Talk:Exponentiation/Archive 2010
![]() | This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page. |
Archive 2005 | ← | Archive 2008 | Archive 2009 | Archive 2010 |
In the axioms for the real numbers, is the operation of raising a negative numbers to fraction exponents really defined? Ever?
If we want the law (x^a)^b = x^(ab) to hold for all real numbers where both sides of the equality are defined, then it is logically inconsistent to say that (-1)^(2/3) = cube root of (-1)^2 = 1. Let x = -1, a = 2/3, b = 3/2. The left side the equality would be 1 and the right side would be (-1)^1 = -1.
In my opinion, rather than add special conditions to that law, it is better say that (-1)^(2/3) is not defined. This doesn't prevent a person from making up their own definition of the operation. But this operation is not "exponentiation". It is helpful to tell a reader than the definition of a fractional exponent suggests a value for (-1)^(2/3) and some calculators and software may give an answer. But the distinction should be explained between the definition of exponent (which obeys (x^a)^b = x^(ab) ) and "homemade" definitions of operations, which may not obey the laws of exponents.
start SA-------
Another opinion:
Is y = x^(1/3) the same as y^3 = x? For what values of x is x^(1/3) defined? It seems nonstandard to say that the graph of y = x^(1/3) is only "half" the graph of y^3 = x. And how are we to compare those two graphs to the graph of the cube root function? The rule (x^a)^b = x^(ab) is commonly understood only for x nonnegative, and I see no reason not simply to say that, and point out that it can fail for negative values of x.
Also, what are we to say about (x^a)^b = x^(ab) when x = -1, a = 2 and b = 1/2? Are we going to disallow (-1)^2 = 1? Or are we going to follow the standard approach,and indicate that the formula (x^a)^b = x^(ab) can sometimes fail for negative x?
97.127.4.156 (talk) 03:27, 2 May 2010 (UTC)Scot Adams
end SA--------
Tashiro (talk) 23:12, 14 July 2009 (UTC)
- It's a matter of context, a lot like the famigerated "zero to the zero" thing. In a strict real-analysis context, where (i) you're excluding complex numbers and (ii) you don't really care about algebraic properties, you're right: The most convenient domain for real-to-real exponentiation is that the base must be positive. In this context, not even the apparently unproblematic expression 0.01.0 is, strictly speaking, defined; the base must be strictly greater than zero. --Trovatore (talk) 01:34, 15 July 2009 (UTC)
- What is wrong with 0.01.0? 203.51.17.5 (talk) 09:19, 3 October 2009 (UTC)
- xy is defined for reals by ey ln x. But the log of 0 is not a real number. The definition can be extended to include −∞ but otherwise, as he says, 0.01.0 is not strictly speaking defined. Dmcq (talk) 09:41, 3 October 2009 (UTC)
- I see. Thankyou. 203.51.17.5 (talk) 12:32, 3 October 2009 (UTC)
- xy is defined for reals by ey ln x. But the log of 0 is not a real number. The definition can be extended to include −∞ but otherwise, as he says, 0.01.0 is not strictly speaking defined. Dmcq (talk) 09:41, 3 October 2009 (UTC)
- What is wrong with 0.01.0? 203.51.17.5 (talk) 09:19, 3 October 2009 (UTC)
I agree with that, so I think the article should be improved to make the context of its statements clear. One one hand it treats the topic "rational powers" under the heading (above it) "real powers of positive numbers". Later, in the section on "Powers of negative real numbers" it states that (-1)^(m/n) = 1 if m is even. This would invalidate the law (x^a)^b = x^(ab) if we take it in the context of the real numbers. One may define some kind of "extended" exponentiation operation where (-1)^(2/3) = 1, but this extension does not obey the same laws as the usual laws of exponents for real numbers. —Preceding unsigned comment added by Tashiro (talk • contribs) 04:44, 15 July 2009 (UTC)
- You have a good point about the section 'Powers of negative real numbers'. Basically the law (x^a)^b = x^(ab) doesn't hold for them along with a lot of other things that are wrong with them. A good example might help plus even more warning. I don't see that the structure of the article otherwise need change though, what exactly were you saying was wrong with the context? Dmcq (talk) 08:44, 15 July 2009 (UTC)
- I think the main thing you are pointing out is that the formula "(x^a)^b = x^(ab)" does not hold for all situations where both sides of the equation are defined, even when x, a, and b are real numbers. Unfortunately, most people learn that identity at a very young age, when they are not taught the hypotheses necessary for the identity to hold, and then they never learn those hypotheses later. I agree the article could be more clear about these hypotheses when it mentions the identity. — Carl (CBM · talk) 12:13, 15 July 2009 (UTC)
- The main problem seems to be that when one allows negative reals then square root has two possible solutions, and saying the cube root of −1 = −1 is saying the output of exponentiation of reals can include negative numbers. The previous sections only treated exponentiation on positive reals producing positive reals. This section would have to include a bit about the possibility of the square root of 1 being either -1 or 1. Dmcq (talk) 18:14, 16 July 2009 (UTC)
- I've had a go at rewriting the section, not an acme of exposition but at least a basis to work from I think. Dmcq (talk) 08:48, 17 July 2009 (UTC)
- The main problem seems to be that when one allows negative reals then square root has two possible solutions, and saying the cube root of −1 = −1 is saying the output of exponentiation of reals can include negative numbers. The previous sections only treated exponentiation on positive reals producing positive reals. This section would have to include a bit about the possibility of the square root of 1 being either -1 or 1. Dmcq (talk) 18:14, 16 July 2009 (UTC)
Note the useful convention used in the J (programming language). Powers of negative numbers are evaluated to be complex numbers.
(-1)^(1%2) 0j1 (-1)^(1%3) 0.5j0.866 (-8)^(1%3) 1j1.73
- I've put a bit at the start of the powers of positive real numbers about this problem. It explains why the section is limited to positive real numbers by referring to this problem. Dmcq (talk) 11:23, 2 May 2010 (UTC)
"Multiplied by itself n times"
The phrase is uniquitous, yet, it is very confusing in case of n = 0. Literally, this means that no exponent exists for n=0. The result is not even a number. Maybe we should return zero? This is not specified anywhere. The "a^n = a · a · ... n times" formula used in definition suffers the same problem. Can you explain why it claims that n must be positive? Why it then admits the possibility to extend to negatives but forgets to discuss zero? To resove the problem, I would like to put it clear that the 'zero number' of multiplication is the multiplicative identity 1. That is, the correct formulation should be: a^n = 1 · a · a ·. Once we know that, the prase "multiplied by itself n times" can be used as abbreviation. --Javalenok (talk) 20:42, 9 October 2009 (UTC)
- There is a section on this 'Exponents one and zero'. If people are worried by this it is easy to find. I think multiplicative identities are just a bit too high falutin' for the level the start of this article is aimed at. Have you found some school text or suchlike that does what you say? Dmcq (talk) 21:14, 9 October 2009 (UTC)
- The phrase "multiplied by itself n times" only occurs in one section of this article: the section on abstract algebra, where it is clearly stated that the definition is being given for nonzero expoenents that that moment. The notation
- "When n is a positive integer, exponentiation corresponds to repeated multiplication:
- just as multiplication by a positive integer corresponds to repeated addition:
- "
- "When n is a positive integer, exponentiation corresponds to repeated multiplication:
- There, it is only claimed that the correspondence is valid when n is a positive integer; this does not claim that exponentiation would be invalid as well, only that the correspondence being described would fail. And that is the lede, which is supposed to be a summary and introduction. The entire remainder of the article servers to explain the things mentioned in the lede.
- I do not see any other examples of "multiplied by itself n times" in the article. So, I don't think that the article suffers from the problem you describe. — Carl (CBM · talk) 21:30, 9 October 2009 (UTC)
- Actually "multiplied by itself n times" is just plain wrong; computing xn requires only multiplications. Corrected now. Marc van Leeuwen (talk) 13:40, 18 March 2010 (UTC)
Rational Exponents
Hello all, I was tutoring a MTH-095 student at the college a few days back and found an erroneous claim in the text, namely that . Using m=2, n=4 I constructed a counterexample, furthermore, if one reduces 2/4 to 1/2 (and then converts to a square root), one will get the same answer as the answer. I noticed the Rational Exponentiation subsections makes no mention of anything like this, I'm not sure if holds in general over . If so, might I suggest this article include that, furthermore this section seems in need of some expansion. The note made in parenthesis is a bit unjustified, it makes sense if one is interpreting the root as a function on real numbers however this isn't necessarily how one would interpret . A math-wiki (talk) 01:28, 6 November 2009 (UTC)
- Could you share your counterexample? Most importantly, is the base a positive real number, as the article says it should be for the identity to hold? At that point in the article, complex number has not been mentioned, and that section is about powers of positive real numbers. Complex exponentiation does not satisfy that identity, if you use principal values to compute it. Also, there is a paragraph in the article starting with "Care needs to be taken when applying the power law identities with negative n-th roots." which is very relevant. This makes me think that you may not have read the entire article. — Carl (CBM · talk) 04:58, 6 November 2009 (UTC)
What's in the article is mathematically correct, though the part in parenthesis is a bit unjustified. I'm just suggesting their should be more about rational exponents. My counterexample was to the erroneous claim in the MTH-095 text, with a=-1, m=2, n=4. Furthermore I'm curious if holds over , as it works when compared to what happens when is reduced to and then converted to . A math-wiki (talk) 06:06, 6 November 2009 (UTC)
- Yes that's true in general where one takes the rational root by starting with a=reiθ with θ in (-π,π] because after taking the mth root θ/m will still be in that range and one doesn't jump to a different branch. Dmcq (talk) 10:49, 6 November 2009 (UTC)
- I apologize; I thought you meant the error was in the text here, instead of in the textbook. Could you explain what extra information about rational exponents you would like to see (even in general terms)? — Carl (CBM · talk) 12:28, 6 November 2009 (UTC)
- Maybe a link to the http://wiki.riteme.site/wiki/Root_of_unity article?
- JWhiteheadcc (talk) 20:39, 21 December 2009 (UTC)
The formula is valid assuming m is odd or assuming n is odd. When both are even, it may fail. This is noted in the article (where m and n must be rel. prime). 216.36.100.202 (talk) 12:51, 2 May 2010 (UTC)Scot Adams
- The real number and the complex number versions give different results, and they both can give different results from what might be considered as the nth root of unity. You have to know exactly what version of nth root you mean. Dmcq (talk) 15:36, 2 May 2010 (UTC)
VBA program code for calculating powers
Excel does not show the answer to a powers question if the answer is really big. It uses scientific notation instead. I made vba which, if without bugs, makes a long text of a powers answer. It is assumed that the power is a 2-digit number.
