Ordinary Differential Equations

Transcription

1 Ordinary Differential Equations Dan B. Marghitu and S.C. Sinha 1 Introduction An ordinary differential equation is a relation involving one or several derivatives of a function y(x) with respect to x. The relation may also be composed of constants, given functions of x, or y itself. The equation y (x) = e x, (1) where y = dy/dx, is of a first order ordinary differential equation, the equation y (x) + y(x) = 0, () where y = d y/dx is of a second order ordinary differential equation, and the equation x y (x) y (x) + 3e x y (x) = (x + 1)y (x), (3) where y = d 3 y/dx 3 is a third order ordinary differential equation. The order of an ordinary differential equation the highest derivative of y in the equation. Definition [1]. The explicit solution of a first-order differential equation is a function y = g(x), a < x < b, (4) defined and differentiable on (a, b), with the property that the equation becomes an identity when y and y are replaced by g and g, respectively. The solution of a differential equation G(x, y) = 0 it is called the implicit solution. 1

2 Example. The explicit solution of the first-order differential equation y (x) = x y(x), (5) is y(x) = c e x /, (6) where c is an arbitrary constant. The differential equation (5) has many solutions. The function (6), with arbitrary c, represents the general solution (the totality of all solutions of the equation). If we consider a definite value of c, for example c = 1, then the solution obtained y(x) = e x / is called a particular solution. First order differential equations.1 Separable equations The equation or g(y)y = f(x), (7) g(y)dy = f(x)dx, (8) is called an equation with separable variables, or a separable equation. The variable x appears only on the right hand side and the function y appears only on the left hand side in Eq. (8). Integrating both sides we obtain g(y)dy = f(x)dx + c. (9) If f and g are continuous functions the general solution of Eq. (7) is obtained evaluating Eq. (9). Example. Solve the equation (y + 1)xdx + (x + 1)ydy = 0. The above equation can be rewritten in the form x x + 1 dx + y dy = 0. y + 1

3 By integration we obtain x ln 1 + x + 1 ln(1 + y ) = c, x With x = 0 and y = 0 we calculate c = 1 ln and x + ln 1 + y = ln(1 + x), x 1. Definition. A first-order differential equation together with an initial condition is called an initial value problem. The initial condition is the condition that at some point x = x 0 the solution y(x) has a prescribed value y(x 0 ) = y 0.. Equations reducible to separable form The first-order differential equation ( y y = g, (10) x) where g is any given function of y/x (g(x) = f(y/x)), can be made separable equation by a simple change of variables. The change of variable is y x = u. The function y = u x and by differentiation we obtain y = u + u x. (11) Combining the equations (11) and (10), and taking into account that g(y/x) = g(u) we obtain u + u x = g(u). By separating the variables u and x, the previous equation takes the form du g(u) u = dx x. 3

4 After integration and replacement of u by y/x the general solution of Eq. (10) is obtained. Example. Solve the equation dy dx = x + 3y 3x + y and 3x + y 0. With the change of function y = ux we obtain or u x + u = + 3u 3 + u u x = u 3 + u and Integrating or u 1 u du = dx x. dx x = u 1 u du ln x = 1 ln u ln 1 u + ln c. 1 + u ( ) 1 u 3 We obtain x 4 (u 1) = c. Replacing u by y/x, the general integral 1 + u will be (x y ) (x y) 3 = c(x + y) 3..3 Exact differential equations A first-order differential equation M(x, y)dx + N(x, y)dy = 0, (1) 4

5 is exact if the left hand side is an exact differential Equation (1) can be rewritten as d u(x, y) = u u dx + dy. (13) x y d u = 0. and by integration the general solution is If there is a function u(x, y) with the properties u(x, y) = c. (14) (a) u x = M, (b) u y = N, (15) then M(x, y)dx + N(x, y)dy = 0 is an exact differential equation. The necessary and sufficient condition for Mdx + Ndy be an exact differential [1] is M y = N x. (16) To find u(x, y) we have the following steps [1]. From Eq. (15.a) if we consider y to be a constant we obtain u = Mdx + k(y), (17) where and k(y) is the constant of integration. k(y) is determined from Eq. (17) deriving u/ y. From Eq. (15.b) we get dk/dy. Example. Solve the equation x x y y + ln(x y) + x x y Writing the equation in the form Eq. (1), we get = 0 x y > 0. x [ x y dy + ln(x y) + x ] dx = 0 (18) x y 5

6 Equation (18) is exact. Consider M = ln(x y)+ x x y, and N = x x y. Then, by differentiation we obtain M y = N x = 1 x y + 1 x y + From Eq. (15.b) we have N = u y x (x y), x (x y). and by integration u = x ln(x y) + k(x). To determine k(x) we differentiate u and apply Eq. (15.a) u x = ln(x y) + x x y + dk dx = M. By simple algebraic manipulations, we find that dk = 0 and, consequently, dx k(x) = c, where c is an arbitrary constant. We obtain the final form x ln(x y) = c..4 Linear differential equations We consider the first-order differential equation y + f(x) y = r(x), (19) which is linear in y and y (f and r may be any given functions of x). If r(x) = 0, x (for all x) the equation is homogeneous. For r(x) 0 the equation is said to be nonhomogeneous. Assuming that f(x) and r(x) are continuous for x I, we need to find a general formula for Eq. (19). Case I. homogeneous equation For the equation y + f(x)y = 0, (0) 6

7 separating variables we have and the solution is dy y = f(x)dx or ln y = f(x)dx + c, y(x) = ce f(x)dx (c = ±e c when y < 0 or y > 0). (1) Case II. nonhomogeneous equation Multiplying Eq. (19) by F (x) = e h(x) where h(x) = f(x)dx. we find Since h = f, we obtain e h (y + fy) = e h r. d dx (yeh ) = e h r. Integrating the above relation we have ye h = e h rdx + c. The general solution of Eq. (19) in the form of an integral may be written [ ] y(x) = e h e h rdx + c, h = f(x)dx. () Example. Solve the differential equation xy + (1 x)y = xe x. We can rewrite the equation in the form ( ) 1 y + x 1 y = e x. Comparing the previous equation to Eq. () we can identify ( ) 1 h = x 1 dx = log x x, 7

8 no constant being added in the integration. Thus, the solution will be ( ) y = e log x+x e log x x e x dx + c ( = ex x x e x ex dx + c), or where c is an arbitrary constant. y = e x ( x + c x.5 Variation of parameters Another way of finding the general solution of linear differential equation ), y + f(x)y = r(x). (3) is the method of variation of parameters. The solution corresponding to a homogeneous equation (r(x) = 0) is v(x) = e f(x)dx. (4) With Eq. (4) we try to determine a function u(x) such that y(x) = u(x)v(x), (5) is the general solution of Eq. (3). This approach is called the method of variation of parameters [1]. Equations (5) and Eq. (3) can be combined into u v + u(v + fv) = r, or u v = r, since v + fv = 0. We find u = r v and by integration u = We obtain the general solution y = uv = v r dx + c. v which is identical with Eq. () of the previous section. ( r v dx + c ), (6) 8

