# Infinity or -1/12?

Recently a very strange result has been making the rounds. It says that when you add up all the natural numbers

1+2+3+4+...

then the answer to this sum is -1/12. The idea featured in a Numberphile video (see below), which claims to prove the result and also says that it's used all over the place in physics. People found the idea so astounding that it even made it into the New York Times. So what does this all mean?

### The maths

First of all, the infinite sum of all the natural number is not equal to -1/12. You can easily convince yourself of this by tapping into your calculator the partial sums

and so on. The get larger and larger the larger gets, that is, the more natural numbers you include. In fact, you can make as large as you like by choosing large enough. For example, for you get

and for you get

This is why mathematicians say that the sum

diverges to infinity. Or, to put it more loosely, that the sum is equal to infinity.

Srinivasa Ramanujan

So where does the -1/12 come from? The wrong result actually appeared in the work of the famous Indian mathematician Srinivasa Ramanujan in 1913 (see this article for more information). But Ramanujan knew what he was doing and had a reason for writing it down. He had been working on what is called the Euler zeta function. To understand what that is, first consider the infinite sum

You might recognise this as the sum you get when you take each natural number, square it, and then take the reciprocal:

Now this sum does not diverge. If you take the sequence of partial sums as we did above,

then the results you get get arbitrarily close, without ever exceeding, the number Mathematicians say the sum converges to , or more loosely, that it equals

Now what happens when instead of raising those natural numbers in the denominator to the power of 2, you raise it to some other power ? It turns out that the corresponding sum

converges to a finite value as long as the power is a number greater than . For every , the expression has a well-defined, finite value. is what’s called a function, and it’s called the Euler zeta function after the prolific 17th century mathematician Leonhard Euler.

So far, so good. But what happens when you plug in a value of that is less than 1? For example, what if you plug in ? Let’s see.

So you recover our original sum, which, as we know, diverges. The same is true for any other values of less than or equal to 1: the sum diverges.

### Extending the Euler zeta function

As it stands the Euler zeta function S(x) is defined for real numbers x that are greater than 1. The real numbers are part of a larger family of numbers called the complex numbers. And while the real numbers correspond to all the points along an infinitely long line, the complex numbers correspond to all the points on a plane, containing the real number line. That plane is called the complex plane. Just as you can define functions that take real numbers as input you can define functions that take complex numbers as input.

One amazing thing about functions of complex numbers is that if you know the function sufficiently well for some set of inputs, then (up to some technical details) you can know the value of the function everywhere else on the complex plane. This method of extending the definition of a function is known as analytic continuation. The Euler zeta function is defined for real numbers greater than 1. Since real numbers are also complex numbers, we can regard it as a complex function and then apply analytic continuation to get a new function, defined on the whole plane but agreeing with the Euler zeta function for real numbers greater than 1. That's the Riemann zeta function.

But there is also another thing you can do. Using some high-powered mathematics (known as complex analysis, see the box) there is a way of extending the definition of the Euler zeta function to numbers less than or equal to 1 in a way that gives you finite values. In other words, there is a way of defining a new function, call it so that for

and for the function has well-defined, finite values. This method of extension is called analytic continuation and the new function you get is called the Riemann zeta function, after the 18th cenury mathematician Bernhard Riemann. (Making this new function give you finite values for involves cleverly subtracting another divergent sum, so that the infinity from the first divergent sum minus the infinity from the second divergent sum gives you something finite.)

OK. So now we have a function that agrees with Euler’s zeta function when you plug in values . When you plug in values , the zeta function gives you a finite output. What value do you get when you plug into the zeta function? You’ve guessed it:

If you now make the mistake of believing that for , then you get the (wrong) expression

This is one way of making sense of Ramanujan’s mysterious expression.

### The trick

So how did the people in the Numberphile video "prove" that the natural numbers all add up to -1/12? The real answer is that they didn’t. Watching the video is like watching a magician and trying to spot them slipping the rabbit into the hat. Step one of the "proof" tries to persuade you of something rather silly, namely that the infinite sum

is equal to

The video doesn’t dwell long on this and seems to imply it’s obvious. But let’s look at it a little closer to see if it makes sense at all. Suppose that the sum has a finite value and call it . Adding to itself you get the infinite sum

But this is just the original sum, implying

Since it follows that which is nonsense. So the assertion that the infinite sum can be taken to equal to 1/2 is not correct. In fact, you can derive all sorts of results messing around with infinite sums that diverge. It’s a trick!

