Navigation

Concept Review

Differential Calculus

Functions, Graphs and Growth Rates

Differential calculus studies how quantities change, and every derivative you will ever compute is the derivative of a function. So before any calculus, you need fluency with the standard library of functions: what each one looks like, where it is defined, how it behaves at the edges of its domain, and how fast it grows. Everything later in the course — limits, asymptotes, optimisation, curve sketching — is a question about the graphs collected here.

Functions, domain and range

A function ff assigns to each input xx in its domain exactly one output f(x)f(x). The range is the set of outputs actually attained. When a formula is given with no domain stated, the natural domain is the largest set of real numbers for which the formula makes sense.

Ingredient in the formulaRequirement
1g(x)\dfrac{1}{g(x)}g(x)0g(x) \ne 0
g(x)\sqrt{g(x)} (any even root)g(x)0g(x) \ge 0
lng(x)\ln g(x), logag(x)\log_a g(x)g(x)>0g(x) > 0
tanx\tan x, secx\sec xxπ2+kπx \ne \dfrac{\pi}{2} + k\pi for every integer kk
cotx\cot x, cscx\csc xxkπx \ne k\pi for every integer kk
arcsing(x)\arcsin g(x), arccosg(x)\arccos g(x)1g(x)1-1 \le g(x) \le 1

Take the intersection of all the requirements: a formula with three ingredients must satisfy all three at once.

The standard library

FunctionNatural domainRangeShape facts
xnx^n, nn even >0>0all reals[0,)[0,\infty)even, U-shaped, through (0,0)(0,0) and (±1,1)(\pm 1,1)
xnx^n, nn odd >1>1all realsall realsodd, increasing, flat at the origin
x1x^{-1}x0x \ne 0y0y \ne 0odd; asymptotes x=0x=0, y=0y=0
x\sqrt{x}[0,)[0,\infty)[0,)[0,\infty)increasing, concave down, vertical tangent at 00
x3\sqrt[3]{x}all realsall realsodd, increasing
axa^x, a>1a>1 (e.g. exe^x)all reals(0,)(0,\infty)increasing; 0\to 0 as xx \to -\infty; a0=1a^0=1
axa^x, 0<a<10 < a < 1all reals(0,)(0,\infty)decreasing
logax\log_a x, a>1a>1 (e.g. lnx\ln x)(0,)(0,\infty)all realsincreasing; \to -\infty as x0+x\to 0^{+}
sinx\sin x, cosx\cos xall reals[1,1][-1,1]period 2π2\pi; sin\sin odd, cos\cos even
tanx\tan xxπ2+kπx \ne \dfrac{\pi}{2}+k\pi, kk integerall realsperiod π\pi, odd, vertical asymptotes
polynomial of degree nnall realsdependsend behaviour follows the leading term anxna_n x^n; at most nn real roots

Rational functions and asymptotes

Let f=P/Qf = P/Q with P,QP,Q polynomials. Factor both and cancel before drawing conclusions.

  • A factor (xc)(x-c) that cancels completely gives a hole at x=cx=c (the point is missing, the graph does not blow up). Its height is the value of the reduced formula at cc.
  • A zero cc of the reduced denominator gives a vertical asymptote x=cx=c. Determine the sign on each side from the signs of the reduced numerator and denominator.
  • End behaviour, comparing degrees: degP<degQ\deg P < \deg Q \Rightarrow horizontal asymptote y=0y=0; degP=degQy=\deg P = \deg Q \Rightarrow y = ratio of leading coefficients; degP=degQ+1\deg P = \deg Q + 1 \Rightarrow a slant asymptote, obtained by polynomial long division; degPdegQ+2\deg P \ge \deg Q + 2 \Rightarrow no linear asymptote (the graph follows the polynomial quotient).

A graph may cross a horizontal or slant asymptote — the asymptote only describes behaviour as x±x \to \pm\infty. It can never cross a vertical asymptote.

Transformations

Let c>0c>0, h,kh,k real.

EquationEffect on the graph of y=f(x)y=f(x)
y=f(x)+ky = f(x) + kshift up kk (down if k<0k<0)
y=f(xh)y = f(x-h)shift right hh (left if h<0h<0)
y=cf(x)y = c\,f(x), c>1c>1vertical stretch by factor cc
y=cf(x)y = c\,f(x), 0<c<10 < c < 1vertical compression
y=f(cx)y = f(cx), c>1c>1horizontal compression by factor 1c\dfrac{1}{c}
y=f(cx)y = f(cx), 0<c<10 < c < 1horizontal stretch by factor 1c\dfrac{1}{c}
y=f(x)y = -f(x)reflect in the xx-axis
y=f(x)y = f(-x)reflect in the yy-axis
y=f1(x)y = f^{-1}(x)reflect in the line y=xy=x (only if ff is one-to-one)

Two more: y=f(x)y = \lvert f(x) \rvert flips the part of the graph below the xx-axis up above it; y=f(x)y = f(\lvert x \rvert) deletes the part with x<0x<0 and replaces it with the mirror image of the x0x \ge 0 part.

Horizontal transformations act on the input, so always factor the argument first: y=f(2x6)=f(2(x3))y = f(2x-6) = f\big(2(x-3)\big) is a horizontal compression by 12\frac{1}{2} followed by a shift right 3 — not right 6. Vertical operations act on the output, in the order they are written: y=cf(x)+ky = c\,f(x)+k stretches by cc first and then shifts by kk, which is not the same graph as y=c(f(x)+k)=cf(x)+cky = c\big(f(x)+k\big) = c\,f(x)+ck. As a book-keeping check, horizontal operations affect only the domain and leave the range alone; vertical operations affect only the range and leave the domain alone.

Even and odd symmetry

These require a domain symmetric about 00 (if xx is in the domain, so is x-x).

  • Even: f(x)=f(x)f(-x) = f(x) for all xx in the domain. Graph is symmetric in the yy-axis. Examples: x2x^{2}, cosx\cos x, x\lvert x \rvert.
  • Odd: f(x)=f(x)f(-x) = -f(x) for all xx in the domain. Graph has 180180^{\circ} rotational symmetry about the origin; if 00 is in the domain then f(0)=0f(0)=0. Examples: x3x^{3}, sinx\sin x, tanx\tan x, 1/x1/x.

Combination rules: even ×\times even == even, odd ×\times odd == even, even ×\times odd == odd (same for quotients). Even ++ even == even, odd ++ odd == odd; a mixed sum is usually neither. The only function that is both even and odd is f(x)=0f(x)=0. Most functions are neither — "neither" is a legitimate and common answer.

