Sei sulla pagina 1di 116

Art of Problem Solving

WOOT
Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others.

Class Transcript 01/21 - Modular Arithmetic B


nsato 7:30:52 pm WOOT 2012-13: Modular Arithmetic B nsato 7:30:57 pm In today's lecture, we will continue our look at number theory and modular arithmetic. The first half of our class will focus on the use of Fermat's Little Theorem and Euler's Theorem. In the second half, we'll cover some more general number theory problems that are typical for this level. nsato 7:31:09 pm We'll start with a nice warm-up exercise. nsato
7:31:14 pm

nsato 7:31:35 pm How can we find the missing digit? nsato 7:32:25 pm We can try to find the sum of the 9 digits. If all 10 digits were present, then their sum would be 0 + 1 + 2 + ... + 9 = 45. However, we can't find the sum of the 9 digits without computing 2^29. Is there an easier way? delta1 mod 9
7:33:00 pm

Dunedubby 7:33:00 pm find the number mod 9 Dunedubby 7:33:00 pm the number mod 9 is the same as the sum of the digits mod 9 A123456789 7:33:00 pm Compute 2^29 (mod 9). aleph0 mod 9
7:33:00 pm

Binomial-theorem 7:33:00 pm Find the number modulo 9 to figure out which number we do not have in our sum brian22 mod 9!
7:33:02 pm

nsato 7:33:06 pm We know that a number and the sum of its digits are congruent modulo 9, so we can reduce 2^29 modulo 9. What do we get? AndroidFusion 5 willwang123 5 ProbaBillity 5 zheng 5
7:34:08 pm

7:34:08 pm

7:34:08 pm

7:34:08 pm

mjseaman1 5 bestbearever 5

7:34:14 pm

7:34:14 pm

Binomial-theorem 7:34:18 pm phi(9)=6, therefore 2^(29)=2^5=32=5(mod 9)

nsato

7:34:20 pm

nsato 7:34:39 pm So what is the missing digit? mcdonalds106_7 4 s.homberg 4 math-fan 4


7:35:25 pm

7:35:25 pm

7:35:25 pm

distortedwalrus 4 zheng 4

7:35:25 pm

7:35:34 pm

PiCrazy31415 4 mjseaman1 4 MathisFun! 4

7:35:34 pm

7:35:34 pm

7:35:34 pm

Binomial-theorem 4!

7:35:34 pm

nsato 7:35:47 pm The missing digit must be congruent to -5 modulo 9, so the missing digit is 4. ProbaBillity 7:36:01 pm and indeed it is 4; 2^29 = 536870912. nsato 7:36:03 pm Indeed, 2^29 = 536870912. All digits are present except 4. nsato
7:36:16 pm

nsato 7:36:28 pm How do we start? ProbaBillity mod 1000


7:37:01 pm

trophies 7:37:01 pm reduce it modulo 1000 bobcat120 7:37:01 take it mod 1000 PiCrazy31415 modulo 1000 bestbearever mod 1000
pm

7:37:01 pm

7:37:01 pm

nsato 7:37:04 pm Finding the last three digits of a number is an indication to work modulo 1000. nsato
7:37:10 pm

nsato 7:37:14 pm How can we reduce this? pgmath 7:37:53 Euler's Theorem ProbaBillity
pm

7:37:53 pm

Euler's theorem nsato 7:38:15 pm We can use Euler's Theorem. trophies 7:38:20 pm 9^400= 1(mod 1000) by euler's theorem A123456789 7:38:20 pm 9^400=1(mod 100), so compute 8^7(mod 400). distortedwalrus 7:38:20 pm note that 9^400 == 1 (mod 1000) nsato
7:38:30 pm

mentalgenius 7:38:44 pm well phi(1000) = 400, so we want 8^7 mod 400 Dunedubby 7:38:44 pm find 8^7 mod phi(1000) nsato
7:38:54 pm

nsato 7:39:02 pm How can we do this? diger 7:40:20 pm eulers theorem again distortedwalrus 7:40:20 pm use euler's theorem again nsato 7:40:22 pm Euler's Theorem doesn't help, because the exponent 7 is far less than phi(400). brian22 7:40:31 8^3*8^3*8
pm

trophies 7:40:31 pm Note that 8^3=512= 112(mod 400). Thus, we have 112*112*8= 112*96= 352(mod 1000) nsato 7:40:44 pm Probably the easiest way to compute this is to look at the first few powers of 8 modulo 400. We do this by starting with 1, then multiplying each term by 8 and reducing modulo 400. nsato 7:41:05 pm This gives us 1, 8, 64, 112, 96, 368, 144, 352. nsato 7:41:16 pm (Or we can combine powers, as suggested above.) mentalgenius 7:41:28 pm 8^3 = 512 112 (mod 400) nsato
7:41:37 pm

nsato 7:41:56 pm How can we compute this residue? nsato 7:42:29 pm We have gone as far as we can with Euler's Theorem. We can try to compute this by looking at the first few powers of 9 modulo 1000, but there is an easier way. nsato 7:42:50 pm Note that 9 is close to 10, and 1000 = 10^3. Can we exploit this fact somehow? trophies 7:43:28 pm binomial theorem! TheLittleOne
7:43:28 pm

use binomial expansion s.homberg 7:43:28 binomial theorem


pm

mentalgenius 7:43:32 pm write 9 as (10 - 1) -- in the expansion, most of the terms cancel out trophies 7:43:33 pm (10-1)^352, use the binomial theorem to find the last three digits delta1 7:43:33 (10-1)^352 nsato
pm

7:43:39 pm

nsato 7:43:52 pm What does this reduce to modulo 1000? mentalgenius 7:44:47 just the last 3 terms PiCrazy31415 081 bestbearever 081 King6997 81
pm

7:44:56 pm

7:44:56 pm

7:44:56 pm

soy_un_chemisto 81 nsato
7:45:01 pm

7:44:56 pm

nsato 7:45:09 pm Therefore, the last three digits of 9^(8^7) are 081. nsato
7:45:39 pm

nsato 7:46:03 pm We could verify that n^5 - n is divisible by 30 for n = 0, 1, 2, ..., 29. But what is an easier way? mjseaman1 7:46:49 pm Look at it mod 2,3,5 trophies 7:46:49 pm show that it is divisible by 2, 3, and 5! willwang123 7:46:49 pm take mods 2, 3, and 5 ProbaBillity 7:46:49 pm we test mod 2, 3, and 5 steve314 7:46:49 pm show its always divisible by 2,3,5 zheng 7:46:52 pm Use mod 2,3,5 to prove divisibility by 2,3 and 5 PiCrazy31415 30 = 2 x 3 x 5
7:46:54 pm

distortedwalrus 7:46:58 pm we want n^5 == n (mod 2), (mod 3), and (mod 5)

nsato 7:47:03 pm We use the prime factorization of 30 = 2 x 3 x 5. Hence, it suffices to show that n^5 - n is divisible by 2, 3, and 5, for all integers n. nsato 7:47:22 pm Is n^5 - n divisible by 2? ProbaBillity 7:48:20 pm yes, we test both 0 and 1 and they work mod 2. bobcat120 7:48:20 pm yes, n^5 and n are the same parity pgmath 7:48:20 pm Yes, n^5 and n have the same parity trophies 7:48:20 pm Yes... we can check 1(mod 2) and 0(mod 2), which both work. zheng 7:48:20 pm Yes. If n is odd we get odd-odd=even. If n is even we get even-even=even Showpar 7:48:20 pm Yes, n^5 and n have the same modulo 2 residue mentalgenius 7:48:20 pm two cases - n even or n odd; if n even, trivial; if n odd, 1-1 0 (mod 2), so yes aleph0 7:48:20 pm n^5 and n must have same parity nsato 7:48:23 pm Yes. By reducing modulo 2, we can check for n = 0 and n = 1. nsato 7:48:26 pm Is n^5 - n divisible by 3? ProbaBillity 7:49:20 pm 0, 1, and 2 work. So it is. trophies 7:49:20 pm Yes, we can check 0(mod 3), 1(mod 3), and 2(mod 3) mjseaman1 7:49:20 pm Yes check n==0,1,2 mod 3 distortedwalrus 7:49:20 pm yes by checking 0, 1, and 2 binmu 7:49:20 pm yes, check 0, 1 and 2(mod3) nsato 7:49:24 pm Yes. By reducing modulo 3, we can check for n = 0, 1, and 2. nsato 7:49:27 pm Is n^5 - n divisible by 5? ProbaBillity 7:50:15 pm 0, 1, 2, 3, and 4 work. Thus 5|n^5 - n. And we are done! pgmath 7:50:15 Yes, by FLT
pm

zheng 7:50:15 pm Fermat's little gives n-n so yes PiCrazy31415 7:50:15 pm yes, fermat's little theorem brian22 7:50:15 pm by Fermat's little theorem Binomial-theorem
7:50:15 pm

mjseaman1 7:50:15 Yes by Fermat

pm

trophies 7:50:15 pm Yes, we can use Fermat's little theorem... n^5=n(mod 5)-n(mod 5)= 0(mod 5) diger 7:50:15 pm fermats little theorem nsato 7:50:18 pm Yes. By reducing modulo 5, we can check for n = 0, 1, 2, 3, and 4. We can also simply cite Fermat's Little Theorem for the prime p = 5. nsato 7:51:04 pm Hence, n^5 - n is divisible by 2 x 3 x 5 = 30 for all integers n. nsato 7:51:09 pm This simple problem illustrates an important principle. If you want to prove that a number is divisible by n, it helps to look at the individual prime powers of n. nsato
7:51:33 pm

nsato 7:52:25 pm We can start by trying to reduce the given expression modulo 2000, but all the bases are already less than 2000. So what else can we do? ProbaBillity 7:53:37 pm divisible by 16 and 125 zheng 7:53:37 pm use mod 16 and mod 125 brian22 7:53:37 mod 16 and 125
pm

pgmath 7:53:37 pm modulo 2^4 = 16 and 5^3 = 125 mjseaman1 7:53:37 pm Reduce it mod 16 and 125. If it is zero mod both, it is divisible by 2000. distortedwalrus 7:53:37 pm note that 2000=2^4*5^3 Binomial-theorem 7:53:37 pm 2000=2*1000=2^4*5^3 look mod 16 and mod 125 nsato 7:53:49 pm The prime factorization of 2000 is 2^4 x 5^3, so we look at the factors 2^4 = 16 and 5^3 = 125 separately. nsato 7:54:00 pm How does the given expression reduce modulo 16? PiCrazy31415 0 modulo 16
7:55:21 pm

diger 7:55:21 pm 9^n - 9^n +12^n -(-4)^n == 0 Binomial-theorem 7:55:21 pm 9^n-9^n+(-4)^n-(-4)^n=0(mod 16) YAY! brian22 0
7:55:21 pm

zheng 7:55:21 pm 9^n-9^n+12^n-12^n=0 distortedwalrus it's 0 (mod 16)


7:55:21 pm

tc1729 7:55:25 pm 9^n - 9^n + 12^n - 12^n = 0 nsato


7:55:31 pm

nsato 7:55:40 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 16. nsato 7:55:51 pm How does the given expression reduce modulo 125? mentalgenius 0 pgmath
7:56:59 pm

7:56:59 pm

bestbearever 7:56:59 0 mod 125 (cancels)

pm

Binomial-theorem 7:56:59 pm (-4)^n-25^n+25^n-(-4)^n so again 0 PiCrazy31415 0 modulo 125


7:56:59 pm

tc1729 7:56:59 pm 121^n - 25^n + 25^n -121^n = 0 (mod 125) trophies 7:56:59 pm 121^n-25^n+25^n-121^n= 0(mod 125). zheng 7:56:59 pm (-4)^n-25^n+25^n-(-4)^n=0 A123456789 7:56:59 pm (-4)^n-25^n+25^n-(-4)^n Showpar 7:56:59 pm (-4)^n-25^n+25^n-(-4)^n ProbaBillity 7:57:03 pm (-4^n) - 25^n + 25^n - (-4^n) 0 (mod 25) DOUBLE YAY! WE DONE! that was really nice. nsato
7:57:07 pm

nsato 7:57:18 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 125. nsato 7:57:27 pm We conclude that 121^n - 25^n + 1900^n - (-4)^n is divisible by 2000 for all positive integers n. nsato 7:57:51 pm Again, all we had to do was look at individual prime powers, which helps break down the expression. nsato
7:58:02 pm

nsato 7:58:14 pm What can we do with this equation? nsato 7:59:22 pm All those fifth powers... does that remind us anything?

Dunedubby 7:59:52 take it mod 30! ProbaBillity 7:59:52 n^5 n (mod 30)

pm

pm

Dunedubby 7:59:52 pm we know that n^5 == n mod 30 nsato 8:00:12 pm We've shown that n^5 is congruent to n modulo 30, so we can reduce the given equation modulo 30. What does that tell us about n? ProbaBillity 8:01:57 n 24 (mod 30)
pm

mentalgenius 8:01:57 n 24 (mod 30)

pm

zheng 8:01:57 pm n=133+110+84+27=24 (mod 30) steve314 8:01:57 n==24 mod 30


pm

brian22 8:01:57 pm 13+20+24+27==22==n mod 30 nsato


8:02:09 pm

nsato 8:03:04 pm That helps narrow down the possible values of n. What else can we do to help find n? ged3.14 8:03:56 pm use units digits along with bounding Showpar 8:03:56 Set bounds on n. ged3.14 8:04:01 then bound n
pm

pm

Dunedubby 8:04:01 pm set an upper bound and lower bound nsato 8:04:02 pm Let's try to find some bounds on n. What's one obvious bound for n? PiCrazy31415 8:04:57 larger than 133 zheng n>133
8:04:57 pm pm

steve314 n>133 Showpar n>133

8:04:57 pm

8:04:57 pm

flappingwings 8:04:57 pm n is definitely greater than 133 diger n>133


8:04:57 pm

bestbearever 8:04:57 greater than 133 Binomial-theorem n>133 yangdongyan n>133

pm

8:04:57 pm

8:04:57 pm

nsato 8:04:59 pm Clearly n > 133. nsato 8:05:11 pm Now we need to find an upper bound on n. nsato 8:05:32 pm The largest term in the left-hand side is 133^5, so we compare all the other terms to 133^5. nsato 8:05:42 pm We know that 110^5 < 133^5. nsato 8:05:53 pm Can we compare 84^5 + 27^5 and 133^5? PiCrazy31415 8:07:17 pm 84^5 + 27^5 < 133^5 flappingwings 8:07:17 pm since 84+27 is less than 133, the former is less than the latter Dunedubby 8:07:17 pm 84^5 + 27^5 < (84 + 27)^5 < 133^5 King6997 8:07:17 pm 84^5+27^5 < 133^5 tc1729
8:07:17 pm

distortedwalrus 8:07:17 pm well it's less than (84+27)^5=111^5 Showpar 8:07:19 pm 84^5+27^5<(84+27)^5=111^5<133^5 binmu 8:07:23 pm yes, 84^5+27^5<133^5 nsato 8:07:26 pm Note that 84^5 + 27^5 < (84 + 27)^5 = 111^5 < 133^5. nsato
8:07:36 pm

Binomial-theorem 8:07:59 pm you can try to aproximate that? nsato 8:08:01 pm What we would like to do now is take the fifth root of both sides. The problem is we would get the fifth root of 3, and it's not clear what to do with this number. mentalgenius approximate it
8:08:33 pm

nsato 8:09:12 pm How could we try to get an approximation on the fifth root of 3? We want an upper bound, and it probably doesn't have to be too sharp. nsato 8:11:18 pm What we could do is look at fifth powers of integers, and see if we can find a ratio of two fifth powers that is close to 3. nsato 8:11:36 pm The first few fifth powers are 1^5 = 1, 2^5 = 32, 3^5 = 243, 4^5 = 1024, 5^5 = 3125, and 6^5 = 7776, and so on. Do we see any fifth powers that approximately have a ratio of 3? flappingwings 4^5 and 5^5 diger 8:12:46 (5/4)^5 PiCrazy31415 5^5 , 4^5
8:12:46 pm

pm

8:12:46 pm

Binomial-theorem

8:12:46 pm

5^5 and 4^5 ProbaBillity 3125/1024 steve314 5^5/4^5


8:12:48 pm

8:12:51 pm

distortedwalrus yeah 5^5/4^5 tc1729 8:12:55 5^5/4^5

8:12:51 pm

pm

nsato 8:13:01 pm We see that 5^5/4^5 = 3125/1024 is approximately 3. nsato 8:13:14 pm Even better, this ratio is slightly larger than 3, which means we can use this ratio to get an upper bound on n: nsato
8:13:27 pm

nsato 8:13:41 pm Therefore, n < 5/4 x 133 = 166 + 1/4. nsato 8:14:06 pm We also know that n > 133 and n is congruent to 24 modulo 30. Do we have enough information to determine n? trophies 8:14:58 pm thus, n must be 144 bestbearever 144 AndroidFusion n = 144 mjseaman1 Yes n is 144
8:14:58 pm

8:14:58 pm

8:14:58 pm

Binomial-theorem 8:14:58 pm n=144, 174, ... but 174>166 so n= 144 mentalgenius yes, n = 144 A123456789 n=144
8:14:58 pm

8:14:58 pm

nsato 8:15:01 pm The only n satisfying 133 < n < 166 + 1/4, and n congruent to 24 modulo 30 is n = 144. nsato
8:15:39 pm

nsato 8:16:03 pm What makes this sum difficult to deal with? ProbaBillity 8:17:17 the absolute values delta1 8:17:17 absolute value
pm pm

diger 8:17:17 pm absolute values math-fan 8:17:17 pm all absolute values

Showpar 8:17:17 Absolute values

pm

Binomial-theorem 8:17:18 absolute values ooops

pm

nsato 8:17:20 pm The absolute value signs make this sum difficult to deal with. How can we deal with the absolute value signs? Is there a convenient way of getting rid of them? tc1729 8:18:27 wlog a_i > b_i
pm

King6997 8:18:27 WLOG ai>bi

pm

A123456789 8:18:27 pm Let all the a_i be larger. distortedwalrus 8:18:27 say a_i>b_i for all i
pm

mentalgenius 8:18:27 pm just assume ai > bi for all i Dunedubby 8:18:27 pm suppose that a_i > b_i for all i, sorry nsato
8:18:35 pm

nsato

8:18:53 pm

nsato 8:19:14 pm Each term in this sum is 1 or 6. How can we use this information? Binomial-theorem 8:20:37 pm find how many 1's and how many 6's maybe? nsato 8:20:41 pm There are 999 terms, so let k of them be equal to 1, so the remaining 999 - k are equal to 6. nsato
8:21:01 pm

nsato

8:21:12 pm

PiCrazy31415 8:22:20 pm a_i and b_i are all in 1,2,3,...,1998 yankeesfan 8:22:20 pm they make up the set {1,2,...,1998} ProbaBillity 8:22:20 pm they make up the numbers from 1 to 1998 Binomial-theorem 8:22:20 pm they're from the set {1,2,..., 1998} Dunedubby 8:22:20 pm numbers from 1 - 1998 diger 8:22:24 pm numbers 1 through 1998 nsato
8:22:26 pm

nsato 8:22:37 pm So can we turn this fact into a related sum? PiCrazy31415 8:25:00 pm a_1 + b_1 +...+ a_999 + b_999 = 999 x 1999 ProbaBillity 8:25:00 pm we know a_1 + b_1 + ... + a_999 + b_999 = 1 + 2 + ... + 1998 = 999*1999 diger 8:25:00 pm a_1 + a_2 + ... +a_999 + b_1 + ... +b_999 = 999*1999 = 1997001 mentalgenius 8:25:00 pm the sum of the ai and bi is 1998*1999/2 nsato
8:25:10 pm

nsato

8:25:25 pm

nsato 8:25:32 pm What can we do with these two equations? brian22 addemup
8:25:54 pm

bestbearever add them willwang123 add them

8:25:54 pm

8:25:54 pm

pgmath 8:25:54 Add them delta1 8:25:54 add them King6997 Add them

pm

pm

8:25:54 pm

nsato 8:25:59 Of course! nsato

pm

8:26:06 pm

nsato 8:26:36 pm Does this equation say anything about the value of k? A123456789 8:27:27 k must be odd.
pm

trophies 8:27:27 pm wait that means that k is odd! AndroidFusion 8:27:27 pm k has to be odd for the right side to be even distortedwalrus it is odd
8:27:27 pm

trophies 8:27:27 pm k=1(mod 2), or in other words, k is odd. steve314 k is odd


8:27:27 pm

Showpar 8:27:27 pm LHS must be even, so k is odd mjseaman1 k is odd


8:27:27 pm

nsato 8:27:31 pm The left-hand side must be even. Therefore, 5k must be odd, which means k must be odd. nsato 8:27:44 pm Let k = 2n + 1 for some integer n. nsato 8:27:47 Now what?
pm

bestbearever 8:28:55 pm combine that with the very first 5994-5k binmu 8:28:55 plug back in
pm

AndroidFusion 8:28:55 pm so 5994 - 5k = 5989 - 10k is congruent to 9 mod 10 ProbaBillity plug it in


8:28:55 pm

willwang123 8:28:55 substitute that in

pm

binmu 8:28:55 pm 5994-5(2n+1)=5994-10n+5==9(mod10) Dunedubby 8:29:00 pm so we're done! - plugging in the first equation 5994 - 5k means that it must end in a 9 nsato 8:29:02 pm We can substitute into the equation above, to get nsato
8:29:06 pm

nsato 8:29:28 pm It looks like we can conclude that this number always ends in 9, but there is a small catch. What's the catch? ProbaBillity 8:30:26 pm we must verify that n is less than 599 A123456789 8:30:26 It could be negative. willwang123 or if n>=599
pm

8:30:26 pm

Calebhe1290 8:30:26 pm 5989<10n wouldn't work nsato 8:30:32 pm What if n is sufficiently high so that 5989 - 10n is negative? For example, if n = 600, then 5989 - 10n = -11, which ends in 1. Binomial-theorem 8:31:45 pm but that's impossible because absolute values impleis it's greater than 0 distortedwalrus 8:31:45 pm what but the sum will always be positive willwang123 8:31:45 pm we assumed wlog that (a_k-b_k) was positive

mjseaman1 8:31:45 pm The sum is positive because it the sum of absolute values each of which is 1 or 6. s.homberg 8:31:45 pm each b_i > a_i, so the sum must be positive Dunedubby 8:31:51 absolute value
pm

nsato 8:31:58 pm The original sum was expressed as a sum of absolute values, which must be nonnegative. And a nonnegative number of the form 5989 - 10n must end in 9, so we are home free. ProbaBillity 8:32:07 pm But it is, since 2n + 1 = k <= 999. nsato 8:32:11 pm That works too. nsato 8:32:24 pm You just want to make sure that you argue this last step clearly. nsato
8:39:10 pm

nsato 8:39:32 pm How can we proceed? tc1729 8:40:22 contradiction?


pm

ProbaBillity 8:40:22 pm proof by contradiction? nsato 8:40:26 pm We want to prove that something can't happen, so let's try argument by contradiction. pgmath 8:40:38 pm d must be a or b Calebhe1290 8:40:38 pm (a,b) is one of (d,2) (d,5) (d,13) trophies 8:40:38 pm we have to prove that one or more of 2d-1, 5d-1, or 13d-1 are not perfect squares. nsato 8:40:41 pm Note that 2 x 5 - 1 = 9, 2 x 13 - 1 = 25, and 5 x 13 - 1 = 64 are all perfect squares. brian22 8:40:57 pm 2d-1=x^2, 5d-1=y^2, 13d-1=z^2 nsato 8:41:00 pm Suppose 2d - 1 = x^2, 5d - 1 = y^2, and 13d - 1 = z^2, for some positive integers x, y, and z. tc1729
8:41:14 pm

nsato 8:41:30 pm Does the first equation 2d - 1 = x^2 tell us anything interesting about x? zheng 8:42:14 x is odd A123456789 x is odd.
pm

8:42:14 pm

Binomial-theorem x is odd diger 8:42:14 x is odd steve314 x is odd


pm

8:42:14 pm

8:42:14 pm

bobcat120 8:42:14 x must be odd AndroidFusion x is odd

pm

8:42:14 pm

nsato 8:42:17 pm The number 2d - 1 is always odd, so x is odd. nsato 8:42:36 pm Since x is odd, what can we do? Calebhe1290 8:43:27 write it as 2n+1 PiCrazy31415 let x = 2n+1
pm

