variability, the most obvious feature in Figure 1 is that there has been no detection of pulsed radio emission from the pulsar since mid 2013 June. It may be that J1023 is still an un-pulsed radio source but the observations reported here are unable to determine if that is the case.¹⁵

Zoom In Zoom Out Reset image size

Inspection of the regular LT timing observations in mid July revealed that pulsations had not been detected since 2013 May 31 at 1500 MHz. A check of the WSRT 350 MHz and 1380 MHz observations showed that the source was last detected on 2013 June 15. Increasing the cadence and duration of observations quickly revealed that the pulsed radio emission was no longer seen at any of these frequencies. Using the telescope and backend parameters given in Table 1 we derive 10σ flux density upper limits of 0.8 and 0.06 mJy for WSRT 350 MHz and LT 1500 MHz observations, respectively. These limits are more than an order of magnitude lower than the typical flux densities of 16 and 0.9 mJy measured between 2009 April and 2013 June. These limits rule out that the sudden disappearance is due to scintillation.

As shown in Archibald et al. (2009, 2013), the eclipse durations are strongly frequency dependent (as well as varying somewhat between orbits). This motivated the higher-frequency GBT and AO observations (the triangle and square symbols in Figure 1, respectively). A full orbit was sampled with the GBT observations and the AO observations spanned orbital phases of 0.5–0.75, away from the low-frequency eclipse region. No pulsed radio emission was detected down to 10σ flux density limits of 0.016 and 0.003 mJy at 2 GHz and 4.5 GHz, respectively. These non-detections confirm that the J1023 system changed to a different state sometime between 2013 June 15 and 2013 June 30. The lack of observable radio pulsations indicates that either the radio pulsar mechanism has been extinguished or the radio pulsations are undetectable because they are scattered in time or completely absorbed by an increase of intra-binary material. This state has persisted until at least 2014 March.

3.2. Gamma-Ray Results

Figure 2 shows that the γ-ray flux has increased dramatically since the radio disappearance. This conclusion is consistently reached using both of the analysis methods described in Section 2. Specifically, the average pre-disappearance γ-ray flux¹⁶ obtained by spectral fitting is 1.05(11) × 10⁻⁸ cm⁻² s⁻¹, while the average post-disappearance flux is 6.3(6) × 10⁻⁸ cm⁻² s⁻¹. In comparison, the fluxes obtained from aperture photometry are lower (because of the restriction to 1–300 GeV photons and because not all >1 GeV photons fall within our aperture) and the background is higher (because the simple aperture cut allows more photons from other sources) but the average flux before the disappearance was 2.22(14) × 10⁻⁹ cm⁻² s⁻¹, while the post-disappearance average flux is 6.9(9) × 10⁻⁹ cm⁻² s⁻¹.

Zoom In Zoom Out Reset image size

To better determine when the γ-ray flux started increasing, we used our aperture-photometry data and combined 2.5 Ms chunks starting every 500 ks. The results are plotted in the lower panel of Figure 1. This procedure unavoidably smoothes all features by 2.5 Ms (about 1 month). Nonetheless, it is clear that the γ-ray flux began rising at approximately the same time that the radio pulsations disappeared. It is unclear, however, whether the flux rose abruptly and then remained constant or whether it rose gradually and is possibly continuing to rise.

To look for spectral changes in a nearly model-independent way, we computed "hardness ratios" before and after the radio disappearance. Examination of the aperture photometry photons revealed that about half of them were above 1.6 GeV, so we simply performed aperture photometry as described above but with separate energy ranges of 1–1.6 GeV and 1.6–300 GeV. A "hardness ratio" would then be a ratio of (exposure-corrected) count rates in these two energy ranges. The hardness ratio before the radio disappearance was 1.10(12), while after the disappearance it was 1.4(3). We do not claim any significant change in hardness. Because of the low number of photons, we were not very sensitive to spectral changes.

