Cheshire Cat knows where infinities go. |

*Solving the mystery of the controversial (and errant) claim that 1+2+3+4+... =-1/12.*

In the previous post on this subject, I somewhat tentatively asserted that the sum of the natural numbers up to infinity (1+2+3+4+ . . .) does not equal -1/12.

So what does it equal? The answer, at least as far as realistic problems I can wrap my head around goes, is -1/12

*plus infinity*.

**Finding -1/12 in the first place**

As the Numberphiles point out, there are many ways to associate the sum of the natural numbers with -1/12. To name just a few . . .

This graph offers a simple way to associate -1/12 with 1+2+3+4+ . . . |

- There's the video Numberphiles made that shows how you can play some potentially dicey games with series manipulations to come up with the equation.

- There's the slightly more rigorous way Euler did it, which the Numberphiles document in a second video on the topic.

- There's the simple way I did it in a previous post that involves nothing more than a little algebra, a trivial amount of calculus, and a graph (like the one here) that you can draw based on the sum of the natural numbers (or any other series consisting of the sum of the natural numbers raised to a non-negative integer power).

- And, of course, there's always the possibility of a straightforward calculation based on a simple physical problem

Fortunately, there's an example that's pretty easy to describe, which relies on the sum of the cubes of the natural numbers 1^3+2^3+3^3+4^3+ . . .

Like the sum of natural numbers, the sum of the cubes can be associated with a small number. Instead of -1/12, it's 1/120. That is, instead of the "equation"

1+2+3+4+ . . . =-1/12

I'm going to talk about the "equation"

1^3+2^3+3^3+4^3+ . . . = 1/120.

The sum comes up in the calculation of the Casimir Effect. Even though the full calculation is nothing to sneeze at, the general set up and the solution behind the mystery of bizarre equations like the ones above are pretty easy to see.

(For those of you who are proficient at math and want to cut to the chase, just read this excellent explanation that Jonathan Dowling wrote up for Mathematics Magazine in 1989).

**The Casimir Effect**

Three quarters of a century ago, Hendrik Casimir proposed that two metal plates suspended in space will attract each other through a previously unimagined force that had nothing to do with gravity, electricity, magnetism or any other mechanism known at that time. The force is tiny and only becomes significant over minute distances, so detecting it is tricky. Nevertheless, nearly twenty years after his proposal, experimentalists were able to measure the attraction Casimir predicted in the lab.

There is some debate about exactly what causes the Casimir Effect, but the most interpretations involve something called vacuum energy. Essentially, the story goes, we are immersed in a sea of particles. An infinite number of particles, in fact. Each carries a tiny amount of energy, but because there is an infinite number of them, the total amount of energy is also infinite.

We aren't usually aware of the energy all around us because we're immersed in it, in the same way you don't notice the pressure of the atmosphere you're immersed in. The air around you exerts pressure (15 pounds on each square inch of your body, actually), but you don't feel it squeezing you because it's the same pressure inside your body as it is outside, and it perfectly balances out. This is also true of the vacuum energy, only instead of 15 pounds per square inch, the pressure is infinite.

One way to detect the vacuum energy would be to keep it out of some region. It would be similar to detecting the pressure of the atmosphere by sucking the air out of a soda bottle to make it collapse. Casimir's plates do almost something like that, except for vacuum energy instead of air.

Infinite vacuum energy outside the plates vs. slightly smaller infinite energy inside. |

As you can see in the sketch, the longest wavelength that can fit between the plates corresponds to the arc at the lowest position. That wavelength has a frequency associated with it, which I will call

**f**. The next curve up has twice the frequency of the first, for a frequency of

**2f**, the third has

**3**

**f**, and so forth. Because light can have any frequency, you can have an infinite ladder of light frequencies between the plates.

Each frequency, in turn contributes some energy to the region between the plate. In fact, the energy per wavelength between the plates is proportional to the frequency cubed, so the total energy between the plates is proportional to

f^3 +(2f)^3 + (3f)^3 + (4f)^3 . . . = f^3*(1 + 2^3 + 3^3 + 4^3 + . . . )

It's reasonable to guess that adding the cubes of all the integers up this way means that the energy between the plates is infinite

But the vacuum energy outside the plates consists of all frequencies, including those that exist between the plates, so the energy outside is also infinity - although it's clearly a larger infinity. To figure out the pressure difference between the inside and outside, you have to subtract one infinity from the other.

Normally, subtracting infinity from infinity is meaningless. But in this problem we know a lot about the infinities. The paper I link to above by Jonathan Dowling goes through two ways to calculate the difference between the two infinities. In both calculations, Dowling shows that for the Casimir Effect problem subtracting infinities results in 1/120. In other words 1 + 2^3 + 3^3 + 4^3 + . . .