It breaks something like 123456789 into 1 23 45 67 89 , does a small multiplication with a 2-digit number from right to left, and noting the carrier from each calculation.
http://www.nakedscience.com/articles/math/Powers%20VBA%20Function%20for%202%20digit%20powers%20of%20small%20numbers%20in%20mathematics.html
links to:
http://www.nakedscience.com/articles/math/powers.htm
and points to a 7zip file of answers up to 99^99 (if my vba is correct!)
Mickipedia2 (talk) 14:10, 13 March 2010 (UTC)Mike
The option: 00=0
This formula is provable as follows:
Every integer x satisfies: 0x = 0x-1+1 = 0x-1 * 01 = 0x-1 * 0 = 0.
There are to ways for denying this proof: a positive way and a negative way:
The negative way is to use the formula a-1=1/a, or the formula ax-y=ax/ay, for claiming that the term 0-1 is meaningless, just as the term 1/0 is meaningless. However, this way of denying the formula 00=0 is wrong, beacuse the very formula a-1=1/a, is just a formula - rather than a definition, and it's not provable when a=0, just as the formula ax-y=ax/ay is not provable when a=0 and x<y.
The positive way for denying the proof (mentioned above) of the formula 00=0, is by claiming that the definition a^(x+1)=(a^x)*a is not universal, or that the function 0^x is defined for positive x only. However, for claiming that, one must prove that the definition a^(x+1)=(a^x)*a, or the function 0^x, can't be applied for all integers, but factually, nobody can prove that! On the other hand, nobody have to prove that the definition a^(x+1)=(a^x)*a is universal, nor do they have to prove that the function 0^x is defined for every integer x, because the default requires to apply every mathematical definition and every arithmetical operation - to every number, unless such an application entails a contradiction.
To sum up:
- The difference between the formula a^(x+1)=(a^x)*a (which proves that 00=0), and the other formulas mentioned above (which prove that 00 can't be 0), is that the the formula a^(x+1)=(a^x)*a is a definition: Nobody have to prove a definition, while they do have to prove formulas which are not definitions.
- The formula 00=0 is well founded, whereas denying it needs a proof/source, but no such a proof/source exists.
Eliko (talk) 00:41, 28 March 2010 (UTC)
- Have you a citation saying what you have said? If not it is WP:Original research. And it isn't even halfway interesting that I can see. Wikipedia is not a repository of junk facts, see WP:What Wikipedia is not Dmcq (talk) 00:54, 28 March 2010 (UTC)
- Stop calling other editors' edits "junk". It's not a proper way for treating arguments. Not every arguments must have s source. If it's a mathematical conclusion then no source is needed. HOOTmag (talk) 01:13, 28 March 2010 (UTC)
- Where are these guidelines on the treatment of arguments and exactly what is the PC way of referring to proposed content without any apparent value? By the way you do need citations for mathematical conclusions, only straightforward routine calculations do not need justification. Dmcq (talk) 01:26, 28 March 2010 (UTC)
- Stop calling other editors' edits "junk". It's not a proper way for treating arguments. Not every arguments must have s source. If it's a mathematical conclusion then no source is needed. HOOTmag (talk) 01:13, 28 March 2010 (UTC)
- Which claim do you need a source for? Is it my claim that the formula a^(x+1)=(a^x)*a entails: 0x = 0x-1+1 = 0x-1 * 01 = 0x-1 * 0 = 0? Or is it the claim that the formula a^(x+1)=(a^x)*a is just a definition? Or is it the claim that nobody can prove that the formula a-1=1/a is valid when a=0? which claim are you referring to? Eliko (talk) 01:53, 28 March 2010 (UTC)
- Have you anything saying your result 00=0? It cannot be proved, it would have to be defined as such. And nobody seems to have any inclination to define it as such. If you think you have a proof it is original research. Dmcq (talk) 02:02, 28 March 2010 (UTC)
- Why do you say that "It cannot be proved, it would have to be defined as such", after I showed in the article (footnote no. 16) that the formula 00=0 is not a definition, but rather is a direct conclusion provable from the definition a^(x+1)=(a^x)*a for exponentiations with integers? Don't you agree that the definition a^(x+1)=(a^x)*a entails: 0x = 0x-1+1 = 0x-1 * 01 = 0x-1 * 0 = 0? Don't you agree that the formula a^(x+1)=(a^x)*a is a definition for exponentiations with integers? Don't you agree that a definition of operation between numbers - should be applied for as many numbers as possible, unless such an application entails a contradiction? Generally, simple obvious conclusions derived directly from definitions are not an "original research". Do you think that the formula 23*99=2201 is an original research? Anyways, the very proof is displayed in the article (footnote no. 16), so the proof can serve as a source, and it's even better than any other source. Eliko (talk) 02:19, 28 March 2010 (UTC)
Eliko, your calculation is flawed in the case x = 0, because in that case x-1 is -1 and so your calculation involves 0-1 which is not defined. As a similar argument, someone could try to say that
using the fact that 0 is definitely equal to 03, in order to try to prove that 0-1 is 0. This has the same sort of problem as your derivation: because 0-2 is not defined, the derivation does not actually find a value for 0-1. If your derivation worked, your method would thus also imply 0-1 = 0, which is not true. — Carl (CBM · talk) 03:33, 28 March 2010 (UTC)
- WP:No_original_research#Routine_calculations gives details about what routine calculations are. Youhave no consensus that you are doing a routine calculation. In fact people can and do disagree with doing a routine calculation for a person's age or date of birth which is something the routing calculation says could be okay. Wikipedia is the wrong place to put proofs which are not already in books or journals. Editors on Wikipedia are not recognized for peer reviewing works. Only results that have been checked elsewhere and can be cited are acceptable. Dmcq (talk) 07:15, 28 March 2010 (UTC)
- By the way what I was classifying as a junk fact was that no calculator returned 0 when 00 was input, not the erroneous proof above. This is an actual fact but there is no citation, it is just some random idea that occurred one day to some editor of Wikipedia. Dmcq (talk) 07:29, 28 March 2010 (UTC)
- I agree with CBM and Dmcq. This addition is not a "routine calculation" - it is unsourced original research and has no place in the article. I have removed it again. Gandalf61 (talk) 08:38, 28 March 2010 (UTC)
- I agree as well. The "proof" begins with the nonstandard axiom that 0n is defined for all integer n, even negative n. Absent a reputable source that uses this definition, this is OR. — Steven G. Johnson (talk) 15:37, 28 March 2010 (UTC)
0⁰ — limits and continuity
Returning to the topic with neverending fascination: I thought of changing the sentence
- In most settings not involving continuity (for instance, those in which only integral exponents are considered), interpreting 00 as 1 simplifies formulas…
to
- In most settings not involving limits, interpreting 00 as 1 simplifies formulas…
and — predictably enough — such a bold change was reverted. Discuss? I remember we had a discussion when we worked out the section to stand as it does now, a long time ago, but I don't recall this specific wording being discussed. (Apologies if it has already been.)
I'd argue that "a setting that involves continuity" also implicitly involves a limit, but the converse is not necessarily true. As the article says, "00 must be handled as an indeterminate form when it is an algebraic expression obtained in the context of determining limits" — and otherwise can be consistently, with no trouble, defined as 1. And defining 00 to be 1 does simply formulae even when continuity is involved; indeed the last few (and especially the last) of the given examples where such a definition helps:
*It greatly simplifies the theory of polynomials and power series that a constant term can be written ax0 for an arbitrary x. For example:
- The formula for the coefficients of a product of polynomials would lose much of its simplicity if constant terms had to be treated specially.
- Identities like and are not valid for x = 0 unless 00 = 1.
- The binomial theorem is not valid for x = 0 unless 00 = 1.
- In differential calculus, the power rule is not valid for n = 1 at x = 0 unless 00 = 1.
most certainly "involve continuity". It seems quite obvious to me that the only argument against defining 00=1 is one involving limits and indeterminate forms; the functions xy and 0y are discontinuous at y=0 no matter what, and the value of 00 is irrelevant to their (dis)continuity. Shreevatsa (talk) 20:05, 16 June 2010 (UTC)
- Notice that in at least the first two examples, the exponent is an integer, not a real number. The third example is less clear, but given the problematic nature of defining negative x raised to the power of fractional n, is arguably the same. The real numbers are inherently about continuity, which is why it makes sense to consider the real number zero, raised to the real number zero, to be undefined. The real number zero raised to the power of the integer zero is another matter, but then there is no issue of continuity in the exponent. --Trovatore (talk) 20:18, 16 June 2010 (UTC)
- Whoops, I miscounted -- the different levels of bulletting confused me. By "first two" I meant the ones starting with "[i]dentities" and "the binomial theorem"; the "third example" is the one about the power rule. --Trovatore (talk) 20:27, 16 June 2010 (UTC)
- I think there's a vagueness of thought here, anticipated by Knuth's joke about 00 being less strongly defined than 0+0. :-) Without going into whether ideas like "the real numbers are inherently about continuity" are a blanket excuse for thinking of limits even if they have no bearing, I'll stick to what the wording should be. The phrase "In most settings not involving continuity [, 00=1] simplifies formulas…" suggests that it simplifies formulae only in the case where everything is discrete, which several of the examples are not. You're talking only of continuity in the exponent, and even there only when taking limits. I'm suggesting that the sentence at top should reflect what the examples given actually show, that it simplifies the situation everywhere (or at least "most settings") where there are no limits. [Here's another situation that does involve continuity where the definition helps: the function x0, which is unnecessarily
discontinuitynot continuous unless you define 00=1. :-)] Shreevatsa (talk) 21:58, 16 June 2010 (UTC)- Well, it doesn't have a discontinuity; it has a point at which it's not defined, which is different. It's continuous on the entire domain of definition. But even there there's no problem unless you're insisting on the exponent being real, as opposed to integer; it's unclear under what circumstances that would naturally arise, with the exponent being constant. --Trovatore (talk) 22:23, 16 June 2010 (UTC)
- You replied only to the thing within brackets, which was not the main point. I've struck it out, and await further clarification of ideas. Shreevatsa (talk) 23:51, 16 June 2010 (UTC)
- Well, it doesn't have a discontinuity; it has a point at which it's not defined, which is different. It's continuous on the entire domain of definition. But even there there's no problem unless you're insisting on the exponent being real, as opposed to integer; it's unclear under what circumstances that would naturally arise, with the exponent being constant. --Trovatore (talk) 22:23, 16 June 2010 (UTC)
- I would say that it is indeterminate at 0 but the limit value is 1. I would reserve the defining as 1 for the straightforward cases where there are no variables involved. Dmcq (talk) 00:07, 17 June 2010 (UTC)
- There are no variables involved in "00", are there? Shreevatsa (talk) 00:22, 17 June 2010 (UTC)
- I would say that it is indeterminate at 0 but the limit value is 1. I would reserve the defining as 1 for the straightforward cases where there are no variables involved. Dmcq (talk) 00:07, 17 June 2010 (UTC)
- Let's put it this way: The important thing to get across is not so much what "simplifies formulas", as what is generally done. In practice, many workers working in the context of continuous quantities (real and complex numbers) use definitions that fail to denote at the point (0,0). Perhaps the most common of these is the definition , where of course the exponentiation on the right-hand side does not mean the same as the exponentiation on the left-hand side. "Simplifies formulas" was a convenient expedient for motivating the choice 00=1 in other contexts than these. Possibly it was not the best wording for that. But the key point is that 00=1 is not standard, or at least not fully standard, in continuous (real and complex) contexts. --Trovatore (talk) 00:16, 17 June 2010 (UTC)
- I don't see why we cannot get across both. The fact that in practice, many people fail to define 00, as well as the fact that defining it to be 1 is perfectly fine everywhere except that you cannot substitute values when you're computing a limit. Shreevatsa (talk) 00:22, 17 June 2010 (UTC)
- Well, no, it actually isn't. But even if it were, that's not our call. We just report what is done. --Trovatore (talk) 00:31, 17 June 2010 (UTC)
- Can you explain "Well, no, it actually isn't"? Shreevatsa (talk) 00:43, 17 June 2010 (UTC)
- Later. I'm at work now. --Trovatore (talk) 00:47, 17 June 2010 (UTC)
- OK, home now. Here's the gist of it: The operations on real numbers are not defined in the same way as the operations on integers, and their definitions have no special cases for when a real happens to be "an integer" (more precisely, the image of an integer under the natural embedding from the integers to the reals). For example, take addition. How do you define addition on the natural numbers? Well, it's inductive, right? 0+0=0, n+(m+1)=(n+m)+1, (n+1)+m=(n+m)+1. Unless I've missed something, those three clauses suffice.
