NNOTE ON EXTENSIONS OF THE BETA FUNCTION
MEHAR CHAND
Abstract.
The classical beta function B ( x, y ) is one of the most fundamental specialfunctions, due to its important role in various fields in the mathematical, physical,engineering and statistical sciences. Useful extensions of the classical Beta functionhas been considered by many authors. In the present paper, our main objective is tostudy the convergence of extensions of classical beta function and introduce modifiedextension of classical beta function. It is interpreted numerically and geometrically inthe view of convergence, further properties and integral presentations are established. Introduction and Preliminaries
In many areas of applied mathematics, various types of special functions become es-sential tools for scientists and engineers. The (Euler’s) classical beta function B ( x, y ) isone of the most fundamental special functions, because of its important role in variousfields in the mathematical, physical, engineering and statistical sciences. During last fourdecades or so, several interesting and useful extensions of the familiar special functions(such as the Gamma and Beta functions, the Gauss hypergeometric function, and soon) have been considered by many authors. In the present work, we are concerned withvarious generalizations of the classical beta function, which can be found in the literature(see [1, 2, 3, 4]).Recently the generalized beta function B δ,ζ ; κ,µp ( α, β ) is introduced by Srivastava et al.,which is the most generalized extension of classical beta function and is defined as(see[1]): B ( δ,ζ ; κ,µ ) p ( α, β )= (cid:90) t α − (1 − t ) β − F (cid:18) δ ; ζ ; − pt κ (1 − t ) µ (cid:19) dt, (1.1)where (cid:60) ( p ) ≥
0; min[ (cid:60) ( α ) , (cid:60) ( β ) , (cid:60) ( δ ) , (cid:60) ( ζ )] >
0; min[ (cid:60) ( κ ) , (cid:60) ( µ )] > F ( . ) isconfluent hypergeometric function, which is special case of the well known generalizedhypergeometric series p F q ( . ).The generalized hypergeometric series p F q ( p, q ∈ N ) is defined as (see [5, p.73]) and[6, pp. 71-75]: Mathematics Subject Classification.
Key words and phrases.
Pochhemmer symbol; gamma function; Beta function; hypergeometricfunction. a r X i v : . [ m a t h . C A ] A p r MEHAR CHAND p F q (cid:20) α , . . . , α p ; β , . . . , β q ; z (cid:21) = ∞ (cid:88) n =0 ( α ) n · · · ( α p ) n ( β ) n · · · ( β q ) n z n n != p F q ( α , . . . , α p ; β , . . . , β q ; z ) , (1.2)where ( λ ) n is the Pochhammer symbol defined (for λ ∈ C ) by (see[6, p.2 and p.5]):( λ ) n := (cid:26) n = 0) λ ( λ + 1) . . . ( λ + n −
1) ( n ∈ N ) (1.3)= Γ( λ + n )Γ( λ ) ( λ ∈ C \ Z − ) , (1.4)and Z − denotes the set of Non-positive integers and Γ( λ ) is familiar Gamma function.When κ = µ , (1.1) reduces to the generalized extended beta function B ( δ,ζ ; µ ) p ( α, β )defined by (see [2, p. 37]) B ( δ,ζ ; µ ) p ( α, β )= (cid:90) t α − (1 − t ) β − F (cid:18) δ ; ζ ; − pt µ (1 − t ) µ (cid:19) dt, (1.5)where (cid:60) ( p ) ≥
0; min[ (cid:60) ( α ) , (cid:60) ( β ) , (cid:60) ( δ ) , (cid:60) ( ζ )] > (cid:60) ( µ ) > µ = 1 reduces to the generalized beta type functionas follows (see[3, p.4602]) B ( δ,ζ ) p ( α, β ) = B ( δ,ζ ;1) p ( α, β )= (cid:90) t α − (1 − t ) β − F (cid:18) δ ; ζ ; − pt (1 − t ) (cid:19) dt, (1.6)where (cid:60) ( p ) ≥
0; min[ (cid:60) ( α ) , (cid:60) ( β ) , (cid:60) ( δ ) , (cid:60) ( ζ )] > δ = ζ is given due to Choudhary et. al.[4] by B p ( α, β ) = B ( α, β ; p ) = B ( δ,δ ) p ( α, β )= (cid:90) t α − (1 − t ) β − exp (cid:18) − pt (1 − t ) (cid:19) dt, ( (cid:60) ( p ) ≥ . (1.7)When we choose p = 0, all the above extensions reduces to the classical beta function B ( x, y ), which is defined by OTE ON EXTENSIONS OF THE BETA FUNCTION 3 B ( α, β ) = (cid:90) t α − (1 − t ) β − dt, ( (cid:60) ( α ) > , (cid:60) ( β ) > . (1.8)It is clear to see the following relationship between the classical beta function B ( α, β )and its extensions: B ( α, β ) = B ( α, β ) = B ( δ,ζ )0 ( α, β )= B ( δ,ζ ;1)0 ( α, β ) = B ( δ,ζ ;1 , ( α, β ) . (1.9)In particular, Euler’s beta function B ( x, y ) has a close relationship to his gammafunction, B ( α, β ) = B ( β, α ) = Γ( α )Γ( β )Γ( α + β ) . (1.10)2. Extensions of Classical Beta Function do not convergent
We claim that the above extensions of classical beta function in equations (1.1), (1.5),(1.6), and (1.7) are not convergent. We prove it mathematically as well as numericallytested in Section 4.First we prove that the extension in equation (1.7) is not convergent as follows.
