I'm new to numerical analysis. I'm still unclear as to why the harmonic series $\sum 1/k $, where k = 1, ... , infinity converges. I would appreciate any help! I can show that the sum does indeed converge in Matlab, I'm just unclear as to the reasoning behind it.


  • 2
    $\begingroup$ In finite precision, there is a smallest representable number. At some point, 1/k is less than this smallest number, and must be rounded to zero. All successive terms are then zero. $\endgroup$ – user14717 Dec 13 '18 at 19:37

Simple situation: fixed point precision. Then there is a smallest number in the system, and once $1/n$ is less than that number, the "numerical harmonic series" no longer changes its value.

Slightly more complicated situation: floating point precision. Then there is a largest representable number $\epsilon_{mach}$ such that the numerical result of $1+\epsilon_{mach}$ is just $1$ (but the same result doesn't hold for $0.1+\epsilon_{mach}$). Once $1/n<\epsilon_{mach}$, certainly the harmonic series up to $n-1$ will exceed $1$ (since the first term is already $1$), so again the series will no longer change. Note that this happens long before the smallest representable number in the system is reached; for example in IEEE double precision, $\epsilon_{mach}$ is around $10^{-16}$ while the smallest representable number is around $10^{-300}$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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