Krasnov-Makarenko-Kiselev Problems and Exercises in The Calculus of Variations

Uploaded by

Santiago Reyes

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF or read online on Scribd

100% found this document useful (2 votes)

489 views221 pages

Krasnov-Makarenko-Kiselev Problems and Exercises in The Calculus of Variations

Uploaded by

Santiago Reyes

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF or read online on Scribd

You are on page 1/ 221

M. L. KRASNOV, G. I. MAKARENKO, A. I. KISELEV PROBLEMS AND EXERCISES IN THE CALCULUS OF VARIATIONS Translated from the Russian by GEORGE YANKOVSKY MIR PUBLISHERS MOSCOWFirst published 1972 Revised from the 197) Russian edition Ha anzauicnom asure © English translation, Mir Publishers, 4975CONTENTS Introduction 2 6 2 ee ee ee ee Preliminary Remarks ae CHAPTER I. THE EXTREMUM OF A FUNCTION OF MANY VARIABLES 4. Absolute Extremum . - -- ee ee ee ee . 2. Conditional Extremum.....-..--.--- CHAPTER IJ. EXTREMA OF FUNCTIONALS 3. The Functional. The Variation of a Functional and Its Properties 2. 2 ee ee ee ee - AD, Elementary Problem in the Caloulus of Variations. Euler’s Equation. . Generalizations of the Ble Calculus of Variations. |. Invariance of Euler's Equi . Field of Extremals 2... 2.2... ee eee - Sufficient Conditions for the Bxtremum of a Functional . Conditional Extremum ... . 40. Moving Boundary Problems .. . . the Discontinuous Problems. One-Sided Varia : . The Hamilton-Jacobi Theory. The Variational Prin: ciples of Mechanics 2... 2... 7 ee CHAPTER Il]. DIRECT METHODS IN THE CALCULUS OF VARIATIONS 43. Euler's Finite-Difference Method 44. Ritz Method. Kantorovich Method .. 2... 2. 45. Variational Methods for Finding Eigenvalues and Bigenfunctions ANSWERS AND HINTS... 2. ..-2.-.....2--- BIBLIOGRAPHY ......-. 2.200222 0005 INDEX Done enINTRODUCTION The modern engineer often has to do with problems that require a sound mathematical background and set skills in the use of diverse mathematical methods. Expanding the mathematical outlook of the engineer contributes ap- preciably to new advances in technology. The calculus of variations is one of the most important divisions of classical mathematical analysis as regards applications. The authors’ aim was to supply the reader with a certain minimum of problems covering the basic divisions of the classical calculus of variations, and they deliberately avoided questions pertaining to the theory of optimal control, The present text was specially designed for students using “Differential Equations and the Calculus of Varia- tions” by L. E. Elsgolts (Mir Publishers, Moscow, 1970) and the reference book “The Calculus of Variations and Integral Equations” by L. Ya. Tslaf. The authorsPRELIMINARY REMARKS 1. If A is an arbitrary set of elements, then the assertion “element @ belongs to (is 2 member of) the set A” is symbolized as follows: 2 € A. The symbolism a €.4 (or a F A) is used to denote that the element ¢ does not belong to (is not a member of) the set 4. Ii 4 and & aro sets, then the assertion “A ts a subset of the set B” (symbolized thus: 4 c &) signifies that any element 2 of tho set A is algo a member of the set J. * f is The union and intersection of two sets A and B are defmed as ollows: the union A) B=(z]rCA or z € B} is the collection of ole- ments x belonging to at least one of the sets A and By _ the intersection A 1) B= {x | a2 €A, x & B}is the sot of elements x lying both in A and in B. 3. If 4 is a certain set of roal numbers, then the least upper bound (supremum) of 4 is the smallest real number M such that a < M for all a 6 A. In other words, Mf is the least upper bound of 4 if for any 2€A wehavec < M, but no matter what ¢ > 0, even if it is arbitra~ rily small, there will be at least one element 6 €.A such that Mf —~ Ba8 PRELIMINARY REMARKS (8) a (2 + y) = az + ay. 3. A linear space R is a normed linear space if to every element zE€R there is associated a nonnegative real number || < Ij called the norm of that element, and: (1) Ua = 0 only when x = 0; 8 i et yl g ten £ \'y 4 (triangle axiom f ris). t~ zil~Ty" ni ‘J or 710) J 6. A sot i of elements 2, y. 2 » wt any nature whatsoever is a metric space if to each pair of elements z, 7 in M there is associated a nonnegative real number p (=, 2 such chat . (4) p (x, y) = 0 if and only if x = y (identity axiom); BEER SEO See Se ny a, - @ (ay 5) hi Je Theaurnber p (2.4) ig called the distance between the clements.x and y- Every normed linear space is a metric space, for it suffices to put P@ y) = Ile—y i ‘The space C [a, d} is the space of all functions y (2) continuous on [a, 3]: = y Hy lls xa ty @) | Tho space Cy la, 5] is the space of all functions y (z) continuous on [2, 6] tonether ‘with their fist derivatives: Wy they = max, Ty 2) |b mare Ly @] The space , la, YI ig the space of all functions y (z) continuous on {e, 6] together with derivatives up to order x inclusive (n is a fixed positive integer): n We log i max | yh @! Sometimes the norm of the element y (z) in C, [a, 4] is defined thus: Hy fis= max {19 (2) fy’ C2) fe oes [x @)T}CHAPTER I THE EXTREMUM OF A FUNCTION OF MANY VARIABLES 4. ABSOLUTE EXTREMUM Let a function f(x, 22, .-.. %) or, briefly, f(z), be given in a domain D of Euclidean n-dimensional space E*. We will say that the function f (x) attains its maximum (minimum) value at a point zy € D ii, no matter what the point z € D, we have F@)F(x)) holds for all points +€ Q(x) () D, Bk ohh, Characteristic of the point of strict maximum (or point of strict minimum) is the fact that Af = f @) — f (#) <0 (or Af > 0) or all €Q (x) 1) D, x x. If for the point 2 there exists a neighbourhood Q (x) such that for all points z €Q (x) | D the condition (2) Sf (e) (or f (=) SF (2) holds, then the point10 CH. 1 THE EXTREMUM OF A FUNCTION OF MANY VARIABLES 2 is simply called the maximum point ini) oir. point (or the minimum Definition 2. The maximum point and the minimum point of a function f(x) are generically termed the extremum Points of that fonetson. & ¥ 8 are 1. Use the definition to find the extremum points of the functions: @) Ff Ge me) = af + 23; tick, xt i560. Har 4) = { tte at > (b) F(a, 22) 4, aipaies0; (0) f Gy m2) = ap 23 in the domain D {zi +23<1}. Theorem { (necessary condition for an extremum). Let @ function f (x), 2 = (a, Za, «+ +1 Bn), be defined in a neighbourhood of the point 2 = (28, af, ..., ah). Jf this point is an extremum point of the function f (x) and if at this point there exist derivatives z f= 1, 2, ..., n), then they are equa lto zero: AF (x60 ie OQ (fad, m) If the function f(a) is differentiable at the extremum. point 2, then its differential is zero at this point: df (2°) = 0. Example 1. Find the extremum points of the function gang by? Solution. The extremum points are among those points for which dz = 0. In our case, dz = 2% dx + 2y dy. The condition dz = 0 holds true in the single point 2 = 0, y=. Indeed, if z= y=0, then dz = 0. Conversely, Jet dz = 0; taking advantage of the arbitrariness of dx and dy, choose dy = 0. Then Q = dz = 2x dx and, by virtue of the arbitrariness of dz, it follows that z = 0. In similar fashion, we find that y == 0. At the point (0, 0) we have 220, at all other points, z= 2? + y?>0. Therefore, the point (0, 0) is the strict minimum point for the function gee gto yf1. ABSOLUTE EXTREMUM 44 If, by including functions that are not differentiable at separate points, we extend the class of functions in which an extremum is sought, we arrive at the following necessary condition for an extremum. If x is an extremum point of the function f (a, Za, . « « Za), then at this point each partial derivative 2 (ied, 2, 0... 0) i is either zero or does not exist. Example 2. We consider the upper nappe of the cone z =z g? + y®, 2 > 0. It is obvious that the function z has a minimum at the point O (0, 0). But z and re do not exist at this point, ° Definition 3. Points at which the necessary condition for an extremum of a function f(z) is fulfilled are called the critical points of the function. The points 2 at which df (2) = 0 are called stationary points of the function f (x). The condition df (x) == 0 is equivalent to the condition (0) FEE 0 (ise 4, 2,02, 0) The existence of a critical point does not guarantees the existence of an extremum of the function. For example, with respect to the function 2 = 2? — y” the point (0, 0) is a stationary point, but there is no extremum of the function z at that point: in any arbitrarily small neighbourhood of the point (0, 0) the function assumes both positive and negative values. 1°. Sufficient conditions for a strict extremum. Definition 4. The quadratic form n A(a)= Ala, Xa, 0.0 Sa) = 2h, wee ie ays ays t, fxd, 2, ory is positive definite (or negative definite) if A (x) >0 (or A (z) <0) for any point z € E*, zs 0, and vanishes only when z==0, that is, when 2, = 2,=...2%a, 0, A quadratic form is nonnegative if it never-assumes negative values. For example, the forms 2} + aj +... +42 CH. THE EXTREMUM OF A FUNCTION OF MANY VARIABLES and (2, +2, -+...-+2,)* are nonnegative forms. The former is positive definite since it vanishes only when ay = ty... == a, = 0, the latter form will not be positive definite since it vanishes, for example, when = 4, t= wd, oy = y=. = Ty = 0. A quadratic form which is positive or negative definite is called a definite quadratic form. A quadratic form that takes on both positive and nega~ tive values is called an indefinite quadratic form. Theorem 2 (sufficient conditions for a strict extremum), Let @ function f (x) be defined and have continuous second derivatives in some. neighbourhood of a point x =x = (29, #3, ..., 22) and let 2 be 2 stationary point of f (x). If the quadratic form ® Gf (ak) A (daa, dit, ...; de)= 3B “peaay 3182) (1) jee that is, the second differential of the function f at the point 2 is a positive definite (negative definite) quadratic form, then the point 2 is a strict minimum point (strict maximum point); if the quadratic form (1) is indefinite, then there is no extremum. at the point 2%, Sylvester’s criterion for positive dejiniteness of a quadratic form. For the quadratic form n A(z) =A (zy, Sy oes Tn) x UR (2) be in which ay = ay, i,j = 1, 2,.... m to be positive definite, it is necessary and sufficient that Ais An Ay a au>0, | 4 *2|>0 G5 Cpr 5] >>0 as O22 14gq Fag gy ag, dag +s Gin1, ABSOLUTE EXTREMUM 43 For the quadratic form (2) to be negative definite, it is necessary and sufficient that au @ an Ae a3 an <0, | “10, | aos on as |< 0 Ae, Gg O33 Gy Q4z ++ an On, Gey s+ Can ae (DPS 0, Gns Ing+-+ Onn The case of n = 2. Let the function f(z, y) be defined and let it have continuous second partial derivatives in some neighbourhood of the point (%», yo) and let (2, yo) be a stationary point; that is, fe (or Yo) = fy (or Yo) = 0 @) Now, if at the point (2, Yo) fish (fau)* > 9 (4) then it is an extremum point, namely a maximum if at that point Fee<0 (fgy-<20) and a minimum if fa>0 (fw>d) But if at the point (x, Yo) feshiv-— (Feu)? <0 then there is no extremum at the point (x, y,). Finally, when Fahey —~ (fau)* = 0 at the point (zp, yo), there may or may not be an extremum at that point. In the latter case, a supplementary investiga- tion is required. Example 3. Consider the functions z == xt + yt, g = = zt — yt, zs gt — y*. The point (0, 0) is a stationary44 CH. 1 THE RXTREMUM OF A FUNCTION OF MANY VARIABLES point fur each of these functions aud for each of them ex Zyy —~ (Zey)* == 0 at that point. It is readily seen that the point (0, 0) is a minimum point for the first function, a maximum point with respect to the second function, and is not an extremum point relative to the third function. Indeed, in all three cases we have 2 (0, 0) = 0, but in the first case the values of the function are positive in any neighbourhood of the point (0, 0), except the point itself, in the second case they are negative, and in the third case the function z = 2* — y* takes on both positive values (for instance, y = 0 when x = 0) and negative values (for example, y 540 when z = 0) arbitrarily close to the coordinate origin. Example 4. Find the extremum of the following function of three variables: foot yt oe — zy + 2 — 2 Solution. Find the stationary points of the given function f. To do this, set up the system of equations of ee arm—yit=d, | oF a Qy— x20, Be 20 | Solving this system, we get mae, yor -¥, fyi. Set up the quadratic form (1) at the point .(—$, $4) Gl gop t)- We have fee 2, fay 1, fae, feet, fy d, fie 9, fiz=0, — fey==0, faze At the point Py) we obtain Oy, = 2, Agr —4, a3 20, gy —~ 4, p30, ag, = 0, gg = 21, ABSOLUTE EXTREMUM 45 so that jst Sse 2-4 ay >0, I aon -| —41 |=3>0 2-410 wt 2 O|m6>0 0 02 Taking advantage of Sylvester's criterion, we conclude that the quadratic form is positive definite and, hence, by Theorem 2, the point P, is a strict minimum point, and 4 f(P)=—¥- Example 5, Find the extremum of the following function of two variables: z= ay (6 — x — y) Solution. We find the stationary points: 2 = 182%y?— 42%? — 82% = 0, z= 12nty— 2aty— Bzty? = 0 } whence z, = 0, y, = 0, and 2, = 3, yz = 2. We obtain two stationary points, P; (0, 0) and P, (3, 2). Let us find the second derivatives of the given function: Bo_ = Bary? — 12z*y? — bzy’, Boy = 1205 — 22t — xy, ey = 3a%y — Baty — 9aty? At the point P, we have Zi. = Sy = ty = 0 so that ex'Zy ~~ Fey)? =O and the question of the existence of an extremum at this point remains open. To resolve this question we have to invoke higher derivatives. At the point P, we have 2, = —144, 2, = —162, ey = —108. Obviously, 225°%yy — Cry)? > 0 and since 2x <0 it follows that at the point P, (3, 2) there is a maximum, and %max = 108 Tost the following functions for maxima and minima. 