8:43:27 pm

Dunedubby 8:43:27 express as 2n + 1 willwang123 say it's 2n+1

pm

8:43:27 pm

binmu 8:43:27 x=2n+1 A123456789 x=2n+1

pm

8:43:27 pm

AndroidFusion 8:43:27 substitute it for 2n+1

pm

nsato 8:43:30 pm Whenever we find some fact like this, we express it algebraically. nsato 8:43:36 pm Let x = 2n + 1 for some integer n. Then 2d - 1 = x^2 = (2n + 1)^2 = 4n^2 + 4n + 1. zheng 8:44:41 d=2n^2+2n+1
pm

ProbaBillity 8:44:41 pm so d = 2n^2 + 2n + 1. Time to plug more things in! Binomial-theorem d=2n^2+2n+1
8:44:41 pm

soy_un_chemisto 8:44:41 then d is 2n^2 + 2n + 1

pm

Dunedubby 8:44:41 pm therefore d = 2n^2 + 2n + 1 nsato 8:44:46 pm Then d = 2n^2 + 2n + 1. What does this tell us about d? esque 8:45:42 d is odd distortedwalrus d is odd soy_un_chemisto d is odd Calebhe1290 d is odd zheng 8:45:42 d is odd
pm

8:45:42 pm

8:45:42 pm

8:45:42 pm

pm

nsato 8:45:51 pm This equation says that d is odd. nsato 8:45:57 pm Then what can we say about 5d - 1 = y^2 and 13d - 1 = z^2?

pgmath 8:46:50 pm 5d -1 and 13d - 1 are both even delta1 8:46:58 pm y and z are even Binomial-theorem y and z are even
8:46:58 pm

Showpar 8:46:58 pm y and z are both even mjseaman1 8:46:58 y and z are even
pm

bestbearever 8:46:58 y and z are even

pm

mentalgenius 8:46:58 pm y and z are both even nsato 8:47:04 pm Since d is odd, 5d - 1 = y^2 and 13d - 1 = z^2 are even. Hence, y and z are even. nsato 8:47:25 pm Again, let's express these statements algebraically, by introducing variables. brian22 8:47:44 y=2k, z=2j
pm

nsato 8:47:52 pm Let y = 2u and z = 2v. Then 5d - 1 = y^2 = 4u^2 and 13d - 1 = 4v^2. nsato 8:48:03 pm What can we do with these equations? delta1 8:49:01 subtract them
pm

A123456789 8:49:01 pm Subtract the first from the second. diger 8:49:01 subtract
pm

binmu 8:49:01 subtract Calebhe1290 subtract

pm

8:49:01 pm

ProbaBillity 8:49:01 subtract them

pm

pgmath 8:49:06 pm subtract the first from the second nsato 8:49:07 pm We can subtract them, to get 8d = 4v^2 - 4u^2, so 2d = v^2 - u^2. nsato 8:49:18 pm What can we do with this equation? brian22 factor!
8:50:15 pm

willwang123 8:50:15 diff of squares binmu 8:50:15 2d=(v+u)(v-u) Calebhe1290 factor the rhs PiCrazy31415 factorize
pm

pm

8:50:15 pm

8:50:15 pm

soy_un_chemisto 8:50:17 factor the right side

pm

nsato 8:50:19 pm We can factor the right-hand side, to get 2d = (v + u)(v - u). nsato 8:50:25 pm Can we say anything interesting about the factors v + u and v - u? steve314 8:51:19 pm they have the same parity Dunedubby 8:51:19 pm they are of the same parity diger 8:51:19 same parity
pm

vjnmath 8:51:19 pm They must be the same parity Binomial-theorem same parity
8:51:19 pm

pgmath 8:51:19 pm they have the same parity ProbaBillity same parity
8:51:24 pm

brian22 8:51:24 same parity

pm

nsato 8:51:28 pm Either v + u and v - u are both even or both odd (since their difference 2u is an even number). nsato 8:51:55 pm What happens if both v + u and v - u are even? binmu 8:52:47 pm 2d is divisible by 4 bestbearever 8:52:47 pm 2d is congruent to 0 mod 4 zheng 8:52:47 pm then we have that d must be even, contradiction distortedwalrus 8:52:47 pm both sides are multiples of 4, so d is even PiCrazy31415 8:52:47 pm then 2d is divisible by 4, but d is odd diger 8:52:47 pm 2d is a multiple of 4 and d is even willwang123 d is even too
8:52:47 pm

cire_il 8:52:47 pm then it means d is even Binomial-theorem 8:52:47 pm then d is even, but that's a contradiction Dunedubby 8:52:47 pm then d is even, a contradiction nsato 8:52:53 pm If both v + u and v - u are even, then 2d = (v + u)(v - u) is a multiple of 4, which is impossible because d is odd. nsato 8:53:03 pm What happens if both v + u and v - u are odd? ProbaBillity 8:53:56 pm another contradiction. LHS is 2d which is even, RHS is odd.

diger 8:53:56 2d is odd

pm

bestbearever 8:53:56 2d must be odd

pm

PiCrazy31415 8:53:56 pm then 2d is odd, contradiction binmu 8:53:56 2d is odd too


pm

Binomial-theorem 8:53:56 pm then 2d must be odd, absurd again tc1729 d is odd


8:53:56 pm

Showpar 8:53:56 pm The RHS is odd, while the LHS is even, contradiction nsato 8:54:00 pm If both v + u and v - u are odd, then 2d = (v + u)(v - u) is odd, which is again impossible because 2d is even. nsato 8:54:22 pm We have obtained a contradiction, so at least one of 2d - 1, 5d - 1, and 13d - 1 is not a perfect square. nsato 8:54:41 pm It can seem simple, but using parity can be a powerful tool. And whenever you find that a number is even (odd resp.), it is usually a good idea to write it in the form 2n (2n + 1 resp.). nsato
8:54:54 pm

nsato 8:55:19 pm You may recall Euclid's proof that there are an infinite number of primes. How can we start? Binomial-theorem 8:55:58 contradiction again!
pm

tc1729 8:55:58 pm assume there are only finitely many PiCrazy31415 8:55:58 proof by contradiction A123456789 8:55:58 Contradiction.
pm

pm

bestbearever 8:55:58 pm assume there are a finite number distortedwalrus 8:55:58 pm we can use a contradiction again nsato 8:56:03 pm We argue by contradiction. zheng 8:56:08 pm assume there are finite primes of the form 3n+2 nsato
8:56:10 pm

nsato 8:56:23 Then what? willwang123 multiply them

pm

8:57:00 pm

mentalgenius 8:57:00 pm multiply them all together and see what you get nsato
8:57:05 pm

nsato 8:57:19 pm What can we say about P? soy_un_chemisto 8:58:14 it's either 0 or 2 mod 3 brian22 8:58:14 pm mod 3, its 0 or 2. ProbaBillity 8:58:20 pm either a multiple of 3 or 2 mod 3 nsato
8:58:21 pm pm

nsato 8:58:28 pm Suppose P reduces to 2 modulo 3. Then what can we say about P? nsato 8:59:14 pm How can we reach a contradiction? zheng 8:59:48 pm It is not divisible by any of the k primes bestbearever 8:59:48 pm it's not divisible by any of the previous primes PiCrazy31415 8:59:48 pm P is relatively prime to p_1, p_2, ... A123456789 8:59:48 pm P is divisible by none of the p_i. willwang123 8:59:48 pm it's not divisible by any of the primes we used nsato 9:00:11 pm So P, by construction, is not divisible by any of the primes p_i we have listed. nsato 9:00:30 pm But these p_i are all the primes that are congruent to 2 modulo 3. nsato 9:00:34 pm What does that say about P? ProbaBillity 9:01:37 pm it can only be divisible by primes congruent to 1 modulo 3 (or 3 itself) diger 9:01:37 pm all prime factors must be in the form 3n or 3n +1 nsato 9:01:44 pm It says that P can only be divisible by primes that are congruent to 1 modulo 3, or the prime 3. nsato 9:01:49 pm But is that possible? mentalgenius 9:02:31 pm the product of numbers congruent to 0 or 1 mod 3 is 0 or 1 mod 3, not 2 mod 3 aleph0 9:02:31 pm no, because then it would be 1 or 0 mod 3 nsato 9:02:36 pm If this is the case, then P is congruent to 0 or 1 modulo 3, but P is congruent to 2 modulo 3. nsato 9:02:58 pm The only case that is left is if P reduces to 0 modulo 3. Unfortunately, there's no good way to handle this case. (We can divide out all the factors of 3, but there's nothing we can say about the resulting quotient.) nsato
9:03:11 pm

Is there any way to fix this argument? Binomial-theorem 9:04:31 pm Consider 3p_1p_2p_3...p_k-1 nsato
9:04:36 pm

tc1729 9:05:07 pm then it has to be 2 mod 3 nsato 9:05:10 pm Then P always reduces to 2 modulo 3, and it is relatively prime to all the p_i, so the argument that we used above now works. nsato 9:05:28 pm More generally, if a and b are relatively prime positive integers, then there are infinitely many primes of the form an + b. This result is known as Dirichlet Theorem, but it is an advanced result and is difficult to prove. nsato 9:06:24 pm The idea of Euclid's proof is also useful for the following problem. nsato
9:06:30 pm

Binomial-theorem 9:07:11 that looks complex

pm

nsato 9:07:12 pm It looks hard; we just have to take things one step at a time. nsato 9:08:19 pm Let's start by looking at small primes. Is the prime 2 in M? nsato 9:08:51 pm Since M contains at least three primes, there is an odd prime in M, say p. Let A = {p}. Then by the condition in the problem, all the prime factors of p - 1 are also in M. binmu 9:09:46 pm so yes 2 is in M zheng 9:09:46 pm Since p is odd, p-1 is even so 2 must be in M Calebhe1290 9:09:46 pm p-1 is even, so 2 must be in M mentalgenius 9:09:52 so 2 has to be in m
pm

nsato 9:09:56 pm But p - 1 is even, so 2 is in M. nsato 9:10:22 pm Can we show that 3 must be in M? Would cases would easily lead to 3 being in M? Dunedubby 3n + 1 prime
9:11:21 pm

math-fan 9:11:21 pm primes that are equal to 1 mod 3 Binomial-theorem p=1(mod 3)


9:11:21 pm

nsato 9:11:28 pm If M contains a prime of the form p = 3n + 1, then take A = {p}. Then all the factors of p - 1 = 3n are also in M, so 3 is in M. nsato
9:12:46 pm

Otherwise, M contains two primes of the form 3n + 2, say p and q. What should we take the set A as in this case? Dunedubby 9:13:20 pm and if not, then the product of two 3n + 2s capu 9:13:33 pm If there is p congruent to 1 mod 3 then done (A= that prime). If not then there are two primes congruent to 2 mod 3 then A= those two and done. Therefore 3 is in M zheng 9:13:33 pm {2, p} or {2,q} diger 9:13:33 A = {p,q} ProbaBillity {p, q}
pm

9:13:33 pm

nsato 9:13:40 pm We can take A = {p, q}. Then all the prime factors of pq - 1 are also in M. mentalgenius 9:13:55 pm pq 1 (mod 3), so 3 is guaranteed to be in M nsato 9:13:56 pm But pq - 1 is congruent to 2 x 2 - 1 = 3, or 0 modulo 3, so pq - 1 is divisible by 3. Therefore, 3 is in M. In either case, 3 is in M. nsato 9:14:39 pm So now we know that 2 and 3 must be in M, which is a great start. What does this tell us? Dunedubby 9:15:42 pm and {2, 3} so 5 is in M as well distortedwalrus 5=2*3-1 is in M
9:15:42 pm

diger 9:15:42 pm 5 must be in M for when A = {2,3} Binomial-theorem 9:15:42 pm 5 is in M, becuase -1+2*3=5 zheng 9:15:42 pm 2*3-1=5 so 5 must be in M nsato 9:15:44 pm Taking A = {2,3}, we see that 5 must be in M. nsato 9:15:56 pm So now we have 2, 3, and 5. Anything else? ProbaBillity 9:16:36 pm 7 is in M, because 3*5 - 1 = 14 which has 7 as a prime factor Dunedubby 9:16:36 and {3, 5} so 7
pm

zheng 9:16:36 pm A={3,5} gives that 7 must be in M diger 9:16:36 pm if A = {3,5} we get 7 soy_un_chemisto 9:16:36 pm 3 * 5 - 1 = 14 so 7 must also be in M binmu 9:16:36 pm 7 is in M, 5*3-1=14 steve314 9:16:36 pm 3*5-1 =14=2*7 so 7 is also in m Calebhe1290 7, {3,5}
9:16:36 pm

yankeesfan 9:16:36 3*5-1=14 so 7

pm

nsato 9:16:39 pm Taking A = {3,5}, we see that 7 must be in M. nsato 9:17:01 pm Hence, the first few primes are in M. We want to prove that M contains all primes. Binomial-theorem 9:17:11 pm we can follow this logic, can't we? nsato 9:17:40 pm We can generate more primes, but we'll need a general approach. Obviously, we can't do this every prime! nsato 9:17:55 pm But primes are hard to get a hold of. nsato 9:17:58 pm What intermediate result can we try proving first? nsato 9:18:15 pm (Here, you might think of Euclid's proof...) PiCrazy31415 M is infinite?
9:18:46 pm

binmu 9:18:46 pm infinite number of primes in M mentalgenius 9:18:46 pm there are an infinite number of primes in M Dunedubby 9:18:46 pm there an infinite number of primes in M diger 9:18:46 pm M has infinite elements aleph0 9:18:46 M is infinite
pm

nsato 9:18:54 pm Let's try proving first that M is infinite. nsato


9:19:03 pm

nsato 9:19:26 pm As a start, let's take A to be all the primes in M, except one, say p_i. (Remember that A must be a proper subset of M.) nsato
9:19:42 pm

nsato

9:20:01 pm

pgmath 9:20:58 pm It is an integer power of p_i diger 9:20:58 pm it is a power of p_i aleph0 9:20:58 pm it's a power of p_i delta1 9:20:58 pm it is some power of p_i nsato
9:21:06 pm

nsato

9:21:28 pm

nsato

9:21:46 pm

nsato 9:22:05 pm To make these equations simpler to work with, we start by omitting the smallest primes. nsato
9:22:17 pm

nsato

9:22:38 pm

nsato

9:22:50 pm

nsato

9:23:01 pm

nsato 9:23:10 pm What can we do with these two expressions for P? steve314 9:23:51 pm set them equal to each other nsato 9:23:55 pm We can equate them. nsato
9:24:02 pm

nsato 9:24:22 pm Let's start analyzing this equation. nsato


9:24:38 pm

nsato 9:24:47 pm Also, P is at least 2 x 3 x 5 x 7. So what can we say about a? mjseaman1


9:27:35 pm

PiCrazy31415 a7 Calebhe1290 a is at least 7

9:27:35 pm

9:27:35 pm

ProbaBillity 9:27:35 a is at least 7 nsato


9:28:05 pm

pm

nsato

9:28:10 pm

nsato 9:28:26 pm So what can we say about b? brian22 b=1


9:29:14 pm

mjseaman1 b=1 delta1 it is 1 capu b=1

9:29:14 pm

9:29:14 pm

9:29:14 pm

PiCrazy31415 b=1

9:29:14 pm

diger 9:29:14 pm divide by 2 to see b=1 nsato 9:29:23 pm Right, because 2^a - 2 has exactly one factor of 2. nsato
9:29:35 pm

nsato 9:29:41 pm How can we analyze this equation? PiCrazy31415 modulo 3 AndroidFusion mod 3
9:30:28 pm

9:30:28 pm

nsato 9:30:29 pm We can try reducing modulo 3. nsato


9:30:35 pm

nsato 9:31:03 pm What does this congruence tell us about a? soy_un_chemisto so a is odd mentalgenius a is odd A123456789 a is odd.
9:31:30 pm

9:31:30 pm

9:31:30 pm

distortedwalrus a is odd Showpar a is odd trophies a is odd

9:31:30 pm

9:31:30 pm

9:31:30 pm

cire_il 9:31:30 a is odd ProbaBillity a = 2n+1

pm

9:31:34 pm

nsato 9:31:38 pm This congruence tells us that a - 1 is even. Let a - 1 = 2n. nsato
9:31:48 pm

nsato 9:31:55 pm What can we do with this equation? distortedwalrus 9:32:44 pm factor as difference of sqaures binmu 9:32:44 pm (2^n-1)(2^n+1) A123456789 9:32:44 pm Factor the right side of the equation. nsato
9:32:47 pm

nsato 9:32:52 pm What does this equation say? delta1 9:33:41 pm the two factors are both powers of 3 mjseaman1 9:33:41 pm Both of the factors on the RHS are powers of 3 nsato 9:33:52 pm This equation says that both 2^n + 1 and 2^n - 1 are powers of 3. nsato 9:33:57 pm But (2^n + 1) - (2^n - 1) = 2. Calebhe1290 9:34:20 pm only 3 and 1 would work

Dunedubby 9:34:31 pm that 2^n + 1 and 2^n - 1 are both powers of 3, so n = 1 A123456789 n=1.
9:34:31 pm

soy_un_chemisto 9:34:31 pm 2^n + 1 and 2^n - 1 must be powers of 3, but this is only satisfied when n = 1 delta1 so n=1
9:34:31 pm

ProbaBillity n =1

9:34:31 pm

nsato 9:34:40 pm The only powers of 3 that differ by 2 are 3 and 1, so n = 1, and a = 3. King6997 but a>=7
9:35:12 pm

binmu 9:35:12 pm but a must be greater than 7 zheng 9:35:12 hoever a>=7
pm

nsato 9:35:14 pm But a is at least 7, which is a contradiction. nsato 9:35:21 pm Therefore, M contains an infinite number of primes. nsato 9:35:27 pm We're actually very close to the end now. nsato
9:35:32 pm

nsato 9:36:06 pm We want to show that p is a factor of some product, minus 1. Does this remind us of any results? nsato 9:37:28 pm We could try using Fermat's Little Theorem. nsato nsato
9:37:37 pm

9:38:37 pm

nsato

9:39:04 pm

Binomial-theorem 9:39:36 pm if p_1=p_2=...p_k=a(mod p)? AndroidFusion 9:39:36 pm let k = p-1 and p1, p2, ... pk be congruent mod p nsato
9:39:48 pm

nsato 9:40:12 pm The primes p_1, p_2, ..., p_{p - 1} must also be distinct from p. But if p was in M, then this argument would not be necessary in the first place. nsato 9:40:50 pm How do we know that there exist p - 1 primes in M that are congruent modulo p?

Dunedubby 9:41:38 pm then use pigeonhole principle - with infinite pigeons AndroidFusion 9:41:38 pm there are an infinite number of primes, so use pidgeon principle Dunedubby 9:41:38 pm by the pigeonhole principle, with infinite pigeons, then at least p-1 must be in some a mod p (where a /== 0) aleph0 9:41:38 pm there are an infinite number, so by pigeonhole AndroidFusion 9:41:38 *pigeonhole principle
pm

mjseaman1 9:41:52 pm M has infinitely many primes, so the pigeonhole principle works nsato 9:41:53 pm We have shown that the set M is infinite. nsato 9:42:00 pm Hence, if we reduce every element in M modulo p, then some nonzero residue must appear an infinite number of times. nsato 9:42:16 pm Then we simply take p - 1 primes p_1, p_2, ..., p_{p - 1} that have this residue. nsato 9:42:35 pm Therefore, M contains every prime p. nsato 9:42:49 SUMMARY
pm

nsato 9:42:51 pm In today's class, we saw how to effectively use Fermat's Little Theorem and Euclid's Theorem. We also saw the power of algebra in number theory problems. We can solve many number theory problems effectively by putting a number in the right form (like expressing an even number in the form 2n), and using factorization. When you see a number theory problem, convert the words into mathematics, and let the equations do the work. nsato 9:43:06 pm Finally, we also saw how to solve problems using argument by contradiction. If a problem asks you to prove that something cannot occur, this is usually a good sign to use contradiction. nsato 9:43:14 pm That's it for today's class. Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others. Copyright 2013 AoPS Incorporated. All Rights Reserved. Contact Us Privacy Terms & Conditions

Art of Problem Solving


WOOT
Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others.

Class Transcript 01/21 - Modular Arithmetic B


Valentin Vornicu 7:31:07 pm WOOT 2012-13: Modular Arithmetic B Valentin Vornicu 7:31:16 pm In today's lecture, we will continue our look at number theory and modular arithmetic. The first half of our class will focus on the use of Fermat's Little Theorem and Euler's Theorem. In the second half, we'll cover some more general number theory problems that are typical for this level. Valentin Vornicu 7:31:24 pm We'll start with a nice warm-up exercise. Valentin Vornicu
7:31:29 pm

Valentin Vornicu 7:31:42 pm How can we find the missing digit? Cogswell 7:32:03 look mod 9 dinoboy 7:32:03 Use mod 9? matholympiad25 mod 9! apple.singer mod 9?
pm

pm

7:32:03 pm

7:32:03 pm

Valentin Vornicu 7:32:24 pm We can try to find the sum of the 9 digits. If all 10 digits were present, then their sum would be 0 + 1 + 2 + ... + 9 = 45. However, we can't find the sum of the 9 digits without computing 2^29. How is mod 9 going to help us here? filetmignon821 7:33:21 pm sum of the digits mod 9 is the same as the number itself mod 9 apple.singer 7:33:21 pm the sum of the digits is congruent to the number itself mod 9 Cogswell 7:33:21 pm number = sum of digits mod 9 sjaelee 7:33:21 pm sum of digits is number mod 9 Valentin Vornicu 7:33:28 pm We know that a number and the sum of its digits are congruent modulo 9, so we can reduce 2^29 modulo 9. What do we get? BOBBOBBOB 7:34:02 2^3=8=-1mod9 krmathcounts 32 mod 9
pm

7:34:02 pm

filetmignon821 7:34:02 pm its the same as 2^5 mod 9, which is 5. antimonyarsenide 5 krmathcounts 5 mod 9 sjaelee 5
7:34:02 pm

7:34:02 pm

7:34:02 pm

Valentin Vornicu

7:34:10 pm

Valentin Vornicu 7:34:12 pm So what is the missing digit? Doink 4


7:35:06 pm

Cogswell 4

7:35:06 pm

apple.singer 4 loquidyE 4 Seedleaf 4 jerrytang 4 nikgran 4

7:35:06 pm

7:35:06 pm

7:35:06 pm

7:35:06 pm

7:35:06 pm

duketip10 4 fireonice 4

7:35:06 pm

7:35:06 pm

Valentin Vornicu 7:35:07 pm The missing digit must be congruent to -5 modulo 9, so the missing digit is 4. Valentin Vornicu 7:35:11 pm Indeed, 2^29 = 536870912. All digits are present except 4. Valentin Vornicu
7:35:20 pm

Valentin Vornicu How do we start? Doink 7:36:02 mod 1000 msinghal mod 1000
pm

7:35:29 pm

7:36:02 pm

yjhan96 7:36:02 mod 1000? Cogswell mod 1000

pm

7:36:02 pm

giratina150 mod 1000

7:36:02 pm

Valentin Vornicu 7:36:05 pm Finding the last three digits of a number is an indication to work modulo 1000. Valentin Vornicu
7:36:09 pm

Valentin Vornicu 7:36:12 pm How can we reduce this? Cogswell 7:36:41 euler's thm
pm

Konigsberg 7:36:41 the euler's theorem filetmignon821

pm

7:36:41 pm

phi(1000)=400, so we find 8^7 mod 400 Konigsberg 7:36:41 pm it repeats every 400 times matholympiad25 7:36:41 pm we find 8^7 mod phi(1000) krmathcounts 7:36:41 pm phi(1000) = 400, and so 9^400 =1 mod 1000 Valentin Vornicu 7:36:49 pm We can use Euler's Theorem. Valentin Vornicu
7:36:55 pm

Valentin Vornicu

7:37:01 pm

Valentin Vornicu 7:37:50 pm Probably the easiest way to compute this is to look at the first few powers of 8 modulo 400. We do this by starting with 1, then multiplying each term by 8 and reducing modulo 400. gurev 352
7:38:05 pm

Valentin Vornicu 7:38:09 pm This gives us 1, 8, 64, 112, 96, 368, 144, 352. Valentin Vornicu
7:38:15 pm

