Derivation of Gamma (N Over 2)
Derivation of Gamma (N Over 2)
Derivation of
Γ ( )& ( )
1
2
Γ
n
2
Boas, Mary L. Mathematical Methods in the Physical Sciences, 3rd ed., 2006. NY: Wiley.
Carr, G.S. Formulas and Theorems in Pure Mathematics, 2nd ed. New York: Chelsea Publishing
Co., 1970.
https://archive.org/details/synopsisofelemen00carrrich
Finney, R.L. and G.B. Thomas, Jr. Calculus, 1990. Addison-Wesley Publishing Co.
O’Brien, Francis J. Jr. 500 Integrals of Elementary and Special Functions, 2009.
ISBN-13: 978-1439219812.
http://www.docstoc.com/docs/23969109/500-Integrals-of-Elementary-and-Special-Functions
____________. “Double Factorials: Selected Proofs and Notes,” Apr. 23, 2009.
http://www.docstoc.com/docs/167385400/Double-Factorials-Selected-Proofs-and-Notes
n n
____________.‟ Identities for cos ( x) & sin ( x) By Complex Numbers,” Apr. 24, 2013.
http://www.docstoc.com/docs/160476032/Identities-for-Cosine--cos%5En(x)-and-Sine-sin
%5En(x)-by-Complex-Numbers.
Sokolnikoff, I. S. Advanced Calculus. New York: McGraw Hill Book Co., 1939.
Introduction
In this research note a detailed derivation is given of the formula for the gamma function,
Γ ( n2 ), n=odd integer≥1 .
We show first the remarkable result,
Γ ( 12 )= √ π ,
and generalize it by
two methods to any odd n. An application of the integration method for the normal or Gaussian
probability integral is also provided.
Γ ( 12 ) Derivation
This demonstration is based on Carr’s 19th century original proof with a little more detail
(see Corollary 2314, p. 362).
Γ ( n )=∫ t e dt , n>0
n−1 −t
0
Use change of variable with same limits:
t=x 2
dt=2 xdx
Then,
∞
2
Γ ( n )=2 ∫ x e dx
2n−1 −x
0
()
∞
1
Γ =2∫ e− x dx
2
2 0
Now multiply
Γ ( 12 ) by itself using a different variable name for the new integrand, and create a
double integral:
()() ()
12
∞ ∞ ∞ ∞
1 1 − ( x 2+ y 2)
Γ Γ =Γ =2∫ e dx 2∫ e dy=4∫ ∫ e
−x2 − y2
dx dy
2 2 2 0 0 0 0
1
See Boas (Ch. 11, “Special Functions,” pp. 538 ff.) for introductory material on the gamma function.
()
∞ ∞
1
2∫ e dx=2 ∫ e− y dy=Γ
−x2 2
,
since 0 0 2 by definition.
x=r cos θ
y=r sin θ
2 2 2
x + y =r
r= √ x 2 + y 2 (absolute value )
y
tan θ=
x
θ=tan −1
y
x () −x 2
Plot of 2e ( x≥0 )
Source: http://www.wolframalpha.com/
dxdy=rdrd θ
0≤r <∞
π
0≤θ≤
2 x
NOTE: Since 0≤x<∞ , then clearly 0≤r <∞ because x=r cos θ starts at 0 (by convention)
and extends indefinitely to + ∞ . With regards to the limits of the angle θ in radians, since
y π
tan θ= , x=0 ,tan θ=∞ , θ=tan −1 ( ∞ ) =
0≤x<∞ , and x then, when 2 , and when
π
−1 0≤θ≤
x=∞,tan θ=0,θ=tan ( 0 )=0. Thus, 2 , meaning we are in the first quadrant of the
circle bounded by x 2 + y 2 =r 2 .
Substituting the polar coordinate variables and evaluating the polar integral,
π π
θ= θ=
( ) (
2 r=∞ 2
()
12
)
∞ ∞
− ( x 2 + y2 ) 2 1 π
Γ =4 ∫ ∫ e dx dy=4 ∫ ∫ re −r
dr dθ=4 ∫ dθ=2 =π
2 0 0 θ =0 r=0 2 θ=0 2
( )
r=∞
∫ re−r dr =1/2
2
r =0
NOTE: The inner integral is solved by change of variable2,
2
u=r ; du=2rdr .
Also, Fubini’s Theorem (see Finney/Thomas, p. 895-9), regarding the order
of integration of iterated integrals, tells us we could have calculated:
π
θ=
2 r=∞
∫ dθ ∫ re−r dr=4 ( 2 )
2 π 1
4 =π .
θ=0 r =0
2
π
θ=
2 2π θ=2 π r=∞
∫ ∫ ()
1
2
4 ∫ dθ= ∫ dθ=2π . dθ re−r dr=2 π
2
=π .
θ=0 θ=0 θ=0 r=0
Note that Thus,
()
2
1
Γ =π
2
Thus, we have demonstrated that or, taking positive square root value only,
Γ ( 12 )= √ π
This is one of the most important relations in applied mathematics, probability & statistics, and
engineering3.
