367

Find a positive integer solution $(x,y,z,a,b)$ for which

$$\frac{1}{x}+ \frac{1}{y} + \frac{1}{z} + \frac{1}{a} + \frac{1}{b} = 1\;.$$

Is your answer the only solution? If so, show why.

I was surprised that a teacher would assign this kind of problem to a 5th grade child. (I'm a college student tutor) This girl goes to a private school in a wealthy neighborhood.

Please avoid the trivial $x=y=z=a=b=5$. Try looking for a solution where $ x \neq y \neq z \neq a \neq b$ or if not, look for one where one variable equals to another, but explain your reasoning. The girl was covering "unit fractions" in her class.

Martin Sleziak
  • 50,316
  • 18
  • 169
  • 342
Low Scores
  • 4,435
  • 4
  • 21
  • 25
  • 3
    What about $a = b = 3$, $x = 6$, $y = 9$ and $z = 18$? Or simpler still $x=y=z=a=8$, $b=2$? – Michael Albanese Jan 30 '13 at 09:41
  • 15
    A $5$-tuple is hardly a ‘pair’! – Brian M. Scott Jan 30 '13 at 09:42
  • 26
    $(2, 4, 8, 16, 16)$ is another solution. – Karolis Juodelė Jan 30 '13 at 09:45
  • 99
    $x\neq y\neq z\neq a\neq b$ is not the same as saying that $x, y, z, a,$ and $b$ are all distinct. For example $1\neq 2\neq 1\neq 2\neq 1$, but they are not distinct. – Michael Albanese Jan 30 '13 at 09:52
  • 3
    http://www.jstor.org/stable/2299023 –  Jan 30 '13 at 13:35
  • 64
    I wouldn't be surprised if a bright fifth grader came up with the trivial x=y=z=a=b=5. Then answer the next question with "No": x=y=z=a=6, b=3. That seems do-able for bright grade fives. – Michael J Swart Jan 30 '13 at 13:47
  • 3
    @Michael, I agree. Low: Why the surprise? We easily found two different solutions, and that answers the question. – GEdgar Jan 30 '13 at 14:13
  • 3
    When I was in 5th grade, I was still learning long division. Crazy that they start algebra so young now. – Nick Hartung Jan 30 '13 at 14:24
  • 3
    If they find any solution other than all=5, then they can just swap which number is assigned to which variable - that works for the "is it the only solution?" part. Sorta. – Izkata Jan 30 '13 at 14:58
  • 1
    More context is needed to judge appropriateness of the difficulty. eg Regular vs honors math, and if it's something that all students are expected to be able to do or a stretch problem. It appears similar in difficulty to weekly challenge/puzzle solving sets that I was regularly given in my honors math class around that grade level; but the typical expectation was that we'd only be able to figure out how to solve 2 or 3 out of 5 of them (and the 'easier' half were generally solvable both via brute force using just pencil and paper and by a more elegant method if you could figure it out). – Dan Is Fiddling By Firelight Jan 30 '13 at 15:40
  • 46
    Is there a reason why 1/5 + 1/5 + 1/5 + 1/5 + 1/5 = 1 is a bad answer? Especially if this is grade school math. For a generic fifth grader is seems that this would be a rather logical introduction to fractions. I know you're looking for 'x≠y≠z≠a≠b' but is there some outside reason as to why? The question, as stated, doesn't seem to have that restriction. Occam's Razor and all that. – grauwulf Jan 30 '13 at 20:13
  • Indeed, is the constraint that the solution consist of distinct integers part of the original fifth grade problem, or is that an embellishment added by you? – Kaz Jan 30 '13 at 21:06
  • 1
    related: http://math.stackexchange.com/questions/109655/how-to-prove-that-sum-n-12012-frac1x-n-1-has-finitely-many-solut – sdcvvc Jan 30 '13 at 21:39
  • @grauwulf I think the point was that s/he assumed a 5th grader 'could' figure that out and was focusing on finding out whether a 5th grader really should be able to figure out another answer. – Brandon Moore Jan 31 '13 at 07:46
  • 3
    From the question it seems clear that the distinct requirement was not given to the 5th graders. Fractions and equations with multiple unknowns is part of the 4-6th grade plan so the problem seems appropriate (though maybe not required for a passing grade). I would expect somewhere between 70-90% to figure out the trivial solution and maybe 20-30% to figure out another solution, thus proving it's not the only one. Less than 1% of 5th graders would probably find a distinct solution. – MatsT Jan 31 '13 at 10:53
  • Interestingly enough, the number of distinct solutions is an upper bound for the number of (isomorphism classes of) groups with exactly 5 conjugacy classes! – yatima2975 Jan 31 '13 at 16:21
  • The actual content of the class would be relevant to the ease of finding examples -- e.g. if the class had recently talked about Egyptian fractions, they might have been taught some ways to find equations like this. –  Feb 01 '13 at 03:54
  • 8
    How old is a 5th grader in your country? – mplungjan Feb 01 '13 at 12:21
  • 5
    Don't have enough rep here (I come from other stackexchange sites) to answer but I came up with another brute force method I didn't see here. Pick the largest fraction with a non-repeating decimal equivalent that's less than your total (1/2) and keep going (1/2 + 1/4 = .75; 1/2 + 1/4 + 1/5 = .95; 1/2 + 1/4 + 1/5 + 1/25 = .99; 1/2 + 1/4 + 1/5 + 1/25 + 1/100 = 1) – Derek Feb 01 '13 at 21:08
  • 1
    A generalized version of the question has received some good answers at http://mathoverflow.net/questions/229514/how-many-solutions-does-frac1x-1-frac1x-2-cdots-frac1x-n-1-ha – Gerry Myerson Jan 28 '16 at 21:45