2. f= (x — 1)? — 2y*46 CH L THE EXTREMUM OF A FUNCTION OF MANY VARIABLES Be fm at yt — Bat day — By" 4a fa (es ye OO, 6. far~ Ht E+d (>, y>d, 2>u. 7. fe tm ay by — Qe ory. 8. fj = sin z-sin y-sin (z@ + y) OSxr ga, Ogy 0, te > 0, .- +5 tm > 0). 10. Show that the function z = (1 + 6”) cos z ~ ye” has an infinitude of maxima and not a single minimum. 11. State whether ihe following condition is sufficient for the minimum of the function f(z, y) at the poiat My (fo. Yo): that this function have a minimum along every straight line passing through the point M,. Consider the example f(z, y) = (¢ — y*) 2 — y*). 12. Show that, unlike a function of one variable, the existence in a domain D ofa unique extremum (maximum or minimum) for a function of two variables does not yet signify that this extremum necessarily yields the maximum or minimum value of the function throughout the domain: (a) 2 = a — y? + 2e-*, oo << a< 00, —oo 0 is sufficiently small. If grad f (x) = 0, then we put == ah — ha (grad f (@*), @n) (ta > 0) and, generally, if grad f (24) 0, then Rete ly (grad f(z"), en) (k= 4, 2, ..., m) (hen > 0) Under definite conditions (see [19]) we get_a monotonic decreasing sequence {f (x”)}. If 2? @ and Zis a minimum point of the function f (x), then grad f (2") + 0 as n-> oo. j owe 6. Find the minimum point of the function Solution. Let us take the point 2° = 1. We have grad f (x) = 22% = 2i xe 0 Therefore eis g?—h-2=1-—- 2h, where h>0 2033948 CH 1 THL EXTREMUM OF A FUNCTION OF MANY VARIABLES Furthermore grad f(a) = 2 (1 — 2h)i If hep, then grad f (2) #0 and a2 mz ot — 2h (AL — 2h) = (1 ~ 2aye Continuing this process, we find a = (1 — 2h)" Clearly, if O<(k<¢4, then z*~-»0 as n—» co. The point z= 0 is the minimum point of the function f(z) = 2. But if k = 1/2, then z’ = 0, grad f (z') = 0 and we obtain the stationary sequence {0} whose limit is zero. Example 7. Find the minimum point of the function F% y) = at ey. . Solution. Let us, for example, take the point (1, 1), that is, 7° = 4, yo = 1. We find grad f (1, 1) = 2i + 2 Since grad f (1, 1) 0, we put ghen go — 22% == 1 — 2h, (h> 0) poy — yh = 1 — 2h We have grad f(c!, y!) =2(1—2h) i+2(1—2h) jx (hae $) and so we take ghee tt Qat-h xx (1—2h)*, ‘ (h>0, he) pox yh Dyt hm (12h) Continuing this process, we obtain a se (1 — 2h)", yt = — ay so that When 0 0 as R=» 902. CONDITIONAL EXTREMUM 49 Thus, the minimum point of the function f (x, y) = 2* + -. y* is the point (0, 0). Using the method of steepest descent, find the minimum point of the function geez? py? — Qe dy +5 2. CONDITIONAL EXTREMUM Suppose we have a function z= f (2, zo, .-., 2) of n variables defined in some domain D of the space £*. Besides, let there be imposed on a1, 22, ..., ty macces~ sory conditions (m << x): Pq (Tay Ly, «+2 Ln) =O, \ ee ee ee (4) re (L4, Lay oe Zn) O which are called the equations of constraint. Let 2 == (2%, at, ..., 2%) be an interior point of the domain D. We say that at the point (z?, 28, ..., 2°) the function f(t, Ze, ..., @) has a conditional mazimum (or conditional minimum) if the inequality Typ tay s+ SSF CL a. - ) w) @) (or f (ty, Ze, «+ +> Mp) BF (@, ah, .-., ah) holds true in some neighbourhood of the point (2?, 22, ..., 28) provided that the points (a, 2, ....2,) and (2%, 2, ..., 2%) satisfy the constraint equations (1). Example 1. The function z = z* + y? has an absolute minimum equal to zero at the point (0, 0). We adjoin the constraint equation cy -—41= 0, that is, we seek the minimum of the z-coordinates of the points of the surface z= 2% -+ y* solely for those values of < and y that satisfy the equation ¢ -+ y — 1 = 0. A conditional minimum cannot be attained at the point (0, 0) because this point does not satisfy the constraint equation. Let us solve the constraint equation z-- y — 1 = 0 for y and substitute the value found, y x, into the equation of the surface. This yields ¢ = 2° ~ (1 —- 2)°, which is a function of one Qn20)0«CH. 1. THE EXTREMUM OF A FUNCTION OF MANY VARIABLES variable. Investigaling it for an extremum, we find Ler}, min = 5. By virtue of the constraint equation, find yer = =. The point (4, 4, +) is the vertex of a parabola obtained in the intersection of the paraboloid z = * + y* cut by the plane z+ y — 1 = 0. We can reason analogously in the more general case as well. Suppose we are seeking the conditional extremum of the function z= f(z, y) given the constraint g (z, y) = 0. Also suppose that for the given values of x and y the equation @ (z, y) = 0 defines y as a single-valued differentiable function y =p (z). Substituting the function yp (x) in place of y into the function f (z, y), we get a function of the single variable x: 2 = f(z, } (z)) = F (z). The absolute extremum of the function F (x) is the desired conditional extremum of the function f(z, y), given the constraint 9 (2, » == Q, This method is not always convenient in a practical sense since it requires the actual solution of the equation (z, y) = 0 for some variable. In seeking the extremal values of the function z = se f (zy, Zp, .+.+, %) in the case of constraints (1), use is made of the Lagrangian method of undetermined multipliers. The Lagrangian multiplier method. Suppose that (1) the functions f(x, z2,.. ., 2) and @; (a, tz, -. +) In) (i= 4, 2, ..., m) have continuous partial derivatives of the first order in the domain D; (2) m<{n and the rank of the matrix eal , is = 1,2, ...m,j=4,2,..., ”, is equal to m at every point of D. A new function (the Lagrangian function) is formed: ma Desf Shui 8) Sh where 4; are undetermined constant multipliers. The function ® (z,, 22. .... %,) is investigated for an absolute extremum; that is the following system of equa-2. CONDITIONAL EXTREMUM 24 tions is set up: a a a 20, Bnd, ..., Bao w from which and from the m equations of constraint 0, 2 = 0, Pm =O are determined the values of the parameters Aq, Ag, ~~ +) Am and the coordinates (z,, %2, -.., Z,) of possible extremum points. The conditions (4) are necessary conditions for an extremum both of the Lagrangian function and the original function 2 = f (2%, 22, 66, Zn)> If the point (a2, 22, ..., 2%) is a conditional extremum point of the function f (a, 2, ..-, %), then it is a sta~ tionary point of the Lagrangian function; that is, at this point = = 0 (i= i, 2, + n). In order to investigate the stationary point (2%, - ++, %) of the Lagrangian function © (z,, zg, ..., 3 for a conditional extremum, it is necessary to set up the quadratic form Tmt B (des, Bay 04 Bina), 2, bij dxy dry (8) that is, the second differential of the Lagrangian function at this point with allowance made for the conditions Fei dat FE dy ©... GEL dz, =O (= 4, 2, 0.2, m) ax (6) Ti the quadratic form (5) is definite, then at the point (af, %, ..., 2%) there will be a strict conditional extremum, namely: a strict conditional maximum if the quadratic form (5) is negative definite, and a strict conditional minimum if the quadratic form (5) is positive definite. But if the quadratic form (5) is indefinite, then the point (zi, 22, ..., #8) is not a conditional extremum point. Thus, the existence at the point (2°, 2%, ..., 2%) of an absolute maximum (minimum) of the Lagrangian function (for the values found: 24, Ag, -.., Am) implies the22 «CH. I. THE EXTREMUM OF A FUNCTION OF MANY VARIABLES existence at that point of a conditional maximum (minimum) of the function z = f (21, 2p, ..-, 2), given the constraints i (yr Lay ey Mp) FO (Em 1, 2,2, m) The absence of an absolute extremum of the Lagrangian function @ (2%, 22, ..., %,) does not yet signify the absence of a conditional extremum of the function F@y Za - + Bn) Example 2. Find the extremum of the function 2 = zy provided y — 2 == 0. Solution. Form the Lagrangian function O (x, y) = ay bay ~ 2) and write out the appropriate system for determining 4 and the coordinates of the possible extremum points: ey —he 0 Fe SY ee, Bnzthad, o) yr 0 From the first equation we find A = y. Substituting into the second, we get -+-y = 0. Thus, zey= of yours whence x = y = 0. We then have 4 = 0. Thus, the appropriate Lagrangian function is of the form ® (x, y) = zy. The function © (z, y) does not have an absolute extremum at the point (0, 0), but there is a conditional extremum of the function z = zy provided that y = a. Indeed, in that case we have z = 2, whence it is seen that there is a conditional minimum at the point (0, 0). Example 8. Find the conditional extremum of the function f(y, 2) = ays 8) subject to the conditions gy, & ¥ 2) eby—2z-—8 moh G2 (2, Ys 2) = omy s—- B= 0 ®2. CONDITIONAL EXTREMUM 2B Solution. Form the Lagrangian function © (x, y, 2) = ayz by (2 by — 2 ~ 3) +g (e~y—2—-8) and write out the system of equations for determining the parameters 4,, Ag and the coordinates of the possible extremum points: a Bayer le=0, Gareth b= 0, 0 22 way —hy—e=0, oo) gym e—3=0, Tom yorrtmB==O y Solving the system of equations (10), we obtain 44 234 44 5 44 dys as dg rey, y=~— 3, Bae eg The second differential of the function ® (x, y, 2) is aD , B&O , FO Dan gat F9. ay + FE ast +229. de dy +2 22. ae de2 Fo -dyde In our case EO = 2c dz dy + Qy de dz + 2x dy de (44) Taking advantage of the constraint conditions (9), we get dz dy—dz=0, } dz — dy — dx 0 whence de = dz, dy = 0. Substituting this into (11), we obtain B (dz) = 2y da®24 CH L THE EXTREMUM OF A FUNCTION OF MANY VARIABLES At the stationary point B == -5 dz?<0, that is, at the point (¢ :~e —%) we havea maximum equal to fm, = 05 =—ar- Example 4. Find the extremum of the function 2 = == cos? z -+ cos* y subject to the condition yo tap Solution. We form the Lagrangian function O@, y) = cos* z+ cos? y+-A (y-z—=) and write out the system of equations for determining the parameter & and the coordinates of the possible extremum points: 2 —2coszsinz—A=0, Bea ~—2cosy sin y-+- = 0, y-2z—h=0 or sin2z= —A, (12) sin 2y =, (43) yr (14) From (12) and (13) we have sin 2z + sin 2y = 0 or 2 sin (x + y) cos (y — z) = 0 (45) According to (14) we have cos (y —2z) = Bz, and therefore from (15) we find that sin (z + y) = 0, whence z-by=skn, k= 0, £4, &2,... (48) Solving equations (14) and (16) simultaneously, we get ket ke, . Barge Yee k=O, ok 4, $2,... (47)2. CONDITIONAL EXTREMUM 2 Let us find the second derivatives of the function © (z, y): or aD ao 3 —2cos2z, aay =0, “age = — 2.008 2y At the points P, (4-4, B44) we have Diq- Diy — (OF)? = 4 008 {lex —{) cos (i+) =2cos2kt=2>0 This means that there is a conditional extremum at the points P,. Furthermore, when k = 2n, FO _ l= VE0 that is, the conditional minimum v2 Zain = 1—- at the points Poa1. Find the conditional extremum in the following problems. 14. f=ay for PH y= i. 15. facx®+ty? for gteet 16. f = 2yz subject to the conditions x+y +2 = 5. zy typ a= 8. 47. f =e when z + y =a. 18. f= 6 — 4g — 3y when P+ Y= 1. 19. f=2—2Qy+ 22 when + yt ea 9. 20. f=sinz sinysinz when t+ y+2=4, z>0, y> 0, s>0. s26 CH. L THE EXTREMUM OF 4 FUNCTION OF MANY VARIABLES 24. Prove the inequality Sr (FEL), nbd, 20, ym0. 22. Find the maximum value of the product ayst of nonnegative numbers z, y, 2, 4, provided that their sum preserves the constant magnitude thypeti= de 23. Find the shortest distance from the point IM (1, 0) to the ellipse 4a7 + 9y? = 36 24. Find the distance between the parabola y = 2* and the straight line z—y = 5. 25. Find the sides of a rectangle of maximum area in- seribed in the circle 2? -- y* = 26. In a sphere of radius insoribe a cylinder having maximum total surface area.CHAPTER II EXTREMA OF FUNCTIONALS 3. THE FUNCTIONAL. THE VARIATION OF A FUNCTIONAL AND ITS PROPERTIES 4°. Definitions of a functional. Proximity of curves. Suppose we have a certain class M of functions y (x). If to each function y (2) € M there is associated, by some law, a definite number J, then we say that a functional J is defined in the class M and we write J == J [y (z)]. The class M of functions y (z) on which the functional J [y (@)] is defined is called the domain of definition of the functional. Example 1. Let Jf = €[0, 1] be the collection of all continuous functions y (x) specified on the interval [0, 1] and let 4 Ju@l= | y@az (t) 0 Then J [y (z)] is a functional of y (z): to every function y(z) €C (0, 1] there is associated a definite value of J [yl]. By substituting concrete functions into (1) in place of yl), we get appropriate values of J [y]. Thus, if y (2) = 1, then L Jttjs: | dedemt 028 CH, UL EXTREMA OF FUNCTIONALS IE y (x) = e*, then i Je] = 5 e daae—t 9 If y (z) = cos az, then 1 J [cos sex] = \ cosnz dz = 0 0 Example 2. Let M = C, la, b] be the class of functions {z) that have continuous derivatives on the interval a, 6] and let J ly @)l = y’ (@), where x, € fa, 0] Q) It is clear that J [y (z)] isa functional defined in the indicated class of functions for the simple reason that to each function of this class there is associated a definite number— the value of the derivative of the function at the fixed point zo. i, for example, a= 4, 6=3 and 2x, = 2, then for y (2) = 2" we have J [x7] = Qzlen2 = 4 For y(@z)=27 +4 we get Jiz*+41] = 4. For y(z) = sIn(t +2) we have Jiln(i+a)l=;ie| = 4. Example 3. Let 47 = C {[—1, 4] be the class of functio: y () continuous on the interval [—1, 1] and let @ (z, y) be a given function defined and continuous fer all —-1 <2 <1 and for all real y. Then Tiy@= \ ole, y(eildz @) will be a functional defined on the indicated class of functions. For example, if @ (2, y) = ter , then for y (z) = t =z we have Jb) = | 3% =0 and for y@)=i+s9, THE FUNCTIONAL. THE VARIATION OF A FUNCTIONAL — 29 we have 4 = | #8 Va— 7i+al= ae =nV5—aretan 2 Example 4, Let M = C,[a, ] be the class of functions (2) having a continuous derivative y’ (z) on the interval @, 6]. Then 5 Jy @=\ VIF wae @ a is a functional