Characterization of Euler's Number by Inequality

From ProofWiki
Jump to navigation Jump to search

Theorem

Let $a$ be a (strictly) positive real number.


Then:

$a = e \iff \forall x \in \R: a^x \ge x + 1$

where $e$ denotes Euler's number.


Proof

Forward Implication

Proved in Exponential Function Inequality.

$\Box$


Reverse Implication

Consider $\map {f_a} x = a^x - x - 1$.

Then we need to prove:

$a = e \impliedby \forall x \in \R: \map {f_a} x \ge 0$


By Linear Combination of Derivatives, Derivative of Power of Constant, Derivative of Identity Function, and Derivative of Constant:

$\map {f'_a} x = a^x \ln a - 1$


By Linear Combination of Derivatives, Derivative of Power of Constant, Derivative of Constant Multiple, and Derivative of Constant:

$\map {f_a} x = a^x \paren {\ln a}^2$


Now, we divide the theorem into two cases:


Case 1: $0 < a \le 1$

Consider $x = 1$.

Then:

\(\ds \map {f_a} 1\) \(=\) \(\ds a^1 - 1 - 1\)
\(\ds \) \(=\) \(\ds a - 2\)
\(\ds \) \(\le\) \(\ds 1 - 2\)
\(\ds \) \(=\) \(\ds -1\)
\(\ds \) \(<\) \(\ds 0\)

Thus the right hand side is not true.

$\Box$


Case 2: $a > 1$

We have:

\(\ds \ln a\) \(>\) \(\ds 0\) Logarithm of 1 is 0 and Logarithm is Strictly Increasing
\(\text {(1)}: \quad\) \(\ds \leadsto \ \ \) \(\ds \paren {\ln a}^2\) \(>\) \(\ds 0\) Square of Non-Zero Real Number is Strictly Positive
\(\text {(2)}: \quad\) \(\ds a^x\) \(>\) \(\ds 0\) Power of Positive Real Number is Positive over Reals
\(\ds \vdash \ \ \) \(\ds a^x \paren {\ln a}^2\) \(>\) \(\ds 0\) Real Number Axiom $\R \text O2$: Usual Ordering is Compatible with Multiplication, using $(1)$ and $(2)$
\(\ds \map {f_a} x\) \(>\) \(\ds 0\)

Hence, by Real Function with Strictly Positive Second Derivative is Strictly Convex, we have that $f_a$ is strictly convex.


Trying to find its minimum, we solve:

$\map {f'_a} x = 0$

Therefore:

\(\ds a^x \ln a - 1\) \(=\) \(\ds 0\)
\(\ds a^x \ln a\) \(=\) \(\ds 1\)
\(\ds a^x\) \(=\) \(\ds \frac 1 {\ln a}\)
\(\ds x\) \(=\) \(\ds \map {\log_a} {\frac 1 {\ln a} }\)
\(\ds x\) \(=\) \(\ds - \map {\log_a} {\ln a}\)


At this point:

\(\ds \map {f_a} x\) \(=\) \(\ds \map {f_a} {- \map {\log_a} {\ln a} }\)
\(\ds \) \(=\) \(\ds \frac 1 {\ln a} + \map {\log_a} {\ln a} - 1\)
\(\ds \) \(=\) \(\ds \frac 1 {\ln a} + \frac {\map \ln {\ln a} } {\ln a} - 1\) Change of Base of Logarithm
\(\ds \) \(=\) \(\ds \frac {1 + \map \ln {\ln a} } {\ln a} - 1\)


It is to be noted that when $a = e$, the minimum is:

$\dfrac {1 + \map \ln {\ln e} } {\ln e} - 1 = 0$

meaning that the only solution to $\map {f_e} x = 0$ is $x = 0$.


Also, from Exponential Function Inequality, we have:

$\forall x \in \R: e^x \ge x + 1$

Substituting $x = \map \ln {\ln a}$:

$\ln a \ge \map \ln {\ln a} + 1$

From Logarithm is Strictly Increasing and Logarithm of 1 is 0:

$\ln a > 0$

Thus:

$\dfrac {\ln a} {\ln a} \ge \dfrac {\map \ln {\ln a} + 1} {\ln a}$

Therefore:

$\dfrac {1 + \map \ln {\ln a} } {\ln a} - 1 \le 0$


Then, we solve the case of equality:

\(\ds \frac {1 + \map \ln {\ln a} } {\ln a} - 1\) \(=\) \(\ds 0\)
\(\ds \frac {1 + \map \ln {\ln a} } {\ln a}\) \(=\) \(\ds 1\)
\(\ds 1 + \map \ln {\ln a}\) \(=\) \(\ds \ln a\)

From earlier, we have that the only solution is $\map \ln {\ln a} = 0$, when $a = e$.


Therefore, for other values of $a$, the minimum is negative.

$\blacksquare$