20 Answers20

415

The perfect number $28=1+2+4+7+14$ provides a solution:

$$\frac1{28}+\frac1{14}+\frac17+\frac14+\frac12=\frac{1+2+4+7+14}{28}=1\;.$$

If they’ve been doing unit (or ‘Egyptian’) fractions, I’d expect some to see that since $\frac16+\frac13=\frac12$, $$\frac16+\frac16+\frac16+\frac16+\frac13=1$$ is a solution, though not a much more interesting one than the trivial solution. The choice of letters might well suggest the solution

$$\frac16+\frac16+\frac16+\frac14+\frac14\;.$$

A little playing around would show that $\frac14+\frac15=\frac9{20}$, which differs from $\frac12$ by just $\frac1{20}$; that yields the solution

$$\frac1{20}+\frac15+\frac14+\frac14+\frac14\;.$$

If I were the teacher, I’d hope that some kids would realize that since the average of the fractions is $\frac15$, in any non-trivial solution at least one denominator must be less than $5$, and at least one must be greater than $5$. Say that $x\le y\le z\le a\le b$. Clearly $x\ge 2$, so let’s try $x=2$. Then we need to solve

$$\frac1y+\frac1z+\frac1a+\frac1b=\frac12\;.$$

Now $y\ge 3$. Suppose that $y=3$; then $$\frac1z+\frac1a+\frac1b=\frac16\;.$$

Now $1,2$, and $3$ all divide $36$, and $\frac16=\frac6{36}$, so we can write

$$\frac1{36}+\frac1{18}+\frac1{12}=\frac{1+2+3}{36}=\frac6{36}=\frac16\;,$$

and we get another ‘nice’ solution,

$$\frac12+\frac13+\frac1{12}+\frac1{18}+\frac1{36}\;.$$

Aarthi
  • 101
  • 9
Brian M. Scott
  • 588,383
  • 52
  • 703
  • 1,170
  • 26
    Oh my gosh Brian. How did you figure that out? It took 5 adults in the room and 50 minutes yet no one got as close as Michael or Karolis! There was a woman in the room who had her B.A. in math from a prestigious university and she couldn't figure this out. Amazing. Is it possible for you to show your reasoning or is this something that "popped" into your head ...like magic? – Low Scores Jan 30 '13 at 09:49
  • 52
    @LowScores: I realized immediately that a perfect number would work **if** it had exactly $5$ proper divisors. $6$ is too small, and I knew that the next perfect number is $28$, so I tried it, and it worked. – Brian M. Scott Jan 30 '13 at 09:53
  • 2
    Think he meant 1/3 for the last term based on the line above. – Anthony Jan 30 '13 at 13:48
  • Fixed it because 7/6 isn't 1. – Aarthi Jan 30 '13 at 15:57
  • 5
    Your answer is the most fitting for lines of reasoning to this problem as a fifth-grader might see it. It's approximately the same approach I used the problem when I first saw it. – Joe Z. Jan 30 '13 at 16:57
  • 32
    I love that we have real professors on here. – temporary_user_name Jan 30 '13 at 18:17
  • 3
    Your solution to try x = 2 ... etc is good. Solving by iteratively eliminating variables is a good technique for 5th graders to learn. – Josh Petitt Jan 31 '13 at 03:46
  • Do the sum of the reciprocals of the factors of all perfect numbers =1? – AAron Jul 06 '15 at 04:18
  • @AAron: The sum of the reciprocals of the factors that are greater than $1$ is always $1$, yes; this follows from the definition of *perfect number*. If you put those fractions over the common denominator $n$, the numerators will be all of the divisors of $n$ except $n$ itself, and a number is perfect precisely when the sum of those divisors is $n$. – Brian M. Scott Jul 06 '15 at 04:22
  • Oh ok, but I am guessing that there are other numbers that fit the bill for this case, right? Or do these numbers have any other significance? – AAron Jul 06 '15 at 04:27
  • @AAron: According to [this answer](http://math.stackexchange.com/a/290575/12042) by **Joe Z.** there are $147$ solutions altogether; the one using the perfect number $28$ was just the first one that occurred to me that had all of the denominators distinct. – Brian M. Scott Jul 06 '15 at 04:33
  • Are there any recommended sites other than Wikipedia where I could study Egyptian fractions? – AAron Jul 06 '15 at 04:37
  • 1
    @AAron: I’ve taken only a quick look, but it appears that you could do a lot worse than to start [here](http://www.maths.surrey.ac.uk/hosted-sites/R.Knott/Fractions/egyptian.html) and [here](http://www.ics.uci.edu/~eppstein/numth/egypt/). – Brian M. Scott Jul 06 '15 at 04:43
  • Thank you very much. Ill have a look at them. Also, I have a problem related to this coming out tomorrow midday so do you mind taking a look at it for me? – AAron Jul 06 '15 at 04:45
  • @AAron: You’re welcome. Once you’ve posted it, leave a note under any of my answers, including a link to the question, and I’ll take a look. – Brian M. Scott Jul 06 '15 at 04:46
304

You could connect it to geometry. Cut the square into equally-sized pieces, then take some of those pieces and cut them into even-smaller ones until you get five pieces. e.g.

enter image description here

enter image description here

After that, you could try some different-sized pieces and have a good chance of getting something like this:

enter image description here

Mark Eichenlaub
  • 6,385
  • 3
  • 30
  • 40
  • 38
    I think this solution is the most impressive, because it seems like the only one that instantly and accessibly conveys to almost everybody that there is an infinite amount of solutions. – gmlime Jan 31 '13 at 09:28
  • 83
    There aren't infinitely many solutions, though. There are 147, as shown in Joe Zeng's answer – Max Jan 31 '13 at 10:41
  • 5
    @gmline This wasn't supposed to be a method for enumerating all the answers or anything like that; just a way to approach it visually that might help for some kids. – Mark Eichenlaub Jan 31 '13 at 17:57
  • 1
    An excellent idea. Cutting out fractions of a unit circle and passing them to the class to put back together in different ways would work really well! – Joshua Shane Liberman Jan 31 '13 at 20:31
  • This is genius. If I were the teacher, I'd hope for this kind of thinking! – awendt Feb 01 '13 at 10:16
  • This is awesome. I thought of so many numerous attempts and thoughts, but this one is the high-five kicker!!! – Stefan Feb 01 '13 at 13:05
  • I feel this answer is the purpose/intention of asking such question to 5th grade students. – Vikram Feb 01 '13 at 16:16
  • I agree. However, I feel 'cakes' are more appropriate a representation for a 5th grader than squares. Most fractions are taught using 'pieces of cakes'. It really IS digestable by halflings :) – PhD Feb 02 '13 at 02:38
  • The simple geometric visual is what makes this solution genius. – But I'm Not A Wrapper Class Nov 12 '13 at 02:09
195

This solution may be too advanced for a fifth-grader, but you can do this problem algorithmically - simply by searching all the possible fractions.