Composition and inverses

(fg)(x)=f(g(x))(f \circ g)(x) = f(g(x)). Its domain is the set of xx in the domain of gg such that g(x)g(x) lies in the domain of ff — read off from the pieces, never from the simplified formula. In general fggff \circ g \ne g \circ f.

ff is one-to-one if f(x1)=f(x2)f(x_1)=f(x_2) forces x1=x2x_1=x_2; graphically, no horizontal line meets the graph more than once (horizontal line test). Every strictly increasing or strictly decreasing function is one-to-one, but the converse fails — a one-to-one function need not be monotone. An inverse f1f^{-1} exists exactly when ff is one-to-one, and it then satisfies f1(f(x))=xfor x in dom(f),f(f1(y))=yfor y in ran(f),f^{-1}(f(x)) = x \quad \text{for } x \text{ in dom}(f), \qquad f(f^{-1}(y)) = y \quad \text{for } y \text{ in ran}(f), with dom(f1)=ran(f)\text{dom}(f^{-1}) = \text{ran}(f) and ran(f1)=dom(f)\text{ran}(f^{-1}) = \text{dom}(f). A function that fails the test can be made invertible by restricting its domain: sin\sin restricted to [π2,π2]\left[-\frac{\pi}{2},\frac{\pi}{2}\right] gives arcsin\arcsin, with domain [1,1][-1,1].

Exponential and logarithm laws

Take a>0a>0, a1a \ne 1. Every exponential is a rescaled exe^x: ax=exlnaa^{x} = e^{x \ln a}, and logax=lnxlna\log_a x = \dfrac{\ln x}{\ln a}.

LawValid when
ax+y=axaya^{x+y} = a^{x}a^{y}all real x,yx,y
(ax)y=axy(a^{x})^{y} = a^{xy}all real x,yx,y
ln(uv)=lnu+lnv\ln(uv) = \ln u + \ln vu>0u>0, v>0v>0
lnuv=lnulnv\ln\dfrac{u}{v} = \ln u - \ln vu>0u>0, v>0v>0
ln(ur)=rlnu\ln(u^{r}) = r \ln uu>0u>0, any real rr
elnu=ue^{\ln u} = uu>0u>0
ln(ex)=x\ln(e^{x}) = xall real xx

Edge case worth memorising: ln(x2)=2lnx\ln(x^{2}) = 2\ln\lvert x \rvert for x0x \ne 0. Writing 2lnx2\ln x silently throws away the whole half x<0x<0 of the domain.

Sinusoids: amplitude, period, phase shift

Put every sinusoid in the standard form y=Asin(B(xC))+D(A0, B0),y = A\sin\big(B(x-C)\big) + D \qquad (A \ne 0,\ B \ne 0), factoring BB out of the argument first: sin ⁣(3x+π2)=sin ⁣(3(x+π6))\sin\!\left(3x+\frac{\pi}{2}\right) = \sin\!\left(3\left(x+\frac{\pi}{6}\right)\right), a shift of π6\frac{\pi}{6} left, not π2\frac{\pi}{2} left.

QuantityValue
amplitudeA\lvert A \rvert
period2πB\dfrac{2\pi}{\lvert B \rvert} (but πB\dfrac{\pi}{\lvert B \rvert} for tan\tan and cot\cot)
phase shiftCC (right if C>0C>0, left if C<0C<0)
midliney=Dy = D
range[DA, D+A][\,D-\lvert A \rvert,\ D+\lvert A \rvert\,]

The same table applies with cos\cos in place of sin\sin. Shift identities that kill a phase: sin ⁣(θ+π2)=cosθ\sin\!\left(\theta+\frac{\pi}{2}\right) = \cos\theta, cos ⁣(θπ2)=sinθ\cos\!\left(\theta-\frac{\pi}{2}\right) = \sin\theta, sin(θ+π)=sinθ\sin(\theta+\pi) = -\sin\theta.

Comparing long-run growth rates

For functions eventually positive, say ff grows faster than gg if limxf(x)g(x)=\lim_{x \to \infty} \dfrac{f(x)}{g(x)} = \infty (equivalently g/f0g/f \to 0). The three facts you need, valid for every p>0p>0, q>0q>0 and every a>1a>1: limx(lnx)qxp=0,limxxpax=0,limx0+xplnx=0.\lim_{x \to \infty}\frac{(\ln x)^{q}}{x^{p}} = 0, \qquad \lim_{x \to \infty}\frac{x^{p}}{a^{x}} = 0, \qquad \lim_{x \to 0^{+}} x^{p}\,\lvert \ln x \rvert = 0.

Slowest to fastest as xx \to \inftyComment
constantsno growth
(lnx)q(\ln x)^{q}, any q>0q>0beaten by every positive power of xx
xpx^{p}, small p>0p>0larger pp always wins
xpx^{p}, large ppbeaten by every exponential
axa^{x}, a>1a>1larger aa wins: ax/bxa^{x}/b^{x} \to \infty when a>b>0a>b>0
n!n!, then nnn^{n} (integer nn)faster than every exponential

To compare two exponentials written differently, put both in the form ecxe^{cx}: 2x=exln22^{x} = e^{x\ln 2} with ln20.693\ln 2 \approx 0.693, so 2x2^{x} beats ex/2e^{x/2}. The practical technique for a limit at infinity is to divide numerator and denominator by the single fastest-growing term that appears. (These facts can be proved later with l'Hopital's Rule, which requires the quotient to be of the indeterminate form 00\frac{0}{0} or \frac{\infty}{\infty}, both functions to be differentiable near the point with the denominator's derivative non-zero there, and the limit of the ratio of derivatives to exist or be ±\pm\infty.)

Curve sketching checklist (no calculus needed)

  1. Domain. 2. Intercepts. 3. Symmetry (even, odd, periodic). 4. Vertical asymptotes and holes, with the sign of ff on each side. 5. End behaviour: horizontal or slant asymptote, or the dominant term. 6. Sign chart from the fully factored form. 7. Plot two or three extra points and join using the shapes from the standard library.

Worked example 1 — a complete rational sketch

Sketch f(x)=x24x2x6f(x) = \dfrac{x^{2}-4}{x^{2}-x-6}.