- Extension to the integers is easy and I won't go through it.
- When you come to the rationals, you have to do a little more, but still pretty simple (p/q+r/s = (ps+rq)/qs). Note that there is no special case here when you want to add, say, 0/1+0/1 — you are still using the new definition, not the one from the naturals or the integers. The meaning is a little different, though not much different.
- When we come to the reals, though, the meaning has changed completely. The reals are not a generalization of the naturals or the integers or the rationals; they're a completely new ball of wax. All of a sudden they model topological intuitions, ones about continuity, whereas the previous structures were entirely algebraic.
- What is a real number? One standard coding is that it's an equivalence class of Cauchy sequences of rational numbers. If you prefer Dedekind cuts, make changes as necessary.
- How do you add two real numbers? Well, you take two Cauchy sequences from the equivalence classes, and add the rationals in them pointwise, using the previous definition of addition on the rationals. You have to prove that the resulting sequence is Cauchy, and that it doesn't matter which representative you pick from each equivalence class.
- How do you add the real number zero to the real number zero? Do you just add the natural numbers zero? Absolutely not!!! You do exactly the same thing you do to add any other two real numbers: You pick representatives from the equivalence class of Cauchy sequences, add them pointwise, and take the equivalence class of that. Turns out that that's again the real number zero. But this is a theorem — it is not completely trivial; it is not essential to this being the "right" definition of addition on the reals.
- Now when we get to exponentiation, the pointwise definition is much harder to make work, partly because a rational raised to a rational power need not be rational (nor indeed even real). With sufficient effort you can do it, but it's much cleaner to just chuck it, and define a completely new function . The domain of definition of this function is x>0.
- Talk pages are not really the place to argue about this (look it up on sci.math if you want to see people arguing endlessly) but because you asked me specifically, I thought it was fair to explain why I see it this way. I am not interested in arguing about it further (most especially with Bo Jacoby); I will insist that the article not in any way deprecate the standard definition on the reals (though certainly it may and indeed must present contrary views, such as Knuth's). --Trovatore (talk) 08:27, 17 June 2010 (UTC)
- Later. I'm at work now. --Trovatore (talk) 00:47, 17 June 2010 (UTC)
- Can you explain "Well, no, it actually isn't"? Shreevatsa (talk) 00:43, 17 June 2010 (UTC)
- Well, no, it actually isn't. But even if it were, that's not our call. We just report what is done. --Trovatore (talk) 00:31, 17 June 2010 (UTC)
- I don't see why we cannot get across both. The fact that in practice, many people fail to define 00, as well as the fact that defining it to be 1 is perfectly fine everywhere except that you cannot substitute values when you're computing a limit. Shreevatsa (talk) 00:22, 17 June 2010 (UTC)
I fully support Shreevatsa's point of view. Note that Trovatore's first argument, the exponent is an integer, not a real number, is wrong, because integers are real numbers. Trovatore's second argument, the key point is that 00=1 is not standard, or at least not fully standard, is not true. A host of formulas in WP use 00=1 as a standard. The fact than some people do not define 00 should not blur the fact that 00 is defined by somebody else. This is analogous to the situation that some people do not define binomial coefficients like or because these are not interpreted as choose functions, but this does not prevent it to be defined in WP without discussion. It goes without saying that some people do not define something. Bo Jacoby (talk) 07:14, 17 June 2010 (UTC).
- Bo, you're just wrong, like you always have been. It is not standard when talking about the reals. The usual definition of exponentiation on the reals is . You can find this in innumerable references. --Trovatore (talk) 08:07, 17 June 2010 (UTC)
I fully agree with Trovatore. x0 is not defined for reals at 0, it has a limit of 1 there, that is proven not a definition, and the function with that value patched gives a continuous function on the reals. 0x is not defined at 0, it has a limit from the positive side of 0 there, that is proven not a definition, and the function with that value patched in gives a continuous function on the non-negative reals. The continuous functions are not the original functions, they are patched versions of the original functions. The original functions were restrictions of xy and that is discontinuous around 00 and is not defined there. The integers are a different matter and defining 00 as 1 is both reasonable and generally so useful that a person would have to explain if they wanted to leave it undefined. Integers are not a restricted set of real numbers any more than the real numbers are a restricted set of complex numbers. Dmcq (talk) 09:59, 17 June 2010 (UTC)
- Trovatore, you say: It is not standard when talking about the reals. Do you mean that 00 is defined only for integer values of the exponent 0? Do you mean that 00 = 1 while 00.0 is undefined? This implies that 0 ≠ 0.0 . I am not the only one to object against that. The usual definition of exponentiation on the reals, , applies for only. For other values of you may use other definitions. The elementary definition of for integer implies that for all values of including (The empty product contains no factors, and so its value does not depend on which factor it doesn't contain). Dmcq, what do you mean by a restricted set? The set of integer numbers is a subset of the set of real numbers, just as the set of real numbers is a subset of the set of complex numbers. Leaving 00 undefined solves no problem because no matter whether 00 is defined or not. Bo Jacoby (talk) 15:36, 17 June 2010 (UTC).
- I do in fact mean that 00 = 1 while 00.0 is undefined. Now one way you could put this is that 0 ≠ 0.0, and this is in fact a consequence of a (perhaps rather literal-minded) interpretation of the standard set-theoretic analysis.
- However, you don't have to put it that way if it bothers you; maybe a less provocative way to express it is that the two notations mean different things. They notate different functions, both called "the exponential function". It's like polymorphism in C or C++ — there are multiple functions called "pow", disambiguated by the types of the inputs. If you don't want to allow zero to be typed, then fine, for us it's disambiguated by context rather than by anything strictly mathematical. --Trovatore (talk) 17:17, 17 June 2010 (UTC)
- The whole mechanism of Indeterminate forms simply will not work if an indeterminate form is given a value. If you try coding it up for a computer the same way one can code up differentiation you find you are stuck with forms that are indeterminate but have a value and there is no simple way out to code something that actually works except to define extra functions that don't have a value which are different from the ones which do and use the ones that don't have a value in the workings. Defining 0.00.0 as 1 breaks too much standard calculus, it just isn't a good thing to do. As the article on indeterminate forms says "Limits involving algebraic operations are often performed by replacing subexpressions by their limits; if the expression obtained after this substitution does not give enough information to determine the original limit, it is known as an indeterminate form." If 00 was 1 then the value would give one enough information to know what the value was and therefore one wouldn't use another method to find the value - and you'd end up with the wrong limit. Instead of just doing normal operations with exponentiation one would have to treat them as black box functions and track everything that was done with them very carefully. It just isn't worth the bother.