Claim 1. If (cid:60) ( α ) > , (cid:60) ( β ) > , (cid:60) ( p ) > , then the extensions of classical beta functionin equation (1.7) does not convergent.Proof. To prove our claim, we write the exponential function in series form, the equation(1.7) reduces to the following form B ( α, β ; p ) = ∞ (cid:88) n =0 ( − p ) n n ! (cid:90) ∞ t α − n − (1 − t ) β − n − dt, (2.1)further using the definition of the beta function in the above equation (2.1), we have B ( α, β ; p ) = ∞ (cid:88) n =0 ( − p ) n n ! B ( α − n, β − n )= B ( α, β ) + − p B ( α − , β −
1) + ... + ( − p ) n n ! B ( α − n, β − n ) + ... to infinity terms . (2.2)In the above equation (2.2), B ( α, β ; p ) is an power series involving B ( α − n, β − n )(where n = 0 , , , ... ). The series B ( α, β ; p ) will be convergent if B ( α − n, β − n ) isconvergent and B ( α − n, β − n ) will be convergent only if (cid:60) ( α − n ) > (cid:60) ( β − n ) > MEHAR CHAND
0. But this is not possible as α and β are finite and n → ∞ . The series B ( α, β ; p )contains many terms, which does not exist. Moreover B ( α, β ; p ) is the sum of termsinvolving B ( α − n, β − n ), which are not convergent, which implies that B ( α, β ; p ) is notconvergent. (cid:3) Example 1.
If we choose α = 5 , β = 7 , p = 3 , then from equation (2.2) , we have B (5 ,
7; 3) = ∞ (cid:88) n =0 ( − n n ! B (5 − n, − n ) , (2.3) The above series B (5 ,
7; 3) is the sum of terms involving the beta function B (5 − n, − n ) and B (5 − n, − n ) does not exit for n > . Therefor B (5 ,
7; 3) is not convergent.
In the following section, we introduce modified extension of classical beta function inequation (3.1). Further, the extension of classical beta function (1.7), Modified extensionof classical beta function (3.1) and the classical beta function are compared and testednumerically to test the convergence of extensions of the classical beta function.3.
Modified extension of beta function
In this section, we introduce modified extension of classical beta function (MECBF).Its convergence is proved mathematically, then numerical results are established andcompared the results with that of the classical beta function and extension of classicalbeta function.We introduce modified extension of classical beta function as follows B m ( α, β ) = B ( α, β ; m )= (cid:90) t α − (1 − t ) β − e mt (1 − t ) dt, (3.1)where (cid:60) ( α ) > , (cid:60) ( β ) > , m ∈ C ; | m | < M (where M is positive finite real number).In our investigation to test the convergence of the above extension of the classical betafunction the following definitions are required. Definition 1 (Ratio Test[9]) . Let (cid:80) ∞ n =1 p n be a series of positive terms, and supposethat p n +1 p n → a limit l as n → ∞ . Then (i) If l < the series (cid:80) p n is convergent, (ii)If l > , then the series is divergent. If l = 1 , the ratio test fails, and the question ofconvergence of (cid:80) p n must be investigated by some other methods. Definition 2 (Leibniz’s Test or Alternating Series Test[9]) . The series (cid:80) ∞ n =1 ( − n +1 a n (where a n > ) is convergent provided (i) a n is a decreasing sequence, (ii) a n → as n → ∞ . Theorem 1. If (cid:60) ( α ) > , (cid:60) ( β ) > , m ∈ C ; | m | < M (where M is positive finite realnumber), then the modified extension of the classical beta function in equation (3.1) isconvergent. OTE ON EXTENSIONS OF THE BETA FUNCTION 5
Proof.
