Professional Documents
Culture Documents
0, then on physical grounds one would expect the solution & for all time from 1= 0 tor and for all x from x= 0 tox = L to be known. To show this mathematically, consider the following: In the desired domain, 4 = x - ct runs from A = L to A = -°, while p = x + ct runs from p = 0 to p = +e, From discussion (B), F(A) and G(u) are known. in the domain 0 < p/A < L. Hence, by hypothesis, (© F(-ct) + G(ct) is known for > 0. (1) F(L-ct) + G(L+e0) is known for t> 0. From (6), since G(j1) is known for argument p1 from 0 to L, F(-ct) is known from argument ct = 0 to ct = L, or F(A) from 4 = 0 to A= -L. From (7), since F(L-ct) is known from L-ct = 0 to L-ct = L (or from ct = L to ct = 0), then G(L+ct) = G(2L+ct-L) is known from argumental values of Lo 2L, i, G(u) from pt = L to p= 2L is known. Mapping procedure (Fig. 8-8) From (B), ® in region A is known. Knowing Gu) from p = 0 to L, FQ) from A = 0 to ~L is known. Knowing F() from 2 = 0 to L, G(u) from pi = L to p= 2L is known. Thus, & is known in region B. Knowing G(u) from p = L to 2L, F(A) from 2 = ~L to ~2L is known. Knowing F() from 4 = 0 to -L, G(u) from 1 = 2L to p = 3L is known. Thus, @ is known in region C. iH Continuing in the forward direction, regions shown in Fig. 8-8 will be 3L mapped out. Proceeding in the one backward direction will map out the aL remaining regions. forward | rf C|BIA aChapter 9 Heat Conduction 9.1 Fundamental Equations If x = thermal conductivity, E = energy content per unit volume, 5 = energy supplied per unit volume per unit time by some external agency, and c = specific heat per unit volume, then Q = -«VT, where Q is the heat flux, and E=cT. From the first law of thermodynamics, OE 7 TVQtS, where -V-Q represents the heat influx. Substituting for E and Q, we have: cl avres. ot Defining s = 7 and A=, we get or = V2 a AWT +s. a) Equation (1) is an inhomogeneous partial differential equation of the para- bolic type (see Section 8.2.1), and is of first-order in time and second-order in space. Theorem: If s is a given function of space and time, and T(r,t-0) is given atr=0, and if for > O either T(r,1) or (Vn at the boundary is given, then T(r,1) is uniquely determined for all > 0.Methods in Mathematical Physics It 345 Proof: Define A(r,t) = T(r,t) - To(r,t), where T;(r,1) and T2(r,1) both satisfy the same conditions above. From (1), oD AW?T(r,) = 5 and Shieh ~AV°72(r,2) = s. Subtracting, we tae AVF = 0, and after sauiiplying by F, we get 0! F -agvg = a(visva- 9). Integrating over all space, (2 dS - 2, ala at a [rv Fas af ‘dt. By hypothesis, either F = 0 or VFn = 0 on the boundary. Integrating over time: Linl - - a fafa i Jud pFa - fl2 Since F= Oat t=0, we ate left win |Z. ay negative quantity. Since - the integral on the left side is a positive quantity, this result is clearly impos- sible unless ¥ = 0. 9.2 Infinite Medium (A) HOMOGENEOUS SOLUTION (s = 0) Consider the homogeneous equation T 1m T _ yep-o, qd at AVT = 0. Performing a Fourier analysis, oie jer Q) Te.) ape lr kne"ak.346 Heat Conduction Substituting into (1): ani ars rer ecs}e a Multiplying by e“™’* for k’ an arbitrary vector in k space and integrating over all space: eae [ak (EMD, er ale = see ar 4 jew a MEF WA}B' -k)=0. qd uD, MRK) =0 = see = -2RFK)). The last step follows since k’ is arbitrary. Expressing F(kf) = fe *™, we have from (2): oe or Te) = Goa lWerek = Teo) ee a Thus, @ MW- = gardener, so that knowledge of 7(r,0) yields knowledge of f(k) and therefore T(r,!) for