Sie sind auf Seite 1von 2

Some Short Problems

Below is a collection of solutions to short problems (some invented, most found online or in a book) that Ive been
meaning to get off my chest.
It is a well-known fact that _
i =1
n
i =
n(n+1)
2
, and a similarly well-known related fact that _
i =1
n
i
3
= |
n(n+1)
2
|
2
. While
there exist relatively simple methods to evaluate these sums (and even easier ones to prove them true), I was curious
about determining a method to prove the relationship between these two sums, that is to prove that the square of
the first sum is the second (not by finding each sum separately, but by modifying one to match the other). That is, to
prove that (1 + 2 + ... + n)
2
= 1
3
+ 2
3
+ ... + n
3
directly. I determined a relatively simple method of doing so that re-
writes the equality into a form that is likeable (getting rid of those nasty cubes). It is as follows:
We must show that (1 + 2 + ... + n)
2
= 1
3
+ 2
3
+ ... + n
3
; this is equivalent to showing that
2[(1-2 + 1-3 + 1-4 + ... + 1-n) + (2-3 + 2-4 + ... + 2-n) + ... + ((n 1) -n)] + 1
2
+ 2
2
+ ... + n
2
=
1
3
+ 2
3
+ 3
3
+ ... + n
3
.
This is the result of expanding the left hand side of the original statement. Proving this is equivalent to proving that
2[(1-2 + 1-3 + ... + 1-n) + ... + ((n 1) -n)] = (1
3
1
2
) + ... + (n
3
n
2
) by rearranging the terms. This is the form
we want the RHS in, for we may write it as a summation in this form: _
i =1
n
i
2
(i 1), which is a sum of n terms. We
will pick n sums from the LHS, and show that each has a value corresponding to one of the n terms on the RHS.
We will take as the i
th
term for the LHS the sum of all the terms in that expansion which multiply an i by some
number less than i: that is, Term
i
= _
k=1
i 1
2 i k. We must first check that this accounts for all the products on the
LHS, then show that their sum is equal to the sum on the RHS. No terms overlap since each product has a unique
largest factor, (ignoring the outside and everpresent multiple of 2), which is true only because the square terms have
been moved to the other side. As well, it is clear that each product is accounted for: therefore, this partition does
account for the entire LHS. As well, note that the sums are equal term by term: for
2 i(1 + 2 + ... + (i 1)) = 2 i|
i (i 1)
2
] = 2
i
2
(i 1)
2
= i
2
(i 1), which is exactly what we need. This completes the proof. .
I have seen other relationships of arithmetic sums that can be proven using two-dimensional area diagrams, and
wondered if it might be done somehow with this relationship: this is worth some thought.
This is problem 110 from Challenging Mathematical Problems with Elementary Solutions, Vol. 2: Blichfeldts
lemma.
0
Let M be a set in the plane with area greater than 1. Show that M contains two distinct points
(x
1
, y
1
) and (x
2
, y
2
) such that x
2
x
1
and y
2
y
1
are integers.
Proof: Consider the plane with the unit lattice. We may consider each lattice square that contains points of M, and
stack them on top of one another: call the resulting diagram X. If any point of X is occupied by multiple points
from the parts of M in the stacked boxes, we are done, since then we have two points in the plane whose coordi-
nates differ by some integer, which is equivalent to the required property. Assume there is no such overlap, i.e. X
has no point that is occupied by multiple points of the original parts of M. Clearly the area of M contained in X
cannot exceed 1, since X is restricted to a 11 box (area 1). But the area of M is greater than 1, and no area overlaps
= the area of X is both greater than and less than or equal to 1, a contradiction. Therefore two such points must
exist. .
The following comes from a Putnam exam, an A2 problem from 2002. It asks us to prove that given 5 points on a
sphere, we can pick a hemisphere so that 4 of the points lie in the same hemisphere. Proof (due to Kent Richey):
Pick 2 points of the 5 and draw the great circle they define. Then either there are 3 points on one side of the great
circle, or there are 2 on one side and one on the other; either way, pick the side with more points (in either case it
will be > 2 points), and pick that hemisphere. Then there are > 4 total points in that hemisphere, which completes
the proof. .