We can write the above equation (3.1) as follows B ( α, β ; m ) = (cid:90) t α − (1 − t ) β − ∞ (cid:88) n =0 [ mt (1 − t )] n n ! dt = ∞ (cid:88) n =0 ( m ) n n ! (cid:90) t α + n − (1 − t ) β + n − dt, (3.2)further, using the definition of beta function, the above equation (3.2) reduces to B ( α, β ; m ) = ∞ (cid:88) n =0 ( m ) n n ! B ( α + n, β + n ) . (3.3)In the above equation, B ( α, β ; m ) is in series form involving B ( α + n, β + n ) (where n = 0 , , , ... ) and in each term of the series, B ( α + n, β + n ) is convergent since (cid:60) ( α + n ) > (cid:60) ( β + n ) > (cid:60) ( α ) , (cid:60) ( β ) >
0, which implies that each termof the series (3.3) exist.Now we shall prove that B ( α, β ; m ) is convergent. m may be greater than or less than0, so there are two cases as follows. Case 1. If m > , then to prove B ( α, β ; m ) is convergent. The equation (3.3) can be written as B ( α, β ; m ) = ∞ (cid:88) n =0 a n , where a n = ( m ) n n ! B ( α + n, β + n ) . (3.4)Further lim n →∞ a n a n +1 = ∞ > . (3.5)By ratio test, B ( α, β ; m ) is convergent for m > . Case 2. If m < , then to prove that the extension of classical beta function B ( α, β ; m ) is convergent. To prove this case, let m = − p (where p > B ( α, β ; m ) = ∞ (cid:88) n =0 ( − p ) n n ! B ( α + n, β + n ) , (3.6) MEHAR CHAND the above equation (3.6) can be written as B ( α, β ; m ) = − ∞ (cid:88) n =0 ( − n − b n , where b n = ( p ) n n ! B ( α + n, β + n ) . (3.7)The series (3.7) is an alternating series, therefor(1) a n > ∀ p > , (cid:60) ( α, β ) > a n − a n +1 = p n n ! B ( α + n, β + n ) × (cid:20) − ( α + n )( β + n )( α + β + 2 n )( α + β + 2 n + 1) (cid:21) > ⇒ a n is decreasing(3) lim n →∞ a n = 0 if p ≤ p n /n ! → n → ∞ only if p ≤ B ( α + n, β + n ) → n → ∞ )All the conditions of Leibniz’s test have been satisfied, therefore B ( α, β ; m ) is convergentfor 0 < p ≤ i.e. − ≤ m < (cid:3) Comparison of numerical values of extended beta function, modifiedextension of the classical beta function and classical beta function
In Table 1, the columns B1 , B2 are the values of the classical beta function B ( x, y );the columns EB1 , EB2 , EB3 , EB4 are the values of the extension of classical beta
Table 1.
Comparison of extensions of classical beta function x y B1 EB1 MB1 EB2 MB2 B2 EB3 MB3 EB4 MB4 0 0 NaN NaN NaN NaN Inf Inf NaN Inf NaN NaN 1 1 1.0000000 NaN 1.0016683 NaN 4.0000000 4.0000000 NaN 4.0035588 NaN 1.0000000 2 2 0.1666667 NaN 0.1670004 NaN 3.2000000 3.2000000 NaN 3.2021899 NaN 0.1666667 3 3 0.0333333 NaN 0.0334048 NaN 2.8444444 2.8444444 NaN 2.8459901 NaN 0.0333333 4 4 0.0071429 NaN 0.0071587 NaN 2.6256410 2.6256410 NaN 2.6268186 NaN 0.0071429 5 5 0.0015873 NaN 0.0015909 NaN 2.4711916 2.4711916 NaN 2.4721336 NaN 0.0015873 6 6 0.0003608 0.0003452 0.0003616 NaN 2.3535158 2.3535158 NaN 2.3542953 0.0003608 0.0003608 7 7 0.0000833 0.0000797 0.0000834 NaN 2.2593751 2.2593751 NaN 2.2600365 0.0000833 0.0000833 8 8 0.0000194 0.0000186 0.0000195 NaN 2.1814656 2.1814656 NaN 2.1820376 0.0000194 0.0000194 9 9 0.0000046 0.0000044 0.0000046 NaN 2.1153606 2.1153606 NaN 2.1158629 0.0000046 0.0000046 10 10 0.0000011 0.0000010 0.0000011 NaN 2.0581887 2.0581887 NaN 2.0586351 0.0000011 0.0000011 function; the columns MB1 , MB2 , MB3 , MB4 are values of modified extension of theclassical beta function, where B1 = B(x , y), EB1 = B(x , y; 0 . , y; 0 . , . , .
25; 0 . , .
25; 0 . , .
25; 0 . OTE ON EXTENSIONS OF THE BETA FUNCTION 7
Table 2.