9 3 Second order differential equation 3.1 Homogeneous linear equations A second-order differential equation which can be written as y + f(x)y + g(x)y = r(x) (7) is said to be linear. It is said to be nonlinear if it cannot be written in the form of Eq. (7). The functions f and g are called the coefficients of the equation (7). If r(x) 0, then Eq. (7) is said to be nonhomogeneous. Otherwise, it is said to be homogeneous and takes the form y + f(x)y + g(x)y = 0. (8) It is called a solution of a differential equation of the second order on an interval J a function y = φ(x) which is defined and two times differentiable on J. Moreover, the equation becomes an identity if φ and its derivative replace the unknown function y and its derivatives, respectively. For the case of homogeneous equations, the following theorem states that solutions of Eq. (8) can be obtained from known solutions by multiplication by constants and by addition. Fundamental Theorem [1]. If a solution of the homogeneous linear differential equation (8) on the interval J is multiplied by any constant, the resulting function is also a solution of Eq. (8) on J. The sum of two solutions of Eq. (8) on J is also a solution of Eq. (8) on that interval. Proof. We assume that φ(x) obeys the conditions to be a solution of Eq. (8) on J. If we replace y by cφ(x) is into Eq. (8), we obtain (cφ) + f(cφ) + gcφ = c[φ + fφ + gφ]. Since φ is a solution of Eq. (8), then φ + fφ + gφ = 0 and we find that cφ is also a solution of Eq. (8). The second part of the theorem can be proved in the same way. Example. The functions y 1 = φ 1 = x and y = φ = x, x R {0} (J R {0}), are two solutions of the equation x y xy + y = 0. The function y 3 = c 1 φ 1 + c φ = c 1 x + c x is also a solution of the equation. 9

10 3. Homogeneous equations with constant coefficients We consider the homogeneous equations of the form y + ay + by = 0, (9) where a, b R are constants, and x R. The solution of the first-order homogeneous linear equation with constant coefficients is an exponential function, We assume that y + ky = 0, y = C e kx. y = e λx, (30) may be a solution of Eq. (9) if λ is properly chosen. Substituting Eq. (30) and its derivatives into Eq. (9), we obtain y = λe λx and y = λ e λx, (λ + aλ + b)e λx = 0. So Eq. (30) is a solution of Eq. (9), if λ is a solution of the equation λ + aλ + b = 0. (31) Eq. (31) is called the characteristic equation of Eq. (9). Its roots are λ 1 = 1 ( a + a 4b), λ = 1 ( a a 4b). (3) From derivation it follows that the functions y 1 = e λ 1x and y = e λ x, (33) are solutions of Eq. (9). This result can be verified by substituting Eq. (33) into Eq. (9). Elementary algebra states that, since a and b are real, the characteristic equation may have 10

11 Case I two distinct real roots, Case II two complex conjugate roots, or Case III a real double root. Example 1. Solve the equation y 5y + y = 0. The characteristic equation of the given differential equation will be λ 5λ + = 0 so that λ 1 = 1, λ =. Then the general solution is y = c 1 e x/ + c e x. Example. The equation y + y + 5y = 0 has the characteristic equation λ + λ + 5 = 0 from which λ 1, = 1 ± i. The general solution will be y = e x (c 1 cos x + c sin x). Example 3. The equation y y + 1 = 0 has the characteristic equation (λ 1) = 0 which gives λ 1, = 1. We obtain the general solution y = e x. 11

12 3.3 General solution. Fundamental system Definition. The general solution of a second order differential equation is a solution which contains two arbitrary independent constants, i.e. the solution cannot be reduced to a form containing only one arbitrary constant or none. A particular solution is a solution obtained from the general solution assigning specific values to the arbitrary constants. We consider the general homogeneous linear equation y + f(x)y + g(x)y = 0, (34) and two solutions y 1 (x) and y (x) of this equation. The Fundamental Theorem states that y(x) = c 1 y 1 (x) + c y (x), (35) is a general solution of Eq. (34), where c 1 and c are two arbitrary constants. Two functions y 1 (x) and y (x) are linearly dependent on an open interval I where both functions are defined, if they are proportional on I (a) y 1 = m y or (b) y = n y 1, (36) for all x I, where m and n are numbers. If the functions are not proportional, they are linearly independent on I. If at least one of the functions y 1 and y is identically zero on I, then the functions are linearly dependent on I. In any other case the functions are linearly dependent on I if and only if the quotient y 1 /y is constant on I. Hence, if y 1 /y depends on x on I, then y 1 and y are linearly independent on I [1]. Example 1. The functions y 1 = 9x and y = 3x are linearly dependent, because the quotient y 1 /y = 3 = const while the functions y 1 = x + x and y = x are linearly independent because y 1 /y = x + 1 const. Two linearly independent solutions of Eq. (34) on I constitute a fundamental system or a basis of solutions on I. 1

13 Theorem [1]. The solution y(x) = c 1 y 1 (x) + c y (x) (c 1, c arbitrary) is a general solution of the differential equation Eq. (34) on an interval I of the x-axis if and only if the functions y 1 and y constitute a fundamental system of solutions of Eq. (34) on I. y 1 and y constitute such a fundamental system if and only if their quotient y 1 /y is not constant on I but depends on x. Example. The equation has the solutions y y 15y = 0 y 1 = e 5x and y = e 3x. These solutions constitute a fundamental system because the ratio y 1 /y is not constant. The general solution is y = c 1 y 1 + c y = c 1 e 5x + c e 3x. 3.4 Complex roots of the characteristic equation. Initial value problem The solutions of the homogeneous linear equation with constant coefficients are y + ay + by = 0 (a, b real) (37) y 1 = e λ 1x and y = e λ x, (38) where λ 1 and λ are the roots of the corresponding characteristic equation λ + aλ + b = 0. (39) In the case of λ 1 λ, the quotient y 1 /y is not constant, and the solutions constitute a fundamental system for all x. The general solution is y = c 1 e λ 1x + c e λ x. (40) 13

14 The solutions of the Eq. (38) are real if the distinct roots of the corresponding characteristic equation are real (Case I). If λ 1 and λ are complex conjugate roots of the form (Case II) λ 1 = p + i q, λ = p i q, then the solutions Eq. (38) are complex y 1 = e (p+i q)x, y = e (p i q)x. The real solutions can be derived from the complex solutions by applying the Euler formulas e iθ = cos θ + i sin θ, e iθ = cos θ i sin θ, for θ = qx. The first solution becomes while the second one is y 1 = e (p+iq)x = e px e iqx = e px (cos qx + i sin qx), y = e (p iq)x = e px e iqx = e px (cos qx i sin qx). From Fundamental Theorem we can conclude that they are solutions of the differential equation Eq. (37). The corresponding general solution is y(x) = e px (A cos qx + B sin qx) (41) where A and B are arbitrary constants. Example 1. Let us consider the second order differential equation with constant coefficients y 4y + 5y = 0 The corresponding characteristic equation is with the roots λ 4λ + 5 = 0, λ 1 = p + iq = + i and λ = p iq = i. 14