Valentin Vornicu 7:38:54 pm We have gone as far as we can with Euler's Theorem. We can try to compute this by looking at the first few powers of 9 modulo 1000, but there is an easier way. Valentin Vornicu 7:38:58 pm Is there anything you notice about the number 9 that may make it easier to compute this power modulo 1000? Doink 7:39:40 9=10-1 krmathcounts -1 mod 10 MathTwo 10-1 fireonice 10-1 yjhan96 10-1=9 msinghal 9=10-1
pm

7:39:40 pm

7:39:40 pm

7:39:40 pm

7:39:40 pm

7:39:40 pm

MathForFun 7:39:40 9==-1 (mod 10) Valentin Vornicu

pm

7:40:09 pm

Valentin Vornicu

7:40:17 pm

Valentin Vornicu 7:40:22 pm Therefore, the last three digits of 9^(8^7) are 081. Valentin Vornicu
7:40:55 pm

Valentin Vornicu 7:41:18 pm We could verify that n^5 - n is divisible by 30 for n = 0, 1, 2, ..., 29. But what is an easier way? duketip10 do 2, 3, 5
7:42:38 pm

briantix 7:42:38 mod 2, 3, 5

pm

jerrytang 7:42:38 pm Take it mod 2, 3, and 5 fireonice 7:42:38 pm factor and show it divides 2,3,5 Cogswell 7:42:38 pm prove that it's divisible by 2, 3, and 5. msinghal 7:42:38 pm show it is divisible for 2, 3, 5 apple.singer 7:42:38 pm show it's divisible by 2, 3, and 5 Valentin Vornicu 7:42:41 pm We use the prime factorization of 30 = 2 x 3 x 5. Hence, it suffices to show that n^5 - n is divisible by 2, 3, and 5, for all integers n. Valentin Vornicu 7:43:03 Is n^5 - n divisible by 2? ashgabat Yes gurev yes
7:43:12 pm pm

7:43:12 pm

apple.singer yes Konigsberg yes it is MathForFun yep

7:43:12 pm

7:43:12 pm

7:43:12 pm

pi.guy3.14 7:43:15 pm 2 is clear, if n is even, even-even=even, if its odd, odd-odd=even Valentin Vornicu 7:43:17 pm Yes. By reducing modulo 2, we can check for n = 0 and n = 1. Valentin Vornicu 7:43:22 Is n^5 - n divisible by 3? gurev yes
7:43:32 pm pm

Konigsberg yes fireonice yes loquidyE yes Doink

7:43:32 pm

7:43:32 pm

7:43:32 pm

7:43:32 pm

yes pi.guy3.14 7:43:32 yes, plug in 0,1,2


pm

Valentin Vornicu 7:43:32 pm Yes. By reducing modulo 3, we can check for n = 0, 1, and 2. Valentin Vornicu 7:43:37 Is n^5 - n divisible by 5? pi.guy3.14 7:44:34 pm yes, plug in 0,1,2,3,4 mrkidney 7:44:34 pm Yes, by Fermat's little theorem sjaelee fermat's tareyza yes
7:44:34 pm pm

7:44:34 pm

duketip10 7:44:34 pm yes. n^4=1 by fermat's little theorem BOBBOBBOB yep


7:44:34 pm

Valentin Vornicu 7:44:40 pm Yes. By reducing modulo 5, we can check for n = 0, 1, 2, 3, and 4. We can also simply cite Fermat's Little Theorem for the prime p = 5. Valentin Vornicu 7:44:46 pm Hence, n^30 - n is divisible by 2 x 3 x 5 = 30 for all integers n. Valentin Vornicu 7:44:51 pm This simple problem illustrates an important principle. If you want to prove that a number is divisible by n, it helps to look at the individual prime powers of n. Valentin Vornicu
7:45:06 pm

Valentin Vornicu 7:45:20 pm We start by trying to reduce the given expression modulo 2000, but all the bases are already less than 2000. So what else can we do? fractals 7:46:11 pm take mod 16 and mod 125 gurev 7:46:11 pm do mod 16 then mod 125 Cogswell 7:46:11 pm split 2000 into 16 and 125 apple.singer 7:46:11 pm factor 2000=2^4*5^3? aerrowfinn72 7:46:11 pm taking mod 125 and 16 filetmignon821 7:46:11 pm 2000=2^4*5^3, so we have to prove that our expression is divisible by 16 and 125. Valentin Vornicu 7:46:13 pm The prime factorization of 2000 is 2^4 x 5^3, so we look at the factors 2^4 = 16 and 5^3 = 125 separately. Valentin Vornicu 7:46:16 pm How does the given expression reduce modulo 16? jeff10 7:47:56 pm 9^n-9^n+12^n-12^n

loquidyE 7:47:56 pm 9^n - 9^n + 12^n - 12^n == 0 (mod 16) Seedleaf 7:47:56 12^n - (-4)^n
pm

filetmignon821 7:47:56 pm it comes out to (-7)^n-(-7)^n+(-4)^n-(-4)^n and everything cancels nicely. Valentin Vornicu
7:48:03 pm

Valentin Vornicu 7:48:17 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 16. Valentin Vornicu 7:48:33 pm Modulo 125 we have a similar situation: Valentin Vornicu
7:48:40 pm

Valentin Vornicu 7:48:45 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 125. Valentin Vornicu 7:49:23 pm We conclude that 121^n - 25^n + 1900^n - (-4)^n is divisible by 2000 for all positive integers n. Valentin Vornicu
7:49:31 pm

Valentin Vornicu 7:50:24 pm What can we do with this equation? Cogswell 7:50:49 pm let's use the problem above: n^5-n (mod 30). loquidyE 7:50:49 pm check it for mod 2, 3, 5, etc. fireonice 7:50:49 take small mods
pm

Valentin Vornicu 7:50:54 pm We can use a previous problem and reduce it modulo 30. This gives us Valentin Vornicu
7:50:58 pm

Valentin Vornicu 7:51:06 pm Now we need to find bounds on n to find its exact value. What's one obvious lower bound on n? Cogswell 133 fireonice 133 fractals 133
7:52:25 pm

7:52:25 pm

7:52:25 pm

Seedleaf 133 Konigsberg 133

7:52:25 pm

7:52:25 pm

antimonyarsenide n > 133 jerrytang 133 duketip10 133

7:52:25 pm

7:52:25 pm

7:52:25 pm

Valentin Vornicu Clearly n > 133.

7:52:29 pm

Valentin Vornicu 7:52:33 pm Now we need to find an upper bound on n. Valentin Vornicu 7:53:22 pm The largest term in the left-hand side is 133^5, so we compare all the other terms to 133^5. Valentin Vornicu 7:53:25 pm We know that 110^5 < 133^5. Valentin Vornicu 7:53:28 pm Can we compare 84^5 + 27^5 and 133^5? Konigsberg 7:53:59 pm yes, it is smaller than 133^5 pi.guy3.14 7:53:59 pm adding them up, they are still smaller aerrowfinn72 7:53:59 pm LHS is definitely smaller fractals 7:53:59 pm 84 + 27 < 133, so the sum is less than 133^5 filetmignon821 7:53:59 pm yes, 84^5+27^5<133^5 matholympiad25 7:53:59 pm Yea, 84^5 + 27^5 < 111^5 < 133^5, so the former is smaller Valentin Vornicu 7:54:03 pm Note that 84^5 + 27^5 < (84 + 27)^5 = 111^5 < 133^5. Valentin Vornicu
7:55:08 pm

Valentin Vornicu 7:55:21 pm What we would like to do now is take the fifth root of both sides. The problem is we would get the fifth root of 3, and it's not clear what to do with this number. Valentin Vornicu 7:55:30 pm So let's see if we can replace 3 by the ratio of two fifth powers. Valentin Vornicu 7:55:35 pm The first few fifth powers are 1^5 = 1, 2^5 = 32, 3^5 = 243, 4^5 = 1024, 5^5 = 3125, and 6^5 = 7776, and so on. Do we see any fifth powers that approximately have a ratio of 3? giratina150 7:56:32 5^5/4^5~~3 pi.guy3.14 4,5
pm

7:56:32 pm

Seedleaf 7:56:32 4^5 and 5^5

pm

Konigsberg 7:56:32 pm 4^5 and 5^5 have a ratio of 1:3

nikgran 4,5

7:56:32 pm

yjhan96 7:56:32 4^5 and 5^5

pm

Valentin Vornicu 7:56:34 pm We see that 5^5/4^5 = 3125/1024 is approximately 3. Even better, this ratio is slightly larger than 3, which means we can use this ratio to get an upper bound on n: Valentin Vornicu
7:57:46 pm

Valentin Vornicu 7:57:54 pm Therefore, n < 5/4 x 133 = 166 + 1/4. Valentin Vornicu So what is n?
7:57:58 pm

nikgran 7:58:17 pm we see n = 144 is the answer Konigsberg 144 fractals 144 briantix 144
7:58:17 pm

7:58:17 pm

7:58:17 pm

filetmignon821 144 mrkidney 144

7:58:17 pm

7:58:17 pm

matholympiad25 n is 144

7:58:17 pm

Valentin Vornicu 7:58:39 pm The only n satisfying 133 < n < 166 + 1/4, and n congruent to 24 modulo 30 is n = 144. Valentin Vornicu 8:03:38 pm We had a little situation with the server. Everything seems to be back to normal now. Valentin Vornicu
8:03:54 pm

Valentin Vornicu 8:04:20 pm To make things easier, let s(N) denote the sum of the digits of N for a positive integer N. Then A = s(4444^4444) and B = s(A). Let C = s(B). Then we want to find C. Valentin Vornicu 8:04:32 pm Is there anything we can say about s(N)? Konigsberg 8:04:51 pm we get the modulo 9, because taking digit sums preserves the remainder when the number is divided by 9 jeff10 8:04:51 pm I am thinking mod 9 MathForFun mod 9
8:04:51 pm

Cogswell 8:04:51 pm it is congruent to N (mod 9) Valentin Vornicu 8:04:54 pm We know that s(N) is congruent to N modulo 9. Valentin Vornicu 8:05:02 pm We can we say that C is congruent to 4444^4444 modulo 9.

Valentin Vornicu 8:05:06 pm How can we compute this residue? alex31415 8:06:13 4444=7 mod 9
pm

aerrowfinn72 8:06:13 pm take 4444 mod 9 first and then take 4444 mod phi(9) loquidyE 8:06:13 pm congruent to 7^4444 jeff10 8:06:13 pm 7^7 (mod 9)=7 (mod 9) Valentin Vornicu
8:06:16 pm

Valentin Vornicu 8:06:19 pm How can we simplify this? Seedleaf 8:06:31 eulers theorem MathTwo euler?
pm

8:06:31 pm

filetmignon821 8:06:31 4444^6=1 (mod 9)

pm

filetmignon821 8:06:31 pm by euler's theorem, this is just 4444^4 mod 9 fractals 8:06:34 phi(9) = 6
pm

Valentin Vornicu 8:06:52 pm We can use Euler's Theorem. Valentin Vornicu


8:06:57 pm

Valentin Vornicu

8:07:33 pm

Valentin Vornicu

8:07:43 pm

Valentin Vornicu 8:07:47 pm Therefore, C is congruent to 7 modulo 9. Valentin Vornicu 8:07:50 pm How can we find the exact value of C? Cogswell bounding
8:08:48 pm

aerrowfinn72 bound it

8:08:48 pm

Valentin Vornicu 8:08:50 pm Let's proceed by finding an upper bound on A, which is the sum of the digits of 4444^4444. What is an upper bound on A? filetmignon821 8:09:44 pm we know that s(n) is less than or equal to 9(log n+1), so we have that s(4444^4444) is at most 9(4444log4444+1) Valentin Vornicu 8:09:49 pm We can try to find the exact number of digits in 4444^4444, but that would involve using logarithms. What if we didn't have a calculator? Can we find an estimate on A, even a crude one? nikgran
8:10:39 pm

10000^4444 Valentin Vornicu


8:10:44 pm

Valentin Vornicu 8:10:50 pm Therefore, A <= 9 x 17776 = 159984. Valentin Vornicu 8:10:57 pm Then we might conclude that B = s(A) is at most s(159984), but this would not be correct. Why not? msinghal 8:12:20 pm A could be, say, 99999 MathTwo 8:12:20 b could be 99999
pm

fractals 8:12:20 pm No, the maximum is s(99999) dpranav 8:12:20 pm you can have different distribution of the digits filetmignon821 8:12:20 pm no you have to add 9 again dpranav 8:12:20 pm you can have 99999 which has a larger sum Valentin Vornicu 8:12:22 pm The function s(N) is not an increasing function. Valentin Vornicu 8:12:28 pm Of the numbers 1, 2, 3, ..., 159984, the number with the greatest s-value is 99999. Valentin Vornicu 8:12:32 pm Therefore, B = s(A) <= 5 x 9 = 45. Valentin Vornicu 8:12:35 pm Now we want a bound on C = s(B). What can we do? loquidyE 8:13:46 pm s(b) is a maximum when b = 39 so the upper bound is 12 nikgran 8:13:46 pm take B = 39, so C = 12 Seedleaf 8:13:46 pm do the same; the highest is 3 + 9 = 12 alex31415 8:13:46 pm S(39)=12 is the highest possible value for C MathTwo 39*
8:13:46 pm

jerrytang 8:13:46 C <= s(39)=12

pm

Porteradams 8:13:46 at most s(39)=12

pm

MathTwo 8:13:46 pm it is at most 12 = s(39) ajb 8:13:46 pm the max is 3+9=12 Valentin Vornicu 8:13:56 pm We can repeat the argument. Of the numbers 1, 2, 3, ..., 45, the number with the greatest s-value is 39. Valentin Vornicu 8:13:59 pm Therefore, C = s(B) <= 3 + 9 = 12.

Valentin Vornicu 8:14:02 pm We have shown that C is congruent to 7 modulo 9, so we conclude that C = 7. Valentin Vornicu
8:14:48 pm

Valentin Vornicu 8:14:59 pm What makes this sum difficult to deal with? Cogswell 8:16:08 absolute values filetmignon821 absolute value
pm

8:16:08 pm

alex31415 8:16:08 pm Absolute value signs matholympiad25 8:16:08 The absolute values! MathForFun 8:16:08 absolute values
pm pm

Konigsberg 8:16:08 pm because it is absolute values Valentin Vornicu 8:16:27 pm The absolute value signs make this sum difficult to deal with. How can we deal with the absolute value signs? Is there a convenient way of getting rid of them? lightbluemathangel a_i > b_i for all i antimonyarsenide WLOG let ai>bi
8:17:39 pm

8:17:39 pm

gurev 8:17:39 pm We can switch the order of the pairs to make ai the bigger one, and the absolute values dissappear Porteradams 8:17:39 pm have ai always greater than or equal to bi Valentin Vornicu
8:17:42 pm

Valentin Vornicu

8:18:01 pm

Valentin Vornicu 8:18:11 pm Each summand in this is 1 or 6. How can we use this information? Valentin Vornicu 8:19:08 pm There are 999 summands, so let k of them be equal to 1, so the remaining 999 - k are equal to 6. Valentin Vornicu
8:19:19 pm

Valentin Vornicu Doink

8:19:23 pm

8:20:46 pm

they are all in the range from 1 to 1998 jeff10 8:20:46 pm They are all less than or equal to 1998 lightbluemathangel 8:20:46 they're all from 1 to 1998
pm

Seedleaf 8:20:46 pm they are distinct and belong to the set 1,2,...1998 Valentin Vornicu
8:20:48 pm

Valentin Vornicu 8:21:24 pm So can we turn this fact into a related sum? jerrytang 8:22:12 pm The sum is 1997001 giratina150 8:22:12 pm sum of (a_i)+sum of (b_i)=1998(1999)/2 filetmignon821 8:22:12 pm they sum to whatever the sum of 1 through 1998 is krmathcounts 8:22:12 pm a1+b1+a2+b2+.....+a999+b999 = 1998*1999/2 Valentin Vornicu
8:22:52 pm

Valentin Vornicu 8:22:54 pm What can we do with these two equations? Cogswell add 'em?
8:23:08 pm

lightbluemathangel 8:23:08 pm add them and then divide by 2 to get the sum of the a_i's krmathcounts add
8:23:08 pm

Valentin Vornicu

8:23:12 pm

Valentin Vornicu 8:23:18 pm We know that the left-hand side is even. Therefore, 5k must be odd, which means k must be odd. Valentin Vornicu 8:23:21 pm Let k = 2n + 1 for some integer n. Valentin Vornicu Now what? numbertheorist17 your done!
8:23:25 pm

8:24:15 pm

Cogswell 8:24:15 pm sum=5994-5(2n+1)=5989-10n nikgran 8:24:15 pm we are done, the sum ends in 9 pi.guy3.14 8:24:15 replace k with n Valentin Vornicu
pm

8:24:17 pm

We can substitute into the equation above, to get Valentin Vornicu


8:24:21 pm

Valentin Vornicu 8:24:26 pm It looks like we can conclude that this number always ends in 9, but what if n is sufficiently high so that 5989 - 10n is negative? For example, if n = 600, then 5989 - 10n = -11, which ends in 1. briantix 8:25:37 pm all the terms are positive ajb 8:25:37 pm Absolute values ensure it is positive apple.singer 8:25:37 pm absolute values. can't be negative gurev 8:25:37 pm How can this be negative there's absolute values? Valentin Vornicu 8:25:38 pm The original sum was expressed as a sum of absolute values, which must be nonnegative. And a nonnegative number of the form 5989 - 10n must end in 9, so we are home free. Valentin Vornicu
8:25:47 pm

Valentin Vornicu 8:25:53 How can we proceed? Cogswell 8:27:21 pm assume the contrary

pm

aerrowfinn72 8:27:21 pm go for proof by contradiction numbertheorist17 contradiction


8:27:21 pm

fractals 8:27:21 pm assume that you can then find contradiction Valentin Vornicu 8:27:22 pm We can argue by contradiction. Valentin Vornicu 8:27:44 pm Note that 2 x 5 - 1 = 9, 2 x 13 - 1 = 25, and 5 x 13 - 1 = 64 are all perfect squares. Valentin Vornicu 8:27:48 pm Suppose 2d - 1 = x^2, 5d - 1 = y^2, and 13d - 1 = z^2, for some positive integers x, y, and z. Valentin Vornicu 8:27:56 pm What does the first equation 2d - 1 = x^2 tell us? gurev 8:28:41 x is odd jerrytang x is odd
pm

8:28:41 pm

lightbluemathangel x^2==1 (mod 2) giratina150 x is odd pi.guy3.14 x is odd

8:28:41 pm

8:28:41 pm

8:28:41 pm

Valentin Vornicu

8:28:44 pm

The number 2d - 1 is always odd, so x is odd. Valentin Vornicu 8:28:50 pm Let x = 2n + 1 for some integer n. Then 2d - 1 = x^2 = (2n + 1)^2 = 4n^2 + 4n + 1. Valentin Vornicu 8:28:53 pm Then d = 2n^2 + 2n + 1. What does this tell us about d? briantix d is odd
8:29:00 pm

Doink 8:29:00 pm d is congruent to 1 mod 2 filetmignon821 d is odd


8:29:00 pm

gurev 8:29:00 pm d is odd, or else 2d-1 is 3 mod 4 Valentin Vornicu 8:29:02 pm This equation says that d is odd. Valentin Vornicu 8:29:06 pm Then what can we say about 5d - 1 = y^2 and 13d - 1 = z^2? Cogswell 8:30:00 pm they are both even aerrowfinn72 both even
8:30:00 pm

fireonice 8:30:00 y and z are even gurev 8:30:00 pm y and z are evem

pm

yjhan96 8:30:00 pm y and z must be even Doink 8:30:00 pm y and z must be even Seedleaf 8:30:00 pm y and z are both even Valentin Vornicu 8:30:02 pm Since d is odd, 5d - 1 = y^2 and 13d - 1 = z^2 are even. Hence, y and z are even. Valentin Vornicu 8:30:08 pm Let y = 2u and z = 2v. Then 5d - 1 = y^2 = 4u^2 and 13d - 1 = 4v^2. Valentin Vornicu 8:30:11 pm What can we do with these equations? msinghal 8:31:05 subrtract them Cogswell 8:31:05 subtract them fireonice subtract yjhan96 subtract
pm

pm

8:31:05 pm

8:31:05 pm

Doink 8:31:05 subtract

pm

Valentin Vornicu 8:31:06 pm We can subtract them, to get 8d = 4v^2 - 4u^2, so 2d = v^2 - u^2. Valentin Vornicu 8:31:12 pm What can we do with this equation? jerrytang
8:31:18 pm

Subtract for difference of squares. ajb 8:31:18 pm subtract them and use difference of squares Valentin Vornicu 8:31:22 pm We can factor the right-hand side, to get 2d = (v + u)(v - u). Valentin Vornicu 8:31:27 pm Can we say anything interesting about the factors v + u and v - u? aerrowfinn72 same parity giratina150 same parity
8:31:43 pm

8:31:43 pm

filetmignon821 8:31:43 pm they have the same parity fractals 8:31:43 pm must be of same parity Cogswell 8:31:43 pm same parity-->d is even. But d is odd. duketip10 8:31:43 same parity jerrytang 8:31:43 Same parity
pm

pm

Valentin Vornicu 8:31:46 pm Either v + u and v - u are both even or both odd (since their difference 2u is an even number). Valentin Vornicu 8:31:51 pm If v + u and v - u are both even, then their product (v + u)(v - u) is a multiple of 4. If v + u and v - u are both odd, then their product (v + u)(v - u) is odd. In particular, (v + u)(v - u) cannot be equal to 2d, because d is odd. Valentin Vornicu 8:31:56 pm We have obtained a contradiction, so at least one of 2d - 1, 5d - 1, and 13d - 1 is not a perfect square. Valentin Vornicu 8:32:03 pm It can seem simple, but using parity can be a powerful tool. And whenever you find that a number is even (odd resp.), it is usually a good idea to write it in the form 2n (2n + 1 resp.). Valentin Vornicu
8:38:45 pm

Valentin Vornicu 8:38:55 pm Let f(n) = 19 x 8^n + 17. What approach can we take here? Seedleaf 8:40:06 pm look at small values of n first fractals 8:40:06 pm find some mod that works filetmignon821 8:40:06 pm prove it's divisible by something ajb 8:40:06 pm Modulo small numbers Valentin Vornicu 8:40:10 pm We can see if there exists a number d > 1 such that f(n) = 19 x 8^n + 17 is always divisible by d, by reducing it modulo d. Valentin Vornicu 8:40:15 pm What numbers d should we try? pi.guy3.14 9 BOBBOBBOB 9 gurev
8:41:33 pm

8:41:33 pm

8:41:33 pm

9? Valentin Vornicu 8:41:35 pm We should try numbers d such that 8^n modulo d simplifies. Valentin Vornicu 8:41:40 pm The first obvious number to try is d = 8. Valentin Vornicu
8:41:52 pm

Valentin Vornicu 8:42:10 pm (9 is a good idea because 8 = - 1 (mod 9)) Valentin Vornicu 8:43:11 pm So reducing modulo 7 doesn't get us anywhere either. Valentin Vornicu
8:43:29 pm

Valentin Vornicu 8:43:48 pm We can look at the first few values of f(n), and see which divisors come up. Valentin Vornicu 8:43:58 pm What is f(0) and its prime factorization? fireonice 8:45:03 36=2^2*3^2 MathForFun 2^2 3^2 Porteradams 2^2(3^2)
pm

8:45:03 pm

8:45:03 pm

loquidyE 8:45:03 36 = 2^2 * 3^2 nikgran 8:45:03 2^2*3^2 Seedleaf 36

pm

pm

8:45:03 pm

Valentin Vornicu 8:45:04 f(0) = 36 = 2^2 x 3^2.

pm

Valentin Vornicu 8:45:08 pm What is f(1) and its prime factorization? jerrytang 13^2; 169
8:46:28 pm

fireonice 8:46:28 169= 13^2 msinghal 8:46:28 169=13^2 duketip10 169=13^2

pm

pm

8:46:28 pm

yjhan96 8:46:28 169=13^2

pm

Valentin Vornicu 8:46:29 f(1) = 169 = 13^2.

pm

Valentin Vornicu 8:46:38 pm What is f(2) and its prime factorization? filetmignon821 1233=3^2*137 Seedleaf
8:47:15 pm

8:47:15 pm

1233, 9*137 giratina150 3^2*137


8:47:15 pm

Valentin Vornicu 8:47:16 f(2) = 1233 = 3^2 x 137.

pm

Valentin Vornicu 8:47:20 pm What is f(3) and its prime factorization? filetmignon821 9745=5*1949 giratina150 5*1949
8:48:00 pm