The gist of it is to use a greedy algorithm - start with the biggest fraction, and continue iterating small fractions until you can't anymore. For example, $\displaystyle \frac 1 2 + \frac 1 3 + \frac 1 7 + \frac 1 {43} + \frac 1 {1806}$ would be the first one you find. The next one would be $\displaystyle \frac 1 2 + \frac 1 3 + \frac 1 7 + \frac 1 {44} + \frac 1 {924}$, then $\displaystyle \frac 1 2 + \frac 1 3 + \frac 1 7 + \frac 1 {45} + \frac 1 {630}$, and so on.

Here's the algorithm in more detail:

  • For the first number, start with $\displaystyle \frac 12$, eventually working your way down to $\displaystyle \frac15$ (which is the smallest that the largest fraction can be, so you can stop there).
  • Subtract this fraction from 1, and use the remaining part to determine what the next few numbers will iterate through.
  • For each of the subsequent fractions, start with the largest unit fraction smaller than both the "remaining part" that's left and the fraction before it, and work down until you reach the smallest unit fraction larger than $\displaystyle \frac1n$ the "remaining part", where your fraction is the $n$th last fraction, and do the same as above, subtracting the unit fraction from the "remaining part" for the next fraction to use.
  • Once you have four fractions, if the "remaining part" for the last fraction can be expressed as a unit fraction, you have a solution. Otherwise, you don't, and continue onwards.

This algorithm will eventually return all the possible unit fraction combinations, starting from $\displaystyle \frac 1 2 + \frac 1 3 + \frac 1 7 + \frac 1 {43} + \frac 1 {1806}$ and ending with $\displaystyle \frac 15 + \frac 15 + \frac 15 + \frac 15 + \frac 15$.


http://joezeng.com/code/fractions/fractions.html $\leftarrow$ This is a list of all of the fractions, dynamically generated using some recursive Javascript that implements the algorithm above. You can view the source code here, which I have MIT-licensed for demonstration purposes.

According to the generator, there are a total of 147 solutions for 5 fractions, and the "minimum unique solution" such that all denominators are distinct and their sum is the lowest possible is $\displaystyle \frac 13 + \frac 14 + \frac 15 + \frac 16 + \frac 1 {20}$. There is another "minimum unique solution", such that the largest denominator is the lowest, which is $\displaystyle \frac 12 + \frac 14 + \frac 1{10} + \frac 1{12} + \frac 1{15}$.

You can also use the generator to generate fraction lists of arbitrary sizes by modifying the initial function call to use 6 levels (or 2, 3, or 4) instead of 5, as well as generate Egyptian fraction expansions for arbitrary fractions (by modifying the first two terms 1, 1 to be other things).

Joe Z.
  • 6,539
  • 3
  • 23
  • 42
  • For 6 terms, the minimum unique solution is 1/3 + 1/4 + 1/6 + 1/10 + 1/12 + 1/15, and my computer lagged out on 7 terms. – Joe Z. Jan 30 '13 at 17:23
  • the Question is one for a specific set of instructions, @Brian showed how simple the question really was. if you look at the fractions Graphically it would seem so simple. you have a pie cut it into 5 pieces that aren't the same. they didn't want the 5th graders to create algorithms they wanted them to "see" the fractions. – Malachi Jan 30 '13 at 17:48
  • 32
    Brian had already presented the best solution; I was simply presenting a more comprehensive, less appropriate one. – Joe Z. Jan 30 '13 at 17:51
  • 2
    Also, this solution would be to the benefit of that one girl with a B.A. in math who couldn't solve it, as mentioned by @LowScores in one comment on Brian's answer. – Joe Z. Jan 30 '13 at 17:55
  • 6
    There are 3462 solutions for six terms, 2320 with all unique terms (72 solutions for five unique terms). There are 294326 solutions with seven terms (245767 unique); the least solution is 3 4 9 10 12 15 18. Beyond seven I'd want a cleverer search strategy (breadth-first for small values would be fine). – Rex Kerr Jan 30 '13 at 19:36
  • I think beyond seven the actual *number* of solutions gets unwieldy. If we're doing a search for the least unique solution, you're right that we could probably employ some sort of BFS strategy. – Joe Z. Jan 30 '13 at 20:10
  • Alternatively, we could try doing the iteration in reverse. One way we can enforce a unique solution is to disallow fractions exactly equal to $\frac 1n$ the remaining part. – Joe Z. Jan 30 '13 at 21:36
  • 17
    "Brian had already presented the best solution; I was simply presenting a more comprehensive, less appropriate one." +1 for this charming analysis of your work! – Mark Harrison Feb 01 '13 at 00:06
  • 1
    @JoeZeng We'd probably need assistance from SO staff to get an answer; but I'd be curious if your answer has a disproportionately high share of up votes from people whose network profiles indicate they're primarily programmers not mathematicians. – Dan Is Fiddling By Firelight Feb 01 '13 at 16:44
  • 1
    It's certainly flooding my reputation count. I've hit the daily limit three days in a row now, and this is the first time I've ever hit the daily limit even once. – Joe Z. Feb 01 '13 at 16:56
  • 7
    I’m very glad that someone took the trouble to do the exhaustive analysis! – Brian M. Scott Feb 01 '13 at 21:38
  • @RexKerr I strongly suspect the minimum unique solution for eight terms is 3 5 9 10 12 15 18 20. – Joe Z. Aug 29 '13 at 04:02
  • @JoeZ. - I don't believe I can check right now, but I wouldn't be surprised given the form of the others. – Rex Kerr Aug 29 '13 at 10:42
  • The link does not work anymore. – Keba Apr 03 '15 at 16:28
  • Yeah, I restructured my site a bit and the code disappeared. I'll put it up on pastebin or something. – Joe Z. Apr 03 '15 at 17:43
120