Factor. x24=(x2)(x+2)x^{2}-4 = (x-2)(x+2) and x2x6=(x3)(x+2)x^{2}-x-6 = (x-3)(x+2), so f(x)=(x2)(x+2)(x3)(x+2).f(x) = \frac{(x-2)(x+2)}{(x-3)(x+2)}. Domain. The denominator vanishes at x=3x=3 and x=2x=-2, so the domain is (,2)(2,3)(3,)(-\infty,-2) \cup (-2,3) \cup (3,\infty).

Reduce. For x2x \ne -2, f(x)=x2x3f(x) = \dfrac{x-2}{x-3}. The cancelled factor gives a hole at x=2x=-2 of height 2223=45=45\dfrac{-2-2}{-2-3} = \dfrac{-4}{-5} = \dfrac{4}{5}.

Vertical asymptote. Only x=3x=3. As x3x \to 3^{-} the numerator 1>0\to 1 > 0 and the denominator 0\to 0^{-}, so ff \to -\infty; as x3+x \to 3^{+} the denominator 0+\to 0^{+}, so f+f \to +\infty.

Horizontal asymptote. Degrees are equal with leading coefficients 11 and 11, so y=1y=1. Does the graph cross it? Solving x2x3=1\dfrac{x-2}{x-3} = 1 gives x2=x3x-2 = x-3, i.e. 2=3-2=-3: impossible, so it never crosses.

Intercepts. f(x)=0f(x)=0 only when x2=0x-2=0, giving (2,0)(2,0); and f(0)=23=23f(0) = \dfrac{-2}{-3} = \dfrac{2}{3}, giving (0,23)\left(0,\frac{2}{3}\right).

Sign. x2x3>0\dfrac{x-2}{x-3} > 0 on (,2)(-\infty,2) and on (3,)(3,\infty), and is negative on (2,3)(2,3).

Position relative to y=1y=1. The sign chart does not settle this — subtract instead: f(x)1=x2x31=(x2)(x3)x3=1x3.f(x) - 1 = \frac{x-2}{x-3} - 1 = \frac{(x-2)-(x-3)}{x-3} = \frac{1}{x-3}. That is positive exactly when x>3x>3, so the curve sits above y=1y=1 on (3,)(3,\infty) and below y=1y=1 on (,3)(-\infty,3), with a hole punched out at (2,45)\left(-2,\frac{4}{5}\right) (and indeed 45<1\frac45 < 1).

Worked example 2 — a growth rate by substitution

Show that limx(lnx)3x=0\lim_{x \to \infty} \dfrac{(\ln x)^{3}}{\sqrt{x}} = 0.

Substitute x=etx = e^{t}, so t=lnxt = \ln x and tt \to \infty as xx \to \infty (legitimate because tett \mapsto e^{t} is continuous, strictly increasing and maps onto (0,)(0,\infty)). Then x=et/2\sqrt{x} = e^{t/2} and (lnx)3x=t3et/2=t3(e1/2)t.\frac{(\ln x)^{3}}{\sqrt{x}} = \frac{t^{3}}{e^{t/2}} = \frac{t^{3}}{\left(e^{1/2}\right)^{t}}. Now apply limttpat=0\lim_{t\to\infty} \dfrac{t^{p}}{a^{t}} = 0 with p=3p=3 and a=e1/21.6487>1a = e^{1/2} \approx 1.6487 > 1. Hence the limit is 0\mathbf{0}.

This convergence is slow, which is exactly why intuition needs the theorem: at x=106x = 10^{6} the ratio is (13.8155)310002.64\dfrac{(13.8155)^{3}}{1000} \approx 2.64, still bigger than 11; at x=1012x = 10^{12} it is (27.631)31060.021\dfrac{(27.631)^{3}}{10^{6}} \approx 0.021.

Common mistakes

  • Wrong: x2=x\sqrt{x^{2}} = x. Right: x2=x\sqrt{x^{2}} = \lvert x \rvert; the square root symbol always returns the non-negative root.
  • Wrong: the domain of x29\sqrt{x^{2}-9} is x3x \ge 3. Right: x29x^{2} \ge 9 means x3\lvert x \rvert \ge 3, so the domain is (,3][3,)(-\infty,-3] \cup [3,\infty).
  • Wrong: ln(a+b)=lna+lnb\ln(a+b) = \ln a + \ln b. Right: only products split: ln(ab)=lna+lnb\ln(ab) = \ln a + \ln b for a,b>0a,b>0. ln(a+b)\ln(a+b) does not simplify at all.
  • Wrong: y=f(2x6)y = f(2x-6) is "compress horizontally by 12\frac{1}{2}, then shift right 66". Right: that sequence produces f(2(x6))=f(2x12)f(2(x-6)) = f(2x-12). Factor first: f(2(x3))f(2(x-3)) — compress, then shift right 33 (or shift right 66 first and then compress).
  • Wrong: f(x)=x2+xf(x) = x^{2}+x is even because x2x^{2} is. Right: f(1)=0f(-1) = 0 but f(1)=2f(1) = 2, so it is neither even nor odd. One term does not decide.
  • Wrong: lnx\ln x is odd because its graph "curls". Right: its domain (0,)(0,\infty) is not symmetric about 00, so f(x)f(-x) is undefined and the function is neither.
  • Wrong: cancelling (x+2)(x+2) removes the problem at x=2x=-2, so the graph is unbroken there. Right: the original function is undefined at x=2x=-2; the graph has a hole at (2,45)\left(-2,\frac{4}{5}\right).
  • Wrong: a graph can never touch its horizontal asymptote. Right: f(x)=sinxxf(x) = \dfrac{\sin x}{x} crosses y=0y=0 at every x=kπx = k\pi with k0k \ne 0, yet y=0y=0 is still its horizontal asymptote.
  • Wrong: with f(x)=x2f(x)=x^{2} and g(x)=xg(x)=\sqrt{x}, f(g(x))=xf(g(x)) = x for every real xx. Right: the composite inherits gg's domain, so f(g(x))=xf(g(x)) = x only on [0,)[0,\infty).
  • Wrong: x100x^{100} grows faster than 2x2^{x} — just check x=10x=10. Right: the two graphs cross where 100lnx=xln2100\ln x = x\ln 2, at x1.007x \approx 1.007 and again at x996x \approx 996; x100x^{100} is larger only strictly between those, and 2x2^{x} wins forever after x996x \approx 996. Growth statements are claims about the long run only.

