17

Motivation: The idea for this question came from an situation I encountered recently. A friend of mine was working on a math problem and tried to compute $ \Gamma\left(\frac{1}{4}\right) $, thinking it might be required to find an exact closed form. I quickly told him that it wasn’t worth the effort, as no 'nice' closed form exists for $ \Gamma\left(\frac{1}{4}\right) $. I also added that, the only $ x \in (0,1) $ with a simple closed form for $ \Gamma(x) $ is $ \frac{1}{2} $ and it is impossible to find a simple closed form for the rest, so he won't waste his time trying to find $\Gamma(x) $ since it isn't required or even possible. But this got me wondering "really? no $x \in (0,1)\setminus\{1/2\}$ has a nice closed form for $\Gamma(x)$?!"


It is well-known that $\Gamma\left(\frac 12\right)=\sqrt \pi$ This made me curious: are there any other values $x\in (0,1)\setminus\{1/2\}$ for which $\Gamma(x)$ has a simple closed form?.

Any parameter $x$ involved should also be expressible in a simple closed form.

I searched on Wikipedia but I couldn’t find any such $x$ , Additionally, the lack of simple closed forms for $\Gamma\left(\frac 14\right)$ and $ \Gamma\left(\frac 13\right)$ seems to suggest that no such $x$ exists. However, I wonder if there could be some very specific, unusual value of $a$ for which we have a simple closed form for $\Gamma(x)$


Simple or nice closed form: is a finite combination of basic elementary functions, expressed in terms of integer arguments..

Basic elementary functions are $x^n, e^x, \sin(x)$ for all $n \in\mathbb{Z}$ and their inverses.

A combination of two functions $f$, $g$ formed through operations such as addition $(f+g)$, multiplication ($f\cdot g$), or composition ( $f\circ g$).


Update: I asked this question on MO here.

John Bentin
  • 20,004
pie
  • 8,483
  • 9
    $\Gamma \left(\frac{1}{4}\right)=\sqrt{\frac{(2 \pi )^{3/2}}{\text{agm}\left(1,\sqrt{2}\right)}}$ or: $\Gamma \left(\frac{3}{4}\right)=\sqrt{\text{agm}\left(1,\sqrt{2}\right) \sqrt{\frac{\pi }{2}}}$ where agm is ArithmeticGeometricMean see here: https://mathworld.wolfram.com/Arithmetic-GeometricMean.html – Mariusz Iwaniuk Jan 13 '25 at 09:16
  • 4
    @MariuszIwaniuk By closed form here I mean that a finite combinations of elementary functions, by combinations of $f, g$ here I mean ($f+g, f\cdot g, af \forall a \in \mathbb{R}, f(g(x))$), I do't want to turn a special function to another one! – pie Jan 13 '25 at 09:25
  • 4
    Maybe this helps: https://functions.wolfram.com/GammaBetaErf/Gamma/03/02/ and https://en.wikipedia.org/wiki/Particular_values_of_the_gamma_function – Mariusz Iwaniuk Jan 13 '25 at 09:44
  • 1
    @MariuszIwaniuk You seem to have good answers in these comments, both positive and negative. They should be an actual answer. – Lee Mosher Jan 14 '25 at 17:03
  • 7
    It’s not really enough to ask for a nice output: if you graph the gamma function, you’ll notice that it’s a one-to-one function from $(0,1)$ to $(1,\infty)$. So for instance there is some $x$ such that $\Gamma(x)=3/2.$ The problem is that there’s nothing nice about this $x$: it’s just as “bad” as a generic gamma value would be. So what you really want is $y=\Gamma(x)$ where both input and output are “nice”. – Semiclassical Jan 14 '25 at 17:06
  • 2
    @Semiclassical That is exactly the point of this question, I wrote "Side note $a$ also should have a simple closed form." – pie Jan 14 '25 at 17:09
  • I think Chowla-Selberg formula also counts, if you consider "special values of modular forms" as "simple closed forms". – Seewoo Lee Jan 14 '25 at 17:44
  • I think that no such $a$ exist but this conjecture can be wrong – pie Jan 15 '25 at 13:16
  • This paper may be helpful: Raimundas Vidūnas, Expressions for values of the gamma function. It contains some very beautiful closed form expressions for the ratio of gamma functions. – Maxime Jaccon Jan 16 '25 at 10:55
  • there is a form involving the zeta function, although this is not finite :( $$\Gamma(1/4) = \frac{\zeta(3/4)}{\zeta(1/4)}\sqrt{\frac{\sqrt{2\pi}(2-\sqrt{2})}{2}}$$ – Maxime Jaccon May 19 '25 at 16:10

2 Answers2

8

If you can accept $\pi$ as a closed-form, then that may be a start as there are actually generalizations of $\pi$ using closed curves other than a circle.