15 For this example p =, q = 1, and from Eq. (41) the answer is y = e x (A cos x + B sin x). Let us consider the values of the solution y(x) and its derivative y (x) at an initial point x = x 0 y(x 0 ) = K, y (x 0 ) = L, (4) The conditions Eq. (4) and the equation Eq. (37) constitute an initial value problem. To solve such a problem we must find a particular solution of Eq. (37) satisfying Eq. (4). Such a problem has a unique solution. Example. Let us consider the initial value problem y 4y + 5y = 0, y(0) =, y (0) = 0. A fundamental system of solutions is e x cos x and e x sin x, and the corresponding general solution is y(x) = e x (A cos x + B sin x), with the initial condition y(0) = A. The derivative y = e x [(A + B) cos x + (B A) sin x)], has the initial value y (0) = A + B. Solving the initial conditions system, y(0) = A =, y (0) = A + B = 0. we get A = 4, B = 1, and the general solution of the differential equation is y = e x ( cos x 4 sin x). (43) 15

16 3.5 Double root of the characteristic equation Now we consider the case when the characteristic equation associated to a homogeneous linear differential equation with constant coefficients has a double root (critical case). If the differential equation takes the general form then the characteristic equation will be y + ay + by = 0, (44) λ + aλ + b = 0. (45) A double root appears if an only if the discriminant of Eq. (45) is zero, that is a 4b = 0, and, then, b = 1 4 a. The double root of the characteristic equation is λ = a/. Then, the first solution of the differential equation is y 1 = e ax/. (46) To find another solution y (x) the method of variation of parameters may be applied. The second solution takes the form y (x) = u(x)y 1 (x) where y 1 (x) = e ax/. Substituting y in the differential equation with b = a /4 we obtain u(y 1 + ay a y 1 ) + u (y 1 + ay 1 ) + u y 1 = 0. The expression in the first parentheses is zero because y 1 is a solution. The second parentheses is also zero because ( y 1 = a ) e ax/ = ay 1. The equation reduces to u y 1 = 0, and a solution is u = x. Consequently, the second solution is ( y (x) = xe λx λ = a ). (47) 16

17 We can observe that the solutions y 1 and y are linearly independent. This case can be summarized by the following theorem Theorem (Double root) [1]. In the case of a double root of Eq. (45) the functions (46) and (47) are solutions of Eq. (44). They constitute a fundamental system. The corresponding general solution is ( y = (c 1 + c x)e λx λ = a ). (48) Example. Solve the following differential equation y 4y + 4y = O. The double root of the characteristic equation is λ = 4. Then, the fundamental system of solutions is e x and xe x and the corresponding general solution is y = (c 1 + c x)e x. All three cases are summarized in the following table: Case Roots of Eq. (45) Fundamental General solution of Eq. (44) system of Eq. (44) I Distinct real e λ1x, e λ x y = c 1 e λ1x + c e λ x λ 1, λ II Complex conjugate e px cos qx y = e px (A cos qx + B sin qx) λ 1 = p + iq, e px sin qx λ = p iq III Real double root e λx, xe λx y = (c 1 + c x)e λx λ = a/ 3.6 Series solutions We consider the general homogeneous linear second-order equation P (x) d y dx + Q(x)dy + R(x)y = 0 (49) dx 17

18 with P (x) 0 in the interval α < x < β. We want to determine a polynomial solution y(x) of Eq. (49). Definition. A functions f(x) can be expanded in power series so that f(x) = a 0 + a 1 (x x 0 ) + a (x x 0 ) +... = a n (x x 0 ) n. (50) n=0 Such functions are said to be analytic at x = x 0 and the series (50) is called the Taylor series of f about x = x 0. The coefficients a n can be computed with the formula a n = f (n) (x 0 )/ n! where f (n) (x) = d n f(x)/dx n. We consider the functions P (x), Q(x), and R(x) as power series about x 0 P (x) = p 0 + p 1 (x x 0 ) +..., Q(x) = q 0 + q 1 (x x 0 ) +..., R(x) = r 0 + r 1 (x x 0 ) +... and y(x) = a 0 + a 1 (x x 0 ) Theorem []. Let the functions Q(x)/P (x) and R(x)/P (x) have convergent Taylor series expansions about x = x 0 for x x 0 < ρ. Then, every solution y(x) of the differential equation P (x) d y dx + Q(x)dy + R(x)y = 0 (51) dx is analytic at x = x 0, and the radius of convergence of its Taylor series expansion about x = x 0 is at least ρ. The coefficients a, a 3,... in the Taylor series expansion y(x) = a 0 + a 1 (x x 0 ) + a (x x 0 ) +... (5) are determined by plugging the series (5) into the differential equation (51) and setting the sum of the coefficients of the like powers of x in this expression equal to zero. Example. Solve the equation Assuming a solution of the form x d y dx + (x + x) dy dx y = 0. y = a k x k, k=0 18

19 we obtain and hence dy dx = k=0 ka k x k 1, d y dx = k=0 k(k 1)a k x k, k(k 1)a k x k + ka k x k+1 + ka k x k a k x k. k=0 k=0 k=0 k=0 The first, third, and fourth summations may be combined to give and hence there follows [k(k 1) + k 1]a k x k = (k 1)a k x k, k=0 k=0 (k 1)a k x k + ka k x k+1. k=0 k=0 In order to combine these sums, we replace k by n in the first and (k + 1) by n in the second, to obtain (n 1)a n x n + (n 1)a n 1 x n. n=0 n=1 Since the ranges of summation differ, the term corresponding to n = 0 must be extracted from the first sum, after which the remainder of the first sum can be combined with the second. In this way we find a 0 + [(n 1)a n + (n 1)a n 1 ]x n. n=1 In order that the previous relation may vanish identically, the constant term, as well as the coefficients of the successive powers of x, must vanish independently, giving the condition and the recurrence formula a 0 = 0 (n 1)[(n + 1)a n + a n 1 ] = 0 (n = 1,, 3,...). 19

20 The recurrence formula is automatically satisfied when n = 1. When n, it becomes a n = a n 1 n + 1 (n =, 3, 4...). Hence, we obtain a = a 1 3, a 3 = a 4 = a 1 3 4, a 4 = a 3 5 = a ,.... Thus, in this case a 0 = 0, a 1 is arbitrary, and all succeeding coefficients are determined in terms of a 1. The solution becomes ( ) y = a 1 x x 3 + x3 3 4 x If this solution is put in the form y = a 1 x = a 1 x ( x! x3 3! + x4 4! x5 5! +... [ ( x x )] 1! + x! x3 3! + x4 4!..., the series in parentheses in the final form is recognized as the expansion of e x, and, writing a 1 = c, the solution obtained may be put in the closed from ( e x ) 1 + x y = c. x In this case only one solution was obtained. This fact indicates that any linearly independent solutions cannot be expanded in power series near x = 0. That is, it is not regular at x = Regular singular points We consider the differential equations ) x d y dx + αxdy + βy = 0 (53) dx 0