Key terms

  • domain
  • range
  • natural domain
  • even function
  • odd function
  • composition of functions
  • one-to-one (injective)
  • horizontal line test
  • inverse function
  • transformation (shift, stretch, reflection)
  • polynomial
  • rational function
  • vertical asymptote
  • horizontal asymptote
  • slant (oblique) asymptote
  • hole (removable point)
  • end behaviour
  • exponential function
  • logarithm
  • period and amplitude
  • growth rate hierarchy
  • dominant term

Practice Problems

Find the natural domain of f(x)=x+3x24,f(x) = \frac{\sqrt{x+3}}{x^{2}-4}, and write your answer in interval notation.

Show hint

There are two separate requirements here: the square root needs a non-negative input, and the denominator is not allowed to be zero. Impose both, then intersect.

Show answer

Step 1 — the square root. We need x+30x+3 \ge 0, i.e. x3x \ge -3. This alone gives [3,)[-3,\infty).

Step 2 — the denominator. We need x240x^{2}-4 \ne 0. Factor: x24=(x2)(x+2)x^{2}-4 = (x-2)(x+2), so x2x \ne 2 and x2x \ne -2.

Step 3 — intersect. Start from [3,)[-3,\infty) and delete the two forbidden points x=2x=-2 and x=2x=2 (both lie inside that interval): [3,2)(2,2)(2,)\boxed{[-3,-2) \cup (-2,2) \cup (2,\infty)}

Check. At x=3x=-3: f(3)=094=0f(-3) = \dfrac{\sqrt{0}}{9-4} = 0, defined — so the left endpoint is correctly included. At x=4x=-4: 1\sqrt{-1} is not real, correctly excluded. At x=2x=2: denominator =0=0, correctly excluded.

Classify each function as even, odd or neither, justifying each answer:

  • (a) f(x)=x43x2+7f(x) = x^{4} - 3x^{2} + 7
  • (b) g(x)=x3sinxg(x) = x^{3}\sin x
  • (c) h(x)=xx2+1h(x) = \dfrac{x}{x^{2}+1}
  • (d) k(x)=lnxk(x) = \ln x
Show hint

Compute f(x)f(-x) and compare it with f(x)f(x) and with f(x)-f(x). Before any of that, check whether the domain is symmetric about 00.

Show answer

(a) Domain is all reals (symmetric). Replace xx by x-x: f(x)=(x)43(x)2+7=x43x2+7=f(x).f(-x) = (-x)^{4} - 3(-x)^{2} + 7 = x^{4} - 3x^{2} + 7 = f(x). Every power of xx present is even, so ff is even.

(b) Domain is all reals. Both factors are odd: (x)3=x3(-x)^{3} = -x^{3} and sin(x)=sinx\sin(-x) = -\sin x. So g(x)=(x)3sin(x)=(x3)(sinx)=x3sinx=g(x).g(-x) = (-x)^{3}\sin(-x) = (-x^{3})(-\sin x) = x^{3}\sin x = g(x). gg is even — odd ×\times odd == even. (Sanity check: g(1)=sin10.841g(1) = \sin 1 \approx 0.841 and g(1)=(1)(sin1)=sin1g(-1) = (-1)(-\sin 1) = \sin 1, equal.)

(c) Domain is all reals since x2+1>0x^{2}+1 > 0 always. Then h(x)=x(x)2+1=xx2+1=h(x),h(-x) = \frac{-x}{(-x)^{2}+1} = \frac{-x}{x^{2}+1} = -h(x), so hh is odd — odd numerator over even denominator. (Check: h(1)=12h(1) = \frac12, h(1)=12h(-1) = -\frac12.)

(d) The domain of lnx\ln x is (0,)(0,\infty), which is not symmetric about 00: for x=2x = 2 in the domain, 2-2 is not. So k(x)k(-x) is undefined and neither condition can hold. kk is neither.

Summary: (a) even, (b) even, (c) odd, (d) neither.

Starting from the graph of y=xy = \sqrt{x}, describe in order the transformations that produce y=2x+3+1,y = -2\sqrt{x+3} + 1, and state the domain and range of the result.

Show hint

Write the formula as 2f(x+3)+1-2f(x+3)+1 with f(x)=xf(x)=\sqrt{x}, and track what each operation does to the domain and to the range separately.

Show answer

Write f(x)=xf(x) = \sqrt{x}, whose domain is [0,)[0,\infty) and range is [0,)[0,\infty). The target is y=2f(x+3)+1y = -2f(x+3)+1.

Step 1 — inside the function: xx+3x \mapsto x+3. y=x+3y = \sqrt{x+3} is a shift 3 units left (since x+3=x(3)x+3 = x-(-3)). Domain becomes x+30x+3 \ge 0, i.e. [3,)[-3,\infty); range still [0,)[0,\infty).

Step 2 — multiply the output by 22. y=2x+3y = 2\sqrt{x+3} is a vertical stretch by factor 2. Range: 2[0,)=[0,)2 \cdot [0,\infty) = [0,\infty).

Step 3 — multiply the output by 1-1. y=2x+3y = -2\sqrt{x+3} reflects in the xx-axis. Range becomes (,0](-\infty,0].

Step 4 — add 11. y=2x+3+1y = -2\sqrt{x+3}+1 is a shift 1 unit up. Range becomes (,1](-\infty,1].

Domain: the vertical operations (steps 2–4) never change the domain, so it stays [3,)\boxed{[-3,\infty)}. Range: (,1]\boxed{(-\infty,1]}.

Check with points. At x=3x=-3: y=20+1=1y = -2\sqrt{0}+1 = 1, the highest point, matching the top of the range. At x=1x=1: 1+3=2\sqrt{1+3}=2, so y=2(2)+1=3y = -2(2)+1 = -3; the curve falls to the right, consistent with a reflected, stretched square root.

Let f(x)=1x1f(x) = \dfrac{1}{x-1} and g(x)=xg(x) = \sqrt{x}. Find formulas for (fg)(x)(f \circ g)(x) and (gf)(x)(g \circ f)(x), and state the domain of each composite.

Show hint

Substitute one formula into the other, then build each domain from two conditions: the inner function must be defined, and its output must be an allowed input for the outer function.

Show answer

The composite fgf \circ g. (fg)(x)=f(g(x))=f(x)=1x1.(f \circ g)(x) = f\big(g(x)\big) = f\big(\sqrt{x}\,\big) = \frac{1}{\sqrt{x}-1}. Domain conditions:

  1. xx must be in dom(g)\text{dom}(g): x0x \ge 0.
  2. g(x)g(x) must be in dom(f)\text{dom}(f), i.e. x1\sqrt{x} \ne 1, which gives x1x \ne 1.

