MATH529: Mathematical methods for the physical sciences IIMay 13, 2021

Final Examination

Solve the following problems (4 course points each). Present a brief motivation of your method of solution. Answers without explanation of solution procedure are not awarded credit.

  1. Use the Laplace transform F(s)={f(t)}=0e-stf(t)dt to solve the problem

    a22ux2=2ut2,x>0,t>0,
    u(0,t)=0,limxux(x,t)=0,t>0,
    u(x,0)=0,v(x,0)=ut(x,0)=-v0,x>0,v+.

    Solution. With notation ut=u/t, utt=2u/t2, use integration by parts to evaluate

    {u'(x,t)}=0e-stut(x,t)dt=[e-stu(x,t)]t=0t+s0e-stu(x,t)dt=-u(x,0)+sU(x,s)
    {u'(x,t)}=0e-stutt(x,t)dt=[e-stut(x,t)]t=0t+s0e-stut(x,t)dt=-ut(x,0)+s[sU(x,s)-u(x,0)].

    Apply to PDE to obtain

    a2d2U(x,s)dx2=s2U(x,s)-su(x,0)-ut(x,0).

    Apply initial conditions to obtain

    a2d2U(x,s)dx2=s2U(x,s)+v0.

    Solve the homogeneous ODE

    a2d2U(x,s)dx2=s2U(x,s),

    to obtain the general solution

    U(x,s)=c1e-sx/a+c2esx/a.

    A particular solution is

    U(x,s)=c1e-sx/a+c2esx/a-v0s2.

    Take the Laplace transform of the boundary conditions:

    limxdU(x,s)dx=0c2=0
    U(0,s)=0c1-v0s2=0c1=v0s2,

    leading to solution

    U(x,s)=v0s2(e-sx/a-1).

    Take the inverse Laplace transforms

    u(t,x)=-1{U(x,s)}=v0[-1{1s2e-sx/a}--1{1s2}].

    From

    {t}=0e-sttdt=-[1se-stt]t=0t+1s0e-stdt=1s2,

    deduce

    -1{1s2}=t.

    Recall that exponential factors in a Laplace transform arise from a delay that can be represented through the unit step function H(t)=1 for t>0, 0 otherwise, as

    {H(t-c)}=0e-stH(t-c)dt=ce-stdt=-1s[e-st]t=ct=e-css,(c>0).

    To obtain an s2 denominator, combine above with integration by parts

    {tH(t-c)}=cte-stdt=-1s[te-st]t=ct+1sce-stdt=ce-css+e-css2=c{H(t-c)}+e-css2,

    to obtain

    e-css2={tH(t-c)}-c{H(t-c)}={(t-c)H(t-c)}(t-c)H(t-c)=-1{e-css2}.

    Gathering the above results

    u(t,x)=v0[-1{1s2e-sx/a}--1{1s2}]=v0[(t-xa)H(t-xa)-t].

  2. Use the Fourier transform F(α)={f(x)}=-eiαxf(x)dx to solve the problem

    2ux2+2uy2=0,0<x<π,y>0,
    u(0,y)=f(y),ux(π,y)=0,y>0,
    uy(x,0)=0,0<x<π.

    Solution. The problem is defined on the half-infinite strip y>0, and the boundary condition at y=0 is on the derivative of the function, uy(x,0)=0. The appropriate transform is the cosine transform

    U(x,β)={u(x,y)}=0cos(βy)u(x,y)dy.

    Integration by parts, assuming u,u/y0, as y leads to

    0cos(βy)2uy2(x,y)dy=[cos(βy)uy(x,y)]y=0y+β0sin(βy)uy(x,y)dy=
    -uy(x,0)+β0sin(βy)uy(x,y)dy=-uy(x,0)+β([sin(βy)u]y=0y-β0cos(βy)u(x,y)dy)=
    -uy(x,0)-β2U(x,β).

    Replacing into Laplace PDE and using boundary condition at y=0, gives

    d2Udx2-β2U=-uy(x,0)=0,

    with solution

    U(x,β)=c1(β)cosh(βx)+c2(β)sinh(βx).