We investigated orbital modulation of the γ-ray emission by using tempo2 (Hobbs et al. 2006) with the fermi plugin to assign the corresponding orbital phase to each probability-tagged photon. We used the weighted H test (Kerr 2011) to look for periodic modulation, obtaining a false positive probability of 0.66—that is, for purely random phases there is a 66% probability of having greater deviation from uniformity than we detect. Testing with a simulated signal in which all source photons are distributed according to a sine wave produced a median false positive probability of 0.08. In other words, even with this very strong modulation, we would have had only a 50% chance of obtaining a detection better than 1.4σ. So although we have tested for orbital modulation, our non-detection rules out only very sharply peaked orbital modulation. We investigated modulation at the pulsar period using tempo2 to assign a pulse phase to each photon, then testing for periodicity as above. The false positive probability we obtained was 0.997. This method has the same lack of sensitivity as our search for orbital pulsations, but it also has another limitation: because our orbital ephemeris for J1023 was last updated before its disappearance in radio, and because J1023 undergoes apparently random orbital period variations (Archibald et al. 2010, 2013), we should expect phase prediction errors as large as ∼0.2 turns to smear out any high harmonic content. It is therefore difficult to rule out even very sharply peaked pulsations.

We report that, sometime between 2013 June 15 and June 30, pulsations from the MSP in the J1023 binary system disappeared at radio frequencies where they are otherwise easily detectable. Concurrent with this disappearance we find that the γ-ray flux of J1023 has increased five-fold. Subsequent optical, UV and X-ray observations have shown that an accretion disk has formed in the binary system (Halpern et al. 2013; Patruno et al. 2014), signifying that J1023 has undergone a transition from a binary MSP to a state that is not accretion driven like an LMXB, but has properties similar to those seen in a state referred to as a quiescent low-mass X-ray binary (qLMXB). The present transition appears similar to that of 2000/2001; the archival study by Archibald et al. (2009) shows the system transitioning from a radio point source and G-type optical colors to having an accretion disk.

We are now fortunate to observe two redback systems that have transitioned multiple times between LXMB and MSP states. Although its 2013 outburst is a uniquely observed event for a known MSP, as a radio pulsar M28I is fairly typical in a population of about 11 known GC redbacks.¹⁷ This suggests that other similar pulsars in the Galactic GC system will also undergo such events in the coming years and decades. These sources remain challenging to study at optical and radio wavelengths, however, because of the relatively large distance to the most massive GCs (∼5–10 kpc). Also, because stellar interactions are common in GCs the evolution of these systems will be modified compared to those in the field and in some cases exchange interactions will result in some GC redbacks hosting neutron stars no longer being orbited by their original binary companions. In contrast, the known field redbacks, of which J1023 was the first discovered, are typically significantly less distant (<2 kpc) and offer better potential for detailed multi-wavelength study. The GC and field redbacks both provide us interesting test beds for studying the accretion process and give us different views on how these systems can form and evolve.

There are now at least 7 redbacks and 17 black widow systems known in the field, thanks in large part to targeted searches of unassociated Fermi sources (Ray et al. 2012). Again, as a radio pulsar, J1023 is no longer very atypical, with several similar examples, such as PSR J2215+5135 (Hessels et al. 2011), which has nearly identical orbital parameters and a likely similar, Roche-lobe-filling companion (Breton et al. 2013). Here as well, given proper radio, optical, and X-ray monitoring, it is expected that it may not be long before these similar systems show transitions like those we have observed from J1023. If, not, then the distinction of J1023 needs to be considered more carefully.