21 which can be rewritten in the form A generalization of Eq. (54) is the equation d y dx + α dy x dx + β y = 0. (54) x d y dx + p(x)dy + q(x)y = 0 (55) dx where p(x) and q(x) can be expanded in series of the form p(x) = p 0 x + p 1 + p x + p 3 x +... q(x) = q 0 x + q 1 x + q + q 3 x + q 4 x... (56) Definition []. Equation (55) is said to have a regular singular point at x = 0 if p(x) and q(x) have series expansions of the form (56). Equivalently, x = 0 is a regular singular point of Eq. (55) if the functions x p(x) and x q(x) are analytic at x = 0. Equation (55) is said to have a regular singular point at x = x 0 if the functions (x x 0 ) p(x) and (x x 0 ) q(x) are analytic at x = x 0. A singular point of Eq. (55) which is not regular is called irregular. Example. Classify the singular points of Bessel s equation of order ν x d y dx + xdy dx + (x ν )y = 0, (57) where ν is a constant [1]. For x = 0 we have P (x) = x = 0. Hence, x = 0 is the only singular point of Eq. (57). Dividing both sides of Eq. (57) by x gives The functions d y dx + 1 ( ) dy x dx + 1 ν y = 0. x x p(x) = 1 and x q(x) = x ν are both analytic at x = 0. Hence Bessel s equation of order ν has a regular singular point at x = 0. 1

22 3.8 Nonhomogeneous linear equations Let us consider a second-order linear nonhomogeneous equation y + f(x)y + g(x)y = r(x). (58) A general solution y(x) of Eq. (58) can be obtained from a general solution y h (x) of the corresponding homogeneous equation y + f(x)y + g(x)y = 0, by adding to y h (x) any particular solution ỹ of Eq. (58) involving no arbitrary constant [1] y(x) = y h (x) + ỹ(x). (59) To show that y(x) is a solution of the nonhomogeneous differential equation we substitute Eq. (59) into Eq. (58). Then the left-hand side of Eq. (58) becomes or (y h + ỹ) + f(y h + ỹ) + g(y h + ỹ). (y h + fy h + gy h ) + ỹ + fỹ + gỹ. The expression in the parentheses is zero because y h is a solution of Eq. (59). The sum of the other terms is equal to r(x) because ỹ satisfies Eq. (58). Hence y(x) is a general solution of the Eq. (58). Theorem [1]. Suppose that f(x), g(x), and r(x) in Eq. (58) are continuous functions on an open interval I. Let Y (x) be any solution of Eq. (58) on I containing no arbitrary constants. Then Y (x) is obtained from Eq. (59) by assigning suitable values to the two arbitrary constants contained in the general solution y h (x) of Eq. (59). In Eq. (59), the function ỹ(x) is any solution of Eq. (58) on I containing no arbitrary constants. Proof. Let set Y ỹ = y. Then y + fy + gy = (Y + fy + gy ) (ỹ + fỹ + gỹ) = r r = 0, that is, y is a solution of Eq. (59) which does not contain arbitrary constants. It can be obtained from y h by assigning suitable values to the arbitrary constants in y h. From this, since Y = y + ỹ, the statement follows.

23 Theorem [1]. A general solution y(x) of the linear nonhomogeneous differential equation Eq. (58) is the sum of a general solution y h (x) of the corresponding homogeneous equation Eq. (59) and an arbitrary particular solution y p (x) of Eq. (58): y(x) = y h (x) + y p (x) (60). Example. Solve the equation y + y = sec x. The homogeneous equation y +y = 0 has the characteristic equation λ + 1 = 0 with roots λ 1 = i and λ = i, so, the general solution of the homogeneous equation is y = c 1 cos x + c sin x. Using the method of variation of parameter we have the following system of equations with the solution Thus by integrating, c 1 cos x + c sin x = 0, c 1 sin x + c cos x = sec x, c 1 = tan x, c = 1. c 1 = ln sec x + A 1, c = x + A, and the general solution is of the nonhomogeneous equation is y = A 1 cos x + A sin x cos x ln sec x + x sin x. 3.9 The method of variation of parameters This method can be applied to solve the nonhomogeneous equation of the form d y dx + p(x)dy + q(x)y = g(x), (61) dx 3

24 once the solutions of the homogeneous equation d y dx + p(x)dy + q(x)y = 0 (6) dx are known. Let y 1 (x) and y (x) be two linearly independent solutions of the homogeneous equation (6). We will try to find a particular solution ψ(x) of the nonhomogeneous Eq. (61) of the form [] ψ(x) = u 1 (x)y 1 (x) + u (x)y (x). (63) The differential equation (61) imposes only one condition on the two unknown functions u 1 (x) and u (x). We may impose an additional condition on u 1 (x) and u (x) such that the left hand side of the nonomogeneous equation be as simple as possible. Computing d dx ψ(x) = d dx [u 1y 1 + u y ] = [u 1 y 1 + u y ] + [u 1y 1 + u y ] we see that d ψ/dx will contain no second-order derivatives of u 1 and u if y 1 (x)u 1(x) + y (x)u (x) = 0. (64) Imposing the condition (64) on the functions u 1 (x) and u (x) the left hand side of the Eq. (61) becomes [u 1 y 1 + u y ] + p(x)[u 1 y 1 + u y ] + q(x)[u 1 y 1 + u y ] = u 1y 1 + u y + u 1 [y 1 + p(x)y 1 + q(x)y 1 ] + u [y + p(x)y + q(x)y ] = u 1y 1 + u y. If u 1 (x) and u (x) satisfy the two equations y 1 (x)u 1 + y (x)u (x) = 0 y 1(x)u 1(x) + y (x)u (x) = g(x), then ψ(x) = u 1 y 1 + u y is a solution of the nonhomogeneous equation (61). We solve the above system of equations as follows [ y1 (x)y (x) y 1(x)y (x) ] u 1(x) = g(x)y (x) [ y1 (x)y (x) y 1(x)y (x) ] u (x) = g(x)y 1 (x). 4

25 The function u 1(x) and u (x) are u 1(x) = g(x)y (x) W [y 1, y ](x) and u (x) = g(x)y 1(x) W [y 1, y ](x), (65) where W [y 1, y ](x) is the Wronskian of the solutions W [y 1, y ](x) = y 1 y y 1 y. Integrating the right-hand sides of Eqs. (65) we obtain u 1 (x) and u (x). Example. (a) Find a particular solution ψ(x) of the equation d y + 4y = 8 sin x (66) dx (b) Find the solution y(x) of Eq. (66) which satisfies the initial conditions y(0) = 1, y (0) = 1. (a) The functions y 1 (x) = cos x and y (x) = sin x are two linearly independent solutions of the homogeneous equation y + 4y = 0 with W [y 1, y ](x) = y 1 y y 1y = (cos x) cos x ( sin x) sin x = 1. Thus, from Eqs. (65), u 1(x) = 8 sin x and u (x) = 8 sin x cos x. (67) Integrating the first equation of (67) gives u 1 (x) = 8 = 4 sin x dx = 4 dx + 4 = 4x + sin x. cos x dx while integrating the second equation of (67) gives u (x) = 4 sin x dx = 4 5 (1 cos x) dx sin x dx = cos x.

