15

The following is an MCQ question, one should answer it without a calculator, $\color{red}{\text{within $3$ minutes}}$.


$\text{Consider the expression}$

$$I(a,b)=\int_{a}^{b}\sqrt{1+x+x^2+x^3+x^4}\text{ d}x.$$

$\text{Which of the following is true?}$

$\space\space\space\space\space\text{I.}$ $I(2,3)+I(3,4)=I(4,5)$

$\space\space\space\space\space\text{II.}$ $I(2,3)+I(3,4)<I(4,5)$

$\space\space\space\space\space\text{III.}$ $16<I(4,5)<25$

$\text{(A) }$ $\text{I only}$

$\text{(B) }$ $\text{II only}$

$\text{(C) }$ $\text{III only}$

$\text{(D) }$ $\text{I and III}$

$\text{(E) }$ $\text{II and III}$


My Attempt (which, I think, is not a feasible way since it is not that accurate and also time consuming):

$\sqrt{1+x+x^2+x^3+x^4}\bigg|_{x=2}=\sqrt{1+2+2^2+2^3+2^4}=\sqrt{1+2+4+8+16}=\sqrt{31}\approx 6$

and

$\sqrt{1+x+x^2+x^3+x^4}\bigg|_{x=5}=\sqrt{1+5+5^2+5^3+5^4}=\sqrt{1+5+25+125+625}=\sqrt{781}\approx 28$

Now considering the integrand, approximately, as the straight line joining the points $(2,6)$ and $(5,28)$. The equation of the line can be found by first determining the slope, $m$:

$$m=\frac{28-6}{5-2}=\frac{22}{3}\approx 7$$

The equation of the line is therefore

$$y-6=7(x-2) \implies y=7x-8$$

now $I(a,b) \approx \int_{a}^{b} (7x-8)\text{ d}x=\frac{7}{2}x^2-8x \bigg|_{x=a}^{x=b} = \dots$

and $I(2,3) \approx \int_{2}^{3} (7x-8)\text{ d}x=9.5$

and $I(3,4) \approx \int_{3}^{4} (7x-8)\text{ d}x=16.5$

and $I(4,5) \approx \int_{4}^{5} (7x-8)\text{ d}x=23.5$

From these approximations, we can say, with low confidence, that $\text{(C)}$ is the correct option, which is not. In fact $\text{(E)}$ is the correct option. Noting that we can do better approximations, but that will cost time.


Attempt $\text{#}2$ (Not sure if it is a correct way):

In the interval $[2,5]$, function $f(x)=\sqrt{1+x+x^2+x^3+x^4}$ is always (positive), is always (increasing), and is always (concaving up). In that case, the inequalities will be the same when we do not consider the radical. So simply we evaluate:

$$\int_{2}^{3}(1+x+x^2+x^3+x^4)\text{ d}x$$

which gives $(x+x^2/2+x^3/3+x^4/4+x^5/5) \bigg|_{x=2}^{x=3}\approx 68$

Similarly, we evaluate:

$$\int_{3}^{4}(1+x+x^2+x^3+x^4)\text{ d}x \text{ and } \int_{4}^{5}(1+x+x^2+x^3+x^4)\text{ d}x$$

which give $\approx 217$ and $\approx 538$, respectively.

Clearly $68+217 < 538$. Hence II is true.

AND

$16^2 = 256 < 538 < 625 = 25^2$. Hence III is true.

Well, this attempt may also require time, but still it would be nice to know if it is a correct way. So, the approach (which might be wrong) is to observe that; between the limits of integration, the integrand is positive, increasing, and concaving up, the we deal with it without the radical. If this is wrong, please support by giving a counter-example.


Your help would be appreciated. Thanks!

