Prove this Kenneth S. Williams inequality
$begingroup$
If $0<a_1le a_2le cdotsle a_n$, then the following inequality holds:
$$frac{1}{2n^2a_n} {sum_{1le i < jle n}^{} {(a_i-a_j)^2}}le frac{a_1+a_2+cdots + a_n}{n}-sqrt [n]{a_1 a_2 cdots a_n }{le frac{1}{2n^2a_1}sum_{1le i < jle n}^{} {(a_i-a_j)^2}}$$This problem was proposed by Kenneth S. Williams, Carleton University, Ottawa in CRUX 247[1977;131] and in CRUX[1978;23,37] it is said
there is a nice simple proof but I can't find this G.Szekeres (OCtober 1977 was published by Rennie in JCMN,NO.12) shorter proof,can help me? Thanks
Maybe Now this inequality have some methods to solve it?such as AM-GM inequality?
real-analysis multivariable-calculus inequality a.m.-g.m.-inequality
$endgroup$
add a comment |
$begingroup$
If $0<a_1le a_2le cdotsle a_n$, then the following inequality holds:
$$frac{1}{2n^2a_n} {sum_{1le i < jle n}^{} {(a_i-a_j)^2}}le frac{a_1+a_2+cdots + a_n}{n}-sqrt [n]{a_1 a_2 cdots a_n }{le frac{1}{2n^2a_1}sum_{1le i < jle n}^{} {(a_i-a_j)^2}}$$This problem was proposed by Kenneth S. Williams, Carleton University, Ottawa in CRUX 247[1977;131] and in CRUX[1978;23,37] it is said
there is a nice simple proof but I can't find this G.Szekeres (OCtober 1977 was published by Rennie in JCMN,NO.12) shorter proof,can help me? Thanks
Maybe Now this inequality have some methods to solve it?such as AM-GM inequality?
real-analysis multivariable-calculus inequality a.m.-g.m.-inequality
$endgroup$
$begingroup$
It's interesting even for $n=3$. Very nice problem!
$endgroup$
– Michael Rozenberg
Dec 13 '18 at 8:21
1
$begingroup$
Here you go maths.ed.ac.uk/cook/iss_no12.PDF
$endgroup$
– Macavity
Dec 16 '18 at 8:21
add a comment |
$begingroup$
If $0<a_1le a_2le cdotsle a_n$, then the following inequality holds:
$$frac{1}{2n^2a_n} {sum_{1le i < jle n}^{} {(a_i-a_j)^2}}le frac{a_1+a_2+cdots + a_n}{n}-sqrt [n]{a_1 a_2 cdots a_n }{le frac{1}{2n^2a_1}sum_{1le i < jle n}^{} {(a_i-a_j)^2}}$$This problem was proposed by Kenneth S. Williams, Carleton University, Ottawa in CRUX 247[1977;131] and in CRUX[1978;23,37] it is said
there is a nice simple proof but I can't find this G.Szekeres (OCtober 1977 was published by Rennie in JCMN,NO.12) shorter proof,can help me? Thanks
Maybe Now this inequality have some methods to solve it?such as AM-GM inequality?
real-analysis multivariable-calculus inequality a.m.-g.m.-inequality
$endgroup$
If $0<a_1le a_2le cdotsle a_n$, then the following inequality holds:
$$frac{1}{2n^2a_n} {sum_{1le i < jle n}^{} {(a_i-a_j)^2}}le frac{a_1+a_2+cdots + a_n}{n}-sqrt [n]{a_1 a_2 cdots a_n }{le frac{1}{2n^2a_1}sum_{1le i < jle n}^{} {(a_i-a_j)^2}}$$This problem was proposed by Kenneth S. Williams, Carleton University, Ottawa in CRUX 247[1977;131] and in CRUX[1978;23,37] it is said
there is a nice simple proof but I can't find this G.Szekeres (OCtober 1977 was published by Rennie in JCMN,NO.12) shorter proof,can help me? Thanks
Maybe Now this inequality have some methods to solve it?such as AM-GM inequality?
real-analysis multivariable-calculus inequality a.m.-g.m.-inequality
real-analysis multivariable-calculus inequality a.m.-g.m.-inequality
edited Dec 13 '18 at 18:05
Michael Rozenberg
101k1591193
101k1591193
asked Dec 7 '17 at 5:00
geromtygeromty
977423
977423
$begingroup$
It's interesting even for $n=3$. Very nice problem!