26 Consequently, ψ(x) = cos x[ 4x + sin x] + sin x( cos x) is a particular solution of Eq. (66). (b) y(x) = c 1 cos x + c sin x + cos x( 4x + sin x) sin x cos x for some choice of constants c 1, c. The constants c 1 and c are determined from the initial conditions Hence, c 1 = 1, c = 3 and 1 = y(0) = c 1 and 1 = y (0) = c. y(x) = cos x + 3 sin x + cos x( 4x + sin x) sin x cos x. 4 Differential equations of arbitrary order 4.1 Homogeneous linear equations A linear differential equation of nth order can be written in the following general form y (n) + f n 1 (x)y (n 1) f 1 (x)y + f 0 (x)y = r(x) (68) where the function r on the right-hand side and the coefficient f 0, f 1,..., f n 1 are any given functions of x, and y (n) is the nth derivative of y. Eq. (68) is said to be homogeneous if r(x) = 0. Then, Eq. (68) becomes y (n) + f n 1 (x)y (n 1) f 1 (x)y + f 0 (x)y = 0. (69) If r(x) 0, Eq. (68) is said to be nonhomogeneous. A function y = φ(x)is called a solution of a differential equation of nth order on an interval I if φ(x) is defined and n times differentiable on I and is such that the equation becomes an identity when we replace the unspecified 6

27 function y and its derivatives in the equation by φ and its corresponding derivatives [1]. Existence and uniqueness theorem [1], [3]. If f 0 (x),..., f n 1 (x) in Eq. (69) are continuous functions on an open interval I, then the initial value problem consisting of the equation Eq. (69) and the n initial conditions y(x 0 ) = K 1, y (x 0 ) = K,..., y (n 1) (x 0 ) = K n, has a unique solution y(x) on I; here x 0 is any fixed point in I, and K 1,..., K n are given numbers. A set of functions, y 1 (x),..., y n (x) are linearly dependent on some interval I where they are defined, if one of them can be represented on I as a linear combination of the other n 1 functions. Otherwise the functions are linearly independent on I. A fundamental system or a basis of solutions of the linear homogeneous equation Eq. (69) is a set of n linearly independent solutions y 1 (x),..., y n (x) of that equation. If y 1,..., y n is such a fundamental system, then y(x) = c 1 y 1 (x) c n y n (x) (c 1,..., c n arbitrary) (70) is a general solution of Eq. (69) on I. The test for linear dependence and independence of solutions can be generalized to nth order equations as follows Theorem [1]. Suppose that the coefficients f 0 (x),..., f n 1 (x) of Eq. (69) are continuous on an open interval I. Then n solutions y 1,..., y n of Eq. (69) on I are linearly dependent on I if and only if their Wronskian W (y 1,..., y n ) = y 1 y... y n y 1 y... y n y (n 1) 1 y (n 1)... y n (n 1) (71) is zero for some x = x 0 in I. (If W = 0 at x = x 0, then W 0 on I). Theorem [1]. Let Eq. (70) be a general solution of Eq. (69) on an open interval I where f 0 (x),..., f n 1 (x) are continuous, and let Y (x) be any solution of Eq. (69) on I involving no arbitrary constants. Then Y (x) is obtained from Eq. (70) by assigning suitable values to the arbitrary constants c 1,..., c n. 7

28 Example. The equation y y y + y = 0. (7) has the solutions y 1 = e x, y = e x, and y 3 = e 3x. The Wronskian is e x e x e 3x W (e x, e x, e 3x ) = e x e x 3e 3x = e 6x 0, e x 4e x 9e 3x which shows that the functions constitute a fundamental system of solutions of Eq. (7). The corresponding general solution is y = c 1 e x + c e x + c 3 e 3x. 4. Homogeneous linear equations with constant coefficients A linear homogeneous equation of order n with constant coefficients y (n) + a n 1 y (n 1) a 1 y + a 0 y = 0, (73) has the correspondent characteristic equation λ n + a n 1 λ n a 1 λ + a 0 = 0. (74) If this equation has n distinct roots λ 1,..., λ n, then the n solutions y 1 = e λ 1x,..., y n = e λnx (75) constitute a fundamental system for all x, and the corresponding general solution of Eq. (73) is If λ is a root of order m, then y = c 1 e λ 1x c n e λn x. (76) e λx, xe λx,..., x m 1 e λx (77) are m linearly independent solutions of Eq. (73) corresponding to that root. 8

29 Example. Consider the differential equation The characteristic equation y + 3y 4y 1y = 0. λ 3 + 3λ 4λ + 1 = 0 has the solutions λ 1 =, λ =, and λ 3 = 3, and the corresponding general solution Eq. (76) is y = c 1 e x + c e x + c 3 e 3x. 4.3 Linear differential equations in state space form The nth-order differential equation a n (t) dn y dt n + a n 1(t) dn 1 y dt n a 0y = 0, can be transformed into a system of n first order equations. With the notations we obtain the system x 1 (t) = y, x (t) = dy/dt,... x n (t) = d n 1 y/dt n 1, dx 1 dt = x, dx dt = x 3,..., dx n 1 dt = x n, and dx n dt = a n 1(t)x n + a n (t)x n a 0 x 1. a n (t) A system of n first-order linear equations has the general form dx 1 dt = a 11(t)x a 1n (t)x n + g 1 (t),. dx n = a n1 (t)x a nn (t)x n + g n (t), dt (78) 9

30 and is said to be nonhomogeneous (g i (t) 0, i = 1,..., n). The system dx 1 dt = a 11(t)x a 1n (t)x n,. dx n dt = a n1 (t)x a nn (t)x n, is said to be homogeneous (g i (t) = 0, i = 1,..., n). The homogeneous linear system with constant coefficients (a ij depend on t) dx 1 dt = a 11x a 1n x n,. dx n = a n1 x a nn x n, dt can be written in matrix notation as (79) do not (80) ẋ = A x, (81) where x = x 1 x. x n and A = a 11 a 1... a 1n a 1 a... a n... a n1 a n... a nn. Theorem (existence-uniqueness theorem) []. There exists one, and only one, solution of the initial-value problem for < t < ẋ = Ax, x(t 0 ) = x 0 = x 0 1 x 0. x 0 n. (8) The dimension of the space of all solutions of the homogeneous linear system of differential equations (81) is n. 30