Hussain-Alqatari
  • 5,738
  • 3
  • 16
  • 44
  • My first thought is to bound the inside of the square root by some easier-to-compute expressions that don't involve square roots. – Benjamin Wang Aug 18 '23 at 10:46
  • 2
    Protip: write integrand as $$\sqrt{1+x+x^2+x^3+x^4} = \sqrt{\frac{x^5-1}{x-1}}$$ –  Aug 18 '23 at 10:47
  • @LuckyChouhan nice observation indeed, but is the resulting integration-by-parts actually easier? – Benjamin Wang Aug 18 '23 at 11:07
  • @BenjaminWang oh no! but I was thinking to approximate as he did in his attempt. I don't think here we should use any tactics of solving integrals. Just use $I(a,b) \approx I(b) - I(a)$. –  Aug 18 '23 at 12:03
  • 1
    I tried to bound $f(x):=\sqrt{1+x+x^2+x^3+x^4}$ using $g(x):=a+bx+x^2$, but doing this naively doesn't work. For example, the following is not tight enough: the upper bound $(a,b)=(\frac34,\frac12)$, and the lower bound $(a,b)=(\frac38,\frac12)$. Note that for the upper bound I needed to use $x^2 > x > 1$. – Benjamin Wang Aug 18 '23 at 16:17
  • 1
    it might be worth putting on a bounty to attract a method that is doable within 3 minutes – Benjamin Wang Aug 24 '23 at 19:09
  • @BenjaminWang before that, I liked to add Attempt 2. Can you review it? – Hussain-Alqatari Aug 24 '23 at 22:45
  • @Hussain-Alqatari see Angelo's comments under Amadeus's answer. I don't think you can remove the radicals like you did in Attempt 2. – Benjamin Wang Aug 25 '23 at 18:48

6 Answers6

6

First of all we will prove that for any $\,x\geqslant\sqrt2\,$ it results that

$\left(\dfrac12x+x^2\right)^{\!2}\!<1+x+x^2+x^3+x^4<\left(x+x^2\right)^2\,.\qquad\color{blue}{(1)}$

Proof :

For any $\,x\in\Bbb R\,$ it results that

$\begin{align}\left(\!\dfrac12x\!+\!x^2\!\right)^{\!2}\!\!&=\!\dfrac14x^2\!+\!x^3\!+\!x^4\!<\!\left(\!1\!+\!\dfrac12x\!\right)^{\!2}\!\!+\!\dfrac12x^2\!+\!\dfrac14x^2\!+\!x^3\!+\!x^4=\\[3pt]&=1\!+\!x\!+\!x^2\!+\!x^3\!+\!x^4\;.\end{align}$

Moreover, for any $\,x\geqslant\sqrt2\,$ it results that

$\begin{align}1\!+\!x\!+\!x^2\!+\!x^3\!+\!x^4&<2x\!+\!x^2\!+\!x^3\!+\!x^4\leqslant x^2\!\!\cdot\!x\!+\!x^2\!+\!x^3\!+\!x^4=\\[3pt]&=x^2\!+\!2x^3\!+\!x^4=\left(x\!+\!x^2\right)^2\;.\end{align}$


Now we will prove that for any $\,x\geqslant2\,$ it results that

$\dfrac38<\sqrt{1+x+x^2+x^3+x^4}-\dfrac12x-x^2 <\dfrac35\;.\quad\color{blue}{(2)}$

Proof :

For any $\,x\geqslant\sqrt2\,$ it results that

$\sqrt{1+x+x^2+x^3+x^4}-\dfrac12x-x^2=$

$=\dfrac{1+x+x^2+x^3+x^4-\left(\dfrac12x+x^2\right)^{\!2}}{\sqrt{1+x+x^2+x^3+x^4}+\dfrac12x+x^2}=$

$=\dfrac{1+x+\dfrac34x^2}{\sqrt{1+x+x^2+x^3+x^4}+\dfrac12x+x^2}\underset{\overbrace{\;\text{ by using }(1)\;}}{>}$

$>\dfrac{1+x+\dfrac34x^2}{\sqrt{\left(x+x^2\right)^2}+\dfrac12x+x^2}=\dfrac{1+x+\dfrac34x^2}{\dfrac32x+2x^2}>$

$>\dfrac{\dfrac9{16}x+\dfrac34x^2}{\dfrac32x+2x^2}=\dfrac{\dfrac38\left(\dfrac32x+2x^2\right)}{\dfrac32x+2x^2}=\dfrac38\;.$

Moreover, for any $\,x\geqslant2\,$ it results that

$\sqrt{1+x+x^2+x^3+x^4}-\dfrac12x-x^2=$

$=\dfrac{1+x+\dfrac34x^2}{\sqrt{1+x+x^2+x^3+x^4}+\dfrac12x+x^2}\underset{\overbrace{\;\text{ by using }(1)\;}}{<}$

$<\dfrac{1+x+\dfrac34x^2}{\sqrt{\left(\dfrac12x+x^2\right)^2}+\dfrac12x+x^2}=\dfrac{1+x+\dfrac34x^2}{x+2x^2}\leqslant$