Yet another method would be to start with $${1\over2}+{1\over3}+{1\over 6}=1$$ Divide by two and add $1\over 2$; this yields $${1\over 2}+{1\over 4}+{1\over 6}+{1\over 12}=1$$ Again, divide by two and add $1\over 2$; this yields $${1\over 2}+{1\over 4}+{1\over8}+{1\over 12}+{1\over 24}=1$$

Per Manne
  • 10,651
  • 2
  • 34
  • 46
43

The number of solutions of $$1={1\over x_1}+{1\over x_2}+\cdots+{1\over x_n},\ \ \ 0\lt x_1\le x_2\le\cdots\le x_n$$ is tabulated, as a function of $n$, at OEIS A002966 but only a few terms are given: $1, 1, 3, 14, 147, 3462, 294314, 159330691$. I don't know whether the number of solutions with all denominators distinct has been tabulated at that site.

Jaideep Khare
  • 19,097
  • 3
  • 38
  • 78
Gerry Myerson
  • 168,500
  • 12
  • 196
  • 359
30

A simple C++ code to find the a-b-c-x-y pairs for the first positive 100 integers.

const double MAX_NUM = 100.0;
const double EPSILON = 0.0000001;
double Sum;
for (double a=1; a<=MAX_NUM; a++)
{
    for (double b=1; b<=a; b++)
    {
        for (double c=1; c<=b; c++)
        {
            for (double x=1; x<=c; x++)
            {
                for (double y=1; y<=x; y++)
                {
                    Sum = 1.0/a + 1.0/b + 1.0/c + 1.0/x + 1.0/y;
                    if (abs(Sum - 1.0) < EPSILON)
                    {
                        std::cout << a << "\t" << b << "\t" << c
                            << "\t" << x << "\t" << y << std::endl;
                    }
                }
            }
        }
    }
}

Its output is:

5       5       5       5       5
6       6       6       4       4
6       6       6       6       3
8       8       4       4       4
8       8       6       4       3
8       8       8       8       2
9       9       9       3       3
9       9       9       6       2
10      5       5       4       4
10      6       5       5       3
10      10      10      5       2
12      6       4       4       4
12      6       6       4       3
12      8       8       3       3
12      8       8       6       2
12      12      4       4       3
12      12      6       3       3
12      12      6       6       2
12      12      12      4       2
14      7       7       7       2
15      5       5       5       3
15      10      4       4       3
15      10      6       3       3
15      10      6       6       2
15      12      10      4       2
15      15      5       3       3
15      15      6       5       2
16      16      8       4       2
18      9       4       4       3
18      9       6       3       3
18      9       6       6       2
18      12      9       4       2
18      18      18      3       2
20      5       4       4       4
20      6       5       4       3
20      8       8       5       2
20      10      10      4       2
20      12      5       3       3
20      12      6       5       2
20      20      5       5       2
20      20      15      3       2
21      7       7       3       3
21      7       7       6       2
21      21      14      3       2
24      8       4       4       3
24      8       6       3       3
24      8       6       6       2
24      12      8       4       2
24      16      16      3       2
24      24      4       3       3
24      24      6       4       2
24      24      12      3       2
28      14      7       4       2
28      21      4       3       3
28      21      6       4       2
28      21      12      3       2
30      10      5       3       3
30      10      6       5       2
30      15      5       5       2
30      15      15      3       2
30      20      4       3       3
30      20      6       4       2
30      20      12      3       2
30      30      10      3       2
33      22      11      3       2
35      14      5       5       2
35      15      14      3       2
36      9       9       4       2
36      18      4       3       3
36      18      6       4       2
36      18      12      3       2
36      36      9       3       2
40      10      8       4       2
40      24      10      3       2
40      40      5       4       2
42      7       4       4       3
42      7       6       3       3
42      7       6       6       2
42      12      7       4       2
42      14      14      3       2
45      9       5       3       3
45      9       6       5       2
45      30      9       3       2
45      36      5       4       2
48      16      4       3       3
48      16      6       4       2
48      16      12      3       2
48      48      8       3       2
54      27      9       3       2
56      42      8       3       2
60      5       5       4       3
60      12      5       5       2
60      15      4       3       3
60      15      6       4       2
60      15      12      3       2
60      20      10      3       2
60      30      5       4       2
60      40      8       3       2
70      7       7       5       2
70      28      5       4       2
72      9       8       4       2
72      24      9       3       2
72      36      8       3       2
78      13      13      3       2
84      14      4       3       3
84      14      6       4       2
84      14      12      3       2
84      84      7       3       2
88      33      8       3       2
90      18      10      3       2
91      78      7       3       2
96      32      8       3       2
99      22      9       3       2
100     25      5       4       2
hkBattousai
  • 4,115
  • 7
  • 36
  • 53
23

My first reaction to the question was $\frac12+\frac14+\frac18+\frac1{16}+\frac1{16}$. sorry, but it seems too natural to need explaining, but also not satisfactory because not all denominators are different.

Then other replies remind me of the "5th grade level" equation $1 = \frac12 + \frac13 + \frac16$.

Well, then the pupil will probably try to use the same equation to divide one of the three fractions and realize that the only one that will provide different denominators is the $\frac16$.

That is $1/6 = \frac{1/6}2 + \frac{1/6}3 + \frac{1/6}6 = \frac1{12} + \frac1{18} + \frac1{36}$

Thus building the solution $1 = \frac12 + \frac13 + \frac1{12} + \frac1{18} + \frac1{36}$


Thinking further about it, I assume that anybody confronted with fractions will easily agree with the simple statement: $1 = \frac1n + \frac{n-1}n$.