### The physics

But how did this curious, wrong result make it into a physics textbook, as shown in the video? Here is where things really get interesting. Suppose you take two conducting metallic plates and arrange them in a vacuum so that they are parallel to each other. According to classical physics, there shouldn't be any net force acting between the two plates.

Illustration of the Casimir effect. Image: Emok.

But classical physics doesn't reckon with the weird effects you see when you look at the world at very small scales. To do that, you need quantum physics, which tells us many very strange things. One of them is that the vacuum isn't empty, but seething with activity. So-called virtual particles pop in and out of existence all the time. This activity gives a so called zero point energy: the lowest energy something can have is never zero (see here for more detail).

When you try to calculate the total energy density between the two plates using the mathematics of quantum physics, you get the infinite sum

This infinite sum is also what you get when you plug the value into the Euler zeta function:

That’s unfortunate, because the sum diverges (it does so even quicker than than ), which would imply an infinite energy density. That’s obviously nonsense. But what if you cheekily assume that the infinite sum equals the Riemann zeta function, rather than the Euler zeta function, evaluated at ? Well, then you get a finite energy density. That means there should be an attractive force between the metallic plates, which also seems ludicrous, since classical physics suggests there should be no force.

But here’s the surprise. When physicists made the experiment they found that the force did exist — and it corresponded to an energy density exactly equal to !

This surprising physical result is known as the Casimir effect, after the Dutch physicist Hendrik Casimir.

Take a moment to take this in. Quantum physics says the energy density should be

That’s nonsense, but experiments show that if you (wrongly) regard this sum as the zeta function evaluated at , you get the correct answer. So it seems that nature has followed the ideas we explained above. It extended the Euler zeta function to include values for that are less than 1, by cleverly subtracting infinity, and so came up with a finite value. That’s remarkable!

The reason why we see and in the Numberphile video and the physics textbook, rather than and is that when you imagine the Casimir effect as happening in one dimension (along a line rather than in 3D), the energy density you calculate is rather than .

So why did the Numberphile people publicise this strange "result"? They certainly know about the analytic continuation that makes the function well-defined, but that was something that was a little too technical for their video. Knowing they had the analytic continuation method, that would make the final result OK, hidden in their back pocket, they went ahead with their sleight of hand. In doing so they got over a million hits and had the world talking about zeta functions and mathematics. For this they should be congratulated. The mathematics of zeta functions is fantastic and what we described here is just the start of a long list of amazing mathematical properties. In bringing mathematics and physics to the public we always have to make choices about what we leave out and what we explain. Where to draw that line is something we all have to leave to our consciences.

David Berman is a Reader in Theoretical Physics at Queen Mary, University of London. He previously spent time at the universities of Manchester, Brussels, Durham, Utrecht, Groningen, Jerusalem and Cambridge as well as a year at CERN in Geneva. His interests outside of physics include football, music and theatre and the arts.

Marianne Freiberger is editor of Plus.

### Is that SO

OMG !! What Have YOU Done?????
Z+Z= the original sum i.e. Z
so you write 2Z=Z
so putting Z=0.5, we get 0.5=1
but why stop here?
It's an infinite sum so at my will I can write (Z+Z = Z+Z) or (Z+Z = Z+Z+Z+.........+Z)
so, Z=Z or Z=nZ
so whatever Z is, Z can be anything else....it's nonsense. RIGHT??
Don't stop. It's not like counting bricks. Whenever you put an end to it, its a divergent series and for that don't assign any value for GOD's sake!
BUT, try to think it this way. The sum is indeed INFINITE. That is the only truth. It doesn't end. But that infinite is a pile of junk. There is a bit of finite part in that infinity. Whenever you stop and calculate, each time it blows up. But there is this (-1/12) part which only accumulates if you allow it to go on forever. And yes we physicists use it in quantum fields and all sort of places. Each time we replace that infinite sum with this very value, we confirm the result! Don't teach things you don't fully understand.