- And integers are not a subset of the reals any more than reals are a subset of the complex numbers. Most of the operations carry over from one to another but there are differences. 5+0i ≥ 3+0i is simply a wrong statement as ≥ is an inapplicable operation with the complex numbers. Personally I have a problem with saying (–2.0)3.0=–8.0 for instance. I would agree that to the power of integer 3 it is okay but to the power of the real number 3.0 it simply is undefined. Dmcq (talk) 17:43, 17 June 2010 (UTC)
This has been a useful discussion. To return to the original topic, and in light of the examples we have, I propose changing the wording to more correctly incorporate the understanding, and mention continuity in the exponent. This would also make it shorter and avoid the need for the parenthetical bit. That is, I propose changing
- In most settings not involving continuity (for instance, those in which only integral exponents are considered), interpreting 00 as 1 simplifies formulas…
to
- In most settings not involving continuity in the exponent, interpreting 00 as 1 simplifies formulas…
Given that some of the examples arguably do involve continuity in the base (especially the one about differentiation!), this seems to be an improvement and clarify matters. As it stands, even the "for instance" is incorrect, since considering only integral exponents does not mean a setting not involving continuity. Shreevatsa (talk) 20:48, 17 June 2010 (UTC)
- This is a constructive proposal that I can support. --Trovatore (talk) 21:31, 17 June 2010 (UTC)
- Great! I've made the change. I also added an interesting paper I found through the indeterminate form article, and another I found while searching... apparently many exercises of the 00 form in calculus textbooks have the answer 1, and there is a reason for that. Shreevatsa (talk) 22:57, 17 June 2010 (UTC)
- Kahan who was the chief architect of the IEEE floating point standard decided pow(0,0) should be 1 early on as the value people would most often want, a kind of engineers rather than mathematicians definition. A particular reason for this was they didn't have a separate integer power function. The 2008 standard has recommended this value and has a rationale which can be applied to any function which in the limit tends to a value in all but a vanishingly small part of a ball round a singularity. My though was it showed intelligent people are better at rationalizing. :) Dmcq (talk) 23:38, 17 June 2010 (UTC)
- To be fair, according to my man page, the rule for pow(x,y) is "If y is 0, the result is 1.0 (even if x is a NaN)." So it's not as if 00 is a special case; they define −π0 to be 0 as well, even though −πy is not defined in any neighborhood of 0. Because, "If x is a finite value less than 0, and y is a finite non-integer, a domain error occurs, and a NaN is returned." So watch out for rounding errors... — Carl (CBM · talk) 02:46, 18 June 2010 (UTC)
- Wow — is there not a
pow(int,int)
? There really ought to be. --Trovatore (talk) 03:10, 18 June 2010 (UTC)- There can't be, really. There is no equivalent of NaN for ints; every operation must succeed (apart from division by 0). But the numbers 2 and 10 000 are representable, and no implementation I know of has an int that can hold the value 210 000. So pow(int,int) would be crippled, and even the domain for which it's defined would be implementation-dependent. However, libraries for arbitrary-precision arithmetic do include integer power functions. — Carl (CBM · talk) 03:34, 18 June 2010 (UTC)
- Well, you could make that same argument about multiplication. I think on a 64-bit machine, if you multiply two unsigneds, you get the product mod 2^64, right? Or is that not defined by the standard? Anyway, you could do the same thing for exponentiation. --Trovatore (talk) 06:35, 18 June 2010 (UTC)
- Yes, the operations "overflow" in some way (which is complicated for signed integers). But for unsigned (nonnegative) integers on a 32-bit system the domain where the pow function doesn't overflow is a tiny triangle from (0,0) to (2,32) to (65 536, 2). The C language was originally implemented in the late 1960s on a 16-bi PDP-11, where the useful domain would have been even smaller. — Carl (CBM · talk) 11:07, 18 June 2010 (UTC)
- Well, you could make that same argument about multiplication. I think on a 64-bit machine, if you multiply two unsigneds, you get the product mod 2^64, right? Or is that not defined by the standard? Anyway, you could do the same thing for exponentiation. --Trovatore (talk) 06:35, 18 June 2010 (UTC)
- There can't be, really. There is no equivalent of NaN for ints; every operation must succeed (apart from division by 0). But the numbers 2 and 10 000 are representable, and no implementation I know of has an int that can hold the value 210 000. So pow(int,int) would be crippled, and even the domain for which it's defined would be implementation-dependent. However, libraries for arbitrary-precision arithmetic do include integer power functions. — Carl (CBM · talk) 03:34, 18 June 2010 (UTC)
- Wow — is there not a
- To be fair, according to my man page, the rule for pow(x,y) is "If y is 0, the result is 1.0 (even if x is a NaN)." So it's not as if 00 is a special case; they define −π0 to be 0 as well, even though −πy is not defined in any neighborhood of 0. Because, "If x is a finite value less than 0, and y is a finite non-integer, a domain error occurs, and a NaN is returned." So watch out for rounding errors... — Carl (CBM · talk) 02:46, 18 June 2010 (UTC)
- Kahan who was the chief architect of the IEEE floating point standard decided pow(0,0) should be 1 early on as the value people would most often want, a kind of engineers rather than mathematicians definition. A particular reason for this was they didn't have a separate integer power function. The 2008 standard has recommended this value and has a rationale which can be applied to any function which in the limit tends to a value in all but a vanishingly small part of a ball round a singularity. My though was it showed intelligent people are better at rationalizing. :) Dmcq (talk) 23:38, 17 June 2010 (UTC)
- Great! I've made the change. I also added an interesting paper I found through the indeterminate form article, and another I found while searching... apparently many exercises of the 00 form in calculus textbooks have the answer 1, and there is a reason for that. Shreevatsa (talk) 22:57, 17 June 2010 (UTC)
- This is a constructive proposal that I can support. --Trovatore (talk) 21:31, 17 June 2010 (UTC)
- The IEEE 754-2008 really is only for floating point numbers. There was some argument during the revision of the standard issued in 2008 that they should have a new recommended function ipow which had an integer power. This would have helped all the people who write pow(x,n) where n is an integer. Currently the n is converted to a floating point number and handed over, then on the other side they check to see if n represents an exact integer first. The proposal didn't get off the ground though as they wanted to change pow incompatibly at the same time. Dmcq (talk) 08:22, 18 June 2010 (UTC)
- I may be wrong there, it looks like the issued standard includes pown and rootn but I haven't the actual issued version to check. I also want to check whether pow(-1,∞) is 1 or NaN. The argument for 1 is that any sufficiently large IEEE floating point number is even! So much for switching a light on and off an infinite number of times. Dmcq (talk) 09:04, 18 June 2010 (UTC)
- Good for google, there's a recent book that goes into it all: Handbook of Floating-Point Arithmetic By Jean-Michel Muller et al. And the preview says pow(-1,∞) is 1 and they have powr as well which gives the results as for ey log x. So everyone's got what they wanted, I think in Wikipedia we call this a POV fork. ;-) Dmcq (talk) 09:13, 18 June 2010 (UTC)
Under the headings 'Integer exponents, Powers of zero' it is written: If the exponent is zero, some authors define 00=1, whereas others leave it undefined, as discussed below. As the undefiners' (that is Trovatore, Carl, and Dmcq)'s objections against 00=1 are in the context of real exponents only, I want to change it to If the exponent is zero, the power is the empty product. So 00=1. Similarily under the heading 'Integer exponents, Exponents one and zero'. Readers not interested in noninteger exponents should not be confused by uncomprehensible reservations. Objections? Bo Jacoby (talk) 07:34, 22 June 2010 (UTC).
- Um. From my mathematical and philosophical point of view, that would be fine. From an encyclopedic, tertiary-source point of view, I'm less sure. What do the sources really say? --Trovatore (talk) 07:47, 22 June 2010 (UTC)
- It would be better to have a bit about power to a natural number 0 but I don't think the reference to the discussion section can be cut out completely. It would confuse too many people if it just baldly said it was 1 there and later on said it wasn't, it needs careful qualification as many people would not see any difference between an integer 0 and a real number 0. And that leads on to the other problem, as Trovatore indicates Wikipedia should really be describing what the sources say and they're a bit confused about the whole business. In this instance I think maybe we could do a bit of a Bourbaki on it saying it is 1 but I think it requires more discussion before sticking in such OR, perhaps advertise on the Maths project for a good statement? Dmcq (talk) 07:58, 22 June 2010 (UTC)
- No, you're missing the whole point. We don't want the article to make a claim that 0^0 "is" an empty product. In the circumstances where 0^0 is interpreted as an empty product, people usually take 0^0 = 1 by convention. In situations where 0^0 does not represent an empty product, such as calculus, they don't. I believe this is what the article already says. It's best to keep all the discussion of 0^0 together in one section; it gets confusing if one place says 0^0 is 1, another place says 0^0 is undefined, and readers have to find a third place for an explanation. — Carl (CBM · talk) 12:37, 22 June 2010 (UTC)
- The closest analogy that I know of is . If this expression is interpreted as an improper Riemann integral, it has a finite value. If it is interpreted as a Lebesgue integral, it does not converge. Usually, our notation is set up to be resiliant to this sort of change. But in some cases like 0^0 and this integral, context is required to know what a certain expression is supposed to represent. — Carl (CBM · talk) 12:52, 22 June 2010 (UTC)
- Not sure what that's about, I thought Lebesgue integration got the same as Riemann integration whenever that is defined,and anyway there's nothing strange about the sinc function that I can see. How about the Dirichlet's function which Riemann integration can't cope with? Dmcq (talk) 13:48, 22 June 2010 (UTC)
- The theorem you are thinking of is that the Lebesgue and Riemann integrals agree for continuous functions on closed, bounded intervals. However, Lebesgue integrals converge absolutely or not at all, so any conditionally-convergent improper Riemann integral diverges as a Lebesgue integral (there are no improper Lebesgue integrals, the same definition works for any measurable domain of integration.) I see now our article on sinc discusses this very issue. Dirichlet's function is OK as an example here, too, but the one I used above is easier to state. — Carl (CBM · talk) 13:56, 22 June 2010 (UTC)
- Thanks, I just had a read of Improper integral and hadn't realized that aspect about them. I guess quite a few first class mathematicians have struggled with integration so I think I'll just excuse myself that! :) Dmcq (talk) 14:26, 22 June 2010 (UTC)
- The theorem you are thinking of is that the Lebesgue and Riemann integrals agree for continuous functions on closed, bounded intervals. However, Lebesgue integrals converge absolutely or not at all, so any conditionally-convergent improper Riemann integral diverges as a Lebesgue integral (there are no improper Lebesgue integrals, the same definition works for any measurable domain of integration.) I see now our article on sinc discusses this very issue. Dirichlet's function is OK as an example here, too, but the one I used above is easier to state. — Carl (CBM · talk) 13:56, 22 June 2010 (UTC)
I completely agree with Carl that it gets confusing if one place says 0^0 is 1, another place says 0^0 is undefined. There are more examples, Carl: The expression , interpreted as a fraction, has a value. When interpreted as an integer, is not defined, (unlike which is defined as an integer). I am pleased that Carl does not also insist that WP mention that some authors leave undefined. Bo Jacoby (talk) 14:34, 22 June 2010 (UTC).
Prune list of computer programs returning 0^0 values
I think we should restrict the list of calculators etc that return a value of 00 to those that document the value they return rather than having people trying them out. I think most of the ones there can be removed citing WP:original research. Especially as a value may change if an underlying library changes and the value isn't guaranteed. We shouldn't be noting things noone else has bothered noting. Dmcq (talk) 22:43, 28 June 2010 (UTC)
- I agree; it looks like it has grown too long. Maybe just the section on the IEEE standard – is that too much cutting? — Carl (CBM · talk) 14:13, 29 June 2010 (UTC)
Programming Languages and the 1-Mandate
The text claims that the C language mandates 00 = 1. The 2007 official Committee Draft ISO/IEC 9899:TC3, linked from the C WG standards page, has this in § 7.12.7.4:
The pow functions compute x raised to the power y. A domain error occurs if x is finite
and negative and y is finite and not an integer value. A range error may occur. A domain error may occur if x is zero and y is zero. A domain error or range error may occur if x
is zero and y is less than zero.