8:48:00 pm

Valentin Vornicu 8:48:01 f(3) = 9745 = 5 x 1949.

pm

Valentin Vornicu 8:48:32 pm Do we see anything promising? msinghal 8:49:55 pm for even n, 9 is a factor. for n=1 mod 4, 13 is a factor ajb 8:49:55 pm Okay, so it doesn't seem there is an n which divides all f(x), but 3^2=9 divides every even one gurev 8:49:55 pm If n is even, this is divisible by 3, which is apparent when the expression is taken mod 9. fractals 8:49:55 pm if n even divisible by 9 dinoboy 8:49:55 pm every f(n) is divisible by a "small prime" Valentin Vornicu 8:49:58 pm We see that 3 divides both f(0) and f(2), so let's work modulo 3 and see what happens. Valentin Vornicu
8:50:06 pm

Valentin Vornicu 8:50:07 pm When is 2^n + 2 divisible by 3? briantix 8:50:38 when n is even fractals n even
pm

8:50:38 pm

Doink 8:50:38 pm when n is even Seedleaf 8:50:38 when n is even


pm

Valentin Vornicu 8:50:45 pm Since 2 is congruent to -1 modulo 3, 2^n + 2 is divisible by 3 if and only if n is even, which means f(n) is divisible by 3 if and only if n is even. Valentin Vornicu 8:50:54 pm Now we must look at f(n) for odd n. Let n = 2t + 1. What modulus can we work with? gurev 5
8:51:31 pm

Konigsberg 5? Cogswell 5

8:51:31 pm

8:51:33 pm

Valentin Vornicu

8:51:34 pm

Since 5 is a factor of f(3), let's work modulo 5. Valentin Vornicu


8:51:43 pm

Valentin Vornicu 8:51:53 pm When is 4^t + 1 divisible by 5? briantix 8:52:25 when t is odd
pm

lightbluemathangel t is odd yjhan96 8:52:25 when t is odd


pm

8:52:25 pm

Konigsberg 8:52:25 pm when t is an odd number fireonice 8:52:25 when t is odd


pm

Valentin Vornicu 8:52:28 pm Since 4 is congruent to -1 modulo 5, 4^t + 1 is divisible by 5 if and only if t is odd. Valentin Vornicu 8:52:41 pm Now we must look at f(2t + 1) for t even. Let t = 2u. Then f(2t + 1) = f(4u + 1). Valentin Vornicu 8:52:44 pm What modulus can we work with? Seedleaf 13 yjhan96 13 gurev 13
8:52:59 pm

8:52:59 pm

8:52:59 pm

Cogswell 13? fireonice 13

8:52:59 pm

8:52:59 pm

Valentin Vornicu 8:53:01 pm There's really only one choice. Since f(1) = 13^2, let's work modulo 13. Valentin Vornicu
8:53:09 pm

Valentin Vornicu 8:53:28 pm Thus, f(4u + 1) is always divisible by 13. Valentin Vornicu 8:53:33 pm To summarize, if n is even, then f(n) is divisible by 3 (and f(n) > 3), Valentin Vornicu
8:53:38 pm

if n is of the form 4k + 3, then f(n) is divisible by 5 (and f(n) > 5), Valentin Vornicu 8:53:43 pm and if n is of the form 4k + 1, then f(n) is divisible by 13 (and f(n) > 13). Valentin Vornicu 8:53:49 pm Therefore, f(n) is composite for all nonnegative integers n. Valentin Vornicu 8:54:06 pm This problem shows that sometime you must try different moduli, even within the same problem, so don't be afraid to experiment. Valentin Vornicu
8:54:25 pm

Valentin Vornicu 8:54:34 pm You may recall Euclid's proof that there are an infinite number of primes. How can we start? briantix 8:54:54 pm hmmm...contradiction? MathForFun contradiction
8:54:54 pm

fireonice 8:54:54 pm assume there are a finite number of primes Porteradams 8:54:54 pm assume the contradiction jerrytang 8:54:54 Contradiction
pm

Valentin Vornicu 8:54:55 pm We argue by contradiction. Valentin Vornicu


8:55:01 pm

Valentin Vornicu Then what?

8:55:07 pm

Seedleaf 8:56:22 pm multiply them together Porteradams 8:56:22 pm multiply all p together giratina150 8:56:22 multiply together pi.guy3.14 product+!
pm

8:56:22 pm

MathForFun 8:56:22 multiply them all! Valentin Vornicu

pm

8:56:26 pm

Valentin Vornicu 8:57:11 pm What can we say about P? fractals 8:57:37 pm it is 2 mod 3 if k is even Seedleaf 8:57:37 pm either P = 0 or 2 mod 3 Valentin Vornicu
8:57:40 pm

Valentin Vornicu 8:58:28 pm Suppose P reduces to 2 modulo 3. Then what can we say about P? Konigsberg 8:59:28 pm then it is in the form of 3n+2 Porteradams 8:59:28 pm it is of the form 3n+2 which was not on the list giratina150 8:59:28 pm it's of the form 3n+2 jerrytang 8:59:28 pm It is of form 3n+2, new prime, contradiction briantix 8:59:28 pm it must have a prime factor that is 2 mod three Valentin Vornicu 8:59:30 pm All primes (other than 3) are congruent to 1 modulo 3 and 2 modulo 3. If P was only divisible by primes congruent to 1 modulo 3, then P itself would be congruent to 1 modulo 3. But P is congruent to 2 modulo 3, so P must be divisible by at least one prime that is congruent to 2 modulo 3. Valentin Vornicu 8:59:36 pm But by construction, P is relatively prime to all primes that are congruent to 2 modulo 3, contradiction. Valentin Vornicu 8:59:55 pm The only case that is left is if P reduces to 0 modulo 3. Unfortunately, there's no good way to handle this case. (We can divide out all the factors of 3, but there's nothing we can say about the resulting quotient.) Valentin Vornicu 8:59:58 pm Is there any way to fix this argument? filetmignon821 9:01:26 pm consider 3p_1p_2...p_k - 1. This must be divisible by some prime of the form 3n+2, but is not divisible by any of the p_i. this forces a contradiction. dinoboy 9:01:26 pm take P = 3p_1p_2...p_k + 2, then its forced to be 2 mod 3 Valentin Vornicu
9:01:36 pm

Valentin Vornicu 9:01:38 pm Then P always reduces to 2 modulo 3, and we can proceed as above. Valentin Vornicu 9:01:44 pm More generally, if a and b are relatively prime positive integers, then there are infinitely many primes of the form an + b. This result is known as Dirichlet Theorem, but it is an advanced result and is difficult to prove. Valentin Vornicu 9:01:49 pm The ideas in Euclid's proof lead to the next few problems. Valentin Vornicu
9:01:58 pm

Valentin Vornicu 9:03:08 pm To get a feel for the problem, let's find the first few terms of the sequence. Valentin Vornicu What is p_2? BOBBOBBOB 3
9:03:14 pm

9:03:22 pm

lightbluemathangel 3 matholympiad25 3 filetmignon821 3

9:03:22 pm

9:03:22 pm

9:03:22 pm

gurev 3

9:03:22 pm

aerrowfinn72 3

9:03:22 pm

Valentin Vornicu 9:03:23 pm p_2 is the largest prime divisor of p_1 + 1 = 3, so p_2 = 3. Valentin Vornicu What is p_3? giratina150 7
9:03:25 pm

9:03:33 pm

filetmignon821 7 aerrowfinn72 7 loquidyE 7 fireonice 7 Seedleaf 7

9:03:33 pm

9:03:33 pm

9:03:33 pm

9:03:33 pm

9:03:33 pm

Valentin Vornicu 9:03:33 pm p_3 is the largest prime divisor of p_1 p_2 + 1 = 7, so p_3 = 7. Valentin Vornicu What is p_4?' BOBBOBBOB 43 aerrowfinn72 43
9:03:36 pm

9:03:42 pm

9:03:42 pm

matholympiad25 43 ajb 43
9:03:42 pm

9:03:42 pm

giratina150 43 fireonice 43 Cogswell 43

9:03:42 pm

9:03:42 pm

9:03:42 pm

Valentin Vornicu 9:03:43 pm p_4 is the largest prime divisor of p_1 p_2 p_3 + 1 = 43, so p_4 = 43. Valentin Vornicu What is p_5? giratina150 139? Cogswell 139
9:03:45 pm

9:04:34 pm

9:04:34 pm

Valentin Vornicu 9:04:35 pm p_5 is the largest prime divisor of p_1 p_2 p_3 p_4 + 1 = 1807 = 13 x 139, so p_4 = 139. Valentin Vornicu 9:04:42 pm We must show that 5 never appears in the sequence. Valentin Vornicu 9:04:44 How can we proceed?
pm

saphireflame 9:05:58 pm proof by contradiction krmathcounts 9:05:58 pm assume 5 does show up? Valentin Vornicu 9:06:01 pm We can argue by contradiction. Valentin Vornicu
9:06:11 pm

Valentin Vornicu 9:06:41 pm What does that say about this expression? Cogswell 9:06:57 pm p_1p_2/dotsp_n-1+1 = 2^Q3^R5^S MathTwo 9:06:57 pm some multiplication of 2s, 3 and 5 briantix 9:06:57 pm so its 2^a*3^b*5^c jerrytang 9:06:57 pm Only primes are 2, 3, 5 Valentin Vornicu
9:06:59 pm

Valentin Vornicu 9:07:02 Can we say more?

pm

filetmignon821 9:08:25 pm this can't be a multiple of 2 or 3, since p_1 and p_2 are 2 and 3, so it must be a power of 5. Seedleaf 9:08:25 pm its only a power of 5 fireonice 9:08:25 pm 5 only because 2 and 3 show up in the sequence apple.singer 9:08:32 pm just 5, as 2 and 3 are p_1 and p_2 and thus will not divide the product +1 Valentin Vornicu
9:08:35 pm

Valentin Vornicu

9:08:45 pm

Valentin Vornicu

9:09:01 pm

Valentin Vornicu

9:09:10 pm

Valentin Vornicu 9:09:12 pm What can we do with this equation?

fractals mod 4 briantix factor it

9:10:18 pm

9:10:18 pm

lightbluemathangel factor out (5-1) giratina150 mod 4

9:10:18 pm

9:10:18 pm

ajb 9:10:18 pm 5^k-1=0 mod 4 briantix 9:10:18 pm both sides must have a 2^2 in them but that's impossible for the left one filetmignon821 9:10:18 pm difference of kth powers forces p_1p_2...p_{n-1} to be divisible by 4 aerrowfinn72 take it mod 4
9:10:18 pm

Valentin Vornicu

9:10:24 pm

Valentin Vornicu 9:10:27 pm This equation tells us that 4 must be a factor of the left-hand side. Valentin Vornicu 9:10:34 pm But this is impossible because p_1 = 2 and all further terms p_2, ..., p_{n - 1} are odd. Valentin Vornicu 9:10:37 pm We have a contradiction, so 5 cannot be a member of this sequence. Valentin Vornicu 9:10:49 pm I'll let you think about why 11 is also not in the sequence. Valentin Vornicu
9:10:55 pm

Valentin Vornicu 9:11:27 pm I actually proposed this problem for the IMO TST. Only the 6 students that actually made the team managed to solve it. Valentin Vornicu 9:11:48 pm Let's start by looking at small primes. Is the prime 2 in P? filetmignon821 9:12:30 pm i bet it is, since P is the set of all primes msinghal 9:12:30 pm yes, all primes are in P Cogswell 9:12:30 pm yes, since it's prime Valentin Vornicu 9:12:36 pm That's not exactly a proof gurev 9:13:13 pm Did you mean is it in M, which is also true? Valentin Vornicu 9:13:30 pm Oops, a small typo. We're looking at 2 being in M. Cogswell 9:14:06 pm let A be the set of all odd primes in M. The result follows.

dinoboy 9:14:06 pm suppose not, take an odd prime in M and let it be A. Then we reach a contradiction so 2 is in M Valentin Vornicu 9:14:08 pm Since M contains at least three primes, there is an odd prime in M, say p. Let A = {p}. Then by the condition in the problem, all the prime factors of p - 1 are also in M. Valentin Vornicu 9:14:39 pm But p - 1 is even, so 2 is in M. Valentin Vornicu 9:14:48 Is the prime 3 in M? gurev Yes
9:15:40 pm pm

Cogswell yes

9:15:40 pm

Valentin Vornicu 9:15:47 pm I need a justification to give you all the 7 points. filetmignon821 9:17:09 pm if we have another prime p in M that is 1 mod 3, then 3 is in M. If p is 2 mod 3, then 2p is 1 mod 3, so 2p-1 is divisible by three. thus, 3 is in M. ajb 9:17:09 pm there must be at least two elements equal to 1 mod 3 or 2 mod 3 by pigeonhole Cogswell 9:17:09 pm take the two primes in M that are not 3. If one is 1 mod 3, the result follows. If they both are 2 mod 3, their product is 1 mod 3, and the result follows Valentin Vornicu 9:17:22 pm If M contains a prime of the form p = 3n + 1, then take A = {p}. Then all the factors of p - 1 = 3n are also in M, so 3 is in M. Valentin Vornicu 9:17:28 pm Otherwise, M contains two primes of the form 3n + 2, say p and q, and we can take A = {p, q}. Then all the prime factors of pq - 1 are also in M. Valentin Vornicu 9:17:31 pm But pq - 1 is congruent to 2 x 2 - 1 = 3, or 0 modulo 3, so pq - 1 is divisible by 3. Therefore, 3 is in M. In either case, 3 is in M. Valentin Vornicu 9:17:35 Is the prime 5 in M? filetmignon821 yeah. 2*3-1=5
9:19:05 pm pm

ajb 9:19:05 pm yes, since 2*3=6, and 6-1=5 loquidyE 9:19:05 yes; 2*3 - 1 = 5 msinghal 9:19:05 yes, 2*3 - 1
pm

pm

Valentin Vornicu 9:19:06 pm We know 2 and 3 are in M, so take A = {2, 3}. Then 2 x 3 - 1 = 5, so 5 is in M. Valentin Vornicu 9:19:11 Is the prime 7 in M? briantix 9:20:00 pm 3*5-1=2*7 so yes apple.singer 9:20:00 yes, take {3,5} antimonyarsenide yes, 7 | 3*5 - 1 gurev 9:20:00 Yeah, 3*5-1
pm pm pm

9:20:00 pm

Seedleaf 9:20:00 3*5 - 1 = 2*7

pm

Cogswell 9:20:00 yes (a={3,5})

pm

filetmignon821 9:20:00 pm yeah, 3*5-1=14, which is divisible by 7 Valentin Vornicu 9:20:01 pm We know 3 and 5 are in M, so take A = {3, 5}. Then 3 x 5 - 1 = 14 = 2 x 7, so 7 is in M. Valentin Vornicu 9:20:20 pm Hence, the first few primes are in M. We want to prove that M contains all primes. Valentin Vornicu 9:20:25 pm What intermediate result can we try proving first? gurev 9:20:43 infinitely many
pm

antimonyarsenide 9:20:43 pm M contains infinitely many primes Valentin Vornicu 9:20:45 pm Let's try proving first that M is infinite. Valentin Vornicu
9:20:50 pm

Valentin Vornicu 9:21:06 pm As a start, let's take A to be all the primes in M, except one, say p_i. (Remember that A must be a proper subset of M.) Valentin Vornicu
9:21:11 pm

Valentin Vornicu

9:21:19 pm

Cogswell 9:22:29 pm the only prime that can divide it is p_i NewAlbionAcademy 9:22:29 pm It isn't divisible by any of them, so it must be a power of p_i gurev 9:22:29 pm it's a power of pi Valentin Vornicu
9:22:47 pm

Valentin Vornicu

9:23:02 pm

Valentin Vornicu

9:23:17 pm

Valentin Vornicu 9:23:31 pm To make these equations simpler to work with, we start by omitting the smallest primes. Valentin Vornicu
9:23:38 pm

Valentin Vornicu

9:25:06 pm

Valentin Vornicu

9:25:37 pm

Valentin Vornicu

9:25:41 pm

Valentin Vornicu 9:26:15 pm What can we do with these two expressions for P? giratina150 equate
9:26:33 pm

Cogswell 9:26:33 pm set them equal to each other Porteradams 9:26:33 set them equal
pm

fireonice 9:26:33 pm set them equal to each other Valentin Vornicu 9:26:35 We can equate them. Valentin Vornicu
pm

9:26:45 pm

Valentin Vornicu 9:26:54 pm Let's start analyzing this equation. Valentin Vornicu
9:27:02 pm

Valentin Vornicu 9:27:04 pm Also, P is at least 2 x 3 x 5 x 7. So what can we say about a? Cogswell 9:28:06 a is at least 6 Porteradams at least 6 pi.guy3.14 at least 6
pm

9:28:06 pm

9:28:06 pm

Valentin Vornicu

9:28:08 pm

Valentin Vornicu

9:28:19 pm

Valentin Vornicu 9:28:23 pm So what can we say about b? Cogswell 9:28:50 it is at least 1. Valentin Vornicu
pm

9:28:52 pm

Valentin Vornicu

9:28:59 pm

Valentin Vornicu

9:29:14 pm

matholympiad25 it is odd. fractals odd

9:29:43 pm

9:29:43 pm

MathForFun It's odd giratina150 odd jerrytang It is odd. fireonice it is odd

9:29:43 pm

9:29:43 pm

9:29:43 pm

9:29:43 pm

Valentin Vornicu

9:29:46 pm

Valentin Vornicu gurev b is 1

9:29:51 pm

9:29:55 pm

MathForFun so b=1

9:29:55 pm

Valentin Vornicu Therefore, b = 1. Valentin Vornicu

9:29:56 pm

9:30:01 pm

Valentin Vornicu 9:30:08 pm How can we analyze this equation? giratina150 9:31:05 mod 2, mod 4
pm

Valentin Vornicu 9:31:10 pm We can try using modular arithmetic. Valentin Vornicu 9:31:14 pm Reducing modulo 2 doesn't get us far. Valentin Vornicu 9:31:17 pm We can try reducing modulo 3. Valentin Vornicu
9:31:28 pm

pi.guy3.14 9:32:24 so a must be odd MathForFun so a is odd fireonice a is odd fractals a is odd Seedleaf a is odd

pm

9:32:24 pm

9:32:24 pm

9:32:24 pm

9:32:24 pm

Valentin Vornicu 9:32:25 pm Hence, a - 1 is even. Let a - 1 = 2n. Valentin Vornicu


9:32:31 pm

Valentin Vornicu 9:32:34 pm What can we do with this equation? briantix 9:33:21 factor the right
pm

matholympiad25 9:33:21 pm factor as a difference of squares MathForFun factor RHS


9:33:21 pm

Cogswell 9:33:21 factor the RHS Valentin Vornicu

pm

9:33:24 pm

Valentin Vornicu 9:33:31 pm This equation says that both 2^n + 1 and 2^n - 1 are powers of 3. Valentin Vornicu 9:33:35 pm But (2^n + 1) - (2^n - 1) = 2. Valentin Vornicu 9:33:43 pm The only powers of 3 that differ by 3 are 3 and 1, so n = 1, and a = 3.

Valentin Vornicu 9:33:48 pm But a is at least 6, which is a contradiction. Valentin Vornicu 9:34:14 pm (Using the Catalan Conjecture was not a good idea - no points would have been awarded for that part of the problem.) Valentin Vornicu 9:34:24 pm Therefore, M contains an infinite number of primes. Valentin Vornicu 9:34:27 pm We're actually very close to the end now. Valentin Vornicu
9:34:32 pm

Valentin Vornicu 9:34:40 pm What result could we try using here? antimonyarsenide 9:36:14 Fermat's little theorem
pm

Valentin Vornicu 9:36:16 pm We could try using Fermat's Little Theorem. Valentin Vornicu
9:36:22 pm

gurev 9:37:35 pm FInding p-1 of them which are the same mod p, you can do this with pigeonhole antimonyarsenide 9:37:35 pm k=p-1, and p1,p2,...,pk are all congruent mod p filetmignon821 9:37:35 pm if k=p-1 and all the p_i are congruent mod p Valentin Vornicu
9:37:39 pm

Valentin Vornicu 9:37:45 pm The primes p_1, p_2, ..., p_{p - 1} must also be distinct from p. But if p was in M, then this argument would not be necessary in the first place. Valentin Vornicu 9:37:49 pm How can we show that such primes exist? msinghal 9:38:34 pm M is infinite, so pigeonhole Cogswell 9:38:36 pm there are infinitely many, so by the "infinite pigeonhole principle" there is an infinity of primes for a given residue mod p Valentin Vornicu 9:38:39 pm We have shown that the set M is infinite. Hence, if we reduce every element in M modulo p, then some nonzero residue must appear an infinite number of times. Valentin Vornicu 9:38:41 pm Then we simply take p - 1 primes p_1, p_2, ..., p_{p - 1} that have this residue. Valentin Vornicu 9:38:46 pm Therefore, M contains every prime p. Valentin Vornicu SUMMARY
9:38:52 pm

Valentin Vornicu 9:38:56 pm In today's class, we saw how to effectively use Fermat's Little Theorem and Euclid's Theorem. We also saw the power of algebra in number theory problems. We can solve many number theory problems effectively by putting a number in the right form (like expressing an even number in the form 2n), and using factorization. When you see a number theory problem, convert the words into mathematics, and let the equations do the work.

Valentin Vornicu 9:39:00 pm Finally, we also saw how to solve problems using argument by contradiction. If a problem asks you to prove that something cannot occur, this is usually a good sign to use contradiction. Valentin Vornicu 9:39:02 pm That's it for today's class. Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others. Copyright 2013 AoPS Incorporated. All Rights Reserved. Contact Us Privacy Terms & Conditions

Art of Problem Solving


WOOT
Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others.