### Skeptical

I like the explanation here distinguishing between the Reimann and Euler zeta functions, but I am still skeptical about their use in characterizing divergent sums.
Even though the Reimann zeta function gives a value for the attraction between two metal plates in a vacuum that matches the value given my QM, it still does not make the sum of the natural numbers or QM any more real. Sure, the both have utility in explaining things: i.e. the force between two plates, but that does not definitively exclude other possibilities.

Galileo sort of argued this for planetary motion to appease the church: you don't have to really believe that the Earth revolves around the sun, but if you want predict where planets will be, you can do the math as if the Earth does revolve around the sun.
And Einstein's general relativity: An elevator sitting in a gravitational field behaves as if we are in an elevator in an accelerating spaceship. Since there is no distinguishing these two scenarios, they are equivalent.
Finally, if Fourier series can represent other functions (I am thinking about X-ray crystallography) then which is more real: the Fourier co-efficients or the values of the real function. If they can both represent the same thing, but it is useful to use one sometimes and the other at other times, can one be more real than the other?
Back to QM and the sum or natural numbers... I wonder whether the treatment of the problem is incorrect (unreal) in both cases, but it cancels each other out to arrive at accurate values. Aren't there other theories still around that describe subatomic events without invoking QM? If these are applied to the Casimir plates, maybe it bypasses the Euler zeta function and uses the Reimann zeta function.

### Infinity or -1/12

Thank you very much for this eloquent and succinct article. It's restored my faith in "intuition".
We are in an era where the rush to publish a "result" for reward or attention has never been greater. It seems rigour is taking a back seat in this new era, so we must stay on our toes if we are to find truth amidst all the noise.

### One Problem

You keep saying this sum is 'wrong.' Of course, it's not. Obviously, if you EVER stop at any point while taking this sum, you'll end up at a positive, finite, and possibly massive number. But the point is, your are taking the sum to infinity, meaning that there ISN'T a stopping point. The sum of all positive integers is -1/12, and it's been proven (the zeta function is real and completely valid); it's not 'wrong' just because it doesn't make sense.

Also, (interesting but perhaps unrelated) if you take the ideas used to calculate this sum, you can find that the sum of all odd positive integers is +1/12 the sum of all positive even integers is -1/6.

### With my little knowledge on

With my little knowledge on maths I think you made a mistake.I was actually bugged about infinity sums for a while but I slowly start to understand them thankst your maths.You said Z+Z=1-1+1-1...1-1+1-1 which equals to Z.Well there's a problem in your maths!You seem to think that Z is actually ending and another Z begins!The Z function never ends and that's why you can't add something on the end of an infinite function!That's why I think (with my basic math knowledge) you shouldn't add a Z function to the end of another Z function!Because you don't know if the Z function will end with a 1 or a -1!The best approach of how much Z+Z is this:
Z+Z=1+1-1-1+1+1-1-1...=2(1-1+1-1+....)=2*1/2=1

Thank for a great night read :)

### Does the result -1/12 work in physics?

The guys in the video mention that they know the answer must be -1/12 because that result allows them to make other accurate calculations regarding string theory. Well, are there any exmaples that show it the answer can or cannot be -1/12?

### 1+2+3+4+5+... forever

So isn't this series basically sigma of n for which n starts from 1 and ends at infinity? In this case, in my opinion, as every terms in this series is positive, this series can immediately be included in the 'comparison tests for convergence and divergence'. French mathematician Nicole Oresme has proved the divergence of the harmonic series 1/n by proving the divergence of a series which has lesser terms than that of the harmonic series. If, as what the numberphile people are saying, 1+2+3+4+... forever does equals to -1/12 doesn't this make the whole concept of comparison test flawed? Even if the Grandi's series which is 1-1+1-1+1-1... does actually equals to a half, the answer 1+2+3+4+.. forever does not make sense as it makes previous proofs and statements made in infinite series incorrect.

Also, i don't understand what the average partial sums are for. Why is there the need to average the partial sums? The average of the partial sums may converge towards a half but that doesn't mean anything that the 'non-average' (as in real) series converge towards a half...

