The Rayleigh-Taylor instability is investigated in a material with a shear modulus m and yield stress s 0 . Incompressible experiments are conducted with constant and impulsive acceleration histories g͑t͒ using a fully characterized material. Two dimensional (2D) perturbations are found to be stabilized for wave number k . rg͞2m and initial amplitude h 0 , s 0 ͞rg, where r is the density. The stability region is larger for 3D perturbations. Unstable modes are found to grow classically during the acceleration phase, but they can recover elastically while coasting. [S0031-9007 (97)05230-7] PACS numbers: 47.20.Bp, 47.50. + dWhen an elastic-plastic plate is accelerated by a lower density fluid, the interface between them is Rayleigh-Taylor (RT) unstable [1-7], but the mechanical strength of the plate mitigates the growth. This occurs when metal plates are accelerated by high explosives (HE) [6,[8][9][10] or electrical currents [11], and in volcanic island formation due to the strength of the earth's crust [12].The effect of a shear modulus m and a tensile yield s 0 on RT stability is reviewed nicely by Robinson and Swegle (RS) [3]. The elastic force ϳ22mk 2 h can overcome the RT buoyancy ϳrgkh and stabilize small amplitude ͑kh ø 1͒ perturbations with wave numbers k . rg͞2m ϵ k c , (1) where r is the plate density and g is the acceleration. These modes may again be destabilized if the pressure drop across the initial amplitude h 0 exceeds the yield stress and drives the plastic flow, i.e., P ء ϵ rgh 0 ͞2s 0 $ P cr .(2) Estimates of the critical scaled pressure P cr for two dimensional (2D) perturbations are ϳ1.15͑͞1 1 e 2kH ͒ by Miles [1], where H is the plate thickness, ϳ1 by Drucker [2], ϳ0.58͓b 1 ͑b 2 1 1͒ 1͞2 ͔ 1͞2 by RS [3], where b ϵ 1 2 e 22kH , and ϳ͑1 2 0.86e 2kH͞ p 3 ͒ ͓͑1 2 e 2kH͞ p 3 ͒ 2 2 ͑k c ͞k͒ 2 ͔ by Nizovtsev and Raevskii (NR) [5,6]. The stability may be enhanced in 3D [2] with P cr ϳ 2 and reduced growth rates [6].Experiments [6,[8][9][10] in which metal plates are accelerated by HE find a P cr , 1 that decreases with l ϵ 2p͞k and show slower growth in 3D than in 2D [5]. Unfortunately, the uncertainties are large, ϳ50%, because the parameter variations are coarse and the material properties are uncertain at the high pressure. For example, to explain the observations, it is necessary to increase the static yield tenfold for weak aluminum (1100-0) but not for strong aluminum (6061-T6) and steel. If there is a strain rate dependence, as with viscosity ͑h͒, a reduced RT growth rate [3] may appear as strength since the experiments are short t ϳ 2͞ p kg. Also, Eq. (1) could not be tested because l c 2p͞k c exceeded the plate width.Numerical simulations [4,6] can treat the nonlinear constitutive properties and complex g͑t͒ profiles. For 10 GPa drive pressures, the simulations obtain P cr ϳ 0.4 at l 2H consistent with the HE experiments. However, the critical wavelength is observed to be much smaller than 4pm͞rg for the experimental conditions, possibly because k c H ø 1. At higher pressures, the values o...