Though J1023 and M28I show many striking similarities in their X-ray behavior (Patruno et al. 2014), the former has yet to enter a truly energetic outburst state in which the accretion flow is reaching the neutron star surface. Indeed, the 2013 June to October activity of J1023 is arguably very similar to its past 2001 state (though no targeted X-ray data was available at that time). An energetic outburst like that seen in M28I (Papitto et al. 2013) can be ruled out for J1023 since 1996 because it would have triggered an all-sky X-ray monitor. M28I has also been observed for ∼200 ks by Chandra in 2008 (Linares et al. 2013), switching several times between a bright "active" (4 × 10³³ erg s⁻¹) and a faint "passive" state (6 × 10³² erg s⁻¹). Therefore M28I can undergo rapid X-ray flux variations similar to those currently seen in J1023.¹⁸ If the X-ray flux variations observed in M28I in quiescence (i.e., the 2008 Chandra data, for which no simultaneous radio observation is available) can be ascribed to the presence of an accretion disk, then there is no specific reason to presume that these two systems are fundamentally different.

Shvartsman (1970) proposed that an active radio pulsar in a contact binary is able to prevent the formation of an accretion disk. However, M28I has shown beyond any doubt a luminous X-ray outburst in 2013 April. In all likelihood J1023 is also capable of undergoing a full accretion episode, although the recurrence time could be on the order of decades. Burderi et al. (2001) proposed that an LMXB can turn into a radio pulsar either cyclically or indefinitely depending on the orbital period of the system. According to Burderi's definition, J1023 falls among the sources that cyclically alternate between an outburst and a radio pulsar state. Ekşİ & Alpar (2005) showed that an active pulsar does not necessarily imply the complete destruction of the accretion disk if the inner disk is truncated at two to three light cylinder radii. It is possible therefore that J1023 is an active pulsar surrounded by an accretion disk truncated outside the light cylinder. We suggest that both J1023 and M28I can probably exhibit three states characterized by (1) an observable radio MSP and a low X-ray luminosity which is rotation powered, (2) the presence of an accretion disk in a qLMXB-like system, and (3) an AMXP, with high X-ray luminosity, powered by accretion, and potentially Type I X-ray bursts.

Regardless of the fact that the radio pulsar is currently not detectable, it appears most likely that J1023's rotation-powered pulsar mechanism, most importantly the relativistic particle wind that is created, is still active. Before its recent state change, the γ-rays from the system were quite likely of magnetospheric origin, in analogy with other similar MSPs detected by Fermi LAT. As suggested by Tam et al. (2010), a portion of the γ-rays could be from an intra-binary shock between the pulsar and its companion star. However, for the γ-ray data preceding 2013 June, Archibald et al. (2013) show that, while there is a marginally significant detection of pulsations from J1023, there is no evidence for orbital modulation.

Searches for orbital-phase-dependent γ-ray variability in other similar systems have also been unsuccessful, again arguing that the observed emission from MSPs is predominantly magnetospheric when the source is in the radio-loud state. Notably, other similar γ-ray MSPs are stable in flux, as are (almost) all young pulsars detected in γ-rays (Abdo et al. 2013). Recently the identification of γ-ray flux variability from PSR J2021+4026 (Allafort et al. 2013) showed that some γ-ray pulsars can undergo mode switches, reminiscent of recent discoveries in the radio (Kramer et al. 2006; Lyne et al. 2010) and Hermsen et al. (2013) showed that for PSR B0943+10 there is a correlated mode switch in both radio and X-rays, providing strong evidence that a sudden and global magnetospheric state change can occur. There is as yet however no evidence that such magnetospheric switching occurs in MSPs. Thus it is highly unlikely, given the strong correlation with the other changes in the properties of the system, including the presence of an accretion disk, that the five-fold increase of the γ-ray flux in the current state can be of magnetospheric origin or precipitated by a change associated with the pulsar itself.

In light of this, it appears more likely that the source of the increased γ-ray flux must be an increased shock within the system. The striking correlation with the timing of the radio shut-off indicates that this rise in γ-rays must be closely related to the reappearance of an accretion disk in the system (Patruno et al. 2014), which is likely the principal driving force behind the observed changes in the radio, optical, X-ray, and γ-ray emission. We suggest that to produce the excess γ-rays requires both an active pulsar wind and an accretion disk, but that there can be no active accretion onto the neutron star, which would quench the pulsar mechanism and hence the relativistic wind.