defined on that class of functions. The functional (4) geometrically describes the arc length of the curve y == y(z) with ends at the points A (a, y(a)) and Bb, y (d)). The variation, ‘or increment, Sy of the argument y (2) of the functional J [y (z)] is the difference between two functions y (z) and y, (z) belonging to the chosen class M' of functions: by = y (x) — Yo () 6) For the class of k-times-diiferentiable functions we have (Sy) = by” (2) (6) We say that the curves y == y (z) and y = y, (z) specified on the interval [a, 4) are close in the sense of proximity of the zeroth order if | y (z) — ys (2) | is small on fa, b]. Geometrically, this means that those curves on the interval (a, 5) are close as regards the ordinates. ‘We will say that the curves y = ¥,@) and y= (2), which are specified on the interval [a, bj, are close in the sense of proximity of the first order if | y(z) —y,(z)] and [ y’ (z) — y; (z) | are small on la, 8]. Geometrically, this means that the curves on the interval [¢, 6] are close both as to ordinates and as to directions of the tangents at the appropriate points. The curves y = y (x) and y = y, (z) are close in the sense of proximity of the kth order if the moduli ly¥@—-u@bhly@~%oOL oor Ey (2) — (2) | are small on [a, 8]. we)30 CH, IL, EXTREMA OF FUNCTIONALS If two curves are close in the sense of proximity of the kth order, they are all the more so close in the sense of proximity of any smaller order. Example 5. The curves y (z) = me, where n is sufficiently large, and y, (z) = 0 on [0, x] are close in the sense of proximity of the zeroth order since the modulus of the difference \¥@—-n@|=|3™| f(a\= max max, 172 (z)-— 7 (2) | (8)32 CH. IL. EXTREMA OF FUNCTIONALS The definition of the distance between curves given on page 30 is, in the sense of the new definition, a distance of the zeroth order. Example 8. Find the first-order distance between the curves f (z) = 2° and j, (x) = 2° on the interval O <2 <1. Solution. We find the derivatives of the given functions, f' (x) = 2z and f, (x) = 32%, and we consider the functions y, (x) = 2? — 2 and ys (xz) = 22 — 32%. Let us find their maximum values on the interval [0, 4]. We have yf = = 22 — 3c%. Equating this derivative to zero, we find the stationary points of the function y, (x): 4, = 0, 2, = Z. Furthermore, ¥; lane = 0, ¥2 \_2 =H The value of ¥, (z) at the right endpoint is y, (1) = 0, whence = —2}= max (2@—2)=24 em Be lA el= as seg Let us now find the zeroth-order distance py) between the derivatives 7 (z) = 22 and fj (x) = 327: Pom max Lye (2)|= max |22—32?| We construct the graph of the function y = | 22 — 3x° | (Fig. 2). From the figure it is evident that B, = 1. Thus,3. THE FUNCTIONAL, THE VARIATION OF & FUNCTIONAL 33 the first-order distance p, between the curves f (z) = 2* and fy (a) = 2 is equal to fs = max (Py; Po) = 4 $8. Find the first-order distance between the curves f () = Ina and f, (z) = z on the interval [e~, e]. 34. Find the second-order distance between the curves f (2) = zand f, (c) = —cos z on the interval [o, $l . 35. Find the 1001st-order distance between the curves { (@) = e and f, (z) = z on the interval [0, 4]. The nth order e-neighbourhood of a curve y= f(z) (@< <2 > 0 there exists a number n > 0 such that for all admissible functions y = y (x) satisfying the conditions ly@—y@i>0 and show that there exists a number 4 > 0 such that | J [y @)] — — J [al |< & as soon as | y (@) ~* | 0 there is an 4 > 0, for example, = a = such that as soon as ply (@), sl 0 (namely, ¢ 0, there will be functions f (2) such that olf, fl 0. We have IF UF IF Ufo (a) |= F (20) Fo (20) | Clearly, if we take » =e, then for 9, ff (2), A @l<4 ‘we will have iW @)—Sh @li 00. On the other hand, the diflerence z J tun CaF [uo @ = | SEB dee F 0 does not depend on x. Thus, as 2 ~» 00, J [y, (¢)] does not tend to J [yo (z)] = 0 and, hence, the given functional is discontinuous in the sense of zeroth-order proximity on the function yo (x) = 0. We leave it to the reader to prove that the functional under consideration is continuous on the function yy (z) == 0 in the sense of first-order proximity. Investigate the following functionals for continuity. 36. J ly @l=y (x), where the functions y (z) € € Cla, b] and x €[a, b] in the sense of zeroth-order proximity. 37. J ty @)] = max | y (z)|, where the functions y (x) axé continuous on the interval fe, 6] in the sense of ze~ roth-order proximity. 0 if y(z) assumes at least one negative value, Jiy@l=4 ¥ sf y(e)=0, 4 if y(z)e0 and y(z) 560 in the sense of zeroth-order proximity. i 39. J ly @ = | |g’ @) Ide, where the fimetions y (2) é have continuous first derivatives on the interval [0, 4]: (a) in the sense of zeroth-order proximity; (b) in the sense of first-order proximity. * 40. J ly @)l = \vi “F ¥'*@) dx on the function yp (2) == se 0, where the function y () €C, (0, nJ: (a) in the sense3. THE FUNCTIONAL THE VARIATION OF A FUNCTIONAL 37 of zeroth-order proximity; (b) in the sense of first-order proximity. A. J ly @) = j (1 + 2y (@))dz on the function yo (2) =20, where’ the function y(z)€C,10, nl in the sense of first-order proximity. Example {2. Show that the functional t J @= | eyitF ea a defined on the set of functions y (z) € C [0, 4] is continnous on the function yo (z) = 2" in the sense of zeroth-order proximity. Solution. Put y (z) = 27-+ an (a), where y (x) € C [0, 4] and @ is arbitrarily small, 4 Tey @l=J ta on (2) =) BVT ae ae o = | VIFF TST Passing to the limit as a —»0, we get from this equation Tim fy @)] = \evTFRarns ta and this signifies the continuity off the functional on the function Yo = 2%. Definition. Let Jf be a normed linear space of the functions y (2). The functional L [y (z)l defined in the space M is called a linear functional if it satisfies the conditions: (4) L ley (2) = ¢-L ly @)I where ¢ is an arbitrary constant, @) Lily @) + y. @) = Lily @) + Lily, @ where y, (z) €M and ys (x) € M.38 CH. IL. EXTREMA OF FUNCTIONALS For example, the functional 5 Ly @l= } tv @)+y@de defined in the space C, a, b] is clearly a linear functional. Here is another definition of the linearity of a functional. A functional Z [y (z)] is said to be linear if it (1) is continuous, and (2) for any y, (x) € M and y, (x) € M satishes the condition Lin @ +42 @)= Llyn @1+ 2 ly. @ 42. Demonstrate that the two foregoing definitions of the linearity of a functional are equivalent. 43. Show that the functional LZ [y ()] = y (z,) is linear. 44. Show that if Z ly (z)] is a linear functional and the ratio wat +0 as jy (z) | +0, then Lly@=9. 3°. Variation of a functional. Let the functional J ly (z)] be given on a set M of functions y (x). The following quantity is termed the increment of the functional J ly (x)l, which corresponds to the increment dy (z) in the argument: AT = Ad ly (= J ly @) + by @I—JS ly @)]_ 0) (8y (2) = 7 @) —y @, where y (@) EM, y @) EM) Example 13. Find the increment in the functional 4 Fig = J u@)y @de 9 defined in the space C,[a, 6] if y (2) = 2, yy @) = 2% Solution. We have £ 4 AJ =F [2] —J [a] = | at80dz— | aetede 0 0 = i (22°—~ 2x) dz=0 ° 45, Find the increment in the functional considered in Example 13 by putting y (z) =e, ¥, @) = 1.3. THE FUNCTIONAL. THE VARIATION OF A FUNCTIONAL 39 Definition. If the increment in the functional, AS = J ly (2) + by) ~ J ly @) can be represented as AJ = Ly @), dyl + BY (), dy) Ih by I where ZL ly (z), Sy] is a linear functional with respect to by and f (y (@), dy) +0 as || dy |] ~-0, then the portion of the increment of the functional that is linear with respect to dy, that is, Z[y (2), Sy] is called the variation of the functional and is denoted by 57. In this case the functional J ly @)] is said to be differentiadle at the point y ©), 46. Show that the variation 6J of the functional J [y (2)] (if the variation exists) is defined in unique fashion. Example 14. Show that the functional > Tu@l=fy@a specified in the space C [a, }] is differentiable at each point y (z) of that space. Solution. AS = I [y+ by} —F Ty] % b Ly = J ye) by(aiae— | vieyde= J sy(e) de 2 Thus, AJ = fey (x) dz. This is a linear functional with respect to by (2). In the given case, the entire increment of the functional reduced to the linear functional with respect to &y (z). This functional is differentiable at every point y (z) and its variation 67 = by (z) dx. 47. Show that every continuovs linear functional J [yl is always differentiable.40 CH, 1, EXTREMA OF FUNCTIONALS Example 15. Show that the functional 8 Tii= | ¥@a defined in the space C [a, 6] is differentiable at every point y @). Solution. We have a a Ars | tye) +dy@Pae— | ve) de 2% a = [eve sy@ar+ | Gyeyrae (4) ‘The first integral in the right member of (14) is a linear functional with respect to dy (z) for every fixed function y (x). Let us estimate the second integral in the right-hand member of (11). We have o a J Gy nyptan= | [sy (e)Pee < (ax [dy (0) ] ee= 2) dy) IP =((b—~a) |! Sy ll) dy i The quantity (6 — a) ff Sy |] +0 as |} Sy |] +0. Thus, the increment AJ of the functional can be represented in the form of a sum of L ly, 6y] and an extra portion having second order of smallness with respect to |} Sy |]. According to the definition, the given functional is differentiable at a point of y (z) and its variation is b SF x2 \ y (2) dy (x) dz 4 48. For the functional J [y (2)] = Sy (2) de put y= 0 4. == 2x, Sy = az* and compare SJ with J for a = 4, 0.4, 0.04.3. THE FUNCTIONAL THE VARIATION OF A FUNCTIONAL a 4 49. For the functional J Ly (2)] = Jay? (2) dx put y = = et, Sy = ax. Compare AJ with 57 for a = 4, 0.1, 0.04. 50. Verify the differentiability of the following functionals: (4) J ly] = y (a) in the space C [a, d]. (2) J Ty] = ve. in the space C, [a, 0). 8) Jly] = VI Fy] in the space C, Ia, 51. (4) Jlyl = |y(@)] in the space C fa, bi. 51. Show that the functional J*[y] is differentiable if J ly) is differentiable. Write down the variation of J? ly). 52. Show that tho functional x Figl= \ fe yla)ae defined in the space C [a, b], where j (x, y) is a continuous function of its arguments having continuous partial derivatives up to order two inclusive in the domain aczqb, co << y << -too, is differentiable and its variation is of the form b 8S == j ED sy (x) dx Example 16. Consider the functional 3 Jtul= | fe, ye), ¥ @)ae defined in the space C, [a, 6} of continuous functions y (2) on the interval [2, 4], the functions having continuous first derivatives. The function f(z, y, y’) is continuous with respect to all its arguments and has continuous partial derivatives up to order two inclusive in the domain acgh, —-wcyc+ow, —wcy <+o0 Let us find the increment AJ of the functional that corresponds to the increment dy (z) of the argument, where2 CH. I. EXTREMA OF FUNCTIONALS dy (z) € C, [a, b]. We have & ATy@l=Vie toy toy) —Te, yy de (12) By Taylor’s formula, f@yt+yy +) — fay y') 28, 2 . , . =F out gry = Re, vy’, by, By) (13) where R (z, y, y’, Sy, 5y’) is the remainder term of Tay- lor’s formula. Putting (13) into (12), we get & Arty = | (ZL by+ gb oy’) ae b +) Rey yoy byez 14) The first term on the right of (14) is linear with respect to Sy and dy’. Suppose all second partial derivatives of the function f (z, y, y’) with respect to y and y’ do not exceed, in absolute value, a constant IM > 0 in the domain bounded with respect to y and y’. Then the following estimate holds true: a JIRG, wy, by dy D1 de ° b S20 § | byl dem 2M (—a) [| By? 2 Here, [I Sy i] = max (1 8y |, | Sy’ }). Thus, the second term in the right member of (14) is of second order relative to || Sy [|. Consequentiy, by definition, the functional J [yl is differentiable in the space C,[a, 6] and its variation is of the form 6 sr t= | (Ff by+ Frey’) de (18) a3, THE FUNCTIONAL. THE VARIATION OF A FUNCTIONAL 43, Example 17. Find the variation of the functional 4 Jil= | (et zy) ae wt Solution. The function f (x, y, y’) = y’e¥ + ay® is obviously continuous in the variables x, y and y’ jointly, and has partial derivatives of all orders with respect to y and y’ that are bounded in any restricted range of the variables y and y’. Therefore the given functional is differentiable in the space C, [~~1, 1] and its variation is, by formula (15), equal to 4 BT = | 1y'e? + 20y) by 7 eby'] dx oe 53. For the functional Jty@) = J (yy + ay”) dx a put y=Inz, by Eb. and compare AJ [y(z)} with 87 [y(@)) for b=4, 0.4, 0.01. 54. For the functional Jty @l= | jy? yy ae 0 put y = 2%, dy = k2*. Compare AJ [y (z)] with 5/7 [y ()] for & = 4, 0.1, 0.04. & 35. For the functional J [y (2)] = |y'* sin 2 dx put y = == sin z, dy = k cos. Compare AJ [y (2)] with 6J [y (2)} for k = —41, 0.3, 0.03, 56. Show that if the function f (@, 4, 4, +. +, Sasa) has continuous second derivatives with respect to all arguments in the domain ez 0 for any nonzero element z. Examples. (1) The expression a J [z, yJ= ) A) z(t) y (8) dt where A (2) is a fixed continuous function, is a bilinear functional, and \4 (f) 2 (t) dt is a quadratic functional in the space C (a, db); also, if A (é)>>0 for all ¢€ la, dl, then this quadratic functional will be positive definite. (2) The expression i [A@2O)+BO 22 Q+C() 2" a) de @ is an example of a quadratic functional defined for all functions in the space C, la, 4]. (8) The integral K(s, tz (s) yds dt Reo ato where X (s, 2) is a fixed function of two variables, is a bilinear functional in C [a, 6). Definition. Let J [y] be a functional defined in some normed linear space. We will say that the functional J [y] has a second variation if its increment AJ = J [y+ dyl — J ly] may be written in the form AJ = Ly [6y) + $ Le (Oy) + BIL yi (18) where L, [Sy] is a linear functional, L, [6y] is a quadratic functional, and 8 ~+0 as || Sy || +0. ‘We will call the quadratic functional ZL, [Sy] the second variation (second differential) of the functional J ly] and denote it by 67.