2
A more general formula useful for
Γ ( n2 )
is Formula 3.326.2 in Gradshteyn & Rhyzik, Table of Integrals, Series
∞
∫ 1 m+1
>0 .
n
x m e −βx dx = γ Γ ( γ ) , γ=
nβ n
and Products (2007): 0 Here, we calculate the inner integral above
Γ (1 ) 1
=
with x =r , m=1 , β=1 ,n=2, γ=1 , and the result is, 2 2 since Γ ( n )=(n−1 )! , and where 0! = 1 by definition.
π
Γ ( m ) Γ ( 1−m )=
3
An alternative derivation comes from Euler’s Reflection Formula (Carr, p. 362), sin mπ with
m=1/ 2 .
Γ ( n+1 )=nΓ ( n )
A Useful Recursion Formula,
This relation is derived by integration by parts, where n is any real number > 0. The
Γ ( n/2 )
solution will provide a means to evaluate for any finite odd integer.
u=t n
n−1
du=nt dt
dv=e−t dt
v=−e−t By Parts: ∫ udv=uv−∫ vdu
Calculate by parts,
|∞ ∞ n
[]
t |∞ ∞
Γ (n+1)=uv ¿| −∫vdu=− t n−1 −t t t n−1 −t
n n ∞
e | +n∫t e dt ¿=− lim t −lim t +n∫t e dt ¿=0+nΓ (n)=nΓ (n) ¿¿
|0 0 0 t→∞ e t→0 e ⏟ 0
|0
¿ Γ (n )
n
t
lim t =0
t →∞ e
NOTE: the limit is well known [See proof in O’Brien, “A General Limit…”
(Appendix)].
Γ ( n/2 ) n.
NOTE: This recursion formula calculates for any odd
Γ ( n+1 )=nΓ ( n )
For example, by ,
Γ ( 32 )=Γ ( 12 +1)= 12 Γ ( 12 )= 12 √ π
.
Likewise,
=(
⏟ )(
n−2 n−4 n−6
2 2 )( 2 ) … [ n−( n−1 )
2 ] Γ( )
1
2
n−1
terms
2
or,
NOTE: See O’Brien, [“Double Factorials…” (Table 1)] where it is shown that for n odd integer,
( n−1 ) !
( n−2 ) !!= , [ n≥+1 ]
2
n−1 n−1
2 2 ( )
!
Γ ()
n ( n−2 ) !!
2
= n−1 √ π
2 2
Then, if this is substituted into , it results:
Γ ( n2 )= 21 ( n−1 ) !
√π [ n odd integer≥1 ]
( )
n−1
n−1
!
2
Γ ( n2 )
This represents a second formulation of published in standard sources. The author’s
version with d. factorials published in 500 Integrals... (sect. 8.339) seems easier to derive and
compute from. Also, see O’Brien, “Double Factorials …” (pp 18-23).
Γ ( 72 )= 52!! √ π = 1⋅3⋅5
3 8
√π Γ ( 72 )= 21 63 !! √ π = 1⋅3⋅5
6 8
√π
, or
Γ ( n )= ( n−1 ) !
NOTE: The recursion formula teaches that for integers.
By integration,
∞
Γ ( 1 )=∫ e−t =1=0!=( n−1 ) !
0
By recursion,
Γ ( n2 )
By Polar Coordinates
Now we show another way to obtain the second formulation directly—by a double
integral in polar coordinates. This approach is more involved but confirms the original
calculation.
Γ ()
3
2
Let’s introduce the integration method for . Sokolnikoff (pp. 375 ff.) does this
integral.
Γ ( 32 )
Derivation
∞
t=x 2
dt=2 xdx
∞
2
Γ ( n )=2 ∫ x 2n−1 e−x dx
0
()
∞
3
Γ =2∫ x 2 e− x dx
2
2 0
() () ()
2
3 3 3
Γ Γ =Γ
2 2 2
∞ ∞ ∞ ∞
¿2∫ x e 2 − x2
dx 2 ∫ y e 2 − y2
dy=4 ∫∫ ( xy )2 e−( x + y ) dx dy
2 2
0 0 0 0
x=r cosθ
y=r sin θ
2 2 2
x + y =r
y
tan θ=
x
θ=tan
−1 y
x()
dxdy=rdrdθ
0≤r<∞
π
0≤θ≤
2
r∧θ
Substituting with the limits for as before, we obtain the reductions,
[ ]
π π
[ )]
2
()
2
3 2
( ) (
∞ ∞
π π
2 2
¿4
Γ (3 )
2 0
2
0
sin 2θ 2
∫ ( cosθ⋅sin θ ) dθ=4∫ 2 dθ ( ) [ Γ ( 3 ) =(n−1)!=2 ]
|2π
| =π ¿
π π π | 4
2 2 2
1−cos 4θ 1 1 |0
¿∫ dθ= ∫ dθ− ∫ cos ( 4θ ) dθ
0
2 20 20
π 1
¿ − ( sin 4θ )
4 8
¿
()
2
3 π
Γ
2
=
4
Γ ( 32 )= 12 √ π
Thus, since , as shown earlier.
sin ( A−B)+sin ( A +B )
sin A cos B= with A=B=θ
2
1−cos(2 A )
sin2 A=
2
[ ]
∞
Γ (3 ) 2 !