Numerical Values of MECBF x y B(x,y,-2.0335) B(x,y,-2.0) B(x,y,-1.0) B(x,y,0.0) B(x,y,1.0) B(x,y,2.0) B(x,y,2.0335) 0.00 0.00 NaN NaN NaN NaN NaN NaN NaN 0.01 0.01 198.27177610 198.29548624 199.06411204 199.96757732 201.03554301 202.30476477 202.35127432 0.02 0.02 98.27460320 98.29781890 99.05065967 99.93608768 100.98330172 102.22848838 102.27412810 0.03 0.03 64.94418565 64.96691911 65.70435720 66.57217010 67.59910183 68.82077895 68.86556756 0.04 0.04 48.28052079 48.30278382 49.02519017 49.87579792 50.88290361 52.08158267 52.12553836 0.05 0.05 38.28360599 38.30541007 39.01314459 39.84694542 40.83466875 42.01084757 42.05398809 0.06 0.06 31.62010537 31.64146162 32.33487366 33.15225447 34.12102693 35.27519021 35.31753282 0.07 0.07 26.86144500 26.88236422 27.56179291 28.36312969 29.31337104 30.44599075 30.48755229 0.08 0.08 23.29333673 23.31382939 23.97960405 24.76526226 25.69738092 26.80891675 26.84971362 0.09 0.09 20.51895277 20.53902903 21.19146950 21.96180446 22.87619790 23.96709764 24.00714582 0.10 0.10 18.30019555 18.31986529 18.95928227 19.71463949 20.61169459 21.68239456 21.72170964 1.0 1.0 0.72102562 0.72477846 0.84887277 1.00000000 1.18459307 1.41068613 1.41909001 2.0 2.0 0.11166177 0.11238923 0.13665458 0.16666667 0.20385173 0.25000000 0.25172603 3.0 3.0 0.02163870 0.02179192 0.02692731 0.03333333 0.04133360 0.05133577 0.05171118 4.0 4.0 0.00455690 0.00459058 0.00572283 0.00714286 0.00892526 0.01116423 0.01124846 5.0 5.0 0.00100158 0.00100918 0.00126512 0.00158730 0.00199311 0.00250456 0.00252383 6.0 6.0 0.00022592 0.00022766 0.00028649 0.00036075 0.00045452 0.00057299 0.00057746 7.0 7.0 0.00005185 0.00005225 0.00006594 0.00008325 0.00010515 0.00013287 0.00013392 8.0 8.0 0.00001205 0.00001214 0.00001536 0.00001943 0.00002458 0.00003112 0.00003137 9.0 9.0 0.00000283 0.00000285 0.00000361 0.00000457 0.00000579 0.00000734 0.00000740 10.0 10.0 0.00000067 0.00000067 0.00000085 0.00000108 0.00000137 0.00000174 0.00000176
MB3 = B(x , .
25; 0 . , y; 0 . , y; 0 .
00) and x , y = 0 : 1 : 10.These values are computed by employing Matlab (R2012a). The values of the exten-sion of the classical beta function are computed by using the equation (2.2) and those ofmodified extension of the classical beta function from equation (3.3). Both the extensionin equations (2.2) and (3.3) are in series form involving beta functions B ( x − n, y − n )(exist for (cid:60) ( x − n > , y − n ) >
0) and B ( x + n, y + n ) (exist for (cid:60) ( x + n > , y + n ) > B ( x − n, y − n ) to find some numerical valuesand in the case of modified extension of classical beta function, sum of first 1000 termshave been taken for x, y = 0 : 1 : 10. The numerical results of the same are establishedin Table 1.From Table 1, it is easily observed that the values of extended beta function in columnsEB1 , EB4 does not exist for x, y = 0 : 1 : 5 (since (cid:60) ( x − n < , y − n ) < x, y = 6 : 1 : 10 (since (cid:60) ( x − n ) > , (cid:60) ( y − n ) < MEHAR CHAND
EB2 , EB3 the values of extended beta function do not exist since in these cases we choose y = 0 .
25 therefor (cid:60) ( y − n ) becomes less than zero, which implies if 0 < x, y < n = 1000) theextended beta function will not converge for any value of x, y ∈ [0 , x and y , sinceit is series having infinite terms and (cid:60) ( x ) , (cid:60) ( y ) can not be infinite.In the case of modified extension of classical beta function, the columns B1 , MB4 areidentical, which implies that if we choose m = 0, we have the values of classical betafunction from the modified extension of the classical beta function. Column B2 , MB2are also identical since m = 0 and at the fix value of y = 0 .
25. If we compare the valuesof B1 with MB1 and those of B2 with MB3, we can see the impact of the value of m onthe classical beta function. Note 1.
