Dr. Marcel B Finan 13 Local Linearity of Multivariable Functions For single variable functions, local linearity is a property of dierentiable functions that roughly says that if you zoom in on a point on the graph of the function (with equal scaling horizontally and vertically), the graph will eventually look like a straight line with a slope equal to the derivative of the function at the point. The advantage of this property is that function values at nearby points can be estimated using the tangent line, i.e., a linear function. The same is true for functions in two variables. A function in two variables is dierentiable at (a, b) if whenever we zoom in over suciently small region around a point (a, b) on the surface of the function we see that the surface is almost a plane containing (a, b). Thus, a dierentiable function of two variables is well approximated locally by a tangent plane,i.e., a linear function in the variables x and y. What is the equation of the tangent plane? Let z = L(x, y) be the equation of this plane. Then z = x+y + with being the xslope, i.e. = f x (a, b) and is the yslope, i.e. = f y (a, b). But z = z f(a, b) = x + y. Hence, from this we conclude that z = f x (a, b)(x a) + f y (a, b)(y b) + f(a, b). Figure 13.1 shows the graph of a function with a tangent plane. Figure 13.1 1 Example 13.1 Find the equation of the tangent plane to the surface z = 4x 3 y 2 + 2y at the point (1, 2, 12). Solution. Since f x (x, y) = 12x 2 y 2 and f y (x, y) = 8x 3 y + 2, we nd f x (1, 2) = 48 and f y (1, 2) = 14. Also, f(1, 2) = 12. Hence, the equation of the tangent plane is given by z = 48(x 1) 14(y + 2) + 12 or z = 48x 14y 64 Now as we mentioned above, the tangent plane is a good approximation of f(x, y) locally at the point (a, b). Hence, for points near (a, b) we can write: f(x, y) f x (a, b)(x a) + f y (a, b)(y b) + f(a, b). This equation is sometimes referred to as a local linearization of f(x, y) at the point (a, b). Example 13.2 Find the local linearization of f(x, y) = 4x 3 y 2 + 2y at the point (1, 2, 12). Estimate f(0.9, 2.1) using this linearization and compare your answer to the true value. Solution. By the previous example we have f(x, y) 48x 14y 64. Thus, f(0.9, 2.1) 48(0.9) 14(2.1) 64 = 8.6. The true value is f(0.9, 2.1) = 4(0.9) 3 (2.1) 2 + 2(2.1) = 8.65956 Although we have conned our discussion and examples to functions of two variables, the above arguments are easily extended to functions having three 2 or more variables. In particular, for a function of three variables f(x, y, z), the local linearization at (a, b, c) becomes: f(x, y, z) f x (a, b, c)(x a) +f y (a, b, c)(y b) +f z (a, b, c)(z c) +f(a, b, c). The Dierential of z = f(x, y) The formula for the local linearization can be used to estimate the change in the value of the function locally: f(x, y) f x (a, b)(x a) + f y (a, b)(y b) + f(a, b) or f(x, y) f(a, b) f x (a, b)(x a) + f y (a, b)(y b). By letting f = f(x, y) f(a, b), x = x a, and y = y b we can write f f x (a, b)x + f y (a, b)y. For arbitrary x and y we can introduce the function df = f x (a, b)dx + f y (a, b)dy. This function is called the dierential of f at (a, b). Note that the dierential of f is a linear function of the variables dx and dy. df represents the change in f locally due to changes in both x and y. Example 13.3 Compute the dierential of f(x, y) = 4x 3 y 2 + 2y at the point (1, 2, 12). Solution. We have df = f x (x, y)dx + f y (x, y)dy = 12x 2 y 2 dx + (8x 3 y + 2)dy. The dierential at (1, 2) is df = 48dx 14dy. Thus if we were to move by dx = 0.1 and dy = 0.1 i.e to (0.9, 2.1), the change in f is: df = 48(0.1) 14(0.1) = 3.2. Note that the actual change is f(0.9, 2.1) f(1, 2) = 8.65956 12 = 3.34044 3