$endgroup$
– Michael Rozenberg
Dec 13 '18 at 8:21
1
$begingroup$
Here you go maths.ed.ac.uk/cook/iss_no12.PDF
$endgroup$
– Macavity
Dec 16 '18 at 8:21
add a comment |
$begingroup$
It's interesting even for $n=3$. Very nice problem!
$endgroup$
– Michael Rozenberg
Dec 13 '18 at 8:21
1
$begingroup$
Here you go maths.ed.ac.uk/cook/iss_no12.PDF
$endgroup$
– Macavity
Dec 16 '18 at 8:21
$begingroup$
It's interesting even for $n=3$. Very nice problem!
$endgroup$
– Michael Rozenberg
Dec 13 '18 at 8:21
$begingroup$
It's interesting even for $n=3$. Very nice problem!
$endgroup$
– Michael Rozenberg
Dec 13 '18 at 8:21
1
1
$begingroup$
Here you go maths.ed.ac.uk/cook/iss_no12.PDF
$endgroup$
– Macavity
Dec 16 '18 at 8:21
$begingroup$
Here you go maths.ed.ac.uk/cook/iss_no12.PDF
$endgroup$
– Macavity
Dec 16 '18 at 8:21
add a comment |
1 Answer
1
active
oldest
votes
$begingroup$
This is not a hard inequality, so I can't imagine what the "longer proof" could be. Here is a proof that only involves easy techniques.
By homogeneity we may assume $sum_{i=1}^na_i=1$. Let
$$S=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_n}sum_{1le i<jle n}(a_i-a_j)^2$$
and
$$T=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_1}sum_{1le i<jle n}(a_i-a_j)^2$$
Obviously, if $a_1=cdots=a_n=1/n$, we have $S=T=0$. It suffices to prove that $S$(resp. $T$) attains its minimum(resp. maximum) at $a_1=cdots=a_n=1/n$. We first show that the minimum of $S$ and the maximum of $T$ are attainable. In fact, the domain of $S$ can be extended to non-negative reals $0le a_1leldots le a_n$, $sum_{i=1}^na_i=1$ since $a_nge1/n>0$ on this set. This set is compact, and $S$ is smooth, so $S$ attains its minimum. On the other hand, $Tto-infty$ as $a_1to0^+$, thus the supremum of $T$ on $0<a_1leldots le a_n$, $sum_{i=1}^na_i=1$ is equal to the supremum of $T$ on $epsilonle a_1leldots le a_n$, $sum_{i=1}^na_i=1$ for some $epsilon>0$. Again this is a compact set and $T$ attains its maximum on it.
The key point of the proof is the following two lemmas:
Lemma 1. Suppose $a_i<sqrt[n]{a_1cdots a_n}$ for some $i$. In this case, we may choose $i$ to be the maximum among all such $i$. Then there exists a pair $a_i',a_{i+1}'$ such that $a_i'+a_{i+1}'=a_i+a_{i+1}$, $a_1leldotsle a_{i-1}le a_i'le a_{i+1}'le a_{i+2}leldotsle a_n$, and
begin{align}S(a_1,ldots,a_i',a_{i+1}',ldots,a_n)<S(a_1,ldots,a_i,a_{i+1},ldots,a_n)
end{align}
Proof. By the maximality of $i$ we have $a_i<a_{i+1}$. Let $f(lambda)=S(a_1,ldots,a_i+lambda,a_{i+1}-lambda,ldots,a_n)$. Taking derivatives if $a_ine0$, we obtain
$$f'(0)=frac{a_{i+1}-a_i}nleft(frac1{a_n}-frac{sqrt[n]{a_1cdots a_n}}{a_ia_{i+1}}right)$$
We show that $f'(0)<0$, hence for some sufficiently small $lambda>0$ we may take $a_i'=a_i+lambda$ and $a_{i+1}'=a_{i+1}-lambda$ which would strictly decrease $S$ as is required. In light of this, we may admit the case $a_i=0$ here where we think $f'(0)=-infty<0$. For $a_ine0$, it reduces to prove $a_ia_{i+1}<a_nsqrt[n]{a_1cdots a_n}$. But $a_i<sqrt[n]{a_1cdots a_n}$ and $a_{i+1}le a_n$. This proves our lemma.
