= Singular perturbation
{title2=$\epsilon\to0^+$}
{wiki}
A <singular perturbation> is a parameter-dependent problem whose limiting problem changes qualitatively as the parameter tends to zero. In an <ordinary differential equation>, a small coefficient of the highest <derivative> can reduce the order of the limiting <differential equation>, preventing it from satisfying all the original <boundary conditions>. Different length or time scales often require different <asymptotic expansions>.<Asymptotic Analysis and Singular Perturbation Theory by John K. Hunter>[§1.1.2, p. 2: regular and singular perturbations]{ref}
Consider the <boundary value problem>
$$
\epsilon y_\epsilon''+y_\epsilon'=0,\qquad 0<x<1,\qquad y_\epsilon(0)=0,\quad y_\epsilon(1)=1,\qquad \epsilon>0.
$$
This <second-order differential equation> is exactly solvable. Setting $v=y_\epsilon'$ gives the <first-order linear differential equation> $\epsilon v'+v=0$, so $v=C e^{-x/\epsilon}$. One <integration> and the two <boundary conditions> give
$$
y_\epsilon(x)=\frac{1-e^{-x/\epsilon}}{1-e^{-1/\epsilon}}.
$$
Both <boundary conditions> hold for every positive $\epsilon$. For numerical evaluation, write the same ratio as $\operatorname{expm1}(-x/\epsilon)/\operatorname{expm1}(-1/\epsilon)$, where $\operatorname{expm1}(t)$ evaluates the <exponential function> minus one without <catastrophic cancellation> near $t=0$. Keep the denominator for finite $\epsilon$: at $\epsilon=0.2$ it is approximately $0.993262$, rather than exactly one.
Setting $\epsilon=0$ gives $y_0'=0$. The <outer solution> selected by the right <boundary condition> is therefore $y_0=1$, which fails the left <boundary condition>. The exact <exponential function> $e^{-x/\epsilon}$ shows how that mismatch is repaired: the solution rises from zero to almost one within a <boundary layer> of width $x=O(\epsilon)$, using <Big O notation>. For each fixed $x>0$, $y_\epsilon(x)\to1$, while $y_\epsilon(0)=0$ always. This <pointwise convergence> fails to be <uniform convergence> to the <outer solution> on $[0,1]$, since $\sup_{0\leq x\leq1}|y_\epsilon(x)-1|=1$ for every $\epsilon>0$.
\Image[/singular-perturbation-boundary-layer-profiles.png]
{title=Exact solutions with a shrinking left boundary layer}
{description=The exact normalized profiles at $\epsilon=0.2$, $0.05$, and $0.01$ satisfy both <boundary conditions>. The dashed <outer solution> misses the left <boundary condition>. Colored dots mark $x=\epsilon$, where most of the initial rise has already occurred.}
The coefficient $\epsilon$ is small, but its product with the <second derivative> need not be small. Direct <differentiation> gives
$$
y_\epsilon'(x)=\frac{e^{-x/\epsilon}}{\epsilon(1-e^{-1/\epsilon})},\qquad
y_\epsilon''(x)=-\frac{e^{-x/\epsilon}}{\epsilon^2(1-e^{-1/\epsilon})}.
$$
At $x=\epsilon X$ with fixed $X\geq0$, these <derivatives> grow as $\epsilon^{-1}$ and $\epsilon^{-2}$ respectively as $\epsilon\to0^+$. Consequently $\epsilon y_\epsilon''$ and $y_\epsilon'$ have exactly equal magnitudes and opposite signs. Even away from the <boundary layer>, their magnitudes remain equal in this example; both become exponentially small at each fixed positive $x$. The <outer expansion> neglects an exponentially small transient there, rather than establishing that one <derivative> term is smaller than the other everywhere.
To predict the width without knowing the exact solution, suppose an order-one change occurs over distance $\delta$. The <derivative> scales are $y'=O(\delta^{-1})$ and $y''=O(\delta^{-2})$. The <dominant balance> $\epsilon/\delta^2\sim1/\delta$ gives $\delta\sim\epsilon$. The resulting <inner variable> $X=x/\epsilon$ makes the <boundary layer> visible at a fixed scale. This is the useful feature of <singular perturbation>: an apparently disappearing term identifies the missing structure when one measures its <derivatives> on the appropriate scale.
Back to article page