31 4.3.1 Solution via the eigenvalue-eigenvector method Consider the linear homogeneous differential system ẋ = Ax. (83) Assuming a solution of the form x(t) = e λt v, v = constant vector. Eq. (83) becomes or λe λt v = e λt Av, Av = λv. (84) The solution of Eq. (83) is x(t) = e λt v if, and only if, λ and v satisfy Eq. (84). A vector v 0 satisfying Eq. (84) is called an eigenvector of A with eigenvalue λ. The eigenvalues λ of A are the roots of the equation det(a λi) = det a 11 λ a 1... a 1n a 1 a λ... a n... a n1 a n... a nn λ = 0. Case I. Distinct eigenvalues The matrix A has n linearly independent eigenvectors v 1,..., v n with distinct eigenvalues λ 1 λ... λ n 1 λ n. For each eigenvalue λ j we have an eigenvector v j and a solution of Eq. (83) is of the form x j (t) = e λ jt v j. There are n linearly independent solutions x j (t) of Eq. (83). Then the general solution of Eq. (83) is given by x(t) = c 1 e λ 1t v 1 + c e λ t v c n e λnt v n. (85) Case II. Complex eigenvalues If λ = α + iβ is a complex eigenvalue of A with eigenvector v = v 1 + iv, then a complex-valued solution of Eq. ( 83) is x(t) = e λt v. 31

32 Lemma []. Let x(t) = y(t) + iz(t) be a complex-valued solution of Eq. (83). Then both y(t) and z(t) are real-valued solutions of Eq. (83). The function x(t) can be written as x(t) = e (α+iβ)t (v 1 + iv ) = e αt (cos βt + i sin βt)(v 1 + iv ) = e αt [(v 1 cos βt v sin βt) + i(v 1 sin βt + v cos βt)]. If λ = α + iβ is an eigenvalue of A with eigenvector v = v 1 + iv, then and y(t) = e αt (v 1 cos βt v sin βt) z(t) = e αt (v 1 sin βt + v cos βt) are two real-valued solutions of Eq. (83). Case III. Equal eigenvalues If the matrix A does not have n distinct eigenvalues, then A may not have n linearly independent eigenvectors. Let us assume that the n n matrix A has only k < n linearly independent eigenvectors. In this case Eq. (83) has only k linearly independent solutions of the form e λt v. To find additional solutions we present the following method as described in []: 1. We pick an eigenvalue λ of A and find all vectors v for which (A λi) v = 0, but (A λi)v 0. For each such vector v e At v = e λt e (A λi)t = e λt [v + t(a λi)v] is an additional solution of Eq. (83). The process is repeated for all eigenvalues of A.. If we still do not have enough solutions, then we find all vectors v for which (A λi) 3 v = 0, but (A λi) v 0. For each such vector v, e At v = e λt [ ] v + t(a λi)v + t! (A λi) v is an additional solution of Eq. (83). 3. We keep proceeding in this fashion until n linearly independent solutions are obtained. 3

33 4.3. Fundamental solution matrix Definition. A matrix X(t) whose columns are x 1 (t),..., x n (t), the n linearly independent solutions of Eq. (81) X(t) = [ x 1 (t) x (t)... x n (t) ]. is called the fundamental solution matrix of Eq. (81) Every solution x(t) can be written in the form x(t) = c 1 x 1 (t) + c x (t) c n x n (t) (86) In the matrix vector form, equation (86) can be written as x(t) = X(t)c, where c is a constant vector. Example []. Find a fundamental matrix solution of the system of differential equations ẋ = x. It can be verified that the three linearly independent solutions of the system are given by e t 1 4 1, e 3t 1 1 and e t Therefore, the fundamental matrix solution for the system is X(t) = e t e 3t e t 4e t e 3t e t e t e 3t e t Theorem []. Let X(t) be a fundamental solution matrix of the differential equation ẋ = Ax. Then where e At is also a fundamental solution matrix.. e At = X(t)X 1 (0). (87) 33

34 We consider the example given in [] and show as to how e At can be computed. In Eq. (81) let A = The eigenvalues are computed from the relation 1 λ 1 1 p(λ) = det(a λi) = det 0 3 λ = (1 λ)(3 λ)(5 λ) λ Thus we have 3 distinct eigenvalues λ = 1, λ = 3, and λ = 5. eigenvectors corresponding to those eigenvalues, respectively, are v 1 = 0 v = v 3 =. 0 0 The The three linear independent solutions of ẋ = Ax are 1 1 x 1 (t) = e t 0 x (t) = e 3t x 3 (t) = e 5t 0 0 The fundamental solution matrix is We compute X 1 (0) = and from the theorem X(t) = e At = X(t)X 1 (0) = e t e 3t e 5t 0 e 3t e 5t 0 0 e 5t 1 = , e t 1 et + 1 e3t 1 e3t + 1 e5t 0 e 3t e 3t + e 5t 0 0 e 5t

35 4.3.3 The nonhomogeneous equation The initial-value problem for a nonhomogeneous equation is ẋ = Ax + f(t), x(t 0 ) = x 0. (88) Applying variation of parameter method, the solution is assumed of the form x(t) = X(t)u(t), where X(t) = [x 1 (t),..., x n (t)], and u(t) = Using this relation Eq. (88) yields Since matrix X(t) satisfies we obtain u 1 (t). u n (t). Ẋ(t)u(t) + X(t) u(t) = AX(t)u(t) + f(t). (89) Ẋ(t) = AX(t), (90) X(t) u(t) = f(t). (91) Matrix X(t) is nonsingular (X 1 (t) exists) and therefore u(t) = X 1 (t)f(t). (9) Integrating this expression between t 0 and t we have u(t) = u(t 0 ) + t = X 1 (t 0 )x 0 + t 0 X 1 (s)f(s) ds (93) t t 0 X 1 (s)f(s) ds. (94) 35

36 Consequently, If X(t) = e At then t x(t) = X(t)X 1 (t 0 )x 0 + X(t) x(t) = e A(t t 0) x 0 + t t 0 X 1 (s)f(s) ds. (95) t 0 e A(t s) f(s) ds. (96) Example. Find the solution of the initial value problem ẋ = [ ] 1 1 x [ e t 0 ], x 0 = [ ] 1 1 From the homogeneous problem we can easily show that the fundamental solution matrix is given by [ e t te X(t) = t 0 e t ]. It is easily verified that Ẋ = AX and X(0) = I. and t 0 X 1 (s) = [ ] 1 s e s. 0 1 [ 1 X(t) X 1 (s)f(s) ds = (et e t ] ). 0 Then from Eq. (95) the solution is given by [ (t 1)e t ] [ 1 x(t) = e t + (et e t ] ) Equilibrium and stability Consider the differential equation ẋ = f(t, x), (97) 36

37 where and f(t, x) = x = x 1 (t). x n (t), ẋ = dx dt, f 1 (t, x 1,..., x n ). f n (t, x 1,..., x n ), is a nonlinear function. In general, Eq. (97)cannot be solved explicitly. However, one can easily determine the qualitative properties of solution of Eq. (97) in the neighborhood of an equilibrium point. The equilibrium points are the values x 0 = x 0 1. x 0 n for which, x(t) = x 0 is a solution of Eq. (97). Observe that ẋ(t) is identically zero if x(t) x 0. The value x 0 is an equilibrium of Eq. (97), if, and only if, f(t, x 0 ) 0. (98) Example.[6] Find all equilibrium values of the system of differential equations The value dx 1 dt = 1 x, x 0 = [ x 0 1 dx dt = x3 1 + x. x 0 ] 37

