Stress and Strain

12.510 Introduction to Seismology 2/20/08 Stress and Strain Reading: Section 2.3 of Stein and Wysession In this lecture, we will examine using Newton’s 2nd Law and a generalized form of Hooke’s Law to characterize the response of a continuous medium to applied forces The stress tensor describes the forces acting on internal surfaces of a deformable, continuous medium. The strain tensor describes the distortion of (or the variation in displacement within) the body In order to produce wavemotion, we need - a spatial change in stress - a way to describe the causal relationship between stress and strain In 1-D, for a purely elastic medium, this relationship turns out to be quite simple: σ =Eε where σ is stress, ε is strain, and E is Young’s modulus (which must have dimensions of stress, as strain is dimensionless. The assumption of a purely elastic medium is one that we will make most of the time in this class. The relationships between stress and strain are dependent on the characteristics of the medium. Sidebar: Another possibility occurs for a viscoelastic medium: σ =ζė where σ is stress, ė is strain rate (δε/δt), and ζ is viscosity. In three dimensions, things become somewhat more complicated. A simple example should convince that the straightforward linear relationship given above can no longer hold. If we stretch a rubber band along its long axis (in this case applying a σx)… Figure 1 12.510 Introduction to Seismology 2/20/08 We can describe the strain along the x-axis in the normal way, l + Δl εx = l but we can see that despite the fact that σy = 0, εy does not equal zero (and a similar case could be made for the z direction) so we must turn to tensors to formulate our Hooke’s Law for three dimensions. We can represent the stress tensor in the form σij , where I represents the direction of the normal vector to the surface upon which the stress acts, and j represents the direction of the stress. For three dimensions, each of these can be broken down into x, y, and z components. Consider the volume element below. Fig. 2 (After Stein & Wyssession, 2003) Strain can be represented by a similar tensor, in the form εkl. The relationship between stress and strain must therefore take the form of a 4th order tensor that we call the elasticity (or stiffness) tensor. The elasticity tensor gives us information about the medium. We can now write Hooke’s Law in a general form. σ ij = cijkl ε kl Later, when we look at the Lamé parameters, remember that they must reside somewhere in the beast cijkl. This equation becomes very complex if it turns out that the elements of c are dependent on stress or strain (i.e. non-linear), but for small deformations we assume that this is not the case. So far we have made the following assumptions (a list that will expand as we proceed) Assumption #1: Our medium is purely elastic 12.510 Introduction to Seismology 2/20/08 Assumption #2: The elements of c are independent of stress and strain Next, we will develop Newton’s second law towards our goal of expressing an equation of motion. Newton’s second law simply states: ∑F i = mai Starting with the left hand side of the equation, we submit that the applicable forces fall neatly into one of two categories: 1. Body Forces: forces such as gravity that work equally well on all particles within the mass- the net force is proportional (essentially) to the volume of the body (fi). 2. Surface Forces: forces that act on the surface of a body- the net force is proportional to the surface area over which the force acts. The surface force Fs acts on a surface element dS which has a unit normal vector n. The forces acting on the surfaces of a volume element can be described by three mutually perpendicular traction vectors using the same convention that we described earlier for the stress tensor on a volume element. We define the traction as the limit of the surface force per unit area at any point as the area becomes infinitesimal r Traction ≡ T = r lim F = (T1 ,T2 ,T3 ) dS ⇒ 0 δS and recall from our volume element (this time using x1, x2, and x3 instead of x, y, and z for coordinate axes) that each traction can be thought of as the sum of forces acting upon an infinitesimal surface, broken down into normal and tangent components, such that r T1 = (T11 ,T12 ,T13 ) r T2 = (T21 , T22 , T23 ) r T3 = (T31 ,T32 ,T33 ) An equivalent notation used in the textbook (Stein and Wysession) gives r T1 = (T1(1) ,T2(1) ,T3(1) ) , etc. where the upper index indicates the surface normal and the lower index gives the force component direction. 