0
$\begingroup$

I'm not really showing how to prove this, since I tried finding the $x$-intercepts/zeros of $f(x) = \sqrt x - \log_{2} x$ , and see that $x = 4, 16$ work but inspection, but I'm not sure how to ensure those are all the zeros.

Once I find the zeros I think I can check the intervals between the zeros and undefined points and do the one-point test in each interval (since if it's positive or negative for one point in the interval then it's positive or negative for all points in the interval).

From there I could prove if it's positive for all $x > 16$. Kindly please help.

$\endgroup$
12
  • 3
    $\begingroup$ I would just look at the derivative of $\sqrt x - \log_2 x$ and show it is positive for $x \gt 16$ MathJax hint: if you put a backslash before common functions you get the right font and spacing, so \log x gives $\log x$ instead of log x which gives $log x$ $\endgroup$ Commented Jun 26 at 2:55
  • $\begingroup$ @RossMillikan Ohh I see. Tysm! But speaking of the zeros of f(x) (just edited the question to clarify what I initially meant by "zeros"), how would I confirm I found all the zeros for such a function? $\endgroup$
    – Bob Marley
    Commented Jun 26 at 3:19
  • 1
    $\begingroup$ You can find it from my answer on math.stackexchange.com/questions/4937833/… on your previous question. $\endgroup$
    – zkutch
    Commented Jun 26 at 3:42
  • 1
    $\begingroup$ This answered your question How do you solve this equation $ log_{2}(x) = \sqrt x$?. Please do not repost. $\endgroup$ Commented Jun 26 at 4:58
  • 2
    $\begingroup$ The down-vote culture on this site is really a drag. It is way over-used (this is a trend from the last couple of years). People want better questions, they should comment on them. The OP is clearly responsive. $\endgroup$
    – Brian Tung
    Commented Jun 26 at 17:02

6 Answers 6

1
$\begingroup$

Define $$f(x) = \sqrt{x} - \log_2 x, \quad x \in (0,\infty). \tag{1}$$ Then $f$ has no discontinuities and is smooth, because $\sqrt{x}$ and $\log_2 x$ are continuous and smooth on this interval. The derivative is $$f'(x) = \frac{1}{2\sqrt{x}} - \frac{1}{x \log 2} = \frac{\sqrt{x} \log 2 - 2}{2 x \log 2}. \tag{2}$$ (Note: for the purposes of this discussion, when the base of the logarithm is unspecified, it is taken to be the natural logarithm.) Consequently, $f$ has the unique critical point $$x = x_0 = \frac{4}{\log^2 2}. \tag{3}$$ As the numerator of $(2)$ is positive when $x > x_0$ and negative when $x < x_0$, it follows that $f$ is increasing on $x \in (x_0 , \infty)$, and decreasing when $x \in (0, x_0)$. Since $$f(x_0) = \frac{2}{\log 2} - 2 \log_2 \frac{2}{\log 2} \approx -0.172143 < 0, \tag{4}$$ and you have already located the zeroes $f(4) = 4(16) = 0$, it follows that these roots must be unique.

$\endgroup$
1
  • $\begingroup$ You may slightly simplify the question to $g(x)=x-2\log_2x$ for $x>4.$ $\endgroup$ Commented Jun 26 at 12:48
1
$\begingroup$

Let $y = \log_{2} \left( 16 x \right) - \sqrt {16 x}$ for all $x > 0$. It is easy to see that $y = \log_{2} \left( x \right) - 4 \left( \sqrt {x} - 1 \right)$. Take its derivative: $\text {d} y / \text {d} x = \log_{2} \left( e \right) / x - 2 \, / \sqrt {x}$, so $\text {d} y / \text {d} x \ge 0$ if and only if $\log_{2} \left( e \right) \ge 2 \sqrt {x}$. However, we know that for all $x > 1$, $2 \sqrt {x} > 2$; meanwhile, $e < 4$, so $\log_{2} \left( e \right) < 2$. In summary, $\log_{2} \left( e \right) < 2 \sqrt {x}$ for all $x > 1$.

So $\text {d} y / \text {d} x < 0$ for all $x > 1$. Now $y = 0$ at $x = 1$, we swiftly conclude that $y < 0$ for all $x > 1$; that is, $\log_{2} \left( x \right) < \sqrt {x}$ for all $x > 16$. $\blacksquare$