all time. Green's function (Go) If T(r,0) = 8%(r - ro) at = 0, then for > 0, T(r,f) = Gol - o,f). From (3) above, ek SL Jee- rete - mF I) - G oeMethods in Mathematical Physics II 347 Substituting into (2): Ten) = aor omen 1 je (xx) EM, iho ja-2)-BM, * Gay a3 i eo a Jeo Evaluating +2 2 i 7 fret ol ale) 5 = Gy)Mersan, and substituting, we have - 3, ad Te.) on (Wye ar. Thus, el Gale-robt) = mau? ‘ar in 3 dimensions. (4) : Go ae eee om (nan? . Properties of Go(|t-rol,t) for t>0 1, Ge = av?Ga , ic. G satisfies the homogeneous equation (1). Proof: For t > 0, T(r,t) = Go. Thus, since T(r) satisfies the homogeneous equation, so does Go. 2. Seate-rapnar =1. Proof: Joate-ranar - sll Poet ee er, sas dxdyde ag? [nay] (4x1 ma348 Heat Conduction 3. Ast + 0*, Go > 8°(r - ro). 4. If T(e,0) is known at = 0, then for #20, T(r) = [Go(r-r,)TW’ Oar’. Proof: @ 5 7 av} “Si 7 av? ete v.07 0dr =0 from property 1 above. Thus, T(t) satisfies the homogeneous equation. (b) Ast +0, T(r) + T(r,0) since Go(r-r’,t) + 8°(r-1’), so that the boundary conditions at t= 0 are satisfied. (c) Asr + ©, T(r,!) + O since Go(r-r’,t) — 0, so that the boundary conditions as r —+ © are satisfied. (B) INHOMOGENEOUS SOLUTION (s # 0) The procedure, in general, is to find a particular solution of the inhomo- geneous equation and the general solution of the homogeneous equation, and then combine to form the most general solution. Particular solution ‘Theorem: If one can find an inhomogeneous Green's function Go(t-To,t-to) such that (6) 5 7 av] Gaeta) = B(r-1.)8(-f), then a particular solution can always be written as: Ted ~ [Gar ese teed Proof: Define Go(r-ro,f-to) = Go(t-o,f-to) for t- fo > 0 and = 0 for t= t <0, where Go is the homogeneous Green's function. Making the transformation f' = t~ fo: Go = 0 for t’ < 0, so that the left- and right-hand sides of (5) are zero (the right-hand side by virtue of the 5-function property). Go = Go for f > 0, so that the left-hand side of (5) vanishes, while the right- hand side again vanishes by virtue of the 6-function property.Methods in Mathematical Physics Il 349 We now seek the behavior as’ 0*. Considering equation (5), from property 3 above, Go = Go > 8(r-0), while the right side ia another delta function that blows up as 5°(r-r0)8(7’ + 0+). Question: Do both sides have the same singularity? Consider, therefore, the following: far [2 -av" aera df B-rB) = Bene). From propetty 3 above, lim Go(-rof’) = 8°(r-r0), so that the left side of the equation becomes ne. q fe we GoAe—Togt’) - afer VWGAr-Tot’) = 1+ I. Term = |Gar-tof)| = Gule-roste) — B%e-r.)as e+ 0. Term II = -Ae(V"Go),-9. + 0 as + 0. Hence, both sides approach 5°(r—ro) as 1’ + 0+, and the proof is complete. ‘We have shown (5) to be satisfied for 1 < 0, for > 0 and for ’ + 0*. Note the dependence of the inhomogeneous Green’s function (Go) upon the homogeneous Green’s function (Go). This dependence hinges on the fact that there is a first-degree derivative in time. This will be seen again when the wave equation is analyzed. Given an initial temperature distribution T(r,0), one can consider this as being built up of many temperature pips, i.e. many 8°(r-r). The resultant T(r,1) for all > 0 can then be found by summing up the diffusion of these pips throughout the infinite medium, i.e. © — Tet) = JGor-r' NT’, O)ar’. If, instead, one is given an external supply of energy s(r,t) at > 0, then the diffusion of heat is expressed by QD Te) =f[G.0-r se Naerar, i.e. the diffusion takes place in a manner similar to