If the rotation-powered pulsar is still active, enshrouding of the system by intra-binary material, causing severe scattering and/or absorption, is the logical explanation for the absence of radio pulsations. This would make J1023 a so-called hidden radio MSP, as hypothesized by Tavani (1991). In this scenario, the MSP is completely enshrouded in the prodigious outflow from its close companion star, rendering the pulsar perpetually eclipsed at radio frequencies (see, e.g., Thompson et al. 1994; Iacolina et al. 2009 for possible mechanisms). At the same time, the system generates strong un-pulsed shock emission in X-rays and γ-rays, due to the interaction of the pulsar wind with the outflow of matter (Tavani 1993). Other recently discovered eclipsing MSPs also exhibit compelling evidence for enshrouding (Romani & Shaw 2011; Ray et al. 2013), though J1023 is the first instance where we see a sudden increase in the high-energy emission, correlated with a radio disappearance. Together, these systems support previous arguments that a non-negligible fraction of MSPs may often be non-detectable at radio wavelengths because of enshrouding. There is also some evidence that this becomes more problematic for the fastest-spinning MSPs (Burderi et al. 2001; Hessels et al. 2006, 2008).¹⁹

Spin-down luminosity is a sensitive function of spin period ($\dot{\rm E} \propto \dot{\rm P}/\rm P^3$) and hence this may explain this trend, all other factors, like orbital separation and companion type, being equal. Indeed this could be one reason for the absence of a discovery of a sub-millisecond pulsar thus far, in spite of reasonable sensitivity to such sources in modern large-scale pulsar surveys, such as the PALFA survey (Cordes et al. 2006), the HTRU survey (Keith et al. 2010) and the aforementioned LAT-directed searches. On the other hand, the existence of many such "hidden" fast radio MSPs is problematic in light of the absence of detections of accreting MSPs with frequencies higher than 619 Hz, in spite of there being no obvious selection effects against finding them with X-ray telescopes (Chakrabarty et al. 2003). J1023 is a 1.7 ms pulsar, the fifth fastest spinner known in the Galactic field. Of the five fastest-spinning MSPs in the field, three are known to eclipse and the other two are isolated.

With the re-appearance of an accretion disk, J1023 also resembles a number of qLMXB systems containing AMXPs. It has been suggested that some AMXPs reactivate as rotation-powered pulsars in quiescence (Stella et al. 1994; Campana et al. 1998; Burderi et al. 2003; Campana et al. 2004). Apart from M28I, no radio pulsations have been detected from any AMXP in quiescence, although this could be a consequence of enshrouding as well (Burgay et al. 2003; di Salvo et al. 2008). In addition, none has yet been detected as γ-ray source with Fermi LAT (Xing & Wang 2013), although they are all significantly more distant than J1023. It is thus plausible that quiescent AMXPs may also be "hidden" MSPs.

Given the nature of the enhanced γ-ray emission that has appeared from J1023, one may consider what similarities exist with the small population of γ-ray binaries, where a compact object is orbited by a massive OB star (e.g., Dubus 2013). Still the best understood high-mass gamma-ray binary is PSR B1259−63, which contains a relatively rapidly rotating (P = 47 ms) young radio pulsar; its pulsar has $\dot{E}$ ∼ 20 times higher than that of PSR J1023+0038 yet has shown γ-rays ∼200 times brighter, achieving near 100% efficiency in converting spin-down power to γ-rays near periastron, when the pulsar's wind is shocked closest to the pulsar. Though the orbital compactness and companion mass in the case of J1023 is quite different when compared with such systems, there may be similarities in some of the physical mechanisms at play. This suggests PSR J1023+0038 could indeed continue to brighten. It may also serve a similar prototype role for a new class of low-mass γ-ray binaries. We note that given its increased γ-ray luminosity, its present γ-ray efficiency, assuming an isotropic luminosity and taking $\dot{E} = 4.3 \times 10^{34}$ erg s⁻¹ (Deller et al. 2012), is 0.14. We therefore expect that the γ-ray luminosity will rise at most another factor of seven, as beyond that it would exceed the available energy from pulsar spin down.