Dividing the equation by $n-1$ and arranging differently will produce $\frac1{n-1} = \frac1n + \frac1{n(n-1)}$, of which $\frac12 = \frac13 + \frac16$ is a special case.

once you have a couple of these equations spelled out, like: $$\frac12 = \frac13 + \frac16 $$ $$\frac13 = \frac14 + \frac1{12} $$ $$\frac14 = \frac15 + \frac1{20} $$ $$\frac15 = \frac16 + \frac1{30} $$

it should be a kid's game to expand $$1 = \frac12 + \frac14 + \frac14$$ into a solution of the given problem.

you might also want to include the case for $n=2$: $$1 = \frac12 + \frac12$$ and start with the expansion of $1$.

Ahaan S. Rungta
  • 7,398
  • 5
  • 25
  • 63
mariotomo
  • 333
  • 1
  • 6
22

First, without context, there is no way to tell if the problem is suitable for a given grade. It doesn't seem likely, but then maybe they did some work on the Egyptian representation of fractions, which is close enough that they might think of it. For me, that was my first thought, and I then came to:

$$\frac{1}{2}+\frac{1}{4}+\frac{1}{8}+\frac{1}{16}+\frac{1}{16}=1$$

Even with that hint, the question about uniqueness is not trivial. You can play with replacing only part of the expression, for example:

$$\frac{1}{8}+\frac{1}{16} = \frac{1}{6}+\frac{1}{48}$$

which yields:

$$\frac{1}{2}+\frac{1}{4}+\frac{1}{6}+\frac{1}{16}+\frac{1}{48}=1$$

but I don't see how 5th grade student can do that other than trial and error.

Other than that, I don't see what “typical fith-grade” reasoning can be used. Perfect numbers seem out of the question.

F'x
  • 1,370
  • 1
  • 14
  • 24
  • 3
    If the goal is simply to find any arbitrary solution, the problem should be solvable if someone starts with the three-number case; it shouldn't take long to find 1/2 + 1/3 + 1/6. From there, one may try to decompose the 1/6 into two parts which are unit fractions. A natural first effort would be to see what's left if one bumps 1/6 to the next smaller unit fraction (1/7). That turns out to leave 1/42--a unit fraction (hooray!) Subtract the next unit fraction smaller than that (1/43) to leave 1/1806. I wouldn't expect students to enumerate the solutions, though. – supercat Jan 30 '13 at 16:26
  • 1
    @supercat: if you decompose the 1/6 in '''three''' parts that are unit fractions, then you get the five fractions you need. see my other answer. according to me easy enough for the fifth grade student. – mariotomo Jan 30 '13 at 22:55
21

Please avoid the trivial $x=y=z=a=b=5$. Try looking for a solution where $x≠y≠z≠a≠b$ or if not, look for one where one variable equals to another, but explain your reasoning. The girl was covering "unit fractions" in her class.

Was this a requirement in the homework problem? The easiest way to prove that there is more than one solution to a problem is to find more than one solution to the problem, even if the additional solution is trivial. A starting point of any arbitrary number of unit fractions that sum to $1$ and that also can be combined into other unit fractions is a very valid starting point.

Case in point: I would propose that this be solved at the fifth-grade level by splitting fractions, with a little special knowledge to produce unique values. $1 = 1/2 + 1/2$. Now we need to come up with three more fractions and differentiate the two we have. We do this by continuing to split, with some clever knowledge of prime factors.

For instance, if a number in the denominator is divisible by both $2$ and $3$, then you can combine two or three of that unit fraction, and when reducing to lowest terms you will get a new unit fraction. $6$ is the simplest example; $1/6 + 1/6 = 1/3$ and $1/6 + 1/6 + 1/6 = 1/2$. This is useful to us: $1/6 + 1/3 = 1/2$, so $1/6 + 1/3 + 1/2 = 1$. Now we just need two more unique fractions. Well, let's split up the largest fraction, $1/2$, into more pieces than we tried before. If we divide it into 12 units, each of those will be $1/24$ of the whole. $2/24 = 1/12, 3/24 = 1/8$, and $6/24 = 1/4$. We can also combine $4$ and $8$ of these, but those two produce the fractions $1/6$ and $1/3$ which we already have. Now, by serendipity (or not), $1 + 2 + 3 + 6 = 12$. That would give us $1/24 + 1/12 + 1/8 + 1/4 = 1/2$. That's four of the five we need, and $1/2$ (the remaining fraction of the whole) isn't spoken for yet and can be produced by recombining $1/3$ and $1/6$, so $1/2 + 1/4 + 1/8 + 1/12 + 1/24 = 1$, and thus $\{2,4,8,12,24\}$ is a valid solution.

This generalizes into the following statement: find five numbers, $j,k,m,n$, and $p$, that are all factors of an arbitrary $z$, and that sum to $z$. Then $1/x=j/z$, $1/y=k/z$, and so on.

A good z-value to try in this case is $100$, which turns the fractions into simple integer percentages. $100$ has the following prime factorization: $2^2*5^2$. Each unique combination of those factors, plus the universal factor $1$, is a factor of $100$ and thus a possible value for the set of $5$ we need. There are 8 factors of $100$, not including $100$; $1, 2, 4, 5, 10, 20, 25$, and $50$. We can now apply a variation of the value-splitting; find numbers in this set that will sum to other numbers in the set. $100=50+50$. $50=25+25$. $25=20+5$, and $5=4+1$. Putting these identities together, $1+4+20+25+50 = 100$. Dividing everything by $100$ and finding lowest terms gives $1/100+1/25+1/5+1/4+1/2 = 1$, so $\{2,4,5,25,100\}$ is another valid solution.