that of a temperature distribution.350 Heat Conduction General solution Theorem: If s(r,t) for t > 0 and T(r,t=0) are given, then for > 0, ® Ten -feur-rare ody + Jarl garrs-roe ed where the first term on the right is the general solution of the homogeneous equation, and the second term is a particular solution of the inhomogeneous equation. Proof: (@® T(c,1) satisfies the inhomogeneous equation oo (3; - av }ren = ste. (ii) At r=, Go(r-r’,t) + 0, while Go(r-r’,1) = Go(r-r’,.) + 0 for t>Oand Go =O fort<0. Hence, T(r) > Oat r=%, (iii) If T(r,0) = 0 and s(r,t) = 0, then 7(r,1) = 0 everywhere and for all r. (iv) Ast + 0,708) + Tte.0)- Hence, T(¢,1) satisfies all the necessary boundary conditions. Consider the expression JafGeee enue nar. Since Go(r-r’-1’) = 0 for r > t, we can rewrite equation (8) as © Tep=JeKe-r ore ode + farfgarr.-es(e nde ‘We consider next the problem of the semi-infinite medium. Question: Why should one get, or even expect to get a new Green's function? ‘Answer: Reflections may occur at the boundaries, and the Green’s function should incorporate these boundary properties in the description.Methods in Mathematical Physics II 351 9.3 Semi-Infinite Medium Boundary conditions (Fig. 9-1): (i) Forall time #2 0, 7-0, = 0 (i.e. in contact with x20 aheat reservoir at temperature Ty=Oatx=0),and ? Tee==,1) = 0. (i) Forx 0, Ste) AVT (2) = sGct) given. (ii) 76,0) given. Fig. 9-1 Consider the following related diffusion problem for the case of an infinite medium (essentially an image problem). Assume that the above temperature distribution and source s(%,f) are reflected in the —x plane, with negative signs for s(x,t) and T(e,t). Then if s’(x,f) and T(t) refer to an infinite medium, T(x,t=0) = T(x,t=0) for x > O and = - T(x,t0) for x < 0. S'(6) = sG6,) for x > O and = - sx) for x <0. From equation (9), Section 9.2, for > 0: QTE) = fo-¥.NTe Od + farfgae-x ers sae. Since Go(x-x’,f’) and s'(2’,’) are even functions in (x - x’) while T(2’,0) is an odd function in x’: T(x) = -T(-x) = =T(0,) = 0. Thus, the solution of the semi-infinite medium problem is the solution of the infinite medium problem for the +x domain. In other words, for a semi- infinite medium, for x > 0: Q) Te) Flex ~ Golxtx’ D]T (t-O)dx + farf[ouc-xs-r) ~ Golaxta’ sr) fd.352 Heat Conduction Semi-infinite Green’s function Let | Tet] = 8(xo-x’) for x > 0 and x’ > 0. The Green’s function G(x %o,t) is defined from T(x) = Geto) = Golx-Xo) - Golx+%o,)- From equation (2): Ten) =[oo-x ore node + farlowxnrsw nue. [Note here that G(«,x’), a function in 3 variables, x, x’ and 1, goes over into a function of 2 variables, (x-2’) and £.] Note that equation (3) has the same form as equation (1) for the case of an infinite medium, except for the domain of integration and the form of the Green’s function. Define the semi-infinite Green’s function G(x’) = G(ex’,f) for 1> O and =O forr<0. Properties of G(x,x’,t) @ ( 7 avila. = 0. This follows since T(,t) satisfies the homo- geneous equation. Gp (5 - av geo. = 8(x-x’)8(1). For 1 < 0, both sides vanish, while for 1> 0, G~= Gand the left side vanishes from (i) above, while the right side vanishes since 8(t #0) = 0. ii) Ast > or, G&D) + B(x-x). (iv) TQ) = Jove NT t-0)dx! + facfosc-vs-ny (xt)de. = Osince Gl-x'1) - Gol’ Thus, the Green’s function G(x-x’,t) obeys the same boundary conditions as does the temperature. Importance of the Green's function: The Green's function is completely determined, in principle, by the particular type of boundary in the