Class Transcript 01/24 - Modular Arithmetic B


Valentin Vornicu 9:00:37 pm WOOT 2012-13: Modular Arithmetic B Valentin Vornicu 9:00:44 pm In today's lecture, we will continue our look at number theory and modular arithmetic. The first half of our class will focus on the use of Fermat's Little Theorem and Euler's Theorem. In the second half, we'll cover some more general number theory problems that are typical for this level. Valentin Vornicu 9:00:49 pm We'll start with a nice warm-up exercise. Valentin Vornicu
9:00:55 pm

Valentin Vornicu 9:01:22 pm How can we find the missing digit? KingSmasher3 mod 9!
9:01:37 pm

vishankjs 9:01:37 look mod 9 v_Enhance Take mod 9

pm

9:01:37 pm

Valentin Vornicu 9:01:45 pm We can try to find the sum of the 9 digits. If all 10 digits were present, then their sum would be 0 + 1 + 2 + ... + 9 = 45. However, we can't find the sum of the 9 digits without computing 2^29. Valentin Vornicu 9:01:52 pm We know that a number and the sum of its digits are congruent modulo 9, so we can reduce 2^29 modulo 9. What do we get? Mousie 5mod9
9:04:57 pm

greensteg2 5 mod 9 Bryan.C 5

9:04:57 pm

9:04:57 pm

aopsqwerty 5 woodstock 5 AkshajK

9:04:57 pm

9:04:57 pm

9:04:57 pm

zhaoamc 5 mod 9 AkshajK

9:04:57 pm

9:04:57 pm

mentalgenius 5

9:04:57 pm

Valentin Vornicu

9:05:01 pm

Valentin Vornicu

9:05:43 pm

So what is the missing digit? baijiangchen 4


9:06:02 pm

ABCDE 9:06:02 pm so the digit is 4 vishankjs 4 Mousie 4? zhaoamc 4 woodstock 4 sdaops 4 cire_il 4
9:06:02 pm

9:06:02 pm

9:06:02 pm

9:06:02 pm

9:06:02 pm

9:06:02 pm

Valentin Vornicu 9:06:03 pm The missing digit must be congruent to -5 modulo 9, so the missing digit is 4. Valentin Vornicu 9:06:06 pm Indeed, 2^29 = 536870912. All digits are present except 4. Valentin Vornicu
9:06:13 pm

Valentin Vornicu How do we start? algebra1337 mod 1000

9:06:19 pm

9:07:42 pm

KingSmasher3 9:07:42 pm last three digits means mod 1000 hawqish 9:07:42 take mod 1000 maxmk 9:07:42 mod 1000 aopsqwerty mod 1000
pm

pm

9:07:42 pm

Valentin Vornicu 9:07:42 pm Finding the last three digits of a number is an indication to work modulo 1000. Valentin Vornicu
9:07:48 pm

Valentin Vornicu 9:07:51 pm How can we reduce this? zhaoamc 9:08:00 pm eulor's totient thm sdaops 9:08:00 pm We want mod 1000, so phi(1000) is useful Porteradams 9:08:00 eulers theorem
pm

Valentin Vornicu 9:08:02 pm We can use Euler's Theorem. Valentin Vornicu


9:08:05 pm

Valentin Vornicu

9:08:47 pm

Valentin Vornicu 9:08:52 pm Probably the easiest way to compute this is to look at the first few powers of 8 modulo 400. We do this by starting with 1, then multiplying each term by 8 and reducing modulo 400. Valentin Vornicu 9:08:56 pm This gives us 1, 8, 64, 112, 96, 368, 144, 352. Valentin Vornicu
9:09:04 pm

Valentin Vornicu 9:09:11 pm We have gone as far as we can with Euler's Theorem. We can try to compute this by looking at the first few powers of 9 modulo 1000, but there is an easier way. Valentin Vornicu 9:09:15 pm Is there anything you notice about the number 9 that may make it easier to compute this power modulo 1000? sdaops 9:09:44 It's 10-1 greensteg2 9 = 10 - 1
pm

9:09:44 pm

ksun48 9:09:44 oh... It's 10-1 Mousie 10 - 1

pm

9:09:44 pm

baijiangchen 9 = (10-1)

9:09:44 pm

GeorgiaTechMan 9:09:44 pm and its congruent to -1 mod 10! KingSmasher3 teethpaste 9=10-1?
9:09:44 pm

9:09:44 pm

Valentin Vornicu

9:09:50 pm

Valentin Vornicu

9:10:58 pm

Valentin Vornicu 9:11:03 pm Therefore, the last three digits of 9^(8^7) are 081. Valentin Vornicu
9:11:11 pm

Valentin Vornicu 9:11:19 pm We could verify that n^5 - n is divisible by 30 for n = 0, 1, 2, ..., 29. But what is an easier way? maxmk 2, 3, 5 AkshajK
9:12:40 pm

9:12:40 pm

SkinnySanta 9:12:40 pm fermats little theorem for 2,3,5 and apply crt

KingSmasher3 9:12:40 divisible by 5,3,2 hawqish 9:12:40 pm take it mod 2, 3 and 5

pm

woodstock 9:12:40 pm Show that n^5 - 5 is divisible by 2, 3, and 5. vishankjs 9:12:40 pm factor and look mod 2,3,5 Valentin Vornicu 9:12:42 pm We use the prime factorization of 30 = 2 x 3 x 5. Hence, it suffices to show that n^5 - n is divisible by 2, 3, and 5, for all integers n. Valentin Vornicu 9:12:45 Is n^5 - n divisible by 2? GeorgiaTechMan yes sdaops Yup ABCDE yes
pm

9:12:52 pm

9:12:52 pm

9:12:52 pm

baijiangchen yes

9:12:52 pm

Valentin Vornicu 9:12:53 pm Yes. By reducing modulo 2, we can check for n = 0 and n = 1. Valentin Vornicu 9:12:56 Is n^5 - n divisible by 3? AkshajK
9:14:03 pm pm

GeorgiaTechMan yes baijiangchen yes

9:14:03 pm

9:14:03 pm

sdaops 9:14:03 pm Mhm, little theorem ABCDE yes ksun48 Yup


9:14:03 pm

9:14:03 pm

Valentin Vornicu 9:14:03 pm Yes. By reducing modulo 3, we can check for n = 0, 1, and 2. Valentin Vornicu 9:14:10 Is n^5 - n divisible by 5? AkshajK
9:14:35 pm pm

baijiangchen 9:14:35 Yes, again by flt

pm

sdaops 9:14:35 pm Little theorem strikes again DVA6102 9:14:35 pm by little theorem, yes vishankjs 9:14:35 pm yes, n^5=n mod 5 Valentin Vornicu
9:14:36 pm

Yes. By reducing modulo 5, we can check for n = 0, 1, 2, 3, and 4. We can also simply cite Fermat's Little Theorem for the prime p = 5. Valentin Vornicu 9:14:42 pm Hence, n^5 - n is divisible by 2 x 3 x 5 = 30 for all integers n. Valentin Vornicu 9:14:46 pm This simple problem illustrates an important principle. If you want to prove that a number is divisible by n, it helps to look at the individual prime powers of n. Valentin Vornicu
9:15:46 pm

Valentin Vornicu 9:15:54 pm We start by trying to reduce the given expression modulo 2000, but all the bases are already less than 2000. So what else can we do? algebra1337 9:17:22 mod 16 and mod 125 anwang16 9:17:22 mod 16? mod 25?
pm

pm

aopsqwerty 9:17:22 pm show it is divisible by 16 and 125 baijiangchen 9:17:22 factor into 2^4 5^3
pm

KingSmasher3 9:17:22 pm prove divisibility by 125 and 16 GeorgiaTechMan 9:17:22 try mod 16 and mod 125
pm

Valentin Vornicu 9:17:24 pm The prime factorization of 2000 is 2^4 x 5^3, so we look at the factors 2^4 = 16 and 5^3 = 125 separately. Valentin Vornicu 9:17:28 pm How does the given expression reduce modulo 16? GeorgiaTechMan 9:18:53 pm 9^n-9^n+(-4)^n -(-4)^n greensteg2 9:18:53 pm 9^n - 9^n + 12^n - (-4)^n teethpaste 9:18:53 12^n - (-4)^n
pm

DVA6102 9:18:53 pm 9^n-9^n+12^n-12^n=0. yes Mousie 9:18:53 12^n - (-4)^n


pm

Valentin Vornicu

9:18:54 pm

Valentin Vornicu 9:18:57 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 16. Valentin Vornicu 9:19:06 pm How does the given expression reduce modulo 125? anwang16 9:20:12 pm (121)^n-(25)^n+25^n-121^n sdaops
9:20:12 pm

(-4)^n -25^n + 25^n - (-4)^n = 0 Mousie 9:20:12 pm 0 as well...terms cancel woodstock 9:20:12 pm 121^n - 25^n + 25^n - 121^n Valentin Vornicu
9:20:15 pm

Valentin Vornicu 9:20:17 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 125. Valentin Vornicu 9:20:21 pm We conclude that 121^n - 25^n + 1900^n - (-4)^n is divisible by 2000 for all positive integers n. Valentin Vornicu
9:20:31 pm

Valentin Vornicu 9:20:42 pm What can we do with this equation? anwang16 9:22:11 mod something?
pm

Valentin Vornicu 9:22:18 pm What mod should we use? GeorgiaTechMan 9:22:31 pm actually, just 2, 3, and 5 suffice SkinnySanta 9:22:31 pm look modulo 30 since n^5 = n mof 30 Valentin Vornicu 9:22:37 pm We can use a previous problem and reduce it modulo 30. This gives us Valentin Vornicu
9:22:40 pm

Valentin Vornicu 9:22:45 pm Now we need to find bounds on n to find its exact value. What's one obvious lower bound on n? hawqish > 133? ABCDE 133
9:24:28 pm

9:24:28 pm

Mousie 9:24:28 oops 133 greensteg2 133 mentalgenius 133 SkinnySanta 133

pm

9:24:28 pm

9:24:28 pm

9:24:28 pm

woodstock 133

9:24:28 pm

GeorgiaTechMan n>133 Valentin Vornicu Clearly n > 133.

9:24:28 pm

9:24:29 pm

Valentin Vornicu 9:24:32 pm Now we need to find an upper bound on n. Valentin Vornicu 9:24:37 pm The largest term in the left-hand side is 133^5, so we compare all the other terms to 133^5. Valentin Vornicu 9:25:21 pm We know that 110^5 < 133^5. Valentin Vornicu 9:25:26 pm Can we compare 84^5 + 27^5 and 133^5? woodstock 9:26:41 pm 84^5 + 27^5 < 111^5 < 133^5 Mousie 9:26:41 pm since 84^5 + 27^5 < 133^5 AkshajK 9:26:41 pm (a+b)^5 > a^5 + b^5 DVA6102 9:26:41 pm 84^5+27^5 < 133^5 Valentin Vornicu 9:26:42 pm Note that 84^5 + 27^5 < (84 + 27)^5 = 111^5 < 133^5. Valentin Vornicu
9:26:48 pm

Valentin Vornicu 9:26:52 pm What we would like to do now is take the fifth root of both sides. The problem is we would get the fifth root of 3, and it's not clear what to do with this number. Valentin Vornicu 9:26:56 pm So let's see if we can replace 3 by the ratio of two fifth powers. Valentin Vornicu 9:27:02 pm The first few fifth powers are 1^5 = 1, 2^5 = 32, 3^5 = 243, 4^5 = 1024, 5^5 = 3125, and 6^5 = 7776, and so on. Do we see any fifth powers that approximately have a ratio of 3? ABCDE 9:28:42 5^5 and 4^5
pm

aopsqwerty 9:28:42 4^5 and 5^5 jhfrost314 5^5/4^5

pm

9:28:42 pm

greensteg2 9:28:42 4^5 and 5^5 pi314159265358979 5^5 and 4^5 hawqish 9:28:42 5^4 and 4^5 baijiangchen (5/4)
pm

pm

9:28:42 pm

9:28:42 pm

GeorgiaTechMan 4 and 5

9:28:42 pm

sdaops 9:28:42 pm Thus n < 133*5/4

zhaoamc 5^5/4^5

9:28:42 pm

Valentin Vornicu 9:28:43 pm We see that 5^5/4^5 = 3125/1024 is approximately 3. Valentin Vornicu 9:28:49 pm Even better, this ratio is slightly larger than 3, which means we can use this ratio to get an upper bound on n: Valentin Vornicu
9:28:52 pm

Valentin Vornicu 9:28:55 pm Therefore, n < 5/4 x 133 = 166 + 1/4. Valentin Vornicu So what is n? Minamoto 144 AkshajK 144
9:28:58 pm

9:29:08 pm

9:29:08 pm

KingSmasher3 144 bobcat120 144 DVA6102 144

9:29:08 pm

9:29:08 pm

9:29:08 pm

Valentin Vornicu 9:29:10 pm The only n satisfying 133 < n < 166 + 1/4, and n congruent to 24 modulo 30 is n = 144. Valentin Vornicu
9:30:25 pm

Valentin Vornicu 9:30:33 pm To make things easier, let s(N) denote the sum of the digits of N for a positive integer N. Then A = s(4444^4444) and B = s(A). Let C = s(B). Then we want to find C. Valentin Vornicu 9:31:43 pm Is there anything we can say about s(N)? DVA6102 9:33:15 pm Can we take it mod 9, since s(x)==x mod 9? sdaops 9:33:15 pm Congruent to N mod 9 maxmk 9:33:15 pm s(N) == N (mod 9) ksun48 9:33:15 9|N-s(N) ABCDE 9:33:15 s(N)==N mod 9
pm

pm

SkinnySanta 9:33:15 s(n) = n mod 9

pm

Valentin Vornicu 9:33:15 pm We know that s(N) is congruent to N modulo 9. Valentin Vornicu 9:33:26 pm So we can we say that C is congruent to 4444^4444 modulo 9. Valentin Vornicu 9:33:30 pm How can we compute this residue? woodstock
9:35:06 pm

4444 = 7 (mod 9) Bryan.C 9:35:06 euler's theorem


pm

hawqish 9:35:06 pm use euler's theorem anwang16 9:35:06 pm Get rid of most of the numbers and end up with 7^4444 which is -2^4444 AkshajK
9:35:06 pm

Valentin Vornicu

9:35:10 pm

Valentin Vornicu 9:35:18 pm We can use Euler's Theorem to simplify this. Valentin Vornicu
9:35:22 pm

Valentin Vornicu

9:36:11 pm

Valentin Vornicu

9:36:16 pm

Valentin Vornicu 9:36:21 pm Therefore, C is congruent to 7 modulo 9. aopsqwerty 9:36:25 but 7=-2 (mod 9)
pm

Valentin Vornicu 9:36:40 pm True we could have used that, and it would have probably been a little faster. Same idea though! Valentin Vornicu 9:36:47 pm Now, how can we find the exact value of C? Minamoto 9:38:16 use bounds aopsqwerty find bounds?
pm

9:38:16 pm

SkinnySanta 9:38:16 pm find upper bounds based on the number of digits sdaops 9:38:16 pm Let's figure out roughly how big C is; if it's a single digit, we know it already. Valentin Vornicu 9:38:18 pm Let's proceed by finding an upper bound on A, which is the sum of the digits of 4444^4444. What is an upper bound on A? Bryan.C 9:39:34 pm 9(log(4444^4444)+1) Valentin Vornicu 9:39:36 pm We can try to find the exact number of digits in 4444^4444, but that would involve using logarithms. What if we didn't have a calculator? Can we find an estimate on A, even a crude one? Minamoto 9:40:00 10000^4444
pm

greensteg2 9:40:00 pm 4444 < 10000 = 10^5, so 4444^4444 < (10^5)^4444 = 10^22220 sdaops
9:40:00 pm

Instead of 4444^4444, use 10000^4444 Valentin Vornicu


9:40:06 pm

Valentin Vornicu 9:40:11 pm Therefore, A <= 9 x 17776 = 159984. Valentin Vornicu 9:40:15 pm Then we might conclude that B = s(A) is at most s(159984), but this would not be correct. Why not? baijiangchen 9:41:36 pm Because that is not max digit sum, for max digit sum use all nines woodstock 9:41:36 pm There are numbers under 159984 which have larger digit sums. SkinnySanta 9:41:36 pm because maximum sum is at 99999 under 159984 Bryan.C 9:41:36 pm sum of digits could be larger jhfrost314 9:41:36 B could be 99999 anwang16 9:41:36 99999 for example
pm

pm

aopsqwerty 9:41:36 s(99999) is bigger

pm

Valentin Vornicu 9:41:38 pm The function s(N) is not an increasing function. Valentin Vornicu 9:41:43 pm Of the numbers 1, 2, 3, ..., 159984, the number with the greatest s-value is 99999. Valentin Vornicu 9:41:47 pm Therefore, B = s(A) <= 5 x 9 = 45. Valentin Vornicu 9:42:38 pm Now we want a bound on C = s(B). What can we do? baijiangchen 9:43:42 pm so 39 => 12, so C<13 Bryan.C 9:43:42 pm C < 12 (39 has greatest sum) Minamoto 9:43:42 pm s(39)=12 is the highest possible XtremeBanana5 9:43:42 pm find the highest s(B) which is when B=39 so 12 bobcat120 9:43:42 pm The largest s(n) is 12 so c=7 anwang16 9:43:42 max is at 39
pm

SkinnySanta 9:43:42 pm maximum is at 39 ==> 12 > C ==> C=7 greensteg2 9:43:42 pm Do the same thing: the maximum would be 39, which is 12. aopsqwerty 9:43:42 pm it is at most s(39)=12 Valentin Vornicu 9:43:53 pm We can repeat the argument. Of the numbers 1, 2, 3, ..., 45, the number with the greatest s-value is 39.

Valentin Vornicu 9:43:56 pm Therefore, C = s(B) <= 3 + 9 = 12. Valentin Vornicu 9:44:01 pm We have shown that C is congruent to 7 modulo 9, so we conclude that C = 7. Valentin Vornicu
9:44:47 pm

Valentin Vornicu 9:45:01 pm What makes this sum difficult to deal with? baijiangchen abs values
9:46:04 pm

AkshajK 9:46:04 pm the absolute values Porteradams 9:46:04 absolute values


pm

greensteg2 9:46:04 pm THE ABSOLUTE VALUES ABCDE 9:46:04 pm the absolute values DVA6102 9:46:04 absolute value
pm

Valentin Vornicu 9:46:07 pm The absolute value signs make this sum difficult to deal with. How can we deal with the absolute value signs? Is there a convenient way of getting rid of them? woodstock 9:47:20 pm Let a_i be greater than b_i. Minamoto 9:47:20 WLOG let a_i>b_i
pm

XtremeBanana5 9:47:20 assume wlog a_i > b_i maxmk 9:47:20 pm assume a_i > b_i for all i Porteradams 9:47:20 let ai>bi for all i anwang16 9:47:20 Assume a_i>b_i Valentin Vornicu
pm

pm

pm

9:47:21 pm

Valentin Vornicu

9:47:25 pm

Valentin Vornicu 9:47:27 pm Each term in this sum is 1 or 6. How can we use this information? hawqish 9:48:32 pm can we just replace it with a 1 or a 6 Valentin Vornicu
9:48:33 pm

There are 999 terms, so let k of them be equal to 1, so the remaining 999 - k are equal to 6. Valentin Vornicu
9:48:37 pm

Valentin Vornicu

9:48:41 pm

sdaops 9:49:58 pm They are distinct numbers from 1 to 1998 baijiangchen 9:49:58 between 1 and 1998
pm

greensteg2 9:49:58 pm They are from the set 1,2,3...,1997,1998 anwang16 9:49:58 all from 1 to 1998 Valentin Vornicu
pm

9:50:01 pm

Valentin Vornicu 9:50:05 pm So can we turn this fact into a related sum? Minamoto 9:51:21 pm they add to (1998)(1999)/2, which is odd Bryan.C 9:51:21 pm sum of all a_i and b_i is 1998*1999/2 Valentin Vornicu
9:51:24 pm

Valentin Vornicu

9:51:30 pm

Valentin Vornicu 9:51:32 pm What can we do with these two equations? AkshajK 9:52:53 add them baijiangchen add anwang16 Add them
pm

9:52:53 pm

9:52:53 pm

hawqish 9:52:53 pm add them together Bryan.C 9:52:53 add them


pm

teethpaste 9:52:53 pm add them or substract sdaops 9:52:53 pm Their sum is even!

Valentin Vornicu

9:52:55 pm

Valentin Vornicu 9:52:59 pm Does this equation say anything about the value of k? greensteg2 9:53:50 Must be odd. Bryan.C 9:53:50 k must be odd woodstock k is odd. jhfrost314 k is odd
pm pm

9:53:50 pm

9:53:50 pm

baijiangchen odd

9:53:50 pm

Minamoto 9:53:50 k must be odd Mousie 9:53:50 it must be odd!


pm

pm

Valentin Vornicu 9:53:50 pm The left-hand side must be even. Therefore, 5k must be odd, which means k must be odd. Valentin Vornicu 9:53:54 pm Let k = 2n + 1 for some integer n. Valentin Vornicu 9:53:57 pm We can substitute into the equation above, to get Valentin Vornicu
9:54:00 pm

sdaops 9:54:05 pm So 5994 - 5k ends in 9! Valentin Vornicu 9:54:12 pm It looks like we can conclude that this number always ends in 9, but what if n is sufficiently high so that 5989 - 10n is negative? For example, if n = 600, then 5989 - 10n = -11, which ends in 1. AkshajK 9:55:11 Absolute value!
pm

sdaops 9:55:11 pm Absolute values to the rescue! XtremeBanana5 9:55:11 pm nope always positive because of absolute values Minamoto 9:55:11 pm but we have absolute values in the original problem jjasonlan 9:55:11 absolute value
pm

Valentin Vornicu 9:55:12 pm The original sum was expressed as a sum of absolute values, which must be nonnegative. And a nonnegative number of the form 5989 - 10n must end in 9, so we are home free. Valentin Vornicu
9:55:20 pm

Valentin Vornicu

9:55:28 pm

How can we proceed? Mousie 9:56:31 pm could we try proof by contradiction? Valentin Vornicu 9:56:32 pm We can argue by contradiction. Valentin Vornicu 9:56:37 pm Note that 2 x 5 - 1 = 9, 2 x 13 - 1 = 25, and 5 x 13 - 1 = 64 are all perfect squares. Valentin Vornicu 9:56:45 pm Suppose 2d - 1 = x^2, 5d - 1 = y^2, and 13d - 1 = z^2, for some positive integers x, y, and z. Valentin Vornicu 9:56:51 pm What does the first equation 2d - 1 = x^2 tell us? Mousie 9:58:16 x must be odd aopsqwerty x is also odd
pm

9:58:16 pm

DVA6102 9:58:16 x is odd number teethpaste x is odd hawqish x is odd jhfrost314 x is odd

pm

9:58:16 pm

9:58:16 pm

9:58:16 pm

Valentin Vornicu 9:58:17 pm The number 2d - 1 is always odd, so x is odd. Valentin Vornicu 9:58:21 pm Let x = 2n + 1 for some integer n. Then 2d - 1 = x^2 = (2n + 1)^2 = 4n^2 + 4n + 1. Valentin Vornicu 9:58:23 pm Then d = 2n^2 + 2n + 1. What does this tell us about d? AmericanPi 9:58:30 pm d must be odd, otherwise 2d-1 would be congruent to 3 mod 4 ABCDE 9:58:30 pm d is odd; mod 4 XtremeBanana5 9:58:30 pm x is odd so x^2 = 1 mod 4 so d is also odd greensteg2 Odd.
9:58:30 pm

Valentin Vornicu 9:59:33 pm This equation says that d is odd. Valentin Vornicu 9:59:37 pm Then what can we say about 5d - 1 = y^2 and 13d - 1 = z^2? greensteg2 Even.
10:00:20 pm

DVA6102 10:00:20 y and z are even sdaops 10:00:20 y and z are even baijiangchen Even

pm

pm

10:00:20 pm

AmericanPi 10:00:20 y and z both even woodstock

pm

10:00:20 pm

y and z must be even. Valentin Vornicu 10:00:21 pm Since d is odd, 5d - 1 = y^2 and 13d - 1 = z^2 are even. Hence, y and z are even. Valentin Vornicu 10:00:24 pm Let y = 2u and z = 2v. Then 5d - 1 = y^2 = 4u^2 and 13d - 1 = 4v^2. Valentin Vornicu 10:00:27 pm What can we do with these equations? AkshajK 10:01:22 subtract them Minamoto subtract
pm

10:01:22 pm

anwang16 10:01:22 pm subtract the first from the second Mousie 10:01:22 subtract them
pm

Bryan.C 10:01:22 subtract them

pm

Valentin Vornicu 10:01:24 pm We can subtract them, to get 8d = 4v^2 - 4u^2, so 2d = v^2 - u^2. Valentin Vornicu 10:01:30 pm What can we do with this equation? greensteg2 10:03:17 pm Difference of squares. aopsqwerty 10:03:17 pm factorize right hand side Mousie 10:03:17 pm factor right side (v - u)(v + u) pi314159265358979 10:03:17 factor difference of squares Minamoto 10:03:17 2d=(v+u)(v-u)
pm pm

Valentin Vornicu 10:03:21 pm We can factor the right-hand side, to get 2d = (v + u)(v - u). Valentin Vornicu 10:03:24 pm Can we say anything interesting about the factors v + u and v - u? XtremeBanana5 10:03:38 pm (v-u)(v+u), but either they are both even or both odd, so they cant multuply to 2*odd aopsqwerty 10:03:38 pm v-u and v+u have the same parity, so d is also divisible by 2, which is a contradiction! AmericanPi 10:03:38 pm v^2-u^2 is either odd or divisible by 4, so there are no odd d that satisfy this equation-- contradiction SkinnySanta same parity
10:03:38 pm

KingSmasher3 10:03:38 must be same parity baijiangchen same parity

pm

10:03:38 pm

vishankjs 10:03:38 same parity jhfrost314 same parity

pm

10:03:38 pm

Valentin Vornicu

10:03:41 pm

Either v + u and v - u are both even or both odd (since their difference 2u is an even number). Valentin Vornicu 10:03:46 pm If v + u and v - u are both even, then their product (v + u)(v - u) is a multiple of 4. If v + u and v - u are both odd, then their product (v + u)(v - u) is odd. In particular, (v + u)(v - u) cannot be equal to 2d, because d is odd. Valentin Vornicu 10:03:48 pm We have obtained a contradiction, so at least one of 2d - 1, 5d - 1, and 13d - 1 is not a perfect square. Valentin Vornicu 10:03:53 pm It can seem simple, but using parity can be a powerful tool. And whenever you find that a number is even (odd resp.), it is usually a good idea to write it in the form 2n (2n + 1 resp.). Valentin Vornicu
10:10:30 pm

Valentin Vornicu 10:11:19 pm Let f(n) = 19 x 8^n + 17. What approach can we take here? AkshajK
10:12:30 pm

pi314159265358979 take mods

10:12:30 pm

petersonb2 10:12:30 pm We need to take modulo of a number. Mousie 10:12:30 pm try a few values? Valentin Vornicu 10:12:35 pm We can see if there exists a number d > 1 such that f(n) = 19 x 8^n + 17 is always divisible by d, by reducing it modulo d. Valentin Vornicu 10:12:40 pm We should try numbers d such that 8^n modulo d simplifies. Valentin Vornicu 10:12:48 pm What's the first obvious choice? baijiangchen DARN 8 and 9 vishankjs 8
10:14:06 pm

10:14:06 pm

Valentin Vornicu 10:14:10 pm The first obvious number to try is d = 8. Valentin Vornicu
10:14:13 pm

Valentin Vornicu 10:14:17 pm So reducing modulo 8 doesn't get us anywhere. What other numbers d should we try? AkshajK vishankjs 7 ksun48 7
10:14:23 pm

10:14:23 pm

10:14:23 pm

Valentin Vornicu 10:14:26 We could also try d = 7. Valentin Vornicu

pm

10:14:30 pm

Valentin Vornicu 10:14:32 pm So reducing modulo 7 doesn't get us anywhere either. Valentin Vornicu
10:14:35 pm

What else can we try? Mousie 10:15:52 pm well we could try plugging in n = 1, 2 and see what they are divisible by... Valentin Vornicu 10:15:55 pm We can look at the first few values of f(n), and see which divisors come up. Valentin Vornicu 10:15:58 pm What is f(0) and its prime factorization? baijiangchen 10:17:21 36 = 2^2 * 3^2 Mousie 10:17:21 pm f(0) = 36 = 2^2*3^2 teethpaste 10:17:21 36 = 2^2*3^2
pm pm

mentalgenius 10:17:21 f(0) = 36 = 2^2 * 3^2 maxmk 10:17:21 pm f(0) = 36 = 2^2*3^2 pseudonymous0 2^3 * 3^2

pm

10:17:21 pm

Valentin Vornicu 10:17:26 f(0) = 36 = 2^2 x 3^2.

pm

Valentin Vornicu 10:17:28 pm What is f(1) and its prime factorization? Mousie 10:18:31 pm f(1) = 169 = 13^2 hawqish 10:18:31 169, 13^2
pm

sdaops 10:18:31 pm 19*8+17 = 152+17 = 169 = 13^2 Minamoto 169=13^2 teethpaste 13^2
10:18:31 pm

10:18:31 pm

pi314159265358979 f(1) = 169 = 13^2 ksun48 10:18:31 169=13^2


pm

10:18:31 pm

Valentin Vornicu 10:18:32 f(1) = 169 = 13^2. Valentin Vornicu Similarly, we get

pm

10:18:39 pm

Valentin Vornicu 10:18:40 f(2) = 1233 = 3^2 x 137. Valentin Vornicu 10:18:43 f(3) = 9745 = 5 x 1949.

pm

pm

Valentin Vornicu 10:18:46 pm Do we see anything promising? maxmk 10:20:05 pm 3 for both even numbers pi314159265358979 10:20:05 pm half of them have a factor of 3 Valentin Vornicu
10:20:08 pm

We see that 3 divides both f(0) and f(2), so let's work modulo 3 and see what happens. Valentin Vornicu
10:20:12 pm

Valentin Vornicu 10:20:15 pm When is 2^n + 2 divisible by 3? algebra1337 10:21:10 when n is even teethpaste 10:21:10 when n is even Bryan.C n is even
10:21:10 pm pm

pm

hawqish 10:21:10 when n is even sdaops Even n

pm

10:21:10 pm

woodstock n is even.