So no 1 here, a domain error may occur for 00, even though the Standard mentions IEEE floating point in other places, making 1 a likely, and NaN a possible outcome, AFAICT. I guess, then, the information of the article is not reliably stipulating the mandate (that 00 = 1!). The article cites the C Rationale to support the claim. The Rationale is from 2003. If there is an obvious truth to be had from the C standard, is there a language lawyer around who could give a definitive answer? GeorgBauhaus (talk) 13:49, 4 October 2010 (UTC)
- Something applicable: starting from the premise in normative Annex F:
Where a binding between the C language and IEC60559 [IEEE 754] is indicated, the IEC60559-specified behavior is adopted by reference, unless stated otherwise.
- F.9.4.4 states that pow(x,±0) returns 1 for any x ,even a NaN. So if a compiler supports IEEE FPT, 00 = 1 in C. Apologies for the noise. GeorgBauhaus (talk) 15:36, 4 October 2010 (UTC)
- I just had a look and it seems that the applicable word in the rationale is 'normative' rather than mandatory or standards conformant. I'll stick it into the article. I think it means the compiler writers should mention the difference in a language standards compliance statement, I'm not sure it even means that much though. Dmcq (talk) 16:17, 4 October 2010 (UTC)
Back power operator?
I've never heard of the 'back power operator' na as written in the first section but I have seen that notation used for Tetration. Is the back power operator a real thing?
57.66.56.195 (talk) 12:44, 29 June 2010 (UTC)
- It definitely isn't something that should be in the leader, it isn't described elsewhere or cited. I don't know why people stick things like that in leads. The see also for tetration is good enough. I'll go and delete it. Dmcq (talk) 13:00, 29 June 2010 (UTC)
more 0^0
Quote: In most settings not involving continuity in the exponent, interpreting 0^0 as 1 simplifies formulas and eliminates the need for special cases in theorems. (See the next paragraph for some settings that do involve continuity.) For example:
may be simplified to: Interpreting 0^0 as 1 simplifies formulas and eliminates the need for special cases in theorems.
Quote: On the other hand, 0^0 must be handled as an indeterminate form when it is an algebraic expression obtained in the context of determining limits.
Expressions involving limits are not algebraic expressions. They are analytical expressions. I suggest: On the other hand, a limit of the form limx,y→0,0 yx = 0^0 must be handled as an indeterminate form:
Can the undefiners accept these tiny improvements? Bo Jacoby (talk) 21:01, 20 July 2010 (UTC).
- Having looked at Analytical expression I think it would be confusing, in fact I'm not at all sure what an analytical expression is really supposed to be. Dmcq (talk) 23:12, 20 July 2010 (UTC)
- Thank you. I agree. What I ment by an 'analytic expression' is that it is an expression belonging to analysis rather than to algebra. I suggest that neither 'analytic expression' nor 'algebraic expression' be used in the article. Bo Jacoby (talk) 04:48, 21 July 2010 (UTC).
- Reinstated the bit giving basic conditions where the 0 to the power of 0 can be defined as 1 without any major problems. Dmcq (talk) 09:22, 21 July 2010 (UTC)
Done. Note that
- 'the functions below are of the form f(t)g(t) with f(t),g(t) → 0 as t → 0+ '
should be
- 'the functions below are of the form f(t)g(t) with f(t),g(t) → 0+ as t → 0+ ' but I leave it to the undefiners to clean this up. Bo Jacoby (talk) 05:18, 21 July 2010 (UTC).
- The third example has g(t) tending to 0 from below. Dmcq (talk) 09:22, 21 July 2010 (UTC)
- You are right. It should be :'the functions below are of the form f(t)g(t) with f(t) → 0+ and g(t) → 0 as t → 0+ ', but I still leave it to the undefiners to clean up. Bo Jacoby (talk) 16:26, 21 July 2010 (UTC).
Quote: 'There is no holomorphic function defined in a neighborhood of 0 that agrees with zz for all positive real numbers z'.
Nor is there any holomorphic function defined in a neighborhood of 0 that agrees with √z for all positive real numbers z. This does not prevent √0 from being defined. So the argument is invalid and I will remove it. Bo Jacoby (talk) 06:56, 21 July 2010 (UTC).
- There was no need to remove the bit before plus its citation. Dmcq (talk) 09:22, 21 July 2010 (UTC)
The article remains unsatisfactory when you undo improvements. Respect the rules. Why talk about holomorphic functions? Bo Jacoby (talk) 10:49, 21 July 2010 (UTC).
- If you will check what I did I left out the statement you were complaining about with the holomorphic function that was uncited but reinstated the bit before which is but you still took out. Dmcq (talk) 15:01, 21 July 2010 (UTC)
- The expression "0^0" itself is the indeterminate form, not the limit. Indeed, our article on indeterminate forms includes 0^0. — Carl (CBM · talk) 11:03, 21 July 2010 (UTC)
- Also, what is this with "the undefiners"? Nobody here is arguing that 0^0 is never defined, or should never be defined. The point is just that in certain settings it is defined, and in other settings it is not. — Carl (CBM · talk) 11:07, 21 July 2010 (UTC)
- When then is undefined, even if is defined. An indeterminate form need not be undefined.
- The article is made difficult for the uninitiated reader by some editors insisting that the article shall report the trivial fact that some authors do not define 0^0. If 0^0 is defined by one or more authors, then it is irrelevant that other authors leave it undefined. Those editors, who undefine what was defined in the very same article they are editing, are the undefiners. Bo Jacoby (talk) 13:00, 21 July 2010 (UTC).
- In the context of calculus, every indeterminate form is undefined: 00, 0/0, 1∞, ∞ − ∞, ∞/∞, 0 · ∞, ∞0, etc. The article does a careful job of explaining the contexts in which the expression "0^0" is defined, and the contexts in which it is not. Basically: when 0^0 is intended to represent an empty product, it is taken by convention to be 1. When 0^0 is intended to represent exp(0 log(0)), it is undefined. When 0^0 is a formal expression obtained by substituting limits of integration, it is undefined.
- 0^0 is not the only form with this property. The form 0·∞ is undefined, and an indeterminate form, in calculus. But in measure theory there is a common convention that 0·∞ = 0.
- All that this article can do is to report the status of things are they are in mathematics. The status of 0^0 is that it is defined in some contexts but not in others, and so that's what this article needs to say. — Carl (CBM · talk) 13:26, 21 July 2010 (UTC)
- Amen to that. Please no more 'undefiners' business. It is a question of what the sources say rather than what we want. If I had my choice I'd talk about integers and reals but the sources don't do that. Dmcq (talk) 16:35, 21 July 2010 (UTC)
- Thank you Carl for repeating the explanation. We agree that some authors define 00, and other authors leave it undefined. How shall this situation be handeled in wikipedia? That is the question.
- When author A says: "It is defined" and author B says: "It is undefined", does that mean that they have a mathematical disargreement? No. It merely means that A defined it and B did not. A has reasons to define it and B has not. Nor has B reasons not to define it.
- The fact, that some authors define 00 and others (including Goethe and Shakespeare) do not define it, is by the undefiners expressed by saying that "in certain settings it is defined, and in other settings it is not". This solution involves a contradiction because the same thing cannot be defined and undefined at the same time. We also need an explanation of "setting". (One undefiner involved in controversial original research, including 00≠00.0 and 0≠0.0, in order to explain when 00 should be defined and when not). The other way to express the fact is simply to say that 00 is defined. The fact that some authors do not define 00 is trivial and irrelevant and useless and goes without saying, and it certainly does not make 00 undefined.
- That the expressions 00 , 0/0, 1∞, ∞−∞, ∞/∞, 0·∞, and ∞0, are indefinite forms means that the following limits , , , , , and are undefined. It says nothing about the expressions themselves. They may be defined (like 00 and 0·∞) or not. Carl writes: "In the context of calculus, every indeterminate form is undefined". It is not always decidable whether we are "in the context of calculus". And the usual expression , for the taylor series for , which is definitely "in the context of calculus", needs 00 to be defined. So, Carl, your statement is not true.
- The undefiners make the article incomprehensible and even nonsensical, to no benefit for the reader. "Any nonzero number raised to the power 0 is 1; one interpretation of these powers is as empty products. The case of 00 is discussed below" is a disaster compared to "Any number raised to the power 0 is 1". An empty product does not depend on the value of the factors which it does not contain! Bo Jacoby (talk) 13:39, 22 July 2010 (UTC).
- Bo, the above is about as well-argued as it's possible to argue your position, and I have to give you credit for that. But the basic fact here is that the article as you prefer it would leave the impression that the reform proposed by Knuth is generally accepted, and in fact it is not. It's true that there aren't a lot of people bothering to argue against it; mostly, those who don't accept it simply ignore it. --Trovatore (talk) 20:23, 23 July 2010 (UTC)
- Trovatore, thanks for giving me credit, that is really nice! I don't mind issuing the WARNING: the reduction is incorrect!!! and so highlighting: The Definition 00 = 1 Used Here Is Not Generally Accepted!!!, but I do mind writing that the value of an empty product a0 depends on a, because that leaves the (possibly false) impression that the wikipedia editors are all insane. Bo Jacoby (talk) 09:01, 24 July 2010 (UTC).
- The expression "a0" does not always represent an empty product. In many texts, the expression "ab" is defined to mean exp(b ln(a)). In such texts, it is a theorem rather then a definition that a2 = a·a, and in such texts the expression 00 has no value, because ln(0) has no value. This has nothing to do with changing the value of an empty product. — Carl (CBM · talk) 11:34, 24 July 2010 (UTC)
- The expression exp(b ln(a)) does not define ab when a=b=0. The empty product does. All the literature that does not define 00 is irrelevant where 00 is discussed. Bo Jacoby (talk) 12:21, 24 July 2010 (UTC).
- If a particular paper or book defines ab as exp(b ln(a)), then for the remainder of that book 00 has no value at all, unless that same book goes out of its way to define 00. This doesn't change if some other book uses a definition that does give 00 a value. This is the sort of thing reflected in footnote 11, which refers to an article on complex analysis directly on the topic of the function xx.
- The expression exp(b ln(a)) does not define ab when a=b=0. The empty product does. All the literature that does not define 00 is irrelevant where 00 is discussed. Bo Jacoby (talk) 12:21, 24 July 2010 (UTC).
- The expression "a0" does not always represent an empty product. In many texts, the expression "ab" is defined to mean exp(b ln(a)). In such texts, it is a theorem rather then a definition that a2 = a·a, and in such texts the expression 00 has no value, because ln(0) has no value. This has nothing to do with changing the value of an empty product. — Carl (CBM · talk) 11:34, 24 July 2010 (UTC)
- Trovatore, thanks for giving me credit, that is really nice! I don't mind issuing the WARNING: the reduction is incorrect!!! and so highlighting: The Definition 00 = 1 Used Here Is Not Generally Accepted!!!, but I do mind writing that the value of an empty product a0 depends on a, because that leaves the (possibly false) impression that the wikipedia editors are all insane. Bo Jacoby (talk) 09:01, 24 July 2010 (UTC).