38 is an equilibrium value if, and only[ if, 1 ] x 0 = 0 and (x 0 1) 3 + x 0 = 0. This 1 yields x 0 = 1 and x 0 1 = 1. Hence is the only equilibrium solution of 1 this system. Stability: Let φ(t) be a known solution of Eq. (97). Suppose that ψ(t) is a second solution with ψ(0) very close to φ(0) such that β(t) ψ(t) φ(t) can be viewed as the disturbance on φ(t). The concept of stability is important in many applications. Consider the equation of motion of a simple pendulum of mass m and length l given by d y dt + g l sin y = 0, where y is the angular displacement from the vertical axis and g is acceleration due to gravity. With the notation x 1 = y and x = dy/dt we have dx 1 dt = x, dx dt = g l sin x 1. (99) The system of Eq. (99) has equilibrium solutions {x 1 = 0, x = 0}, and {x 1 = π, x = 0}. If we disturb the pendulum slightly from the equilibrium position {x 1 = 0, x = 0}, then it will oscillate with small amplitude about x 1 = 0. If we disturb the pendulum slightly from the equilibrium position {x 1 = π, x = 0}, then it will either oscillate with very large amplitude about x 1 = 0, or it will rotate around and around. The two solutions have very different properties, and, intuitively, we would say that the equilibrium value {x 1 = 0, x = 0} is stable, while the equilibrium point {x 1 = π, x = 0} is unstable. In the case when f(t, x) does not depend explicitly on t i.e. f = f(x) the differential equations are called autonomous. 4.5 Phase-plane Let us consider a two dimensional system dx dt = f(x, y), dy dt = g(x, y). (100) 38

39 Every solution x = x(t), and y = y(t) of Eq. (100) defines a curve in the three-dimensional space {t, x, y}. For example the solution of the system of differential equations dx dt = y, dy dt = x, is x = cos t, y = sin t. This solution describes a helix in three-dimensional space {t, x, y}. Every solution x = x(t), and y = y(t), of Eq. (100), for t 0 t t 1, also defines a curve in the x y plane. This curve is called the orbit, or trajectory, of the solution x = x(t), y = y(t), and the xy plane is called the phase-plane of the solutions of Eq. (100). In the general case let x(t) be a solution of the vector differential equation ẋ = f(x), x = x 1. x n, f(x) = f 1 (x 1,..., x n ). f n (x 1,..., x n ) (101) on the interval t 0 t t 1. As t runs from t 0 to t 1, the set of points (x 1 (t),..., x n (t)) trace out a curve C in the n-dimensional space x 1, x,..., x n. This curve is called the orbit of the solution x = x(t), for t 0 t t 1, and the n-dimensional space x 1,..., x n is called the phase-space or state-space of the solution of Eq. (101). 4.6 Linear approximation at equilibrium points [5] Consider again the Eq. (100) dx dt = f(x, y), dy dt = g(x, y), with f(0, 0) = g(0, 0) = 0 as the equilibrium point. Using Taylor expansion about this point, we can write f(x, y) = ax + by + P (x, y), g(x, y) = cx + dy + Q(x, y), where P (x, y) = O(r ) and Q(x, t) = O(r ) as r = x + y 0, and a = f (0, 0), x b c = g (0, 0), x d = f (0, 0), y (10) = g (0, 0). y (103) 39

40 The linear approximation of Eq. (100) in the neighbourhood of the origin is defined as the system ẋ = ax + by, ẏ = cx + dy, or where [ a b A = c d ] ẋ = A x (104), x = [ x y ], ẋ = [ ẋ ẏ ]. (105) The solutions of Eq. (104) are geometrically similar to those of Eq. (100) near the origin unless one (or more) of the eigenvalues of A is zero or has zero real part. The two linearly independent solutions are of the form x = u e λt, (106) where u = [ r s ] 0. (107) Then ẋ = λ u e λt, and equations (104) and (106) yield (A λi) u = 0 (108) where I is the identity matrix. With u 0 and Eq. (108), we have det(a λi) = 0, or a λ c b d λ = 0. (109) 40

41 The two eigenvalues are given by the solution of the quadratic equation λ (a + d)λ + (ad bc) = 0. (110) The solutions of Eq. (108) are the eigenvectors: u 1 corresponding to λ 1, and u corresponding to λ. The general solution of Eq. (104) is x = C 1 u 1 e λ 1t + C u e λ t, for λ 1 λ. (111) Using the nonsingular linear transformation Eq. (104) becomes x 1 = Sx; S = [u 1 u ], (11) ẋ 1 = SAS 1 x 1 = Bx 1, (113) where B is diagonal or in Jordan form. The topological character of the transformed equilibrium point at the origin is not affected in the new variable x 1 = [x 1, y 1 ] T. The equations in the new coordinates are simpler. Case I. λ 1 λ 0 and λ 1, λ R (real) We can choose S so that ẋ 1 = λ 1 x 1, ẏ 1 = λ y 1, and then the equation for the phase paths is The solutions are dy 1 dx 1 = λ λ 1 y 1 x 1. y 1 = C x 1 λ /λ 1, where C = arbitrary. The origin is a node (Figure 1) when λ /λ 1 > 0. The node is stable when λ 1, λ < 0 (Figure 1) and unstable when λ 1, λ > 0. 41

42 y x Figure 1: Stable node y x Figure : Saddle point The origin is a saddle-point (Figure ) when λ /λ 1 < 0. Case II. λ 1 = λ = λ (b and c not both zero) We can choose S so that ẋ 1 = λx 1 + y 1, ẏ 1 = λy 1, λ R, and then the equation for the phase paths is The solutions are dy 1 dx 1 = λy 1 λx 1 + y 1. y 1 = 0, x 1 = 1 λ y 1 log e y 1 + Cy 1 where C = arbitrary. The origin is a inflected node, stable if λ < 0 (Figure 3) and unstable if λ > 0. 4

43 y x Figure 3: Stable inflected node Case III. λ 1 = λ = α + iβ with β 0 We can choose S so that the equations become ẋ 1 = αx 1 βy 1, ẏ 1 = βx 1 + αy 1. With z(t) = x 1 (t)+i y 1 (t) = r(t)e iθ(t) we have ż = (α+iβ)z, and r(t) = z(t). The equations in polar coordinates are ṙ = αr, θ = β. The origin is a stable spiral (or focus) if α < 0, β 0 (Figure 4), and an unstable spiral if α > 0, β 0. y x Figure 4: Stable spiral The origin is a center if α = 0, β 0, (Figure 5). We can sumarize all the above cases in the following table [5], Figure 6 43

44 y x Figure 5: Center (1) λ 1, λ real, unequal, same sign Node () λ 1 = λ (real)b 0, c 0 Inflected node (3) λ 1, λ complex, non-zero real part Spiral (4) λ 1 0, λ = 0 Parallel lines (5) λ 1, λ real, different sign Saddle point (6) λ 1, λ pure imaginary Center q (3) (3) () () Δ= 0 (6) (1) (1) (4) (4) p Stable points Unstable points (5) Figure 6: General classification Example. Classify the equilibrium point at (0,0) for the system ẋ = e x 3y 1, ẏ = x(1 y ). 44

