Difference of two numbers divided by the logarithm of their quotient
Three-dimensional plot showing the values of the logarithmic mean.
In
.
This calculation is applicable in
engineering problems involving
heat and
mass transfer .
Definition
The logarithmic mean is defined as:
M
lm
(
x
,
y
)
=
lim
(
ξ
,
η
)
→
(
x
,
y
)
η
−
ξ
ln
(
η
)
−
ln
(
ξ
)
=
{
x
if
x
=
y
,
y
−
x
ln
y
−
ln
x
otherwise,
{\displaystyle {\begin{aligned}M_{\text{lm}}(x,y)&=\lim _{(\xi ,\eta )\to (x,y)}{\frac {\eta -\xi }{\ln(\eta )-\ln(\xi )}}\\[6pt]&={\begin{cases}x&{\text{if }}x=y,\\[2pt]{\dfrac {y-x}{\ln y-\ln x}}&{\text{otherwise,}}\end{cases}}\end{aligned}}}
for the positive numbers x, y .
Inequalities
The logarithmic mean of two numbers is smaller than the arithmetic mean and the generalized mean with exponent greater than 1. However, it is larger than the geometric mean and the harmonic mean , respectively. The inequalities are strict unless both numbers are equal.
2
1
x
+
1
y
≤
x
y
≤
x
−
y
ln
x
−
ln
y
≤
x
+
y
2
≤
(
x
2
+
y
2
2
)
1
/
2
for all
x
>
0
and
y
>
0.
{\displaystyle {\frac {2}{\displaystyle {\frac {1}{x}}+{\frac {1}{y}}}}\leq {\sqrt {xy}}\leq {\frac {x-y}{\ln x-\ln y}}\leq {\frac {x+y}{2}}\leq \left({\frac {x^{2}+y^{2}}{2}}\right)^{1/2}\qquad {\text{ for all }}x>0{\text{ and }}y>0.}
[1] [2] [3] [4]
Toyesh Prakash Sharma generalizes the arithmetic logarithmic geometric mean inequality for any
n belongs to the whole number as
x
y
(
ln
x
y
)
n
−
1
(
n
+
ln
x
y
)
≤
x
(
ln
x
)
n
−
y
(
ln
y
)
n
ln
x
−
ln
y
≤
x
(
ln
x
)
n
−
1
(
n
+
ln
x
)
+
y
(
ln
y
)
n
−
1
(
n
+
ln
y
)
2
{\displaystyle {\sqrt {xy}}\ \left(\ln {\sqrt {xy}}\right)^{n-1}\left(n+\ln {\sqrt {xy}}\right)\leq {\frac {x(\ln x)^{n}-y(\ln y)^{n}}{\ln x-\ln y}}\leq {\frac {x(\ln x)^{n-1}(n+\ln x)+y(\ln y)^{n-1}(n+\ln y)}{2}}}
Now, for n = 0 :
x
y
(
ln
x
y
)
−
1
ln
x
y
≤
x
−
y
ln
x
−
ln
y
≤
x
(
ln
x
)
−
1
ln
x
+
y
(
ln
y
)
−
1
ln
y
2
x
y
≤
x
−
y
ln
x
−
ln
y
≤
x
+
y
2
{\displaystyle {\begin{array}{ccccc}{\sqrt {xy}}\left(\ln {\sqrt {xy}}\right)^{-1}\ln {\sqrt {xy}}&\leq &{\dfrac {x-y}{\ln x-\ln y}}&\leq &{\dfrac {x(\ln x)^{-1}\ln x+y(\ln y)^{-1}\ln y}{2}}\\[4pt]{\sqrt {xy}}&\leq &{\dfrac {x-y}{\ln x-\ln y}}&\leq &{\dfrac {x+y}{2}}\end{array}}}
This is the arithmetic logarithmic geometric mean inequality. similarly, one can also obtain results by putting different values of n as below
For n = 1 :
x
y
(
1
+
ln
x
y
)
≤
x
ln
x
−
y
ln
y
ln
x
−
ln
y
≤
x
(
1
+
ln
x
)
+
y
(
1
+
ln
y
)
2
{\displaystyle {\sqrt {xy}}\left(1+\ln {\sqrt {xy}}\right)\leq {\frac {x\ln x-y\ln y}{\ln x-\ln y}}\leq {\frac {x(1+\ln x)+y(1+\ln y)}{2}}}
for the proof go through the bibliography.