$\leqslant\dfrac{1+x+\dfrac34x^2+\dfrac14\left(x^2-4\right)+\dfrac15x(x-2)}{x+2x^2}=$

$=\dfrac{\dfrac35x+\dfrac65x^2}{x+2x^2}=\dfrac{\dfrac35\left(x+2x^2\right)}{x+2x^2}=\dfrac35\;.$


From $\,(2)\,$ it follows that

$\sqrt{1\!+\!x\!+\!x^2\!+\!x^3\!+\!x^4}<\dfrac35\!+\!\dfrac12x\!+\!x^2\;,\;\;$ for any $\,x\geqslant2\;.\quad\color{blue}{(3)}$

$\sqrt{1\!+\!x\!+\!x^2\!+\!x^3\!+\!x^4}>\dfrac38\!+\!\dfrac12x\!+\!x^2 \;,\;\;$ for any $\,x\geqslant2\;.\quad\color{blue}{(4)}$


Now we will estimate the values of the definite integrals $\,I(2,3)+I(3,4)\,$ and $\,I(4,5)\,.$

$\begin{align}I(2,3)+I(3,4)&=\displaystyle\!\!\int_2^4\!\!\sqrt{1+x+x^2+x^3+x^4}\,\mathrm dx\!\!\underset{\overbrace{\;\text{ by using }(3)\;}}{<}\\[3pt]&<\int_2^4\!\left(\dfrac35+\dfrac12x+x^2\right)\mathrm dx=\\[3pt]&=\left[\dfrac35x+\dfrac14x^2+\dfrac13x^3\right]_2^4=\\[3pt]&=\dfrac{12}5+4+\dfrac{64}3-\dfrac65-1-\dfrac83=\dfrac{343}{15}\;.\end{align}$

$\begin{align}I(4,5)&=\displaystyle\!\!\int_4^5\!\!\sqrt{1+x+x^2+x^3+x^4}\,\mathrm dx\!\!\underset{\overbrace{\;\text{ by using }(4)\;}}{>}\\[3pt]&>\int_4^5\!\left(\dfrac38+\dfrac12x+x^2\right)\mathrm dx=\\[3pt]&=\left[\dfrac38x+\dfrac14x^2+\dfrac13x^3\right]_4^5=\\[3pt]&=\dfrac{15}8+\dfrac{25}4+\dfrac{125}3-\dfrac32-4-\dfrac{64}3=\dfrac{551}{24}\;.\end{align}$

Since $\;\dfrac{343}{15}<\dfrac{344}{15}=\dfrac{344\!\cdot\!8}{15\!\cdot\!8}=\dfrac{2752}{120}<\dfrac{2755}{120}=\dfrac{551}{24}$

it follows that

$I(2,3)+I(3,4)<\dfrac{343}{15}<\dfrac{551}{24}<I(4,5)\;.$

Hence, the statement II is true.

Moreover ,

$\begin{align}I(4,5)&=\displaystyle\!\!\int_4^5\!\!\sqrt{1+x+x^2+x^3+x^4}\,\mathrm dx\!\!\underset{\overbrace{\;\text{ by using }(3)\;}}{<}\\[3pt]&<\int_4^5\!\left(\dfrac35+\dfrac12x+x^2\right)\mathrm dx=\\[3pt]&=\left[\dfrac35x+\dfrac14x^2+\dfrac13x^3\right]_4^5=\\[3pt]&=3+\dfrac{25}4+\dfrac{125}3-\dfrac{12}5-4-\dfrac{64}3=\dfrac{1391}{60}\;.\end{align}$

Consequently ,

$16<22<\dfrac{551}{24}<I(4,5)<\dfrac{1391}{60}<24<25\,.$

Hence the statement III is true too.