& CH. I. EXTREMA OF FUNCTIONALS The second variation of a functional (if it exists) is defined uniquely. Example 19. Find the second variation of the functional 1 Jiyl= } cy +y') dx o defined in the space C, [0, 4] of functions y (2). Solution. We have AS == F [y+ dy] — JF [y) £ =) Ut o+ G+ oy Pay] de 0 t = | ey Sy +x (by) + 3y"? Oy! + By (by')® Q t + (Gy) da = | (Ray by + By"* by’) de 0 t t + | te Guy 8y Gy'Vide+ | yee (19) 0 o For a fixed y (x), the first term in the right-hand member of (49) is a functional linear with respect to dy (x); the second term of the right-hand member is a quadratic functional. Finally, the third term of the right-hand member allows for the obvious estimate | \ (yy de] <(eoax | Sy’)? j 18y'| ate<| | ay’ | dz ly I? 0 o Qo {the norm in the sense of the space C, (0, 1]), whence it is seen that this term can be represented in the form B -|| Sy {j*, where B —»-O as [{ dy || 0. According to the definition, the given functional has a second variation 6 and it is equal to L GF 2) fe (Gy + Sy! (6y')'] de o3, THE FUNCTIONAL, THE VARIATION OF A FENCTIONAL 49 62. Prove that a quadratic functional is differentiable and find its second variation. 63. Write down the second variation of the functional eF@, where F (y) is a twice differentiable functional. 64. Show that functionals of the form b Jul= | F@,y ye in the space C; la, 2] are twice differentiable if the integrand F has continuous derivatives up to order three inclusive. Find the expression for the second variation. We introduce the function @ (x) = J [y+ aby). The second variation &*/ of the functional J [y] is also defined in terms of the second derivative of the function (a) at the point @ = 0: 2 sae SO (2) OF = aor hex Both of these definitions coincide for functionals of the integral type, which are the ones that we will be consider- ing for the most part. Find the second variations. a 65. Til= | Fe, yy, 1 ye) de. 68. Fil= |) Fle, ys a, te, 2) dx dy. 6 & 67. Fass soy val = J PGE yey ooo Yas Yo oes Yad a. 2 6. Extremum of a functional. Necessary condition for an extremum. We say that a functional J [y (z)] attains a maximum on a curve y = Yo (z) if the values of the functional J [y (z)] on any curve close to y = yp (x) do not exceed J [yp (z)], that is, AT = J ly (@)) —F ly @ <0 4053950 CH. 1, EXTREMA OF FUNCTIONALS Mf AJ <0, and AJ = 0 only when y (x) = yp (a), then we a, at a strict maximum is attained on the curve y = = Yo (2). The curve y = yo (z) on which a minimum is attained is defined in similar fashion. In this case AJ 0 on all curves close to the curve y = yp (z). Example 20. Show that the functional t Ju @l=) @rvyde Q attains a strict minimum on the curve y (z) = 0, Solution. For any function y (z) continuous on [0, 4] we bave 4 4 L ATF ly (a —J 10] = | Gy) de | dex | warm Q 0 0 equality occurring only when y (x) == 0. Strong and weak extrema. We say that a functional J ly (@)] attains a strong relative maximum on a curve y = == Yo (x) if for all admissible curves y = y (z) located in an eneighbourhood of zeroth order of the curve y = Yo (x) we have J ly @I< J bye @) A strong relative minimum. of the functional is defined simi- ly. We say that the functional J [y (z)] attains a weak relative maximum. on a curve y = Yo (z) if for all admissible carves y = y (2) located in an e-neighbourhood of the first order of the curve y = Yq (z) we have Fly @I 0 on the interval [—1, 4], and / lyl = 0 only when y’ (x) = 0, that is y (x) == C = constant. The function y (z) = C lies in the class C,[—4, 4] of functions having on the interval {—1, 1] a continuous first derivative but does not satisfy the given boundary conditions. Hence, J {yl > 0 for all y (z) €C,{—4, 1] that satisfy the conditions y (—1) = —1, y (i) = 1. Thus, the functional has a lower bound but it is not attained on the curves y (x) € €C,[—1, 1]. Indeed, let us consider the one-parameter family of curves z arctan — a Ya(Q)=———z-, «>0 arctan These curves satisfy the boundary conditions yo (~—1i) = = —1, Yq (1) =4. In the limit, as « 0, we get the function 0 if z=0, +1 Orci or ¥ (z) = sgn (Fig. 3). This function belongs to the class of functions that are piecewise differentiable on the interval [—4, 1]. We have —-1itf -t0 Although the function v (z) does not have a second derivative, it satisfies the appropriate Euler equation. Indeed, since F(z, y, y)=y*(1—~y'), by putting y =v (z) we get Euler’s equation Ww oP +E [8 ry) 0 @ But according to the definition of the function v (x) we will have, on [—1, 1], Fy = —2e* (1 — v’) = 0 and, hence, also é Fy = 0; and although the Euler equation (6) is formally of the second order and v” (x) does not exist, the substitution of v (z) into the Euler equation converts it to an identity. Theorem 2. Let y = y (x) be @ solution of the Euler equation Fy — fe Fy =0 If the function F (z, y, y') has continuous partial derivative up to order two inclusive, then at all points (x, y) at whid Pyy (%, y @, y’ (@)) #0 Q the function y = y (x) has a continuous second derivatin.4. AN ELEMENTARY PROBLEM. EULER'S EQUATION 59 Corollary. The extremal y = y (z) can have a salient point only at those points where Fyy = 0. Thus, in Example 4, Pyy = op vanishes at points on the axis; the extremal has a salient point at the point t= Theorem 3 (Bernstein). Suppose we have an equation y= F(a, yy’) @) If the functions F, Fy, Fy are continuous at each endpoint @, y) for every finite 4 and if there exist a constant k >0 and functions a==a(z,y) 20, B= y RO bounded in every finite portion of the plane such that Fy @ yy) >, @) [F@ yy) |< oy? + B (40) then one and only one integral curve y = @ (2) of equation (8) passes through any two points (a, 4) and (b, B) of the plane that have distinct abscissas (@ s+ b). Example 5. Prove that through any two points of the plane with distinct abscissas there passes one and only one extremal of the functional { en Bb (y’2_. 1) dz Solution. The Euler equation of the given functional is of the form y’ = 2y (1 + y”) and Theorem 3 is applicable. Indeed, in the given case, F(z, y, y') = 2y (1 + y) and F,=2(+y) blak Furthermore, [Fa wy)l=layd+y) | <2lyly?+2 [yl go that a= BP =2[y| p90.60 CH EXTREMA OF FUNCTIONALS Example 6. Show that it is not possible to pass an extre mal of the functional Jigs | @+ViTY) de through just any two points of a plane having distinct abscissas. Solution. The Euler equation is of the form yf = 2y (t + 7? (41) and Theorem 3 is not applicable since the condition (10) is not fulfilled [F (z, y, y’) increases faster with respect to y’ than the second power of y’]. However, this does not yet imply that it is not possible to draw an extremal through Just any two points with distinct abscissas. Putting y’ = p, y= pe in equation (141), we get @ peawi+ ey? or pap Gage 7 8Y Ini ting, we find— 1 y—C or tegrating, Vie ? C-y)Viryiat dy Vi=G=F a7 eo (12) where C is a real constant. It is easy to verify that for all values of y that satisfy, for example, the conditions 0 V3, there is no admissible value of the constant C for which the right member of (12) will be real. 80. Show that one and only one extremal of the functional whence Jay@=) VTP ye ae passes through any two points in the plane.4, AN BLEMENTARY PROBLEM. EULER'S EQUATION 61 Example 7. Prove that any equation Y@=7eyy) (13) is an Euler equation for some functional Jw@l=) Fre, wy) a (14) (1) How is the function F (a, y, y’) determined from the function f (z, y, y’)? (2) Find all functionals for which the straight lines y@=C2t+C, are extremals. Solution. We will seek the functional for which the Euler equation Py Pye Fyyy —Fyyy =9 (45) coincides with equation (43). This means that there must be an identity with respect to 2, y, y’: Fam Frye Py —Fyy fe sy!) 20 Differentiating this identity with respect to y’, we get PyystPyyy lt Pye tt lyy ty =o Put u = Fyy. Then for the function u we get a partial differential equation: Sty Btiet hy und (16) Thus, finding the functional, that is, finding the function F (z, y, y’), reduces to integrating the linear partial differential equation (16) and to subsequent quadrature. Let us consider the second question. In this case, the Buler equation must be of the form y” = 0 and for the function u we get—by virtue of (16)—the equation ou ou ety an! (17) Let us integrate this equation. The characteristic equation is of the form a t yO2 CH, II. EXTREMA OF FUNCTIONALS Integrating this system, we get y= 6, y= Cr+, whence C, = y— ay’. Therefore, the general solution of equation (17) is uz, y, y= Oy’, y — ay’) where © is an arbitrary differentiable function of its arguments. From this we have z Fe, ye 2) =2(2, +28 (z, w+ | GO, pte de (18) oO where a (z, y) and 6 (z, y) are arbitrary functions of their arguments satisfying the relation on oy oa It is evident from the solution that there exists an infinitude of variational problems for which equation (13) is Euler’s equation. Elementary cases of integrability of Euler's equation. 1°. F does not depend on y’: F == F (u, y). In this case Euler's equation is of the form Fy, (z, y) =0 (49) The solution of this finite equation does not cortain any arbitrary elements and therefore, generally speaking, does not satisfy the boundary conditions y (a2) = A, y (8) = ~ Only in exceptional cases where the curve (19) passes through the boundary points (a, A) and (6, B) does there exist a curve on which an extremum can be attained. Example 8. Find the extremals of the functional ny Siy@n= | y@x—y) de, yO)=0, ¥(F)=F 0 Solution. Euler’s equation is of the form 22 —~ 2y = 0, that is, y = 2. Since the boundary conditions are satisfied,4. AN ELEMENTARY PROBLEM, EULER'S EQUATION 63 it, foliows that on the straight line y =a the integral \ y (2e—y) dz can attain an extremum. For other boun- e dary conditions, for example, y (0) = 0, u(¥) a4, the extremal y == z does not pass through the boundary points (0, 0) and (F > 4) so that for these boundary conditions the variational problem does not have a solution. 2°. F is linearly dependent on y’, that is, FayY=MayrVG@yy Euler's equation in this case has the form eM an “ae = (20) This equation, as in the caso of 1°, is a finite and not a differential equation. The curve defined by the equation ao 2 == 0 does not, generally speaking, satisty the boundary conditions and, hence, the variational problem does not, as a rule, have any solution in the class of continuous functions. we Se 0 in some domain D of the ay- plane, then the expression F (x, y, y') = M (z, y) dz+ +N te, y) dy is a total differential ‘and the functional 2 2 Jiy@=) PG wy )de= | (Mast ay) @ (a, A) does not depend on the path of integration: the value of the functional J [y (z)l is the same on admissible curves. The variational problem becomes meaningless. Example 9. Investigate the following functional for an extremum: b Sly @= | G+ aay’) az, ye) =A, yO) =P a64 CH. TL. EXTREMA OF FUNCTIONALS Solution. Here, F is linvarly dependent on y’. We have om an aman Sy tte ae ey, and Fay =O Hence, the integrand (y* — 2zyy’) dz is a total differential. Consequently, the integral does not depend on the path of integration: g @ B) Jiy@n= | @aeteeydy= | dia) @ @, A) a= oy ES = BB? — at? no matter along what curve y (z) passing through the points (a, A) and (b, B) we integrate. The variational problem is meaningless. 3°. F depends only on y’, that is, F = F (y’). Euler's equation is of the form Fyyy" = 0 4) In this case, the extremals are all possible straight lines y= Cz + Cy where C, and C, are arbitrary constants. Example 10. Find the extremals of the functional v Jy @l= | VTFEE ae, y@\—4, yO=B 2 This functional defines the length of the curve joining the points (a, A) and (0, B). Geometrically, the problem. reduces to finding the shortest line connecting the given points. Solution. Euler's equation has the form y” (x) = 0. The general solution is y () = Cyz + Cy The extremal that satisfies the boundary conditions y (4) = = A and y (6) = B is obviously a straight line passing through the points (@, A) and (, B): Be, y= God ea) tA4. AN ELEMENTARY PROBLEM EULER'S EQUATION 6 4°, F does not depend on y, that is, F = F (x, y'). In this case the Euler equation x Fy (z, y’) = 0, whence Fy @, 9) = Cy (22) where C, is an arbitrary constant. Equation (22) is a first-order differential equation. Inte- grating it, we find the extremals of the problem. Example 11. From among the curves connecting the points A (1, 3) and B (2, 5), find the curve on which an extremum of the functional 2 Jty@~ | y @d+sy' @) de i can be attained. Solution. Since F is not dependent on y, Euler’s equation is of the form ay (, y’) = 0 or & (4 + 2z2y’) = 0, whence 1+ 227y’ = Cy Then ya Sgh so that y= Sy C,, where C} =: tof : Thus, the extremals are a family of hyperbolas. Let us isolate the extremal passing through the given points. To determine the constants Cf and C, we form the system Be Ci+ Cy, whence CY =e C,=7. The sought-for extremal is y@)=7T—+- 5°. F does not depend explicitly on 2; that is, F = rie y’). In this case, Euler’s equation assumes the orm. Fy Py -y’ — Fyy-y’ = 0 5053966 CH. 1 EXTREMA OF FUNCTIONALS Multiplying both parts of this equation by y’, we get an exact derivative in the left member, that is, =e F-y xX Xx Fy) = 0, whence Fay Fy = Cy (23) where C, is an arbitrary constant. This equation can be inte- grated by solving for y’ and separating variables or by introducing a parameter. Example 12. (Least resistance to a gas flow.) To determine the shape of a solid moving in a flow of gas with least yh Fig.t4 resistance. For the sake of simplicity, we will consider a solid of revolution (Fig. 4). Solution. Assuming