Derivation
Mean value theorem of differential calculus
From the
there exists a value
ξ in the
interval between
x and
y where the derivative
f ′ equals the slope of the
secant line :
∃
ξ
∈
(
x
,
y
)
:
f
′
(
ξ
)
=
f
(
x
)
−
f
(
y
)
x
−
y
{\displaystyle \exists \xi \in (x,y):\ f'(\xi )={\frac {f(x)-f(y)}{x-y}}}
The logarithmic mean is obtained as the value of ξ by substituting ln for f and similarly for its corresponding derivative :
1
ξ
=
ln
x
−
ln
y
x
−
y
{\displaystyle {\frac {1}{\xi }}={\frac {\ln x-\ln y}{x-y}}}
and solving for ξ :
ξ
=
x
−
y
ln
x
−
ln
y
{\displaystyle \xi ={\frac {x-y}{\ln x-\ln y}}}
Integration
The logarithmic mean can also be interpreted as the area under an exponential curve .
L
(
x
,
y
)
=
∫
0
1
x
1
−
t
y
t
d
t
=
∫
0
1
(
y
x
)
t
x
d
t
=
x
∫
0
1
(
y
x
)
t
d
t
=
x
ln
y
x
(
y
x
)
t
|
t
=
0
1
=
x
ln
y
x
(
y
x
−
1
)
=
y
−
x
ln
y
x
=
y
−
x
ln
y
−
ln
x
{\displaystyle {\begin{aligned}L(x,y)={}&\int _{0}^{1}x^{1-t}y^{t}\ \mathrm {d} t={}\int _{0}^{1}\left({\frac {y}{x}}\right)^{t}x\ \mathrm {d} t={}x\int _{0}^{1}\left({\frac {y}{x}}\right)^{t}\mathrm {d} t\\[3pt]={}&\left.{\frac {x}{\ln {\frac {y}{x}}}}\left({\frac {y}{x}}\right)^{t}\right|_{t=0}^{1}={}{\frac {x}{\ln {\frac {y}{x}}}}\left({\frac {y}{x}}-1\right)={}{\frac {y-x}{\ln {\frac {y}{x}}}}\\[3pt]={}&{\frac {y-x}{\ln y-\ln x}}\end{aligned}}}
The area interpretation allows the easy derivation of some basic properties of the logarithmic mean. Since the exponential function is monotonic , the integral over an interval of length 1 is bounded by x and y . The homogeneity of the integral operator is transferred to the mean operator, that is
L
(
c
x
,
c
y
)
=
c
L
(
x
,
y
)
{\displaystyle L(cx,cy)=cL(x,y)}
.
Two other useful integral representations are
1
L
(
x
,
y
)
=
∫
0
1
d
t
t
x
+
(
1
−
t
)
y
{\displaystyle {1 \over L(x,y)}=\int _{0}^{1}{\operatorname {d} \!t \over tx+(1-t)y}}
and
1
L
(
x
,
y
)
=
∫
0
∞
d
t
(
t
+
x
)
(
t
+
y
)
.
{\displaystyle {1 \over L(x,y)}=\int _{0}^{\infty }{\operatorname {d} \!t \over (t+x)\,(t+y)}.}
Generalization
Mean value theorem of differential calculus
One can generalize the mean to n + 1 variables by considering the mean value theorem for divided differences for the n -th derivative of the logarithm.
We obtain
L
MV
(
x
0
,
…
,
x
n
)
=
(
−
1
)
n
+
1
n
ln
(
[
x
0
,
…
,
x
n
]
)
−
n
{\displaystyle L_{\text{MV}}(x_{0},\,\dots ,\,x_{n})={\sqrt[{-n}]{(-1)^{n+1}n\ln \left(\left[x_{0},\,\dots ,\,x_{n}\right]\right)}}}
where
ln
(
[
x
0
,
…
,
x
n
]
)
{\displaystyle \ln \left(\left[x_{0},\,\dots ,\,x_{n}\right]\right)}
denotes a
divided difference
of the logarithm.
For n = 2 this leads to
L
MV
(
x
,
y
,
z
)
=
(
x
−
y
)
(
y
−
z
)
(
z
−
x
)
2
(
(
y
−
z
)
ln
x
+
(
z
−
x
)
ln
y
+
(
x
−
y
)
ln
z
)
.
