Sunteți pe pagina 1din 8

PROMYS problems solutions

Rafail Ketsetsidis
1. Lets call each letter on the diamond a node. We observe that by travers-
ing the diamond, as stated in the problem, by going from each node either
down-left or down-right, we can always spell ABRACADABRA. This
means that we just have to count in how many distinct ways we can go
from the top to the bottom of the diamond, by going either down-left
or down-right from each node. Lets call the nodes that are up-left and
up-right from a node, its left and right parent respectively. We observe
that the number of ways to visit a node, beginning from the top of the
diamond, is equal to the sum of the number of ways to visit its parents.
The solution is in the following image:
Figure 1: The solution
There are 252 ways to spell ABRACADABRA by traversing the diamond from
top to bottom.
1
2. We have that:
3(33)=327=3*913=3*(10-1)13
Using Newtons binomial theorem we have:
3*(10-1)13 = 3*((-1)13 + 13*(-1)12*10 + 78*(-1)11*100) (mod
1000) = 3*(-1 + 130 - 7800) (mod 1000) = 987 (mod 1000) This
means that the last three digits of 3(33) are 987. By examples we
suppose that all numbers after t(n) including t(n), have the same last
n digits. Which means that the last two digits of 3(3(33)) are 87.
3. Lets call k the number of consecutive integers used to construct n, which
means that if there is a solution for n and k, then n can be written as
the sum of k consecutive integers. For example, for n=12, there exists a
solution such that k=3 (3+4+5=12). It is obvious that for any n, there
exists a solution for k=1. It is obvious that the sum of k consecutive
positive integers modulo k is either 0 or k/2 if k is odd or even respectively
(we exclude the number that is divisible by k and create the sums (1 mod
k) + (k-1 mod k) etc. If k is even, k/2 will have no match and will be
left over). That means that for every pair of solutions (n,k), if k is odd,
n=0(mod k) else n=k/2(mod k). Well prove that this is sucient for
a solution. Lets suppose we have a pair (n,k) such that n=0(mod k)
and k is odd, OR n=k/2(mod k) and k is even. Consider k consecutive
positive integers, starting from n. In order to reach a solution, well need
to subtract on each step exactly 1 from every element, until to reach the
solution. Since the elements are k, we have:
n + (n+1) + (n+2) + . . . + (n+k-1) - a*k = n (whereas a is a
positive integer. We just have to prove that it exists.)
n*(k-1) + k*(k-1)/2 - a*k = n
2*k*n - 2*n + k*(k-1) = 2*a*k
2*n - 2*n/k + k - 1 = 2*a
There are two posibilities:
1. If k is odd, n=0(mod k) so n=b*k, where b is a positive integer:
2*b*k - 2*b*k/k + k - 1 = 2*a
2*b*k - 2*b + k - 1 = 2*a
We just have to prove that k-1 is even, or, k is odd, which by hy-
pothesis is true.
2
2. If k is even, n=k/2(mod k) so n=b*k+k/2, where b is a positive integer:
2*b*k + 2*k/2 - 2*b*k/k - (2*k/2)/k + k - 1 = 2*a
2*b*k + k - 2*b - 1 + k - 1 = 2*a
2*b*k + 2*k - 2*b - 2 = 2*a
b*k + k - b - 1 = a
4. What causes the paradox, is that given gure (see Figure 3) is misleading.
Figure 2: The gure I drew
On the given gure, AD is shorter than BC, while it is stated that AD=BC.
This causes A and B to be on dierent sides of PD, while from trying to create
the gure (see Figure 2) it is obvious that this is impossible. This means that,
while triangles PAD and PBC are equal, and angles PAB and PBA are equal,
angle x is not equal to PAD - PAB, because x is out of PAD, and thus the
argument is incorrect.
3
Figure 3: The gure provided
5. After a computer simulation to reach more digits, we observe a pattern
on the chessboard. It appears that some squares reappear diagonicaly.
Furthermore, we observe that the squares adjacent to those placed diag-
onicaly (for example the squares of numbers inside the cyan squares, but
not inside the red ones) have all their elements incremented by a number,
starting from the squares diagonicaly (2=0+2, 3=1+2 etc.). This leads us
to consider an induction proof. Consider T(n) a nxn square that fullls
the restrictions. Then a solution will exist for T(2*n) as such:Since each
line and each column of T(n) contains every number from 0 to 2n-1, by
adding 2n to each element of T(n), the restrictions are kept, because
not only the new elements are unique, but due to the fact that T(n) is
a solution, T(2*n) will also have the minimum elements at each line and
column. We will then prove that each element of T(2*n) at coordinates
(x,y) will have value x XOR y.
