1. Jun 17, 2004

### AKG

In Hilbert's famous paradox of the Grand Hotel, we have a hotel with an infinite number of rooms and an infinite number of guests, and we can create a vacancy by having each guest move over to the next room. However, I don't see how this works. For one, each individual guest moves, and each move by a guest creates a vacancy (when he leaves his room) and then eliminates a vacancy (when he occupies the next room). Each individual move changes the number of vacancies by zero. Why should an infinite number of such moves be any different? The sum of a countably infinite number of zeroes is zero, so how is the vacancy created?

Also, why is it permissible to say that all of those guests who move over actually do find a room (leaving one vacancy) and, there isn't always going to be one guest with no room (even if we can't say he's the "last" guest) but it is not permissible to do the following:

0 = 0 + 0 + 0 + 0 + ...
0 = (1 - 1) + (1 - 1) + (1 - 1) + ...
0 = 1 - 1 + 1 - 1 + 1 - 1 + ...
0 = 1 + (-1) + 1 + (-1) + 1 + (-1) + ...
0 = 1 + (-1 + 1) + (-1 + 1) + (-1 + 1) + ...
0 = 1 + 0 + 0 + 0 + ...
0 = 1 ?

2. Jun 17, 2004

### AKG

The error here is that the associative law cannot be applied freely to infinite sums unless they are absolutely convergent. In fact, it is possible to show that in any field, 0 is not equal to 1.

(Source)

Why is it that we cannot do this with the associative law, but we can essentially do something similar with the guests and rooms. Actually, what does it mean to apply the associative law to an infinite sum? We are applying it to an infinte number of finite sums, are we not? Anyways, assuming someone clears that up, why is that something we cannot do, but we can essentially shift the association of guests and rooms, i.e. if you think of each room as +1, and each guest as -1, then originally we have (+1-1) + (+1-1) + ... and have zero vacancies. Then we shift the brackets over, and get 1 vacancy, as shifting the brackets over is similar to associating a guest with the next room.

3. Jun 17, 2004

### mathman

Hilbert: As you know, every real hotel has only a finite number of rooms. Once you get to infinite, funny things happen. It is no different from saying the number of integers is the same as the number of even integers.

Series: Your second line (1-1+1.....) is not absolutely convergent, so it is not surprising that rearranging terms gives a different answer.

4. Jun 17, 2004

### jcsd

It's an interting problem, but the very fact your moving from sum whose value is a defined, to one whose value is undefined back to one which is defined, eems simlair to divison by zero to me.

5. Jun 17, 2004

### robert Ihnot

If it is difficult to see the vacancy, look at it like this: The hotel manager tells every guest to move from room n to room 2n. Thus 1 goes to 2, 2 goes to 4, 3 goes to 6, and what do you know: We have empty rooms 1, 3, 5,.....2n+1, +++, why we have half the rooms vacant!!!! This way of working is a lot faster than just creating only 1 new vacancy per move!

6. Jun 17, 2004

### AKG

I haven't rearranged the terms at all. Please explain why the associative law cannot be applied to infinite sums that aren't absolutely convergent. Actually that sounds strange to me (it's what was quoted from Wikipedia). Could you explain exactly what's meant by that? And please also explain why we can essentially do an analogous thing with the people and rooms, but not with numbers.

7. Jun 17, 2004

### AKG

Or better yet, prove to me that all guests can find a room after shifting over one. You can say that for all guests originally in room n, there must be a room n+1, but I can say that for every term t_n in the series such that t_n = -1, there is a term t_(n+1) = 1, so they can always cancel out.

8. Jun 17, 2004

### robert Ihnot

As for the sum, so as 0=1; the problem with that is that an alternating series converges to diffent sums depending on how the terms are grouped, as you have shown. If the absolute value of the terms was convergent we have a different matter, but here, of course, the absolute value term by term is infinite. The error here is taking an infinite series, calling its sum 0, the making it an alternating series 1 -1, 1, -1 and then rearranging the terms, well, now the sum is not 0.

9. Jun 17, 2004

### AKG

This is the same "error" Hilbert seems to be making. And you're not saying why it's wrong, simply that it's wrong. Personally, I can't see how it makes any sense that the order in which terms are added should make a difference just because the terms are infinite and the series is not absolutely convergent. As long as you still ultimately add the same terms, why would order matter? But anyways, I'll take it that it does matter without an explanation for now. But then tell me why a reordering does matter, and thus is not allowed with 1 -1 + 1 + ... but is allowed with the guests and rooms.

In the attached image, consider the black dots to be guests, and the squares to be rooms. The red lines show the original associations of rooms to guests, and the blue lines show the associations after the move. Shifting association like this seems similar to shifting the brackets around with the series I presented. And, of course, if anyone can give a reason as to why the associative law cannot be used, or that a non-absolutely-convergent series cannot have the associative law applied to some of the terms, that would be nice.

EDIT: Maybe another way of putting it: we know that if we have a finite number of terms, we can freely associate the terms in the series and perform additions. The same is true with finite rooms/guests, i.e. if we have a 6 room hotel (more like a motel), we can puts guests 1 through 6 in rooms 1 through 6, or put guest 1 in room 2, and mix it up in general. In both cases, we can freely associate terms or room/guests. Now we can't freely associate them with an infinite series that's not absolutely convergent. Along the same lines, what makes Hilbert or you or anyone think they can freely associate guests like that with an infinite number of rooms?

ANOTHER EDIT: Please, also keep in mind the question : if each move made by a guest from one room to another can neither create nor eliminate a vacancy, how can an infinite number of them?

