While constructing some exponential curve, I have been trying to solve the following:


for $z$. I have tried using WolframAlpha, but it was unable to get any answer for $z$ within the standard computation limit, while Symbolab couldn't arrive at any answer. I'm not sure how to solve this by hand. I would initially try to reduce it to the form

$$\left(v^{\log _b\left(a-m\right)}+m\right)^z-m^z=a-m$$

but now, I'm not sure how to reduce this any further to get an answer for $z$.

In context, I can assume that $v$, $b$, $a$ and $m$ are positive real numbers, and that $a > m$.

I couldn't seem to be able to apply any log rules (since the bases would be different; e.g. $\log(x)+\log(y)=\log(xy)$ only holds when both $\log(x)$ and $\log(y)$ were to have the same base, which in this case they don't), so I don't know how to continue.


1 Answer 1


I suppose (and more) that you will need some numerical method; however, there a few things that xe could do considering that we look for the zero's of function $$f(z)=c ^z-m ^z-d \quad \text{where} \quad c=\left(v^{\log _b\left(a-m\right)}+m\right)\quad \text{and} \quad d=(a-m)$$ We have $$f'(z)=c^z \log (c)-m^z \log (m)\quad \text{and} \quad f''(z)=c^z \log ^2(c)-m^z \log ^2(m)$$ The first derivative cancels at $$z_*=-\frac{\log \left(\frac{\log (c)}{\log (m)}\right)}{\log (c)-\log (m)}$$ At this point, we need to assume that $$f(z_*) <0 \quad \text{if} \quad f''(z_*) >0$$ or that $$f(z_*) >0 \quad \text{if} \quad f''(z_*) <0$$

If this is the case, for an approximation, you could build a Taylor expansion around $z_*$ and get as possible estimates $$z_\pm=z_*\pm \sqrt{-2\frac {f(z_*)}{f''(z_*)}}$$

Then use Newton method to polish the root(s) if any does exist in the real domain.


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.