- Bo, the above is about as well-argued as it's possible to argue your position, and I have to give you credit for that. But the basic fact here is that the article as you prefer it would leave the impression that the reform proposed by Knuth is generally accepted, and in fact it is not. It's true that there aren't a lot of people bothering to argue against it; mostly, those who don't accept it simply ignore it. --Trovatore (talk) 20:23, 23 July 2010 (UTC)
- I think that the underlying issue here is that you are arguing that the expression "0^0" should always be read as an empty product. But that's not right at all; the notation "x^0" can have all sorts of meanings, only some of which involve empty products. — Carl (CBM · talk) 13:59, 24 July 2010 (UTC)
- A book that defines ab as exp(b ln(a)), but does not define 00, is as irrelevant for the discussion of 00 as any other book that does not define 00. (It goes without saying that most books does not at all define ab). The notation x0 may have more interpretations for x≠0 than for x=0, and so what? Bo Jacoby (talk) 20:16, 24 July 2010 (UTC).
- Not irrelevant at all. An author who defines ab as exp(b ln(a)) is perfectly entitled to ask, what happens when a=0 and b=0? The answer is that, following that definition, 00 has no value. We have a reference for this phenomenon already in the article: footnote 11 is a paper directly on the topic of the function xx where x is real but the value is allowed to be complex. — Carl (CBM · talk) 00:48, 25 July 2010 (UTC)
- The square root of x may be defined by exp(ln(x)/2). Does that mean that the square root of zero is sometimes undefined? Bo Jacoby (talk) 06:34, 25 July 2010 (UTC).
- Can you cite anyone who actually defines it that way? Of course, if it is defined that way, then sqrt(0) is undefined. But in fact it's not defined that way. --Trovatore (talk) 07:34, 25 July 2010 (UTC)
- Yes, Carl just taught us that the square root x1/2 = exp(log(x)/2). Bo Jacoby (talk) 09:30, 25 July 2010 (UTC).
- I didn't say x1/2. I said the square root of x. Can you cite anyone who defines the square root of x as exp(ln(x)/2)? --Trovatore (talk) 19:10, 25 July 2010 (UTC)
- Yes, Carl just taught us that the square root x1/2 = exp(log(x)/2). Bo Jacoby (talk) 09:30, 25 July 2010 (UTC).
- Can you cite anyone who actually defines it that way? Of course, if it is defined that way, then sqrt(0) is undefined. But in fact it's not defined that way. --Trovatore (talk) 07:34, 25 July 2010 (UTC)
- The square root of x may be defined by exp(ln(x)/2). Does that mean that the square root of zero is sometimes undefined? Bo Jacoby (talk) 06:34, 25 July 2010 (UTC).
- Not irrelevant at all. An author who defines ab as exp(b ln(a)) is perfectly entitled to ask, what happens when a=0 and b=0? The answer is that, following that definition, 00 has no value. We have a reference for this phenomenon already in the article: footnote 11 is a paper directly on the topic of the function xx where x is real but the value is allowed to be complex. — Carl (CBM · talk) 00:48, 25 July 2010 (UTC)
- A book that defines ab as exp(b ln(a)), but does not define 00, is as irrelevant for the discussion of 00 as any other book that does not define 00. (It goes without saying that most books does not at all define ab). The notation x0 may have more interpretations for x≠0 than for x=0, and so what? Bo Jacoby (talk) 20:16, 24 July 2010 (UTC).
- I think that the underlying issue here is that you are arguing that the expression "0^0" should always be read as an empty product. But that's not right at all; the notation "x^0" can have all sorts of meanings, only some of which involve empty products. — Carl (CBM · talk) 13:59, 24 July 2010 (UTC)
- (←) If no other provision is made, then yes that power would be undefined. Here is a quote from a different book: Classical Complex Analysis by Mario González:
- "For we define by the equality . The principal value ... is defined by .... For , we define , while is not defined. "
- The author there went out of his way to ensure that he did not define a value for 0^0, even as he defined powers of 0 for other exponents. — Carl (CBM · talk) 12:15, 25 July 2010 (UTC)
- Mario González apparently defines . Most authors do not define . Some authors define . Bo Jacoby (talk) 17:12, 25 July 2010 (UTC).
- Indeed. Now, I don't think the issue of 0^{-1} warrants much discussion in this article. However, it would be equally inappropriate for the article to simply say "0^{-1} equals ∞" or "0^{-1} equals 0" as it would be for the article to simply say "0^0 = 1". In each case we are talking about a convention that some authors follow and some do not. — Carl (CBM · talk) 17:21, 25 July 2010 (UTC)
- No, my point is that it goes without saying what some authors do not define. Nor should it be reported in wikipedia that we have now found an author who even defines 1/0=0. The article must make sense. It is incorrect to translate: "some authors do not define 00" into: "00 is sometimes not defined" because that does not make sense. Bo Jacoby (talk) 18:37, 25 July 2010 (UTC).
- It is explicitly not defined otherwise the reasoning of indeterminate forms doesn't work. Your reasoning would follow even for 0/0=1. See removable singularity for another instance of the same kind of thing, you couldn't have a removable singularity if the value was actually defined as something else. As indeterminate form says 'Limits involving algebraic operations are often performed by replacing subexpressions by their limits; if the expression obtained after this substitution does not give enough information to determine the original limit, it is known as an indeterminate form.' If 00 was defined as 1 then an expression of that form would have the value 1 and therefore would give enough information to determine the original limit. Unfortunately that value could easily be wrong. Basically you'd have to forget about treating exponentiation like any other operation and treat it as a black box function instead. Dmcq (talk) 20:01, 25 July 2010 (UTC)
- 'Limits involving algebraic operations are often performed by replacing subexpressions by their limits', but sometimes this replacement is not justified. Repeating: the reduction is incorrect!!! This reduction is incorrect no matter whether 00 is defined or not. Undefining 00 does not help. Defining 00 does not help either. Whether you define 00 or undefine 00, the reduction is still incorrect. Such reductions are nevertheless often performed, and so 00 is called an indeterminate form to remind you that when you get 00 as a result of you should be warned that you have made an incorrect reduction. Try to understand this, because we are going in circles here. Bo Jacoby (talk) 21:20, 25 July 2010 (UTC).
- Exactly how do you know it is an incorrect reduction if the value of 00 is defined? What you are trying to do would cause bits of established practical mathematics used for the last few hundred years or so to no longer be valid. That doesn't sound like a good idea to me. As to ew log(z) the value of that for zero arguments is normally defined where possible using limits the same as 1/(1/x) would normally be considered to be zero when x is zero. I'd guess the book made the mistake by only considering positive real values of w. Dmcq (talk) 21:52, 25 July 2010 (UTC)
- I know it is an incorrect reduction because So the singularity isn't removable. This has nothing to do with if the value of 00 is defined or not. Bo Jacoby (talk) 23:08, 25 July 2010 (UTC).
- Instead of throwing away the workings of indeterminate forms it is much easier to simply evaluate the expression passing down the limits and if the value is undefined you know either there is no limit or there was an invalid reduction somewhere. That is easy to follow and people are happy to follow it. If 00 is defined as 1 then the value is not undefined. In anything except where the exponent is an integer there just is no point in defining 00 as 1 because it causes trouble. As you say it is not a removable singularity and those are the only ones where people really feel happy defining a value to fill in the gap. When it is an integer exponent it is not a variable and filling the gap at 0 of x0 is something people are happy with and the obvious value to define the value of that at 0 is 1. If one had 0/x even many people would be quite happy to define that as 0 when x is 0 if it was going to fill a gap and simplify the work rather than having special cases. That doesn't mean 0/0 should in general be defined as 0. Dmcq (talk) 11:18, 26 July 2010 (UTC)
- I am sorry, Dmcq, but you are completely mistaken. Non-removable singularities are not marked by undefining the function value. Consider the sign function defined by sgn(x)=+1 when x>0, sgn(x)=0 when x=0, and sgn(x)=−1 when x<0. This function has a non-removable singularity at x=0, because . Yet sgn(0) is defined. Bo Jacoby (talk) 11:41, 26 July 2010 (UTC).
- Instead of throwing away the workings of indeterminate forms it is much easier to simply evaluate the expression passing down the limits and if the value is undefined you know either there is no limit or there was an invalid reduction somewhere. That is easy to follow and people are happy to follow it. If 00 is defined as 1 then the value is not undefined. In anything except where the exponent is an integer there just is no point in defining 00 as 1 because it causes trouble. As you say it is not a removable singularity and those are the only ones where people really feel happy defining a value to fill in the gap. When it is an integer exponent it is not a variable and filling the gap at 0 of x0 is something people are happy with and the obvious value to define the value of that at 0 is 1. If one had 0/x even many people would be quite happy to define that as 0 when x is 0 if it was going to fill a gap and simplify the work rather than having special cases. That doesn't mean 0/0 should in general be defined as 0. Dmcq (talk) 11:18, 26 July 2010 (UTC)
- I know it is an incorrect reduction because So the singularity isn't removable. This has nothing to do with if the value of 00 is defined or not. Bo Jacoby (talk) 23:08, 25 July 2010 (UTC).
- Exactly how do you know it is an incorrect reduction if the value of 00 is defined? What you are trying to do would cause bits of established practical mathematics used for the last few hundred years or so to no longer be valid. That doesn't sound like a good idea to me. As to ew log(z) the value of that for zero arguments is normally defined where possible using limits the same as 1/(1/x) would normally be considered to be zero when x is zero. I'd guess the book made the mistake by only considering positive real values of w. Dmcq (talk) 21:52, 25 July 2010 (UTC)
- 'Limits involving algebraic operations are often performed by replacing subexpressions by their limits', but sometimes this replacement is not justified. Repeating: the reduction is incorrect!!! This reduction is incorrect no matter whether 00 is defined or not. Undefining 00 does not help. Defining 00 does not help either. Whether you define 00 or undefine 00, the reduction is still incorrect. Such reductions are nevertheless often performed, and so 00 is called an indeterminate form to remind you that when you get 00 as a result of you should be warned that you have made an incorrect reduction. Try to understand this, because we are going in circles here. Bo Jacoby (talk) 21:20, 25 July 2010 (UTC).