∫ f (r )dr= 2
= =1 ,
2
0
The inner integral, is solved by change of variable,
2
u=r ;du=2rdr ;r=√ u,
and the definition of the gamma function, or the general integral
∞
∫ x m e−βx dx ,
n
Γ ()
n
2
Derivation
t=x 2
dt=2 xdx
∞
2
Γ ( n )=2 ∫ x 2n−1 e−x dx
0
()
∞
n
Γ =2∫ x n−1 e−x dx
2
2 0
()
n 2
∞ ∞ ∞ ∞
n−1 − ( x2 + y2 )
=2 ∫ x n−1 e− x dx 2 ∫ y n−1 e− y dy=4 ∫∫ ( xy ) e
2 2
Γ dx dy
2 0 0 0 0
[ ] [ ]
π π
2 2
( ) ( )
∞ ∞
−r 2 2
=4 ∫ ( r cosθ⋅r sin θ )
0
n−1
∫ re0
dr dθ=4 ∫ ( cosθ⋅sin θ ) n−1
∫ r 2n−1 e−r dr dθ
0
0
π π
2 2
( )
n−1
Γ ( n) Γ ( n ) sin 2 θ
=4
2 0
∫ ( cosθ⋅sin θ )n−1 dθ=4
2 0 2
∫ dθ
π
2
Γ ( n)
n ∫
=4 ( sin2 θ )n−1 dθ
2 0
The inner integral is solved by either change of variable and the gamma function, or the footnote
2 integral with m=2 n−1 , β=1 ,n=2 ,γ=n .
n−1
At this point we are faced with expanding and integrating ( sin 2θ ) where n is odd
integer, and, therefore, n–1 is even integer. This integral is known and published but we will
evaluate it.
n
This sine expansion is adapted from the form given in O’Brien, ‟ Identities for cos ( x)
n
& sin ( x) By Complex Numbers” [p. 13]:
1
( sin 2θ )n−1 = ¿ ( n−1 ¿ ) ¿ ¿
2
n−1
¿
We must integrate:
π
2
Γ ( n)
4
2 n ∫ ¿¿ ¿
0
It is left to the reader to show that the second term integrates to sin and vanishes when the limits
cos ( n−1−2 k ) 2θ
are substituted [the only variable in the second term is ], leaving the solution
as:
( )
2
n Γ (n)
Γ =4 ¿¿
2 2 n
Γ ( n2 )= 21 ( n−1 ) !
√π [ n=1,3,5,7, … ]
( )
n−1
n−1
!
2
( n ¿) ¿¿ ¿
¿
1
∞ − x2
I=∫ e 2
dx
−∞
Different approaches can be used for the evaluation of this integral. Here we use the
1
− x2
2
Plot of even function f (x )=f (−x )=e (−∞< x <∞ )
1 1 2
∞ − x2 ∞
− x
∫e 2
dx=2 ∫ e 2
dx
Area of even function: −∞ 0
Source: http://www.wolframalpha.com/
Since the curve is symmetric around 0, we can double one–half of the curve:
4
For proof, see Finney/Thomas, p. 540 (“Techniques of Integration”).
1 1 2
∞
− x2 ∞
− x
I=∫ e 2
dx=2∫ e 2
dx
−∞ 0 .
1
Let a = 2 in the exponent to simplify the algebra:
1
∞ − x2 ∞
2
I=2∫ e 2
dx=2∫ e −ax
dx
0 0
Γ ( 12 )
Proceeding as earlier for , create a double integral, transform to polar coordinates,
and solve inner/outer integrals to arrive at the solution.
}
∞ ∞ ∞∞
−a (x 2+y 2)
I =2∫ e dx2∫ e dy=4∫ ∫ e
2 −ay 2
¿ ¿¿ [ a=1/2] ¿
2 −ax
dxdy
0 0 00
1
∞
− x2
∴∫ e 2
dx=√ 2π
−∞
————
NOTE: This integral can be solved without a double integral in polar coordinates since we know
1
∞
− x2
Γ ( 12 )=√ π . I=2∫ e 2
dx
The reader can see that if 0 is submitted into general integral,
∞
1 m=1
∫ x m e−βx dx= nβ γ Γ ( γ ) , γ=
n
,
0 n with parameters m=0 , β=1/2,n=2 , γ=1/2 , it
then results,
(
[]) 1
1
− x2 1 Γ
∞ ∞
− x2 2
∫e 2
dx=2 ∫ e 2
dx=2 =√ 2 π .
2( )
1
−∞ 0 1 2
2
NOTE: The solution seen in standard textbooks on probability theory sets up the calculation as:
(∫ )
θ=2 π r=∞ 1
− r2
I =∫
2
e 2
rdr dθ=2 π
θ=0 r=0
I=√ 2 π
Clearly the size of the radius r is 0 to ∞. The value of θ ranges from 0 to 2 π but may be set up
as we did above since,
π
θ2 2 2π
###