We use gravity and topography data to study crust and mantle structure in Phoebe Regio and Devana Chasma on Venus. We find that Phoebe is a region of thickened crust and hotter than normal mantle. These results are inconsistent with a cold, downwelling model and may support a mantle plume origin for Phoebe. However, the pattern of thermal anomalies is unusual, with temperature maxima on the periphery of the highland rather than in its center. This may reflect development of thermal boundary layer instabilities as the plume head interacts with the lithosphere. The hot mantle anomaly beneath Devana Chasma is discontinuous near latitude 10 North, coinciding with a 600 km offset in the rift zone trend. Devana should be regarded as two separate rifts, one propagating south from Beta Regio and the other propagating north from Phoebe Regio. Our results support the view that Venus remains a geologically active world.
 For planets other than Earth, orbital gravity observations provide our only quantitative constraints on the subsurface density anomalies associated with topographic, tectonic, and volcanic structures. Features with small ratios of gravity anomalies to topography are compensated at shallow depths and thus presumed to be related to density anomalies in the crust. Features with large gravity/topography ratios are either uncompensated or are compensated at large depths by density anomalies in the mantle.
 In this work, we examine compensation mechanisms for the Phoebe Regio and Devana Chasma region of Venus (Figure 1). Phoebe Regio is a highland region, with peak elevations of 1.5 to 2 km, and is covered by tessera terrain. Tessera is a highly deformed type of tectonic unit on Venus, with two or more sets of closely-spaced, intersecting faults. Tessera covers nearly 10% of the surface of Venus. In Phoebe Regio, fault structures in the tessera are interpreted as extensional in origin, with the amount of strain increasing at higher elevations [Hansen and Willis, 1996]. Devana Chasma is a rift system that begins at the northern edge of Phoebe Regio, continues northward to Beta Regio, and terminates on the north side of the Beta Regio highland. In the region between Phoebe and Beta, Devana has a maximum width of about 200 km and a total extension of 10 to 20 km [Foster and Nimmo, 1996; Rathbun et al., 1999; Connors and Suppe, 2001].
2. Gravity and Topography Inversion
Grimm and Phillips  and Herrick and Phillips  applied two-layer inversion methods to isolate the roles of shallow and deep compensation mechanisms in various parts of Venus. In these models, the shallow layer was assumed to be an Airy-compensated crust and the deep layer included viscous flow in the mantle. We follow the inversion method of Kiefer et al. , which we summarize briefly here. The inversion uses the observed gravity and topography to estimate the strength of density anomalies in two shells located at different depths. The inversion is performed in the spherical harmonic domain, inverting on a term-by-term basis for the crustal thickness variations and mantle density anomalies necessary to exactly reproduce the observed gravity and topography. The choice of the cut-off wavelength in the inversion involves a trade-off between horizontal resolution and vertical sensitivity. The maximum horizontal resolution is achieved for high degree spherical harmonics, whereas sensitivity to deep mantle structure is maximized at low degree harmonics [Richards and Hager, 1984]. An important goal of this study is to place constraints on the density structure in the upper mantle of the study region. Our results are complete through spherical harmonic degree 40, corresponding to a resolving half-wavelength of 475 km. The gravity and topography fields for this study are based on the work of Konopliv et al.  and Rappaport et al. .
 The shallow shell is assumed to represent variations in crustal thickness which are supported by a combination of Airy isostasy and elastic flexure. The flexure model [Turcotte et al., 1981] is an addition to the original inversion model [Kiefer et al., 1996]. Lithospheric flexure provides support for crustal anomalies and thus increases the associated gravity anomaly relative to the isostatic case. Spectral admittance models for Phoebe Regio have been developed by Simons et al.  and by Barnett et al.  for study domains similar to that modeled here. Simons found an elastic thickness of 20–30 km and a crustal thickness of 20–40 km. Barnett found an elastic thickness of 16–23 km, based on an assumed crustal thickness of 16 km. The results in this paper assume a mean crustal thickness of 25 km and an elastic lithosphere thickness of 20 km. Sensitivity tests for crustal thickness (15–40 km) and elastic thickness (15–30 km) show that our results are not sensitive to the precise choice of parameter values. Figure 2a shows regional variations in crustal thickness and should be interpreted as differences between the actual crustal thickness and the assumed global mean thickness of 25 km. These results assume a crustal density of 2900 kg m−3 and a mantle density of 3300 kg m−3.
 Density anomalies in the deep shell are assumed to drive viscous flow in the mantle. The viscous flow is calculated using a propagator matrix formulation [Richards and Hager, 1984] assuming whole mantle flow and free-slip boundary conditions. The depth-dependent viscosity model includes a 100 km thick high viscosity layer at the surface overlying an isoviscous interior [Kiefer and Hager, 1991]. The surface topography produced by the flow is calculated from the vertical normal stress, with the effects of the elastic lithosphere included in the force balance. The geoid anomaly produced by the deep shell is the sum of contributions from the mantle density anomalies and from the induced topography at the surface and at the core-mantle boundary. In principle, the mantle density anomalies can be different at different depths. For the sake of a well-posed inversion, we assume that the density anomalies have the same spatial pattern at all depths, so that our results should be considered as a vertical average through the mantle. We assume that the density anomalies are distributed from 100 km depth (the base of the upper thermal boundary layer) to 700 km depth. Density anomalies are presumably also present at larger depths but do not make much contribution to the observed geoid and topography. Mantle density anomalies are assumed to be thermal in origin and are converted to temperature anomalies assuming a thermal expansion coefficient of 3 · 10−5 K−1. Figure 2b shows the difference between the local temperature and the global mean mantle temperature.