So dom(fg)=[0,1)(1,)\text{dom}(f \circ g) = \boxed{[0,1) \cup (1,\infty)}.

The composite gfg \circ f. (gf)(x)=g(f(x))=1x1.(g \circ f)(x) = g\big(f(x)\big) = \sqrt{\frac{1}{x-1}}. Domain conditions:

  1. xx must be in dom(f)\text{dom}(f): x1x \ne 1.
  2. f(x)f(x) must be in dom(g)\text{dom}(g): 1x10\dfrac{1}{x-1} \ge 0. Since the numerator 11 is positive, this fraction is non-negative exactly when x1>0x-1 > 0 (it is never equal to 00).

So x>1x > 1, and dom(gf)=(1,)\text{dom}(g \circ f) = \boxed{(1,\infty)}.

Check. (fg)(4)=141=11=1(f\circ g)(4) = \dfrac{1}{\sqrt{4}-1} = \dfrac{1}{1} = 1, and (gf)(2)=11=1(g \circ f)(2) = \sqrt{\dfrac{1}{1}} = 1. At x=0x=0: (fg)(0)=101=1(f\circ g)(0) = \dfrac{1}{0-1} = -1 is fine, but (gf)(0)=1(g\circ f)(0) = \sqrt{-1} is not — the two composites genuinely have different domains, and fggff \circ g \ne g \circ f.

Solve for xx: log2(x)+log2(x2)=3.\log_{2}(x) + \log_{2}(x-2) = 3. State clearly which candidate solutions are valid and why.

Show hint

Write down the domain restrictions before you combine the logarithms — that is what decides the fate of the second root.

Show answer

Step 1 — domain. Both logarithms must have positive arguments: x>0x > 0 and x2>0x-2>0. Together: x>2x > 2. Any candidate root not exceeding 22 must be rejected.

Step 2 — combine. For x>2x>2 both arguments are positive, so the product law applies: log2(x(x2))=3.\log_{2}\big(x(x-2)\big) = 3.

Step 3 — remove the logarithm. By definition of log2\log_2, x(x2)=23=8.x(x-2) = 2^{3} = 8.

Step 4 — solve the quadratic. x22x8=0(x4)(x+2)=0x=4  or  x=2.x^{2} - 2x - 8 = 0 \quad \Rightarrow \quad (x-4)(x+2) = 0 \quad \Rightarrow \quad x = 4 \ \text{ or } \ x = -2.

Step 5 — test against the domain. x=2x = -2 fails x>2x>2 (indeed log2(2)\log_2(-2) is undefined), so it is extraneous and is rejected. x=4x=4 satisfies 4>24>2.

Verification. log24+log22=2+1=3\log_{2}4 + \log_{2}2 = 2 + 1 = 3. Correct.

Answer: x=4\boxed{x = 4} is the only solution.

For y=2sin ⁣(3x+π2)1y = 2\sin\!\left(3x + \dfrac{\pi}{2}\right) - 1, state the amplitude, period, phase shift, midline and range. Then find every xx in [0,2π)[0,2\pi) at which the maximum value is attained.

Show hint

Factor the 33 out of the argument before reading off the phase shift. Then look for an identity that removes the phase shift entirely.

Show answer

Step 1 — factor the argument. 3x+π2=3(x+π6),soy=2sin ⁣(3(x+π6))1.3x + \frac{\pi}{2} = 3\left(x + \frac{\pi}{6}\right), \qquad \text{so} \qquad y = 2\sin\!\left(3\left(x+\frac{\pi}{6}\right)\right) - 1.

Step 2 — read off the data.

  • Amplitude: 2=2\lvert 2 \rvert = 2.
  • Period: 2π3\dfrac{2\pi}{3} (the input is multiplied by 33, compressing the period by 13\frac13).
  • Phase shift: π6\dfrac{\pi}{6} to the left.
  • Midline: y=1y = -1.
  • Range: the sine factor lies in [1,1][-1,1], so 2sin()[2,2]2\sin(\cdot) \in [-2,2] and y[21, 21]=[3,1]y \in [-2-1,\ 2-1] = \boxed{[-3,1]}.

Step 3 — simplify. Using sin ⁣(θ+π2)=cosθ\sin\!\left(\theta + \frac{\pi}{2}\right) = \cos\theta with θ=3x\theta = 3x, y=2cos(3x)1.y = 2\cos(3x) - 1. (Check at x=0x=0: the original gives 2sinπ21=21=12\sin\frac{\pi}{2}-1 = 2-1 = 1; the simplified form gives 2cos01=12\cos 0 - 1 = 1. Agreed.)

Step 4 — locate the maxima. yy is largest when cos(3x)=1\cos(3x) = 1, i.e. when 3x=2kπx=2kπ3,k=0,±1,±2,3x = 2k\pi \quad \Rightarrow \quad x = \frac{2k\pi}{3}, \qquad k = 0, \pm 1, \pm 2, \dots Restricting to 0x<2π0 \le x < 2\pi gives k=0,1,2k = 0,1,2: x=0, 2π3, 4π3with maximum value y=2(1)1=1.\boxed{x = 0,\ \frac{2\pi}{3},\ \frac{4\pi}{3}} \qquad \text{with maximum value } y = 2(1)-1 = 1. (k=3k=3 gives x=2πx = 2\pi, which is excluded.) The three maxima are exactly one period apart, as they must be.

For f(x)=2x23x2x24f(x) = \frac{2x^{2}-3x-2}{x^{2}-4} find the domain, any holes, all asymptotes and both intercepts, and describe the behaviour of the graph on each side of the vertical asymptote.

Show hint

Factor the numerator and the denominator first. A common factor changes the story completely.

Show answer

Step 1 — factor. For the numerator, look for (2x+1)(x2)=2x24x+x2=2x23x2(2x+1)(x-2) = 2x^{2}-4x+x-2 = 2x^{2}-3x-2. Correct. The denominator is (x2)(x+2)(x-2)(x+2). So f(x)=(2x+1)(x2)(x2)(x+2).f(x) = \frac{(2x+1)(x-2)}{(x-2)(x+2)}.

Step 2 — domain. The original denominator vanishes at x=2x = 2 and x=2x = -2: domain=(,2)(2,2)(2,).\text{domain} = (-\infty,-2) \cup (-2,2) \cup (2,\infty).

