MIT OpenCourseWarehttp://ocw.mit.edu 18.034 Honors Differential Equations Spring 2009 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms. LECTURE 23. THE DIRAC DISTRIBUTION Impulse signals: Dirac’s idea. Let a> 0 be small. Let fa(t) be identically zero except for the ∞ a interval t ∈ [0,a] and fa(t)dt = fa(t)dt =0. If the integral is not very small then fa(t) −∞ 0 must be quite large in the interval t ∈ [0,a], and the function describes the “impulsive” behavior. An “impulsive function” means a signal which acts for a very short time but produces a large effect. The physical situation is exemplified by a lightning stroke on a transmission line or a hammer blow on a mechanical system. fa(t)= 0 a Figure 23.1. Graph of a typical impulsive function In the early 1930’s the Nobel Prize winning physicist P. A. M. Dirac developed a controver-sial method for dealing with impulsive functions. Let a → 0+. The function fa(t)/( fa(t) dt) approaches to, say, δ(t) which takes zero for t =0 and the integral of δ(t) over any interval con-taining 0 is the unity. The function δ(t) is called “the Dirac delta function”. To formulate the Dirac delta function, let a> 0 and let 1/a 1/a, t ∈ [0,a)fa(t)= 0, elsewhere, ∞ a so that fa(t) dt = fa(t) dt =1for all a> 0. −∞ 0 a If fa(t) → δ(t) as a → 0+, then Figure 23. 2. Graph of fa(t) ∞ (23.1) δ(t)=0 for t ,δ(t) dt =1=0 . −∞ (This is often used as the definition of δ(t) in elementary differential equations textbooks.) It is easy to show that a 11 − e−sa L[fa(t)] = e −st dt = → 1 as a → 0. 0 a sa In this sense, δ(t) describes the effect of (23.2) L[δ(t)]=1. Let us say that no ordinary function with the property (23.1) exists, and whatever else δ(t) may be, it is not a function of t! However, says Dirac, one formally treats δ(t) as if it were a function and gets the right answer. 1 In the late 1940’s, the French mathematician Laurent Schwartz∗ succeeded in placing the delta function on a firm mathematical foundation. He accomplished this by enlarging the class of all functions so as to include the delta function, called it the class of distributions. Here, we first explore the usefulness of δ(t) in (23.1), and then we indicate its mathematical meaning. Examples. Let us consider the initial value problem y + y = δ(t), y(0) = y (0)=0. Taking the transform, 1 Ly = , y(t)= h(t) sin t. s2 +1Here, we use (23.2). The solution y is continuous for all t ∈ (−∞, ∞) and it satisfies the differential equation every-where except t =0.At t =0, however, it satisfies neither the differential equation nor the initial condition. It is not even differentiable at t =0. Indeed, y(0+) = 1 and y(0−)=0. The unit impulse signal δ(t) produces a jump of magnitude 1 in y(t) at t =0. Let us now consider y + y = fa(t)= 1/a, t ∈ [0,a) y(0) = y (0) = 0. 0, elsewhere, 1 We write fa(t)= (h(t) − h(t − a)). Taking the transform then gives a 1 1 1 − e−sa Ly = s2 +1 a s and thus, 1 1 ya(t)= h(t)(1 − cos t) − h(t − a)(1 − cos(t − a)) a a ⎧ ⎪ ⎪ 0, (−∞, 0] = ⎪ ⎨1 − cos t , t ∈ (0,a) a cos(t − a) − cos t ⎪ ⎪ ⎪ ⎩ , [a, ∞). a As a → 0+, the second interval t ∈ (0,a) vanishes, and the third interval t ∈ [a, ∞) tends to [0, ∞) and the functional expression on this interval tends to sin t. In other words, as a → 0 the solution ya(t) gives the same solution as was obtained with δ(t). We note that an uncritical use of δ(t) as in (23.1) and (23.2) gives the correct answer as obtained by a conventional passage to the limit. Moreover, the conventional method is much more difficult. It is where the usefulness of δ(t) lies. Next, let us consider y +2y +2y = δ(t), y(0) = y (0)=0. Taking the transform, 1 1 Ly = s2 +2s +2 = (s − 1)2 +1, y(t)= h(t)e −t sin t. This example illustrates another feature of impulsive signals that it induces a lasting effect. ∗Laurent Schwartz wass awarded a Fields medal (a mathematical equivalent of the Nobel prize) in 1950 for his creation of the theory of distributions. Lecture 23 2 18.034 Spring 2009input D2 +2D +2 Figure 23.3. The effect of impulsive signals. Theory of distributions. We conclude the lecture with a very brief description of the germ of Laurent Schwartz’s brilliant idea. A function is characterized by giving its value at each t.A distribution δ(t) is characterized, not by its value at t, but by giving its value δ{φ} on a suitable class of functions φ, called test functions. Test functions are assumed to have derivatives of all orders and to vanish outside of a finite interval. Let us define δ as ∞ (23.3) δ(t)φ(t) dt = φ(0) −∞ for any test function φ. We note here that we cannot speak of the value of δ(t) at t. The only ∞ meaningful quantity is δ(t)φ(t) dt. The distribution δ(t) is never used alone, but only in com-−∞ bination with functions. If δ(t) were a function and if the integral in (23.3) were an ordinary integral, then by a change of variables, ∞ ∞ δ(t − c)φ(t) dt = δ(t)φ(t + c) dt. −∞ −∞ By (23.3) the right side gives the value of φ(t + c) at t =0, that is, φ(c). Now we take it as the definition of the left side. Thus, δ(t − c) is defined as ∞ δ(t − c)φ(t) dt = φ(c) −∞ for any test function φ. Similarly, if δ(t) were a continuously differentiable function, then an integration by parts would yield ∞ ∞ δ(t)φ(t) dt = − δ(t)φ(t) dt. −∞ −∞ By (23.3) the right side is −φ(0), and again we take it to define δ(t) as ∞ δ(t)φ(t) dt = −φ(0) −∞ for any test function φ. With the same line of thought, δ(n)(t − c) is defined as ∞ δ(n)(t − c)φ(t) dt =(−1)nφ(n)(c) −∞ for any test function φ. For a<b, let us define b a δ(t − c)φ(t) dt = ∞ −∞ (h(t − a) − h(t − b))δ(t − c)φ(t) dt = φ(c), 0, if a � c � b otherwise. Lecture 23 3 18.034 Spring 2009 With the choice φ(t)=e−st, where s is a constant, we obtain L[δ(t − c)] = ∞ e −stδ(t − c) dt = e−sc , c � 0 0 0, c<0 when c � 0. When c<0 the result is zero. If c =0then the above gives the formula Lδ(t)=1, which agrees with (23.2). These are definitions,
View Full Document
Unlocking...