Pretty much any number with $4$ or more factors other than $1$ and itself is a candidate z-value. It helps to have a few extra factors laying around, as was shown from the example above; a number with only $4$ unique factors, say $20$ $(2, 4, 5, 10)$, is unlikely to have those factors all sum to the original number as well (in this case the sum of those four factors is $21$; close but no cigar). It also helps if the prime factorization of the number (a graspable concept for a fifth grader, with multiplication, division and primes under her belt) includes more than one prime factor; this allows you to do the "divide by one, then multiply by the other" trick to produce unique denominators.

The are multiple, but not infinite, total solutions to this problem; the numbers involved are naturally bound by the two conditions; that natural numbers are used, and that only reciprocals of those natural numbers are allowed (no numerators other than 1). Other answers have empirically found all possible answers.

KeithS
  • 2,709
  • 1
  • 13
  • 22
20

There are many ways to obtain many solutions. Here is one systematic way to obtain solutions.

First look at a class of solutions such that $x \leq y \leq z \leq a \leq b \leq 10$ (other solution can be obtained as permutation of these).

This implies that $2 \leq x \leq 5$. So now start with $x=2$. This now means that $3 \leq y \leq 8$. Choose $y=3$. This now means that $7 \leq z \leq 10$. You will quickly find that no solution exists such that $x \leq y \leq z \leq a \leq b \leq 10$. Then choose $y=4$. Again you will find that no solution exists such that $x \leq y \leq z \leq a \leq b \leq 10$.

Going through this you will find that if we want $x=2$, and $x \leq y \leq z \leq a \leq b \leq 10$, then $$[2, 5, 10, 10, 10]; [2, 6, 9, 9, 9]; [2, 8, 8, 8, 8]$$ are the only solutions with $x=2$ and $x \leq y \leq z \leq a \leq b \leq 10$.

Now set $x=3$ and get bounds for the remaining variables to see that $$ [3, 3, 9, 9, 9]; [3, 4, 6, 8, 8]; [3, 5, 5, 6, 10]; [3, 6, 6, 6, 6]; $$ are the only solutions with $x=3$ and $x \leq y \leq z \leq a \leq b \leq 10$.

Here are the solutions such that $x \leq y \leq z \leq a \leq b \leq 10$ $$[x,y,z,a,b] \in \{[2, 5, 10, 10, 10]; [2, 6, 9, 9, 9]; [2, 8, 8, 8, 8]; [3, 3, 9, 9, 9]; [3, 4, 6, 8, 8]; [3, 5, 5, 6, 10];\\ [3, 6, 6, 6, 6]; [4, 4, 4, 8, 8]; [4, 4, 5, 5, 10]; [4, 4, 6, 6, 6]; [5, 5, 5, 5, 5]; \} $$

There are $114$ distinct solutions i.e. without permutations such that $$x \leq y \leq z \leq a \leq b \leq 100$$ and can be found here.

  • 1
    You can edit the code in my answer to preclude answers that have numbers above a certain bound. – Joe Z. Jan 30 '13 at 23:39
19

$$\frac{1}{2}+\frac{1}{3}+\frac{1}{8}+\frac{1}{30}+\frac{1}{120}=1$$
In general, if you want $n$ distinct unit fractions to sum up to $1$ , you can use a formula
(which I found almost by luck) and find a direct solution

$$\frac{1}{\frac{2!}{1}}+\frac{1}{\frac{3!}{2}}+\frac{1}{\frac{4!}{3}}+...+\frac{1}{\frac{n!}{n-1}}+\frac{1}{n!}=1$$
It is accepted at the Monthly and it will be published.

Konstantinos Gaitanas
  • 8,745
  • 4
  • 26
  • 42
11

Note that $1=\dfrac12+\dfrac12.$ Then we can use $\dfrac1n=\dfrac1{(n+1)}+\dfrac1{n(n+1)}$ to expand this up to five terms. In more general,

$$\color{Green}{\dfrac{1}{pqr}=\dfrac{1}{pr(p+q)}+\dfrac{1}{qr(p+q)}.}$$

For distinct 4-tuples, I got that $$1=\dfrac12+\dfrac14+\dfrac16+\dfrac1{12}$$ $$1=\dfrac12+\dfrac13+\dfrac17+\dfrac1{42}$$ $$1=\dfrac12+\dfrac13+\dfrac19+\dfrac1{18}$$ $$1=\dfrac12+\dfrac13+\dfrac1{10}+\dfrac1{15}.$$ Then continue inductively .....

Similarly, $1=\dfrac13+\dfrac13+\dfrac13$ is a special case of

$$\color{Blue}{\dfrac{1}{pqr}=\dfrac{1}{pq(p+q+r)}+\dfrac{1}{qr(p+q+r)}++\dfrac{1}{pr(p+q+r)}.}$$

For example, $\dfrac1{2n}=\dfrac1{2(n+3)}+\dfrac1{n(n+3)}+\dfrac1{2n(n+3)}$ for any positive integer $n.$

Bumblebee
  • 16,357
  • 4
  • 37
  • 69
11

I am posting another solution since one of my students found this method.
(He is 16 years old but as you see he is gifted)
Start with $\frac{1}{2}+\frac{1}{3}+\frac{1}{6}$. No write $\frac{1}{6}=\frac{3}{18}=\frac{1}{18}+\frac{2}{18}=\frac{1}{9}+\frac{1}{18}$ and you get $\frac{1}{2}+\frac{1}{3}+\frac{1}{9}+\frac{1}{18}=1$. Proceed like this: Write the last fraction (which will have an even denominator) of the sum as $\frac{1}{2n}=\frac{1}{6n}+\frac{2}{6n}=\frac{1}{3n}+\frac{1}{6n}$ and so on!