Step 3 — reduce, and find the hole. For x2x \ne 2, f(x)=2x+1x+2f(x) = \dfrac{2x+1}{x+2}. Since (x2)(x-2) cancelled completely, x=2x=2 is a hole, not an asymptote. Its height is the reduced formula at x=2x=2: 2(2)+12+2=54.Hole at (2,54).\frac{2(2)+1}{2+2} = \frac{5}{4}. \qquad \text{Hole at } \left(2, \frac{5}{4}\right).

Step 4 — vertical asymptote. The reduced denominator vanishes only at x=2x = -2, so x=2x=-2 is the single vertical asymptote.

Step 5 — horizontal asymptote. Numerator and denominator of the original have equal degree 22, with leading coefficients 22 and 11, so y=2y = 2. Does the graph meet it? 2x+1x+2=2\dfrac{2x+1}{x+2} = 2 gives 2x+1=2x+42x+1 = 2x+4, i.e. 1=41=4: impossible, so it never crosses.

Step 6 — intercepts. f(x)=0f(x) = 0 when 2x+1=02x+1=0, i.e. x=12x = -\frac12 (which is in the domain): the point (12,0)\left(-\frac12,0\right). And f(0)=24=12f(0) = \dfrac{-2}{-4} = \dfrac12, so the yy-intercept is (0,12)\left(0,\frac12\right).

Step 7 — sides of x=2x=-2. Use the reduced form 2x+1x+2\dfrac{2x+1}{x+2}; near x=2x=-2 the numerator is close to 2(2)+1=3<02(-2)+1 = -3 < 0.

  • As x2x \to -2^{-}: x+20x+2 \to 0^{-}, so negativenegative+\dfrac{\text{negative}}{\text{negative}} \to +\infty.
  • As x2+x \to -2^{+}: x+20+x+2 \to 0^{+}, so negativepositive\dfrac{\text{negative}}{\text{positive}} \to -\infty.

Numerical check. f(2.1)=3.20.1=32f(-2.1) = \dfrac{-3.2}{-0.1} = 32 (large positive) and f(1.9)=2.80.1=28f(-1.9) = \dfrac{-2.8}{0.1} = -28 (large negative). Consistent.

Let f(x)=x2+1x1f(x) = \dfrac{x^{2}+1}{x-1}. Find all asymptotes, and determine whether the graph approaches the slant asymptote from above or from below as x+x \to +\infty and as xx \to -\infty.

Show hint

The degree of the numerator is exactly one more than the degree of the denominator. Divide.

Show answer

Step 1 — vertical asymptote. The denominator vanishes at x=1x=1, and the numerator there is 12+1=201^{2}+1 = 2 \ne 0, so there is a vertical asymptote at x=1x=1 (no cancellation, hence no hole).

Step 2 — long division. Since deg(numerator)=deg(denominator)+1\deg(\text{numerator}) = \deg(\text{denominator}) + 1, expect a slant asymptote. Write x2+1=(x1)(x+1)+2x^{2}+1 = (x-1)(x+1) + 2, because (x1)(x+1)=x21(x-1)(x+1) = x^{2}-1 and x21+2=x2+1x^{2}-1+2 = x^{2}+1. Therefore f(x)=x+1+2x1.f(x) = x+1+\frac{2}{x-1}.

Step 3 — the slant asymptote. As x±x \to \pm\infty, 2x10\dfrac{2}{x-1} \to 0, so the graph approaches the line y=x+1.y = x+1. There is no horizontal asymptote (the function is unbounded at both ends).

Step 4 — which side? The signed gap is exactly f(x)(x+1)=2x1.f(x) - (x+1) = \frac{2}{x-1}.

  • For x>1x > 1 this is positive, so as x+x \to +\infty the graph lies above y=x+1y=x+1.
  • For x<1x < 1 this is negative, so as xx \to -\infty the graph lies below y=x+1y=x+1.

The gap is never zero, so the graph never crosses its slant asymptote.

Step 5 — behaviour at the vertical asymptote and intercepts. As x1+x \to 1^{+}, 2x1+\dfrac{2}{x-1} \to +\infty so f+f \to +\infty; as x1x \to 1^{-}, ff \to -\infty. There are no xx-intercepts because x2+1>0x^{2}+1 > 0 for all xx; the yy-intercept is f(0)=11=1f(0) = \dfrac{1}{-1} = -1.

Numerical check. f(10)=101911.222f(10) = \dfrac{101}{9} \approx 11.222 while x+1=11x+1 = 11: the gap is 0.222=290.222 = \dfrac{2}{9}, positive and small, exactly as predicted. f(10)=101119.182f(-10) = \dfrac{101}{-11} \approx -9.182 while x+1=9x+1 = -9: the gap is 0.182=211-0.182 = \dfrac{2}{-11}, negative.

Show that f(x)=2x+1x3f(x) = \dfrac{2x+1}{x-3} is one-to-one on its domain, find a formula for f1f^{-1}, and state the domain and range of both ff and f1f^{-1}.

Show hint

Rewriting the fraction as "constant plus a multiple of 1x3\dfrac{1}{x-3}" makes both the injectivity and the range immediate.

Show answer

Step 1 — domain of ff. The denominator vanishes at x=3x=3, so dom(f)={x:x3}\text{dom}(f) = \{x : x \ne 3\}.

Step 2 — rewrite. Force the numerator to contain x3x-3: f(x)=2x+1x3=2(x3)+7x3=2+7x3.f(x) = \frac{2x+1}{x-3} = \frac{2(x-3) + 7}{x-3} = 2 + \frac{7}{x-3}. (Check: 2(x3)+7=2x6+7=2x+12(x-3)+7 = 2x-6+7 = 2x+1. Correct.)

Step 3 — one-to-one. Suppose f(a)=f(b)f(a) = f(b) with a,b3a,b \ne 3. Then 2+7a3=2+7b37a3=7b3a3=b3a=b.2 + \frac{7}{a-3} = 2 + \frac{7}{b-3} \Rightarrow \frac{7}{a-3} = \frac{7}{b-3} \Rightarrow a-3 = b-3 \Rightarrow a = b. So ff is one-to-one and an inverse exists.

Step 4 — range of ff. Since 7x3\dfrac{7}{x-3} takes every non-zero real value (and never 00), f(x)=2+7x3f(x) = 2 + \dfrac{7}{x-3} takes every real value except 22. So ran(f)={y:y2}\text{ran}(f) = \{y : y \ne 2\} — which is exactly the statement that y=2y=2 is the horizontal asymptote and is never attained.

