A SELF-RECURRENCE METHOD FOR GENERALIZING KNOWN SCIENTIFIC RESULTS

Florentin Smarandache University of New Mexico 200 College Road Gallup, NM 87301, USA E-mail: smarand@unm.edu A great number of articles widen known scientific results (theorems, inequalities, math/physics/chemical etc. propositions, formulas), and this is due to a simple procedure, of which it is good to say a few words: Let suppose that we want to generalizes a known mathematical proposition P(a ) , where a is a constant, to the proposition P(n ) , where n is a variable which belongs to subset of N . To prove that P is true for n by recurrence means the following: the first step is trivial, since it is about the known result P(a ) (and thus it was already verified before by other mathematicians!). To pass from P(n ) to P(n + 1) , one uses too P(a ) : therefore one widens a proposition by using the proposition itself, in other words the found generalization will be paradoxically proved with the help of the particular case from which one started! We present below the generalizations of Hölder, Minkovski, and respectively Tchebychev inequalities.

1. A GENERALIZATION OF THE INEQUALITY OF HÖLDER One generalizes the inequality of Hödler thanks to a reasoning by recurrence. As particular cases, one obtains a generalization of the inequality of Cauchy-BuniakovskiScwartz, and some interesting applications. Theorem: If ai( k ) ∈ R + and pk ∈]1, +∞[ , i ∈ {1, 2,..., n} , k ∈ {1, 2,..., m} , such 1 1 1 that:, + + ... + = 1 , then: p1 p2 pm

∑∏ a
i =1 k =1

n

m

(k ) i

≤∏
k =1

m

⎛ n (k ) ∑a ⎜ ⎝ i =1 i

( )

pk

⎞ pk with m ≥ 2 . ⎟ ⎠

1

Proof: For m = 2 one obtains exactly the inequality of Hödler, which is true. One supposes that the inequality is true for the values which are strictly smaller than a certain m . Then:,

1

∑∏ a
i =1 k =1

n

m

(k ) i

⎛ m−2 n pk ⎛⎛ ⎜ ⎛ ( m −1) ( m) ⎞ k ⎞ = ∑ ⎜ ⎜ ∏ ai ⎟ ⋅ ( ai ⋅ ai ) ⎟ ≤ ⎜ ∏ ⎜ ∑ ( ai( k ) ) ⎜ i =1 ⎝ ⎝ k =1 ⎠ ⎠ ⎜ k =1 ⎝ i =1 ⎝
n m−2

⎞ ⎟ ⎟ ⎠

1 pk

p ⎞ n p⎞ ⎟ ⎛ ( m −1) (m) ⋅ ai ) ⎟ ⎟ ⋅ ⎜ ∑ ( ai 1 i = ⎝ ⎠ ⎟ ⎠ 1

where but

1 1 1 1 + + ... + + = 1 and ph > 1 , 1 ≤ h ≤ m − 2 , p > 1 ; p1 p2 pm − 2 p

∑(a
n i =1

( m −1) i

) ⋅(a )
p (m) i

p

⎛ n ≤ ⎜ ∑ ai( m −1) ⎝ i =1

((

)

p t1

)

⎞ ⎛ n (m) ⎟ ⋅ ⎜ ∑ ai ⎠ ⎝ i =1

1 t 1

1

((

)

p t2

)

⎞ ⎟ ⎠

t

2

where

1 1 + = 1 and t1 > 1 , t 2 > 2 . t1 t 2 It results from it:

∑(a
n i =1

( m −1) i

) ⋅(a )
p (m) i

p

⎛ ≤ ⎜ ∑ ai( m −1) ⎝ i =1
n

(

)

pt1

⎞ ⎟ ⎠

1 pt 1

⎛ ⋅ ⎜ ∑ ai( m ) ⎝ i =1
n

(

)

pt2

⎞ ⎟ ⎠

1 pt 2

with

1 1 1 + = . pt1 pt 2 p

1 1 1 + + ... + = 1 is true and one p1 p2 pm has p j > 1 for 1 ≤ j ≤ m and it results the inequality from the theorem. Let us note pt1 = pm −1 and pt 2 = pm . Then
Note: If one poses p j = m for 1 ≤ j ≤ m and if one raises to the power m this

inequality, one obtains a generalization of the inequality of Cauchy-BuniakovskiScwartz:
n m ⎛ n m (k ) ⎞ a ≤ ai( k ) ∑∏ ⎠ ∏ ∑ ⎜ ⎝ i =1 k =1 i ⎟ k =1 i =1 m

( ).
m

Application: Let a1 , a2 , b1 , b2 , c1 , c2 be positive real numbers. Show that: 6 6 6 (a1b1c1 + a2 b2 c2 )6 ≤ 8(a16 + a2 )(b16 + b2 )(c16 + c2 ) Solution: We will use the previous theorem. Let us choose p1 = 2 , p2 = 3 , p3 = 6 ; we will obtain the following:
a1b1c1 + a2b2 c2 ≤ (a + a ) (b + b ) (c + c ) ,
2 1 3 1 6 1 1 2 2 2 1 3 3 2 1 6 6 2

2

or more: 2 3 3 2 6 (a1b1c1 + a2 b2 c2 )6 ≤ (a12 + a2 ) (b13 + b2 ) (c16 + c2 ), and knowing that 3 2 6 (b13 + b2 ) ≤ 2(b16 + b2 ) and that 2 3 6 2 4 6 (a12 + a2 ) = a16 + a2 + 3(a14 a2 + a12 a2 ) ≤ 4(a16 + a2 ) since
2 4 6 2 a14 a2 + a12 a2 ≤ a16 + a2 (because: − a2 − a12

(

) (a
2

2 1

2 + a2 ≤ 0)

)

it results the exercise which was proposed.