that the gas density is sufficiently small and the molecules are mirror reflected from the surface of the solid, for the normal component of the pressure we will have the following expression: p = 2ov* sin? 6 (24) Here, p is the density of the gas, v the velocity of the gas relative to the solid, @ the angle between the velocity and its tangential component. The pressure is perpendicu- lar to the surface so that it is possible to write down the force component along the z-axis acting on a ring of width4. AN ELEMENTARY PROBLEM. EULER'S EQUATION 67 (1 + y’)* de and radius y (z) in the form aF = 2pov* sin? 9 [Ory (+ y')]sin Ode (25) The total force acting in the positive z-direction is t Fe j Amcpo? sin? Gy (Ay) de 26) 0 In order to simplify the problem, let us assume that in Qe yf . Sema SY Then the force of resistance will be i Px bexoo? | y'ty de (27) 0 ‘The problem consists in finding a function y (z) for which F assumes the smallest possible value, and yQ)=0, yQ=R (28) Euler's equation for the functional (27) is of the form y—3-£ (wy) =0 (29) The particular solution y == 0 of this equation is unaccep- table due to the houndary conditions (28). Equation (29) may be rewritten in the form y -» Byy’y” = 0 @0) Multiplying both members of (30) by y’, we notice that the left member is (y'y)’. Integrating, we get yy = C whence Yonge and y= (Ceo) (31) Using the boundary conditions (28), we get RB Cramp, Con 068 CH. I] EXTREMA OF FUNCTIONALS whence y=R(> i)" Thus, the contour with given endpoints for which the resistance of the solid is minimal is a parabola of degree 3/4. Example 13. Find the extremal of the functional e Jiy@n=) Ee ae a which extremal passes through the given points (a, A) and (0, B) lying in the upper half-plane. Solution. Since the integrand does not contain x expli- eitly, Euler's equation yields, by (23), VIFF ay He y vVirye Simplifying, we get yVI7 7% =C,, where C, = a Integrating the latter equation, we find @ + C,)* + y= == €3, which is a family of circles centred on the z-axis, The desired extremal is that one which passes through the given points. The problem has a unique solution since one and only one semicircle centred on the z-axis passes through any two points lying in the upper half-plane. Note. According to Fermat’s principle, the path of a light ray moving in an inhomogeneous two-dimensional medium with velocity v (x, y) is an extremal of the functional mt vee Til= J vz, y) ad 0 If the velocity of light, v, is proportional to y, then, as may be seen from the example examined above, the light rays are ares of circles whose centres lie on the z-axis. Let there be given a curve g. The optical path length q is the time 7 (g) during which the curve is traversed with the velocity of light, uv (z, y)-4, AN ELEMENTARY PROBLEM. EULER’S EQUATION 69 Wo will consider the upper half-plane y > 0 as an optical medium in which the velocity of light at every point is equal to the ordinate of that point, v = y. As we have al- ready seen, the light rays in this medium will be semicircles with centres on the z-axis. It may be shown that the part AD of the semicircle g, one of the ends of which lies on the z-axis, has an infinite optical path length (Fig. 5). We will ¥ Fig. 5 therefore call points on the z-axis infinite points. We will consider the semicircles with centres on the z-axis to be Straight lines, the optical path lengths of the arcs of such semicircles to be their lengths, and the angles between the tangents to the semicircles at their intersections to be the angles between such straight lines. We obtain a flat geometry in which many of the propositions of ordinary geometry are preserved. For example, one and only one straight line can be drawn through two points (only one semicircle centred on the z-axis can be drawn through two points on the semicircle). Two straight lines will be said to be parallel if they have a common point at infinity (that is, two semicircles tangent to each other at a point B lying on the z-axis). Then it is possible to draw, through the given point A not lying on the straight line g, two straight lines q, and g, parallel to g. The straight lines that pass through point A and lie in the vertical angles I and III intersect the straight line g. The straight lines lying in the angles IT and IV do not intersect g.7” CH. Ty EXTREMA OF FUNCTIONALS We have obtained what is called the Poincaré model of Lobachevskian geometry in the plane (Fig. 6). yh | | | | I Fig. 6 Find the extremals of the following functionals. LJ BL. Sty @) = | oy t t+ ey’ }aes ya=4, ¥Q)=B. 8: w& 4 » J Ty (a) = j (e¥-L ay’) da; y(0)=0, y (1) =a. % 83. oO wh Jw =) wt-vyés yO=t, o(F)=¥. 0 8 S Ty =) Ww) da VO=41, y@)= 1. a L 85. JIy (@p =} @+y7 dei yO)=1, yt) =2. 0 4 86. Sty =) G+y) da; VO)=O, yt=t. rs OF 87. Ty @= |W? Fade yO=e, y)—t.4 AN ELEMENTARY PROBLEM, EULER'S EQUATION vie 4 88. J [yG)) = | Qe) des yO) =1, y() =e. 9 v 89. J (y@) = | (ey ty" de. » 3 90. Fey) =) (yr) de 91. Show that the linear functional b Fiy@= | wey +a) y-ray de a where p(x) €Cy la, I, ¢(@) €C la, b], r(z)€Cle, BI, does not have any extrema. 92. Given the functional b Jiy@= |r, y, y)ae and the boundary conditions y (a) = A, y (0) = B. Show that if to the integrand F (2, y, y"}dz we add the total differential of any function u == u (x, y), then Euler’s equation will remain unchanged. , 98. Sty Gy) am | Gey 2ye") de. 3 R/2 9%. Jty@l= | y—v?—Sycosh 2) de; o y (0) = 2, u($) =2e0sh 95. Find the extremals of the functional by Jiu@l= | ya at72 CH. If EXTREMA OF FUNCTIONALS and show that for x = 1 two points lying on different sides of the y-axis cannot be joined by an extremal. Variational problems in parametric form. In a number of problems it is more convenient and sometimes simply necessary to make use of a parametric representation of the lines, @ f= Q } ; O33) For example, in the functional Jo= | edy—yde c the integrand is positive homogeneous of first degree. Indeed, ere Fay a= —ye and clearly Pe, y, kt ky) = RF (@, y, 2, 9) If the line 2: vest 0—constant. 5. GENERALIZATIONS OF THE ELEMENTARY PROBLEM OF THE CALCULUS OF VARIATIONS 1°. Funetionals dependent on higher derivatives. Sup- pose we have the functional xa Jy @l= | FG, yz), y'(@), -.., ¥™(a))de (A) where F is an n-+ 2 times differentiable function with respect to all arguments, y (z)€C,[xp, 2], and the boundary conditions have the form YA0) = Yor Y" (a) = Yay oo oy YOO (eq) SY G-D, 2 YQ) te Yi (a) eH YOO ey) Ser YD } ) The extremals of the functional (1). given conditions (2). are the integral curves of the Euler-Poisson equation: By iye dehy = HIE Fm a0 Example 1. Find the extremal of the functional 1 Tiy@l= | 802% —y%) de 0 yOu O=1, 70) =0, yy A) = 25 Solution. The Euler-Poisson equation is of the form 36022 +2 (— 2y') = 0 or FV (a) = 1802? and the general solution is ga) —Z + Ce + Coat + Cet Cy6 CH LL, EXTREMA OF FUNCTIONALS Using the boundary conditions, we get Cua, = —3, Cat, C= 0 The desired extremal is y@afeppe—ates Let us consider a problem in which not all the conditions (2) are given on the boundary, but a smaller number of conditions, so that in the general solution of Euler’s equation there remain certain free constants after using the boundary conditions. To solve such a problem it is necessary to find a variation of the functional (1), transform it taking into account the given boundary conditions and, equating the variation to zero, obtain supplementary conditions on the boundary. Example 2. Find the curve y = y (x) that extremizes the functional o Jul=% | wae ® = subject to the conditions y (@) = 0, y (0) = 0 6) Solution. The Euler-Poisson equation has the form y™ (@} = 0 The general solution is y@)=C,+ C+ Ce + Ca 6) contains four arbitrary constants C; (¢ = 1, 2, 3, 4, and the given boundary conditions (5) are insufficient for their determination. Therefore, according to the foregoing, we find a variation of the functional (4). We have a a= Jy" yar ” a5. GENERALIZATIONS OF THE ELEMENTARY PROBLEM TT Integrating (7) twice by parts, we get & BF fy] =u" (2) by" (2) B— | 9” (@) dy’ (@) de z =y (2) by (2) fk —y" (2) by @R ob +) @iy@ ede 0) a Expression (8) should vanish on the extremal y (2) of the functional (4). From the arbitrariness of the function $y (z) it follows that yY (z) = 0. This is the Euler-Poisson function for the functional (4). But if the integral in the right member of (8) vanishes, then the boundary expression Iu" 2) 6y (@)~ 9" @) by @)IE must also vanish identically. Since dy (a) = dy (b) = 0 (the ends are fixed), we find that it must be true that y” (&) by’ (b) — y” (2) dy’ (2) = 0 Since the quantities 6y’ (d) and dy’ (a) are arbitrary, we necessarily obtain y” (@) = 0, y" (®) = 0 (9) The conditions (9), together with conditions (5), unam- biguously isolate the extremal from the family (6): y (c) == 0. 2°. Functionals dependent on wz functions. For a functional that is dependent on m functions ¥, (z), ..., ¥m (2) TY) Yor sey Yel 4 = JP] Yet os tm YoU os valde (10) xo given boundary conditions of the form Yn (%) =the Yala) = UP = 1,2,....m) At)bd CH. DL EXTREMA OF FUNCTIONALS the extremals are found from the following system of second- order differential equations: ad Fy iy=0 (F=t 2, ..., m) (42) called a system of Euler’s equations. Example 3. Find the extremals of the functional 2 Jiy(@),2@)= J (y?p a2") de for the boundary conditions yQ=4, y@)=2, 21) 0, 22) = 1 Solution. In this case the system of equations (12) is of the following form: y’=0, } B— 2" ox Solving this system, we find y= Cr+ C, c= Ce + Ce* By virtue of the boundary conditions we have vd G=1, G=0, Gagiy, Cy so that the desired extremal, y=a, az SOR (2 —1) =~ sink T is a space curve which is the intersection of two cylindrical surfaces. | Example 4. Find the extremals of the functional * J ly @), 2(@)J= J (2ye—2y?-4-y2—2) de ¥() =0, y@)=4, 2) = 0, z@)=—45, GENERALIZATIONS OF THE ELEMENTARY PROBLEM 79 Solution. The system of equations (12) has the form y+ 2y—z=0, } 2’ —-y=0 whence, eliminating the function z, we get y8O + 24” + y = 0 The general solution of this equation is y (z) = C, cose + C,sinz + 2 (C; cos x + C, sin z) By virtue of the boundary conditions y (0) = 0, y (y= 4 we get C, = 0, C= -+ and, hence, ¥ (2) = Cy sin2 + Ce sinz—= cos x We find the function z from the condition that z= y” + 2y and we have z= Cysinz +, (2 cos 2 + 2xsin z) + (2 sin z — x cos z) The constants C, and C,, are found from the boundary conditions z (0) = 0, z (m) = —41; this yields C, = 0 and C, is arbitrary. Then z=Cpsinz-+4+(2¢in zz 0092) The family of extremals is y=Czsinz —= cos 2, a= Cysin2 +4 (Qsine—ze0sz) \ where Cz is an arbitrary constant. 3°. Functionals dependent on functions of several independent variables. Let us consider a functional of the form J [z(z yy= {yr (z, Y, 3, 2, Fi) dedy (43) D80 CH. I. EXTREMA OF FUNCTIONALS where F is a three-times differentiable function of its arguments, and let us assume that a function z = 2(z, y) is being sought that is continuous together with its derivatives up to order two inclusive in the domain D and assumes on the boundary T of D given values and yields an extremum of the functional (43). Tf an extremum of the functional (43) is realized on the surface s = z (z, y), then the function z = z(z, y) satisfies the Euler-Ostrogradsky equation Pima Php Fh =0 “4 where 2 {F;} and Fe UFd ase total partial derivatives with respect to x and y, respectively: a a: a ‘ ag Po) = Feat Fos Get Pop E+ Pp <45) a a: a, Gy Pda Ft Fat he Et hat (16) Here, we put # =D, = = q for the sake of brevity. Equation (14) is a necessary condition for an extremum of the functional (13). It is a second-order partial differential equation; we seek the solution z = z(z, y) that takes on the specified values on the boundary I, Example 5. Write down the Euler-Ostrogradsky equation for the functional rec. on= {§(() (Be) eee Solution, We have F (z, y, 2, p, g =p*—¢ so that, by (14), we got —~ 2 @p) — (24) = 0 or oe a 4 oat Gy? For the functional F [2 (1, XH, -.+, Zp)] = JJ e- ] Plea ze, oes ans 8 Die Pes eves Pa) dtydey Db w+. day (47)5. GENERALIZATIONS OF THE ELEMENTARY PROBLEM 8 where py = & (k= 1, 2,..., 7), the necessary condition of an extremum is expressed by the following Euler-Ostro- gradsky equation: a Fe~)) 5 Fp) =0 (48) teed or, expanded, n te 4 Fe~ 3) (Fey Foy Det DS) Fo, Bt) =0 (49) tend j=t The solution of this equation—the function z= 2 (x, ta, ..- ++ +1 %)—on the boundary I’ of the z-dimensional domain D must satisfy the given boundary conditions. Example 6. Find the conditions for which the function 2 (ty, Za, - 2 oy fs which assumes specified values on o! the boundary I‘ of the domain Q, minimizes the Dirichlet integral De= ff iP (EY depen... den * Solution. In this case, F= 5 (#)* that is, F does eet not explicitly depend on a, zg, ..., Za, % Hence, 2 when t=], Fe=Fen=Fp,, Foo { 9 wnen tay and by (49) we got 2 > a7 or Az=0 fat (an n-dimensional Laplace equation), Note. If derivatives of the function (z,-y) up to order appear under the integral sign, then the Euler-Ostrogradsky 6053982 CH. if. EXTREMA OF PUXCTIONALS equation becomes AoE I-E+ Fad +2 ghgy Peg) + Ge Fagh HHI EE Fay. =O 20) Example 7. Write down the Euler-Ostrogradsky equation for the functional Jie (z, = SN [Cay + (ae) +2 (By) 24 Jena Solution. We have Fea ( Se) + (Se) +? (Gy) Fe w According to (20), we find —2t ia (2S) +e (2-H) +2afe5 (2 aba) =° or az oz ain sat? ara +aa=1 (, ¥) This latter equation is briefly written thus: AAz = f (2, y) Find the extremals of the following functionals. i 99. Jiy = | + 2y7 +7) ars o yO=0, y()=0, yQ=4, y(t)=—sinbt.5. GENERALIZATIONS OF THE ELEMENTARY PROBLEM 8 9 100. Fty(ayi= | (240y—y") de; —t y(t) 4, yO)=0, y'(—1)= —45, ¥O)=0, y(—1)=16, 9" (0) =0. 