**doesn't equal 1/120, it equals 1/120**

*plus infinity*.

It takes a fair amount of math for Dowling to come to that conclusion. And that means we have a choice when faced with a problem that looks like the Casimir Effect - we can do the same calculations he does, or when we see 1 + 2^3 + 3^3 + 4^3 + . . . we can just replace it with 1/120 and go about our business.

That is, provided you make sure to ignore the infinity outside the plates, you can

*pretend*that

1 + 2^3 + 3^3 + 4^3 + . . . = 1/120

**and you'll get the correct answers.**

**Back to -1/12**

The Casimir Effect described here is a three dimensional problem. It's not surprising then that the relevant series involves cubed terms. One-dimensional problems, on the other hand, involve the sum of the natural numbers 1+2+3+4+ . . .

The Numberphiles point out that the relation 1+2+3+4+ . . .= -1/12 (though not true) is useful for string theory. I assume that's that case because strings are one-dimensional. But if you were to set up a Casimir Effect type of problem in one dimension, you'd be able to apply the same sort of calculation that Dowling used in three dimensions, and find that

1+2+3+4+ . . .= -1/12 + infinity

There's no analogous effect in two dimensions, by the way. You can see this because the same sorts of approaches that lead to 1+2+3+4+ . . .=-1/12 and 1 + 2^3 + 3^3 + 4^3 + . . . = 1/120 also produce the relation 1+2^2 +3^2+4^2+ . . . =0.

In other words the vacuum energy inside and outside a 2-d analog of a Casimir Effect-type problem would exactly balance out. Which means there is no Casimir Effect in Flatland (or in 4-d, 6-d, etc.).

**Why Do All Those Other Calculations Leave Out the Infinity?**

In the longer Numberphile video, Ed Copeland explains that the technique he used to calculate 1+2+3+4+ . . .=-1/12 involved mathematically moving around a troublesome point called a singularity, which erases the infinity on the way. That seems believable, but it's hard for me to have an intuitive feel for it.

I suspect the series manipulation they presented in their original video works because the things they use in the calculation have their own infinities snipped out in some way, leading to a self-consistent, but still deceptive, result.

The graphical way I use to find factors associated with the infinite sums works, I think, because it effectively includes both the infinities as you're setting it up.

In the graph here, for example, the stepped lines represent the sum 1+2+3+4+ . . . as you build it up one piece at a time.

1

1+2 = 3

1+2+3 = 6

etc.

This is analogous to adding up the discrete energies in the interior portion of a one-dimensional Casimir Effect problem.

The smooth curved line represents adding up the continuous distribution of frequencies in the vacuum energy of open space. The areas between the curve and the stepped lines are what's left over after you subtract the continuous lined from the stepped line. All of these sections are positive, so they still add up to infinity.

However, if you extend the graph to the left of zero you get something that looks like this.

I've colored portions of the graph that are exactly equal in size. The blue section on the right is the same size as the blue section on the left. And the same is true for the yellow, red, and purple portions. Because they are on opposite sides of the y-axis, the respective colors exactly cancel each other out.

However, there's a little portion between 0 and -1 that doesn't have a matching section to balance it out. I've zoomed in on the graph and colored it green. The area of that region exactly equals -1/12.

All you need to do to calculate the difference between the two infinities is find the size of this little orphaned section.

The same procedure works for 1+2^3+3^3+4^3+ . . . or any other series of the natural numbers raised to a non-negative integer.

To summarize the procedure:

- pick an appropriate series
- find the generating function for the sequence produced by the partial sums of the series
- integrate the continuous version of the generating function from -1 to 0

1 dimension -> 1+2+3+4+ . . . ---> -1/12

3 dimensions -> 1+2^3+3^3+4^3+ . . . ---> 1/120

5 dimensions -> 1+2^5+^5+4^5+ . . . ---> -1/252

7 dimensions -> 1+2^7+3^7+4^7+ . . . ---> 1/240

9 dimensions -> 1+2^9+3^9+4^9+ . . . ---> -1/132

11 dimensions -> 1+2^11+3^11+4^11+ . . . ---> 691/32760

13 dimensions -> 1+2^13+3^13+4^13+ . . . ---> -1/12

*For all even numbers of dimensions, the net difference is zero

(My reasons for extending the graph to negative

*n*and

*x*are a bit sketchy. For one thing, it works. I guess I could also say something about negative energy states and the negative energy sea, but I'm not keen on going there at the moment.)

One of the interesting features of the graph, by the way, is it seems to me to qualitatively explain why you can also solve Casimir Effect problems by simply assuming a cut-off that ignores the higher frequencies. After all, if the portion that accounts for the -1/12 contribution is near the origin, you can choose pretty much any convenient and reasonable cutoff without changing the result, and you don't have to worry at all about subtracting infinities or relying on counter-intuitive infinite sums. You just have to know what error to include as a result of cutting it off though ( that's also pretty easy to figure out from this type graph, but I'll explain it some other time).