It isnt made clear in the problem whether points on the boundary are included; we may see that they must be with
the following example. Consider the following arrangement of the 5 points (we will use Earth as an example of a
sphere for some useful referene points): one point at the north pole and one at the south, two points on opposite
sides of the equator (i.e. one point on the equator and the other on the other side of the diameter of that 3rd point
through the center of the sphere), and a fifth point in a random location (not one of the four already chosen). The
reader may check that there is no way to draw a hemisphere with 4 points contained in it if the boundary of the
hemisphere is not included; thus, we assume that boundary points are included.
This last problem comes to us from http://www.mast.queensu.ca/~mikeroth/putnam/putnam.html, a site for
Putnam practice problem sets (for some college I believe...). The problem reads: Given any n + 1 integers between
1 and 2 n, show that one of them is divisible by another. To do so, well use the Pidgenhole Principle: note that
each integer x e[1, n] has a corresponding positive integer 2 x e[2, 2 n] with the property that if we choose x to
belong to our set of n + 1 integers, 2 x may not belong (else we are done). Thus there are n pairs of integers, 1 and 2,
2 and 4, 3 and 6, up to n and 2 n so that if one belongs to the set the other one may not. Thus we may at best pick
one number from each pair: but that makes only n total integers, so that we are required to pick 2 numbers from the
same pair, the desired result. .
I thought it interesting that other issues may occur: for exmaple, if we chose 3 and 9, which dont have any issue
according to the pairing used above, could still cause a problem. This doesnt show a flaw in the proof, but rather
indicated to me that n + 1 may not be the smallest number for which this property is required. In my investigation I
considered the smallest number x
n
with this property for each n. It turns out that n + 1 is minimal for all evens (2n),
while odds (2n 1) have minimal value n + 1. My first notes were that x
n
is non-decreasing, as the set of integers
from 1 to n is built recursively, so that if a set is possible with the first i integers it is also available with the first i + 1
integers. A primitive lower bound I came up with was x
n
>
n
, since we may always consider the set of the first n
primes which, by definition, dont divide one another. However, if we consider all the numbers in the range
||
n+1
2
|, n| as a set, they are all too large to divide one another; there are exactly |
n+1
2
| integers in this range, which
gives the lower bound of |
n+1
2
|+1. But this is exactly the maximum value established using the Pidgenhole Principle,
and consequently represents the minimum such value. (We encourage the reader to check that it gives the right value
for both even and odd values of n; note that |X| here represents the smallest integer larger than X.)
Some Short Problems
Below is a collection of solutions to short problems (some invented, most found online or in a book) that Ive been
meaning to get off my chest.
It is a well-known fact that _
i =1
n
i =
n(n+1)
2
, and a similarly well-known related fact that _
i =1
n
i
3
= |
n(n+1)
2
|
2
. While
there exist relatively simple methods to evaluate these sums (and even easier ones to prove them true), I was curious
about determining a method to prove the relationship between these two sums, that is to prove that the square of
the first sum is the second (not by finding each sum separately, but by modifying one to match the other). That is, to
prove that (1 + 2 + ... + n)
2
= 1
3
+ 2
3
+ ... + n
3
directly. I determined a relatively simple method of doing so that re-
writes the equality into a form that is likeable (getting rid of those nasty cubes). It is as follows:
We must show that (1 + 2 + ... + n)
2
= 1
3
+ 2
3
+ ... + n
3
; this is equivalent to showing that
2[(1-2 + 1-3 + 1-4 + ... + 1-n) + (2-3 + 2-4 + ... + 2-n) + ... + ((n 1) -n)] + 1
2
+ 2
2
+ ... + n
2
=
1
3
+ 2
3
+ 3
3
+ ... + n
3
.
This is the result of expanding the left hand side of the original statement. Proving this is equivalent to proving that
2[(1-2 + 1-3 + ... + 1-n) + ... + ((n 1) -n)] = (1
3
1
2
) + ... + (n
3
n
2
) by rearranging the terms. This is the form
we want the RHS in, for we may write it as a summation in this form: _
i =1
n
i
2
(i 1), which is a sum of n terms. We
will pick n sums from the LHS, and show that each has a value corresponding to one of the n terms on the RHS.
We will take as the i
th
term for the LHS the sum of all the terms in that expansion which multiply an i by some
number less than i: that is, Term
i
= _
k=1
i 1
2 i k. We must first check that this accounts for all the products on the
LHS, then show that their sum is equal to the sum on the RHS. No terms overlap since each product has a unique
largest factor, (ignoring the outside and everpresent multiple of 2), which is true only because the square terms have
been moved to the other side. As well, it is clear that each product is accounted for: therefore, this partition does
account for the entire LHS. As well, note that the sums are equal term by term: for
2 i(1 + 2 + ... + (i 1)) = 2 i|
i (i 1)
2
] = 2
i
2
(i 1)
2
= i
2
(i 1), which is exactly what we need. This completes the proof. .
