A lot of calculators actually agree with me saying that it is defined and the result equals 1, which makes sense to me because:

$$ (-1)^{2.16} = (-1)^2 \cdot (-1)^{0.16} = (-1)^2\cdot\sqrt[100]{(-1)^{16}}\\ = (-1)^2 \cdot \sqrt[100]{1} = (-1)^2 \cdot 1 = 1$$

However, there are certain calculators (WolframAlpha among them) which contest this answer, and instead claim it is equal to:

enter image description here

Graphing this as an exponential function was not possible.

What's going on?

  • 2
    $\begingroup$ Careful! $\endgroup$
    – J.G.
    Sep 14 '19 at 18:51
  • 1
    $\begingroup$ Wolfram is taking the root before raising it to the 16th power. The principle 100th root of $-1$ has an angle of $180^o/100 = 1.8^o$. Then sixteen of these angles angles added together gives you $28.8^o$ $\endgroup$
    – David P
    Sep 14 '19 at 18:52
  • 1
    $\begingroup$ For most people (who know about the nuances involved), the meaning of $(-1)^{2.16}$ depends on whether $2.16$ is interpreted as a real number or as a specific fraction (presumably $\frac{54}{25},$ and not $\frac{216}{200}$ or $\frac{108}{50}).$ $\endgroup$ Sep 14 '19 at 18:54
  • 2
    $\begingroup$ How do you define a non-integer power of a negative number? For positive reals and rational powers there is a unique positive real root, and a real function can be defined. $0^0$ is notoriously problematic and you might want to think about negative powers of zero . Where the power is $\frac pq\gt 0$ and $q$ is odd, the function can be extended to the reals. $\endgroup$ Sep 14 '19 at 19:01
  • 2
    $\begingroup$ It's a matter of definition. If $q = \frac ab; a\in \mathbb Z; b\in \mathbb N; \gcd(a,b) = 1$ and if $k < 0$ then is it a valid definition to define $k^q $ as $\sqrt[b]{k^a}$? It can be argued that is valid if $a$ is even but I'd say it has too many issues and handwaving and is inconsistant when we do bring in complex analysis to be acceptable. I'd say that definition holds if $k > 0$ but if $k < 0$ we will need to consider principal complex roots. By squaring $(-1)$ you are deliberating and artificially avoiding them. I'd say that is no good. $\endgroup$
    – fleablood
    Sep 14 '19 at 19:01


is a set of $25$ numbers corresponding to $k=0,...24$ as the exponential above has period $25$.

Choosing $k=12$ shows that $1$ is indeed in this set, though it doesn't correspond to the usual "principal" value which is for $k=0$ and which in this case, gives $\exp({\frac{4}{25}i\pi})$ and this is what Wolfram Alpha gave

Edit later - just to make it clear, here $\exp(z)=\sum{\frac{z^n}{n!}}$ is the uniquely defined usual entire exponential function


If we stay strictly in the real domain, then exponentiation is defined by the formula $x^y=\exp(y\ln(x))$.

Since the logarithm is defined only on positive numbers, exponentiation is also well defined only on positive bases $x$.

With this definition $(-1)^{2.16}$ is undefined.

However, since it also comes from rational exponentiation by continuity, which is merely an extension of integer exponentiation where things like $(-1)^n$ are perfectly defined, it is possible to extend exponentiation to negative real numbers as well, but with limitation to rational exponents.

If we call $\mathbb Q_{odd}=\{\frac pq\in\mathbb Q\mid p\in\mathbb Z, q\in\mathbb N^*, \gcd(p,q)=1\text{ AND } q\text{ odd}\}$

Then for $x<0$ we have $x^r$ defined for any $r\in\mathbb Q_{odd}$ by $$x^r=x^{\frac pq}=\sqrt[q]{x^p}=\left({\sqrt[q]{x}}\right)^p$$

Since the $q-$root is an odd function, it is perfectly defined for negative numbers.

Note that we explicitly require that $\gcd(p,q)=1$ to avoid contradictions like $-1=(-1)^{\frac 13}=(-1)^{\frac 26}=\sqrt[6]{(-1)^2}=1$

Considering this $(-1)^{2.16}=(-1)^{\frac {54}{25}}=1$ would be a reasonable answer.

Now it is also possible to take some high, and examine the situation on the complex domain.

$z_1^{z_2}=\exp(z_2\ln(z_1))$ is defined by the same formula that in the real case, but this time the logarithm has an extended definition.

The price to pay is that since $\exp$ is a $2i\pi-$periodic function, its inverse, the logarithm becomes multi-valued, meaning there will be more than one value per arithmetic expression.

We define $$\ln(z)=\ln|z|+i(\arg(z)+2k\pi)\quad k\in\mathbb Z$$

Let's examine the consequences for $(-1)^{2.16}$

$\ln(-1)=\underbrace{\ln(1)}_0+i(\underbrace{\arg(-1)}_{\pi\text{ from }e^{i\pi}=-1}+2k\pi)=(2k+1)i\pi$

Also since $2.16$ is in fact rational $\frac {54}{25}$ as we already seen, we won't get an infinite number of values for the expression, yet still $25$ different ones.

Let's call them for $k=0\cdots24$ $$z_{[k]}=\exp(\frac{54(2k+1)}{25}i\pi)=\exp(\frac{108k}{25}i\pi+\frac{54}{25}i\pi)=z_{[0]}\,\phi^k$$

With $\begin{cases}z_{[0]}=e^{(2.16i\pi)} & \text{is called the principal value at }k=0\\\phi=\exp(\frac{108}{25}i\pi) & \text{is a cyclic multiplicative factor (in this case a rotation)}\end{cases}$

As noted by Conrad, when $k=12$ then $z_{[12]}=1$ the answer found in the rational exponentiation case is effectively among the results, although it is not the principal value.


The problem is that the exponential function: $$f:\mathbb{R}\rightarrow\mathbb{R}, f(x) = ab^x$$ is only defined for $b\in(0,\infty)$ and $a\neq0$ so $(-1)^x$ is not an exponential function, thus it doesn't have its proprieties.

On the other hand $(-1)^{2.16} = e^{2.16\log(-1)} = e^{2.16\pi i} = \cos(2.16\pi) + i \sin(2.16\pi)$ which gives the same value as WolframAlpha.

The graphing is not working because the function $(-1)^x = \cos(x\pi) + i \sin(x\pi)$ takes real values $\iff \sin(x\pi) = 0 \iff x\in\mathbb{Z}$.


I don't think you are allowed to split negative numbers up like that. Since $-1=e^{i\pi}$ then $$(-1)^{2.16}=e^{i\pi\cdot 2.16}$$ which is not a real number.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.