101. twen={ wt: ¥()=¥0 O=v ¥@=%, 9 b)=y. 402. 7 wer=| Yt yy) de y@j=A, ¥ (e)= dy y@=B, yy O=B, 4 108. Jiy(ai}= | (y? 2y) de; 6 y(O0)=0, y(i)=sinht, y'(O)=4, yy’ (1)=coshd. 104. Find the extremal of the functional t Jyi=¥ | wae ° ubject to the conditions y(0)=0, y’ (O)=0, 9 (A). Hh 105. J [y(z), 2(2)) = j (22 — Ay? -b y’® — 2/2) das 2 yO=0, y(Z)=t 2@=0, 2(Z)—t. 4 2 106. J ty (zs 2a) | (2ey—y2+ 45) de; ai yit)=0, y(—A)=2, sat, e(—t)=—4, o84 CH. Ir, EXTREMA OF FUNCTIONALS 107. F{y (a), 4 @)) -j (y? 2? — 2ya) day yO=0, y($)=1, 2Q=0, 2($)=1. 4 108. J ty), s(ay= | (y? + 2'2 42g) dar; o 3 yO)=1, yQ=s, 2(0)e0, 2(t)=1. 109. Show that the Euler equations Jor tho functional 5 Tin t= | Fey y' a 2) de 2 admit the following first integrals: (4) Hac if F does not contain y; (2) Foy ad eal if F does not contain z. Write down the Euler-Ostrogradsky equations for the following functionals. 140. J [(z, y= i [(E)'+(E) verte] axay. 414. Fie@ y= | f (35+ $3) an ay. D Gat Gy 142. J [2 (a4, te, «.., tn)] SI Mace 20a) 6 (gy vvey Zp) Pp Qef Cy, woes tn) ] dan dig -.. da. P 113. Derive the differential equation for minimal sur- ‘aces.6 INVARIANCE OF EULER'S EQUATION 85 414, Find the extremal of the functional 144 Thalz, y= S ‘ e'W'sin 2, dz dy 00 given the conditions z (x, 0)= 0, ¢(2, 1)== 1. 6, INVARIANCE OF EULER'S EQUATION If a functional , Fin=J PCy yee can be transformed by replacing the independent variable or by a simultaneous replacement of the desired function and the independent variable, then the extremals of the functional are, as before, found from the Euler equation set up for the transformed integrand. Therein lies the invariance of Euler’s oquation. Let 2 = 2(u, v), y = y (u,v), and fu 2p Yu Yo 0 Then 5 F(z, y, y')dz = ‘ F [z@, v), y(z, v), Bee] (Sq-+ typ) du cd fo (w, B, Vy) dw and the extremals of the original functional are determined from the Euler equation for the functional Sow», vy) du: ®,— £0, =0 Example 1. Find the extremals of the functional tt Jin=|VFTr de, whee r=r(9) On86 GH. Il. EXTREMA OF FUNCTIONALS Solution. The Euler equation for this functional is r ad re pase (yap) 9 ‘The change of variables r= rcosg, y= rsin@ yields Vari dg= Vir yds and we arrive at a functional of the form b Jial= | VIF yee for which the Euler equation is y” = 0 so that y= Or +C, Hence, the extremals of the original functional are. given by the equations rsin p= Cyr cos p + Cz where C, and C2 are arbitrary constants. Example 2. Find the extremals of the functional Ing Jivi= | ety?) ae 0 Solution. Euler’s equation for this functional is of the form yy’ ey = 0 ‘We make a change of variables: a=lnu, {yao Then the original functional becomes z 2 Jil = J (e7 muy! elt uy) OE | (ve — v2) du i a and for it the Euler equation v” + v= 0 is readily integra- ted: v= C,cosu +Czsinu6. INVARIANCE OF EULER'S EQUATION 87 Passing to the original coordinates z, y, we get the equation of the extremals in the form y= C, cose - C, sin &* 115. Find the extremals of the functional O41 Js j rsing Vrier dp Po. 116. Show that the extremals of the functional Oo Im J f(irsing) VP=r? dp Po can always be found in quadratures. 447. Find the extremals of the functional b Jal VPTR VIN a As in the case of one variable, the Euler-Ostrogradsky equation is invariant with respect to transformations of coordinates. Example 3. Write down the Laplace equation a. at pe te =O @) in polar coordinates. ‘olution. Consider the functional Dists, i= | | +a) deay G The Euler-Ostrogradsky equation for the functional is pre- cisely equation (1). Let us pass from Cartesian coordinates (2, y) to polar coordinates (0, @): == 9 cos, y= o sin g. e have 6p ap * eo SiR Gp cos @ az = OOS P, P= SING “Fe py88 CH. U1, EXTREMA OF FUNCTIONALS whence D{z(e, 9] ~{[lloBre)'+ (B48) Joanie =) J (08 + 428) do do e Forming the Euler-Ostrogradsky equation for the latter integral, we arrive at the Laplace equation in polar coordi- pates 4 : py tee tT Plo tap =O 7. FIELD OF EXTREMALS A family of curves y= y (z, c) forms a proper field in @ given domain D of the zy-plane if through every point (z, y) of the domain there passes one and only one curve of the family y= y (2, ¢). The gradient p(x, y) of the tangent to 4 curve of the family y= y(z,c) passing through the point (z, y) is called the slope of the field at the point (z, y). A family of curves y= y (w, ¢) forms a central field in a domain D of the xy-plane if these curves cover the entire domain D without seli-intersections and issue from a single point (x, ¥,) lying outside D. The point (zp, yo) is called the centre of the pencil of curves. Example 1. Inside the circle 2? -+-y* <4, the family of curves y= Ce*, where C is an arbitrary constant, in particular, C== 0, forms a proper field, since these curves do not intersect anywhere and through every point (z, y) of the circle there passes one and only one curve of the family (Fig. 7). The slope of the field at an arbitrary point (z, y) is equal to O(a, y) = CeFmy Example 2. The family of parabolas y= (« + C}* inside the circle x* + y*> <1 does not form a proper field since different curves of the family intersect inside the circle and do not cover the entire domain (Fig. 8).COM90 CH. I. EXTREMA OF FUNCTIONALS Example 3. The family of curves y= Cz forms ti field in the domain t > 0. y * forms a central Do the following families of curves form fields (proper or central) in the indicated domains? 118. y= C-tana; Oe, ~FSu0 o7. FIELD OF EXTREMALS 4 125. Do the same for the functional wh Jil= | WP 428+ 4d 0 Let the curve y= y (z) be an extremal of the functional my Jil= | F(a uy) ee xo oe extremal passes through the points A (zo, y,) and Zr, Yr). We say that the extremal y= y(z) is included in the Proper field of extremals if there is a family of extremals y= y(z,C) that forms a field containing the extremal y= y (2) for some value C= C,: this extremal y= y (x) does not lio on the boundary of the domain D in which the family y= y (x, C) forms a field. Tf a pencil of extremals centred at the point (zo. yo) in the neighbourhood of the extremal y= y (x) passing through the samo point forms a field. then we say that a central feld has been found that includes the given extremal y= y(z). For the parameter of the family y = y (x, C) we take the slope of the tangent to the curves of the pencil at the point (x». 4). Example 5. Let us consider the elementary variational problem for the functional 2 Fiyl= | (sin? 2) de 0 ‘a) Let y (0)== 1, y (2)= 1. The family of extrémals of ant ‘ith 5 onal As determined by the equation y= = Cyz + Cy. The given boundary conditions are satisfied by the extremal y= 1. This extremal is included in the proper field of extremals y= C2, where C, is an arbitrary constant. | (b) Let y (0) = 9, y (2)== 4. The extremal corresponding to these boundary conditions is the straight line y= 2z, which is included in the central field of extremals y,=92 CH. IL EXTREMA OF FUNCTIONALS == Cyz (C, is an arbitrary constant) with centre at the point ° ample 6. Consider the elementary variational problem Jil= Jy'-(22-dy’) de, y(—t) =0, y=} The solution of the Euler equation is of the form y= == 2? — Cyz + Cy. The extremal of this problem, y= z? ++ y Fig. 9 +£~—3 » may be included in the proper field of extremals y = 2° + > + Cz (Fig. 9). Show that the extremals of the following elementary variational problems may be included in the extremal field (proper or central).7. BIELD OF EXTREMALS 98 tae 126. J iy]= | (y’*— 2zy) das y(0)=y (1) =0. o 4 127. Jivl= | QFyty? ax yO=1, y=e 0 128. Tiyl= | W—v) de (@>0, aed); 0 y(0)=0, y(a)=0. 2 129. Ful=] w+ 2%de yQ=1, yQ)=3. ° Definition. Suppose we have a family © (z, y, C) =0 of plane curves. The C-discriminant of this family is a locus of points defined by the system of equations D(z, ¥, C)=0, oO (zy 9 } (4) oc ~ In the general case, the C-discriminant includes the envel- opes of the family, the locus of nodal points, and the locus of cusps. The envelope of the family ® (x, y, C) = 0 is a curve, each point of which touches a certain curve of the given family and each portion of which is touched by an infinitude of curves of the family. If we have a pencil of curves with centre at the point A (Zp, Yo), then the centre of the pencil belongs to the C-discriminant. Example 7, Find the C-discriminant of the family of curves y = (2 — C)*. Solution, In this case, equations (4) have the form Ve =O} 2(t—C)=0 whence y = 0. It is easy to verify that the curve y = 0 is the envelope of the given family. Indeed, at any point 2 = Zp the curve y = 0 has a common tangent with the cor-4 CH. 0, EXTREMA OF FUNCTIONALS responding curve of the family y = (x — z,)*. Furthermore, no matter how small a portion of the curve y = 0 we take, an infinitude of curves of the given family are tangent to it. In the given instance, the C-discriminant consists of the envelope alone. Find the C-diseriminants of the given families in the following examples. 130. y = Cz + C% 131. y(C~2)—~—C?= 0. 182. (c — CP + Y= 1. If an are AB of the curve y = y (x) has a point A* (different from A) in common with the C-discriminant of gh | 4RD Fig. 40 the pencil y = y (x, C) with centre at the point A, which pencil contains the given curve, then the point A* is said to be conjugate to the point A. Example 8. Let us consider a one-parameter family of curves y = € sinc, The C-diseriminant of this family is defined by the equations y~Csing =0, } —sinz= 0 That is, it is a discrete set of points (km, 0), & = 0, +4, +2 (intersection points of the sine curve with the z-axis). For instance, taking C = 2, we get the curve y = 2sin 2, which belongs to the given pencil of sine curves centred at the point O (0, 0). If the other end B (Fig. 10) of the are2, FIELD OF EXTREMALS 95 of the curve y = 2 sin z has the abscissa x € (x, 2m), then the are OB will contain another point [in addition to the point O (0, 0)] belonging to the C-discriminant, namely the point O* (, 0), which will be conjugate to the point O (0, 0). If 0 <2 < x, then there will be no points conjugate to the point O (0, 0) on the are OB. 133. Given a family of curves y= C+(x — 1) 2. Find the point conjugate to the point O (0, 0). 134. Given the family of curves y == C-sinb x. Find the point conjugate to the point O (0, 0). 1°. Jacobi’s sufficient condition for including an extremal in a central field of extremals. In order to be able to include an are AB of an extremal in a central field of extremals with centre at the point A (xp. yo), it is sufficient that the point A4*, conjugate to the point A, not lie on the are AB. Example 9. Let us consider the functional Jiy a= | ytd?" —1) da, Qo y@)=0, yla)=0 Verify the possibility of including the extremal y = 0 in a central field of extremals with centre at the point 0 (0,0). Solution. The Euler equation of the given functional is of the form y” + 9y = 0. Its general solution is y (%} = = C, sin 32 + Cz cos 32. li ax es » k an integer, then the extremal that satisfies the given boundary conditions is the straight line y = 0. If we consider the one-parameter family of extremals y, = == C, sin 3z, then, as it is easy to verify, the C-discriminant of this family consists of the points (%, 0) , kan integer. Therefore, if a,96 CH. Il. EXTREMA OF FUNCTIONALS then on the extremal y = 0 there will be at least one point conjugate to the point O (0, 0), and Jacobi’s sufficient condition is not fulfilled. In that case, the extremals y = == C, sin 32 do not form a field. Analytic form of the Jacobi condition. Suppose we have an elementary variational problem: Fy@= | Fey, y) az, me y(%)=Yo Y(a)=y If the solution u = u (z) of the Jacobi equation (Fu —& Far) u— 2 Fyyu')=0 (2) that satisfies the condition u (z,) = 0 vanishes at some other point of the interval z) > 0. Hence there is no point conjugate to the point O (0, 0) on the arc OB of the extremal. Therefore, it may be included in the central field of extremals with centre at the point O (0, 0). It is easy to verify that the desired extremal is the straight line y = da, which clearly can be included in the central field of extremals y = Cyz. Example 14. Is the Jacobi condition fulfilled for the extremal, passing through the points A (0, 0) and B (a, 0), of the functional Jiyl= { (yy? +e) dz (o# (n+4) x)? g Solution. The Jacobi equation has the form u” + 4u = == 0. Its general solution is u (2) = Cy sin 2z + Cz cos 22 From the condition u (0) = 0 we find that Cz = 0 so that u(z) = C, sin 2c. If aF: then the solution of the Jacobi equation w=, sin 22 vanishes at the point 2 = = which lies in the interval [0, ¢], and on the are of the extremal y = 0 (0 <2 > = there does not exist a central field of extremals that includes the given extremal. In the following examples, show where the Jacobi condition is fulfilled. zt 135. Ftyl= | (Adey-ty?-+2%) ax; “4 ¥(—t=—2, y(t=0. 136. Jiyl= | y2+9y'—~Sa)de; y(0)=0, y (a) =0. 0 1053998. CM. IL EXTREMA OF FUNCTIONALS 4 137. Jiyl= | d+ yds; yO)=yh) =9. a 138. Ji) | yer a y(O)=4, y()=b (a>0). 6 20 139. Fiyl= | G2) da; y(0)=0, y 2mm o 140. Show that if the integrand of the functional b Jul=)F@y)de z does not contain y explicitly, thea each extremal can always be included in the field of extremals. Note. The Jacobi condition is necessary for attaining an extremum of the functional J [y (2)]; that is, on the extremal AB that realizes an extremum, the point conjugate to A cannot lie in the interval 2»<( 2 2. Hence, the extremal y (7) = 0 does not give the given functional a minimum since there are curves close to y (z)s=0 on which the values of the func~ tional are negative. Now let us take the family of curves Ya (2) = 4 sin 4 « that have proximity of any order with respect Lo the curve y (z) =z 0. It is readily seen that 5 as ore a J lyn) == {2+ ix ae~ | ¥ py osttadra Pt. > 0 a 0 Hence, the extremal y (x) «= 0 does not give the functional a maximum either. 