Step 5 — invert. Set y=2x+1x3y = \dfrac{2x+1}{x-3} and solve for xx: y(x3)=2x+1yx3y=2x+1yx2x=3y+1x(y2)=3y+1x=3y+1y2(y2).\begin{aligned} y(x-3) &= 2x+1 \\ yx - 3y &= 2x + 1 \\ yx - 2x &= 3y + 1 \\ x(y-2) &= 3y+1 \\ x &= \frac{3y+1}{y-2} \qquad (y \ne 2). \end{aligned} Swapping names, f1(x)=3x+1x2.\boxed{f^{-1}(x) = \frac{3x+1}{x-2}}.

Step 6 — domains and ranges. dom(f)={x3},ran(f)={y2},dom(f1)={x2},ran(f1)={y3},\text{dom}(f) = \{x \ne 3\}, \quad \text{ran}(f) = \{y \ne 2\}, \quad \text{dom}(f^{-1}) = \{x \ne 2\}, \quad \text{ran}(f^{-1}) = \{y \ne 3\}, confirming dom(f1)=ran(f)\text{dom}(f^{-1}) = \text{ran}(f) and ran(f1)=dom(f)\text{ran}(f^{-1}) = \text{dom}(f).

Check. f(4)=91=9f(4) = \dfrac{9}{1} = 9, and f1(9)=27+17=287=4f^{-1}(9) = \dfrac{27+1}{7} = \dfrac{28}{7} = 4. Also f(0)=13=13f(0) = \dfrac{1}{-3} = -\dfrac13 and f1 ⁣(13)=1+1132=0f^{-1}\!\left(-\dfrac13\right) = \dfrac{-1+1}{-\frac13-2} = 0. Both round trips return the start.

Order the following six functions from slowest to fastest growing as xx \to \infty, and justify the three closest calls with an explicit limit: A=(lnx)5,B=x1/3,C=xlnx,D=x2,E=ex/2,F=2x.A = (\ln x)^{5}, \quad B = x^{1/3}, \quad C = \sqrt{x}\,\ln x, \quad D = x^{2}, \quad E = e^{x/2}, \quad F = 2^{x}.

Show hint

Compare functions in pairs by looking at the limit of their ratio, and rewrite both exponentials in the common form ecxe^{cx} before comparing them.

Show answer

The order is A, B, C, D, E, F\boxed{A,\ B,\ C,\ D,\ E,\ F} (slowest to fastest). Justification, pair by pair:

AA vs BB. By the log-versus-power fact limx(lnx)qxp=0\lim_{x\to\infty}\dfrac{(\ln x)^{q}}{x^{p}} = 0 with q=5q=5, p=13p=\frac13: limx(lnx)5x1/3=0,\lim_{x\to\infty}\frac{(\ln x)^{5}}{x^{1/3}} = 0, so AA grows slower than BB. Any positive power of xx eventually beats any power of lnx\ln x.

BB vs CC. CB=x1/2lnxx1/3=x1/6lnx,\frac{C}{B} = \frac{x^{1/2}\ln x}{x^{1/3}} = x^{1/6}\ln x \to \infty, since both factors tend to \infty. So CC grows faster than BB.

CC vs DD. CD=x1/2lnxx2=lnxx3/20\frac{C}{D} = \frac{x^{1/2}\ln x}{x^{2}} = \frac{\ln x}{x^{3/2}} \to 0 by the log-versus-power fact with q=1q=1, p=32p=\frac32. So CC is slower than DD.

DD vs EE. Power versus exponential: ex/2=(e1/2)xe^{x/2} = \left(e^{1/2}\right)^{x} with e1/21.6487>1e^{1/2} \approx 1.6487 > 1, so x2ex/20,\frac{x^{2}}{e^{x/2}} \to 0, and EE beats DD.

EE vs FF. Put both in the form ecxe^{cx}: E=e0.5xE = e^{0.5x} and F=2x=exln2F = 2^{x} = e^{x\ln 2} with ln20.6931\ln 2 \approx 0.6931. Then FE=ex(ln20.5)=e0.1931x,\frac{F}{E} = e^{x(\ln 2 - 0.5)} = e^{0.1931x} \to \infty, so FF is the fastest.

A warning about "eventually". At x=100x=100 the actual values are A2071A \approx 2071, B4.64B \approx 4.64, C46.1C \approx 46.1, D=104D = 10^{4}, E5.2×1021E \approx 5.2\times10^{21}, F1.3×1030F \approx 1.3\times10^{30} — so AA is still comfortably ahead of BB and CC. Solving 5ln(lnx)=13lnx5\ln(\ln x) = \frac13\ln x shows BB only overtakes AA near x1027x \approx 10^{27}. The ordering is an asymptotic statement, never a claim about the values you can see on a calculator.

Evaluate the following, justifying each step with the growth hierarchy (no l'Hopital's Rule):

(a) limx3x+x1023x+5x/2\lim_{x\to\infty} \dfrac{3^{x}+x^{10}}{2\cdot 3^{x}+5^{x/2}}

(b) limx0+xlnx\lim_{x\to 0^{+}} x\ln x

Show hint

For (a), first decide which of the three terms grows fastest, then divide top and bottom by it. For (b), the substitution t=1/xt = 1/x turns the problem into a familiar comparison at infinity.

Show answer

(a) First identify the fastest-growing term. Rewrite 5x/2=(51/2)x=(5)x5^{x/2} = \left(5^{1/2}\right)^{x} = (\sqrt{5})^{x} with 52.2361<3\sqrt{5} \approx 2.2361 < 3. So among 3x3^{x}, x10x^{10} and (5)x(\sqrt5)^{x}, the dominant term is 3x3^{x}: it beats the power x10x^{10}, and it beats the smaller-base exponential.

Divide numerator and denominator by 3x3^{x} (legal, since 3x>03^{x} > 0): 3x+x1023x+5x/2=1+x103x2+(53)x.\frac{3^{x}+x^{10}}{2\cdot 3^{x}+5^{x/2}} = \frac{1 + \dfrac{x^{10}}{3^{x}}}{2 + \left(\dfrac{\sqrt{5}}{3}\right)^{x}}.