The most sensible thing to me is this: the ordering or association of a series does not matter. 1 + 1 - 1 + 1 - 1 + ..., strange as it sounds, should depend on whether there can be a pairing between the +1's and -1's or not. Essentially, the infinity of terms would have to be either odd or even (yes, it's weird, but no one said we're dealing with non-weird stuff). The arrangment should not matter. If we do not know whether it's even or odd, it's like asking what sin(x) is as x approaches infinity, and let's say x is in the set of all multiples of pi/2. Now if we say that a hotel is occupied, then moving the guests around shouldn't affect the occupied state. If we can actually pair the guests to the room, then it's occupied.

Of course, this isn't perfect, I'd have to think about it some more. But I think it's a little better. If anything, it at least provides us a criteria to determine if there are vacancies or not, whereas the normal approach does not. Whether the hotel is full or "half"-empty, we can always draw a one-to-one correspondence between rooms and guests, and in fact we can always draw a "five-to-one" correspondence too. So what criteria is there?

#### Attached Files:

• ###### hilbert.JPG
File size:
5.4 KB
Views:
228
Last edited: Jun 17, 2004
10. Jun 17, 2004

### master_coda

The reason it's wrong to rearrange the terms is because if you assume that an infinite number of additions is associative, then you can show that 1=0 which is a contradiction. So your assumption that addition is associative in the infinite case must have been wrong.

Anyway, a lot of things break down when you just take the finite case and naively extend it to the infinite case.

11. Jun 17, 2004

### AKG

That's like saying if you rearrange the guests and create one vacancy out of none, you have a contradiction, so you can't rearrange the guests. Again, you're not showing why it's wrong, only that it's wrong and leads to a contradiction, not explaining what's fundamentally wrong with applying the associative law like that. And what is really being done that's considered "association in the infinite case?" I'm applying the associative law to only a finite nubmer of terms (2 terms), simply doing it an infinite number of times. It seems to me that a series that is absolute convergent allows this sort of association, so it's not that we can't use the association law an infinite number of times. It simply leads to a contradiction in certain cases. Again, I think it's contradictory to state that 1=0 is a contradiction, but creating 1 vacancy in a hotel with zero vacancies is not a contradiction.

12. Jun 17, 2004

### Hurkyl

Staff Emeritus
If everybody moves to the next room, does anybody move into the first room?

As for your questions about infinite sums, allow me to stress that there is more involved in infinite sums than addition. If you start with the first number, then add the second number, then the third number, and so on, you will never have added an infinite number of terms; such an approach is simply inadequate.

13. Jun 17, 2004

### master_coda

Creating a vacany out of none isn't a contradiction, unless you assume that a hotel with an infinite number of rooms is supposed to behave exactly like a hotel with a finite number of rooms - there's no reason to make that assumption, so we don't. On the other hand, 1=0 is a contradiction according to any useful definition of numbers.

The reason addition is not associative in the infinite case is because defining it to be associative in an infinite case is because it cannot be done without rendering numbers useless.

14. Jun 17, 2004

### hello3719

Suppose we arrange the sum 1-1+1-1+1...

by grouping them in packs of two we do get (1-1) +(1-1)+...
but this sum yields zero if and only if the number of terms is EVEN, Since this is an infinite sum then there is an infinite of terms. But infinite isn't a number so we can't judge if it is even or not, meaning that the grouping of these terms is inconclusive.

15. Jun 18, 2004

### wisky40

starting with this identity 0=0, then 0=0+0+0+0+.........

it's true that when every 0 is broken into (1-1), the alternating ones (positives & negatives) are even. I think AKG forgot one (-1) of the last pair.

I can work it out in this way also:
0= n0
0= n(1-1)
0= (n-1+1)(1-1)
0= (n-1)(1-1)+(1-1)
0=1+(n-1)(1-1)-1
0=1+0+0+0+...............................-1=1-1=0 AKG is not showing (-1) from the last pair.

16. Jun 18, 2004

### matt grime

I have yet to see any compelling reason from AKG why the 1=0 sum paradox is equivalent to the Hilbert hotel. The best so far is that they 'seem' the same.

Let us prove that there is not problem in the Hilbert hotel:

let S be the set of people who do not find a new room after rearrangement. If S is non;empty it has a least element, s, say. However by COnstruction s was asked to move to room s+1 # so S is empty.

17. Jun 18, 2004

### jcsd

But by definition there is no 'last pair'.

18. Jun 18, 2004

### AKG

Let $v_n$ represent the number of vacancies created by the room change made by the $n^{th}$ guest. A guest cannot enter an occupied room. No more than one guest can be in a room at any one time. If a guest is in a room, and moves to another room, then the room he/she was in becomes vacant by his or her leaving, and the vacant room he or she enters becomes occupied. $\forall n \in \mathbb{N},\ v_n = 0$. Now, we move all guests, so the number of vacancies created is:

$$\sum _{n=1} ^{\infty} v_n = \sum _{n=1} ^{\infty} 0 = 0$$

19. Jun 18, 2004

### matt grime

That sum is not allowed or rather it won't be if only finitely many of the entries are zer, as you must want in order to do the paradoxical thing., so your model does not hold. Algebric sums must be finite, or they are formal series that do not represent a natural number. Try again.

Last edited: Jun 18, 2004
20. Jun 18, 2004

### AKG

What's not allowed? Formal series do not represent a natural number? Why is that? What do they represent? Surely they can represent a real number. Now would you suggest then that the 1 of the reals is not the zero of the 1 of the naturals?