- It is explicitly not defined otherwise the reasoning of indeterminate forms doesn't work. Your reasoning would follow even for 0/0=1. See removable singularity for another instance of the same kind of thing, you couldn't have a removable singularity if the value was actually defined as something else. As indeterminate form says 'Limits involving algebraic operations are often performed by replacing subexpressions by their limits; if the expression obtained after this substitution does not give enough information to determine the original limit, it is known as an indeterminate form.' If 00 was defined as 1 then an expression of that form would have the value 1 and therefore would give enough information to determine the original limit. Unfortunately that value could easily be wrong. Basically you'd have to forget about treating exponentiation like any other operation and treat it as a black box function instead. Dmcq (talk) 20:01, 25 July 2010 (UTC)
- No, my point is that it goes without saying what some authors do not define. Nor should it be reported in wikipedia that we have now found an author who even defines 1/0=0. The article must make sense. It is incorrect to translate: "some authors do not define 00" into: "00 is sometimes not defined" because that does not make sense. Bo Jacoby (talk) 18:37, 25 July 2010 (UTC).
- Indeed. Now, I don't think the issue of 0^{-1} warrants much discussion in this article. However, it would be equally inappropriate for the article to simply say "0^{-1} equals ∞" or "0^{-1} equals 0" as it would be for the article to simply say "0^0 = 1". In each case we are talking about a convention that some authors follow and some do not. — Carl (CBM · talk) 17:21, 25 July 2010 (UTC)
- Mario González apparently defines . Most authors do not define . Some authors define . Bo Jacoby (talk) 17:12, 25 July 2010 (UTC).
- If you would have a look at the section sign function#Generalized signum function you will see that some people use a version where it cannot have a defined value at 0. How about going and telling them they must accept the value 0 at 0 because signum is defined that way? Dmcq (talk) 12:51, 26 July 2010 (UTC)
(←) No, the point is that neither undefining nor redefining sgn(0) changes the fact that lim sgn(−1/n) ≠ lim sgn(+1/n) even if lim −1/n = lim +1/n, where the limits are for n→∞. Undefining 00 solves no problem. I am not an expert on the Algebra_of_generalized_functions so I do not know which problem undefining sgn(0) without undefining (sgn(0))2 is supposed to solve. Are we having a sincere discussion or are you just playing a game? Bo Jacoby (talk) 15:59, 26 July 2010 (UTC).
- You seem to have problems about discussing this without making personal attacks so I shall once more absent myself from the discussion. Dmcq (talk) 16:33, 26 July 2010 (UTC)
- It is not an attack. It is a question. Bo Jacoby (talk) 16:57, 26 July 2010 (UTC).
- Since we have asked you nicely not to try turning things into a battleground of a versus B with your undefiners and you continue I will not answer any of your queries here as per WP:DENY. All these points have been gone through before and I see no point in discussing them again with someone who will not be WP:CIVIL When you wish to discuss things in a reasonable manner say so. Dmcq (talk) 13:56, 22 July 2010 (UTC)
- This is a civil discussion about the formulation of the article, not a battleground. Bo Jacoby (talk) 22:15, 22 July 2010 (UTC).
- Okay I see you have stopped referring to 'the undefiners'. Thanks. Dmcq (talk) 15:38, 25 July 2010 (UTC)
- This is a civil discussion about the formulation of the article, not a battleground. Bo Jacoby (talk) 22:15, 22 July 2010 (UTC).
still more 0^0
I have the feeling that the present discussion does not start from the main issue. To me, the point is not: should one give a meaning to the expression , and if so, what should it be? To me, this is a fake probelm: not motivated, not at all urgent; an encyclopaedic aberration, so to speak. Putting the problem this way, the motivation only seems to have a typographical-combinatorial source, and a natural answer would be then: no, why on the Earth should one give a meaning to any typographical combination. We may as well ask ourselves: "should one give a meaning to a polygon with vertices? should one give a numeric value to ?" The point is, that not all that one can write necessarily must have a sense. In mathematics, introducing new objects and new definitions should have better motivation than just "why not?".
My point about the debated point is, instead, that e.g. the cardinality of mappings from the set to the set is a well defined number, and we certainly need a notation for this, even in the case . So the real and practical questions are: how to denote the number of maps from the empty set to itself? How to denote a polynomial or a power series? &c. The very natural answer (in fact, universally agreed) is: by using the convention --pma 13:16, 30 July 2010 (UTC)
- No one disputes that 0^0 is 1 when interpreted as the cardinality of maps from the empty set to the empty set. But that is not what the real-to-real-power exponential function is. When we say , we certainly don't mean that the cardinality of the set of maps from a set of cardinality to one of cardinality is ! That would be simply gibberish. And the meaning of the exponential function is not based on case logic; there is no special case for zero.
- So the conclusion must be that the real-to-real-power exponential function is simply a different function from the natural-to-natural-power one. They're notated the same, and called by the same name, but that does not make them the same.
- They form a nice commutative diagram with large ranges of the inclusion map, and so you could if you wanted to make a grand unified exponential function that incorporates them both. That's what Bo wants to do. But in practice no one does that. For the very good reason that the grand unified exponential function isn't good for anything. --Trovatore (talk) 20:24, 30 July 2010 (UTC)
- Trovatore writes: "So the conclusion must be". That conclusion is Trovatore's original research and not the conclusions of the references.
- Trovatore writes: "in practice no one does that". Thats what everybody do. There is no warning in the article that the exponentiation function is really several functions "notated the same, and called by the same name".
- Trovatore wrote: "I didn't say x1/2. I said the square root of x". The article on square root says: "For all non-negative real numbers x and y, and ".
- Trovatore wrote: "Notice that in at least the first two examples, the exponent is an integer, not a real number", and "unless you're insisting on the exponent being real, as opposed to integer". Are the integers not real numbers? The article real number says: "Real numbers include both rational numbers and irrational numbers", and the article rational number says: "a rational number is any number that can be expressed as the quotient a/b of two integers, with the denominator b not equal to zero. Since b may be equal to 1, every integer is a rational number".
- Trovatore wrote: "The reals are not a generalization of the naturals or the integers or the rationals". [citation needed]. The article Rational number#Real numbers and topological properties says: "The rationals are a dense subset of the real numbers". This means that the reals are a generalization of the rationals.
- Dmcq wrote: "many people would not see any difference between an integer 0 and a real number 0". Nor does the article on zero.
- Dmcq wrote: "Integers are not a restricted set of real numbers any more than the real numbers are a restricted set of complex numbers". [citation needed]. What is a restricted set?
- For notation, let be the function defined as . In analysis, people often use the notation for this function, instead of . But all the properties that we associate with the notation have to be proved as theorems about f; they are not part of the definition. For example, it has to be proved that equals . Because the definition of is , and the exp and ln functions themselves have complicated definitions. In particular, exp is not defined in terms of powers of e.
- When x is a positive real number, every function I have seen written as gives the same value. But when x is zero, there are different conventions the people follow. For example, I gave a reference above that defines . That is perfectly internally consistent as long as the author does not also try to prove that whenever is defined. That is, under that author's definitions the notation does not always represent the reciprocal of x, when that notation is read as referring to the real exponentiation operation.
- The issue with is similar. In the definitions used by some authors, it is 1, while the definitions used by author authors cause the expression to have no value when 0 is substituted for x and y. Only sometimes does the notation refer to an empty product. That's what the article already says. — Carl (CBM · talk) 13:36, 31 July 2010 (UTC)
- Just curious, does your author define exp and log without using exponentiation with integer exponent like exp(x) = limn→∞(1+x/n)n or exp(x) = Σnxn/n! ?
- Yes, the function f can be defined such that f(x,y) = xy whenever both sides are defined, and f(0,0) is undefined while 00 is defined. This f(x,y) does not merit treatment on wikipedia. The article is about xy.
- That "other authors cause the expression to have no value when 0 is substituted for x and y" is not a definition at all. Definitions are useful, but undefinitions are useless.
- The article says that "some authors define 00=1, whereas others leave it undefined", which is true, but it goes without saying that others leave it undefined. That is unimportant and confusing information. The reader may be interested in useful definitions but is certainly uninterested in useless undefinitions.
- Bo, no it isn't what everyone does. The grand unified exponential function is your original research. What everyone does is define different functions in the different contexts. They just don't bother to warn you about it, because it's usually not important. --Trovatore (talk) 17:56, 31 July 2010 (UTC)
- Trovatore told us that "some authors define 00=1, whereas others leave it undefined". Trovatore did not report any single author entertaining the two conflicting function definitions simultaneously. The context-dependent interpretation is an original attempt (by Trovatore, Carl, and Dmcq) to unify the conflicting authors. (Please, provide also a reference for the claim that integers are not reals). Bo Jacoby (talk) 21:14, 31 July 2010 (UTC).
- A common definition of ln and exp in calculus first defines ln in terms of an integral of 1/x, then defines exp as the inverse of ln. There is no need to reference integer exponentiation to define real number exponentiation.
- Bo, no it isn't what everyone does. The grand unified exponential function is your original research. What everyone does is define different functions in the different contexts. They just don't bother to warn you about it, because it's usually not important. --Trovatore (talk) 17:56, 31 July 2010 (UTC)
- The claim "This f(x,y) does not merit treatment on wikipedia. The article is about xy." is confused. In the book I pointed out, f(x,y) is xy because xy is defined as f(x,y) there. The notation on xy does not specify a function, it's just notation. — Carl (CBM · talk) 19:47, 31 July 2010 (UTC)
- Thanks. Bo Jacoby (talk) 21:14, 31 July 2010 (UTC).
- When the function which is not defined for (x,y)=(0,0) is not longer called xy on wp, then the notation xy specifies a function, as it is expected to. Bo Jacoby (talk) 21:14, 31 July 2010 (UTC).
- Please do not split up my comments.
- I do not know what you mean by "When the function which is not defined for (x,y)=(0,0) is not longer called xy on wp". This function is called xy in books. It's one of the several definitions of the exponential function.
- Also, the claims made in the article are not original to editors here. For example, the text in footnote 13 explicitly addresses the issue. And we also have several references to authors who explicitly say that 00 has no value in their definitions, including footnote 11 in the article and another one I provided above.