Angelo
  • 13,952
  • 4
    By a long distance, this answer is the most accurate one to this question. – Sarvesh Ravichandran Iyer Aug 19 '23 at 06:55
  • 1
    For me, I got the same lower bound as you did more easily. Essentially, we are trying to bound $$f(x):=x^4+x^3+x^2+x+1$$ by $$g(x; b, a) := (x^2+bx+a)^2 = x^4 + 2bx^3+(b^2+2a)x^2+2bax+a^2.$$ After substituting $(b, a) = (1/2, 3/8)$ we get that the $x^3$ and $x^2$ terms all have coefficient $1$, while the $x$ term has coefficient $2(1/2)(3/8) < 1$ and the constant term is $(3/8)^2<1$. So for $x\ge 0$ we have $$x^2+\frac12x+\frac38 < \sqrt{f(x)}.$$ Sadly, I had a lot of trouble getting a good upper bound like you did. Anyway, my method might help provide a solution that is closer to "3 minutes". – Benjamin Wang Aug 20 '23 at 17:23
2

Three minutes is a short time. What we need is some simple, plausible reasoning and a bit of luck. Here is an attempt.

Using the finite geometric series formula we have \begin{align*} \int_{a}^{b}\sqrt{1+x+x^2+x^3+x^4+x^5}\,dx=\int_{a}^b\sqrt{\frac{x^5-1}{x-1}}\,dx \end{align*}

We consider $f(x)=\sqrt{\frac{x^5-1}{x-1}}$ and calculate for $x\in\{2,3,4,5\}$ \begin{align*} \color{blue}{f(2)}&=\sqrt{31}\color{blue}{\approx 5.5}\tag{$5<\sqrt{31}<6$}\\ \color{blue}{f(3)}&=\sqrt{121}\color{blue}{=11}\\ \color{blue}{f(4)}&\approx\sqrt{\frac{1}{3}2^{10}}=\sqrt{\frac{1024}{3}}\color{blue}{\approx 18}\tag{$18^2=324$}\\ \color{blue}{f(5)}&\approx\sqrt{\frac{1}{4}5^5}\approx\frac{1}{2}25\sqrt{5}\color{blue}{\approx 28}\tag{$\sqrt{5}\approx 2.2$} \end{align*}

Now we do some linear approximation \begin{align*} \color{blue}{I(n,n+1)}&\approx f(n)+\frac{1}{2}\left(f(n+1)-f(n)\right)\color{blue}{=\frac{1}{2}\left(f(n+1)+f(n)\right)}\\ \\ I(2,3)&\approx\frac{1}{2}\left(11+5.5\right)=8.25\\ I(3,4)&\approx\frac{1}{2}\left(18+11\right)=14.5\\ I(4,5)&\approx\frac{1}{2}\left(28+18\right)=23\\ \end{align*}

We obtain \begin{align*} \color{blue}{I(2,3)+I(3,4)}&\approx 8.25+14.5=22.75\color{blue}{< 23\approx I(4,5)}\tag{II.}\\ \color{blue}{16}&\color{blue}{<23\approx I(4,5)<25}\tag{III.} \end{align*} and conclude $\color{blue}{\mathrm{(E)}}$ is the solution.

Markus Scheuer
  • 112,413
  • 1
    In that way, you cannot deduce that $I(2,3)+I(3,4)<I(4,5)$ because, since the function $f(x)$ is convex on $[2,5]$, it follows that $I(2,3)!+!I(3,4)<\dfrac{f(2)!+!f(3)}2!+!\dfrac{f(3)!+!f(4)}2=23.01697…$ and $I(4,5)<\dfrac{f(4)+f(5)}2=23.20628…$ You only can deduce that $I(2,3)!+!I(3,4)<23.20628…>I(4,5)$, but it does not imply that $I(2,3)!+!I(3,4)<I(4,5),.;$ – Angelo Aug 19 '23 at 08:50
  • @Angelo: A little difficult to see without a calculator. – Markus Scheuer Aug 19 '23 at 09:01
  • 1
    For that reason we cannot estimate a number if we cannot find a lower bound and an upper bound. In any case, even if the numbers are not approximately equal (for example $6$ and $8$), from $I(2,3)+I(3,4)<6$ and $I(4,5)<8$, it is impossible to compare $I(2,3)+I(3,4)$ and $I(4,5)$, in fact you need to find a lower bound for $I(4,5)$ which is greater than $6$. – Angelo Aug 19 '23 at 09:28
2

Remark: The following can not be done during three minutes.

First, III is true.

We have $$I(4, 5) \ge \int_4^5 x^2\,\mathrm{d} x = \frac{61}{3} > 16,$$ and $$I(4, 5) \le \int_4^5 (x^2 + x/2 + 1)\,\mathrm{d} x = \frac{283}{12} < 25$$ where we use $$(x^2 + x/2 + 1)^2 = x^4 + x^3 + (2 + 1/4)x^2 + x + 1 \ge 1 + x + x^2 + x^3 + x^4.$$


Second, II is true.