141. In the functional b T= | Fe. yy) ae Jet us suppose the integrand F has bounded partial derivatives of third order with respect to the variables y, y’ in any bounded range of y and y’. Show that if y = y («) and ye100 CH IL EXTREMA OF FUNCTIONALS y = y (2) + 1 (z) are two close extremals, then the function 1 (z) satisfies the Jacobi equation to within a quantity of higher order compared with the first-order distance between these extremals: ‘ 1 . Py Fy — ay Fut Fy) #9 2°. Sufficient conditions of Legendre. A sufficient condition for including an extremal of the functional xa Jtyl= ‘ F(z,y,y') da, x0 U(%)=Yo, y(t) = ys in the field of extremals is the fulfilment of Legendre’s strong condition. It consists in demanding the fulfilment of the inequality Fyy > 9 at all points of the extremal under consideration (that is, for all z € [zp, z]). Example 13. Given the functional 2 Jivi= | w+y") dss a yO=1, yQ=5 The extremals are the straight lines y = Cyr +-C,. The desired extremal satistying the given boundary conditions is the straight line y = 2z + 1. the given case, Py = 12y* -- 2, and at all points of the extremal y= 2z-+4 we have Fyy = 50>0. Legendre's strong condition is fulfiiled and, hence, the extremal y = 22 +1 may be included in the extremal field. This is also evident directly. The extremal y = 2x -+- 4 Hes in the one-parameter family of extremals y = 22 +4 (x a parameter) that form a proper field.7. FIELD OF EXTREMALS 404 Example 14. Given the functional 4 ) wy? +12y) de, m4 y=, yA) =4 Solution. The Euler equation for this functional is of the form wy” + acy’ — 12y = 0 Its general solution is y= Cy + Con ‘The extremal y = 2° satisfies the given boundary conditions. It cannot be included in the field. The only one- parameter family of extremals that contains it is the family y = az’. This field does not cover the region containing the point with abscissa « = 0 (the extremals of this family do not pass through points of the y-axis with ordinates different from zero). In the given case, Fyy = 22? and the Legendre condition is not ed forz = 0. Check the possibility of including an extremal in the feld for the following functionals. 1 142. J (yl== J G2 yy) des y (0) =0, y(t) =0. 148. rn fren ¥(0)=0, y@)=b>0. xt 144. Ttyl= | ny) ViFy* eae; ° yo) Yo, Y(t)=yr, my) > 0. 145. J [y= j (6y'2—y"4) da; 0 yQ=0, y@)=b, «>0, b>0,402 CH. I. EXTREMA OF FUNCTIONALS 8 SUFFICIENT CONDITIONS FOR THE EXTREMUM OF A FUNCTIONAL We consider the most element variational problem for the functional ry , Tlyl= fre, yy’) da, (4) =o ¥@) = Yo y= (2) 1°. Sufficient conditions of Weierstrass. The Weierstrass function E (z, y, p, y’) is a function defined by the equation E (zu. py) = F(z, y, y') ~ F (ew, y, p) ~~ —p)Fp(z,yp) (8) where p = p (z, y) is the slope of the extremal field of the variational problem (4)-(2) under consideration at the point (2, y). Sufficient conditions for a weak extremum, A curve C gives the functional (1) a weak extremum if: 1. The curve C is an extremal of the functional (1) which satisfies the boundary conditions (2); that is, for the funetional (1) et 5° solution of Euler's equation that satisfies the conditions (2). 2. The extremal C may be included in the field of extremals; this will occur in particular if the Jacobi condition is fulfilled, 3. The Weierstrass function E (z, y, BP, y') must preserve sign at all points (a, y) close to the extremal C and for values of y’ close to p (x, y). The functional J [y] will have a mazimum on C if 8 <0 and a minimum HESS. Sufficient conditions for a strong extremum. al curve C gives the functional (1) a strong extremum ci x She are & Be tememe af Se amen \ij tres ati the boundary conditions (2). . ee extremal C may be included in the field of extremals. 3. The Weierstrass function E (z, y, p, y') preserves sign at all points (x, y) close to the extremal C and for arbitrary values of y’. There will be @ maximum for E <0 and a minimum for FE > 0. Note. The Weierstrass condition is necessary for the existence of an extremum in the following meaning: if for8 SUFFICIENT CONDITIONS FOR THE EXTREMUM 103 certain values of y’ at the points of the extremal the function £ has opposite signs, then a strong extremum is not attained, If this property occurs for values of y’ arbitrarily close to p, then a weak extremum is not attained oither. Example {. Test for an extremum the functional t Jil=| wry 3ds, yO=0, y()=2 o Solution. The Euler equation for the given functional is of the form y’y” = 0 so that the extremals are the straight Umes y (z) = Cie + Cy. The straight line y = 22 is the extremal that satisfies the given boundary conditions. The slope of the field at the points of this extremal is p = 2. It is obvious that the given extremal y = 2z is included in, the central field of extremals y = Cz with centre at the point O (0, 0). It is also easy to verify that in the given case the Jacobi condition is fulfilled. In the given case, the Jacobi equation has the form £ (6y’n"} == 0, where we have y’ = 2 by virtue of the equation of the extremal. Hence, the Jacobi equation is of the form u” (z) = 0, whence wu (z) = Cyt + + C2. From the condition wu (0) = 0 we get Cy, = 0. Since for C, == 0, except the point z = 0, this solution u = Cy does not vanish anywhere, the Jacobi condition is fulfilled. We form the Weierstrass function: E (x, yy Bs’) = YY? by — p? — p — (y’—p) (Bp?-+1) = (y’ ~ p(y’ + 2p) The first factor is always nomnogatins far arbitrary y’. while She seuout Une eins lot vee ME ff ttn eZ. thle all the conditions for the existence of a weak minimum are falfilled. However, as is readily evident, ify’ < —4, the function E will be negative, and the sufficient condition for a strong extremum is not fulfilled, since under the eondi- tions of a strong’extremum it is required that the Weler- strass funetion & preserve sign for all values of y’. Taking into account the note on page 102,"we conclude that in this tase there is no strong extremum.104 CH. If. EXERCMA OF FUNCTIONALS Example 2. Investigate for an extremum the functional 1 : 4 Tul= { (s=2y-- $y") da, é y(0)=0, y(l)=0 Solution. The Euler equation for this functional is of the form y” = 2. The extremals are parabolas y == 27 + + Cx + Cz. The extremal satisfying the boundary conditions is y = 2? — 2, We set up the Jacobi equation (u') == == 0 or wu” =0. Its goneral solution is wu (z}) = Cyz + Cy. The condition wu (0) = 0 yields Cz = 0, and since for C, 0 u(z) = Cyt does not vanish anywhere on the interval [0, 4] except at the point = 0, the Jacobi condition is fulfilled and, hence, the extremal y = = 2* — ¢ may be included in the central field of extremals with centre at the point 0 (0, 0), namely: y = 22+ Cx (Fig. 11). The Weierstrass function is of the form E (ey Pry’) =¢ Gy — py, Fie. 14 whence it is seen that for arbi- e trary values of y’ we will have E= + (y’ — p)* > 0. Hence, on the extremal y = 2* — x the given functional attains a strong minimum equal to Jit — l= 4. Test for extrema the following functionals. t 146. Jiyl= fer (v +54) des yO=4, y=e, o& SUFFICIENT CONDITIONS FOR THE EXTREMUM 405 107. J tylex | evy'tdey yO)=0, y(t)=n& rp dx; y(t)=t, yQ)=4 t 148. J [yl = 149. J (y]= | 3 yO)=0, y(a)=b, a>0, b> 0. oy Bota ota le 5 150. J fyl= \ d+) y*da; y(0)=0, y(=1. eS 151. Jyi= | yds yO=4, ¥(F)=1- 152. J [yl | y Gey’) day y(~t)=4, ¥Q)= tar ot} i oy 453. J [y|= jotty an y(—=—1, yl) =3. 2. Sufficient * conditions of Legendre. Let a function , (zy, y’) have a continnous partial derivative (z, y, y’) and let the extremal C be included in the sae of extremals. If on the extremal C we have Py.) Pe then a weak minimum is attained on the curve C; if Fy-y <0 on the extremal C, then a weak maximum of the 1 Fnctional (4) is attained on it. These conditions are known as Legendre’s strong conditions. When F,,- (z, y, y') SO at points (z, y) close to the extremal fois for arbitrary, values of y’, we have a strong minimum; when Fy-y (2, y, y’) <0 for the indicated values of the arguments, we have a strong maximum. Example 3. Test for an extremum the functional L Jin= | @*—ey)dz, ¥)=0, y= —2 Q (w any real number).406 CH, UL. EXTREMA OF FUNCTIONALS Solution. Since the integrand depends only on y’, the extremals are the straight lines y == C,x -+ C,. The extremal satisfying the boundary conditions is the straight line y = =x 2x, which may be included in the central field of extremals y = Cz. On this extremal, the slope of the field is = —2, We then find F,-,. = 6y'. On the given extremal we have Fy. = —12-<0, that is, a weak maximum of the functional is attained on the curve y = —2z. The sign of Fy-y- is not preserved in the case of arbitrary values of y’; hence, the sufficient conditions of a strong maximum are not fulfilled. In the given case, the Weierstrass function E£ (a, y, p, y’) is of the form E (ey, py’) = (y — py? y+ 2p) and for certain values of y’ it has opposite signs. Taking into account the note on page 102, we find that there is no strong maximum. Example 4. Test for an extremum the functional 2 Til J e7+3)de, yO)=0, y(2m4 8 Solution. The extremals are the straight lines y = Cyz + -- Co. The extremal that satisfies the boundary conditions is the straight line y = 5 It can be included in the central field of extremals y= Cz. In the given case, Pyy (a, y, y’)= =e’ > 0 for any values of y’. Hence, the functional has a strong minimum on the extremal y = + : Example 5. Test for an extremum the functional (iar 0) =0 = Ftyl | a dz, y(Q)=0, yla)=y Solution. The integrand does not depend explicitly on 2; hence, we get F —y’- Fy’ = C, or, in our case, ViEve ee E, Vi ViVi8. SGFFICIENT CONDITIONS FOR THE EXTREMUM 4107 whence 4 rad Vivir oF Ves = es where Ong. Set y =coty. We then have i yer Cysin® fae SE (A cos t) Furthermore, 4; deme A oe ASIA 2 Cy sinh at Cot 2 cot = Integrating, we get aa, { Coed 4 sin) Cy Thus, za Cy (f~sin#) + Cr, y= C,(1—cos2) } are the parametric equations of a family of cycloids. From the condition y (0) = 0 we find that C, = 0. The pencil of cycloids f= € (E~sind), } y=C (1—cost) forms a central field with centre at the point O (0, 0), which field includes the extremal z= R(t—sint), \ y= R(t{—cos#) where R is determined from the condition of passage of a cycloid through the second boundary point B (a, y,) if a@o y (198)408 CH. I EXTREMA OF FUNCTIONALS for any values of y’. Hence, for a0). oun Zi 158. J(yl= | y’*da: y(0)=p>0, y(t)=a>0. 6 159. Test for an extremum the functional 4 Jii=Veyt+ysraten yO)=0, yM)=4 6 for various values of the parameter e.8, SUFFICIENT CONDITIONS FOR THE EXTREMUM 409 Example 6. (Euler's problem.) A rod of length J rests on its ends and is subject to a pressure P. The rod is deflected longitudinally at a given value of P (Buler’s critical force). Tt is required to determine the smallest force P that yields a longitudinal deflection. Solution. Let E be the modulus of elasticity, I is the smallest moment of inertia of the cross sections of the rod, pis the radius of curvature, and is the angle of the tangent with the axis, The potential energy of deflection is given by the formula D4 BI \ 3 When an end of the rod dips by the quantity Om \ (1—cos g) dS the potential energy of the rod decraases by Ug == Paz: Pl P j cos ¢ dS Q If the potential energy prior to the deformation was zero, then after the deformation it is expressed by the formula L U=U,—Us= | ($-Bl-4-+ Poosq) aS— PI 0 Since p = s and (in the case of small values of 9) cos 9 wi ©, it follows that va [er(se)—ew asm | [20 ( 2) In the case of equilibrium, the potential energy assumes a minimal value. Therefore, the solution of the problem440 CH. IL. EXTREMA OF FUNCTIONALS reduces to determining the minimum of the integral Tile i [ez (42)'— Pe] ax Ge In the given case, Fa BI (32)? py and Euler’s equation assumes the form qy’ -ap=0, where ae fe The complete integral of this equation is g@ = C, sin wx + Cy cos ar Since we have tan 9 ~ @ for small values of » and, besides, tan @ = y’, it follows that y’ = C, sin ae + Cy cos ax whence : y@=— Lxc0s oe + Easin as ~¢ If the lower end of the rod lies at the coordinate origin, then for a = 0 we will have y = 0 and, hence, C, = C =0 an y(a= a sin at Let us see whether the Legendre and Jacobi conditions are fulfilled. The Legendre condition is clearly fulfilled: Fra I >0 Jacobi’s equation is Els" + Ps = 0 or 2” + ag = 0 and z (0) = 0. Therefore, the solution of Jacobi'’s equation is a= Asin az§. SUFFICIENT CONDITIONS FOR THE EXTREMUM qit The function z vanishes for 2, = ant, 2,...) so that the Jacobi condition will be fulfilled it 2 > = » whence P>Z er The smallest value of Euler's critical force will be Pain EI Here, Cy. ne y= -Z sin is the equation of the deflection curve. 