I. Integrals

First define the $\pi_k$ as,

\begin{align} \pi_2 &= 2\int_0^1\frac1{(1-x^2)^{1/2}}\,dx = 3.1415926535\dots\\[4pt] \pi_3 &= 3\int_0^1\frac1{(1-x^3)^{2/3}}\,dx = 5.2999162508\dots\\[4pt] \pi_4 &= 4\int_0^1\frac1{(1-x^4)^{3/4}}\,dx = 7.4162987092\dots\\ \end{align}

Alternatively,

\begin{align} \pi_2 &= \;\Gamma\big(\tfrac12\big)\,\Gamma\big(\tfrac12\big)\,=\,\pi\\[4pt] \pi_3 &= \frac{\;\Gamma\big(\tfrac13\big)\,\Gamma\big(\tfrac13\big)}{\;\Gamma\big(\tfrac23\big)\,\Gamma\big(\tfrac33\big)} = \frac{\sqrt3\,\Gamma^3\big(\tfrac13\big)}{2\pi}\\[4pt] \pi_4 &= \frac{\sqrt2\,\Gamma\big(\tfrac14\big)\,\Gamma\big(\tfrac24\big)}{\,\Gamma\big(\tfrac34\big)\,\Gamma\big(\tfrac44\big)}=\frac{\Gamma^2\big(\tfrac14\big)}{\sqrt\pi}\\ \end{align}

Incidentally,

  • The half-period $\dfrac{\pi_3}{2\sqrt3} = \omega_2 \approx 1.529954$ is also called the omega_2 constant.
  • The half-period $\dfrac{\pi_4}{2\sqrt2} = \varpi = L \approx 2.622057$ is also called the lemniscate constant.

II. As arc lengths

Note that,

  • $\pi_2 \approx 3.14159$ is arc length for the unit circle.
  • $\pi_3\sqrt3 \approx 9.179724$ is arc length for the unit trefoil.
  • $\pi_4/\sqrt2 \approx 5.244115$ is arc length for the unit lemniscate.

III. Role in $x^k+y^k=1$

Note that,

  • $\pi_2$ for the trigonometric functions and $x^2+y^2=1$, with $\color{blue}{2\sin^{-1}(1)=\pi_2}$.
  • $\pi_3$ for the Dixon elliptic functions and $x^3+y^3=1$, with $\color{blue}{3\operatorname{sm}^{-1}(1)=\pi_3}$.
  • $\pi_4$ for the lemniscate elliptic functions and $x^4+y^4=1$, with $\color{blue}{4\sqrt2\,\operatorname{sl}^{-1}(1)=\pi_4}$.

$$\sin^2(z)+\cos^2(z)=1$$ $$\operatorname{sm}^3(z)+\operatorname{cm}^3(z)=1$$ $$\big(1+\operatorname{sl}^2(z)\big)\big(1+\operatorname{cl}^2(z)\big)=2$$

with the latter two being the Dixon and lemniscate functions, respectively. The $x^4+y^4=1$ is obeyed by the hyperbolic lemniscate functions.


IV. Conclusion

Therefore, if we allow the $\pi_k$ as closed-forms and fundamental constants in $x^k+y^k=1$, then $\Gamma\big(\tfrac12\big)$ can be expressed by $\pi_2$, while $\Gamma\big(\tfrac13\big)$ by $\pi_3$, and $\Gamma\big(\tfrac14\big)$ by $\pi_4$ using the formulas in the first section.