10:21:10 pm

greensteg2 10:21:10 pm When n is even. PERFECT mentalgenius n is even


10:21:10 pm

Valentin Vornicu 10:21:13 pm Since 2 is congruent to -1 modulo 3, 2^n + 2 is divisible by 3 if and only if n is even, which means f(n) is divisible by 3 if and only if n is even. Valentin Vornicu 10:21:18 pm Now we must look at f(n) for odd n. Let n = 2t + 1. What modulus can we work with? anwang16 5
10:22:02 pm

Valentin Vornicu 10:22:05 pm Since 5 is a factor of f(3), let's work modulo 5. Valentin Vornicu
10:22:58 pm

Valentin Vornicu 10:23:00 pm When is 4^t + 1 divisible by 5? Mousie 10:24:10 when t is odd
pm

pokey0099 10:24:10 when t is odd AkshajK 10:24:10 when t is odd Bryan.C t is odd
pm

pm

10:24:10 pm

woodstock n is odd. maxmk

10:24:10 pm

10:24:10 pm

t is odd Minamoto t is odd


10:24:10 pm

teethpaste 10:24:10 when t is odd

pm

Valentin Vornicu 10:24:10 pm Since 4 is congruent to -1 modulo 5, 4^t + 1 is divisible by 5 if and only if t is odd. Valentin Vornicu 10:24:58 pm Now we must look at f(2t + 1) for t even. Let t = 2u. Then f(2t + 1) = f(4u + 1). Valentin Vornicu 10:25:02 pm What modulus can we work with? Mousie 10:26:15 now try 13? Minamoto 13
pm

10:26:15 pm

baijiangchen 13 vishankjs 13 Bryan.C 13

10:26:15 pm

10:26:15 pm

10:26:15 pm

jhfrost314 13

10:26:15 pm

pi314159265358979 13 AmericanPi 13

10:26:15 pm

10:26:15 pm

Valentin Vornicu 10:26:16 pm There's really only one choice. Since f(1) = 13^2, let's work modulo 13. Valentin Vornicu
10:26:20 pm

Valentin Vornicu 10:27:09 pm Thus, f(4u + 1) is always divisible by 13. Valentin Vornicu 10:27:33 pm To summarize, if n is even, then f(n) is divisible by 3 (and f(n) > 3), Valentin Vornicu 10:27:36 pm if n is of the form 4k + 3, then f(n) is divisible by 5 (and f(n) > 5), Valentin Vornicu 10:27:39 pm and if n is of the form 4k + 1, then f(n) is divisible by 13 (and f(n) > 13). Valentin Vornicu 10:27:42 pm Therefore, f(n) is composite for all nonnegative integers n. Valentin Vornicu 10:27:45 pm This problem shows that sometime you must try different moduli, even within the same problem, so don't be afraid to experiment. Valentin Vornicu
10:27:51 pm

Valentin Vornicu 10:28:44 pm You may recall Euclid's proof that there are an infinite number of primes. How can we start? Mousie 10:28:53 contradiction?
pm

hawqish 10:28:53 pm probably using contradiction, because it's proving infinite stuff greensteg2 10:28:53 Use contradiction.
pm

Valentin Vornicu 10:28:54 pm We argue by contradiction. Valentin Vornicu


10:28:57 pm

Valentin Vornicu Then what?

10:30:43 pm

SkinnySanta 10:31:09 pm consider either product of p_i plus three, or product of p_i plus 1 depending on parity of k Minamoto 10:31:09 pm take the largest three and multiply them; they're of the form 3n+2 and are then no longer the largest, done teethpaste 10:31:09 pm u multiply them together and add 1? hawqish 10:31:09 pm i think you can multiply them, and then p1p2p3....pk + 1 must also be a prime number.., that's how the regular proof goes, i think Valentin Vornicu
10:31:11 pm

Valentin Vornicu

10:31:15 pm

Valentin Vornicu 10:31:20 pm Suppose P reduces to 2 modulo 3. Then what can we say about P? DVA6102 10:32:34 pm we have a power of 2 mod 3 when we multiply all of the p_ks sdaops 10:32:34 pm It must have some prime factor that's 3n+2, since a bunch of 3n+1's would have a product that's 1 mod 3 greensteg2 10:32:34 pm It is another prime of the form 3n +2 that is prime with our assumptions, and we are done. Bryan.C 10:32:34 pm not divisibly by any of the p_i Valentin Vornicu 10:32:35 pm All primes (other than 3) are congruent to 1 modulo 3 and 2 modulo 3. If P was only divisible by primes congruent to 1 modulo 3, then P itself would be congruent to 1 modulo 3. But P is congruent to 2 modulo 3, so P must be divisible by at least one prime that is congruent to 2 modulo 3. Valentin Vornicu 10:33:41 pm But by construction, P is relatively prime to all primes that are congruent to 2 modulo 3, contradiction. Valentin Vornicu 10:33:43 pm The only case that is left is if P reduces to 0 modulo 3. Unfortunately, there's no good way to handle this case. (We can divide out all the factors of 3, but there's nothing we can say about the resulting quotient.) Valentin Vornicu 10:33:45 pm Is there any way to fix this argument? maxmk 10:33:52 pm consider 3p_1p_2p_3...p_n + 2.

Valentin Vornicu

10:33:56 pm

Valentin Vornicu 10:34:59 pm Then P always reduces to 2 modulo 3, and it is relatively prime to all the p_i, so the argument that we used above now works. Valentin Vornicu 10:35:35 pm More generally, if a and b are relatively prime positive integers, then there are infinitely many primes of the form an + b. This result is known as Dirichlet Theorem, but it is an advanced result and is difficult to prove. Valentin Vornicu 10:35:36 pm The ideas in Euclid's proof lead to the next few problems. Valentin Vornicu
10:35:40 pm

Valentin Vornicu 10:36:36 pm To get a feel for the problem, let's find the first few terms of the sequence. Valentin Vornicu What is p_2? AkshajK 3
10:37:22 pm

10:38:23 pm

pi314159265358979 3 sdaops 3
10:38:23 pm

10:38:23 pm

A123456789 3. shandongboy p2 = 3? Mousie 3?

10:38:23 pm

10:38:23 pm

10:38:23 pm

Minamoto 3

10:38:23 pm

Valentin Vornicu 10:38:25 pm p_2 is the largest prime divisor of p_1 + 1 = 3, so p_2 = 3. Valentin Vornicu What is p_3? greensteg2 p_3 = 7 ABCDE 7
10:38:28 pm

10:38:36 pm

10:38:36 pm

Minamoto 7 woodstock 7 sdaops 7 DVA6102 7.

10:38:36 pm

10:38:36 pm

10:38:36 pm

10:38:36 pm

greensteg2 7 jhfrost314 7

10:38:36 pm

10:38:36 pm

Valentin Vornicu

10:38:37 pm

p_3 is the largest prime divisor of p_1 p_2 + 1 = 7, so p_3 = 7. Valentin Vornicu What is p_4? shandongboy 43 baijiangchen 43 woodstock 43 jhfrost314 43 Mousie 43 ksun48 43
10:38:39 pm

10:39:37 pm

10:39:37 pm

10:39:37 pm

10:39:37 pm

10:39:37 pm

10:39:37 pm

Valentin Vornicu 10:39:38 pm p_4 is the largest prime divisor of p_1 p_2 p_3 + 1 = 43, so p_4 = 43. Valentin Vornicu What is p_5? teethpaste 139
10:39:40 pm

10:40:46 pm

GeorgiaTechMan no 139! aopsqwerty 139

10:40:46 pm

10:40:46 pm

Valentin Vornicu 10:40:47 pm p_5 is the largest prime divisor of p_1 p_2 p_3 p_4 + 1 = 1807 = 13 x 139, so p_4 = 139. Valentin Vornicu 10:40:51 pm We must show that 5 never appears in the sequence. Valentin Vornicu 10:40:55 How can we proceed? AkshajK 10:42:13 contradiction
pm pm

Valentin Vornicu 10:42:14 pm We can argue by contradiction. Valentin Vornicu


10:42:17 pm

Valentin Vornicu 10:42:20 pm What does that say about this expression? Mousie 10:43:18 pm contains powers of 2, 3, and 5 only baijiangchen 10:43:18 divisible by only 2 3 5 Valentin Vornicu
pm

10:43:19 pm

Valentin Vornicu 10:43:22 Can we say more?

pm

sdaops 10:43:32 pm Furthermore, it's odd, so there aren't any factors of 2 anwang16 its odd
10:43:35 pm

Valentin Vornicu

10:43:38 pm

Valentin Vornicu 10:44:09 pm Can we say anything else? v_Enhance 10:44:24 pm power of five, since 2,3 are taken SkinnySanta 10:44:24 pm that the expression is a power of 5 since it isn't divisible by 2 or 3 AmericanPi 10:44:24 pm 2 and 3 aren't factors maxmk 10:44:24 pm p2 = 3, so it is a power of 5 anwang16 10:44:24 pm Also not divisible by 3, so its only 5 Valentin Vornicu
10:44:53 pm

Valentin Vornicu

10:44:55 pm

Valentin Vornicu

10:45:00 pm

Valentin Vornicu 10:45:47 pm What can we do with this equation? AkshajK 10:46:05 pm however this is divisible by 4 SkinnySanta 10:46:05 pm factor as a difference of powers, so 4|product contradiction GeorgiaTechMan 10:46:05 pm but 5^k-1 is divisible by 4! AmericanPi mod 4
10:46:05 pm

Mousie 10:46:05 pm when k is even you can factor 5^k - 1 and you see that 4 is a factor...contradiction ABCDE 10:46:05 pm divisible by 5-1=4, which means that two of the primes are 2 which is a contradiction A123456789 10:46:05 pm The left side of the equation cannot be divisible by 4, so this is a contradiction. sdaops 10:46:05 pm That product must be divisible by 4 hawqish 10:46:05 pm 5^k - 1 is always divisible by 4, so there's a contridiction, 4 isn't prime Valentin Vornicu
10:46:06 pm

Valentin Vornicu 10:46:12 pm This equation tells us that 4 must be a factor of the left-hand side. Valentin Vornicu 10:46:15 pm But this is impossible because p_1 = 2 and all further terms p_2, ..., p_{n - 1} are odd. Valentin Vornicu 10:46:18 pm We have a contradiction, so 5 cannot be a member of this sequence. Valentin Vornicu
10:46:24 pm

Valentin Vornicu 10:46:51 pm This is a problem I proposed for the IMO TST in 2003. Only the 6 guys who actually made the team got a 7 on it. Valentin Vornicu 10:48:32 pm Let's start by looking at small primes. Is the prime 2 in M? ksun48 yeah!
10:49:26 pm

shandongboy yes?

10:49:26 pm

AkshajK 10:49:26 pm yes, because the expression can be even sdaops 10:49:26 pm Yes, since we have at least 2 odd primes we can multiply together Valentin Vornicu 10:49:31 pm Since M contains at least three primes, there is an odd prime in M, say p. Let A = {p}. Then by the condition in the problem, all the prime factors of p - 1 are also in M. Valentin Vornicu 10:49:34 pm But p - 1 is even, so 2 is in M. Valentin Vornicu 10:49:36 Is the prime 3 in M?
pm

sdaops 10:50:40 pm Yes; if there's a prime that's 1 mod 3, use that one, otherwise take 2 * a prime that's 2 mod 3 baijiangchen 10:50:40 pm Either a prime is 3n+1, so yes, or two primes are 3n+2, so yes Minamoto 10:50:40 pm yes, all primes are 2 mod 3, so we have 2*2-1 henrypickle 10:50:40 pm take the subset of all p congruent to 1 mod 3? Valentin Vornicu 10:50:44 pm If M contains a prime of the form p = 3n + 1, then take A = {p}. Then all the factors of p - 1 = 3n are also in M, so 3 is in M. Valentin Vornicu 10:50:58 pm Otherwise, M contains two primes of the form 3n + 2, say p and q. Valentin Vornicu 10:51:01 pm We can take A = {p, q}. Then all the prime factors of pq - 1 are also in M. Valentin Vornicu 10:51:06 pm But pq - 1 is congruent to 2 x 2 - 1 = 3, or 0 modulo 3, so pq - 1 is divisible by 3. Therefore, 3 is in M. In either case, 3 is in M. Valentin Vornicu 10:51:09 Is the prime 5 in M?
pm

Mousie 10:52:31 yes 2*3 - 1

pm

AmericanPi 10:52:31 yes, take 2 and 3 baijiangchen 2*3-1

pm

10:52:31 pm

sdaops 10:52:31 2*3-1=5 ksun48 10:52:31 yeah 2*3-1

pm

pm

Valentin Vornicu 10:52:32 pm We know 2 and 3 are in M, so take A = {2, 3}. Then 2 x 3 - 1 = 5, so 5 is in M. Valentin Vornicu 10:53:26 Is the prime 7 in M? Minamoto 10:53:37 yes, 7|3*5-1 sdaops 10:53:37 3*5-1=14
pm pm pm

AmericanPi 10:53:37 yes, use 3 and 5 jhfrost314 10:53:37 yes 3*5-1=14

pm

pm

Valentin Vornicu 10:53:38 pm We know 3 and 5 are in M, so take A = {3, 5}. Then 3 x 5 - 1 = 14 = 2 x 7, so 7 is in M. Valentin Vornicu 10:53:54 pm Hence, the first few primes are in M. We want to prove that M contains all primes. Valentin Vornicu 10:53:58 pm What intermediate result can we try proving first? GeorgiaTechMan 10:54:48 pm we can prove theres an infinite number of primes in it now. Minamoto 10:54:48 prove M is infinite mentalgenius M is infinite
pm

10:54:48 pm

sdaops 10:54:48 pm Prove that M is infinite baijiangchen 10:54:48 pm It contains infinitely many primes Valentin Vornicu 10:54:49 pm Let's try proving first that M is infinite. Valentin Vornicu
10:54:52 pm

Valentin Vornicu 10:55:56 pm As a start, let's take A to be all the primes in M, except one, say p_i. (Remember that A must be a proper subset of M.) Valentin Vornicu
10:56:00 pm

Valentin Vornicu

10:56:05 pm

sdaops 10:56:40 pm It's a power of p_i baijiangchen 10:56:40 p_i ^ n = that thing Valentin Vornicu
pm

10:56:42 pm

Valentin Vornicu

10:57:13 pm

Valentin Vornicu

10:57:21 pm

Valentin Vornicu 10:57:26 pm To make these equations simpler to work with, we start by omitting the smallest primes. Valentin Vornicu
10:57:30 pm

Valentin Vornicu

10:57:36 pm

Valentin Vornicu

10:57:42 pm

Valentin Vornicu

10:57:48 pm

Valentin Vornicu

10:58:58 pm

What can we do with these two expressions for P? baijiangchen Equate them!
10:59:43 pm

Valentin Vornicu 10:59:44 We can equate them. Valentin Vornicu

pm

10:59:49 pm

Valentin Vornicu 10:59:53 pm Let's start analyzing this equation. Valentin Vornicu
10:59:57 pm

Valentin Vornicu 11:00:01 pm Also, P is at least 2 x 3 x 5 x 7. So what can we say about a? sdaops 11:00:57 a is at least 7 woodstock a >= 7
pm

11:00:57 pm

Valentin Vornicu

11:00:59 pm

Valentin Vornicu

11:01:02 pm

Valentin Vornicu 11:01:08 pm So what can we say about b? jhfrost314 b=1


11:01:51 pm

sdaops 11:01:51 In fact, b is 1 SkinnySanta b =1

pm

11:01:51 pm

Valentin Vornicu

11:01:53 pm

Valentin Vornicu

11:01:56 pm

Valentin Vornicu

11:02:02 pm

Valentin Vornicu Valentin Vornicu Therefore, b = 1. Valentin Vornicu

11:02:04 pm

11:02:08 pm

11:02:40 pm

Valentin Vornicu 11:02:44 pm How can we analyze this equation? Valentin Vornicu 11:03:04 pm (Sorry guys we're going to go a few minutes over -- hope you don't mind baijiangchen 11:03:11 pm um Catalan's conjecture but you won't like that Valentin Vornicu 11:03:37 pm Yes, you didn't get the points. There was one guy who said that and I didn't give him the full 7 points for it. He was unable to prove it when I asked him about it the next day ;) Valentin Vornicu 11:04:08 pm Only use name theorems you know how to prove is the guidelines after which the grading is done usually. Valentin Vornicu 11:04:15 pm We can try using modular arithmetic. What modulus can we try? jhfrost314 mod 3
11:04:22 pm

Valentin Vornicu 11:04:23 pm We can try reducing modulo 3. Valentin Vornicu


11:04:28 pm

Valentin Vornicu 11:04:39 pm What does this congruence tell us about a? baijiangchen so, a is odd hawqish a is odd jhfrost314 its odd
11:05:38 pm

11:05:38 pm

11:05:38 pm

DVA6102 11:05:38 a==1 mod 2 woodstock a is odd. Mousie odd

pm

11:05:38 pm

11:05:38 pm

teethpaste odd?

11:05:38 pm

Valentin Vornicu 11:05:39 pm This congruence tells us that a - 1 is even. Let a - 1 = 2n. Valentin Vornicu
11:05:43 pm

Valentin Vornicu 11:05:47 pm What can we do with this equation? Mousie 11:06:58 factor right side
pm

difference of squares
11:06:59 pm

Valentin Vornicu

Valentin Vornicu 11:07:05 pm This equation says that both 2^n + 1 and 2^n - 1 are powers of 3. Valentin Vornicu 11:07:19 pm But (2^n + 1) - (2^n - 1) = 2. Valentin Vornicu 11:07:24 pm The only powers of 3 that differ by 2 are 3 and 1, so n = 1, and a = 3. greensteg2 11:07:28 Contradiction. hawqish 11:07:28 pm which is impossible Valentin Vornicu 11:07:29 pm But a is at least 7, which is a contradiction. Valentin Vornicu 11:07:32 pm Therefore, M contains an infinite number of primes. Valentin Vornicu 11:07:35 pm We're actually very close to the end now. Valentin Vornicu
11:07:38 pm pm

Valentin Vornicu 11:07:44 pm What result could we try using here? sdaops 11:08:33 pm Look at the entire set modulo an allegedly missing prime Valentin Vornicu 11:08:35 pm We could try using Fermat's Little Theorem. Valentin Vornicu
11:08:38 pm

aopsqwerty 11:09:49 pm P=something to the power of p-1 GeorgiaTechMan 11:09:49 pm there are at least p-1 numbers congruent to a certain residue mod p which is obvious Valentin Vornicu
11:10:00 pm

Valentin Vornicu 11:10:06 pm The primes p_1, p_2, ..., p_{p - 1} must also be distinct from p. But if p was in M, then this argument would not be necessary in the first place. Valentin Vornicu 11:10:11 pm How can we show that such primes exist? SkinnySanta 11:11:01 pm pigeonhole since M is infiinite GeorgiaTechMan pigeonhole baijiangchen Pidgeonhole!
11:11:01 pm

11:11:01 pm

sdaops 11:11:01 pm Shoot holes in some pigeons (or is it the other way around?) Valentin Vornicu
11:11:03 pm

We have shown that the set M is infinite. Valentin Vornicu 11:11:06 pm Hence, if we reduce every element in M modulo p, then some nonzero residue must appear an infinite number of times. Valentin Vornicu 11:11:09 pm Then we simply take p - 1 primes p_1, p_2, ..., p_{p - 1} that have this residue. Valentin Vornicu 11:11:13 pm Therefore, M contains every prime p. Valentin Vornicu SUMMARY
11:11:18 pm

Valentin Vornicu 11:11:20 pm In today's class, we saw how to effectively use Fermat's Little Theorem and Euclid's Theorem. We also saw the power of algebra in number theory problems. We can solve many number theory problems effectively by putting a number in the right form (like expressing an even number in the form 2n), and using factorization. When you see a number theory problem, convert the words into mathematics, and let the equations do the work. Valentin Vornicu 11:11:23 pm Finally, we also saw how to solve problems using argument by contradiction. If a problem asks you to prove that something cannot occur, this is usually a good sign to use contradiction. Valentin Vornicu 11:11:26 That's it for today's class.
pm

Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others. Copyright 2013 AoPS Incorporated. All Rights Reserved. Contact Us Privacy Terms & Conditions

Art of Problem Solving


WOOT
Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others.

