Ubiquitous Low-Velocity Layer Atop the 410-km Discontinuity in the Northern Rocky Mountains

Receiver functions (RF) from three 30-station IRIS-PASSCAL small aperture arrays (2-15 km station spacing) operated for ten months each in the northern Rocky Mountains show a ubiquitous negative polarity P to S arrival (NPA) just preceding the 410-km discontinuity arrival. Data from the arrays was divided into NW, SE and SW backazimuths and stacked to form nine quadrant stacks (QS). Remarkably, the NPA is apparent in 8 of the 9 QS, with 7 of the 8 displaying a similar dipole shape (paired negative and positive swings). Each QS contains clear P to S arrivals from the 410- and 660-km discontinuities and display the correct moveout. To model the NPA, a "double gradient slab" model consisting of five parameters is used: top gradient thickness and shear wave velocity drop; a constant velocity layer; bottom gradient thickness and shear wave velocity increase. Model misfit is assessed via a grid search over the model space using a reflectivity code to calculate synthetic seismograms. Assessment of model likelihood is done by calculating 1- and 2-D marginal probability density functions (PDF). Model parameters for each QS are well resolved and uncorrelated, with the exception of the anti-correlation of the top and bottom gradients. To define an average model, the probability distributions of each QS for each parameter are multiplied to form summary 1-D marginal PDF from which 90% probability bounds are calculated. These probability bounds are: the top gradient is < 8 km with a velocity decrement of 0.3-0.5 km/s; the constant velocity layer thickness is < 5 km; and, the bottom gradient is 29-37 km with a velocity increase of 0.4-0.6 km/s. The effective width of the low velocity layer atop the 410 (herein called the 410-LVL) is characterized as the layer thickness plus half the two gradient widths. Thus, the 410-LVL is found to have a mean thickness of 26 km and a mean shear wave velocity decrement of 8.3%. These results contrast with 410-LVL widths of 25-90 km and shear velocity drops of 4-6% reported by other researchers. Physical explanations for the origin of this 410-LVL suggest that temperature and/or compositional effects are unlikely causes. The positive Claypeyron slope associated with the 410 km discontinuity is expected to enhance thermal convection; thus, no thermal boundary layer nor ponding of thermal heterogeneity is expected (as opposed to the 660). While a layer of oceanic crust could produce a low velocity anomaly, again the 410 km discontinuity is not expected to trap chemical heterogeneity: thus, ascribing the 410-LVL to subducted slab would be serendipitous. Dipping layers or anisotropic velocity contrasts cannot cause the NPA and furthermore the tangential QS are quiet with respect to the radial components. Thus, via a process of elimination, we are drawn to the "water-filter" model [Bercovici and Karato, 2003] which predicts a layer of partial melt atop the 410 as a result of melting of hydrous wadsleyite as it upwells across the 410. A simple global water-filter model calculation that assumes a 1 mm/yr upwelling rate suggests that the melt layer would have 0.1%-1.5% melt porosity and a thickness of 2-20 km. We note that for plausible cm/yr scale up-flow across the phase transition, much thicker melt layers could develop. We interpret our findings of a 26 km wide 410-LVL with an 8.3% shear wave velocity drop (corresponding to a 3.8% melt porosity using the melt-velocity scaling of Kreutzmann et al., 2004) as consistent with the "water-filter" model.