It is also noted that if p = 0 , the extension of classical beta function should bereduced to the classical beta function i.e. B ( α, β ; 0) = B ( α, β ) . But the same not depictin the graph cited in [4, Fig. 1, p. 31] . The graph depicts the values of extension of theclassical beta function B (1 , .
25; 0) = B (1 , .
25) = 13 . when p = 0 , x = 1 , y = 0 . and B (10 , .
25; 0) = B (10 , .
25) = 2 . (approximately) when p = 0 , x = 10 , y = 0 . ,which are not the correct values of the classical beta function. The correct values of theclassical beta function at these point are B (1 , .
25) = 4 and B (10 , .
25) = 2 . . Italso can be easily observed from the figure that when p (cid:54) = 0 the graph of extended betafunction totally different from that of the classical beta function. i.e. the behavior andnature of the curves differ from that of classical beta function. Note 2.
From the above discussion, it is very clear that the extension of classical betafunction in equation (1.7) do not convergent. The extensions in equations (1.1) , (1.5) and (1.6) are further extension of the extended beta function (1.7) , therefor all theseextensions also can not be convergent. Note 3.
Due to the lack of the convergence of the extension of the classical beta function,all the associated results established by Chaudhary et al. [4] do not convergent. Numerical results and their interpretation of modified extension ofthe classical beta function
The numerical results of modified extension of classical beta function (MECBF) havebeen calculated in this section by employing the Matlab. We choose the values of variable x, y and parameter m as x, y ∈ [0 ,
10] and m ∈ [ − . , . B ( x, y ; m )does not exist at x = y = 0 and it is also tested that B ( x, y ; m ) does not exist for m < − . m > . B ( x, y ; m ) → ∞ as x, y → B ( x, y ; m ) → x, y → ∞ , which implies that the behavior of modifiedextension of classical beta function is the same as that of classical beta function, whichcan be observed from Table 2 and Figure 1. Also B ( x, y ; 0) represents the values ofclassical beta function B ( x, y ). OTE ON EXTENSIONS OF THE BETA FUNCTION 9 B ( x , y , m ) B(x,y,-2)B(x,y,-1)B(x,y,0)B(x,y,1)B(x,y,2)
Figure 1.
Graphical presentation of of MECBF for y=1.6.
Geometrical interpretation of modified extension of the classicalbeta function If β = 1 , m ∈ [ − ,
2] and x ∈ [0 , B ( x, y ; m ) inFigure 1 is decreasing. In figure 2, we choose m = 2 and x, y ∈ [0 , x, y → B ( x, y ; m ) → ∞ and B ( x, y ; m ) → α, β → ∞ . We also check theeffect of m on the modified extension of classical beta function. For this purpose, we fixthe values of x and y as shown in figure 3, then we plot the graph which depicts that B ( x, y ; m ) is an increasing function as the values of m increase. It is very clear fromFigure 1 that for the graph of classical beta function and modified extension of classicalbeta function remains concave upward (or convex downward) for different values of α, β and m . The value of m does not effect the nature of classical beta function, the maineffect of the value of m is that it just push the curve up or drag down the curve fromthe curve of the classical beta function as shown in the Figure 1 the same behavior canbe observed from the Table 2.From the above proof of radius of convergence of series and further more numericallyinvestigation of the power series in Table 1, we find that the interval of convergence ofthe series is ( − . , . B ( x, y ; m ) is convergent for | m | < M where M is positive real number slightly greater than 2.0335. B ( x , y , m ) Figure 2.
3D presentation of of MECBF.
Note 4.
From the above discussion, it is easy to conclude that the value of (cid:60) ( m ) lies inthe interval [ − . , . i.e. − . ≤ (cid:60) ( m ) ≤ . . Note 5.
In the sequel of this paper, | m | < M represents the circle of convergence and M is the radius of convergence of equation (3.1) , where M is sightly grater than from . . Integral representation of the MECBF function
The integral representation of the modified extended beta function is important tocheck both the extension is natural and simple and for use later. It is also importantto investigate the relationship between the classical beta function and the modified ex-tension of the classical beta function. In this connection we first provide a relationshipbetween them. The following integral formula is useful for further investigation [7] (cid:90) ∞ x m exp( − βx n ) dx = Γ( γ ) nβ γ , (cid:18) γ = m + 1 n (cid:19) . (7.1) Theorem 2 (Relation between modified extension of the classical beta function and theclassical beta function) . If (cid:60) ( α + s ) > , (cid:60) ( β + s ) > , m ∈ C ; | m | < M (where M is positive finite real number slightly greater than 2.0335), then we have the following OTE ON EXTENSIONS OF THE BETA FUNCTION 11 -2 -1.5 -1 -0.5 0 0.5 1 1.5 20.040.060.080.10.120.140.160.180.20.22 m B ( x , y , m ) B(2.1,2.1,m)B(2.2,2.2,m)B(2.3,2.3,m)B(2.4,2.4,m)B(2.5,2.5,m)
Figure 3.