Regardless of J1023's upcoming X-ray behavior, when it returns to an MSP state then it will be possible to compare the radio pulsar timing before and after the current active state. J1023 benefits greatly in this regard from its relative brightness and the high cadence of our recent radio monitoring observations. It also does not suffer from the contaminating influence of acceleration in a GC gravitational potential, as in the case of M28I. After being spun-up to millisecond periods, an accreting neutron star should enter a spin-down phase where the mass transfer rate decreases due to the progressive detachment of the donor radius from its Roche lobe. During this final evolutionary phase, the neutron star magnetosphere expands substantially and a large fraction of the neutron star rotational energy is lost via a propeller-like mechanism (Tauris 2012). It is possible therefore that what we are observing in J1023 is indeed this very last phase of its life as an qLMXB/LMXB. If this were the case and a propeller-like mechanism does occur before the source re-appears as a periodic emitter at any wavelength, then it should be possible to observe a strong spin down during this active phase, i.e., well in excess of the pulsar magnetic dipole spin down. Even if no accretion-powered X-ray pulsations are detected, this test can be performed once the active phase is over by comparing the pre- and post-active phase radio timing solutions.

In conclusion, we have demonstrated a new stage in the back-and-forth transitioning that seems to characterize the redback class of MSPs. For the first time, we have seen a bright radio MSP become undetectable while at the same time the system becomes γ-ray bright. We argue that the pulsar mechanism remains active in this system, but that the radio pulsations are obscured. While our higher-frequency radio observations have also been unsuccessful, there is perhaps potential for detecting pulsed X-rays (not generated by accretion) that will further confirm the continued activity of the pulsar itself. The detection or lack of accretion-powered X-ray pulsations in forthcoming X-ray observations can easily disprove or strengthen such an hypothesis. Similarly, planned radio interferometric observations to look for a continuum radio source are also important in this regard.

The analogy with the wider-orbit, much more massive companion γ-ray binaries is tantalizing. A multi-wavelength campaign that follows J1023 back into its radio MSP state may better resolve the state transition and associated radio/optical/X-ray/γ-ray phenomenology. Long-term, phase-coherent timing will also shed light on the accretion torques experienced while the radio pulsar was undetectable.

We thank H.A. Krimm for kind support in the use of the Swift/BAT data. A.P. acknowledges support from the Netherlands Organization for Scientific Research (NWO) Vidi fellowship. A.M.A. and J.W.T.H. acknowledge funding for this work from an NWO Vrije Competitie grant. J.W.T.H. also acknowledges funding from an NWO Vidi fellowship and ERC Starting Grant "DRAGNET" (337062). The WSRT is operated by ASTRON with support from NWO. Pulsar observations with the Lovell Telescope are funded through a consolidated grant from STFC. The Fermi LAT is a pair conversion telescope designed to cover the energy band from 20 MeV to greater than 300 GeV. It is the product of an international collaboration between NASA and DOE in the U.S. and many scientific institutions across France, Italy, Japan, and Sweden. The GBT is operated by the National Radio Astronomy Observatory (NRAO). NRAO is a facility of the NSF operated under cooperative agreement by Associated Universities, Inc. The Arecibo Observatory is operated by SRI International under a cooperative agreement with the NSF (AST-1100968), and in alliance with Ana G. Méndez-Universidad Metropolitana, and the Universities Space Research Association. The Fermi LAT Collaboration acknowledges support from a number of agencies and institutes for both development and the operation of the LAT as well as scientific data analysis. These include NASA and DOE in the United States; CEA/Irfu and IN2P3/CNRS in France; ASI and INFN in Italy; MEXT, KEK, and JAXA in Japan; and the K. A. Wallenberg Foundation, the Swedish Research Council, and the National Space Board in Sweden. Additional support from INAF in Italy and CNES in France for science analysis during the operations phase is also gratefully acknowledged.