Lemma 2. Suppose $a_j>sqrt[n]{a_1cdots a_n}$ for some $j>1$. In this case, we may choose $j$ to be the minimum among all such $j$. Then there exists a pair $a_{j-1}',a_j'$ such that $a_{j-1}'+a_j'=a_{j-1}+a_j$, $a_1leldotsle a_{j-2}
le a_{j-1}'le a_j'le a_{j+1}leldotsle a_n$, and
$$T(a_1,ldots,a_{j-1}',a_j',ldots,a_n)>T(a_1,ldots,a_{j-1},a_j,ldots,a_n)$$
Proof. Completely similar to the proof of Lemma 1.
Finally, we note that $a_1gesqrt[n]{a_1cdots a_n}$ implies $a_1=a_2=cdots=a_n$; so is $a_nlesqrt[n]{a_1cdots a_n}$. If the maximum(resp. minimum) of $S$(resp. $T$) is not attained at $a_1=cdots=a_n=1/n$, Lemma 1(resp. Lemma 2) would immediately yield a contradiction. This completes our proof.
$endgroup$
add a comment |
Your Answer
StackExchange.ifUsing("editor", function () {
return StackExchange.using("mathjaxEditing", function () {
StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix) {
StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
});
});
}, "mathjax-editing");
StackExchange.ready(function() {
var channelOptions = {
tags: "".split(" "),
id: "69"
};
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function() {
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled) {
StackExchange.using("snippets", function() {
createEditor();
});
}
else {
createEditor();
}
});
function createEditor() {
StackExchange.prepareEditor({
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader: {
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
},
noCode: true, onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
});
}
});
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2555054%2fprove-this-kenneth-s-williams-inequality%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
1 Answer
1
active
oldest
votes
1 Answer
1
active
oldest
votes
active
oldest
votes
active
oldest
votes
$begingroup$
This is not a hard inequality, so I can't imagine what the "longer proof" could be. Here is a proof that only involves easy techniques.
By homogeneity we may assume $sum_{i=1}^na_i=1$. Let
$$S=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_n}sum_{1le i<jle n}(a_i-a_j)^2$$
and
$$T=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_1}sum_{1le i<jle n}(a_i-a_j)^2$$
Obviously, if $a_1=cdots=a_n=1/n$, we have $S=T=0$. It suffices to prove that $S$(resp. $T$) attains its minimum(resp. maximum) at $a_1=cdots=a_n=1/n$. We first show that the minimum of $S$ and the maximum of $T$ are attainable. In fact, the domain of $S$ can be extended to non-negative reals $0le a_1leldots le a_n$, $sum_{i=1}^na_i=1$ since $a_nge1/n>0$ on this set. This set is compact, and $S$ is smooth, so $S$ attains its minimum. On the other hand, $Tto-infty$ as $a_1to0^+$, thus the supremum of $T$ on $0<a_1leldots le a_n$, $sum_{i=1}^na_i=1$ is equal to the supremum of $T$ on $epsilonle a_1leldots le a_n$, $sum_{i=1}^na_i=1$ for some $epsilon>0$. Again this is a compact set and $T$ attains its maximum on it.