$\endgroup$
5
  • $\begingroup$ Sorry kindly could you further elaborate on the change of variable from 16x to x in your argument? I have a hard time understanding/feeling comfortable w/ change of variable stuff like this. $\endgroup$
    – Bob Marley
    Commented Jun 26 at 4:26
  • $\begingroup$ @BobMarley, let $u = 16 x$. Then $\log_{2} \left( 16 x \right) < \sqrt {16 x}$ for all $x > 1$ mathematically equates to $\log_{2} \left( u \right) < \sqrt {u}$ for all $u > 16$. Now, clearing memory, redefine that $u = x$. $\endgroup$
    – Simon
    Commented Jun 26 at 4:32
  • $\begingroup$ Gotcha, thanks! $\endgroup$
    – Bob Marley
    Commented Jun 26 at 4:59
  • $\begingroup$ Out of curiosity, could you explain your thought process of choosing this way of solving the problem with the whole 16x replacing x and all? This was quite a creative solution, and I was able to follow the proof/logic, but I can't quite see where it stems from. How did you conjure such a solution/proof basically? $\endgroup$
    – Bob Marley
    Commented Jun 26 at 6:45
  • $\begingroup$ @BobMarley, letting $u = 16 x$ highlights the case where $u = 16$, and it makes the calculations somewhat easier, and that is desirable. This is the motivation behind the substitution. As you solve more problems and read more answers from others, you will build your own intuition as to what step to take. It's hard to explain how I knew $u = 16 x$ would work. $\endgroup$
    – Simon
    Commented Jun 26 at 7:37
1
$\begingroup$

First, observe that $e^2<9<16=2^4$ which implies $2<4\log 2$, also for $t\ge 16$ we get $$2\sqrt{t}<4\sqrt{t}\log 2$$ also, $4\le \sqrt t,\;$ so we get $$2\sqrt{t}<4\sqrt{t}\log 2\le t\log 2$$

then \begin{align*} \frac1{\log 2}\frac1t&<\dfrac12\dfrac1{\sqrt t}\quad \text{when } t\ge 16\\ \dfrac1{\log 2}\int_{16}^x\dfrac1t\,\mathrm dt&<\frac12\int_{16}^xt^{-1/2}\,\mathrm dt\quad \text{when } x> 16\\ \dfrac1{\log 2}\left(\log x-\log 16\right)&<\sqrt{x}-\sqrt{16}\\ \dfrac{\log x}{\log 2}-\dfrac{\log 16}{\log 2}&<\sqrt{x}-4\\ \log_2x-4&<\sqrt{x}-4 \end{align*}

Therefore $x>16\implies\log_2 x<\sqrt{x}$.

$\endgroup$
10
  • $\begingroup$ Sorry I'm not following how you got $2\sqrt{t}<4\log 2\le\sqrt{t}\log 2<t\log 2$ for $ t \geq 16$, especially the $2\sqrt{t}<4\log 2$ part since t can get arbitrarily large and then $2\sqrt{t}$ can exceed $4\log 2$, right? $\endgroup$
    – Bob Marley
    Commented Jun 26 at 3:25
  • $\begingroup$ I have edited my post, the term $4\log 2$ is not needed. $\endgroup$ Commented Jun 26 at 3:32
  • $\begingroup$ sorry still not following :). Also, how is sqrt(t) <= 4? $\endgroup$
    – Bob Marley
    Commented Jun 26 at 3:34
  • $\begingroup$ I'm specifically lost on this part of your answer: also for $t\ge 16$ we get $\sqrt{t}\le 4$ and $$2\sqrt{t}<\sqrt{t}\log 2<t\log 2$$ then \begin{align*} \frac1{\log 2}\frac1t&<\dfrac12\dfrac1{\sqrt t}\quad t\ge 16\\\end{align*} $\endgroup$
    – Bob Marley
    Commented Jun 26 at 3:38
  • $\begingroup$ when $x=16$, $\text{log}_2x=\sqrt x$; your final implication should be $x>16$. $\endgroup$ Commented Jun 26 at 4:39
1
$\begingroup$

Let $x=2^{2t+2}.$ Then the inequality takes the form $2^t>t+1,$ $t>1.$ For $t=1$ we get the equality. So it suffices to show that $f(t)=2^t-t$ is increasing for $t\ge 1.$ We have $f'(t)=(\log 2)\, 2^t-1.$ Thus $f'(t)$ is increasing. Hence $$f'(t)>f'(1)=2\log 2 -1=\log 4-1>0,\quad t>1$$

$\endgroup$
1
$\begingroup$

Consider the function $$f(x) = \frac{\log_2(x)}{\sqrt{x}}$$ Taking the derivative, $$f'(x) = \frac{2 - \ln(x)}{x^{3/2} \ln(4)}$$ We see $f'(x) < 0$ for $x >e^2$, so $f(x)$ is decreasing for $x > e^2$.

Since $16 > e^2$ and $f(16) = 1$, $f(x) < 1$ for $x > 16$.

$\endgroup$
0
$\begingroup$

By @zkutch (and edited by @5xum):

As you already wrote in the comments, $x=4$ and $x=16$ are both solutions, so this is just a proof that they are the only solutions.

Let's consider $f(x)=\sqrt{x} - \log_2 x$ function.

