Write , with . Under the Cole-Hopf transformation , direct differentiation givesSince , Burgers' equation is equivalent toThe ratio is therefore a function alone. Multiplying by leaves unchanged and removes this scalar freedom. Thus one may choose the normalization for whichThis is the heat equation. The positive sign in corresponds to the negative conservative flux used here; the more usual positive-flux convention has the opposite sign.
Integrate the initial Cole-Hopf transformation and choose a continuous positive normalization:Split the heat kernel convolution at zero, and complete the square in the positive half-line contribution. DefineThe resulting heat equation solution isDifferentiating the moving endpoints gives and . These terms cancel. Hence , and the viscous Burgers step solution with negative flux isThe lower limit of is , as printed in the original PDF. The converted TeX's does not give the correct convolution or endpoint cancellation.
For the vanishing-viscosity limit, the conservative flux is and the characteristic speed is . If , the characteristics from the right move left and collide with those on the left. The Rankine-Hugoniot condition gives shock speed . ThusThe viscous transition has width , and its centre value is exactly by the midpoint symmetry of a viscous Burgers step.
If , the right characteristics move right and separate from the left ones. The limit is a rarefaction wave,This follows either from the Burgers Riemann problem with negative flux or the large-argument Gaussian-tail asymptotics of the exact formula. The two signs therefore give a shock and a spreading rarefaction respectively, not two shocks.
Set and write . Substitution into the modified Burgers equation with cubic flux givesAfter one integration,A regular constant limiting state has . The state at the negative end sets , and the nonzero state at the positive end givesThus the first-order equation is . Put . Its equation is logistic,The heteroclinic travelling wave connecting the prescribed endpoints is consequentlywhere is an arbitrary translation. Taking the sign from is essential; squaring the equation alone loses the negative-front solution. A fixed level propagates with , also obtained from the Rankine-Hugoniot condition for flux .
Articles by others on the same topic
There are currently no matching articles.