Graph of MECBF for fixed value of x, y and m = − . relation (cid:90) ∞ m s − B ( α, β ; m ) dm = ( − s Γ( s ) B ( α + s, β + s ) . (7.2) Proof.
Multiplying both sides of equation (3.1) by m s − , then integrate w.r.t. m from m = 0 to m = ∞ , we have (cid:90) ∞ m s − B ( α, β ; m ) dm = (cid:90) ∞ m s − (cid:20)(cid:90) t α − (1 − t ) β − exp( mt (1 − t )) dt (cid:21) dm, (7.3)interchanging the order of integration, the above equation (7.3), reduces to (cid:90) ∞ m s − B ( α, β ; m ) dm = (cid:90) t α − (1 − t ) β − (cid:20)(cid:90) ∞ m s − exp( mt (1 − t )) dm (cid:21) dt, (7.4)further using the formula given in equation (7.1), after simplification the above equa-tion (7.4) reduces to (cid:90) ∞ m s − B ( α, β ; m ) dm = ( − s Γ( s ) (cid:90) t α + s − (1 − t ) β + s − dt, (7.5)using the definition of classical beta function, we have the required result. (cid:3) Remark 1.
By setting s = 1 , the result in equation (7.2) reduces to (cid:90) ∞ B ( α, β ; m ) dm = − B ( α + 1 , β + 1) , (cid:60) ( α ) > − , (cid:60) ( β ) > − , (7.6) gives the interesting relation between classical beta function and modified extended betafunction. Remark 2.
All the derivatives of the modified extension of classical beta function (MECBF)with respect to the parameter m can be expressed in terms of the function as ∂ n ∂ n m B ( α, β ; m ) = B ( α + n, β + n ) , (cid:60) ( α + n ) > , (cid:60) ( β + n ) > . (7.7) Theorem 3 (Integral representations of the modified extension of the classical betafunction) . If (cid:60) ( x ) > , (cid:60) ( y ) > , m ∈ C ; | m | < M (where M is positive finite realnumber slightly greater than 2.0335), then we have the following relation B ( α, β ; m ) = 2 (cid:90) π cos α − θ sin β − θ × exp( m cos θ sin θ ) dθ, (7.8) B ( α, β ; m ) = (cid:90) ∞ u α − (1 + u ) α + β exp (cid:0) mu/ (1 + u ) (cid:1) du, (7.9) B ( α, β ; m ) = 12 (cid:90) ∞ u α − + u β − (1 + u ) α + β × exp (cid:0) mu/ (1 + u ) (cid:1) du, (7.10) B ( α, β ; m ) = ( c − a ) − α − β (cid:90) ca ( u − a ) α − ( c − u ) β − × exp (cid:18) m ( u − a )( c − u )( c − a ) (cid:19) du, (7.11) B ( α, β ; m ) = 2 − α − β (cid:90) − (1 + t ) α − (1 − t ) β − × exp (cid:0) m (1 − t ) / (cid:1) du, (7.12) OTE ON EXTENSIONS OF THE BETA FUNCTION 13 B ( α, β ; m ) = 2 − α − β × (cid:90) ∞−∞ exp (cid:20) ( α − β ) x + m x (cid:21) dx (cosh x ) α + β , (7.13) B ( α, β ; m ) = 2 − α − β (cid:90) ∞ cosh(( α − β ) x ) × exp (cid:20) m x (cid:21) dx (cosh x ) α + β , (7.14) B ( α, β ; m ) = 2 − α − β × (cid:90) ∞−∞ exp (cid:20)
12 ( α − β ) x + m x (cid:21) dx (cosh x/ α + β , (7.15) B ( α, β ; m ) = 2 − α − β (cid:90) ∞ cosh(( α − β ) x/ × exp (cid:104) m x (cid:105) dx (cosh x/ α + β . (7.16) Proof.