problem, independent of initial temperature distribution and independent of heat source. Once the form of the Green's function is known, one can plug any given s(x,t) and 7(r,0) into the problem in order to solve for T(x,f) at some later time t > 0.Methods in Mathematical Physics II 353 9.3.1 Age of the Earth (Treatment due to Lord Kelvin) Assumptions (i) Attime t= 0 the earth starts to condense. (ii) Non-curvature of the earth, i.e. a semi-infinite medium. (iii) T= 0 at the surface of the earth today. (A fair assumption, since it is small compared with T(¢ = 0) = 120°C.) (iv) Uniform temperature of 1200°C as the melting point of rock at f= 0. Given (i) T= 0) = 1200°C. (Actually somewhat greater due to pressure.) (ii) A= We = 0.005 cm’fec. 6 (iii) s= 5 = ae = 10" Keo, (iv) At present time, \F = 32% = 327/10%em. [S From equation (3), Section 9.3, Q) Tee) = Joc moyae + faloo-vpnwe nde. Assuming T(’,0) = To = 1200°C and constant, and s(’,’) = so = constant, TOs) = 1: fow-x.nae + sJarlee-r sna. Differentiating, and evaluating at x = 0: if |SGG-x'.1), fel Pa. @ i LC 7 to aeeres |354 Heat Conduction Using the properties of Ge’): |? |. BGoK-x's)| oo S ax e = | 2Gole-w') AGolety’, 9 Gale) ox ar’ ee Thus, OGD gy POLED gy 2, f Gee af Se ae = apa le he eee (many? Substituting into (2), Te eS x L (nar)? f or Ta_, 250ft | ae) i One now compares the value of |22 | as calenlated from (3) using known no constants with the known current value of |S 1 = 10° years = 3 x 10'Ssec: . Expecting a value of 0 = (w5-10°-3-10")” = (3750-10) = 3-7:10%, so that calculations from (3) yield er) 1200 2610n 10 ee | 21-10° 21-10" 3000’ while the current known value is ae se jo km 10%em 300° esMethods in Mathematical Physics II 355 When Lord Kelvin first computed t, he neglected so which we see contributes the greater portion to er L in the ratio SO - 5t0 1. 9.3.2 Temperature Variation of the Earth’s Surface In contrast with the previous problem for which the gradient O7/Ax at the surface of the earth extends through a distance of kilometers, we are here concerned with a gradient that extends through centimeters, or metets, i.e. we are seeking a daily or yearly variation in temperature, with boundary conditions that are functions of time. Boundary conditions @ T=A sinot at x = 0. (If not a sine function itself, then at least a Fourier series in sines.) (ii) T(&,1-0) given at r= 0. (ii) 5@,1) known for 12 0. Whenever the boundary temperature is a function of time, as in boundary condition (i), itis only necessary to find a particular solution Tp that satis- fies the homogeneous equation 0T,/6t = AV"Tp. Since Tp is given at x = 0, if we define T’= T'- Tp, then conditions for T° are: (iv) T’=Oatx=0. (v) T’ will satisfy OT/ot= AV°T" + S. (vi) T’ is known at t= 0. In this manner we have reduced an inhomogeneous boundary condition to its corresponding homogeneous boundary condition for a semi-infinite medium. We seek, then, a solution of the homogeneous equation: OT, or Let T(x=0,1) = Ae™, the im: our boundary conditions at x () = Aav'7,. ry part of which we seek in order to satisfy356 Heat Conduction Assuming a particular solution 7)(x,t) = F(x)e'®" and substituting into (1), we have iaF (x)e™ = 2V7F (xe, or oc @ SP. Bre = FOFw. Choosing @ > 0, solutions to (2) are given by @) FG) = Bexp[(in/4)(a/2)x] + C exp[-(iny4y(a/2)"x], where B and C have yet to be determined. Since e™ = (1+i//7, and since at x=, exp|(in/4)(o/)'7x] = 2, B must = 0, since F(@) = 0. Also, F(Q)=C=A, so that C= A. We finally