Regardless of the way you choose to do the problem, the infinities are accounted for. The Numberphiles chose approaches that get rid of them implicitly (some people would say they get rid of them deceptively). Dowling does the hard math that explicitly deals with the infinities. And I include them in the set up of the problem in a way that lets you account for them without having to do any difficult math. But any way you slice it -

1+2+3+4+ . . . = -1/12

*+ infinity*

**Bonus Nonsense: Fun with Series Manipulations**

Just to annoy the mathematicians, it occurred to me that there's an even quicker way "prove" 1+2+3+4+ . . .= -1/12 (even though it really doesn't) than the Numberphiles used.

If you are willing to accept the fact that 1-2+3-4+5-. . . = 1/4, then let S = 1+2+3+4+ . . .

Multiply S by 4.

4*S =1+8+12 +. . . .

Subtract 4*S from S, but line them up like this

1+2 +3+4+5+ 6. . .

- 4 -8 -12 . . .

= 1-2+3-4+5-6 . . . = 1/4

so

(1-4)S = -3*S = 1/4

or

S = (1/4)(-1/3) = -1/12

***

You can use the above result to find the value of Grandi's series too.

Add (1-4)S, as written above, to (1-4)S, but shift the series like this

1-2+3-4+5-6 . . .

+( 1 -2+3-4+5 . . .)

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

But (1-4)S+(1-4)S = -6S = -6(-1/12) = 1/2

So 1-1+1-1+1-1 . . .= 1/2, as the Numberphiles also show.

I can also "prove" that

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

and

2+3+4+5+ . . .= -7/12

and other similar things, but I've had enough of this stuff for now.

:-) Well done, Buzz Skyline!

ReplyDeleteI think this blog illustrates well to a public beside mathematical experts what the relation between the divergent series and their finite assigned numbers are.

Taken with humour, your "bonus nonsense series manipulations" are great. We can use them to "prove" so many weird things, e.g.:

Assuming that "1+2+3+4+... = -1/12" and "2+3+4+5+... = -7/12", we easily see that their difference is

1+2+3+4+5+...

- (2+3+4+5+...)

= 1

on the left-hand side and -1/12 - (-7/12) = 1/2 on the right-hand side.

That's a nice "proof" for 1 = 1/2 from which you can prove everything you want. ;-)

Bernd, This is a perfect example for that you can prove anything with infinity mathematics.

ReplyDeleteIn my view, infinity does not exist (in the real world), nor does 0.

Infinity = anything / 0

so

0 = anything / infinity.

Physical argument: the smallest thing (measure) is the planck dimension (planck lenght, planck time etc.) so there cannot be an infinite number of since the beginning of time (the big bang).

Although arguments from physics have been used here, these doubtful equations "1+2+3+4+... = -1/12" etc. are not about physics, but about mathematics. And in mathematics, zero is one of the most important numbers. Infinity is not a (real or complex) number, but it can be treated consistently in many ways, e.g. by using proper limits of numbers getting larger and larger without bounds.

DeleteIn mathematics, there is no fundamental limit like the Planck dimension. So we may ask in a well-defined way, what is the result if we sum the terms of a series up to infinity. For a divergent series as discussed here, there is no finite answer, unless you properly regularize it. But you may play this game for convergent series without problems.

And even in physics, the Planck dimension is just the point where gravity as we know it from general relativity breaks down. We just don't know what's beyond. (It's like trying to describe in mathematical terms what the physics inside the horizon of a black hole is. Still, something must be there, even if we fail to describe it with our tools.)

I really want to understand how -1/12 is related to the sum of natural numbers, but I can't for the life of me and these arguments don't help at all.

DeleteMy main concern is that you are talking about the area under the curve x(x+1)/2 and use it to make an argument about the sum of natural numbers. The point is that this area is more than just the sum of natural numbers, its the sum of the sum of every natural numbers plus the values of x(x+1)/2 evaluated at every other real numbers that were not taken into account by the sum of the sum of natural numbers.

Second, these two sums-the area (or the integral) under the curve and the sum of natural numbers-share a common property: both add an infinite amount of elements. True, but the cardinality of both infinities is different:there are more terms added in calculating the integral than in calculating the sum. How both can be compared? The same problem occurs for the area under the curve for x = -1 to 0.

Again, I really don't understand how -1/12, the definite integral of x(x+1)/2 for x goes from -1 to 0, is related to the sum of natural numbers from 0 to infinity since both sums are totally different things.

Sorry, this was a general comment on the article not a reply I. Fabian comment.