Note that $f(x) := \sqrt{1 + x + x^2 + x^3 + x^4}$ is convex on $x \ge 0$. We have $I(2, 3) \le \frac12[f(2) + f(3)]$ and $I(3, 4) \le \frac12[f(3) + f(4)]$. Thus, we have $$I(2, 3) + I(3, 4) \le \frac12 f(2) + f(3) + \frac12 f(4) = \frac12\sqrt{31} + 11 + \frac12\sqrt{341}.$$

To find a lower bound of $I(4, 5)$, note that \begin{align*} 1 + x + x^2 + x^3 + x^4 &= (x^2 + x/2)^2 + \frac34 x^2 + x + 1\\ &= (x^2 + x/2)^2 + \frac34 (x^2 + x/2) + \frac58 x + 1\\ &= (x^2 + x/2 + 3/8)^2 + \frac58 x + \frac{55}{64}. \end{align*} We hope to find a constant $c_1 > 3/8$ such that $$ (x^2 + x/2 + 3/8)^2 + \frac58 x + \frac{55}{64} \ge (x^2 + x/2 + c_1)^2, \quad \forall x \in [4, 5]$$ that is $$g(x) := \frac58 x + \frac{55}{64} - (c_1 - 3/8)(2x^2 + x + 3/8 + c_1)\ge 0, \quad \forall x \in [4, 5]. \tag{1}$$ Since $g(x)$ is concave. (1) is equivalent to $g(4) \ge 0, g(5) \ge 0$. We have $c_1 \le -\frac{55}{2} + \sqrt{781}$. We choose $c_1 = 4/9$. We have $$\sqrt{1 + x + x^2 + x^3 + x^4} \ge x^2 + x/2 + 4/9, \quad \forall x \in [4, 5].$$

We have $$I(4, 5) \ge \int_4^5 (x^2 + x/2 + 4/9)\,\mathrm{d} x = 829/36.$$

Thus, $$I(4, 5) - I(2, 3) - I(3, 4) \ge 829/36 - \left(\frac12\sqrt{31} + 11 + \frac12\sqrt{341}\right) > 0.$$

We are done.

River Li
  • 49,125
  • Good answer ! I like it very much ! Do you think it is possible to find a way of solving the exercise in just three minutes ? I cannot understand how to do it. Could you try to find a way ? – Angelo Aug 19 '23 at 19:13
  • @Angelo I cannot find it. It should be a tricky way. – River Li Aug 19 '23 at 22:44
1

The trick is that $\sqrt{1+x+x^2+x^3+x^4} = x^2 \sqrt{\frac1{x^4}+\frac1{x^3}+\frac1{x^2}+\frac1x +1}$ which is approximately $x^2$.

For the first two statements we need to compare $I(2,4)$ with $I(4,5)$ and we see that with this approximation that $I(4,5)-I(2,4) = \frac{61}3-\frac{56}3 = \frac53$ which we can see is almost $2$. This gives us II. as true. For the final question we noted that $\frac{56}3 = 18 \frac13$ so III. is true as well.