12.510 Introduction to Seismology 2/20/08 This set of nine terms that describes the surface forces can be grouped into the stress tensor σji. The tensor’s rows are comprised of the three traction tensors, presented below in such a way as to demonstrate the equivalence of various methods of notation. σ ji σ xy σ xz ⎤ ⎡σ 11 σ 12 σ 13 ⎤ ⎡T (1) ⎤ ⎡T1(1) T2(1) T3(1) ⎤ ⎢ ⎥ ⎢ ⎥ ⎥ σ yy σ yz ⎥ = ⎢⎢σ 21 σ 22 σ 23 ⎥⎥ = ⎢T (2) ⎥ = ⎢T1(2) T2(2) T3(2) ⎥ σ zy σ zz ⎥⎦ ⎢⎣σ 31 σ 32 σ 33 ⎥⎦ ⎢⎣T (3) ⎥⎦ ⎢⎣T1(3) T2(3) T3(3) ⎥⎦ ⎡σ xx ⎢ = ⎢σ yx ⎢σ zx ⎣ Backing up a bit, we return to our discussion of Newton’s 2nd Law, this time separating out body forces (fi) from surface forces (Ti) ∑F i = mai which we can expand to: ∫ f dV + ∫ T dS = ma i V i i S Looking at this equation, we see that we are adding a surface integral to a volume integral, so we should want to convert one to the other form. Since we have mass on the right hand side (which is an integration of density over volume), we should elect to convert the surface integral into a volume integral, which can be done using the Gauss (Divergence) Theorem. For the time being, we will leave this idea to simmer, but will return to collect it when we develop equations of motion at the end of this lecture. Before we continue, let’s pause for a moment to collect our bearings: If we look again at the stresses, we recall that we have defined them as a force per unit area, so we can think of them as similar to pressures, with units of Pa. 1 atm = 1 bar = 1000 mbar = 105 Pa Strain is defined as a change in some dimension (length, volume, angle) over that same dimension, so it is by definition dimensionless. This means that our elasticity/ stiffness tensor cijkl must also have dimensions of stress (Pa). To develop a feeling for some typical values, we present a brief table of values for lithostatic stress at various depths. Lithostatic stress ≡ The normal stress due to the weight of the overlying rock (overburden) 12.510 Introduction to Seismology 2/20/08 Depth 410 km 660 km CMB Center of the Earth Typical values for lithostatic stress 14 GPa 23 GPa 130 GPa 360 GPa We can see that these values are quite extreme. Actually, for seismic propagation we are more interested specifically in non-lithostatic stresses, but more on that later. The stress tensor gives a traction vector acting on an arbitrary surface element within the medium. We can show this by presenting the stress tensor on an infinitesimal tetrahedron, where one face is a surface element with a normal vector n that is not aligned with any coordinate axis. A traction is applied to this surface. The other three faces of the tetrahedron are each perpendicular to a coordinate axis and to each other. Fig. 3 (After Stein & Wyssession, 2003) We can decompose the traction into components acting on our perpendicular surfaces. 12.510 Introduction to Seismology 2/20/08 r T = (T1 ,T2 ,T3 ) Note that in doing this we have implicitly made a third assumption, namely that Assumption #3: Volume elements within the medium can be treated as continuums Each of our Ti also have three components (a total of nine elements), so we can rewrite the stress tensor as: Ti dS = σ 1i ΔS1 + σ 2i ΔS 2 + σ 3i ΔS 3 We can also see that there is a relationship between angles (and surfaces), specifically that ΔS i = cos γ i dS , where γ i is the angle between the normal vector and the ith coordinate axis. This means that we can rewrite our stress tensor as a sum of scalar products. Ti = σ 1i ⋅ n̂1 + σ 2i ⋅ n̂ 2 + σ 3i ⋅ n̂3 or in Einstein summation Ti = σ ji ⋅ n̂ j Likewise we can look at the elements of the stress tensor σ ji = Ti ( j ) In the absence of body forces, the stress tensor is treated as symmetric (σij=σji), so there are only six independent elements. The diagonal elements represent the normal stress and the off-diagonal elements the shear stress. Assumption #4: The stress tensor is symmetric (σij=σji) A symmetric tensor can also be rotated into a principal coordinate frame such that the tractions become parallel to the normals (i.e. the shear stresses go to zero) In other words, Ti = σ ji ⋅ n̂ j = σ ij ⋅ n̂ j = λn̂i which is a straightforward eigenvalue/eigenvector problem. (σ ij − λδ ij )n̂ j = 0 12.510 Introduction to Seismology 2/20/08 σ 12 σ 13 ⎤ ⎡σ 11 − λ ⎢ det ⎢ σ 21 σ 22 − λ σ 23 ⎥⎥ = 0 ⎢⎣ σ 31 σ 32 σ 33 − λ ⎥⎦ Solving the determinant for zero gives a cubic equation, therefore three solutions for λ (eigenvalues) which can be plugged back in to get the principal stress axes (eigenvectors). ⎡σ 1 0 σ = ⎢⎢ 0 σ 2 ⎢⎣ 0 0 0⎤ 0 ⎥⎥ σ 3 ⎥⎦ In seismology, we typically order these such that σ1 > σ2 > σ3 (i.e. σ1 is the most compressive stress). Some situations in which the sigma values have specific relationships are of particular interest. 1. Uniaxial stress (σ1≠0, σ2 = σ3 = 0) 2. Plane stress (σ1 ≠ 0, σ2 = 0, σ3 ≠ 0) 3. 4 Pure shear stress (σ1 = -σ3, σ2 = 0) (This is actually a special case of plane stress) Isotropic stress (σ1 = σ2 = σ3 = P) (also termed lithostatic/hydrostatic stress) As we stated earlier, in seismic propagation, we are interested in the non-lithostatic stresses, i.e. that which is left over when we subtract out the lithostatic stress. Deviatoric Stress ≡ the remaining stress after the effect of the mean stress (P=1/3 (σ1 + σ2 + σ3)) has been removed. σ 12 σ 13 ⎤ ⎡σ 11 − P ⎢ σ ij′ = ⎢ σ 21 σ 22 − P σ 23 ⎥⎥ ⎢⎣ σ 31 σ 32 σ 33 − P ⎥⎦ 12.510 Introduction to Seismology 2/20/08 12.510 Introduction to Seismology 2/20/08 12.510 Introduction to Seismology 2/20/08 12.510 Introduction to Seismology 2/20/08 12.510 Introduction to Seismology 2/20/08