DeleteDon't interpret too much into the curve x(x+1)/2. It simply is an interesting feature that this curve is related in some way to both sides of the problem:

DeleteOn the one hand, the partial sums of the series 1+2+3+4+... are given by the value x(x+1)/2 for x = 1, 2, 3, 4, ..., so this polynomial smoothly interpolates between the partial sums.

On the other hand, the finite area between the curve x(x+1)/2 and the x-axis is -1/12. And -1/12 also is the value obtained by several rigorous mathematical methods which regularize the divergent series 1+2+3+4+... and assign a value to it in a consistent way.

Understanding these rigorous methods requires quite some mathematical knowledge, in particular about analytic continuation. So it is nice that the curve x(x+1)/2 offers this simple picture which kind of interpolates between the divergent series and the finite value -1/12.

Talking about natural numbers and physics always make me feel a little bit uncomfortable, because people forget the basics. The whole world we call it physics has only one relation to the numbers we are using every day. This relationship is called - A Model. For example a set of real objects (apples) may be modeled by getting one apple from a basket and its existence is defined by the number 1. So we all know, that ONE apple exists, is picked out etc. Same logic if we get two or more apples. In this example, the number of objects we consider as a set or subset is modeled by the natural numbers. So in this case, the natural numbers represent the real number of objects. Very simple ans looks at first glance worthless to be mentioned. Indeed, if we do not forget this simple technique when talking about more complex calculations. For example - calculating the number of infinite series of number like the sum of all inf. number of naturals must result to one number, that we consider as correct. I.e 1=1 and 1+2=3 and 1+2+3=6 and so on. So if we get back to the example with the apples, we know, that each number we calculate of a series of apples represent the number of apples we can count. Well, does the same logic is applicable to a decimal series? No, because whatever we calculate as a convergence of any series, called also a limit of a sequence is a number, which does not correspond to a real number of countable objects. In the Physics we call that - fuzzy, undefined, a sigma value etc. This junction from a definite to a undefinite like a sequence leads mostly to a error conclusions and calculations. I will stop at this point and also hope, that we do not have to forget the basic concepts we build our knowledge on. Otherwise, the stupidity is an infinite number as Einstein once has been said. Regards to all audience.

ReplyDeleteWell, knowing physicist, they would add a dark constant or variable, just to make their equation right.

ReplyDeleteThey should noted it as Dk+1+2+3+4+5+6....=-1/12, where Dk is dark number, it is whatever it needs to be to make equation right.

In the Bonus Nonsense section:

ReplyDeleteTypo: 4*S = 1 + 8 + 12 + ... Should be 4*S = 4 + 8 + 12 + ...

You wrote "Because they are on opposite sides of the y-axis, the respective colors exactly cancel each other out". Why would areas on opposite sides of the y-axis cancel each other out? I know that areas on the opposite sides of the x-axis would cancel each other out, because one would be a negative area and the other would be positive, but areas to the left of the y-axis and above the x-axis are positive, just like the areas to the right of the y-axis and above the x-axis.

ReplyDeleteI like the way Buzz tries to understand maths. Still I could not see how "the opposite sides of the x-axis would cancel each other out".

DeleteHere is what may bring us closer:

- the positive branch of the hyperbola generates the sums of positive integers up to INF

- the negative branch of the hyperbola generates the sums of negative integers up to -INF

- the logic says that the two sums should except signs equal

- but the integrals of the two branches of the generating function differ by the value of -1/12

- thus the -1/12 can be undestood as a difference between the two infinities in the positive and the negative domains

ps.: This all should be proven, e.g. from the perspective of analytic continuation, but this goes beyond my expertise.

It is great that you write interesting articles like this, but it would be even better if you responded to posts that point out what seem to be errors or problems in your article.

ReplyDeleteI don't get it.

ReplyDeleteWhat's the link between area under the curve between certain limits and the sum of the series within the same limits ?

For example, area under the curve , (1/2)x(x+1) from 1to 10 ( = 765/4= 191.25 sq units ) . In which way is it related to the sum of the numbers 1 to 10 ( which is 55 units of length ) ?

But there seems to exist a connection between arc length and sum of the series

For example,arc length of curve (1/2)x(x+1) from 1to10 is 54.94836 units. Its very nearly equal to the sum from 1 to 10 which is 55 .

Arc length of the same curve between -1 and 0 is =. = 1.0402288 length units .

-[1+ (0.99)+(0.98) + (0.97) + ......+0.01+ 0.00) = -1-(50x99) = -49.51length units

And the area under the curve between -1 and 0 is = -1/12 = 0.0833333 sq units

And, so the question remains.

What's the idea of bringing in area under the curve when we are

talking of summation of numbers between limits which is linked more with arc length ?

But then the coincidence of the sum to infinity with this number -1/12 which is the number we get for the area under the curve also needs to have some kind of explanation too.

Can some one explain these discrepancies, please?