4.1. Crustal Structure and Composition
 With the exception of crust beneath Attu island, receiver function results that are independent of active-source constraints indicate that the crust of the Aleutian island arc is 38.5 ± 2.9 km thick, where 2.9 km is the RMS variation of individual measurements, and not much larger than the 2.6 km average uncertainty. Thus, crustal thickness remains relatively constant despite lateral changes in arc properties, including the westward transition from a continental to oceanic overriding plate, changes in obliquity and speed of convergence, increasing along-strike extension of the arc westward [Freymueller et al., 2008], and changes in composition of primitive lavas or the parental arc magma [Kelemen et al., 2003]. Our sampling of the arc is sparse and constrained to the locations of the islands, so there may be significant variability in crustal thickness at interstation wavelengths shorter than 50–100 km that would not be observed. Nonetheless, our data does reveal that there is comparatively little long-wavelength variation in crustal thickness along much of the arc. We would expect an increase in crustal thickness from the oceanic to continental sections due to thickness changes in preexisting crust, if magmatic productivity remained constant despite large changes in preexisting crustal thickness along the arc. Arc-normal convergence rate varies by <10% from ADK east to SDPT [Syracuse and Abers, 2006]. The observed consistency in crustal thickness implies that some process, either variation in magma production rate or differential erosion or foundering of the crust, must modulate crustal thickness along the arc. Age estimates for volcanic rocks in the Aleutians from 40Ar/39Ar dating show differences in ages of eruptive events along the arc [Jicha et al., 2006] indicating that magma production may not be uniform. On the other hand, most island summits lie at elevations within 1000 m of sea level; thus, similar crustal thicknesses are balanced by similar elevations, perhaps indicating that erosion and isostasy act to keep crustal thickness roughly constant. We note that the small average velocity changes inferred from active-source data, roughly ±0.2 km/s, correspond to average density changes of perhaps 100–200 kg/m3, too small to significantly affect isostatic balance more than 1–2 km in crustal thickness.
 The average Vp/Vs of 1.77 indicates an intermediate to mafic crustal composition in the Aleutian island arc [Zandt and Ammon, 1995] and is higher than typical continental crust [Christensen, 1996; Brocher, 2005]. However, lower Vp/Vs appears to be present in the east than the west (Figure 6) more similar to that of continental crust [Christensen and Mooney, 1995; Christensen, 1996]. To first order, silica content is inversely proportional to Vp/Vs [Zandt and Ammon, 1995], at least in rocks with >55 wt % SiO2 and bearing quartz [Hacker et al., 2003]. The change implies that the crust to the west is more mafic than that to the east. This variation in average crustal Vp/Vs corresponds with the change from continental to oceanic crust in the overriding plate, which should lead to a westward decrease in average silica content since the continental crust is expected to contain more evolved continental material.
 The other possible contribution to Vp/Vs is the composition of primary magmas, which can lead to changes in the crustal composition [Kay and Kay, 1994; Kelemen et al., 2003]. Specifically, the concentration of MgO decreases and SiO2 increases westward along the intraoceanic part of the arc [Kay and Kay, 1994; Kelemen et al., 2003]. This should correspond to a westward decrease in Vp/Vs along the arc, to the extent that higher SiO2 leads to lower Vp/Vs. This is the opposite of what we observe. Although Vp/Vs increases with decreasing SiO2 for many bulk compositions, rocks containing <55 wt % SiO2 have minimal quartz and show a weak positive correlation between Vp/Vs and SiO2 controlled by other phases [Behn and Kelemen, 2006]. For example, rocks with greater abundance of olivine and orthopyroxene tend to have lower Vp/Vs than rocks dominated by plagioclase or garnet [Hacker et al., 2003]. If most of the intruded lavas are sufficiently primitive that quartz is minor to absent, this trend would be expected. Primitive Aleutian arc magmas show SiO2 ranging between 45 and 60 wt % in our study area and thus straddle both trends [Kelemen et al., 2003]. At least in the upper and middle crust, it is reasonable to expect more silicic compositions consistent with the lower observed Vp/Vs.