Class Transcript 01/25 - Modular Arithmetic B


Valentin Vornicu 7:30:29 pm WOOT 2012-13: Modular Arithmetic B Valentin Vornicu 7:30:35 pm In today's lecture, we will continue our look at number theory and modular arithmetic. The first half of our class will focus on the use of Fermat's Little Theorem and Euler's Theorem. In the second half, we'll cover some more general number theory problems that are typical for this level. Valentin Vornicu 7:30:39 pm We'll start with a nice warm-up exercise. Valentin Vornicu
7:31:03 pm

Valentin Vornicu 7:32:06 pm How can we find the missing digit? trophies 7:33:29 take it mod 9! BOGTRO 7:33:29 take it mod 9
pm

pm

superpi83 7:33:29 take it mod 9 archer4 7:33:29 take it mod 9

pm

pm

negativebplusorminus Find 2^29 mod 9 swimmer 7:33:29 pm find 2^29 (mod 9) frtennis1 7:33:29 Take it mod 9
pm

7:33:29 pm

Valentin Vornicu 7:33:35 pm We can try to find the sum of the 9 digits. If all 10 digits were present, then their sum would be 0 + 1 + 2 + ... + 9 = 45. However, we can't find the sum of the 9 digits without computing 2^29. Valentin Vornicu 7:33:39 pm We know that a number and the sum of its digits are congruent modulo 9, so we can reduce 2^29 modulo 9. What do we get? archer4 5
7:34:25 pm

trophies 7:34:25 it is 5(mod 9) swimmer 5

pm

7:34:25 pm

cire_il 7:34:25 5 mod 9

pm

negativebplusorminus frtennis1 5 (mod 9)

7:34:25 pm

7:34:25 pm

Valentin Vornicu

7:34:34 pm

Valentin Vornicu 7:34:40 pm So what is the missing digit? trophies 7:35:43 pm the missing digit is 4! addictedtomath 4 cheetah17 4 Wolfkit 4
7:35:43 pm

7:35:43 pm

7:35:43 pm

slidersknob 4 Fersolve 4 JRY 4

7:35:43 pm

7:35:43 pm

7:35:43 pm

davidzhang 4

7:35:43 pm

Valentin Vornicu 7:35:45 pm The missing digit must be congruent to -5 modulo 9, so the missing digit is 4. Valentin Vornicu 7:35:48 pm Indeed, 2^29 = 536870912. All digits are present except 4. Valentin Vornicu
7:35:51 pm

Valentin Vornicu How do we start?

7:36:29 pm

trophies 7:36:44 pm We can simplify it mod 1000 negativebplusorminus 7:36:44 pm We want 9^8^7 mod 1000; split this up into its prime factors. exmath89 7:36:44 use mod 1000
pm

thecmd999 7:36:44 pm last three digits is mod 1000 archer4 7:36:44 pm take it mod 1000 dasobson 7:36:44 modulo 1000?
pm

Valentin Vornicu 7:36:47 pm Finding the last three digits of a number is an indication to work modulo 1000. Valentin Vornicu
7:37:13 pm

Valentin Vornicu 7:37:15 pm How can we reduce this? gundraja 7:37:57 pm Use euler's theorem. trophies 7:37:57 pm use euler's theorem thecmd999 7:37:57 pm note that phi(1000)=400

exmath89 7:37:57 phi(1000)=400

pm

AndrewKwon97 7:37:57 find 8^7 mod 400

pm

Valentin Vornicu 7:38:00 pm We can use Euler's Theorem. Valentin Vornicu


7:38:03 pm

Valentin Vornicu

7:38:29 pm

Valentin Vornicu 7:38:34 pm Probably the easiest way to compute this is to look at the first few powers of 8 modulo 400. We do this by starting with 1, then multiplying each term by 8 and reducing modulo 400. Valentin Vornicu 7:38:37 pm This gives us 1, 8, 64, 112, 96, 368, 144, 352. Valentin Vornicu
7:38:59 pm

Valentin Vornicu 7:39:25 pm We have gone as far as we can with Euler's Theorem. We can try to compute this by looking at the first few powers of 9 modulo 1000, but there is an easier way. Valentin Vornicu 7:39:30 pm Is there anything you notice about the number 9 that may make it easier to compute this power modulo 1000? negativebplusorminus 10-1 patrickhompe =10-1 frtennis1 9 = 10 - 1
7:39:46 pm

7:39:46 pm

7:39:46 pm

thecmd999 9=10-1

7:39:46 pm

thecmd999 7:39:46 binomial theorem Valentin Vornicu

pm

7:39:51 pm

Valentin Vornicu

7:40:13 pm

Valentin Vornicu 7:40:22 pm Therefore, the last three digits of 9^(8^7) are 081. Valentin Vornicu
7:40:29 pm

MathWhizzz 7:41:34 pm shwo divisibility by 2 , 3, 5 negativebplusorminus Consider it mod 2,3,5


7:41:34 pm

apple413 7:41:34 pm take mods of 2, 3, 5 trophies 7:41:34 pm show that it is divisible by 2,3, and 5 Valentin Vornicu 7:41:45 pm We could verify that n^5 - n is divisible by 30 for n = 0, 1, 2, ..., 29. Valentin Vornicu 7:41:54 pm But we can do it easier: we use the prime factorization of 30 = 2 x 3 x 5. Hence, it suffices to show that n^5 - n is divisible by 2, 3, and 5, for all integers n. Valentin Vornicu 7:41:58 Is n^5 - n divisible by 2?
pm

JRY 7:42:41 pm yes, n^5 and n have the same parity negativebplusorminus Same parity, so yes
7:42:41 pm

Fersolve 7:42:41 pm yes, odd-odd=even, even-even=even Waffle15 7:42:41 pm yes, since both terms always have the same parity MathWhizzz 7:42:41 pm both terms are 1 mod 2 or both terms are 0 mod 2 so difference is 0 bobbyq yes
7:42:41 pm

Valentin Vornicu 7:42:47 pm Yes. By reducing modulo 2, we can check for n = 0 and n = 1. Valentin Vornicu 7:42:50 Is n^5 - n divisible by 3?
pm

MathWhizzz 7:43:52 pm yes, n^2=1 mod 2, so n^4*n=n mod 3, so n^5-n=0mod3 thecmd999 7:43:52 pm yes; 1^5=1, 2^5=2, 0^5=0 mod 3 gundraja 7:43:52 pm Yes, since n^5 and n are in the same congruence class (mod 3). Valentin Vornicu 7:43:58 pm Yes. By reducing modulo 3, we can check for n = 0, 1, and 2. Valentin Vornicu 7:44:01 Is n^5 - n divisible by 5? swimmer Fermat's
7:44:44 pm pm

archer4 7:44:44 pm yes; by euler's totient formula screenhunter372 7:44:44 pm yes by the same principle, only n= 0, 1, 2, 3, 4, and 5 superpi83 7:44:44 pm yes, because by Euler's Theorem n^5==n (mod 5) trophies 7:44:44 pm yes, we can use fermat's little theorem to prove it gundraja 7:44:44 pm Yes for the same reason as before. Valentin Vornicu 7:44:50 pm Yes. By reducing modulo 5, we can check for n = 0, 1, 2, 3, and 4. We can also simply cite Fermat's Little Theorem for the prime p = 5. Valentin Vornicu
7:44:54 pm

Hence, n^5 - n is divisible by 2 x 3 x 5 = 30 for all integers n. Valentin Vornicu 7:44:57 pm This simple problem illustrates an important principle. If you want to prove that a number is divisible by n, it helps to look at the individual prime powers of n. Valentin Vornicu
7:45:02 pm

Valentin Vornicu 7:45:16 pm We start by trying to reduce the given expression modulo 2000, but all the bases are already less than 2000. So what else can we do? superpi83 7:46:12 pm take the expression mod 16 and mod 125 thecmd999 7:46:12 split into primes
pm

MathWhizzz 7:46:12 pm mod 16 and mod 125 addictedtomath 7:46:12 pm show divisibility by 2^4 and 5^3 slidersknob 7:46:12 pm reduce by modulo 16 and 125 exmath89 7:46:12 2000=2^4*5^3 JRY 7:46:12 pm mod 16 and 125 Valentin Vornicu 7:46:19 pm The prime factorization of 2000 is 2^4 x 5^3, so we look at the factors 2^4 = 16 and 5^3 = 125 separately. Valentin Vornicu 7:46:22 pm How does the given expression reduce modulo 16? superpi83 7:47:21 pm 25^n-25^n+(-4)^n-(-4)^n=0 yay gundraja 7:47:21 pm 9^n - 9^n + (-4)^n - (-4)^n = 0. negativebplusorminus It's 0. Valentin Vornicu
7:47:21 pm pm

7:47:23 pm

Valentin Vornicu 7:47:29 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 16. Valentin Vornicu
7:47:44 pm

Valentin Vornicu 7:47:49 pm Therefore, 121^n - 25^n + 1900^n - (-4)^n is divisible by 125. Valentin Vornicu 7:48:31 pm We conclude that 121^n - 25^n + 1900^n - (-4)^n is divisible by 2000 for all positive integers n.

Valentin Vornicu

7:48:36 pm

Valentin Vornicu 7:49:09 pm What can we do with this equation? superpi83 7:50:12 pm take it in a convenient mod that limits the possibilities for n tc1729 7:50:12 pm find bounds on n BOGTRO 7:50:12 pm Take it mod 5, should give some useful stuff due to FLT screenhunter372 reduce (mod 5) ?
7:50:12 pm

Valentin Vornicu 7:50:25 pm We can take mods. In fact, we can use a previous problem and reduce it modulo 30. This gives us Valentin Vornicu
7:50:32 pm

Valentin Vornicu 7:51:05 pm Now we need to find bounds on n to find its exact value. What's one obvious lower bound on n? thecmd999 7:51:26 134 i suppose BOGTRO 133 superpi83 133... swimmer 133 < n frtennis1 133
7:51:26 pm pm

7:51:26 pm

7:51:26 pm

7:51:26 pm

negativebplusorminus 133 Valentin Vornicu Clearly n > 133.

7:51:26 pm

7:51:27 pm

Valentin Vornicu 7:51:54 pm Now we need to find an upper bound on n. Valentin Vornicu 7:51:57 pm The largest term in the left-hand side is 133^5, so we compare all the other terms to 133^5. Valentin Vornicu 7:52:01 pm We know that 110^5 < 133^5. Valentin Vornicu 7:52:04 pm Can we compare 84^5 + 27^5 and 133^5? BOGTRO 7:52:59 pm 84^5+27^5<(84+27)^5=111^5 swimmer 7:52:59 pm (84 ^ 5 + 27 ^ 5) < (84 + 27) ^ 5 < 133 ^ 5 superpi83
7:52:59 pm

84^5+27^5<(84+27)^5=111^5<133^5 BOGTRO 7:52:59 pm 84^5+27^5<(84+27)^5=111^5<133^5 yugrey 7:52:59 pm 84^5+27^5<(84+27)^5<133^5 frtennis1 7:52:59 pm 84^5 + 27^5 < 133^5 Valentin Vornicu 7:53:07 pm Note that 84^5 + 27^5 < (84 + 27)^5 = 111^5 < 133^5. Valentin Vornicu
7:53:12 pm

Valentin Vornicu 7:53:17 pm What we would like to do now is take the fifth root of both sides. The problem is we would get the fifth root of 3, and it's not clear what to do with this number. Valentin Vornicu 7:53:23 pm So let's see if we can replace 3 by the ratio of two fifth powers. Valentin Vornicu 7:53:31 pm The first few fifth powers are 1^5 = 1, 2^5 = 32, 3^5 = 243, 4^5 = 1024, 5^5 = 3125, and 6^5 = 7776, and so on. Do we see any fifth powers that approximately have a ratio of 3? thecmd999 5^5/4^5 BOGTRO 5 and 4
7:53:55 pm

7:53:55 pm

superpi83 7:53:55 pm 5^5/4^5 seems to work swimmer 4, 5


7:53:55 pm

gundraja 7:53:55 4^5 and 5^5. screenhunter372 4 and 5 tc1729 7:53:55 5^5/4^5

pm

7:53:55 pm

pm

Valentin Vornicu 7:53:56 pm We see that 5^5/4^5 = 3125/1024 is approximately 3. Valentin Vornicu 7:54:05 pm Even better, this ratio is slightly larger than 3, which means we can use this ratio to get an upper bound on n: Valentin Vornicu
7:54:17 pm

Valentin Vornicu 7:54:21 pm Therefore, n < 5/4 x 133 = 166 + 1/4. Valentin Vornicu So what is n? thecmd999 it's 144 BOGTRO 144
7:54:25 pm

7:55:15 pm

7:55:15 pm

cheetah17 144

7:55:15 pm

screenhunter372

7:55:15 pm

144 superpi83 144 swimmer 144


7:55:15 pm

7:55:15 pm

MathWhizzz 144

7:55:15 pm

negativebplusorminus 144

7:55:15 pm

Valentin Vornicu 7:55:16 pm The only n satisfying 133 < n < 166 + 1/4, and n congruent to 24 modulo 30 is n = 144. Valentin Vornicu
7:55:22 pm

Valentin Vornicu 7:55:38 pm To make things easier, let s(N) denote the sum of the digits of N for a positive integer N. Then A = s(4444^4444) and B = s(A). Let C = s(B). Then we want to find C. Valentin Vornicu 7:56:04 pm Is there anything we can say about s(N)? yugrey 7:56:54 pm It is congruent to n mod 9. superpi83 7:56:54 pm congruent to N mod 9 BOGTRO 7:56:54 s(n)=n mod 9
pm

mathepic 7:56:54 pm s(N) == N (mod 9) gundraja 7:56:54 s(n) = n (mod 9). tc1729
pm

7:56:54 pm

Fersolve mod 9

7:56:54 pm

exmath89 7:56:54 s(N)==N mod 9

pm

Valentin Vornicu 7:57:00 pm We know that s(N) is congruent to N modulo 9. Valentin Vornicu 7:57:07 pm So we can we say that C is congruent to 4444^4444 modulo 9. Valentin Vornicu 7:57:11 pm How can we compute this residue? thecmd999 7:58:23 pm 4444 is congruent to 16 or -2 mod 9 so then we use euler negativebplusorminus 4444 == 7 mod 9 tc1729 7:58:23 4444 = 7 mod 9
pm 7:58:23 pm

bobbyq 7:58:23 pm 4444==-2 (mod 9) Valentin Vornicu


7:58:26 pm

Valentin Vornicu 7:58:32 pm How can we simplify this? mathepic Euler's


7:59:14 pm

superpi83 7:59:14 pm reduce exponent mod phi(9)=6 swimmer 7:59:14 euler's totient BOGTRO 7:59:14 7^6=1 (mod 9) bobbyq 7:59:14 euler's theorem yugrey 7:59:14 7^6=1 mod 9 tiger21 7:59:14 Euler's theorem
pm

pm

pm

pm

pm

Valentin Vornicu 7:59:16 pm We can use Euler's Theorem. Valentin Vornicu


7:59:19 pm

Valentin Vornicu

7:59:23 pm

Valentin Vornicu

7:59:58 pm

Valentin Vornicu 8:00:06 pm Therefore, C is congruent to 7 modulo 9. Valentin Vornicu 8:00:09 pm How can we find the exact value of C? MathWhizzz 8:01:12 find bounds on C superpi83 8:01:12 find bounds
pm

pm

BOGTRO 8:01:12 pm find the number of digits in B gundraja 8:01:12 pm Figure out how big C is. exmath89 8:01:12 use some bounds
pm

frtennis1 8:01:12 pm Get a bound on it's size panjia123 8:01:12 estimate its value
pm

Valentin Vornicu 8:01:14 pm Let's proceed by finding an upper bound on A, which is the sum of the digits of 4444^4444. What is an upper bound on A? mathepic 8:02:49 9 floor log N
pm

MathWhizzz 8:02:49 pm num digits < 4444log4444 < 4444*4 Valentin Vornicu 8:02:53 pm We can try to find the exact number of digits in 4444^4444, but that would involve using logarithms. What if we didn't have a calculator? Can we find an estimate on A, even a crude one? ProfessorBeast 8:04:22 pm well we know that 4444<10000=10^5? Valentin Vornicu
8:04:25 pm

frtennis1

8:04:41 pm

Valentin Vornicu 8:04:46 pm Therefore, A <= 9 x 17776 = 159984. Valentin Vornicu 8:04:56 pm Then we might conclude that B = s(A) is at most s(159984), but this would not be correct. Why not? panjia123 8:06:27 pm s(159984) < s(99999) etc. EricMathPath09 8:06:27 s(159899) > s(159984)
pm

apple413 8:06:27 pm because if it were 999999, s(A) is larger frtennis1 8:06:27 pm for example 99999 < 159984, yet S(99999) > S(159984) ProfessorBeast 8:06:27 pm because, for example, A could be 99999 which has a greater digit sum chchend 8:06:27 pm it would be at most s(99999) Valentin Vornicu 8:06:32 pm The function s(N) is not an increasing function. Valentin Vornicu 8:06:36 pm Of the numbers 1, 2, 3, ..., 159984, the number with the greatest s-value is 99999. Valentin Vornicu 8:06:39 pm Therefore, B = s(A) <= 5 x 9 = 45. Valentin Vornicu 8:06:43 pm Now we want a bound on C = s(B). What can we do? swimmer 8:08:13 pm greatest is 39 = 12 superpi83 8:08:13 C<=s(39)=12
pm

davidzhang 8:08:13 pm s(45) < s(39) = 12 so the bound on C is 12 superbob1235 8:08:13 pm s(B) has a maximum at 12, when B = 39, so C <= 12 mathepic 8:08:13 pm B = 39 yields a maximum value of 12 for C apple413 8:08:13 39 is largest s(B)
pm

Valentin Vornicu 8:08:24 pm We can repeat the argument. Of the numbers 1, 2, 3, ..., 45, the number with the greatest s-value is 39.

Valentin Vornicu 8:08:27 pm Therefore, C = s(B) <= 3 + 9 = 12. Valentin Vornicu 8:08:31 pm We have shown that C is congruent to 7 modulo 9, so we conclude that C = 7. Valentin Vornicu
8:08:36 pm

Valentin Vornicu 8:09:39 pm What makes this sum difficult to deal with? exmath89 8:09:59 absolute values
pm

thecmd999 8:09:59 absolute values

pm

negativebplusorminus The absolute values. panjia123 8:09:59 absolute values JRY 8:09:59 pm absolute values shandongboy 8:09:59 Absolute values
pm

8:09:59 pm

pm

Valentin Vornicu 8:10:01 pm The absolute value signs make this sum difficult to deal with. How can we deal with the absolute value signs? Is there a convenient way of getting rid of them? panjia123 8:11:39 WLOG a_i >= b_i screenhunter372 assume a > b?
pm

8:11:39 pm

exmath89 8:11:39 pm we can assume a_i>b_i for all i swimmer 8:11:39 let ai > bi?
pm

frtennis1 8:11:39 pm Define WLOG that a_i > b_i tc1729 8:11:39 wlog a_i > b_i
pm

Valentin Vornicu

8:11:42 pm

Valentin Vornicu

8:11:49 pm

Valentin Vornicu 8:11:51 pm Each term in this sum is 1 or 6. How can we use this information? exmath89 8:13:13 pm let x=number of 1s and y=number of 6s Fersolve
8:13:13 pm

k(1)+(999-k)6 mod 10 Valentin Vornicu 8:13:17 pm There are 999 terms, so let k of them be equal to 1, so the remaining 999 - k are equal to 6. Valentin Vornicu
8:13:20 pm

Valentin Vornicu

8:13:23 pm

panjia123 8:14:17 pm they're in {1,...,1998} apple413 8:14:17 1<=a, b<=1998


pm

swimmer 8:14:17 pm numbers from 1 to 1998 gundraja 8:14:17 pm They range from 1,..., 1998 and they are distinct. ProfessorBeast 8:14:17 pm the 999 pairs contain the numbers 1-1998 once each Valentin Vornicu
8:14:21 pm

Valentin Vornicu 8:14:23 pm So can we turn this fact into a related sum? negativebplusorminus 8:15:52 Their sum is 999(1000)/2
pm

exmath89 8:15:52 pm a_1+a_2+...+a_999+b_1+b_2+...b_999=1999(999) gundraja 8:15:52 pm a_1 + b_1 + ... + a_999 + b_999 = 999*1999. JRY 8:15:52 pm a_1+b_1+a_2+b_2+...+a_999+b_999=1+2+...+1998 apple413 8:15:55 pm the sum of all ai, bi =1997001 Valentin Vornicu
8:15:57 pm

Valentin Vornicu

8:16:01 pm

Valentin Vornicu 8:16:04 pm What can we do with these two equations? qfoam 8:17:22 add them MathWhizzz add!
pm

8:17:22 pm

trophies add them apple413 add them frtennis1 add them

8:17:22 pm

8:17:22 pm

8:17:22 pm

negativebplusorminus add screenhunter372 add them hikerboy 8:17:22 Add them! tc1729 8:17:22 add them

8:17:22 pm

8:17:22 pm

pm

pm

Valentin Vornicu

8:17:23 pm

Valentin Vornicu 8:17:31 pm Does this equation say anything about the value of k? trophies 8:17:44 pm now we realize that k must be odd! screenhunter372 k must be odd swimmer odd Fersolve k is odd dooba It's odd
8:17:44 pm

8:17:44 pm

8:17:44 pm

8:17:44 pm

apple413 it's odd

8:17:44 pm

Valentin Vornicu 8:17:45 pm The left-hand side must be even. Therefore, 5k must be odd, which means k must be odd. Valentin Vornicu 8:17:48 pm Let k = 2n + 1 for some integer n. Valentin Vornicu 8:17:52 pm We can substitute into the equation above, to get Valentin Vornicu
8:17:55 pm

Valentin Vornicu 8:18:55 pm It looks like we can conclude that this number always ends in 9, but what if n is sufficiently high so that 5989 - 10n is negative? For example, if n = 600, then 5989 - 10n = -11, which ends in 1. BOGTRO 8:19:38 pm we assumed that a_i>b_i ProfessorBeast 8:19:38 pm we said at the beginning that all ai-bi must be positive frtennis1 8:19:38 pm the absolute values mean that it must be positive superpi83 8:19:38 pm the original sum is positive

slidersknob 8:19:38 absolute values

pm

Valentin Vornicu 8:19:39 pm The original sum was expressed as a sum of absolute values, which must be nonnegative. And a nonnegative number of the form 5989 - 10n must end in 9, so we are home free. Valentin Vornicu
8:19:44 pm

Valentin Vornicu 8:19:51 How can we proceed? shandongboy Contradiction?