Figure 2a shows that Phoebe Regio is topographically elevated primarily because it is a region of thickened crust. This includes both the main tessera unit (281 E, 12 S) as well as Chimon-mana Tessera (275 E, 3 S) to the northwest of Phoebe. The maximum crustal thickening in Phoebe is 12.6 km. Several other tessera highlands have thickened crust, including Alpha, Ovda, and Tellus Regiones [Grimm, 1994]. Past studies have interpreted tessera highlands in two very different ways. Bindschadler et al.  proposed that tessera units formed by crustal compression over cold, downwelling mantle. On the other hand, Phillips and Hansen  proposed that tessera formed as volcanically thickened crust over hot, upwelling mantle plumes, with the crust later experiencing extensional collapse. Long-wavelength admittance modeling demonstrates the presence of active mantle dynamics in Phoebe but does not indicate whether the flow is upwelling or downwelling [Simons et al., 1997].
 There is no evidence for a cold downwelling beneath Phoebe Regio (Figure 2b). The mantle temperature in central Phoebe is slightly higher than the global mean value. Temperature maxima are reached at the southern, northeastern, and western corners of Phoebe Regio. The temperature maximum in the south is associated with the shield volcano Yunya-mana Mons (285 E, 18 S). In the west, the temperature maximum occurs along a fracture belt. At the western edge of the fracture belt is the volcano Uretsete Mons (261 E, 12 S), whose lava flows are more than 500 km across. The temperature maximum is centered on the fracture belt (270 E, 9 S), 1000 km east of Uretsete Mons. In the northeast, the temperature maximum is associated with the Devana Chasma rift (288 E, 1 S). The positive temperature anomalies as well as the geologic indicators such as shield volcanos and rifts are broadly consistent with a mantle plume origin for Phoebe. In classical mantle plume models, e.g. [Kiefer and Hager, 1991], the maximum thermal anomaly is along the axis of the plume. An inversion similar to Figure 2 for Beta Regio does produce a plume-like thermal structure, with a single temperature maximum in the center of the highland. One possibility is that the thermal highs on the periphery of Phoebe are related to boundary layer instabilities in the mantle plume head [Moore et al., 1999]. An alternative interpretation, in which the three high temperature maxima around Phoebe formed independently, can not be ruled out. We prefer the single plume interpretation because of its simplicity.
 Hot mantle is present along the length of Devana Chasma, consistent with the interpretation that this is a rift zone. An important observation is that the rift thermal structure is discontinuous near 10 North latitude. Figure 1 shows that Devana strikes primarily north–south. Between 7 and 10 North latitude, Devana strikes primarily east–west to connect the two major arms of the rift system. Measured from the center of faulting in each rift segment, the offset in the rift is approximately 600 km. In the central region, the density of faulting is considerably lower than in the northern and southern segments (Figure 3). We interpret the discontinuity in mantle density structure, the offset in rift trend, and the decrease in fault density as indicating that Devana is actually two distinct rift systems. One arm of the rift propagates southward from Beta Regio and is driven by the stress regime created by the Beta mantle plume [Kiefer and Hager, 1991]. The other arm propagates northward from Phoebe Regio. The decrease in mantle temperature anomaly from (288 E, 1 S) northward along the rift is consistent with propagation of the Phoebe segment of Devana Chasma toward the north (Figure 2b). Initially, the two rifts formed independently. When the two rift arms approached within a critical distance of the other arm, each rift tip could interact with the stress field associated with the other rift [Pollard and Aydin, 1984]. The rifts then turned and began propagating toward each other.
 The continued existence of strong (>50 K) temperature anomalies in the study region indicates the presence of geologically young structures in this part of Venus. As an estimate of the required cooling time to eliminate mantle thermal anomalies, the hot thermal anomalies in the upper 700 km of the mantle will be advected to the base of the upper thermal boundary layer in less than 50 million years. Thermal diffusion through a 100 km thick thermal boundary layer takes an additional 100 million years. Thus, unless mantle thermal anomalies are actively maintained by convective flow, they will become undetectable in less than 150 million years. The hot thermal anomalies shown in Figure 2b must be younger than this, indicating that Venus remains a geologically active world at the present day.
 The regional plains in the study region include two large volcanic structures, Tuulikki Mons (275 E, 10 N) and Aruru corona (262 E, 9 N). Both were important volcanic features at some time, as shown by the crustal thickening (Figure 2a). However, the absence of mantle thermal anomalies at these two features (Figure 2b) suggests that both features are now geologically dead. Aruru has a peak to trough topography at harmonic degree 40 of 720 meters. If supported primarily by mantle dynamics, this would correspond to a gravity anomaly of 35–40 mGal, which would be easily detectable in the available gravity data. For Tuulikki, the long-wavelength topography is 390 meters, corresponding to a gravity anomaly of 20–25 mGal if dynamically supported. The non-detection of these features in Figure 2b is statistically reliable. Thus, our results can help to distinguish past and present geological activity on Venus.
 We thank Robbie Herrick and Pat McGovern for helpful discussions and Sue Smrekar and an anonymous individual for formal reviews.This research was conducted at the Lunar and Planetary Institute, which is operated by the Universities Space Research Association under contract NASW-4574 with the National Aeronautics and Space Administration.K.P. was supported by the LPI Summer Intern Program.This is Lunar and Planetary Institute Contribution No. 1145.