2. A GENERALIZATION OF THE INEQUALITY OF MINKOWSKI Theorem : If p is a real number ≥ 1 and ai( k ) ∈ R+ with i ∈ {1, 2,..., n} and k ∈ {1,2,..., m} , then:

⎛ n ⎛ m (k ) ⎞ p ⎞ ⎜ ∑ ⎜ ∑ ai ⎟ ⎟ ⎜ i =1 ⎝ k =1 ⎠ ⎟ ⎝ ⎠

1 p

⎛ m ⎛ n (k ) ⎞ p ⎞ ≤ ⎜ ∑ ⎜ ∑ ai ⎟ ⎟ ⎜ k =1 ⎝ i =1 ⎠ ⎟ ⎝ ⎠

1/ p

Demonstration by recurrence on m ∈ N*. First of all one shows that: ⎛ n (1) p ⎞ ∑a ⎠ ⎜ ⎝ i =1 i ⎟

( )

1 p

p⎞ ⎛ n ≤ ⎜ ∑ ai(1) ⎟ ⎠ ⎝ i =1

( )

1/ p

, which is obvious, and proves that the inequality

is true for m = 1 . (The case m = 2 precisely constitutes the inequality of Minkowski, which is naturally true!). Let us suppose that the inequality is true for all the values less or equal to m ⎛ n ⎛ m +1 ( k ) ⎞ p ⎞ ⎜ ∑ ⎜ ∑ ai ⎟ ⎟ ⎜ i =1 ⎝ k =1 ⎠ ⎟ ⎝ ⎠
1 p

⎛ n (1) p ⎞ ≤ ⎜ ∑ ai ⎟ ⎝ i =1 ⎠
1 p

1 p

⎛ n ⎛ m +1 ( k ) ⎞ p ⎞ + ⎜ ∑ ⎜ ∑ ai ⎟ ⎟ ⎜ i =1 ⎝ k = 2 ⎠ ⎟ ⎝ ⎠
1 p

1 p

p⎞ ⎛ n ≤ ⎜ ∑ ai(1) ⎟ ⎠ ⎝ i =1

( )

⎛ m +1 ⎛ n ( k ) ⎞ p ⎞ + ⎜ ∑ ⎜ ∑ ai ⎟ ⎟ ⎠ ⎠ ⎝ k = 2 ⎝ i =1

3

⎛ m +1 ⎛ n ( k ) ⎞ p ⎞ and this last sum is ⎜ ∑ ⎜ ∑ ai ⎟ ⎟ ⎠ ⎠ ⎝ k =1 ⎝ i =1 m +1.

1 p

therefore the inequality is true for the level

3. A GENERALIZATION OF AN INEQUALITY OF TCHEBYCHEV
k) Statement: If ai( k ) ≥ ai(+ 1 , i ∈ {1,2,..., n − 1} , k ∈ {1,2,..., m} , then:

1 n m (k ) 1 m n (k ) ∑ ∏ ai ≥ n m ∏ ∑ ai . n i =1 k =1 k =1 i =1

Demonstration by recurrence on m . Case m = 1 is obvious:
1 n (1) 1 n (1) ∑ ai ≥ n ∑ ai . n i =1 i =1

In the case m = 2 , this is the inequality of Tchebychev itself:
(1) (1) (1) (2) (2) (2) If a1 ≥ a2 ≥ ... ≥ an and a1 ≥ a2 ≥ ... ≥ an , then: (1) (2) (1) (2) (1) (1) (2) a1(1) a1(2) + a2 a2 + ... + an an a (1) + a2 + ... + an a (2) + ... + an ≥ 1 × 1 n n n

One supposes that the inequality is true for all the values smaller or equal to m . It is necessary to prove for the rang m + 1 : 1 n m +1 ( k ) 1 n ⎛ m ( k ) ⎞ ( m +1) ⋅a . ∑ ∏ ai = n ∑ ⎜ ∏ ai ⎟ ⎠ i n i =1 k =1 i =1 ⎝ k =1 ⎛1 n m ⎞ ⎛1 n ⎞ ⎛ 1 m n ⎞ ⎛1 n ⎞ This is ≥ ⎜ ∑ ∏ ai( k ) ⎟ ⋅ ⎜ ∑ ai( m +1) ⎟ ≥ ⎜ m ∏ ∑ ai( k ) ⎟ ⋅ ⎜ ∑ ai( m +1) ⎟ ⎝ n i =1 k =1 ⎠ ⎝ n i =1 ⎠ ⎝ n k =1 i =1 ⎠ ⎝ n i =1 ⎠ and this is exactly
1 n
m +1

∏∑a
k =1 i =1

m +1 n

(k ) i

(Quod Erat Demonstrandum).

{Translated from French by the Author.}
Reference: F. Smarandache, Généralisations et Généralités, Ed. Nouvelle, Fès, Morocco, 1984.

4