45 Using Taylor expansion for the exponential function, the linearized system of equations about (0,0) is or in matrix form ( ẋ ẏ ẋ = x 3y, ) = ( ẏ = x, ) ( x y ). are real with different sign. The equi- The eigenvalues are λ 1, = 1 ± 17 librium is a saddle point. 5 Partial differential equations The word ordinary in ordinary differential equation distinguishes it from partial differential equation (PDE), involves partial derivatives of two or more independent variables. For a first order partial differential equation, a unified general theory exists; however, this a case for higher order partial differential equations. Generally speaking, the second order PDEs may be classified into three following categories, viz., elliptic, hyperbolic, and parabolic types. 5.1 Normal forms of elliptic, hyperbolic, and parabolic Equations Consider a linear second order differential operator for the function u(x, y) given by L(u) = a u x + b u x y + c u y, (114) where a, b, and c are either constants or functions of x and y. A corresponding quasilinear PDE may be represented by L(u) + g(x, y, u/ x, u/ y) = L(u) +... = 0, (115) where g(x, y, u/ x, u/ y) is not necessarily linear and does not contain any second derivative. 45

46 Let us introduce the transformations ξ = αx + βy, Therefore, L(u) in Eq. (114) takes the form η = γx + δy. (116) L(u) = (aα + bαβ + cβ ) u ξ +(aαγ + b(αδ + βγ) + cβδ) u ξ η +(aγ + bγδ + cδ ) u η. (117) If the transformed operator is desired to be of the form u, then we need ξ η aα + bαβ + cβ = 0, (118) aγ + bγδ + cδ = 0. (119) If a = c = 0, then the trivial transformation ξ = y and η = y provides the desired form. For the non-trivial case either a or c or both are non-zero. Let us say a 0, thereby implying that α 0, γ 0. Dividing Eq. (118) by β and Eq. (119) by δ, we obtain two quadratic equations in (α/β) and (γ/δ). These yield α/β = 1 a { b ± b 4ac}, (10) γ/δ = 1 a { b ± b 4ac}. (11) The ratios α/β and γ/δ must be different (by choosing positive sign in Eq. (10) and negative sign in Eq. (11)) so that the transformation given by Eq. (116) is non-singular. Further b 4ac should be positive. Therefore, L(u) reduces to the form u if and only if ξ η b 4ac > 0, (1) 46

47 and this case is said to be hyperbolic. Then the transformation Eq. (116) takes the form ξ = ( b + b 4ac)x + ay, Then the PDE given by Eq. (115) reduces to η = ( b b 4ac)x + ay. (13) 4a(b 4ac) u u + g(ξ, η, ξ η ξ, u ) = 0. (14) η If b 4ac = 0, then L is termed as parabolic. In this case Eq. (10) and Eq. (11) reduce to a single equation and α/β = b/a forces the coefficient of u/ξ in Eq. (117) to vanish. Further, since b = 4ac or b/a = c/b, the u coefficient of also vanish. Thus the transformation (c.f. Eq. (13)) ξ η ξ = bx + ay, can be used to transform Eq. (115) into η = x (arbitrary), (15) a + g( ) = 0. (16) η This is the normal form of a parabolic quasilinear PDE. For the final case, b 4ac < 0, and the operator L(u) is said to be elliptic. In this case it is not possible to eliminate the coefficients of u ξ or u. Nevertheless, if we use the transformation η then L(u) = a ξ = ay bx, η = t (arbitrary), (17) 4ac b ( ) u ξ + u, and the general PDE has the form η ( ) u a ξ + u + g(ξ, η, u η ξ, u ) = 0. (18) η 47

48 For the linear case u ξ + u = 0, (19) η which is the well-known Laplace s equation. Once a PDE has been reduced to its normal form, the method of characteristic may be effectively used to find its solution. However, in the following we discuss the solution of a particular hyperbolic equation, known as the wave equation by the use of separation of the variables which is a popular approach in engineering. The equation c u(x, t) x ü(x, t) = 0, c = constant, (130) is a partial differential equation. The following notation was used The initial conditions are ü(x, t) = u(x, t) t. The boundary conditions are u(x, 0) = f(x), u(x, 0) = g(x). (131) u u (0, t) = (l, t) = 0. (13) x x We seek the solution of Eq. (130) in the form of a product of a function of time and a function of position u(x, t) = U(x)ϕ(t). (133) Introducing (133) into (130), we replace Eq. (130) by the system of two ordinary equations ϕ + β c ϕ = 0, (134) d U dx + β U = 0, (135) 48

49 where β is for the time being an undetermined parameter. The solution of Eqs. (134) and (135) is ϕ(t) = A sin ωt + B cos ωt, (136) U(x) = C sin βx + D cos βx, (137) where ω = βc. We first consider the second boundary conditions (13). They imply that C = 0 and The latter condition is satisfied if Dβ sin βl = 0. β n = nπ, (n = 0, 1,,..., ). (138) l It is evident that every value of β n is associated with a particular solution of Eq. (130), viz. u n (x, t) = (A n sin ω n t + B n cos ω n t)d cos β n x. (139) The general solution of (130) takes the form u(x, t) = u n (x, t) = (A n sin ω n t + B n cos ω n t)d cos β n x. (140) n=0 n=0 The constants A n, B n are to be found from the initial conditions (131) i.e. f(x) = DB n cos β n x, g(x) = Dω n A n cos β n x. (141) The functions n=0 n=0 U n (x) = D cos β n x, (14) are the eigenfunctions of the problem. They are orthogonal, i.e. l 0 l 0 U n (x)u m (x) dx = 0, if n m, U n(x) dx = l D, if n = m, (143) 49

50 as can easily be verified by integration. The constant D is arbitrary. Assume that D = /l. Then l Un(x) dx = 1 and the eigenfunctions U n (x) = 0 l cos β nx = l nπx cos, (144) l are called normalized eigenfunctions. Making use of the normalized eigenfunctions we can rewrite relations (141) in the form f(x) = l B n cos β n x, g(x) = n=0 l ω n A n cos β n x. (145) n=0 To find the coefficient B n we multiply the first equation (145) by cos β n x and integrate with respect to x from 0 to l. Then, making use of the orthogonality relations, we obtain l B n = f(x) cos β n x dx, (n = 1,,..., ), l B 0 = 1 l 0 l 0 f(x) dx. (146) Similarly we have A n = 1 l g(x) cos β n x dx, (n = 1,,..., ), cβ n l 0 A 0 = 0. (147) Introducing the values of A n, B n into Eq. (140), we arrive at the final solution. Example. We consider next the equation c u ü = 0, (148) x with assuming homogeneous initial conditions (u(x, 0) = u(x, 0) = 0) and the boundary conditions u(0, t) = 0, u (l, t) = P (t). (149) x 50