have for Ty: Txt) = Aexp[-(iry4X(o/ 2)x] exp [iar] = Aexp[-(@/22)(1+i)x + ior] = Aexp[-(@/2A)7x] exp [i€eor-(a/22)7x] s Since we seek the imaginary part as the physical solution, 4) Txt) = Aexp[-(@/22)"7x] sin [or-(@/22)x]. [Note that 7}(x,t) has a part that decays in the +x direction, and includes phase changes that accompany varying changes in x. Note also that the equation 87)(z,t) /ot = AV77)(x,2) is a real equation when is real. If Tp is complex, one can always equate real and imaginary parts. The Schrédinger equation, however, is not real since 4 is complex.] Yearly variation For soil, 0.002cm? | _2n___2n Se Te 910 acc (ay 20)” = [273-107-410]? cm = (12-104) = 710% em,Methods in Mathematical Physics Il 357 Atx=0, Tp = Asinayt (Fig. 9-2a). T, Atx= 1 meter, (oy 2A)x = 0.7 = 1/4. ef si. Ty(2,) = (A/2)sin(wyt-1/4) (Fig. 9-2b). 0 Atx= 4 meter, (@20)?x=n. Fi exp [oy2a)*x] = 12". T,Ax,1) = (4/2)sin(wyt-n). Thus, at a depth of 4 meters, the earth is coldest in summer and hottest in winter (Fig. 9-2c). Daily variation Yearly and daily phase changes will be equal when Vaoaxa = Vay. or When Xa ve, 1 % (@a/o,)" = 59° Thus, x= 1 meter (yearly) —- x= Sem (daily) to get 1/4 phase change, and x=4 meters (yearly) = x= 20cm (daily) to get m phase change.Chapter 10 The Eigenvalue Problem 10.1 Eigenvalues and Eigenfunctions In Section 8.1 we considered the particular example of the wave equation type of differential equation, Qa We-=S =o. v For a one-dimensional vibrating string, © is the displacement and v” = T (tensionYp (density). For the case of 2 dimensions, we are dealing with a vibrating membrane. We seek the characteristic frequency of the string/ membrane. Assuming (r,t) = ‘P(r)e" and substituting into (1), we have mead Vrpyeitt 4 2 OF ior ae Setting A= w/v’, we get: 2) -VAM(r) = AM(r) Equation (2) is typical of an eigenvalue problem, where one regards -V? as the operator, ¥ as the eigenfunction, and 2 as the eigenvalue. (A) One-dimensional string For the one-dimensional vibrating string, equation (2) becomes FM) _ ae MO.Methods in Mathematical Physics II 359 Solutions are given by ‘P(x) = A sinkx + B coskx, where 4 =k? Since boundary conditions demand that = 0 at x = 0, we see that B= 0. In addition, since boundary conditions also demand that Y= 0 at x = L, it follows that A sinkL=0 =+ k= nn/L(n=0,1...%). Thus, due to the boundary conditions, 4 = k?= (n°x”y/L’ has certain discrete values, namely the eigenvalues. Substituting for k, we have M(x) = A sinnna/L. 52 ATX 1 2 Note that [¥2(a)de = fei Oe = “fa ~ cos2nmx/L)dx = ° 7 : Thus, since we want to normalize the Yn, we choose A = (2/L)”, yielding for Wx(2): QB) B(x) = (2/L)"sinnnx/L, where An = k7= (n° YL? (n= 1,2...2), satisfying —V¥,(x) = An'Pa(x). Characteristics of the eigenvalue problem 1. ~ d¥dx" is a good Hermitian operator for all functions with boundary conditions ¥,, = 0 at x = 0 and x = L, i.e. Ta fe 4 | fu [- Joa for (--& oa, where ® and are in the class of functions admitted by the boundary conditions of the problem. Proof: [ott fe oar - oof oun360 The Eigenvalue Problem 4 2, Pensa dE = Ban. 3. The ,(x) (n = 1,2...2°) form a complete and ortho- normal set for any function ‘P(x) satisfying W(x) = Oatx=Oandx=L. (B) Two-dimensional circular membrane Fig. 10-1 ‘The eigenvalue equation [from (2)] is given by -V"H(p,9) = A¥(0,9), where A = @7/v under the condition that '¥ = 0 at p = po (radius of the membrane) (Fig. 10-1). To determine the eigenvalues and the eigen- functions, we can expand 'Y(p,p) as ¥(p,9) = )fn(p)e'"®, since on a circle the e”®, or equivalently, the Fourier series, form a complete set, and any function g(p) can be expanded in a Fourier series. Substituting into the eigenvalue equation: ELS velo Multiplying by e“"® and integrating over the circle: probe (aPs bs) dp’ p dp Using the orthonormality properties of the e"” yields to), 1 do), am 4) -|fe) - 0. O wow fe) ae d d & Set Vip =2,so that 5 >= WRG = apt hag md @ becomesMethods in Mathematical Physics I! 361 Dividing through by 2 we artive at Bessel’s equation with integral m values: : Te, 1A (mga 0. The m are necessarily integers since they arose from expanding ¥(p,@) into a Fourier series. In Section 7.5.4 we discussed the case for m being integer- valued, for which the second solution of Bessel’s equation is given by the Hankel functions H}, or Hx . Thus, the most general solution of the above vibrating membrane differential equation (5) is Ju(2) = AJn(2) + BHn(2)- Since we want fin(z) to be finite at p = 0 (i.e. at z = 0), while on the other hand Hn(z) possesses a logarithmic singularity at z = 0, we must have B= 0. Thus, (© — fu(Z) = AJn(2). Since ¥(posp) = O at p = po for all g, it follows that Jn(/Xpo) = 0. Itis this boundary condition which gives 2 its discreteness property. We now seek the roots of Jn(z) = 0. For m = 0 (Fig. 10-2), roots are Zo1, Z02, 203, etc. For m # 0 (Fig. 10-3), roots are Zmn, i.e. Zm1, Zm2, Zm3, etc . The origin is not included as a root since at the origin, z = 0, or VAp = 0. If p # 0, then 2. must = 0, which, however, means no vibration. Hence, roots are Znn, where n = 1,2... Jof2) InZ) Fig. 10-2 Fig. 10-3362 The Eigenvalue Problem Corresponding to roots Zmn, there are the eigenvalues an = Can / 0)", and the corresponding eigenfunctions (8) Yna(059) = Aln( Tenn ple™*. Degeneracy For the case of +m and -m (but ¥ 0), if we recall that J+m(2) differs from J-m(z) only by the factor (~1)", then it follows that both J+m(z) and J-m(z) will have identical roots, giving rise to identical eigenvalues. The Yin(P,?)s however, do not depend only on Jn(z), so that we have different eigen- functions that give rise to the same eigenvalue, i.e. degeneracy. Tabulating the roots in increasing order, we have: Zo = 2.405; 2 = 3.832; z21 = 5.135; zon = 5.520; 231 = 6.379; ete. From A = @7/v”, we note that Amn has a frequency, ®mn = VRmn 0 = Zmn/Pos associated with it. Thus, the eigenvalues of the circular membrane are given by Zmn0 (ee We now seek the vibration patterns for the above characteristic frequencies. m=0, n=1 This is a non-degenerate case, with only one eigenfunction, Wo1 = AJo(201P/Po). Note that for p < po (i.e. within the boundary of the problem), zo1p/po < zo1, and Jo(z) has no roots for z < zo (Fig. 10.2). Hence there is anode only at the boundary, and the amplitude of the oscillation is the amplitude of Jo(z). (9) mn = m=1, n=1 Here we have degeneracy (m = #1). Weir = AUi(zup/po)e"®. C21) + 2-1) = Adi(ZP/po)cosp. C211) - G2DP-11) = Adi(Zup/po)sing. Since the left-hand sides of both equations are separately eigensolutions, Ii(211/po)cosep and J(z119/po)sing are also eigensolutions.Methods in Mathematical Physics I 363 In fact, any linear combination of these degenerate solutions will also be eigenfunctions. Hence, for m= #1, n= 1: Wxis(P.p) ~ Ji(2up/p.)sin(p-8), where 0 is an arbitrary constant. oe Question: What will the vibration patterns look like, and vy; where else but at the boundary will ¥ = 0? From the form of Y above, we see that Y = 0 at