The key point of the proof is the following two lemmas:
Lemma 1. Suppose $a_i<sqrt[n]{a_1cdots a_n}$ for some $i$. In this case, we may choose $i$ to be the maximum among all such $i$. Then there exists a pair $a_i',a_{i+1}'$ such that $a_i'+a_{i+1}'=a_i+a_{i+1}$, $a_1leldotsle a_{i-1}le a_i'le a_{i+1}'le a_{i+2}leldotsle a_n$, and
begin{align}S(a_1,ldots,a_i',a_{i+1}',ldots,a_n)<S(a_1,ldots,a_i,a_{i+1},ldots,a_n)
end{align}
Proof. By the maximality of $i$ we have $a_i<a_{i+1}$. Let $f(lambda)=S(a_1,ldots,a_i+lambda,a_{i+1}-lambda,ldots,a_n)$. Taking derivatives if $a_ine0$, we obtain
$$f'(0)=frac{a_{i+1}-a_i}nleft(frac1{a_n}-frac{sqrt[n]{a_1cdots a_n}}{a_ia_{i+1}}right)$$
We show that $f'(0)<0$, hence for some sufficiently small $lambda>0$ we may take $a_i'=a_i+lambda$ and $a_{i+1}'=a_{i+1}-lambda$ which would strictly decrease $S$ as is required. In light of this, we may admit the case $a_i=0$ here where we think $f'(0)=-infty<0$. For $a_ine0$, it reduces to prove $a_ia_{i+1}<a_nsqrt[n]{a_1cdots a_n}$. But $a_i<sqrt[n]{a_1cdots a_n}$ and $a_{i+1}le a_n$. This proves our lemma.
Lemma 2. Suppose $a_j>sqrt[n]{a_1cdots a_n}$ for some $j>1$. In this case, we may choose $j$ to be the minimum among all such $j$. Then there exists a pair $a_{j-1}',a_j'$ such that $a_{j-1}'+a_j'=a_{j-1}+a_j$, $a_1leldotsle a_{j-2}
le a_{j-1}'le a_j'le a_{j+1}leldotsle a_n$, and
$$T(a_1,ldots,a_{j-1}',a_j',ldots,a_n)>T(a_1,ldots,a_{j-1},a_j,ldots,a_n)$$
Proof. Completely similar to the proof of Lemma 1.
Finally, we note that $a_1gesqrt[n]{a_1cdots a_n}$ implies $a_1=a_2=cdots=a_n$; so is $a_nlesqrt[n]{a_1cdots a_n}$. If the maximum(resp. minimum) of $S$(resp. $T$) is not attained at $a_1=cdots=a_n=1/n$, Lemma 1(resp. Lemma 2) would immediately yield a contradiction. This completes our proof.
$endgroup$
add a comment |
$begingroup$
This is not a hard inequality, so I can't imagine what the "longer proof" could be. Here is a proof that only involves easy techniques.
By homogeneity we may assume $sum_{i=1}^na_i=1$. Let
$$S=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_n}sum_{1le i<jle n}(a_i-a_j)^2$$
and
$$T=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_1}sum_{1le i<jle n}(a_i-a_j)^2$$
Obviously, if $a_1=cdots=a_n=1/n$, we have $S=T=0$. It suffices to prove that $S$(resp. $T$) attains its minimum(resp. maximum) at $a_1=cdots=a_n=1/n$. We first show that the minimum of $S$ and the maximum of $T$ are attainable. In fact, the domain of $S$ can be extended to non-negative reals $0le a_1leldots le a_n$, $sum_{i=1}^na_i=1$ since $a_nge1/n>0$ on this set. This set is compact, and $S$ is smooth, so $S$ attains its minimum. On the other hand, $Tto-infty$ as $a_1to0^+$, thus the supremum of $T$ on $0<a_1leldots le a_n$, $sum_{i=1}^na_i=1$ is equal to the supremum of $T$ on $epsilonle a_1leldots le a_n$, $sum_{i=1}^na_i=1$ for some $epsilon>0$. Again this is a compact set and $T$ attains its maximum on it.