Konstantinos Gaitanas
  • 8,745
  • 4
  • 26
  • 42
11

Surprised that no-one proposed $1,-1,1,-1,1$ and a whole bunch of variations around it.

dustin
  • 7,933
  • 11
  • 42
  • 81
Andrey Regentov
  • 239
  • 1
  • 3
9

Here are ALL solutions (72 in total) for x<y<z<a<b

{2,3,7,43,1806},{2,3,7,44,924},{2,3,7,45,630},{2,3,7,46,483},{2,3,7,48,336},{2,3,7,49,294},{2,3,7,51,238},{2,3,7,54,189},{2,3,7,56,168},{2,3,7,60,140},{2,3,7,63,126},{2,3,7,70,105},{2,3,7,78,91},{2,3,8,25,600},{2,3,8,26,312},{2,3,8,27,216},{2,3,8,28,168},{2,3,8,30,120},{2,3,8,32,96},{2,3,8,33,88},{2,3,8,36,72},{2,3,8,40,60},{2,3,8,42,56},{2,3,9,19,342},{2,3,9,20,180},{2,3,9,21,126},{2,3,9,22,99},{2,3,9,24,72},{2,3,9,27,54},{2,3,9,30,45},{2,3,10,16,240},{2,3,10,18,90},{2,3,10,20,60},{2,3,10,24,40},{2,3,11,14,231},{2,3,11,15,110},{2,3,11,22,33},{2,3,12,13,156},{2,3,12,14,84},{2,3,12,15,60},{2,3,12,16,48},{2,3,12,18,36},{2,3,12,20,30},{2,3,12,21,28},{2,3,14,15,35},{2,4,5,21,420},{2,4,5,22,220},{2,4,5,24,120},{2,4,5,25,100},{2,4,5,28,70},{2,4,5,30,60},{2,4,5,36,45},{2,4,6,13,156},{2,4,6,14,84},{2,4,6,15,60},{2,4,6,16,48},{2,4,6,18,36},{2,4,6,20,30},{2,4,6,21,28},{2,4,7,10,140},{2,4,7,12,42},{2,4,7,14,28},{2,4,8,9,72},{2,4,8,10,40},{2,4,8,12,24},{2,4,9,12,18},{2,4,10,12,15},{2,5,6,8,120},{2,5,6,9,45},{2,5,6,10,30},{2,5,6,12,20},{3,4,5,6,20}

ZaMoC
  • 236
  • 4
  • 9
6

$$1={1\over 3}+{1\over 2}+{1\over 6}={1\over 3}+\left({1\over 4}+{1\over 4}\right)+{1\over 6}+\left({1\over5}-{1\over 5}\right)={1\over3}+{1\over4}+{1\over5}+{1\over6}+{1\over 20}$$

So we have

$${1\over3}+{1\over4}+{1\over5}+{1\over6}+{1\over20}=1$$

gymbvghjkgkjkhgfkl
  • 7,762
  • 6
  • 15
  • 61
6

You can try taking $x, y, z, a$ each with value larger than $5$ which may or may not be equal. Now find $b$ such a way that the given relation is satisfied. But this question is way beyond fifth grade student (my opinion).

Phani Raj
  • 2,230
  • 1
  • 13
  • 27
  • 37
    As long as it’s clearly understood to be a somewhat open-ended and challenging question, I think that it’s a great question at the $5$-th grade level. – Brian M. Scott Jan 30 '13 at 10:03
3

There are $147$ such $5$-tuples (see Gerry Myerson's answer).
There are $72$   $5$-tuples among them with all distinct $x,y,z,a,b$ (see J42161217's answer).
And Joe Z. had exhaustive list of solutions (see his answer).
Since his link doesn't work now, I'll post the list of solutions here for completeness.

Solutions are sorted by $x \leqslant y \leqslant z \leqslant a \leqslant b$ (and solutions with all distinct $x,y,z,a,b$ are marked as bold):