CyclotomicField
  • 11,925
  • 1
  • 14
  • 32
  • 3
    I don't think it's so obvious that the approximation is good enough that you could use it directly, without trying to create a upper bound as well – Amadeus Aug 19 '23 at 04:04
  • Your approximation is not good, actually $I(4,5)-I(2,4)\approx0.36$ which is completely different from what you extimated. Also see this picture where it is clear that $x^2$ is not a good approximation. – Angelo Aug 19 '23 at 04:20
  • @Amadeus it just can't be close to $1$, and it's close to $2$ and we're good to go. – CyclotomicField Aug 19 '23 at 12:37
  • @Angelo It's good enough to determine that $I(4,5) > I(2,4)$ and $16 < I(4,5) < 25$ since the bounds are so large. Note that we don't need to consider values near $0$ at all so your graph is overdrawn. It's accuracy is better if you restrict yourself to the interval $[2,5]$. The context clue in the question is $16 < I(4,5) < 25$ also strongly suggests this approximation. With no calculator, graphs, and a 3 minute time limit I think it's ideal. – CyclotomicField Aug 19 '23 at 12:48
  • 1
    @CyclotomicField, $I(4,5)-I(2,4)$ is not $\frac53$, but $\approx0.36$. Actually $I(2,4)\approx22.68$ is very close to $I(4,5)\approx23.04$ and your function $g(x)=x^2$ is not a good approximation to prove that $I(2,4)<I(4,5)$. Look at this picture and pay close attention to it. Do you really think $g(x)=x^2$ is a good approximation? Also note the great difference between the integral of the function $f(x)$ and the integral of your approximation $g(x)=x^2$. It is much greater than the small difference between $I(2,4)$ and $I(4,5)$. – Angelo Aug 19 '23 at 13:57
  • @Angelo Approximations need to be evaluated against the allowed tolerance. Getting $18$ instead of $23$ doesn't matter if I just need to know it's between $16$ and $25$. Unless you have an alternative estimate that works and is simpler than $x^2$ I'd argue it's optimal. – CyclotomicField Aug 19 '23 at 17:20
  • 2
    @CyclotomicField, getting $18$ instead of $23$ doesn't matter if we just need to know it's between $16$ and $25$, but with your approximation we cannot prove that $I(2,4)<I(4,5)$ because the difference of this two definite integrals is very small. – Angelo Aug 19 '23 at 18:57
0

For III, $\int_{4}^{5}\sqrt{x^4}\text{ d}x<\int_{4}^{5}\sqrt{1+x+x^2+x^3+x^4}\text{ d}x<\int_{4}^{5}\sqrt{(x^2+\frac{x}{2}+1)^2}\text{ d}x$ It's a very good approximation because the first two largest terms are conserved as is, on solving you should get a range for III somewhere between $(19,24)$.

Amadeus
  • 1,139
  • Actually $$\int_2^4f^2(x)dx<\int_4^5f^2(x)dx$$ does not imply $$\int_2^4f(x)dx<\int_4^5f(x)dx;.$$ A counterexample is $f(x)=x^2+10$. Also see this picture. – Angelo Aug 19 '23 at 04:53
  • @Angelo oh whoops, yeah the interval is different I forgot – Amadeus Aug 19 '23 at 06:04
  • The interval length I mean – Amadeus Aug 19 '23 at 06:17
  • Amadeus, if the function $f(x)$ is not increasing, it is possible to get counterexamples even though the two intervals have the same length. A counterexample is any continuous function $f:[0,3]\to\Bbb R$ such that $$f(x)=\begin{cases}4+x^2&\text{if }x\in[0,1]\6x-11&\text{if }x\in[2,3]\end{cases}$$ Even there are counterexamples for sums of only two numbers: $4^2+5^2<1^2+7^2$, but $4+5>1+7,.;$ – Angelo Aug 19 '23 at 08:10
  • @Angelo Outside of the interval mistake, what I said was not wrong just incomplete, which was intended as it was supposed to be a short answer with gaps the reader would fill in themselves. What I claimed was, for the given $f$ in the given interval, if $f(x_1)<f(x_2)$ we also have $(f(x_1))^2<(f(x_2))^2$ – Amadeus Aug 19 '23 at 08:54
  • You also wrote that it was sufficient to compare the integrals of $f^2(x)$ but actually in that way we cannot deduce anything about the integrals of $f(x)$. – Angelo Aug 19 '23 at 08:59
  • @Angelo this is trivially true. How you get from there to only comparing the integral of the squares is by taking $x_1$s in $(2,4)$ and $x_2$s in $(4,5)$ and noticing that $f(x_1)<f(x_2)$ is always true. I assume going further requires some real analysis that I don’t know, but its not hard to see why it will be true – Amadeus Aug 19 '23 at 08:59
  • 1
    For this reason, I wrote in my previous comment “ if the function is not increasing, …”. It obvious that if the function is increasing and the two intervals have the same length, then you can deduce that the integral on the interval on the left is less than the integral on the interval on the right (we do not need to square the function). – Angelo Aug 19 '23 at 09:07
0

This can be done in 3 minutes if you know the dirty method :

All body knows that $f(x)=x^2$ is convex and the inverse function of $f$ is concave so we have :

$$\sqrt{\frac{x^5-1}{x-1}}\geq \sqrt{f'(a^2)\left(\sqrt{\frac{x^5-1}{x-1}}-a^2\right)+f(a^2)}$$

A lower bound can be find with $f^{-1}(x)=\sqrt{x}$

Barackouda
  • 3,879