- This conversation seems to be going on and on with no concrete results. Several people have explained why they do not agree with the proposed wording changes you mentioned at 21:01, 2010-7-20, so it is clear there isn't agreement for it. Given that, I personally am going to exist this thread at this point. I'm all for discussing new info, but not for rehashing the same stuff over and over. — Carl (CBM · talk) 01:02, 1 August 2010 (UTC)
- I hold high regards for Carl and can hardly believe that Carl does not understand, but then I must explain more carefully, and don't blame me for continuing this discussion unnecessarily. Carl wrote: "xy does not specify a function, it's just notation". This notation, xy, can, depending on context, specify either of at least two functions, according to Carl. One of these functions is the one defined by Mario González: "For z≠0, f(z,w)= ew Log(z). For w≠0, f(0,w)=0, while f(0,0) is not defined". Another function is g(x,y) defined by g(z,0)=1 for complex (or real or integer) z, g(z,n+1)=z·g(z,n) for integer n≥0, g(z,n−1)=g(z,n)/z for z≠0 and integer n, exp(z)=limn→∞(1+z/n)n, log(x) defined by exp(log(x))=x for (real) x>0, g(x,z)=exp(z log(x)). Now f(0,0) is not defined while by definition g(0,0)=1. "The function which is not defined for (x,y)=(0,0)" is the González function f(x,y). If the wikipedia article ceases to use the notation xy for the González function f(x,y) and reserves the notation xy for the function g(x,y), then the notation xy specifies only one function, as it is expected to.
- Extensions to g(x,y), such as y a square matrix, can still be called xy , but restrictions to g(x,y), such as undefining g(0,0), can no longer be called xy, because once 00 is defined it cannot be undefined.
- That f(0,−1)=0 while g(0,−1) is undefined is another complication. 0−1 is not usually defined to be 0.
- The point of view that 00 is defined or undefined depending on context or interpretation needs an exact definition of 'context'.
- The undefining of 00 serves no purpose and solves no problem, even if Dmcq thinks that the undefining is necessary in order to maintain that 00 is an indeterminate form.
- I share Carl's frustration that this communication is difficult. We must also share the blame. I am not the only editor who have complained that the present wording of the article is unsatisfactory, although presently I am the only one to argue this point of view. The undefiners have a history of reverting contributions, and I do not engage in edit wars, so the current state of the article is the one that the undefiners are happy with, while I am not.
- The point of view that 0≠0.0 needs to be defended or abandoned. Is 0.0 a real number? If yes, is it rational or irrational? If rational, is it an integer or not? If an integer, is it the integer 0 or not? Bo Jacoby (talk) 18:29, 1 August 2010 (UTC).
Limits of powers
The subsection Limits of powers is not conforming to Wikipedia:Encyclopedic style#Tone. It needs to be improved. Bo Jacoby (talk) 06:37, 21 July 2010 (UTC).
- Could you be more specific please? It doesn't strike me as particularly needing attention. In fact overall my feeling about it is that I'm quite glad that someone who knows their maths contributed to Wikipedia and wrote it. Dmcq (talk) 17:52, 25 July 2010 (UTC)
- "One may ask", "we consider", "We view", "we obtain", and "It should be borne in mind" are examples of non-encyclopedic tone. I request the subsection to be improved - not to be removed. Bo Jacoby (talk) 18:21, 25 July 2010 (UTC).
Rational powers of negative bases.
It's important to distinguish between notation, representation, numerals and numbers. The roman numeral "V", "five", "10/2", and "5" are different notation for the same real number. A real number can be either rational or irrational. Rational numbers are defined as the result of dividing two integers, they are not the two integers themselves. So five is a rational number but not a fraction. A fraction is not any kind of number, it is a representation of a number. There is only one fractional representation, but there are several fractional notations ( , 1:2). Every rational has an infinite number of representations as fractions. The rational represented by 3/2 is not a fraction. 3/2 and 6/4 are different fractional representation of the same rational.
There are several common rational power representations:
- Reduced fraction representation:
- Arbitrary fraction representation:
- Decimal representation:
They all represent raising the base to the same rational. If the notation is to be self-consistent, they all must represent the same mathematical operation:
When a is negative, there are two distinct mathematical interpretations:
- Numerator precedence:
- Denominator precedence:
For example, let a = -1:
- Numerator precedence:
- Denominator precedence:
The choice seems arbitrary. Now let a = -1, and let the numerator and denominator = 2, the result must be -1.
- Numerator precedence:
- Denominator precedence:
The denominator must have precedence for the notation to be consistent. To my knowledge this convention is used by most computer algebra systems. —Preceding unsigned comment added by NOrbeck (talk • contribs) 06:44, 24 July 2010 (UTC)
- Real powers of negative real numbers aren't properly defined because of the problems. In the above your preferred definition used the square root of miunus 1 which isn't a real number and in complex numbers you come across the problem of multiple roots. You can define the cube root of a negative number okay because the three is an explicit number in cube root. I think I ought to point out in the text of the section Exponentiation#Negative nth roots that cube roots and other such odd roots are well defined for negative numbers but that is different from saying the 1/3 power is defined.
- At the bottom is the problem of integers being treated as actually being a subset of the real numbers rather than being isomorphic to a subset of the real numbers. Personally I'd distinguish between (-1.0)1.0 and (-1)-1. However people don't distinguish between these in general so we are led to the situation like the previous discussion or the stuff you wrote above. Thankfully the new floating point standard for computers does provide functions to distinguish between them so we have pow(-2.0,3.0) which tries to always give a result returns -8.0, but the better defined functions powr(-2.0,3.0) will return an error, and pown(-2.0,3) returns -8.0, the 3 here is an integer. IEEE floating point can't represent 1/3 exactly but rootn(-8.0,3) will return -2.0 Dmcq (talk) 09:58, 24 July 2010 (UTC)
- I didn't mention it, but my intention was merely to describe how and why TI calculators and other CAS systems, choose to define exponentiation on negative bases. On my calculator the ability to handle negative bases is a kludge that can be turned on and off. Usually, but not always, an answer is better than an error message. NOrbeck (talk) 03:18, 25 July 2010 (UTC)
On the revert it said: "That subsection is explicitly in the positive real numbers section." I couldn't find where it mentioned that raising a negative base to a rational power is anti-commutative, try searching for it. Neither did I find any mention of the ambiguity in rational power notation, notional definition has nothing do with mathematical definition. For example, Microsoft created a font that used the same glyph to represent the letter l as the number 1. Problems occurred because people were unaware that one symbol had two meanings, not because of the definitions of l and 1.
Also, the phrase "rational in lowest terms" is technically meaningless because a rational is not a fraction. I don't think "fraction in lowest terms" is an improvement, since the operand is not a fraction. It's a conundrum. —Preceding unsigned comment added by NOrbeck (talk • contribs) 04:22, 25 July 2010 (UTC)
- About the calculators and CAS, I was going to prune down that section because most of it is just what people found by testing their calculator or whatever and is not documented in WP:reliable sources. A little of that might be okay but it sounds like you would be going well over the top into WP:original research territory. We should only be putting in things that people have noticed and put into a WP:reliable source not stuff we just notice ourselves. Dmcq (talk) 08:50, 25 July 2010 (UTC)
- Raising a negative base to a rational power is not anti-commutative so it shouldn't be mentioned. By ambiguous notation do you mean that 2/4 = 1/2 for instance? If you look at the section Exponentiation#Negative nth roots I pointed to above I believe it deals with this. I think most people know rational in lowest terms has a 'represented' implicitly in it so sticking it in is just noise. Dmcq (talk) 08:50, 25 July 2010 (UTC)
- The most useful interpretation of (−1)x is ei π x. It is continuous and defined for all real or complex values of x. The interpretation (−1)1/(2n+1) = −1 and (−1)1/(2n) = 1 is not really useful even if it is described in elementary math books. Bo Jacoby (talk) 10:40, 28 July 2010 (UTC).
Mizar definition of powers
I was looking through Integer and Rational Exponents in the Mizar formalized mathematics system. It has different natural, integer, rational and real powers and propositions matching them up. Proposition 44 defines integer power of 0 as 1, but definition 8 only defines real powers where the base is positive, not zero or negative.
I was wondering how reliable this source would be considered to be in Wikipedia terms and how this sort if stuff could be incorporated in the article? Perhaps a formalized maths section devoted to it since it isn't exactly how people normally go round it. I'm a bit afraid people would incorporate too much from such documents as they tend to be very long and include nerdishly tiny propositions. Dmcq (talk) 14:05, 31 July 2010 (UTC)
- The material is published, in a sense, in the "Journal of Formalized Mathematics". However, I'm not sure how we would want to use it in this article. Perhaps we could point out in a single sentence that the formalization used in Mizar maintains a distinction between the integer exponentiation operations and the real-number exponentiation, and proves they yield the same values for certain inputs. I think that more than a sentence or two would be too much weight, but it is an interesting example from a well-known formalization project. — Carl (CBM · talk) 19:49, 31 July 2010 (UTC)
Integer and Rational Exponents assumes in definition 1 that , and confirmes it in proposition 4. Also by proposition 44. The statement after definition 4 is: "Let a, b be real numbers. Then is a real number." This indicates that is defined, but not how it is defined. Why not ask the author? Bo Jacoby (talk) 09:16, 4 August 2010 (UTC).
History
At the moment the history makes it sound like exponentiation wasn't thought of till the Arabs started on it. Diophantus had abbreviations for squares and cubes, and probably the Babylonians thought of them even before Euclid. I haven't access to anything at the moment so if anyone else has references about that it would make the history a bit more balanced. Dmcq (talk) 22:07, 28 August 2010 (UTC)
Terminology section
I believe the terminology section is redundant. All it is saying is the normal rules of grammar as in 'an adjective noun1 of a noun2' means adjective applies to noun1. This applies to:
- A positive power of a number
- A negative power of a number
- A non-negative power of a number
- An even power of number
- An odd power of a number
I don't believe we need a section on this. And more to the point I don't believe there are textbooks that point this obvious fact out. Even more we don't need it spelled out at length. A negative power of a number does not mean the same as a power of a negative number because not only English but also other languages distinguish between such things quite easily and well. Dmcq (talk) 16:10, 2 November 2010 (UTC)
- I do not say that a negative power of a number can mean a power of a negative number, but that it can mean that the result is negative. It is related to the ambiguity that power is the result, but can also be used for the exponent, like in the phrases above: 2^0 is odd, unlike higher integer powers of 2, which are even, so 2^0 is an odd power of 2, but in the other meaning, an even power of 2.--Patrick (talk) 20:56, 2 November 2010 (UTC)
- Have you actually ever seen power being used in such a way or a book that describes this problem? The adjectives are being used in the standard way, we don't need to describe usual English sentence structure. There might be some point if there was something unusual but there isn't. Dmcq (talk) 22:46, 2 November 2010 (UTC)
- Anyway I've had another think. You seem to think it could easily be confusing so possibly others will be confused, which is very bad. The explanation you gave in the article was quite long and nowhere near as good as the quick one you gave above so I'll copy the one here back into the article. Dmcq (talk) 23:03, 2 November 2010 (UTC)