### +1-1+1-1+1-1 . . . in cycles

One of the guys in the vid says there's a proof for Z (the Grandi series) = 1/2. I'll leave that to the more advanced, but to me that makes sense intuitively. In the absence of a graph for now, imagine a horizontal wavy line ascending with every +1 to a peak at 1 on the y axis and descending with -1 again to 0. Repeat, say with six 1's, so for three cycles. Now draw another horizontal straight line skimming along from peak to peak, and clearly you will have a figure falling into two equal halves above and below the wavy line. Each of these two halves will also equal the halves created by yet another horizontal straight line drawn from .5 on the y axis.

Now for the smart bit. Do an Oresme on the series, rearrange the +1's and the -1's into different, though still zero sum cycles: +1+1+1-1-1-1. This time the amplitude will rise to 3 and the horizontal through the mid points will be at 1.5, so the sum of the series will be 3/2.

This hocus pocus, if you want to call it that, also has an echo in nature. I've just been learning about Cepheid variable stars, and the close direct relationship between luminosity and the period over which luminosity varies from minimum to maximum and back. Luminosity, energy emitted per unit of time, could be represented by the area under the wavy line maybe. How the values 1/2 and 3/2 would relate to the natural parameters of luminosity and period I don't know, but any comments gratefully received.

### 1-1+1-1+1-...=1/2

In order to find the sum of this series, first of all, the sequence of the partial sums must converge. However, the sequence of partial sums does not converge. Rather, it oscillates between 0 and 1.
Thus, insofar as the sequence does not converge there is no reason to talk about a unique sum of the series. In this case we can prove anything we want: 1, 1/2...

### I saw a demonstration of this

I saw a demonstration of this sum recently. It went like this:
S=1-1+1-1+...
=1-(1-1+1-...)
the periods indicating a continuation to infinity, it is assumed that:
S=1-S
2S=1
S=1/2

Then again, the alternating harmonic series converges to ln2 but by rearranging the terms in the equation, it also converges to (ln2)/2 and (ln2)/4 and so forth. The problem in your reasoning is that you take a partial sum as reference. I know that it's counter-intuitive just like 0.00000...1= 0 but the reason it gives that result is only because the sum is infinite. The moment it becomes finite, these results all go wrong.

### Cesaro convergence

Dear David and Marianne,

Thank you for the interesting explanations. I would remark, though, that the summation "1-1+1-1+... = 1/2" is not the source of fallacy in the Numberphile video. That series, while not convergent, is still "Cesaro summable" [which means, that the average value of its first n partial sums converges in the limit as n is taken to infinity].

The Cesaro sum of series has many nice properties, e.g., allowing a certain class of rearrangements of the terms, allowing the sums to be added when Cesaro summable series are added term by term, etc. The fallacies enter elsewhere, in the various manipulations of divergent and non-Cesaro summable series. Such manipulations can be used to arrive at any value at all for the sum of all the natural numbers.

Perhaps others have already made this point, as I've not viewed all of the linked and related sites.

Like others, I too would have wished to see some indication in the Numberphile video of the "tongue in cheek" nature of the "proof," without losing the huge entertainment value. Fallacies (like "proving" that 1 = 0 with a hidden divisioon by zero in the algebra) have always been a source of fun and exploration in school mathematics.

Here, the video is understandable to many middle grades students as well as high school students; this is great! But their teachers then need to know it is "sleight of hand," and encourage their students to question the "authority" behind it. Teaching mathematics by authority rather than through understanding is a continuing, unfortunate trend in our schools, at least in the USA, and I cannot help but feel an opportunity here might be being missed.

Gerald Goldin
Rutgers University

### A proof without Cesaro convergence

Disclaimer: I'm just a 7th grader who's smart, so don't trust me on this.

Alright, so there actually is a different proof that 1-1+1-1+1...=1/2. First, let's consider this way of solving this convergent series, namely 1+1/2+1/4+1/8.... We already know this sums to 2.

Let us call this series S. Then 2S=2+1+1/2+1/4+1/8... = 2+S.

2S=2+S
S=2

We can use similar methods to work with divergent series and other series. If 1-1+1-1+1-1... equals S, then S+S=1-1+1-1+1...+1-1+1-1+1-1...

Shift the second S over by one term, and all terms after the first in the first S get canceled out, leaving us with 2S=1 and S=0.5.

