2.30   ODE No. 30

  1. Problem in Latex
  2. Mathematica input
  3. Maple input

xa1y(x)2xa+y(x)=0 Mathematica : cpu = 0.289385 (sec), leaf count = 197

{{y(x)xa((1)ac1xΓ(a+1)Ia1(2x)+(1)a+1ac1Γ(a+1)Ia(2x)+(1)ac1xΓ(a+1)Ia+1(2x)+xΓ(1a)Ia1(2x)+xΓ(1a)I1a(2x)aΓ(1a)Ia(2x))2((1)ac1Γ(a+1)Ia(2x)+Γ(1a)Ia(2x))}}

Maple : cpu = 0.097 (sec), leaf count = 54

{y(x)=xa+1(Ka+1(2x)_C1+Ia+1(2x))1x(Ka(2x)_C1+Ia(2x))1}

Hand solution

y+xa1y2xa=0y=xaxa1y2(1)=P(x)+Q(x)y+R(x)y2

This is Ricatti first order non-linear ODE. Using standard transformationy=uuR(x)=xa+1uu

Hence

y=(a+1)xauu+xa+1uuxa+1(u)2u2

Comparing to (1) gives

xaxa1y2=(a+1)xauu+xa+1uuxa+1(u)2u2xaxa1(xa+1uu)2=(a+1)xauu+xa+1uuxa+1(u)2u21xa1xax2a+2(u)2u2=(a+1)uu+xuux(u)2u21x(u)2u2=(a+1)uu+xuux(u)2u21=(a+1)uu+xuu(2)xu+(1+a)uu=0

In standard form u+1x(1+a)u1xu=0 or u+p(x)(1+a)u+q(x)u=0. We see that p(x) is not analytic at x=0 (the expansion point). So we can’t use power series solution, and will use Forbenius series. Power series, which is u=n=0cnxn is used when the expansion point is not singular point. (i.e. p(x) and q(x) are analytic there). Forbenius series u=xrn=0cnxn is used when there is a removable singular point (called also regular singular point), as in this case. Starting withu=xrn=0cnxn=n=0cnxn+r Hence u=n=0(n+r)cnxn+r1u=n=0(n+r)(n+r1)cnxn+r2

Substituting in (2) givesxn=0(n+r)(n+r1)cnxn+r2+(1+a)n=0(n+r)cnxn+r1n=0cnxn+r=0n=0(n+r)(n+r1)cnxn+r1+(1+a)n=0(n+r)cnxn+r1n=0cnxn+r=0

Dividing out xrn=0(n+r)(n+r1)cnxn1+(1+a)n=0(n+r)cnxn1n=0cnxn=0 Each term should have xn1 in it. So we adjust the last termn=0(n+r)(n+r1)cnxn1+(1+a)n=0(n+r)cnxn1n=1cn1xn1=0 Expanding the second termn=0(n+r)(n+r1)cnxn1+n=0(n+r)cnxn1+n=0a(n+r)cnxn1n=1cn1xn1=0 Hence for n=0(n+r)(n+r1)cnxn1+(n+r)cnxn1+a(n+r)cnxn1=0r(r1)c0+rc0+arc0=0

Since c00 thenr(r1)+r+ar=0 Hence r= a or r=0. Now for n1 (n+r)(n+r1)cnxn1+ (n+r)cnxn1+ a(n+r)cnxn1 cn1xn1=0(n+r)(n+r1)cn + (n+r)cn + a(n+r)cn  cn1 =0((n+r)(n+r1) + (n+r) + a(n+r))cn  =cn1cn  =cn1(n+r)(n+r1) + (n+r) + a(n+r)

For r=0, we obtain(3)cn  =cn1n(n1) + n + an For r=a(4)cn  =cn1(na)(na1) + (na) + a(na) There are two solutions. Looking at (3) for now, for n=1c1=c0 1 + a For n=2c2  =c14 +2a=c0 1 + a12(2 +a) For n=3c3=c23(2) + 3 +3a=c23(3 +a)=c0 1 + a12(2 +a)13(3 +a) And so on. Since the solution is assumed to be xrn=0cnxn and we are looking at case r=0 thenur=0(x)=n=1cnxn=c0+c1x+c2x2+=c0x0+c0 1 + ax+c0 1 + a12(2 +a)x2+c0 1 + a12(2 +a)13(3 +a)x3+(5)=c0(x0+1 1 + ax+1 (1 + a)12(2 +a)x2+1 (1 + a)12(2 +a)13(3 +a)x3+)