    Cosine transform of the boundary condition at x=0 leads to

    U(0,β)=F(β)=0cos(βy)f(y)dy=c1(β),

    specifying that c1(β)=F(β), the cosine transform of the boundary value f(y). Compute the x-derivative

    dUdx(x,β)=β[c1(β)sinh(βx)+c2(β)cosh(βx)].

    The boundary condition at x=π specifies

    dUdx(π,β)=0=β[c1(β)sinh(βπ)+c2(β)cosh(βπ)]c2(β)=-c1(β)sinh(βπ)cosh(βπ)=-F(β)tanh(βπ).

    The solution is obtained by taking the inverse cosine transform of

    U(x,β)=F(β)[cosh(βx)-tanh(βπ)sinh(βx)]
    u(x,y)=2π0F(β)[cosh(βx)-tanh(βπ)sinh(βx)]cos(βy)dβ=2π0F(β)cosh[β(x-π)]cos(βy)cos(βπ)dβ.
  3. Find all solutions of the equation z8-2z4+1=0. Write the roots in both Cartesian and polar form.

    Solution. Introduce w=z4 to obtain w2-2w+1=(w-1)2=0 with double root w1,2=1. The equation z4-1=0z4=e2πi, has roots zk=eikπ/2 for k=0,1,2,3. The original equation z8-2z4+1=0 as double roots at each zk.

  4. Sketch the region defined by -1Im(1/z)<1. Is this region a domain?

    Solution. With z=x+iy,

    Im(1x+iy)=Im(x-iyx2+y2)=-yx2+y2.

    The inequality -y/(x2+y2)<1 leads to -y<x2+y2x2+(y+12)2-14>0, the exterior of a circle of radius 1/2 centered at (0,-1/2), excluding the circle. The inequality -1-y/(x2+y2) leads to x2+y2-y0x2+(y-12)2-140, the exterior of a circle of radius 1/2, centered at (0,1/2). The region is not a domain since it is not an open set.

  5. Is f(z)=x2-x+y+i(y2-5y-x) an analytic function? Is it differentiable along the curve y=x+2?

    Solution. With z=x+iy, f(z)=u(x,y)+iv(x,y),

    u(x,y)=x2-x+y,v=y2-5y-x

    the first Cauchy-Riemann condition is not verified, i.e.,

    ux=2x-1vy=2y-5,uy=1=-vx

    so f is not analytic. On the curve C:y=x+2 the function restriction is a quadratic polynomial in x

    fC(x)=x2-x+x+2+i[(x+2)2-5(x+2)-x]=x2+2+i(x2-2x-6),

    and is differentiable, fC'(x)=2x+2i(x-1).

  6. Evaluate the integral

    C2zz2+3dz

    for C defined as:

    1. |z|=1;

      Solution. Assume C traversed in positive (CCW) direction. The integrand has isolated pole singularities at z1,2=±i3 not within the unit circle, hence

      C2zz2+3dz=0
    2. |z-2i|=1.

      Solution. The curve C is a circle of radius 1 centered at (0,2). The singularity z1=i3 is within the CCW-traversed contour, the singularity z2=-i3 is outside. The integral is therefore

      C2zz2+3dz=Cf(z)dz=2πiRes[f(z=i3)].

      From

      2zz2+3=1z-i3+1z+i3

      observe Res[f(z=i3)]=1.

  7. Expand

    f(z)=1z(1-z)2

    in a Laurent series valid for:

    1. 0<|z|<1;

      Solution. Differentiation term-by-term of the convergent series for |z|<1

      11-z=1+z+z2+z3+1(1-z)2=1+2z+3z2+4z3+

      and multiplying by 1/z (|z|>0z0) gives

      f(z)=1z+2+3z+4z2+5z3++(k+2)zk+
    2. |z|>1.

      Solution. Substitute w=1/z, |w|<1, and use above series to obtain

      1z(1-z)2=w(1-1w)2=w3(w-1)2=w3(1-w)2=w3+2w4+3w5+.

      Substitue back z=1/w and deduce

      f(z)=1z3+2z4+3z5++k-2zk+.