Because $$f'(x)=\frac{1}{2\sqrt{x}} - \frac{1}{x\ln 2}=\frac{\sqrt{x}\ln 2 - 2 }{2x \ln 2},$$ $f$ has a minimum at $x=\frac{4}{\ln^2 2}\approx 8.325$. From left of minimum on set $\left(0,\frac{4}{\ln^2 2}\right)$ $f$ is strictly decreasing and on set $\left(\frac{4}{\ln^2 2},+\infty \right)$ strictly increasing.

On other hand we know $f(4)=f(16)=0$ and as $4<\frac{4}{\ln^2 2}<16$ then on set $(4,16)$ function $f$ is strictly negative and on set $(0,4)\cup(16,+\infty)$ strictly positive.



By @BrianTung:

An alternative sledgehammer route is via the Lambert $W$ function, on both the principal branch $W_0$ and the other real branch $W_{-1}$.

$$ \log_2 x = \sqrt{x} $$ $$ 2\log_2 \sqrt{x} = \sqrt{x} $$ $$ 2\frac{\ln \sqrt{x}}{\ln 2} = \sqrt{x} $$ $$ \ln \sqrt{x} = \left(\frac12 \ln 2\right) \sqrt{x} $$ $$ \sqrt{x} = e^{\left(\frac12 \ln 2\right) \sqrt{x}} $$ $$ \sqrt{x} e^{\left(-\frac12 \ln 2\right) \sqrt{x}} = 1 $$ $$ \sqrt{x} e^{\left(\frac12 \ln \frac12\right) \sqrt{x}} = 1 $$ $$ \left(\frac12 \ln \frac12\right) \sqrt{x} e^{\left(\frac12 \ln \frac12\right) \sqrt{x}} = \left(\frac12 \ln \frac12\right) $$ $$ \left(\frac12 \ln \frac12\right) \sqrt{x} = W \left(\frac12 \ln \frac12\right) $$ $$ x = \left[ \frac{W \left(\frac12 \ln \frac12\right)}{\frac12 \ln \frac12} \right]^2 $$

(There's probably a faster way to get there, but I'm not the best Lambert manipulator.) Fortunately, we have an identity: When $z \geq 1/e, W_0(z \ln z) = \ln z$, so the principal branch $W(z) = W_0(z)$ gives us

$$ x = \left[ \frac{\ln \frac12}{\frac12 \ln \frac12} \right]^2 = 2^2 = 4 $$

What about the other value? That comes from the other branch, for which there is a corresponding identity: When $0 < z \leq 1/e, W_{-1}(z \ln z) = \ln z$. This would seem difficult to apply to the current equation, since $1/2 > 1/e$. However, note that we have the equality

$$ \frac12 \ln \frac12 = \frac14 \ln \frac14 $$

so we can now also obtain, from $W(z) = W_{-1}(z)$,

$$ x = \left[ \frac{W \left(\frac14 \ln \frac14\right)}{\frac14 \ln \frac14} \right]^2 $$ $$ x = \left[ \frac{\ln \frac14}{\frac14 \ln \frac14} \right]^2 = 4^2 = 16 $$

as desired.


To be sure, there's a certain unsatisfying roundabout-ness to this approach, but it's another way of tackling it.

$\endgroup$
3
  • $\begingroup$ @zkutch How'd you exactly prove f(x) is positive for x-interval $(0, 4)$? I can understand from the calc reasoning that since for x-interval $(0, \frac{4}{(\ln2)^2})$ $f(x)$ is strictly decreasing (that is from right to left relative to x-axis of graph), that means from left to right it'll be strictly increasing. Hence past the $f(4) = 0$ point, $f(x)$ is always positive as a result. Same idea for x-interval $(16, \infty)$... $\endgroup$
    – Bob Marley
    Commented Jun 26 at 16:34
  • $\begingroup$ @zkutch... However, I was trying to show this though inequality manipulations. Starting with $0 < x < 4 \iff 0 < \sqrt x < 2 \iff 0 < \log_2 x < 2$, we have $-\log_2 x > -2$ and $\sqrt x < 2$. Combing these 2 I get $\sqrt x - \log_2 x > \sqrt x - 2$, but all I have here as far as I can tell is $\sqrt x - 2 < 0$. Hence I'm stuck. Similar issue I come across for x-interval $x > 16$. Kindly if you could help me with this alternative method (and also clarify if my calc reasoning was correct here) I'd be grateful. $\endgroup$
    – Bob Marley
    Commented Jun 26 at 16:45
  • $\begingroup$ @zkutch also (2nd follow up questions), I just want to confirm how you showed the function was strictly increasing/decreasing to the left and right, respectively, of the minimum (basically an inequality question too). Also, how'd you show the 2nd derivative was positive to hence show it was indeed a minimum (also inequality question)? $\endgroup$
    – Bob Marley
    Commented Jun 26 at 16:45

Not the answer you're looking for? Browse other questions tagged .