p = 0. ~ Since J(z) is zero at the origin (Fig. 10.3) and at the root zi (Le. when p = po), nodes occur at p = po for all @, and along the nodal line @ = @ (Fig. 10-4). Fig. 10-4 m= £2, n=1 In this case we have degeneracy again. t2ip ae, 2 nodal lines P22,1 = Al2(221p/po)e®. Writing as before: TY (2)M.21 + V-2,] = Ads(2219/p.)eos20. eo C2DP¥ a1 ~ V2] = Adse1P/po)sin29. xs Since J2(z21p/po)cos2p and J2(z21p/po)sin2 are eigensolutions, a linear combination of the two will Fig. 10-5 also be eigensolutions. Thus, 2.405p,/5520 W22,1(P,9) & Ja(Z21P/po)sin(2p-8), so that Ps2,1(p,p) = 0 at p= 0/2, or p= 0/2+ 7/2, which results in the two nodal lines shown in Fig. 10-5. m=0, n=2 As was the case for m = 0, there is Fig. 10-6 degeneracy. Wor = AJo(Z02P/Po) = Jo(Zo2P/Po) = 0 at z= Zo1, 02, etc. (Fig. 10-2), ie. Woz = 0 when 202P/Po = 202. = _P = Po, and ‘oz = 0 when 202P/po = 201, oF when p = zo1Po/z02 = 2.405po/5.520. The nodal lines for all ~ ate shown in Fig. 10-6. Nodal lines for m # 0 appear as a result of degeneracy. The location of the nodal line in these cases is extremely arbitrary, due to cylindrical symmetry.364 The Eigenvalue Problem 10.2 Harmonic Oscillator/Free Particle in a Sphere (A) HARMONIC OSCILLATOR In the — oscillator ae (Section 7.3.1), we saw that by setting p>-ih® andé -+ th Z in the Schrddinger equation, E= Ee Viwe ee (-£ Eveyly Substituting V = kx7/2, multiplying through by 2/fia = 2(m/k)""/M , and let- ting z = (Vmk/fn)” x, we arrive at ay 2EY ae aoe ye Setting fiz) = '¥(z)e"” results in the Hermite equation w ate. 2 19, maf) =0, where 26/ffia = (22+ 1) = E=(4+1/2)he. Solutions of the Hermite equation were given as f(z) = H(z), so that Y,(2) = e*H,(z), and unnor- alized solutions of the harmonic oscillator are given by Q — ¥(ax) = &**HA(0x). Since JHtsz\Hale)e* de = By", we have as normalized solutions: a @)— Wa(ox) = (al eH Ko2), 1 vihere 2 = ox = | :Methods in Mathematical Physics It 365 The eigenvalue equation is given by @ -£e- vly = Ebay 2m where () Ex = (n+ 12)(/m)”. * - Fig. 10-7 ©) Sala) Yom x)A(0X) = Ban. Oo 2 x Fig. 10-8 Fig. 10-9 Forn=0: Woolox) = Ae”, For n= 1: Yoi(ax) = Bre®*”, For n= 2: Wax) = Ce**(Dx - F). Note that the lowest eigenvalue, Eo, corresponds to the case of no node (Fig. 10-7), while the next highest eigenvalue, E1, corresponds to the case of 1 node (Fig. 10-8), and the next highest eigenvalue, Ez, corresponds to the case of 2 nodes (Fig. 10-9). These are to be compared with the case of the circular membrane. (B) FREE PARTICLE IN A SPHERE a V=0 fora free particle, Schrédinger’s equation becomes E vine) = EW(r), where Y(t) = Oat r= a (radius of sphere) and is ae, irweset e~ HF, we have for Schrédinger’s equation: -VY(r) = V(r) = AY(r).366 The Eigenvalue Problem Expanding ‘Y(r,0,g) in terms of the complete set of spherical harmonics: ae ¥r0.9) - YVR n(0.9)- be Be waroe) + Puce = FSF (Piao) : 2¥inf0.9) , 1 Halts] Eola in 20°" 08 Paint | + Len RO) =0. Substituting OF in( 8, FYin( 0, eS sing Mn) 2. ® 10+ 1¥in(0,0): Tg ee ee Ropaow = 0. Employing the orthonormality relations for ¥im(0,9), we have 1 dy dRir) Le : o Fare A fe )-0. Setting Rin) -A2, Substituting in (7): ap, 1 _ [0 oto - a 4 HD, 1HO, py) C2 BIO - 0.Methods in Mathematical Physics I! 367 If we now let z= kr, we have dQ), 1 fe) [ a : ® ae * za at z fe) 2 which we recognize as Bessel's equation with A= + 1/2. Solutions are given by Sk) = const. Jnyal2) + Rin) = const, ue) Roots of Ji+12(kr) = 0 are given by kin, where n= 1,2... = Ain=kin- Finally then, we have for the eigenfunctions: ©) Yaslr89) = const Ae” v.50). 