Professional Documents
Culture Documents
Jca 10 06
Jca 10 06
Classical
A nalysis
Volume 10, Number 1 (2017), 59–66 doi:10.7153/jca-10-06
Abstract. All beta-type functions, i.e. the functions B f : (0,∞)2 → (0,∞) of the form
f (x) f (y)
B f (x,y) =
f (x + y)
for some f : (0,∞) → (0,∞) , which are p -homogeneous, are determined. Applying this result,
we show that a beta-type function is a homogeneous mean iff it is the harmonic one. A refor-
mulation of a result due to Heuvers in terms of a Cauchy difference and the harmonic mean is
given.
1. Introduction
For a given f : (0, ∞) → (0, ∞) , the function B f : (0, ∞)2 → (0, ∞) defined by
f (x) f (y)
B f (x, y) = , x, y > 0,
f (x + y)
is called the beta-type function, and f is called its generator ([7]). The notion the
beta-type function arises from the well-known relation between the Euler Beta function
B : (0, ∞)2 → (0, ∞) and the Euler Gamma function Γ : (0, ∞) → (0, ∞)
Γ (x) Γ (y)
B (x, y) = , x, y > 0.
Γ (x + y)
2xy
H (x, y) = , x, y > 0.
x+y
Mathematics subject classification (2010): 33B15, 26B25, 39B22.
Keywords and phrases: Beta function, Gamma function, beta-type function, pre-mean, mean, homo-
geneity, functional equation.
c , Zagreb 59
Paper JCA-10-06
60 M ARTIN H IMMEL AND JANUSZ M ATKOWSKI
A related companion of the beta-type function is the Cauchy difference Cg : (0, ∞)2 →
R defined by
Cg (x, y) = g (x + y) − g (x) − g (y)
for a function g : (0, ∞) → R. The relationship
B f = exp ◦ −Clog ◦ f
2. Main result
and p
xy
B f (x, y) = b , x, y > 0.
x+y
f (tx)
ϕt (x) := , x > 0.
t p f (x)
stating that ϕt is an exponential function. Hence (see, for instance, [1] p. 39), for every
t > 0, there exists a unique additive function αt : R → R such that
Since the right hand side is symmetric in x and t, so is the left hand side; thus
eαx (t) x p f (t) = f (xt) = f (tx) = eαt (x)t p f (x) , x,t > 0.
α1 (x) = 0, x > 0.
and, consequently,
f (x) = f (1) x p eαx (1) , x > 0.
Putting, for convenience, λ : (0, ∞) → R,
we have
f (x) = f (1) x p eλ (x) , x > 0. (2.3)
Inserting this into (2.1), we obtain,
f (1) (tx) p eλ (tx) f (1) (ty) p eλ (ty) f (1) x p eλ (x) f (1) y p eλ (y)
p λ (t(x+y)) = tp , x, y,t > 0,
f (1) [t (x + y)] e f (1) (x + y) p eλ (x+y)
that reduces to
whence
is additive. From (2.3) and the assumed regularity of f we get that ω is continuous or
Lebesgue measurable. Thus, ω , being additive and continuous or measurable, is of the
form ([6], p. 129, see also [1])
ω (x) = ω (1) x, x > 0,
and hence, by (2.4),
λ (tx) − λ (x) = (λ (t) − λ (1)) x, x,t > 0,
whence
λ (tx) = λ (x) + (λ (t) − λ (1)) x, x,t > 0.
The symmetry in t and x of the left hand side implies that
λ (x) + (λ (t) − λ (1)) x = λ (t) + (λ (x) − λ (1))t, x,t > 0,
whence
λ (x) (1 − t) + λ (1)t = λ (t) (1 − x) + λ (1) x, x,t > 0.
Subtracting λ (1) from both sides yields
λ (x) (1 − t) + λ (1)t − λ (1) = λ (t) (1 − x) + λ (1) x − λ (1) , x,t > 0,
whence
λ (x) (1 − t) − λ (1)(1 − t) = λ (t) (1 − x) − λ (1) (1 − x), x,t > 0,
and, consequently,
λ (x) − λ (1) λ (t) − λ (1)
= , x,t > 0, x = 1 = y.
1−x 1−t
It follows that there exists c ∈ R such that
λ (x) − λ (1)
= −c, x > 0, x = 1,
1−x
whence,
λ (x) = c (x − 1) + λ (1) , x > 0,
and we obtain
λ (x) = cx + d, x > 0,
where d := λ (1) − c. Inserting this function λ into (2.3), we obtain
f (x) = f (1) ed x p (ec )x , x > 0,
whence, setting
a := ec , b := f (1)ed ,
we get
f (x) = bx p ax , x > 0,
and
xy p
B f (x, y) = b , x, y > 0,
x+y
which proves (ii). The implication (ii) =⇒ (i) is obvious.
H OMOGENEOUS B ETA - TYPE FUNCTIONS 63
3. Applications to pre-means
M (x, x) = x, x ∈ I;
M is called a pre-mean in I ([8]), if it is reflexive and M I 2 ⊆ I;
M is called a mean in I, if
R EMARK 2. Obviously, every mean is a pre-mean, but, in general, not vice versa.
Indeed, the function M : (0, ∞)2 → (0, ∞) defined by
2x2 + y2
M (x, y) =
x + 2y
(iii) the beta-type function coincides with the harmonic mean, i.e.
2xy
B f (x, y) = , x, y > 0.
x+y
Proof. Assume (i). By Theorem 1 and remark 3, its generator f is of the form
for some a, b ∈ (0, ∞). Since B f is reflexive, that is B f (x, x) = x for all x ∈ (0, ∞) .
Substituting here x = 2 and using Theorem 1 (ii), yields
f (2) f (2) b · 2 · 2
2 = B f (2, 2) = = = b,
f (2 + 2) 2+2
64 M ARTIN H IMMEL AND JANUSZ M ATKOWSKI
For another result connecting harmonic mean and the Euler Gamma function see
[3].
and p
xy
Cg (x, y) = log − d, x, y > 0.
x+y
since, using the definition of beta-type function, we have, for all x, y > 0,
eg(tx)+g(ty)−g(t(x+y)) = t −peg(x)+g(y)−g(x+y) .
REFERENCES
[1] J. A CZ ÉL, Lectures on Functional Equations and Their Applications, Academic Press, New York and
London, 1966.
[2] J. A CZ ÉL , J. G. D HOMBRES , Functional Equations in Several Variables, Encyclopedia of Mathematics
and its Applications, Cambridge University Press, 1989.
[3] H. A LZER, A harmonic mean inequality for the Gamma function, J. Comput. Appl. Math. 87 (1997),
195–198.
[4] K. J. H EUVERS , Another logarithmic functional equation, Aeq. Math., 58 (1999), 260–264.
[5] P L . K ANNAPPAN, Functional Equations and Inequalities with Applications, Springer Monographs in
Mathematics, Springer, New York, 2009.
[6] M. K UCZMA , A. G IL ÁNYI , An Introduction to the Theory of Functional Equations and Inequalities,
2009, Birkhäuser Verlag AG, Basel – Boston – Berlin.
66 M ARTIN H IMMEL AND JANUSZ M ATKOWSKI
[7] M. H IMMEL , J. M ATKOWSKI , Directional convexity and characterizations of Beta and Gamma func-
tions (submitted).
[8] J. M ATKOWSKI , Convergence of iterates of pre-mean type mappings, ESAIM: Proceedings and Surveys,
ECIT 2012, Witold Jarczyk, Daniele Fournier-Prunaret, João Manuel Gonçalves Cabral, November
2014, Vol. 46, 196–228.