Similar methods are used in the final proof.

### summation of 1 - 1 + 1 - 1 ....

Hi,

I believe you have made a mistake in doing Z + Z = 2Z = Z

for 2Z to equal Z you have to assume that one of the Z sums ends at -1 so that the other can conveniently start at 1. This is a misconception as these are infinite series that have no end.

The only conclusion you can draw from Z + Z is as follows

Z = 1 - 1 + 1 - 1 + 1 - .......
+
Z = 1 - 1 + 1 - 1 + 1 - .......

which is equal to 2 - 2 + 2 - 2 + 2 - .......

Christian Whittaker email me at cpwhittaker@hotmail.co.uk if you are unsure (or disagree) with this

take out a factor of 2 you get 2( 1 - 1 + 1 - 1 + 1 - .....) = 2Z in which case you are back where you started

So because 2Z is not infact Z, your theory is irrelevant

### Thank you!

Thanks very much for this post - I feel like this is the follow up that like Numberphile is missing!

I'm not quite sure what Numberphile were thinking... I'm all for promoting discussion and argument, but even I have "limits" (-: Seriously though, I hope that Numberphile come back with a follow up video at some stage explaining themselves!

### Closure Property of Natural Numbers

Ok so I have read the article and watched the shorthand proof on numberphile. However don't the natural numbers have a closure property, meaning if you add any of the natural numbers together you must get a natural number? So how can it be that when we add all of them, we get -1/12, which does not belong to the natural numbers? I think my idea of adding all the natural numbers doesn't match with what is going on in the article, so if anyone could fill in the gaps that would be great.

### The closure of the natural

The closure of the natural numbers under addition means that the sum of any two natural numbers is a natural numbers. We can apply this principle again and again (finitely many times) to see that the sum of any finite number of natural numbers is a natural number. This however says nothing of the sum of an infinite number of naturals, which we have in the case of an infinite series; in fact we see that this need not exist at all.

Similarly, whilst rational numbers are closed under addition also, the sequence S given in the article for pi/6 gives an example of a series of rational numbers converging to an irrational number (the rational numbers are also not closed in another sense, as a sequence of rational numbers can converge to an irrational number).

### (pi^2)/6

The limit of a series of rationals can be irrational. (Pi^2)/6 is an accumulation point in Q. It element of the epsilon neighborhood of the partial sums Sn for a lage n. Sn is still part of Q for all n.

### Very good, the naturals are

Very good, the naturals are closed under addition. And indeed the point of this article is that the sum of all naturals is not -1/12. Only the so called regularised sum where "an infinity" has been subtracted- that "infinity" is obviously not a natural.

### Almost there

Ok thanks for the response, sorta wanna get my head around this incase I tell people about it. I like how Numberphile made a separate video for the sequence S1 = 1/2, cheeky way to get more views. Where this all loses me is where the extended function gets its values from when you plug in values less than 1 (somewhere in the article it says less than -1, is this a mistake, what about -1

### yep it should be less than or

yep it should be less than or equal to -1.

So exactly where does it get its value from? This is the technical bit about analytic continuation. There is a way, beyond the scope of this article to extend functions of complex numbers once you know the value of the function elsewhere. I understand why the numberphile people didn't want to bring that up because it is hard to explain nontechnically and indeed to give them credit where it is due, they have a discussion about it in an adjoining video and a written bit about it now too. So basically there is a bit of maths using function of complex numbers that allow you to assign a value to a function once you know its values elsewhere. The point of the artice was that this was too hidden for my taste which is why we still get posts like the one above where they want to still claim that the sum of all the integers is -1/12. it is not. Only a function that is related to that sum has that value. Where by related I mean that function has the same value as that sum when that sum is convergent.

### The Numberphile people do explain the 1/2 result

Z = 1 - 1 + 1 - 1 + 1 - 1 ......
1 - Z = 1 -[ 1 - 1 + 1 - 1 + 1 - 1 ....... ] = 1 - 1 + 1 - 1 ..... = Z
1 - Z = Z
2Z = 1
Z = 1/2

### 1 - Z = Z ; implying: (z+z=1) derives from nonsense

statement is logically faulty ;
substitute : z=zero
result : one minus zero is zero !!!
facit : give me a buck (=1) for each such transaction !