Now take limits term by term:

  • x103x0\dfrac{x^{10}}{3^{x}} \to 0 by the power-versus-exponential fact (p=10p=10, a=3>1a=3>1).
  • (53)x0\left(\dfrac{\sqrt5}{3}\right)^{x} \to 0 because 0<530.745<10 < \dfrac{\sqrt5}{3} \approx 0.745 < 1, and rx0r^{x}\to0 for 0<r<10 < r < 1.

Hence limx3x+x1023x+5x/2=1+02+0=12.\lim_{x\to\infty}\frac{3^{x}+x^{10}}{2\cdot 3^{x}+5^{x/2}} = \frac{1+0}{2+0} = \boxed{\frac{1}{2}}.

Numerical check at x=60x=60: 3604.239×10283^{60} \approx 4.239\times10^{28}, 60106.05×101760^{10} \approx 6.05\times10^{17}, 5309.31×10205^{30} \approx 9.31\times10^{20}. The quotient is 4.239×10288.478×1028+9.31×10200.50000\dfrac{4.239\times10^{28}}{8.478\times10^{28} + 9.31\times10^{20}} \approx 0.50000. Consistent.

(b) Substitute t=1xt = \dfrac{1}{x}, so that x=1tx = \dfrac1t and t+t \to +\infty as x0+x \to 0^{+}. Then xlnx=1tln ⁣(1t)=1t(lnt)=lntt.x\ln x = \frac{1}{t}\ln\!\left(\frac{1}{t}\right) = \frac{1}{t}\big(-\ln t\big) = -\frac{\ln t}{t}. By the log-versus-power fact with q=1q=1, p=1p=1, lntt0\dfrac{\ln t}{t} \to 0 as tt \to \infty. Therefore limx0+xlnx=0.\lim_{x\to 0^{+}} x\ln x = \boxed{0}.

Numerical check: at x=103x = 10^{-3}, xlnx=0.001×(6.9078)0.0069x\ln x = 0.001 \times (-6.9078) \approx -0.0069; at x=106x=10^{-6}, xlnx1.4×105x\ln x \approx -1.4\times10^{-5}. Approaching 00 from below, as the negative sign predicts.

Let f(x)=exex2f(x) = \dfrac{e^{x}-e^{-x}}{2}.

  • (a) Show that ff is odd.
  • (b) Show that ff is one-to-one, and find an explicit formula for f1f^{-1}.
  • (c) State the domain and range of f1f^{-1}.
Show hint

For (b), multiply the equation y=f(x)y = f(x) through by 2ex2e^{x} and treat the result as a quadratic in the single unknown u=exu = e^{x} — then remember that uu must be positive.

Show answer

(a) The domain is all reals, which is symmetric about 00. Replace xx by x-x: f(x)=exe(x)2=exex2=exex2=f(x).f(-x) = \frac{e^{-x}-e^{-(-x)}}{2} = \frac{e^{-x}-e^{x}}{2} = -\,\frac{e^{x}-e^{-x}}{2} = -f(x). So ff is odd. (Consistently, f(0)=112=0f(0) = \frac{1-1}{2} = 0.)

(b) One-to-one. exe^{x} is strictly increasing, and exe^{-x} is strictly decreasing, so ex-e^{-x} is strictly increasing. A sum of strictly increasing functions is strictly increasing, hence ff is strictly increasing and therefore one-to-one.

Inverting. Set y=exex2y = \dfrac{e^{x}-e^{-x}}{2} and multiply both sides by 2ex2e^{x} (positive, so nothing is lost): 2yex=e2x1.2ye^{x} = e^{2x} - 1. Let u=exu = e^{x}, so u>0u > 0 and u22yu1=0.u^{2} - 2yu - 1 = 0. By the quadratic formula, u=2y±4y2+42=y±y2+1.u = \frac{2y \pm \sqrt{4y^{2}+4}}{2} = y \pm \sqrt{y^{2}+1}. Because y2+1>y2=yy\sqrt{y^{2}+1} > \sqrt{y^{2}} = \lvert y \rvert \ge y, the choice u=yy2+1u = y - \sqrt{y^{2}+1} is negative, which is impossible for u=exu=e^{x}. So ex=y+y2+1x=ln ⁣(y+y2+1).e^{x} = y + \sqrt{y^{2}+1} \quad \Rightarrow \quad x = \ln\!\left(y+\sqrt{y^{2}+1}\right). Renaming the variable, f1(x)=ln ⁣(x+x2+1).\boxed{f^{-1}(x) = \ln\!\left(x+\sqrt{x^{2}+1}\right)}.

(c) As xx\to\infty, exe^{x}\to\infty and ex0e^{-x}\to0, so f(x)f(x)\to\infty; as xx\to-\infty, f(x)f(x)\to-\infty. Being continuous and strictly increasing, ff attains every real value exactly once, so ran(f)\text{ran}(f) is all of the reals. Therefore dom(f1)=ran(f)=all reals,ran(f1)=dom(f)=all reals.\text{dom}(f^{-1}) = \text{ran}(f) = \text{all reals}, \qquad \text{ran}(f^{-1}) = \text{dom}(f) = \text{all reals}. This agrees with the formula directly: x2+1>x\sqrt{x^{2}+1} > \lvert x \rvert, so x+x2+1>0x+\sqrt{x^{2}+1} > 0 for every real xx and the logarithm is always defined.

Check. f(ln3)=3132=8/32=43f(\ln 3) = \dfrac{3 - \frac13}{2} = \dfrac{8/3}{2} = \dfrac{4}{3}. Going back: f1 ⁣(43)=ln ⁣(43+169+1)=ln ⁣(43+53)=ln3f^{-1}\!\left(\frac43\right) = \ln\!\left(\frac43 + \sqrt{\frac{16}{9}+1}\right) = \ln\!\left(\frac43+\frac53\right) = \ln 3, since 259=53\sqrt{\frac{25}{9}} = \frac53. The round trip works.

Bonus consistency check: the inverse of an odd one-to-one function must be odd, and indeed f1(x)=ln ⁣(x2+1x)=ln(x2+1x)(x2+1+x)x2+1+x=ln1x2+1+x=f1(x).f^{-1}(-x) = \ln\!\left(\sqrt{x^{2}+1}-x\right) = \ln\frac{\left(\sqrt{x^{2}+1}-x\right)\left(\sqrt{x^{2}+1}+x\right)}{\sqrt{x^{2}+1}+x} = \ln\frac{1}{\sqrt{x^{2}+1}+x} = -f^{-1}(x).