 In addition to the Moho boundary, a midcrustal boundary is observed at two stations in the central portion of the arc at ∼20 km depth with a Vp/Vs of ∼1.70 in the overlying section. The average Vp/Vs in the upper crust is lower than the total crustal Vp/Vs, suggesting that the Vp/Vs in the lower crust is higher. Such stratification is expected for many crustal differentiation scenarios that place more silicic rocks in the upper crust over a lower crust that is mafic. Additionally, elevated temperatures, presence of melt, and other factors that might be relevant to the lower crust beneath the active volcanic line would also serve to increase Vp/Vs in the deep part of the crust.
4.2. Previous Studies and Comparison to Other Arcs
 A small number of previous active-source seismic studies in the Aleutians provide a basis for comparison to this study. Shillington et al.,  present a P wave velocity model based on a sparse along-arc seismic refraction profile, 20–40 km south of the active volcanic line (but within the arc platform). The “unconstrained” Moho depths (those independent of the active-source data) are generally in agreement with the results from Shillington et al. —their study predicts an average crustal thickness from 35 to 37 km, while the average crustal thickness from this study is 38.5 km. This agreement is independent of our input P wave velocity assumption: even a very low crustal Vp of 6.5 km/s on average does not yield thicknesses less than 35 km (Figure 6). Our results more closely agree with the results of Shillington et al.  and Van Avendonk et al.  than the 30 km thick crust obtained by Fliedner and Klemperer [1999, 2000] in their analysis of the same profile, or equally thin crust (∼30–32 km) seen on the cross lines [Holbrook et al., 1999; Lizarralde et al., 2002]. Six of the eight coincident stations have Moho error ranges that overlap with the depth range predicted by the Shillington et al.  active-source data. The mid-crustal layer observed at approximately 20 km depth in Shillington et al.,  also is observed in this study intermittently.
 The Vp/Vs of the lower crust using active-source data by Shillington et al.  yields somewhat lower values of Vp/Vs (1.7–1.75) than implied by our study. The active-source profile lies seaward of the active arc (and seaward of the arc crust sampled by receiver functions in this study). While crustal thickness is not expected to change considerably over the arc platform, there could be significant changes in temperature. Thus, lower Vp/Vs may be expected there owing to the absence of melt and lower temperatures [e.g., Takei, 2002]. Additionally, at lower temperatures, alpha quartz is stable at lower-crustal depths; even small (<5%) amounts can reduce Vp/Vs of the bulk composition substantially [Ohno et al., 2006; Shillington et al., 2013]. The Vp/Vs estimates from Fliedner and Klemperer  agree with our results in the eastern section of the arc, but since their crustal thicknesses do not, it is not clear they can be compared.
 Zandt and Ammon  characterize Poisson's ratio for different types of continental crust worldwide. While their sampling of island arcs is very limited and does not include the Aleutians, they suggest that the average arc has a Vp/Vs of 1.91 ± 0.16. This is higher than our measurements, but within their uncertainties. Studies elsewhere differ. The Izu-Bonin system exhibits substantial along-strike variations in average crustal thickness, from ∼20 km in the Bonin arc to ∼30 km in Izu [Kodaira et al., 2007a]; no published information on Vp/Vs exists for this area. This arc contains a thick region of relatively felsic material; it has been suggested that it may represent a more mature arc system than the Aleutians [Tatsumi et al., 2008]. In addition, large variations in crustal thickness (>10 km) are seen in the Izu-Bonin system on short (between volcanoes) scales [Kodaira et al., 2007a, 2007b], which are not seen here. These rapid variations in arc thickness have been interpreted to imply that focused crustal accretion and the generation of relatively felsic crust are occurring beneath large basaltic volcanoes in this arc. In Costa Rica, the crustal thickness ranges from 27.2 to 37.9 km while in Nicaragua it ranges from 24.6 to 43.5 km including the forearc and backarc, and where differences in preexisting crust are known to exist [MacKenzie et al., 2008]. Central America shows more crustal thickness variability than we observe in the Aleutians, which have comparable differences in preexisting crust. However, studies in the Aleutians do not have a sampling density comparable to these studies, and thus would miss volcano-spacing wavelength structure. But larger-scale variations are not observed.