### substitute : z=zero is logically faulty

Anonymous proves their own point by using nonsense themselves? huh?
a buck? might as well give you an imaginary goat instead.

### So, I will say it

So, I will say it again..

Writing Z=.. and then assuming it is a well defined quantity is an assumption you cannot make.

The sum is nonconvergent. However you can very simply analytically continue the function that the sum represents beyond the radius of convergence but that is a continuation. The sum itself is not defined, it is not convergent. So if its nonconvergent don't write Z=..
It doesn't equal anything...

### The Numberphile folks have a

The Numberphile folks have a video that goes through basically the same area, talks about the sleight-of-hand in the shorter video, discusses analytic continuation and all that good stuff, and it's linked at the end of the short version. Well worth a watch: www.youtube.com/watch?v=E-d9mgo8FGk

### yes indeed, this is all good

yes indeed, this is all good stuff and explains what lies behind it all.

### The sum S(-3) is wrong

The expression for S(-3) two lines below the two plates draw is wrong. It reads S(-3)= 1/2^3 + 1/3^3 + 1/4^3 +... and should be S(-3)=1/2^-3 + 1/3^-3 + 1/4^-3 +...
Thanks.

### Thanks for pointing that out!

Thanks for pointing that out! We've corrected it.

It's cooooool.

### Infinity or -1/12?

You can't add two infinite series the way you did.

[ Z+Z ≠ 1-1+1-1+1-....+1-1+1-1+1-… .]

However,

Z+Z = 2-2+2-2+2-… ≈ 1,

which is consistent with, Z≈ 1/2.

### Not sure how to answer you.

Not sure how to answer you. Non-convergent series are unpleasant and subtle things most 1st year mathematics text books will tell you the convergence properties of this sum. Instead let me just quote from Abel in 1828, "Divergent series are the invention of the devil and it is shamful to base on them any demonstration whatsoever". Hardy's excellent book on the subject explains the rights and wrongs; the good the bad and the ugly.

### Congratulated?

I appreciate this piece, and it's helped me make sense of a lot of what's going on here. However, I have to part ways in calling for the Numberphile folks to be "congratulated." As a math teacher, and one who struggles every day to counter the deeply-ingrained notion that math makes no sense whatsoever, I can't stand it when that notion is spread across a wide audience and further ingrained into our culture. Sure, those who are already somewhat mathematically inclined are intrigued and want to know more. But those who are not see a video like this and say to themselves, "Further confirmation that math makes absolutely no sense." So I'm not about to congratulate them.

Matt E.
Sharon, MA

Thank you for your comments, its something that many math teachers have expressed to me including lecturers at university which is what prompted us to write the piece; Its clear to me where I would draw the moral line and it isn't in the same place they have .

### Isn't it more important to

Isn't it more important to recognise that maths is about adventure, discovery, fun? This is simply an example of an important aspect of in mathematics - you take some concept, abstract it and extend it, and see where it leads. Shouldn't more maths teaching and learning emphasise this?

A nice example that I use in my classroom, far simpler than analytic continuation and one that survives the journey more intact, concerns the index laws. You take the well understood concept that a times a times a... n times is a^n, and uncover the fact that a^m times a^n = a^(m+n). Then generalise the concept of power so you can consider things that "don't make sense" like a^(1/2) or a^(-3). These are not meaningful under the initial view that "power means repeated multiplication", but the extension and subsequent exploration lead to important and very meaningful results. Learning the index laws can be an exercise in rote memorisation, or it can be a wonderful journey of discovery, where seeming "nonsense" becomes clarified and empowering!

Numberphile demonstrate this side of maths, and should absolutely be congratulated.

### Non mathematician's view

Well, I think the Numberphile guys made an entertaining video and I wish that my school maths teachers had been even half as good with their explanations as these guys are. For me it was the teachers, with their leaps of faith, blasting through the set books and omitting whole chapters (homework : study chapter so and so and do the exercises) who instilled the fear of maths and a sense of futility into all but the three or four kids, out of a class of 25, who could figure out what was going on. Maths teachers take a look in the mirror!

Nick
Berlin, Germany