The key point of the proof is the following two lemmas:
Lemma 1. Suppose $a_i<sqrt[n]{a_1cdots a_n}$ for some $i$. In this case, we may choose $i$ to be the maximum among all such $i$. Then there exists a pair $a_i',a_{i+1}'$ such that $a_i'+a_{i+1}'=a_i+a_{i+1}$, $a_1leldotsle a_{i-1}le a_i'le a_{i+1}'le a_{i+2}leldotsle a_n$, and
begin{align}S(a_1,ldots,a_i',a_{i+1}',ldots,a_n)<S(a_1,ldots,a_i,a_{i+1},ldots,a_n)
end{align}
Proof. By the maximality of $i$ we have $a_i<a_{i+1}$. Let $f(lambda)=S(a_1,ldots,a_i+lambda,a_{i+1}-lambda,ldots,a_n)$. Taking derivatives if $a_ine0$, we obtain
$$f'(0)=frac{a_{i+1}-a_i}nleft(frac1{a_n}-frac{sqrt[n]{a_1cdots a_n}}{a_ia_{i+1}}right)$$
We show that $f'(0)<0$, hence for some sufficiently small $lambda>0$ we may take $a_i'=a_i+lambda$ and $a_{i+1}'=a_{i+1}-lambda$ which would strictly decrease $S$ as is required. In light of this, we may admit the case $a_i=0$ here where we think $f'(0)=-infty<0$. For $a_ine0$, it reduces to prove $a_ia_{i+1}<a_nsqrt[n]{a_1cdots a_n}$. But $a_i<sqrt[n]{a_1cdots a_n}$ and $a_{i+1}le a_n$. This proves our lemma.
Lemma 2. Suppose $a_j>sqrt[n]{a_1cdots a_n}$ for some $j>1$. In this case, we may choose $j$ to be the minimum among all such $j$. Then there exists a pair $a_{j-1}',a_j'$ such that $a_{j-1}'+a_j'=a_{j-1}+a_j$, $a_1leldotsle a_{j-2}
le a_{j-1}'le a_j'le a_{j+1}leldotsle a_n$, and
$$T(a_1,ldots,a_{j-1}',a_j',ldots,a_n)>T(a_1,ldots,a_{j-1},a_j,ldots,a_n)$$
Proof. Completely similar to the proof of Lemma 1.
Finally, we note that $a_1gesqrt[n]{a_1cdots a_n}$ implies $a_1=a_2=cdots=a_n$; so is $a_nlesqrt[n]{a_1cdots a_n}$. If the maximum(resp. minimum) of $S$(resp. $T$) is not attained at $a_1=cdots=a_n=1/n$, Lemma 1(resp. Lemma 2) would immediately yield a contradiction. This completes our proof.
$endgroup$
add a comment |
$begingroup$
This is not a hard inequality, so I can't imagine what the "longer proof" could be. Here is a proof that only involves easy techniques.
By homogeneity we may assume $sum_{i=1}^na_i=1$. Let
$$S=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_n}sum_{1le i<jle n}(a_i-a_j)^2$$
and
$$T=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_1}sum_{1le i<jle n}(a_i-a_j)^2$$
Obviously, if $a_1=cdots=a_n=1/n$, we have $S=T=0$. It suffices to prove that $S$(resp. $T$) attains its minimum(resp. maximum) at $a_1=cdots=a_n=1/n$. We first show that the minimum of $S$ and the maximum of $T$ are attainable. In fact, the domain of $S$ can be extended to non-negative reals $0le a_1leldots le a_n$, $sum_{i=1}^na_i=1$ since $a_nge1/n>0$ on this set. This set is compact, and $S$ is smooth, so $S$ attains its minimum. On the other hand, $Tto-infty$ as $a_1to0^+$, thus the supremum of $T$ on $0<a_1leldots le a_n$, $sum_{i=1}^na_i=1$ is equal to the supremum of $T$ on $epsilonle a_1leldots le a_n$, $sum_{i=1}^na_i=1$ for some $epsilon>0$. Again this is a compact set and $T$ attains its maximum on it.