3°. Figuratrix. Suppose we have a functional & Jul= | Fey) a We will consider z and y to be parameters and we will examine the function Y = F (z, y, y’) as a function of the argument y’. The graph of this function in the plane of the variables (y’, Y) is called a figurairiz. It is easy to show that the Weierstrass function £ (z,y,p,y’) is the difference between the figuratrix and the tangent to it drawn at the point with abscissa y’ = p. The fact that the Weierstrass function preserves sign for certain values of y’ means that the figuratrix lies above the tangent or under it for the indicated values of y’. In that case we have a weak extremum. If the figuratrix lies to one side of the tangent for all values of y’ and for values of the parameters x and y close to the points of the extremal, then we have a strong extremum. In these terms, the sufficient Legendre condition reads as follows: if for all points (a, y) close to the extremal the figura- trig is everywhere convex or concave, then we have a strong extremum. Example 7. Test for an extremum the functional a Jil=Jy2dz(@>0), yO)=0, ylmb, b>0 oO412, Of IL EXTREMA OF FUNCTIONALS Solution. The extremals are the straight lines y = Cyr + ~r Cy. The desired extremal is defined by the equation y= Ze. It is included in the central field of oxtremals. In the given case, the figuratrix is a parabola, Y = y’* (Fig. 13). It is easy to see that the figuratrix lies entirely yh | Fig¥43 above the tangent drawn to it at the point p = 2 for any @ and 5, a0. Hence, the extremal y= 2x gives the functional a stroug minimum. Example 8. Test for an extremum the functional Jil=|y%de, y(0)=0, y@)=d, b>0 o Solution. The sought-for extremal is the straight line y= 2x, which is included in the central field of extremals y == Cz with centre at the point O (0, 0). The figuratrix is a cubic parabola, Y = y® (Fig. 14). For values of y’8. SUFFICIENT CONDITIONS FOR THE EXTREMUM 13 sufficiently close to the value p = 4, the figuratrix lies above the tangent to it drawn at the point with abscissa y= 2. From Fig. 44 it may be seen that the figuratrix ze + 5 + 2b intersects the tangent at the point with abscissa y’ = — = ¥ and lies above the tangent to the left of this point. Hence, a weak minimum is attained on the extremal y =z Note that if p = 0 (this corresponds to the case b = 0, the extremal is a portion of the z-axis), then the tangent to the figuratrix is the y’-axis, and the point O (0, 0) itself is an inflection point of the figuratrix. Taking into account the note on page 102, we see that in an arbitrarily small neighbourhood of the point O(0, 0) the figuratrix has positive and negative ordinates. Hence, the Weierstrass function E has opposite signs for values of y’ arbitrarily close to p = 0 and, consequently, in this case a weak extremum is not attained either. $0599444 CH. IL EXTREMA OF FUNCTIONALS _ Example 9. Show that the extremal y = 0 of the varia~ tional problem t Jil=)@%—w) ar, yQ=yit)=0 0 gives the functional a weak minimum. Solution. The Legendre condition in this case yields Fryy ly=o = (2 — 6yy')lyeo = 2 > 0 In other words, a weak minimum is attained on the extremal y = 0, We will show that a strong minimum is not attained YA Fig. 45 on it. Let us construct the figuratrix ¥ = y'? — yy? for values of y >> 0 (Fig. 15). From the figure it is clear that the tangent to the figuratrix drawn at the point with abscissa p = 0 intersects the iguratrix at the point y’ = +. Thus, for the points (c, y), where y >> 0, that axe close to the points of the extremal y = 0, the Weierstrass function Z is positive for values of y’ less than $+ and is negative for y’ >t. According to the note on page 102, there is no strong minimum. A similar situation océurs for y <0 as well. This example is noteworthy in that fulfilment of the condition Py-y > 0 on the extremal for arbitrary y’ does not imply the existence of a strong extremum.8, SUFFICIENT CONDITIONS FOR THE EXTREMUM 415 Using a figuratrix, test the following functionals for extrema. 160. J [y)= | (1-+-2)yde; y(0)=0, y(t) = ~2. 164. F [y} = tian ote y (Ly!) day y(— day (2) = 1. 162. J [y)== \ (1-—e#) da; a y(0)=0, y(a)=b (a2>0, b> 0). 163. J m= J (By? —y'4 yy’) da; y(0)=0, y@j=b (@>0, b>0). Note. A sufficient condition for an extremum with respect to the second variation. Nonnegativity of the second variation is necessary but not sufficient for a functional J [y] to attain a minimum on a given curve. Example 10. Let us consider the functional Tui=| ¥@e—ye@la 0 in the Space ¢ (0, 1). Euler's equation is of the form F, = 0 ory = 0. The second variation of the functional on the extremal y= 0, Orci, 4 87 (0, Sy}= | v(dy)P de 0 is positive for every 6ys% 0. However, the functional J [y] also assumes negative values in any neighbourhood of zero; for a given e > 0 it suffices to take the function —a+e, 00. ge16 CH. I EXTREMA OF FUNCTIONALS Definition. A guadratic functional L, (k) specified on some normed space is said to be strong-positive if there exists a constant k >> 0 such that Lz) > RR IP for all h. é. sufficient condition for a minimum. For @ functional J (y] defined in a normed space to have a minimum at a stationary point y = yo, tt ws sufficient thet when y = yo the second variation be strong-positive, that is, that the following condition be fulfilled: 87 [yo, dy] S * | Sy I? where k = constant, k > 0. 4°. Let us seek the extremum of the functional Sts Yar oes Yad a si =| PE My Ye Yas Ue Uh unde mo dependent on nm functions y (2). yz (u), -- +1 Yo (2) with the boundary conditions Yr (Zo) = Ynor Yn (@,) = Yar (= 1, 2,..., 2”) (5) Legendre’s strong condition consists in the requirement that the following inequalities hold true Fyy > Pray Pagug 0 Puy Fug 0, Fre Fags (6) F, 4BMy Fyus oe . Pug Faguz- ++ Pages at all points of the indicated extremal of the functional (4). Jacobi's strong condition consists in the requirement that the interval [x , z,] not contain a point conjugate to the point x. Legendre’s strong condition (6) in conjunction with Jacobi's strong condition ensure the existence of at least a weak minimum of the functional (4).8, SUFFICIENT CONDITIONS FOR THE EXTREMUM 447 Example 11. Test for an extremum the functional 1 Sty, al= | (y?+2) az, (2) 0 yO)=0, 2(0)=0, yon, sta? f ® Solution. Buler's equations for the functional (7) are y =O, 27 =O so that y (z= C+ Coz, } a(t) = Cy+ Cz Utilizing the conditions (8), we get Cy=0, Cyo=i, Cy=0, Cp =2 The desired extremal y@)=z, } 2(z) == 2x is a straight line passing through the origin. We have Pyy = 2, Pye =0, Poy =0, Poy =2 Legendre’s strong condition is fulfilled: Fyy Fy) |2 | ror -| 5 f=4>9 ho) Let us now check to see whether Jacobi’s strong condition is fulfilled or not. One of the definitions of a conjugate point is the following (see [7]). Suppose we have a family of extremals of the functional (4) issuing from an initial point (29, Yo, . « - +» Yno) in close-lying but linearly independent directions. The point z* € [z, 2) is said to be conjugate to the point o if there exists a sequence of extremals issuing from an nitial point and arbitrarily close to the given extremal ch that each of these extremals intersects the given extre- (%) Fyy =2>0,448 CH. IL EXTREMA OF FUNCTIONALS mal, and the abscissas of the intersection points converge to the point 2*, In the given example, the extremals are the straight lines (9). All the extremals issuing from the point (0, 0, 0) intersect the extremal (9) only at that point. Hence, the interval [0, 1] of variation of x does not contain a point conjugate to the point x» = 0. Thus, both Legendre’s strong condition and Jacobi’s strong condition are fulfilled, so that the extremal (9) gives the functional (7) a weak mini- Test the following functionals for extrema. 164. J [y@), 2(@))= i Vt yt Fe de; 1(0)=0, y(4)=2, 2(0)=0, 2(1)=4. 165. J [y (2), so] (y2+ 2° = 4x) der; o y(O)=0, y(d)==t, 2(0)=0, 2 (1) =O, 9, CONDITIONAL EXTREMUM 4°. Isoperimetrie problem. Let there be given two fune- tions F (x, y, y’) and G (gz, y, y’). From among all curves y = y (x) € C; [zo, %] along which the functional Kui= |e, yy) de assumes a given value 7, determine the one for which the functional eh Jul=) Fey vee * assumes an extremal value. With respect to the functions F and @ we assume that they have continuous first and second partial derivatives for xz <2
You might also like
PG Math entrance papers kashmir university 2010-2020 (KU)
No ratings yet
PG Math entrance papers kashmir university 2010-2020 (KU)
125 pages
Answers To Problems For A Course in Real Analysis by Hugo Junghenn
No ratings yet
Answers To Problems For A Course in Real Analysis by Hugo Junghenn
10 pages
[Translations of Mathematical Monographs 26] Gennadiĭ Mikhaĭlovich Goluzin - Geometric Theory of Functions of a Complex Variable (Translations of Mathematical Monographs, Vol. 26) (1969, AMS Bookstore)
100% (1)
[Translations of Mathematical Monographs 26] Gennadiĭ Mikhaĭlovich Goluzin - Geometric Theory of Functions of a Complex Variable (Translations of Mathematical Monographs, Vol. 26) (1969, AMS Bookstore)
684 pages
Module IV (QB)
No ratings yet
Module IV (QB)
10 pages
Pontryagin's Maximum Principle
No ratings yet
Pontryagin's Maximum Principle
21 pages
L'HAna Des Inf Pet An Ann Tra Wit Sou Mat by Joh Ber
100% (1)
L'HAna Des Inf Pet An Ann Tra Wit Sou Mat by Joh Ber
357 pages
DPOCexam2008midterm Solution
No ratings yet
DPOCexam2008midterm Solution
12 pages
Functional Ito Calculus and Stochastic Integral Representation of Martingales
No ratings yet
Functional Ito Calculus and Stochastic Integral Representation of Martingales
33 pages
The Jordan-H Older Theorem
No ratings yet
The Jordan-H Older Theorem
6 pages
M.L. Krasnov, G. Yankovsky-Problems and Exercises in The Calculus of Variations-Central Books LTD (1975) PDF
No ratings yet
M.L. Krasnov, G. Yankovsky-Problems and Exercises in The Calculus of Variations-Central Books LTD (1975) PDF
112 pages
Fcalc03 PPT 04
No ratings yet
Fcalc03 PPT 04
100 pages
A "More Topological" Proof of The Tietze-Urysohn Theorem - Brian Scott
No ratings yet
A "More Topological" Proof of The Tietze-Urysohn Theorem - Brian Scott
3 pages
MATH4030 Differential Geometry, 2017-18 Solutions To Midterm
No ratings yet
MATH4030 Differential Geometry, 2017-18 Solutions To Midterm
5 pages
Penggunaan Turunan
No ratings yet
Penggunaan Turunan
34 pages
Hilbert-Spaces MATH231B Appendix1of1
No ratings yet
Hilbert-Spaces MATH231B Appendix1of1
28 pages
Functional Analysis
No ratings yet
Functional Analysis
93 pages
Complex Analysis
No ratings yet
Complex Analysis
40 pages
Calculus of Variations
No ratings yet
Calculus of Variations
167 pages
Function On Sevral Variables
No ratings yet
Function On Sevral Variables
8 pages
Banach Space
No ratings yet
Banach Space
13 pages
Extremum Values - Eklenmis
No ratings yet
Extremum Values - Eklenmis
30 pages
Slides 10-2023
No ratings yet
Slides 10-2023
32 pages
Calculus of Variations
No ratings yet
Calculus of Variations
15 pages
Introduction of Frechet and Gateaux Derivative
No ratings yet
Introduction of Frechet and Gateaux Derivative
6 pages
MMC2020 Solutions
100% (1)
MMC2020 Solutions
9 pages
Referensi 1 - Purcell
No ratings yet
Referensi 1 - Purcell
64 pages
Stability Theory of Ordinary Differential Equations 10.1007 - 978-1-4614-1806-1 - 106 PDF
No ratings yet
Stability Theory of Ordinary Differential Equations 10.1007 - 978-1-4614-1806-1 - 106 PDF
19 pages
Solutions Manual Corrections Linear Algebra Olver
No ratings yet
Solutions Manual Corrections Linear Algebra Olver
7 pages
The Variational Approach To Optimal Control
100% (1)
The Variational Approach To Optimal Control
48 pages
Functions of Several Variables
No ratings yet
Functions of Several Variables
4 pages
Functional Equations - Reid Barton - MOP 2006
No ratings yet
Functional Equations - Reid Barton - MOP 2006
2 pages
7-Calculus of Variations
No ratings yet
7-Calculus of Variations
23 pages
Calculus of Variations
No ratings yet
Calculus of Variations
30 pages
RG Overview of Complex Analysis and Applications
No ratings yet
RG Overview of Complex Analysis and Applications
8 pages
Calculus of Variations Solution Manual Russak
100% (2)
Calculus of Variations Solution Manual Russak
240 pages
Calculus of Variations & Optimal Control - Sasane
No ratings yet
Calculus of Variations & Optimal Control - Sasane
63 pages
Econ 271 Reading - Aut09Win10
No ratings yet
Econ 271 Reading - Aut09Win10
3 pages
The Hopf Bifurcation: Maple Solution
100% (2)
The Hopf Bifurcation: Maple Solution
5 pages
1 Continuous Functions On Compact Sets and Rela-Ted Extremum Problems
No ratings yet
1 Continuous Functions On Compact Sets and Rela-Ted Extremum Problems
10 pages
Functional Analysis by R. Vittal Rao: Lecture 6: Open and Closed Sets - June 8, 2012
No ratings yet
Functional Analysis by R. Vittal Rao: Lecture 6: Open and Closed Sets - June 8, 2012
3 pages
Midterm Exam Solutions
100% (1)
Midterm Exam Solutions
26 pages
Complete Quadratic Lyapunov Functionals Using Bessel LegendreInequality
No ratings yet
Complete Quadratic Lyapunov Functionals Using Bessel LegendreInequality
6 pages
Maxima -Minima 20-Dec-2024 12-04 pm
No ratings yet
Maxima -Minima 20-Dec-2024 12-04 pm
20 pages
Bifurcation
100% (1)
Bifurcation
5 pages
Phase Plane Analysis PDF
No ratings yet
Phase Plane Analysis PDF
27 pages
Linear Algebra and Applications: Numerical Linear Algebra: David S. Watkins
No ratings yet
Linear Algebra and Applications: Numerical Linear Algebra: David S. Watkins
107 pages
Calculus of Variations
100% (2)
Calculus of Variations
22 pages
01 Directional Derivatives and Gradient
No ratings yet
01 Directional Derivatives and Gradient
114 pages
Cov Main
No ratings yet
Cov Main
33 pages
The Advance Calculus of One Variable-Meredith Corporation (1971)
No ratings yet
The Advance Calculus of One Variable-Meredith Corporation (1971)
112 pages
Caculus of Variations
No ratings yet
Caculus of Variations
4 pages
Erwin Kreyszig: Books
50% (4)
Erwin Kreyszig: Books
2 pages
Metric Spaces PDF
No ratings yet
Metric Spaces PDF
33 pages
Finite Difference Spectral Approximations For The Time-Fractional Diffusion Equation
No ratings yet
Finite Difference Spectral Approximations For The Time-Fractional Diffusion Equation
20 pages
Lecture-Notes-2021 PG 27, 40
No ratings yet
Lecture-Notes-2021 PG 27, 40
62 pages
Duhamel's Principle
No ratings yet
Duhamel's Principle
14 pages
Rudin
No ratings yet
Rudin
26 pages
Solution Tu's Manifolds Exer1 - 6
No ratings yet
Solution Tu's Manifolds Exer1 - 6
1 page
Lipschitz Functions: Lorianne Ricco February 4, 2004
No ratings yet
Lipschitz Functions: Lorianne Ricco February 4, 2004
3 pages