Since Γ(n)=(n1)! and a(1+a)(2+a)(n+a)=Γ(a+n+1)Γ(a) Then(1+a)(2+a)(n+a)=Γ(a+n+1)aΓ(a) And (5) can now be written as(6)yr=0(x)=c0n=11n!aΓ(a)Γ(a+n+1)xn But modified Bessel function of first kind is BesselI(a,z)=n=01n!1Γ(a+n+1)(z2)2n+a So if we let z=2x we obtainBesselI(a,2x)=n=01n!1Γ(a+n+1)(2x2)2n+a=n=01n!1Γ(a+n+1)(x)2n(x)a=n=01n!1Γ(a+n+1)xn(x)a

Hence(7)1xaBesselI(a,2x)=n=01n!1Γ(a+n+1)xn If we now compare (6) and (7), we see that if we set c0, which is arbitrary, to be c0=1aΓ(a), then we obtainur=0(x)=1aΓ(a)n=01n!aΓ(a)Γ(a+n+1)xn=n=01n!1Γ(a+n+1)xn

But this is (7). Hence we found the first solution, which is (8)ur=0(x)=1xaBesselI(a,2x)

The above was for  r=0. Now we find the second solution for r=a. From (4)

cn  =cn1(na)(na1) + (na) + a(na)

For n=1

c1  =c0a(1a)+ (1a) + a(1a)=c0 (1a) 

For n=2

c2  =c1(2a)(1a) + (2a) + a(2a)=c142a=c0 (1a) 12(2a)

For n=3

c3  =c2(3a)(2a) + (3a) + a(3a)=c23(3a)=c0 (1a) 12(2a)13(3a)

And so on. Since the solution is assumed to be xrn=0cnxn then

ur=a=xan=0cnxn=n=0cnxna=c0xan=01 n!(1(1a)1(2a)1(3a)1(na))xna

But as we found above, we obtain that (1a)(2a)(na)=Γ(a+n+1)aΓ(a), therefore

ur=a=c0n=01 n!aΓ(a)Γ(a+n+1)xna

Modified Bessel function of second kind is BesselK(a,z)=π21sin(aπ)(BesselI(a,z)BesselI(a,z)). The above should result in 1xaBesselK(a,2x) for z=2x by setting c0 to appropriate arbitrary value. I need to work out this final manipulation later. Hence we find ur=a(x)=1xaBesselK(a,2x). Therefore, the solution isu=C11xaBesselI(a,2x)+C21xaBesselK(a,2x) But ddx1xaBesselI(a,2x)=1x1+aBesselI(1+a,2x)ddx1xaBesselI(a,2x)=1x1+aBesselK(1+a,2x)

Henceu=C11x1+aBesselI(1+a,2x)C21x1+aBesselK(1+a,2x) And from y=xa+1uuy=x1+aC11x1+aBesselI(1+a,2x)C21x1+aBesselK(1+a,2x)C11xaBesselI(a,2x)+C21xaBesselK(a,2x) Let C=C2C1 hencey=x1+a1x1+aBesselI(1+a,2x)C1x1+aBesselK(1+a,2x)1xaBesselI(a,2x)+C1xaBesselK(a,2x) Ory=x1+ax12BesselI(1+a,2x)Cx12BesselK(1+a,2x)BesselI(a,2x)+CBesselK(a,2x)=x12+aBesselI(1+a,2x)Cx12+aBesselK(1+a,2x)BesselI(a,2x)+CBesselK(a,2x)

Verification

eq:=diff(y(x),x)+x^(-a-1)*y(x)^2-x^a = 0; 
num:=x^(1/2+a)*BesselI(1+a,2*sqrt(x))-_C1*x^(1/2+a)*BesselK(1+a,2*sqrt(x)); 
den:=BesselI(a,2*sqrt(x))+_C1*BesselK(a,2*sqrt(x)); 
my_sol:=num/den; 
odetest(y(x)=my_sol,eq); 
0