Split square T(2*n) into four squares, up-left, up-right, down-left, down-right,
as in the above gure. At the up-left square, 0<=x<n and 0<=y<n, at the
up-right square, n<=x<2*n and 0<=y<n, at the down-left square, 0<=x<n
and n<=y<2*n, at the down-right square, n<=x<2*n and n<=y<2*n. Lets
examine the down-right square. As already stated, each one of its elements will
have coordinates n<=x<2*n and n<=y<2*n. Since n will always be a power
4
Figure 4: The computer-generated pattern
Figure 5: Induction
5
of 2, by the properties of the XOR operation, x XOR y = (x-n) XOR (y-n).
This means that the down-right square will be equal to the up-left square, as
already proven. Lets now examine the up-right square. As already stated, each
one of its elements will have coordinates n<=x<2*n and 0<=y<n. Once again,
because n is a power of two and because y<n, using the properties of the XOR
operation, x XOR y = (x-n) XOR y + n. This means that the up-right square
is equal to the up-left square plus n, which complies to the previous proof. In
the same way, the down-left square is equal to the up-left square plus n, which
concludes the solution. Thus, the number that will appear at the 2014th row
and 1777th column will be 2014 XOR 1777 = 303.
6. Well employ a Dynamic Programming-like train of thought. Since
2n=100. . . 00(with n+1 digits), all scattered numbers below 2n will
have at most n digits. We observe that the binary expansion of a
scattered number will start with 10. Lets call f(n) the number of
ways to write a n-digit (base 2) number. Observe that f(1)=1 and
f(2)=1. If we take o the rst two digits of a scattered number (which
will always be 10) we will be left with another scattered number, or
a scattered number with zeroes in front of it. This means that for
n>2, f(n)=f(n-2)+f(n-3)+. . . +f(3)+f(2)+f(1)+1. But f(n-1)=f(n-3)+f(n-
4)+. . . +f(3)+f(2)+f(1)+1. So f(n)=f(n-1)+f(n-2). But since the rst two
numbers of this sequence are both 1 (f(1)=1 and f(2)=1), this means that
the f(n) is equal to the nth Fibonacci number. In conclusion, the number
of scattered numbers below 2n is equal to the nth Fibonacci number.
7. It is obvious that a convex polyehdron can be represented as a planar
graph, with the same amount of faces, edges and vertices. Now consider
the dual of this graph, which will have as vertices the rst graphs faces,
as edges the rst graphs edges and as faces the rst graphs vertices. We
just need to prove that the nal graph has at least two vertices with the
same degree. But it is known that every simple graph with at least two
vertices has at least two vertices with the same degree. Here is the proof:
Lets call n the number of vertices of a graph. All degrees of a graph are
between 1 and n-1, inclusive (because a vertice can connect with at most
all the other vertices, which are n-1, and at least with one other), but there
are n vertices. Due to the pidgeonhole principle, there are at least two
vertices with the same degree. This means that the convex polyhedron
has at least two faces with the same number of edges.
6
8. If (a-(a mod b))/b=c, a has c positive multiples below or at c. For ex-
ample 2 has (100-0)/2=50 multiples below or at 100. In order to nd
all the prime-like numbers below 100, we need to do the same with 2,3,5
and 99 (100-1=99 because we want the numbers below 100) and sub-
tract that from 99. But we will double-count their common-multiples. So
what we want to do is add the multiples of 2,3 and 5 below 100, sub-
tract the multiples of 6,10,15 because we added them twice and add the
multiples of 30, because we subtracted them twice. And that gives us:
99-(98/2+99/3+95/5-96/(2*3)-90/(2*5)-90/(3*5)+90/(2*3*5) = 99-73 =
26. In the same manner, the prime-like numbers below 1000 are 266. Now,
very prime-likes are those who are not divisible by 2,3,5,7,11,13. But that
doesnt mean we cant use the same thought! We just have to adjust the
calculations, and, voila! The very prime-like numbers below 90000 are
17261 (excuse any miscalculations, the way I did it was similar to the