8:20:54 pm

pm

yugrey 8:20:54 Contradiction.

pm

BOGTRO 8:20:54 contradiction

pm

Valentin Vornicu 8:20:59 pm We can argue by contradiction. Valentin Vornicu 8:21:03 pm Note that 2 x 5 - 1 = 9, 2 x 13 - 1 = 25, and 5 x 13 - 1 = 64 are all perfect squares. Valentin Vornicu 8:21:09 pm Suppose 2d - 1 = x^2, 5d - 1 = y^2, and 13d - 1 = z^2, for some positive integers x, y, and z. Valentin Vornicu 8:21:15 pm What does the first equation 2d - 1 = x^2 tell us? gundraja x is odd.
8:21:54 pm

screenhunter372 x is odd

8:21:54 pm

martian179 8:21:54 x must be odd shandongboy x is odd? trophies x is odd

pm

8:21:54 pm

8:21:54 pm

Valentin Vornicu 8:21:58 pm The number 2d - 1 is always odd, so x is odd. Valentin Vornicu 8:22:01 pm Let x = 2n + 1 for some integer n. Then 2d - 1 = x^2 = (2n + 1)^2 = 4n^2 + 4n + 1. Valentin Vornicu 8:22:04 pm Then d = 2n^2 + 2n + 1. What does this tell us about d? MathWhizzz d is odd thecmd999 d is odd archer4 d is odd
8:22:21 pm

8:22:21 pm

8:22:21 pm

slidersknob odd exmath89 d is odd swimmer d is odd

8:22:21 pm

8:22:21 pm

8:22:21 pm

Valentin Vornicu 8:22:30 pm This equation says that d is odd. Valentin Vornicu 8:22:32 pm Then what can we say about 5d - 1 = y^2 and 13d - 1 = z^2? exmath89 8:23:50 pm y and z are both even negativebplusorminus y and z are both even. superpi83 both even
8:23:50 pm 8:23:50 pm

apple413 8:23:50 pm y and z are both even thecmd999 8:23:50 y and z are even ProfessorBeast y, z are even?
pm

8:23:50 pm

sinnoel 8:23:50 pm y and z are even mathepic 8:23:50 y, z are even


pm

Valentin Vornicu 8:23:55 pm Since d is odd, 5d - 1 = y^2 and 13d - 1 = z^2 are even. Hence, y and z are even. Valentin Vornicu 8:24:01 pm Let y = 2u and z = 2v. Then 5d - 1 = y^2 = 4u^2 and 13d - 1 = 4v^2. Valentin Vornicu 8:24:04 pm What can we do with these equations? superpi83 8:25:16 subtract them apple413 8:25:16 subtract them addictedtomath subtract chchend Subtract
pm

pm

8:25:16 pm

8:25:16 pm

Fersolve 8:25:16 pm take the difference MathWhizzz 8:25:16 subtract maybe


pm

AndrewKwon97 8:25:16 pm subtract the first from the second and factor slidersknob subtract
8:25:16 pm

Valentin Vornicu 8:25:22 pm We can subtract them, to get 8d = 4v^2 - 4u^2, so 2d = v^2 - u^2. Valentin Vornicu 8:25:26 pm What can we do with this equation? exmath89 factor RHS swimmer factor
8:26:16 pm

8:26:16 pm

slidersknob 8:26:16 pm factor the right side

superpi83 8:26:16 pm v^2-u^2==2 (mod 4), contradiction ProfessorBeast 8:26:16 pm factor as difference of squares? swimmer 8:26:16 pm 4 | v^2 - u^2 - contradiction gundraja 8:26:16 pm d is even if 2d = v^2 - u^2 for integers u,v. Valentin Vornicu 8:26:21 pm We can factor the right-hand side, to get 2d = (v + u)(v - u). Valentin Vornicu 8:26:25 pm Either v + u and v - u are both even or both odd (since their difference 2u is an even number). Valentin Vornicu 8:26:30 pm If v + u and v - u are both even, then their product (v + u)(v - u) is a multiple of 4. If v + u and v - u are both odd, then their product (v + u)(v - u) is odd. In particular, (v + u)(v - u) cannot be equal to 2d, because d is odd. Valentin Vornicu 8:26:48 pm We have obtained a contradiction, so at least one of 2d - 1, 5d - 1, and 13d - 1 is not a perfect square. Valentin Vornicu 8:26:59 pm It can seem simple, but using parity can be a powerful tool. And whenever you find that a number is even (odd resp.), it is usually a good idea to write it in the form 2n (2n + 1 resp.). Valentin Vornicu
8:27:10 pm

Valentin Vornicu 8:27:22 pm Let f(n) = 19 x 8^n + 17. What approach can we take here? thecmd999 8:28:06 take mod something
pm

superpi83 8:28:06 pm find convenient mods Valentin Vornicu 8:28:10 pm We can see if there exists a number d > 1 such that f(n) = 19 x 8^n + 17 is always divisible by d, by reducing it modulo d. Valentin Vornicu 8:28:14 pm What numbers d should we try? 19oshawott98 mod 8?
8:29:04 pm

Valentin Vornicu 8:29:09 pm We should try numbers d such that 8^n modulo d simplifies. Valentin Vornicu 8:29:12 pm The first obvious number to try is d = 8. Valentin Vornicu
8:29:16 pm

Valentin Vornicu 8:29:20 pm So reducing modulo 8 doesn't get us anywhere. What other numbers d should we try? swimmer 7
8:29:23 pm

negativebplusorminus 7

8:29:23 pm

Valentin Vornicu 8:29:27 We could also try d = 7. Valentin Vornicu

pm

8:29:31 pm

Valentin Vornicu 8:30:31 pm So reducing modulo 7 doesn't get us anywhere either. Valentin Vornicu 8:30:34 What else can we try?
pm

Valentin Vornicu 8:31:07 pm (Are we really going to try all these other mods you're suggesting?) thecmd999 8:31:22 pm n=0 and n=1 give 36 and 169 which share no factors in common Valentin Vornicu 8:31:25 pm We can instead look at the first few values of f(n), and see which divisors come up. Valentin Vornicu 8:31:27 pm What is f(0) and its prime factorization? 19oshawott98 2^2*3^2
8:32:31 pm

panjia123 8:32:31 36 = 2^2 3^2 frtennis1 8:32:31 36 = 2^2 * 3^2 addictedtomath (2^2)(3^2) shandongboy 2^2 * 3^2 ProfessorBeast 36 = 2^2*3^2

pm

pm

8:32:31 pm

8:32:31 pm

8:32:31 pm

Valentin Vornicu 8:32:38 f(0) = 36 = 2^2 x 3^2.

pm

Valentin Vornicu 8:32:40 pm What is f(1) and its prime factorization? gundraja 13^2.
8:33:29 pm

frtennis1 8:33:29 169 = 13^2 addictedtomath 13^2

pm

8:33:29 pm

panjia123 8:33:29 169 = 13^2 sinnoel 13^2


8:33:29 pm

pm

ProfessorBeast f(1)=13^2 thecmd999 169=13^2

8:33:29 pm

8:33:29 pm

Valentin Vornicu 8:33:30 f(1) = 169 = 13^2.

pm

Valentin Vornicu 8:33:41 pm What is f(2) and its prime factorization? Valentin Vornicu 8:33:44 f(2) = 1233 = 3^2 x 137.
pm

Valentin Vornicu 8:34:34 pm Finally f(3) = 9745 = 5 x 1949. Valentin Vornicu 8:34:38 pm Do we see anything promising?

negativebplusorminus 8:36:29 pm Try mod 3 and see if we can knock out all the even ones. exmath89 8:36:29 pm when n is even, f(n)==0 mod 9 Fersolve 8:36:29 pm even ns are divisible by 9 * Valentin Vornicu 8:36:32 pm We see that 3 divides both f(0) and f(2), so let's work modulo 3 and see what happens. Valentin Vornicu
8:36:35 pm

Valentin Vornicu 8:36:38 pm When is 2^n + 2 divisible by 3? qfoam 8:37:50 pm when n is even; 2^n = (-1)^n superpi83 8:37:50 when n is even
pm

superbob1235 8:37:50 When n is even. 19oshawott98 when n is even

pm

8:37:50 pm

dooba 8:37:50 pm When n is even frtennis1 8:37:50 when n is even shandongboy n is even?
pm

8:37:50 pm

chchend 8:37:50 pm When n is even, -1^n+2= 0 mod 3 Valentin Vornicu 8:37:53 pm Since 2 is congruent to -1 modulo 3, 2^n + 2 is divisible by 3 if and only if n is even, which means f(n) is divisible by 3 if and only if n is even. Valentin Vornicu 8:38:06 pm Now we must look at f(n) for odd n. Let n = 2t + 1. What modulus can we work with? negativebplusorminus 5 gundraja mod 5. exmath89 mod 5
8:38:57 pm 8:38:57 pm

8:38:57 pm

Valentin Vornicu 8:39:01 pm Since 5 is a factor of f(3), let's work modulo 5. Valentin Vornicu
8:39:08 pm

Valentin Vornicu 8:39:10 pm When is 4^t + 1 divisible by 5?

19oshawott98 when t is odd lilyidi 8:40:56 when t is odd dooba t is odd

8:40:56 pm

pm

8:40:56 pm

MathWhizzz t odd JRY 8:40:56 t is odd bobbyq t is odd

8:40:56 pm

pm

8:40:56 pm

Valentin Vornicu 8:40:56 pm Since 4 is congruent to -1 modulo 5, 4^t + 1 is divisible by 5 if and only if t is odd. Valentin Vornicu 8:41:03 pm Now we must look at f(2t + 1) for t even. Let t = 2u. Then f(2t + 1) = f(4u + 1). Valentin Vornicu 8:41:05 pm What modulus can we work with? Fersolve 13? gundraja mod 13. JRY 13
8:41:49 pm

8:41:49 pm

8:41:49 pm

exmath89 mod 13 panjia123 13 chchend 13?

8:41:49 pm

8:41:49 pm

8:41:49 pm

negativebplusorminus 13

8:41:49 pm

Valentin Vornicu 8:41:56 pm There's really only one choice. Since f(1) = 13^2, let's work modulo 13. Valentin Vornicu
8:42:00 pm

Valentin Vornicu 8:42:04 pm Thus, f(4u + 1) is always divisible by 13. Valentin Vornicu 8:42:35 pm To summarize, if n is even, then f(n) is divisible by 3 (and f(n) > 3), Valentin Vornicu 8:42:37 pm if n is of the form 4k + 3, then f(n) is divisible by 5 (and f(n) > 5), Valentin Vornicu 8:42:40 pm and if n is of the form 4k + 1, then f(n) is divisible by 13 (and f(n) > 13). Valentin Vornicu 8:42:42 pm Therefore, f(n) is composite for all nonnegative integers n.

Valentin Vornicu 8:42:47 pm This problem shows that sometime you must try different moduli, even within the same problem, so don't be afraid to experiment. Valentin Vornicu
8:50:05 pm

Valentin Vornicu 8:50:15 pm You may recall Euclid's proof that there are an infinite number of primes. How can we start? frtennis1 8:51:04 pm assume that there are a finite amoung apple413 8:51:04 pm contradiction: let p equal the largest prime number in the form 3n+2 and use it to construct another prime Valentin Vornicu 8:51:09 pm We argue by contradiction. Valentin Vornicu
8:51:14 pm

Valentin Vornicu Then what?

8:51:18 pm

exmath89 8:52:38 pm consider p_1*p_2*...*p_k+1 superpi83 8:52:38 pm consider their product +-1 addictedtomath 8:52:38 pm Take the product of the p_1, ..., p_k panjia123 8:52:38 pm multiply them and add 1? MathWhizzz 8:52:38 pm consider the product minus 1 Valentin Vornicu
8:52:40 pm

Valentin Vornicu 8:52:43 pm What can we say about P? panjia123 8:53:41 pm congruent to 2^k+1 mod 3 gundraja 8:53:41 pm P = 2^k + 1 (mod 3). dooba 8:53:41 pm Well, it's not divisible by any p_i frtennis1 8:53:41 pm we can take it mod 3 Valentin Vornicu
8:53:43 pm

Valentin Vornicu 8:53:54 pm Suppose P reduces to 2 modulo 3. Then what can we say about P? apple413 8:55:06 pm it's a prime of the form 3n+2 exmath89 8:55:06 pm then P is in the form 3n+2 BOGTRO 8:55:06 pm it is prime yet not in the list of primes

Waffle15 8:55:06 pm it's in the form 3n + 2 dooba 8:55:06 pm It's another prime of the form 3n+2 panjia123 8:55:10 pm it's of the form 3n+2 and not divisible by any primes of the form 3n+2 Valentin Vornicu 8:55:13 pm All primes (other than 3) are congruent to 1 modulo 3 and 2 modulo 3. If P was only divisible by primes congruent to 1 modulo 3, then P itself would be congruent to 1 modulo 3. But P is congruent to 2 modulo 3, so P must be divisible by at least one prime that is congruent to 2 modulo 3. Valentin Vornicu 8:55:35 pm But by construction, P is relatively prime to all primes that are congruent to 2 modulo 3, contradiction. Valentin Vornicu 8:56:02 pm The only case that is left is if P reduces to 0 modulo 3. Unfortunately, there's no good way to handle this case. (We can divide out all the factors of 3, but there's nothing we can say about the resulting quotient.) Valentin Vornicu 8:56:06 pm Is there any way to fix this argument? AndrewKwon97 8:57:43 pm if k is odd, can't we just multiply the product of a_i by 2, then get 2p_1p_2...p_k+1, which is now 2(mod3) Valentin Vornicu
8:57:46 pm

Valentin Vornicu 8:57:58 pm Then P always reduces to 2 modulo 3, and it is relatively prime to all the p_i, so the argument that we used above now works. Valentin Vornicu 8:58:13 pm (There's other options like AndrewKwon97 described) Valentin Vornicu 8:58:20 pm More generally, if a and b are relatively prime positive integers, then there are infinitely many primes of the form an + b. This result is known as Dirichlet Theorem, but it is an advanced result and is difficult to prove. Valentin Vornicu 8:58:27 pm The ideas in Euclid's proof lead to the next few problems. Valentin Vornicu
8:58:49 pm

Valentin Vornicu 8:59:00 pm To get a feel for the problem, let's find the first few terms of the sequence.a Valentin Vornicu What is p_2? chchend 3 superpi83 3
8:59:06 pm

9:00:15 pm

9:00:15 pm

EricMathPath09 3 patrickhompe 3 swimmer 3 yugrey 3 Fersolve 3

9:00:15 pm

9:00:15 pm

9:00:15 pm

9:00:15 pm

9:00:15 pm

Valentin Vornicu 9:00:18 pm p_2 is the largest prime divisor of p_1 + 1 = 3, so p_2 = 3. Valentin Vornicu What is p_3?
9:00:20 pm

negativebplusorminus 7 gundraja 7
9:00:32 pm

9:00:32 pm

EricMathPath09 7 screenhunter372 7 frtennis1 7

9:00:32 pm

9:00:32 pm

9:00:32 pm

superbob1235 7 swimmer 7

9:00:32 pm

9:00:32 pm

Valentin Vornicu 9:00:36 pm p_3 is the largest prime divisor of p_1 p_2 + 1 = 7, so p_3 = 7. Valentin Vornicu What is p_4? EricMathPath09 43 gundraja 43
9:00:38 pm

9:01:42 pm

9:01:42 pm

MathWhizzz 43 Wolfkit 43

9:01:42 pm

9:01:42 pm

exmath89 43 swimmer 43 superpi83 43

9:01:42 pm

9:01:42 pm

9:01:42 pm

Valentin Vornicu 9:01:43 pm p_4 is the largest prime divisor of p_1 p_2 p_3 + 1 = 43, so p_4 = 43. Valentin Vornicu What is p_5? EricMathPath09 139 frtennis1 139
9:01:53 pm

9:02:21 pm

9:02:21 pm

Valentin Vornicu 9:02:25 pm p_5 is the largest prime divisor of p_1 p_2 p_3 p_4 + 1 = 1807 = 13 x 139, so p_4 = 139. Valentin Vornicu 9:02:28 pm We must show that 5 never appears in the sequence. Valentin Vornicu 9:02:30 How can we proceed? screenhunter372 contradiction
pm

9:03:19 pm

Valentin Vornicu 9:03:22 pm We can argue by contradiction. Valentin Vornicu 9:03:37 pm (No the sequence is not increasing although it might appear to be so right now) Valentin Vornicu
9:03:44 pm

Valentin Vornicu 9:03:50 pm What does that say about this expression? EricMathPath09
9:05:01 pm

qfoam 9:05:01 2^a*3^b*5^c

pm

Wolfkit 9:05:01 2^n*3^m*5^o

pm

Valentin Vornicu

9:05:10 pm

Valentin Vornicu 9:05:12 Can we say more?

pm

superpi83 9:05:28 pm the expression is a power of 5 sinnoel 9:05:28 pm it is a power of 5 yugrey 9:05:28 pm It's a power of 5. BOGTRO 9:05:28 pm 5 must be the only prime divisor BOGTRO 9:05:28 pm since p_1=2 and p_2=3 negativebplusorminus 9:05:28 But the expression is odd. Valentin Vornicu
9:05:35 pm pm

Valentin Vornicu

9:05:39 pm

Valentin Vornicu

9:05:42 pm

Valentin Vornicu

9:05:45 pm

Valentin Vornicu 9:06:54 pm What can we do with this equation? superpi83


9:07:10 pm

then product=5^k-1==0 mod 4 contradiction! negativebplusorminus 9:07:10 pm But that has a factor of 4 in it. exmath89 9:07:10 factor 5^k-1
pm

BOGTRO 9:07:10 pm 5^k-1 is a multiple of 4, which is impossible apple413 9:07:10 take mod 4 superbob1235 4 | 5^k - 1
pm

9:07:10 pm

swimmer 9:07:10 pm then 4 | p1...pn-1, contradiction screenhunter372 9:07:10 which is a contradiction Valentin Vornicu
pm

9:07:16 pm

Valentin Vornicu 9:07:18 pm This equation tells us that 4 must be a factor of the left-hand side. Valentin Vornicu 9:07:22 pm But this is impossible because p_1 = 2 and all further terms p_2, ..., p_{n - 1} are odd. Valentin Vornicu 9:07:27 pm We have a contradiction, so 5 cannot be a member of this sequence. Valentin Vornicu
9:08:02 pm

Valentin Vornicu 9:08:25 pm I actually proposed this problem for the IMO TSTs back then. Only the 6 guys who actually made the team got 7s on the problem. Valentin Vornicu 9:08:38 pm Let's start by looking at small primes. Is the prime 2 in M? yugrey 9:09:32 pm Yes 2 is certainly in M. Valentin Vornicu 9:09:41 You need a reason!
pm

exmath89 9:10:30 pm prod(p)-1 is even by taking odd primes Valentin Vornicu 9:10:35 pm Since M contains at least three primes, there is an odd prime in M, say p. Let A = {p}. Then by the condition in the problem, all the prime factors of p - 1 are also in M. Valentin Vornicu 9:10:41 pm But p - 1 is even, so 2 is in M. Valentin Vornicu 9:10:43 Is the prime 3 in M?
pm

superpi83 9:12:19 pm yes, take one prime that is 1 mod 3 or 2 primes that are 2 mod 3 yugrey 9:12:19 pm By pigeonhole two of tehse WLOg p_1 and p_2 are congruent mod 3

AndrewKwon97 9:12:19 pm if there is another prime 1(mod3) then we immediately get 3 is in M, if there is another prime 2(mod3) then 2 times that prime -1 is divisible by 3, so yes Valentin Vornicu 9:12:23 pm If M contains a prime of the form p = 3n + 1, then take A = {p}. Then all the factors of p - 1 = 3n are also in M, so 3 is in M. Valentin Vornicu 9:12:38 pm Otherwise, M contains two primes of the form 3n + 2, say p and q. In this case we can take A = {p, q}. Then all the prime factors of pq - 1 are also in M. Valentin Vornicu 9:13:16 pm But pq - 1 is congruent to 2 x 2 - 1 = 3, or 0 modulo 3, so pq - 1 is divisible by 3. Therefore, 3 is in M. In either case, 3 is in M. Valentin Vornicu 9:13:19 Is the prime 5 in M? bobbyq 9:13:54 pm yes, since 2 and 3 are in M superbob1235 Yes. A = {2,3} Wolfkit 9:13:54 Yes, 2*3-1=5
9:13:54 pm pm

pm

JRY 9:13:55 pm yes, let A={2,3}. M must contain 2x3-1=5 Valentin Vornicu 9:14:02 pm We know 2 and 3 are in M, so take A = {2, 3}. Then 2 x 3 - 1 = 5, so 5 is in M. Valentin Vornicu 9:14:05 Is the prime 7 in M? negativebplusorminus {3,5} so yes exmath89 9:15:05 yes, take {3,5} superbob1235 Yes. A = {3,5} EricMathPath09 Yes: A = {3,5}
pm pm

9:15:05 pm

9:15:05 pm

9:15:05 pm

Wolfkit 9:15:05 pm Yes, 3*5-1=14=2*7 Valentin Vornicu 9:15:07 pm We know 3 and 5 are in M, so take A = {3, 5}. Then 3 x 5 - 1 = 14 = 2 x 7, so 7 is in M. Valentin Vornicu 9:15:10 pm Hence, the first few primes are in M. We want to prove that M contains all primes. Valentin Vornicu 9:15:13 pm What intermediate result can we try proving first? yugrey 9:16:12 pm M has infinitely many primes. frtennis1 9:16:12 pm there are an infinite amount of primes in M AndrewKwon97 it is not finite shandongboy Infinite
9:16:12 pm

9:16:12 pm

Valentin Vornicu 9:16:14 pm Let's try proving first that M is infinite. Valentin Vornicu
9:16:18 pm

Valentin Vornicu

9:16:55 pm

As a start, let's take A to be all the primes in M, except one, say p_i. (Remember that A must be a proper subset of M.) Valentin Vornicu
9:16:59 pm

Valentin Vornicu

9:17:05 pm

yugrey 9:18:02 pm Then P/p_1-1 is relatively prime to all the primes in P except p_1 so it is a pwoer of p_1 negativebplusorminus 9:18:02 IT must be a power of p_i. superpi83 9:18:02 pm it must be a power of p_i gundraja 9:18:02 pm It must be divisible by p_i. frtennis1 9:18:02 pm it is a power of p_i Valentin Vornicu
9:18:11 pm pm

Valentin Vornicu

9:18:55 pm

Valentin Vornicu

9:19:13 pm

Valentin Vornicu 9:19:22 pm To make these equations simpler to work with, we start by omitting the smallest primes. Valentin Vornicu
9:19:25 pm

Valentin Vornicu

9:20:10 pm

superpi83 9:21:44 also P=3^(b+1)+3

pm

negativebplusorminus Similarly P=3^(b+1)+3 Valentin Vornicu

9:21:44 pm

9:22:11 pm

superpi83 9:23:37 pm no solution because of some lemma that says the only consecutive prime powers are 8 and 9 EricMathPath09 9:23:37 pm It is well-known that there are no solutions, but I have no idea how to prove that. yugrey 9:23:37 pm Mod 3 and letting a+1=2c works, but MIHAILESCU is way cooler. Valentin Vornicu 9:24:04 pm You actually have to prove it, just stating "Mihailescu" doesn't give you the points, unless you know how to prove that theorem. Valentin Vornicu 9:24:45 pm So how do we analyze this? AndrewKwon97 mod 3
9:25:39 pm

exmath89 9:25:39 pm take mod 3 and mod 4 negativebplusorminus 9:25:39 Mod 3; then a+1 is even. Valentin Vornicu 9:25:49 pm We can try reducing mod 3. Valentin Vornicu
9:26:22 pm pm

Valentin Vornicu

9:27:19 pm

negativebplusorminus Now mod 4. superpi83 mod 4


9:27:33 pm

9:27:33 pm

Valentin Vornicu 9:28:06 pm I'm not sure that's going to get us anywhere. It will tell us something about the parity of b but it's not useful. AndrewKwon97 9:29:06 pm or factor difference of squares yugrey 9:29:06 pm Or just subtract the 1 from both sides... superbob1235 9:29:06 pm (2^(k+1) - 1) (2^(k+1) + 1) = 3^(b+1) dinoboy 9:29:06 pm Move the 1 onto the other side and do difference of squares. Fersolve 9:29:06 pm take diff of squares 2^k -1 is divis by power of 3? Valentin Vornicu
9:29:45 pm

Fersolve k=0

9:30:43 pm

AndrewKwon97 k=0 archer4 k=0 gundraja k = 0. exmath89 k=0

9:30:43 pm

9:30:43 pm

9:30:43 pm

9:30:43 pm

Valentin Vornicu

9:31:50 pm

Valentin Vornicu 9:31:58 pm Therefore, M contains an infinite number of primes. Valentin Vornicu 9:32:03 pm We're actually very close to the end now. Valentin Vornicu
9:32:08 pm

Valentin Vornicu 9:33:09 pm What result could we try using here? superpi83 9:34:20 fermat's little thm
pm

dinoboy 9:34:20 pm fermat's little theorem superpi83 9:34:20 pm at least one residue class mod p must be infinite by pigeonhole. take p-1 of those primes and apply FLT Valentin Vornicu 9:34:24 pm We could try using Fermat's Little Theorem. Valentin Vornicu
9:34:32 pm

Valentin Vornicu 9:34:36 pm The primes p_1, p_2, ..., p_{p - 1} must also be distinct from p. But if p was in M, then this argument would not be necessary in the first place. Valentin Vornicu 9:34:40 pm How can we show that such primes exist? dinoboy 9:35:46 pm I mean pidgeonhole principle Fersolve 9:35:46 pm pigeon hole on infinite number of primes negativebplusorminus pigeonhole!
9:35:46 pm

exmath89 9:35:46 pm take pk+1 since there are infinitely many of them Valentin Vornicu 9:35:49 pm We have shown that the set M is infinite.p Valentin Vornicu 9:35:53 pm Hence, if we reduce every element in M modulo p, then some nonzero residue must appear an infinite number of times. Valentin Vornicu 9:36:16 pm Then we simply take p - 1 primes p_1, p_2, ..., p_{p - 1} that have this residue. Valentin Vornicu 9:36:18 pm Therefore, M contains every prime p. Valentin Vornicu SUMMARY
9:36:24 pm

Valentin Vornicu 9:36:28 pm In today's class, we saw how to effectively use Fermat's Little Theorem and Euclid's Theorem. We also saw the power of algebra in number theory problems. We can solve many number theory problems effectively by putting a number in the right form (like expressing an even number in the form 2n), and using factorization. When you see a number theory problem, convert the words into mathematics, and let the equations do the work. Valentin Vornicu 9:36:31 pm Finally, we also saw how to solve problems using argument by contradiction. If a problem asks you to prove that something cannot occur, this is usually a good sign to use contradiction. Valentin Vornicu 9:36:36 pm That's it for today's class. Copyright AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you cannot share the contents of this file with others. Copyright 2013 AoPS Incorporated. All Rights Reserved. Contact Us Privacy Terms & Conditions

Potrebbero piacerti anche