I have seen other relationships of arithmetic sums that can be proven using two-dimensional area diagrams, and
wondered if it might be done somehow with this relationship: this is worth some thought.
This is problem 110 from Challenging Mathematical Problems with Elementary Solutions, Vol. 2: Blichfeldts
lemma.
0
Let M be a set in the plane with area greater than 1. Show that M contains two distinct points
(x
1
, y
1
) and (x
2
, y
2
) such that x
2
x
1
and y
2
y
1
are integers.
Proof: Consider the plane with the unit lattice. We may consider each lattice square that contains points of M, and
stack them on top of one another: call the resulting diagram X. If any point of X is occupied by multiple points
from the parts of M in the stacked boxes, we are done, since then we have two points in the plane whose coordi-
nates differ by some integer, which is equivalent to the required property. Assume there is no such overlap, i.e. X
has no point that is occupied by multiple points of the original parts of M. Clearly the area of M contained in X
cannot exceed 1, since X is restricted to a 11 box (area 1). But the area of M is greater than 1, and no area overlaps
= the area of X is both greater than and less than or equal to 1, a contradiction. Therefore two such points must
exist. .
The following comes from a Putnam exam, an A2 problem from 2002. It asks us to prove that given 5 points on a
sphere, we can pick a hemisphere so that 4 of the points lie in the same hemisphere. Proof (due to Kent Richey):
Pick 2 points of the 5 and draw the great circle they define. Then either there are 3 points on one side of the great
circle, or there are 2 on one side and one on the other; either way, pick the side with more points (in either case it
will be > 2 points), and pick that hemisphere. Then there are > 4 total points in that hemisphere, which completes
the proof. .
It isnt made clear in the problem whether points on the boundary are included; we may see that they must be with
the following example. Consider the following arrangement of the 5 points (we will use Earth as an example of a
sphere for some useful referene points): one point at the north pole and one at the south, two points on opposite
sides of the equator (i.e. one point on the equator and the other on the other side of the diameter of that 3rd point
through the center of the sphere), and a fifth point in a random location (not one of the four already chosen). The
reader may check that there is no way to draw a hemisphere with 4 points contained in it if the boundary of the
hemisphere is not included; thus, we assume that boundary points are included.
This last problem comes to us from http://www.mast.queensu.ca/~mikeroth/putnam/putnam.html, a site for
Putnam practice problem sets (for some college I believe...). The problem reads: Given any n + 1 integers between
1 and 2 n, show that one of them is divisible by another. To do so, well use the Pidgenhole Principle: note that
each integer x e[1, n] has a corresponding positive integer 2 x e[2, 2 n] with the property that if we choose x to
belong to our set of n + 1 integers, 2 x may not belong (else we are done). Thus there are n pairs of integers, 1 and 2,
2 and 4, 3 and 6, up to n and 2 n so that if one belongs to the set the other one may not. Thus we may at best pick
one number from each pair: but that makes only n total integers, so that we are required to pick 2 numbers from the
same pair, the desired result. .
I thought it interesting that other issues may occur: for exmaple, if we chose 3 and 9, which dont have any issue
according to the pairing used above, could still cause a problem. This doesnt show a flaw in the proof, but rather
indicated to me that n + 1 may not be the smallest number for which this property is required. In my investigation I
considered the smallest number x
n
with this property for each n. It turns out that n + 1 is minimal for all evens (2n),
while odds (2n 1) have minimal value n + 1. My first notes were that x
n
is non-decreasing, as the set of integers
from 1 to n is built recursively, so that if a set is possible with the first i integers it is also available with the first i + 1
integers. A primitive lower bound I came up with was x
n
>
n
, since we may always consider the set of the first n
primes which, by definition, dont divide one another. However, if we consider all the numbers in the range
||
n+1
2
|, n| as a set, they are all too large to divide one another; there are exactly |
n+1
2
| integers in this range, which
gives the lower bound of |
n+1
2
|+1. But this is exactly the maximum value established using the Pidgenhole Principle,
and consequently represents the minimum such value. (We encourage the reader to check that it gives the right value
for both even and odd values of n; note that |X| here represents the smallest integer larger than X.)
2 Some Short Problems.nb

Das könnte Ihnen auch gefallen