The key point of the proof is the following two lemmas:
Lemma 1. Suppose $a_i<sqrt[n]{a_1cdots a_n}$ for some $i$. In this case, we may choose $i$ to be the maximum among all such $i$. Then there exists a pair $a_i',a_{i+1}'$ such that $a_i'+a_{i+1}'=a_i+a_{i+1}$, $a_1leldotsle a_{i-1}le a_i'le a_{i+1}'le a_{i+2}leldotsle a_n$, and
begin{align}S(a_1,ldots,a_i',a_{i+1}',ldots,a_n)<S(a_1,ldots,a_i,a_{i+1},ldots,a_n)
end{align}
Proof. By the maximality of $i$ we have $a_i<a_{i+1}$. Let $f(lambda)=S(a_1,ldots,a_i+lambda,a_{i+1}-lambda,ldots,a_n)$. Taking derivatives if $a_ine0$, we obtain
$$f'(0)=frac{a_{i+1}-a_i}nleft(frac1{a_n}-frac{sqrt[n]{a_1cdots a_n}}{a_ia_{i+1}}right)$$
We show that $f'(0)<0$, hence for some sufficiently small $lambda>0$ we may take $a_i'=a_i+lambda$ and $a_{i+1}'=a_{i+1}-lambda$ which would strictly decrease $S$ as is required. In light of this, we may admit the case $a_i=0$ here where we think $f'(0)=-infty<0$. For $a_ine0$, it reduces to prove $a_ia_{i+1}<a_nsqrt[n]{a_1cdots a_n}$. But $a_i<sqrt[n]{a_1cdots a_n}$ and $a_{i+1}le a_n$. This proves our lemma.
Lemma 2. Suppose $a_j>sqrt[n]{a_1cdots a_n}$ for some $j>1$. In this case, we may choose $j$ to be the minimum among all such $j$. Then there exists a pair $a_{j-1}',a_j'$ such that $a_{j-1}'+a_j'=a_{j-1}+a_j$, $a_1leldotsle a_{j-2}
le a_{j-1}'le a_j'le a_{j+1}leldotsle a_n$, and
$$T(a_1,ldots,a_{j-1}',a_j',ldots,a_n)>T(a_1,ldots,a_{j-1},a_j,ldots,a_n)$$
Proof. Completely similar to the proof of Lemma 1.
Finally, we note that $a_1gesqrt[n]{a_1cdots a_n}$ implies $a_1=a_2=cdots=a_n$; so is $a_nlesqrt[n]{a_1cdots a_n}$. If the maximum(resp. minimum) of $S$(resp. $T$) is not attained at $a_1=cdots=a_n=1/n$, Lemma 1(resp. Lemma 2) would immediately yield a contradiction. This completes our proof.
$endgroup$
This is not a hard inequality, so I can't imagine what the "longer proof" could be. Here is a proof that only involves easy techniques.
By homogeneity we may assume $sum_{i=1}^na_i=1$. Let
$$S=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_n}sum_{1le i<jle n}(a_i-a_j)^2$$
and
$$T=frac{a_1+cdots+a_n}n-sqrt[n]{a_1cdots a_n}-frac1{2n^2a_1}sum_{1le i<jle n}(a_i-a_j)^2$$
Obviously, if $a_1=cdots=a_n=1/n$, we have $S=T=0$. It suffices to prove that $S$(resp. $T$) attains its minimum(resp. maximum) at $a_1=cdots=a_n=1/n$. We first show that the minimum of $S$ and the maximum of $T$ are attainable. In fact, the domain of $S$ can be extended to non-negative reals $0le a_1leldots le a_n$, $sum_{i=1}^na_i=1$ since $a_nge1/n>0$ on this set. This set is compact, and $S$ is smooth, so $S$ attains its minimum. On the other hand, $Tto-infty$ as $a_1to0^+$, thus the supremum of $T$ on $0<a_1leldots le a_n$, $sum_{i=1}^na_i=1$ is equal to the supremum of $T$ on $epsilonle a_1leldots le a_n$, $sum_{i=1}^na_i=1$ for some $epsilon>0$. Again this is a compact set and $T$ attains its maximum on it.
