Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm studying for a test and I'd like to know how justify why the only $k$-step method of order $k$ with stiff decay is BDF. By definition of stiff decay(Ascher & Petzold) a method has stiff decay if

\begin{equation}|y_n-g(t_n)|\rightarrow \ 0,\qquad \text{as }h_nRe(\lambda)\rightarrow -\infty,\end{equation}


\begin{equation} y'=\lambda(y-g(t)),\end{equation}

and $g(t)$ is an arbitrary bounded function. Assuming stiff decay and considering the definition of the general LMM I don't see why this forces $\beta_j=0$ for $j>0$. Thanks for your time.

share|cite|improve this question
I do not have that book. Can you tell us what BDF and LMM are? – Amzoti Aug 8 '13 at 0:19
@Amzoti My apologies LMM refers to linear multistep methods, those of the form $\sum\limits_{j=0}^k\alpha_jy_{n-j} =\sum\limits_{j=0}^k\beta_j f(t_{n-j},y_{n-j})$. BDF stands for backward differentiation formulae, linear multistep methods where $\beta_j=0$ for $j>0$. – Dave Aug 8 '13 at 1:27
There should also be an $h$ in front of the RHS summation. – Dave Aug 9 '13 at 2:29
up vote 1 down vote accepted

Applying a linear multistep method to the equation $y' = \lambda(y-g(t))$ yields $$ \sum_{j=0}^k \alpha_jy_{n−j} = \sum_{j=0}^k h \lambda \beta_j(y_{n-j}-g(t_{n−j})), $$ which we can rewrite as $$ \sum_{j=0}^k \beta_j(y_{n-j}-g(t_{n−j})) - \frac1{h\lambda} \sum_{j=0}^k \alpha_jy_{n−j} = 0. $$ In the limit that $h\lambda \to -\infty$, the second term goes to zero, so this becomes $$ \sum_{j=0}^k \beta_j(y_{n-j}-g(t_{n-j})) = 0, $$ which we can re-arrange as $$ y_n - g(t_n) = \frac{1}{\beta_0} \sum_{j=1}^k \beta_j(y_{n-j}-g(t_{n-j})). $$ Since this needs to be zero for all values of $y_{n-j} - g(t_{n-j})$, the coefficients $\beta_j$ ($j=1,\dots,k$) need to be zero.

share|cite|improve this answer
Okay that's where I'm confused then because when applied to the general method we have \begin{equation}\sum\limits_{j=0}^k \alpha_jy_{n−j}=\sum\limits_{j=0}^k h \lambda \beta_j(y_{n-j}-g(t_{n−j})).\end{equation} So where do the left hand side terms go? – Dave Aug 9 '13 at 2:27
I added some details on the first step. – Jitse Niesen Aug 9 '13 at 13:14

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.