The result (7.8) can be easily obtained by setting t = cos θ , to prove (7.9) choose t = u/ (1 + u ), (7.10) can be easily obtained by applying the symmetric property inequation (7.9) then adding new one and (7.9), the result in equation (7.11) is obtainedby taking t = ( u − a ) / ( c − a ), setting a = − , c = 1 in equation (7.11) gives the resultin equation (7.12) and to prove the result in equation (7.13) put u = tanh x in (7.12).The results in equation (7.14), (7.15) and (7.16) can be easily obtained from the result(7.13). (cid:3) Remark 3 (Useful inequalities) . If (cid:60) ( α ) > , (cid:60) ( β ) > , then we have the followinginequality | B ( α, β ; m ) | ≤ . B ( α, β ) (7.17) follows from the integral representation (7.9) . Since the function exp( mu/ (1+ u ) ) attainsits maximum value . at u = 1 and m = 2 . . The equality is verified with thehelp of numerical results by using Matlab. Properties of the modified extension of the classical beta function
Theorem 4 (Functional Relation) . If (cid:60) ( α ) > , (cid:60) ( β ) > , m ∈ C ; | m | < M (where M is positive finite real number slightly greater than 2.0335), then we have the followingrelation B ( α, β + 1; m ) + B ( α + 1 , β ; m ) = B ( α, β ; m ) . (8.1) Proof.
LHS of the above equation (8.1) is equal to (cid:90) { t α − (1 − t ) β + t α (1 − t ) β − } e mt (1 − t ) dt (8.2)after simplification the above equation (8.2) reduced to (cid:90) { t α − (1 − t ) β − } e mt (1 − t ) dt = B ( α, β ; m ) . (8.3) (cid:3) If we choose m = 0, we get the usual relation for the beta function from (8.1). Theorem 5 (Symmetry) . If (cid:60) ( α ) > , (cid:60) ( β ) > , m ∈ C ; | m | < M (where M is positivefinite real number slightly greater than 2.0335), then we have the following relation B ( α, β ; m ) = B ( β, α ; m ) . (8.4) Proof.
From equation (3.3) we have B ( α, β ; m ) = ∞ (cid:88) n =0 ( m ) n n ! B ( α + n, β + n ) (8.5)Since usual beta function is symmetric i.e. B ( α, β ) = B ( β, α ). Using this property inthe right hand side of the equation (8.5), then we have B ( α, β ; m ) = ∞ (cid:88) n =0 ( m ) n n ! B ( β + n, α + n ) = B ( β, α ; m ) . (8.6) (cid:3) Theorem 6 (First Summation Relation) . If (cid:60) ( α ) > , (cid:60) (1 − β ) > , m ∈ C ; | m | < M (where M is positive finite real number slightly greater than 2.0335), then we have thefollowing relation B ( α, − β ; m ) = ∞ (cid:88) n =0 ( β ) n n ! B ( α + n, m ) . (8.7) Proof.
The LHS of the equation (8.7) can be written as B ( α, − β ; m ) = (cid:90) t α − (1 − t ) − β exp( mt (1 − t )) dt, (8.8)using the binomial series expansion (1 − t ) − β = ∞ (cid:88) n =0 ( β ) n n ! t n in the above equation (8.8)and then interchanging the order of summation and integration, the above result (8.8)reduced to the following form OTE ON EXTENSIONS OF THE BETA FUNCTION 15 B ( α, − β ; m )= ∞ (cid:88) n =0 ( β ) n n ! (cid:90) t α + n − exp( mt (1 − t )) dt (8.9) ⇒ B ( α, − β ; m ) = ∞ (cid:88) n =0 ( β ) n n ! B ( α + n, m ) . (8.10) (cid:3) Theorem 7 (Second Summation Relation) . If (cid:60) ( α ) > , (cid:60) ( β ) > , m ∈ C ; | m | < M (where M is positive finite real number slightly greater than 2.0335), then we have thefollowing relation B ( α, β ; m ) = ∞ (cid:88) n =0 B ( α + n, β + 1; m ) . (8.11) Proof.
The LHS of the equation (8.7) can be written as B ( α, β ; m ) = (cid:90) t α − (1 − t ) β − exp( mt (1 − t )) dt, (8.12)using the binomial series expansion (1 − t ) β − = (1 − t ) α ∞ (cid:88) n =0 t n , ( | t | <
1) and inter-changing the order of summation and integration, the above equation (8.12) reducesto B ( α, β ; m ) = ∞ (cid:88) n =0 (cid:90) t α + n − (1 − t ) β exp( mt (1 − t )) dt (8.13) ⇒ B ( α, β ; m ) = ∞ (cid:88) n =0 B ( α + n, β + 1; m ) . (8.14) (cid:3) Theorem 8 (Separation) . If (cid:60) ( α ) > , (cid:60) ( β ) > , m ∈ C ; | m | < M (where M is positivefinite real number slightly greater than 2.0335), then B ( α, β ; m ) can can be separated intoreal and imaginary part of m as follows B ( α, β ; r cos θ )= (cid:90) t α − (1 − t ) β − exp( rt (1 − t ) cos θ ) dt, (8.15) B ( α, β ; r sin θ )= (cid:90) t α − (1 − t ) β − exp( rt (1 − t ) sin θ ) dt, (8.16) where r = (cid:112) x + y = | m | < M and θ = tan − y/x. Proof.