above) and the very prime-like numbers below 1010 are 1918081917 (I
know that it was asked as an approximatation, but I thought that since I
could calculate it, why not?). But we cannot hope to calculate the very
prime-like numbers below 10100! Well have to settle down for an ap-
proximatation. Let f(x) the oor function. That means we can write (a-(a
mod b))/b=c as f(a/b)=c which is much easier. Let also N=10100-1 for
ease of calculations. As we did before, we have that the number of very
prime-likes below 10100 is:
N - (f(1/2) + f(1/3) + f(1/5) + f(1/7) + f(1/11) + f(1/13)) + (f(1/6) + f(1/10) +
f(1/14) + f(1/15) + f(1/21) + f(1/22) + f(1/26) + f(1/33) + f(1/35) + f(1/39)
+ f(1/55) + f(1/65) + f(1/77) + f(1/91) + f(1/143)) - (f(1/30) + f(1/42) +
f(1/66) + f(1/70) + f(1/78) + f(1/105) + f(1/110) + f(1/130) + f(1/154) +
f(1/165) + f(1/182) + f(1/195) + f(1/231) + f(1/273) + f(1/286) + f(1/385)
+ f(1/429) + f(1/455) + f(1/715) + f(1/1001) +) + (f(1/210) + f(1/330) +
f(1/390) + f(1/462) + f(1/546) + f(1/770) + f(1/858) + f(1/910) + f(1/1155)
+ f(1/1365) + f(1/1430) + f(1/2002) + f(1/2145) + f(1/3003) + f(1/5005)) -
(f(1/2310) + f(1/2730) + f(1/4290) + f(1/6006) + f(1/10010) + f(1/15015)) +
(f(1/30030))
which is almost equal to:
N - (1/2 + 1/3 + 1/5 + 1/7 + 1/11 + 1/13) + (1/6 + 1/10 + 1/14 + 1/15
+ 1/21 + 1/22 + 1/26 + 1/33 + 1/35 + 1/39 + 1/55 + 1/65 + 1/77 + 1/91
+ 1/143) - (1/30 + 1/42 + 1/66 + 1/70 + 1/78 + 1/105 + 1/110 + 1/130 +
1/154 + 1/165 + 1/182 + 1/195 + 1/231 + 1/273 + 1/286 + 1/385 + 1/429 +
1/455 + 1/715 + 1/1001) + (1/210 + 1/330 + 1/390 + 1/462 + 1/546 + 1/770
+ 1/858 + 1/910 + 1/1155 + 1/1365 + 1/1430 + 1/2002 + 1/2145 + 1/3003
+ 1/5005) - (1/2310 + 1/2730 + 1/4290 + 1/6006 + 1/10010 + 1/15015) +
(1/30030) (for approximatation +-32)
which is equal to:
5760*N/30030 +-32 = 0.191808192*N +-32
7
If N is 10100-1, there are about 0.1918*(10100-1) very prime-likes below it.
9. If we do not care about leaving an empty committee, the answer is the
following: Each person can join either one or two (a pair of) committees.
The number of ways to join only one committee is three and the number
ways to join a pair of committees is three. Thus the arrangement of
people in committees so that no person is in all three committees, but
a committee can be empty, has 610 posibilities. We must then count
how many of those ways leave empty committees. If a committee is left
empty, each person has three options: to go to either of the remaining
two, or to both. Thus the number of ways to leave a committee empty
must be 3*310. But we have double-counted the posibilities to leave two
committees empty, which can be done in 3 ways. Thus, the nal answer is
that 10 people can join 3 committees so that no person is in all three and
so that no committee is empty in 610 - 3*310 + 3 = 60289032 ways.
10. The short answer is yes, the ea will catch the wallaby, supposing of course
that the rubber band does not break and the ea does not get tired. Lets
name the ratio of one mile to one inch a ( a=63360 ). In order to solve the
problem, well count based on the rubber band (the wallaby will always
be at 1 and the ea will start at 0). At step S, (after the wallaby has
jumped), the ea will jump 1/(a*S) on the rubber band. For example, if
a is 2, the following will happen (The moves of the wallaby are included
as an increase in S):
The ea jumps to the 1/2 of the rubber band.
The ea jumps another 1/4, landing to the 3/4 of the rubber band.
The ea jumps another 1/6, landing to the 5.5/6 of the rubber band.
The ea can jump another 1/8, landing to the 1.041667 of the rubber
band, which means that at this step, it catches the wallaby.
It is clear that at each step, the ea will land on the 1/a + 1/(2*a) + . . . +
1/(S*a) of the rubber band, which is equal to 1/a*(1 + 1/2 + . . . + 1/S) which
is also equal to H(S)/a, where H(S) is the Sth harmonic number. Since the
harmonic numbers have no limit, and because a is constant, we can infer that
H(S)/a will for some S surpass 1, meaning that the ea will at some step catch
the wallaby.
8

S-ar putea să vă placă și