{\displaystyle L_{\text{MV}}(x,y,z)={\sqrt {\frac {(x-y)(y-z)(z-x)}{2{\bigl (}(y-z)\ln x+(z-x)\ln y+(x-y)\ln z{\bigr )}}}}.}
Integral
The integral interpretation can also be generalized to more variables, but it leads to a different result. Given the simplex
S
{\textstyle S}
with
S
=
{
(
α
0
,
…
,
α
n
)
:
(
α
0
+
⋯
+
α
n
=
1
)
∧
(
α
0
≥
0
)
∧
⋯
∧
(
α
n
≥
0
)
}
{\displaystyle S=\{\left(\alpha _{0},\,\dots ,\,\alpha _{n}\right):\left(\alpha _{0}+\dots +\alpha _{n}=1\right)\land \left(\alpha _{0}\geq 0\right)\land \dots \land \left(\alpha _{n}\geq 0\right)\}}
and an appropriate measure
d
α
{\textstyle \mathrm {d} \alpha }
which assigns the simplex a volume of 1, we obtain
L
I
(
x
0
,
…
,
x
n
)
=
∫
S
x
0
α
0
⋅
⋯
⋅
x
n
α
n
d
α
{\displaystyle L_{\text{I}}\left(x_{0},\,\dots ,\,x_{n}\right)=\int _{S}x_{0}^{\alpha _{0}}\cdot \,\cdots \,\cdot x_{n}^{\alpha _{n}}\ \mathrm {d} \alpha }
This can be simplified using divided differences of the exponential function to
L
I
(
x
0
,
…
,
x
n
)
=
n
!
exp
[
ln
(
x
0
)
,
…
,
ln
(
x
n
)
]
{\displaystyle L_{\text{I}}\left(x_{0},\,\dots ,\,x_{n}\right)=n!\exp \left[\ln \left(x_{0}\right),\,\dots ,\,\ln \left(x_{n}\right)\right]}
.
Example n = 2 :
L
I
(
x
,
y
,
z
)
=
−
2
x
(
ln
y
−
ln
z
)
+
y
(
ln
z
−
ln
x
)
+
z
(
ln
x
−
ln
y
)
(
ln
x
−
ln
y
)
(
ln
y
−
ln
z
)
(
ln
z
−
ln
x
)
.
{\displaystyle L_{\text{I}}(x,y,z)=-2{\frac {x(\ln y-\ln z)+y(\ln z-\ln x)+z(\ln x-\ln y)}{(\ln x-\ln y)(\ln y-\ln z)(\ln z-\ln x)}}.}
Connection to other means
Arithmetic mean :
L
(
x
2
,
y
2
)
L
(
x
,
y
)
=
x
+
y
2
{\displaystyle {\frac {L\left(x^{2},y^{2}\right)}{L(x,y)}}={\frac {x+y}{2}}}
Geometric mean :
L
(
x
,
y
)
L
(
1
x
,
1
y
)
=
x
y
{\displaystyle {\sqrt {\frac {L\left(x,y\right)}{L\left({\frac {1}{x}},{\frac {1}{y}}\right)}}}={\sqrt {xy}}}
Harmonic mean :
L
(
1
x
,
1
y
)
L
(
1
x
2
,
1
y
2
)
=
2
1
x
+
1
y
{\displaystyle {\frac {L\left({\frac {1}{x}},{\frac {1}{y}}\right)}{L\left({\frac {1}{x^{2}}},{\frac {1}{y^{2}}}\right)}}={\frac {2}{{\frac {1}{x}}+{\frac {1}{y}}}}}
See also
References
Citations
^
B. C. Carlson (1966). "Some inequalities for hypergeometric functions" . Proc. Amer. Math. Soc . 17 : 32–39. .
^
B. Ostle & H. L. Terwilliger (1957). "A comparison of two means". Proc. Montana Acad. Sci . 17 : 69–70.
^
Tung-Po Lin (1974). "The Power Mean and the Logarithmic Mean". The American Mathematical Monthly . 81 (8): 879–883. .
^
Frank Burk (1987). "The Geometric, Logarithmic, and Arithmetic Mean Inequality". The American Mathematical Monthly . 94 (6): 527–528. .
Bibliography
Oilfield Glossary: Term 'logarithmic mean'
Weisstein, Eric W. "Arithmetic-Logarithmic-Geometric-Mean Inequality" . MathWorld .
Stolarsky, Kenneth B.: Generalizations of the logarithmic mean , Mathematics Magazine, Vol. 48, No. 2, Mar., 1975, pp 87–92
Toyesh Prakash Sharma.: https://www.parabola.unsw.edu.au/files/articles/2020-2029/volume-58-2022/issue-2/vol58_no2_3.pdf "A generalisation of the Arithmetic-Logarithmic-Geometric Mean Inequality , Parabola Magazine, Vol. 58, No. 2, 2022, pp 1–5