0
$\begingroup$

I found the following approximation for the function $$f=\ln(x+1)$$ $$f\simeq\Psi\left(x+\dfrac{3}{2}\right)-2+\gamma+\ln(2)$$ where $\Psi(x)$ is the 'Digamma' function: $\Psi(x)=\dfrac{\dfrac{d}{dx}\Gamma(x)}{\Gamma(x)}$ and $\gamma$ is the Euler - Mascheroni constant. The following inequality holds: $$\left|\ln(x+1)-\Psi\left(x+\dfrac{3}{2}\right)-2+\gamma+\ln(2)\right|\lt 0.036$$ Is this a known result? Thanks

$\endgroup$
1
$\begingroup$

I'm not sure why there is a $(-2+\gamma+\ln 2)$, because $\left|\ln(x+1) - \Psi(x+\frac32)\right| < 0.0365$ itself is already satisfied.

The approximation comes from the series expansion of $\exp\left(\Psi(x+\frac12)\right)$, which states, for $x > 1$,

\begin{align} \Psi\left(x + \frac12\right) = \ln\left( x + \frac{1}{4!\cdot x} - \frac{37}{8\cdot 6!\cdot x^3} + o\left(\frac1{x^5}\right) \right) \end{align} and thus $\Psi(x+\frac32)$ is approximately $\ln(x+1)$. The largest difference happens at $x=0$ which is $\Psi(\frac32) = 2 - \ln 4 - \gamma \approx 0.03648997$.

$\endgroup$
0
$\begingroup$

I doubt your approximation. First I guess you missed a factor $2$ for $\ln 2$, the approximation should be $$f(x) =\ln(x+1)-\Psi(x+3/2)-2+\gamma+2\ln 2 \cdot$$ If you omit the $2$ you have $$f(0)= 0 -(2-\gamma-2\ln2) -2+\gamma+ \ln 2 = -4 + 2\gamma + 3\ln 2 \approx -0.766$$ while with the factor $2$ you have $$f(0)= 0 -(2-\gamma-2\ln2) -2+\gamma+ 2\ln 2 = -4 + 2\gamma + 4\ln 2 \approx -0.0723$$ which is much smaller but still violates your given bound.

PS: A Taylor series at $x=0$ computed with Maple is $$\ln(x+1)-\Psi(x+\tfrac{3}{2}) = (-2+\gamma+2\ln2)+(5-\tfrac{1}{2}\pi^2)x+(-\tfrac{17}{2}+7\zeta(3))x^2 +O(x^3)$$

$\endgroup$
  • $\begingroup$ In fact, I omissed to write that the inequality holds for $x \gt 1$ $\endgroup$ – Riccardo.Alestra May 31 '16 at 10:01
  • $\begingroup$ But in any case you have to add the factor $2$, otherwise you have $f(2)\approx -0.734$! But even with the factor you have $f(2)\approx-0.041$ $\endgroup$ – gammatester May 31 '16 at 10:05

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.