Since m ∈ C , so let m = x + iy where x, y ∈ R and also let x + iy = r cos θ + ir sin θ ⇒ r = (cid:112) x + y and θ = tan − y/x , then from equation (3.1), we have B ( α, β ; x + iy )= (cid:90) t α − (1 − t ) β − exp( re iθ t (1 − t )) dt, (8.17)after simplification the above equation (8.17) reduces to B ( α, β ; r cos θ + ir sin θ )= (cid:90) t α − (1 − t ) β − exp( rt (1 − t ) cos θ ) dt + i (cid:90) t α − (1 − t ) β − exp( rt (1 − t ) sin θ ) dt, (8.18)equating the real and imaginary parts of m only, we have the required results. (cid:3) The modified extended beta distribution
It is expected that there will be many applications of the modified extension of theclassical beta function, like there were of the generalized gamma function. One applica-tion that springs to mind is to Statistics. For example, the conventional beta distributioncan be extended, by using our modified extension of the classical beta function, to vari-ables p and q with an infinite range. It appears that such an extension may be desirablefor the project evaluation and review technique used in some special cases.We define the extended beta distribution by f ( t ) = t p − (1 − t ) q − exp( mt (1 − t )) B ( p, q ; m ) , < t < , otherwise . (9.1)A random variable X with probability density function (pdf) given in equation (9.1)will be said to have the extended beta distribution with parameters p and q , −∞
B x ( p, q ; 0) = B x ( p, q ), where B x ( p, q ) is the incomplete betafunction [7] defined as B x ( p, q ) = B x ( p, q ; 0) = x p p F ( p, − q ; p = 1; x ) . (9.8)It is to be noted that the problem of expressing B x ( p, q, m ) in terms of other specialfunctions remains open. Presumably, this distribution should be useful in extendingthe statistical results for strictly positive variables to deal with variables that can takearbitrarily large negative values as well.10. conclusion The (Euler’s) classical beta function B ( x ; y ) play an important role in various fields inthe mathematical, physical, engineering and statistical sciences. The extensions of theclassical beta function is also important for further investigation in the respective areas,provided these extensions of beta function should be convergent for each parameterinvolving. The modified extension of classical beta function is convergent and if wechoose m = 0, the classical beta function is obtained. It is very clear from Figure 1 that for the graph of classical beta function and modified extension of classical beta functionremains concave upward (or convex downward) for different values of α, β and m . Thevalue of m does not effect the nature of classical beta function, the main effect of thevalue of m is that it just push the curve up or drag down the curve from the curve ofthe classical beta function as shown in the Figure 1 the same behavior can be observedfrom the Table 2. Both modified extension of classical beta function and classical betafunction i.e. B ( α, β ; m ) & B ( α, β ) → x, y → ∞ ; B ( α, β ; m ) & B ( α, β ) → ∞ as x, y → B ( α, β ; m ) & B ( α, β ) does not exist at x = y = 0. References [1] H. M. Srivastava, A. Cetinkaya, and ˙I. O. Kiymaz, A certain generalized Pochhammer symbol andits applications to hypergeometric functions, Appl. Math. Comput. 226 (2014), 484-491.[2] R. K. Parmar, A new generalization of Gamma, Beta, hypergeometric and confluent hypergeometricfunctions, Matematiche (Catania) 69 (2013), no. 2, 33-52.[3] E. ¨ O zergin, M. A. ¨ O zarslan, and A. Altin, Extension of gamma, beta and hypergeometric functions,J. Comput. Appl. Math. 235 (2011), no. 16, 4601-4610.[4] M. A. Chaudhry, A. Qadir, M. Rafique and S. M. Zubair, Extension of Euler’s beta function, J.Comput. Appl. Math. 78 (1997), no. 1, 19-32.[5] E. D. Rainville, Special Functions, Macmillan Company, New York, 1960; Reprinted by ChelseaPublishing Company, Bronx, New York, 1971.[6] H. M. Srivastava and J. Choi, Zeta and q-Zeta Functions and Associated Series and Integrals,Elsevier Science Publishers, Amsterdam, London and New York, 2012.[7] I.S. Gradshteyn and I.M. Ryzhik, Table of Integrals, Series, and Products, Seventh edition, ElsevierAcadmic Press publication.[8] V.K. Rohatgi. An Introduction to Probability Theory and Mathematical Statistics (Wiley, NewYork, 1976).[9] J.A. Green. Sequence and Series, Routledge & Kegan Paul Ltd., Broudway House, Carter Lane,London (1958). Department of Applied Sciences, Guru Kashi University, Bathinda-1513002 (India)
E-mail address ::