  • 1
    That is interesting! However the only reason that I accepted $\pi$ as a closed form is that $\pi =2\arcsin(1)$, The only functions allowed are $x^n, e^x, \sin(x)$ and their inverses – pie Jan 19 '25 at 12:08
  • 2
    Nice! Where did you find these (or did you create the generalizations on your own)? – Maxime Jaccon Jan 19 '25 at 14:25
  • 3
    @MaximeJaccon They are actually already known in the literature. In fact, I used the same notation $\pi_k$. Multiplied by a suitable square root, $\pi_3$ and $\pi_4$ are known as the omega-2 constant and lemniscate constant respectively. – Tito Piezas III Jan 19 '25 at 14:31
  • 1
    @pie I guess that certainly narrows things. Though personally I would love to find elementary expressions for $\pi_3$ and $\pi_4$ (and automatically $\Gamma\big(\tfrac13\big)$ and $\Gamma\big(\tfrac14\big)$), the data above suggests their relatives live in $x^k+y^k=1$ for $k>2$, while trigonometric functions are only for $k=2$. – Tito Piezas III Jan 19 '25 at 14:57
  • 2
    @pie Regarding your statement why you accepted $\pi_2$, I dug deeper into the Dixon elliptic functions $\operatorname{sm}(z)$ and $\operatorname{cm}(z)$. Consider the striking analogies, $$2\sin^{-1}(1)=\pi_2$$ $$3\operatorname{sm}^{-1}(1)=\pi_3$$ partly a consequence of the relationships they obey, $$\sin^2(z)+\cos^2(z)=1$$ $$\operatorname{sm}^3(z)+\operatorname{cm}^3(z)=1$$ Perhaps the elliptic Dixon $\operatorname{sm}(z)$ and $\operatorname{cm}(z)$ deserve to be on equal footing with the trigonometric $\sin(z)$ and $\cos(z)$. – Tito Piezas III Jan 19 '25 at 19:18
  • 1
    Great answer - I suggest reading this related question I asked in the same period https://math.stackexchange.com/q/5019801/1135547 – Davide Masi May 19 '25 at 16:42
6

No explicit evaluations of other gamma values are known. See the recent paper by Raimundas Vidūnas, Expressions for values of the gamma function. (available online)

The author writes

Values of the gamma function are of broad interest. ... An easy consequence of the reflection formula is $\Gamma(\frac{1}{2}) = \sqrt \pi$. ... No explicit evaluations of other gamma values are known.

Unfortunately, per the conditions of your question, only $\Gamma(\frac{1}{2})$ is known. The form of known representations of the gamma function and terms are discussed soon after.

Some gamma terms occur as values of hypergeometric functions at special points and as period integrals. ... Conversely, some gamma values at rational points can be expressed in terms of the elliptic integrals.

Nonetheless, some beautiful relations between specific gamma values exist (see the paper above).

$$\frac{\Gamma(\frac{1}{2}) \Gamma(\frac{2}{3})}{\Gamma(\frac{3}{4}) \Gamma(\frac{5}{12})} = \frac{\sqrt{\sqrt 3 +1}}{2^{1/4} 3^{3/8}}$$

$$\frac{\Gamma\left(\frac15\right)\Gamma\left(\frac4{15}\right)}{\Gamma\left(\frac13\right)\Gamma\left(\frac2{15}\right)}=\frac{\sqrt2\,\,\sqrt[20]3}{\sqrt[6]5\,\sqrt[4]{5-\frac{7}{\sqrt{5}}+\sqrt{6-\frac{6}{\sqrt{5}}}}}=\frac{\phi \,\, \sqrt[20]3 \,\, \sqrt{\!\sqrt 3 \cdot \sqrt[4] 5-\phi^{3/2}}}{\sqrt 2 \,\, \sqrt[24] 5}$$

Hope this helps :)

  • 1
    https://www.jstage.jst.go.jp/article/kyushujm/59/2/59_2_267/_pdf/-char/en – kimchi lover Jan 17 '25 at 13:58
  • I see that the paper was published in 2004, and it makes me wonder if there have been any developments in the past two decades regarding simple closed forms for other gamma values within $(0,1)-{1/2}$ – pie Jan 18 '25 at 21:08
  • The wikipedia article on the Gamma function gives "It might be tempting to generalize the result that $\Gamma(\frac{1}{2}) = \sqrt \pi$ by looking for a formula for other individual values $\Gamma(r)$ where $r$ is rational ... However these numbers $\Gamma(r)$ are not known to be expressible by themselves in terms of elementary functions." It has been proven that some Gamma values are transcedental and algebraically independent of $\pi$. – Maxime Jaccon Jan 18 '25 at 21:14
  • See Waldschmidt, M. (2006). "Transcendence of Periods: The State of the Art," which proves the algebraic independence and transcendentality of some $\Gamma(r)$ :) – Maxime Jaccon Jan 18 '25 at 21:15