The key point of the proof is the following two lemmas:
Lemma 1. Suppose $a_i<sqrt[n]{a_1cdots a_n}$ for some $i$. In this case, we may choose $i$ to be the maximum among all such $i$. Then there exists a pair $a_i',a_{i+1}'$ such that $a_i'+a_{i+1}'=a_i+a_{i+1}$, $a_1leldotsle a_{i-1}le a_i'le a_{i+1}'le a_{i+2}leldotsle a_n$, and
begin{align}S(a_1,ldots,a_i',a_{i+1}',ldots,a_n)<S(a_1,ldots,a_i,a_{i+1},ldots,a_n)
end{align}
Proof. By the maximality of $i$ we have $a_i<a_{i+1}$. Let $f(lambda)=S(a_1,ldots,a_i+lambda,a_{i+1}-lambda,ldots,a_n)$. Taking derivatives if $a_ine0$, we obtain
$$f'(0)=frac{a_{i+1}-a_i}nleft(frac1{a_n}-frac{sqrt[n]{a_1cdots a_n}}{a_ia_{i+1}}right)$$
We show that $f'(0)<0$, hence for some sufficiently small $lambda>0$ we may take $a_i'=a_i+lambda$ and $a_{i+1}'=a_{i+1}-lambda$ which would strictly decrease $S$ as is required. In light of this, we may admit the case $a_i=0$ here where we think $f'(0)=-infty<0$. For $a_ine0$, it reduces to prove $a_ia_{i+1}<a_nsqrt[n]{a_1cdots a_n}$. But $a_i<sqrt[n]{a_1cdots a_n}$ and $a_{i+1}le a_n$. This proves our lemma.
Lemma 2. Suppose $a_j>sqrt[n]{a_1cdots a_n}$ for some $j>1$. In this case, we may choose $j$ to be the minimum among all such $j$. Then there exists a pair $a_{j-1}',a_j'$ such that $a_{j-1}'+a_j'=a_{j-1}+a_j$, $a_1leldotsle a_{j-2}
le a_{j-1}'le a_j'le a_{j+1}leldotsle a_n$, and
$$T(a_1,ldots,a_{j-1}',a_j',ldots,a_n)>T(a_1,ldots,a_{j-1},a_j,ldots,a_n)$$
Proof. Completely similar to the proof of Lemma 1.
Finally, we note that $a_1gesqrt[n]{a_1cdots a_n}$ implies $a_1=a_2=cdots=a_n$; so is $a_nlesqrt[n]{a_1cdots a_n}$. If the maximum(resp. minimum) of $S$(resp. $T$) is not attained at $a_1=cdots=a_n=1/n$, Lemma 1(resp. Lemma 2) would immediately yield a contradiction. This completes our proof.
answered Dec 7 '17 at 11:11
Cave JohnsonCave Johnson
3,8481326
3,8481326
add a comment |
add a comment |
Thanks for contributing an answer to Mathematics Stack Exchange!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
Use MathJax to format equations. MathJax reference.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f2555054%2fprove-this-kenneth-s-williams-inequality%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
there is a nice simple proof but I can't find this G.Szekeres (OCtober 1977 was published by Rennie in JCMN,NO.12) shorter proof,can help me? Thanks
$begingroup$
It's interesting even for $n=3$. Very nice problem!
$endgroup$
– Michael Rozenberg
Dec 13 '18 at 8:21
1
$begingroup$
Here you go maths.ed.ac.uk/cook/iss_no12.PDF
$endgroup$
– Macavity
Dec 16 '18 at 8:21