1) [2, 3, 7, 43, 1806]
2) [2, 3, 7, 44, 924]
3) [2, 3, 7, 45, 630]
4) [2, 3, 7, 46, 483]
5) [2, 3, 7, 48, 336]
6) [2, 3, 7, 49, 294]
7) [2, 3, 7, 51, 238]
8) [2, 3, 7, 54, 189]
9) [2, 3, 7, 56, 168]
10) [2, 3, 7, 60, 140]
11) [2, 3, 7, 63, 126]
12) [2, 3, 7, 70, 105]
13) [2, 3, 7, 78, 91]
14) [2, 3, 7, 84, 84]
15) [2, 3, 8, 25, 600]
16) [2, 3, 8, 26, 312]
17) [2, 3, 8, 27, 216]
18) [2, 3, 8, 28, 168]
19) [2, 3, 8, 30, 120]
20) [2, 3, 8, 32, 96]
21) [2, 3, 8, 33, 88]
22) [2, 3, 8, 36, 72]
23) [2, 3, 8, 40, 60]
24) [2, 3, 8, 42, 56]
25) [2, 3, 8, 48, 48]
26) [2, 3, 9, 19, 342]
27) [2, 3, 9, 20, 180]
28) [2, 3, 9, 21, 126]
29) [2, 3, 9, 22, 99]
30) [2, 3, 9, 24, 72]
31) [2, 3, 9, 27, 54]
32) [2, 3, 9, 30, 45]
33) [2, 3, 9, 36, 36]
34) [2, 3, 10, 16, 240]
35) [2, 3, 10, 18, 90]
36) [2, 3, 10, 20, 60]
37) [2, 3, 10, 24, 40]
38) [2, 3, 10, 30, 30]
39) [2, 3, 11, 14, 231]
40) [2, 3, 11, 15, 110]
41) [2, 3, 11, 22, 33]
42) [2, 3, 12, 13, 156]
43) [2, 3, 12, 14, 84]
44) [2, 3, 12, 15, 60]
45) [2, 3, 12, 16, 48]
46) [2, 3, 12, 18, 36]
47) [2, 3, 12, 20, 30]
48) [2, 3, 12, 21, 28]
49) [2, 3, 12, 24, 24]
50) [2, 3, 13, 13, 78]
51) [2, 3, 14, 14, 42]
52) [2, 3, 14, 15, 35]
53) [2, 3, 14, 21, 21]
54) [2, 3, 15, 15, 30]
55) [2, 3, 15, 20, 20]
56) [2, 3, 16, 16, 24]
57) [2, 3, 18, 18, 18]
58) [2, 4, 5, 21, 420]
59) [2, 4, 5, 22, 220]
60) [2, 4, 5, 24, 120]
61) [2, 4, 5, 25, 100]
62) [2, 4, 5, 28, 70]
63) [2, 4, 5, 30, 60]
64) [2, 4, 5, 36, 45]
65) [2, 4, 5, 40, 40]
66) [2, 4, 6, 13, 156]
67) [2, 4, 6, 14, 84]
68) [2, 4, 6, 15, 60]
69) [2, 4, 6, 16, 48]
70) [2, 4, 6, 18, 36]
71) [2, 4, 6, 20, 30]
72) [2, 4, 6, 21, 28]
73) [2, 4, 6, 24, 24]
74) [2, 4, 7, 10, 140]
75) [2, 4, 7, 12, 42]
76) [2, 4, 7, 14, 28]
77) [2, 4, 8, 9, 72]
78) [2, 4, 8, 10, 40]
79) [2, 4, 8, 12, 24]
80) [2, 4, 8, 16, 16]
81) [2, 4, 9, 9, 36]
82) [2, 4, 9, 12, 18]
83) [2, 4, 10, 10, 20]
84) [2, 4, 10, 12, 15]
85) [2, 4, 12, 12, 12]
86) [2, 5, 5, 11, 110]
87) [2, 5, 5, 12, 60]
88) [2, 5, 5, 14, 35]
89) [2, 5, 5, 15, 30]
90) [2, 5, 5, 20, 20]
91) [2, 5, 6, 8, 120]
92) [2, 5, 6, 9, 45]
93) [2, 5, 6, 10, 30]
94) [2, 5, 6, 12, 20]
95) [2, 5, 6, 15, 15]
96) [2, 5, 7, 7, 70]
97) [2, 5, 8, 8, 20]
98) [2, 5, 10, 10, 10]
99) [2, 6, 6, 7, 42]
100) [2, 6, 6, 8, 24]
101) [2, 6, 6, 9, 18]
102) [2, 6, 6, 10, 15]
103) [2, 6, 6, 12, 12]
104) [2, 6, 7, 7, 21]
105) [2, 6, 8, 8, 12]
106) [2, 6, 9, 9, 9]
107) [2, 7, 7, 7, 14]
108) [2, 8, 8, 8, 8]
109) [3, 3, 4, 13, 156]
110) [3, 3, 4, 14, 84]
111) [3, 3, 4, 15, 60]
112) [3, 3, 4, 16, 48]
113) [3, 3, 4, 18, 36]
114) [3, 3, 4, 20, 30]
115) [3, 3, 4, 21, 28]
116) [3, 3, 4, 24, 24]
117) [3, 3, 5, 8, 120]
118) [3, 3, 5, 9, 45]
119) [3, 3, 5, 10, 30]
120) [3, 3, 5, 12, 20]
121) [3, 3, 5, 15, 15]
122) [3, 3, 6, 7, 42]
123) [3, 3, 6, 8, 24]
124) [3, 3, 6, 9, 18]
125) [3, 3, 6, 10, 15]
126) [3, 3, 6, 12, 12]
127) [3, 3, 7, 7, 21]
128) [3, 3, 8, 8, 12]
129) [3, 3, 9, 9, 9]
130) [3, 4, 4, 7, 42]
131) [3, 4, 4, 8, 24]
132) [3, 4, 4, 9, 18]
133) [3, 4, 4, 10, 15]
134) [3, 4, 4, 12, 12]
135) [3, 4, 5, 5, 60]
136) [3, 4, 5, 6, 20]
137) [3, 4, 6, 6, 12]
138) [3, 4, 6, 8, 8]
139) [3, 5, 5, 5, 15]
140) [3, 5, 5, 6, 10]
141) [3, 6, 6, 6, 6]
142) [4, 4, 4, 5, 20]
143) [4, 4, 4, 6, 12]
144) [4, 4, 4, 8, 8]
145) [4, 4, 5, 5, 10]
146) [4, 4, 6, 6, 6]
147) [5, 5, 5, 5, 5]


To obtain these 5-tuples, one can use following steps:
- for $x$: loop from $2$ to $5$;
- for $y$: loop from $x$ to $\displaystyle\Biggl\lfloor{\frac{4}{1-\frac{1}{x}}}\Biggr\rfloor$;
- for $z$: loop from $y$ to $\displaystyle\Biggl\lfloor\frac{3}{1-\frac{1}{x}-\frac{1}{y}}\Biggr\rfloor$;
- for $a$: loop from $z$ to $\displaystyle\Biggl\lfloor\frac{2}{1-\frac{1}{x}-\frac{1}{y}-\frac{1}{z}}\Biggr\rfloor$;
- check whether appropriate integer $b$ exists.

Oleg567
  • 16,708
  • 2
  • 36
  • 74
0

Note that $\frac{4!}{4!}=1$. Now write $4!=24$ as $1+2+3+6+12$ (basically the divisors of $24$) Then we have $$1=\frac{4!}{4!}=\frac{1+2+3+6+12}{24}=\frac{1}{24}+\frac{1}{12}+\frac{1}{8}+\frac{1}{4}+\frac{1}{2}$$

C.S.
  • 5,364
  • 14
  • 31
0

If you allow 2 fractions to be equal, as the question allowed, you can just use powers of $2$ to generate a solution for any number of fractions since:

$1/2 +1/2^2+...+1/2^n + 1/2^n = 1$

for all $n>1,$ containing $n+1$ fractions, so we need $n=4$

$1/2 + 1/4 + 1/8+ 1/16 + 1/16 =1$

This may be an easier solution to discover at school.

In addition with this idea you can easily create a solution with distinct fractions from the case of $1/2+1/3+1/6=1$ (which can easily be found by trial and error) using this idea by dividing by $4$ and using the $n=2$ case.

Thus $1/8+1/12+1/24=1/4$

and $1/2+1/4+1/4=1$ giving

$1/2+1/4+1/8+1/12+1/24=1.$

Ivan
  • 575
  • 4
  • 8