10.3 The Variational Principle In Section 5.3.2 it was pointed out that a Hermitian operator H satisfies Su nea - [fe mea) where dr includes the range of the particular problem. igenvalue characteristics For HY = WP: (A) All the eigenvalues, , are real. Proof: Sv" H¥de = Af¥'¥dr, from which _ Serva ye - Pena Seva | fever cr (fi war) Seva (B) If (@): HY = AW and (b): HY’ = 2°V', then. fP""Wdr = 0, which says that eigenfunctions ‘¥’ and are orthogonal.368 The Eigenvalue Problem @ Fora #2’: Proof: Multiplying (a) by ‘Y”" and integrating: @ — fernvar = aferwar. Multiplying (b) by ‘Y” and integrating: ® frnva-rfewa - [few var) Substituting (fvnwar] = fre" var into (4): 6) Setmedr = xfiv'war. Subtracting (5) from (2), we have (A - ¥)f'#""Wadr = 0, so that f'¥"War = 0 ifhen, (ii) For = 2’: In this case we have degeneracy, since two eigenfunctions and" give rise to the same eigenvalues 4 = 2’, and Y and Y’ are not necessarily orthogonal. However, one can use the Schmidt orthogonalization method, discussed in Section 6.4.1, to make them orthogonal. Since the character- istics above hold for H real or complex, if H is a real Hermitian operator, and if HY = AXP, then HYP" = 2X8", Since this equation is satisfied for the real and imaginary parts, separately, of any complex eigenfunction, we will always choose ¥ to be real. Functionals F(Y) Definition: For every given function ¥, there is a unique value for the func- tional F(Y). fervar An example of a functional is the expression <‘P|H|¥> = f ont: edt Functional Variation 5F(¥) Definition: 5F(Y) = F(Y + VY) — F(#), where terms of order o~r) are neglected. Example (i): If FOP) = f¥ar, then BFW) = fovsevyar - far = [[2vow+owy er = 2fvordr.Methods in Mathematical Physics I 369 Example (ii): WFC) = - fev*war. Fe) = — Sverre nde + fovwyrac -- fervyynas + foveyar. ‘We saw in Section 10.1 that V” will be a good Hermitian operator if we set W =O at the surface boundary. Thus, FW) = ~ fevevar = Sivas where both expressions (a) and (b) represent the same functional. @ BFC) = - forvevar - fev'owd. Due to the Hermitian property of V”, fevover = favVWar, so that BFW) = -2 fowv'war. 2 feowvanar - 2 Jov-vvar 2 fowvey nds - 2 fow-vvar. (b) BFW) = 2fv¥-VEYdr As above, if 8 = 0 on the boundary, then 3F(#) = -2 fowv'War, in agreement with (a) above. 10.3.1 Distribution of the Eigenvalues Let HYY; = 4;, where H is the Hermitian operator for the particular Y; that satisfies the boundary conditions ; = 0 on the boundary. Let us arrange the A such that A1 < Az < A3 $24 S... ie. a spectrum of eigen- values. Define F(¥), the functional in question, as Yd fea where the domain of integration is the allowed domain of the problem (such as the length of a string, area of a membrane, etc.), and where ¥ satisfies the proper boundary conditions. (1) SPREE? =370 The Eigenvalue Problem Theorem (i): If HY; = AY; and if ¥ = V;, then
= ri.
Proof:
yr cee ee elie
"Sitar Svar
Theorem (ii): If 8<‘Y|H\> = 0, then HY = for all satisfying the boundary con-
ditions of (1) only.
[In Section 10.1 we discussed the vibrating circular membrane and showed
that the roots Zmn for m # 0, i.e. equivalent to case (2) above, are greater
than Zon for m = 0, ie. equivalent to case (1) above, and Amn & rn & Zan -]
‘Thus, the effect of constraints is to raise the minimum 41, and as a result,
Ory S$ Ady
Theorem (vi): Let HY; = 411, where A1 is the lowest eigenvalue. Then
the lowest value of You might also like