10.6a Arcsine, Arccosine, and Arctangent

None of the six circular functions is one-to-one, so none has an inverse until its domain is restricted. Covers why the restriction is necessary and how the intervals are chosen, the resulting properties of arccosine, arcsine, arctangent and arccotangent, the asymmetric cancellation rules that make arccos of cos x fail to be x outside a specific interval, and the substitution technique that turns any composition of a circular function with an inverse one into an ordinary triangle problem.

Subject: Trigonometry · 67 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. Arcsine, Arccosine, and Arctangent

Title

Trigonometry · Section 10.6

How a periodic function is forced to have an inverse

2. What this lesson gives you

Objectives

None of the six circular functions is one-to-one, so strictly speaking none of them has an inverse at all. This lesson does what Section 5.2 did for the squaring function: cut the domain down until the function is one-to-one, invert that, and then live with the consequences.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 819-826 — the pages these objectives are drawn from

You have spent five sections building the circular functions and getting comfortable with the fact that they repeat forever. That repetition is now going to cause a problem, and the whole of this lesson is the fix.

The problem is simple to state. An inverse has to undo a function, and to undo something you have to be able to tell where it came from. A repeating function destroys exactly that information: the cosine of one number and the cosine of that number plus a full turn are identical, so given the answer you cannot say which one you started with.

The fix is to throw most of the function away. Keep one carefully chosen stretch on which the function never repeats itself, invert that, and accept the consequences. Almost everything that feels strange about inverse trigonometric functions, including why a calculator answers one way when you expected another, traces back to that one decision.

3. Before anything new: how many angles have cosine one half?

Warm-up

Discussion prompt

In Section 10.2 you solved equations like cosine of t equals one half and found infinitely many solutions. A calculator's inverse cosine button gives back exactly one number. Write down how those two facts can both be true.

Write something down before you reveal the answer. Even a rough sentence will do, because the value of this question is in noticing the tension, not in resolving it perfectly.

There really are infinitely many angles with a cosine of one half, and you found them all in Section 10.2 as two families. There really is only one number that a calculator hands back. Both statements are true, and they are not in conflict once you see that they are answers to different questions.

Solving an equation asks which inputs give this output, and the answer may be a whole family. Applying a function asks what does this function do to this input, and a function is by definition only allowed one answer. So the inverse cosine must choose a single representative from each family, and everything in this lesson follows from how it chooses.

4. Why a restriction comes first

Section

Part 1

5. No circular function is one-to-one

Concept

Figure (svg): The cosine curve over two full periods with a dashed horizontal line at one half crossing it at four marked points.

A horizontal line drawn anywhere inside the range meets the graph over and over. Until the domain is cut down, there is nothing to invert.

A function has an inverse exactly when it is one-to-one: no two inputs may share an output. Graphically that is the horizontal line test, and every periodic function fails it as badly as a function can.

\[ \cos\left(\tfrac{\pi}{3}\right) = \cos\left(-\tfrac{\pi}{3}\right) = \cos\left(\tfrac{7\pi}{3}\right) = \cdots = \tfrac12 \]

Every one of those inputs would have to come back from the inverse at the same time, which no function can do. Something has to be given up.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 819 — the opening paragraph

Look at the dashed line and count the crossings. There are four in the two periods shown, and the picture could be extended in both directions forever, adding two more crossings each time.

A function that has an inverse must pass the horizontal line test: no horizontal line may cross its graph more than once. The reason is worth saying out loud. If two different inputs share an output, then when the inverse is handed that output it would have to return both inputs at once, and that is precisely what a function cannot do.

So the statement is not that the circular functions are awkward to invert. It is that, as they stand, they have no inverse at all. Something genuine has to be surrendered before the word arccosine means anything.

6. The square root is this same trick, already learnt

Concept

Figure (svg): The parabola y equals x squared, dashed for negative x and solid for x at least zero, with the square-root curve as its mirror image in the dashed line y equals x; dots mark the points two, four and four, two.

Squaring is not one-to-one either. The square root inverts only the right-hand half, which is why the root of a square comes back positive.

Squaring is not one-to-one either, and the fix was never to abandon the square root. The domain was cut down to the numbers that are not negative, and the square root inverts only that half.

\[ (-2)^2 = 2^2 = 4 \qquad\text{but}\qquad \sqrt{4} = 2 \]

The surprise that follows is the one worth remembering: putting a number in and taking it out again does not always return the number you started with.

\[ \sqrt{(-2)^2} = 2 \ne -2 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 823 — the comparison the book draws in its remarks

You already own this idea, which is why the book points at it. Squaring is not one-to-one either, because a number and its negative square to the same thing, so strictly speaking the squaring function has no inverse.

Nobody responds to that by abandoning square roots. Instead the domain is quietly cut down to the numbers that are not negative, where squaring is one-to-one, and the square root is defined to invert that half. The dashed left arm of the parabola in the picture is the part that was discarded, and the square root curve is the mirror image of the solid right arm in the dashed line.

Now notice the price. Square minus two and you get four; take the square root and you get two, not minus two. Putting a number in and pulling it back out did not return the number. Hold onto that sentence: it is going to reappear, word for word in structure, when you meet the arccosine of a cosine.

7. Three demands on the interval that is kept

Concept

Figure (svg): The cosine curve over two periods drawn dashed, with the piece from zero to pi drawn thickly inside a shaded vertical band and its two endpoints marked.

On this one stretch the cosine falls steadily from 1 to minus 1: one-to-one, and still reaching every value it ever reached.

restricted domain — An interval on which the function is one-to-one while still attaining every value in its range. The inverse is defined as the inverse of that restricted function, so the inverse's range is exactly that interval.

For cosine the interval from zero to pi satisfies all three, and the book chooses it. The choice is a convention, not a theorem, but it is the convention every calculator and every later course uses.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 819 — restricting cosine to the interval from zero to pi

There are many intervals on which the cosine is one-to-one, so the choice needs criteria, and the three in the list are the ones that matter. The first is non-negotiable: without it there is nothing to invert. The second is about not wasting anything: if you kept only the stretch from zero to a quarter turn, the cosine there would only produce values from zero to one, and the arccosine would refuse any negative input.

The third is about comfort rather than necessity. You could stitch together a one-to-one function from scattered pieces, but its inverse would be a broken, awkward object. Keeping one unbroken stretch keeps the inverse a single smooth curve.

The interval from zero to pi meets all three, and the shaded band in the picture is it. Do not lose sight of the fact that this is a convention. A different unbroken stretch would also work, as the next few slides make you check. But every calculator, every table and every later course uses this one, so it is the one to learn.

8. Why sine cannot reuse the cosine's interval

Worked example

Show that the interval from zero to pi, which works for cosine, fails for sine.

\[ \text{is } \sin(x) \text{ one-to-one on } [0, \pi]\,? \]

Test two inputs with the same reference angle

Why: Pi over six and five pi over six both sit in the interval and both have reference angle pi over six.

\[ x_1 = \tfrac{\pi}{6}, \qquad x_2 = \tfrac{5\pi}{6}, \qquad x_1 \ne x_2 \]

Evaluate the sine at each

Why: Both angles are in the upper half of the circle, so both sines are positive.

\[ \sin\left(\tfrac{\pi}{6}\right) = \tfrac12, \qquad \sin\left(\tfrac{5\pi}{6}\right) = \tfrac12 \]

Figure (svg): The sine curve from just below zero to just past pi, with a dashed horizontal line at one half meeting it at two marked points, at pi over six and at five pi over six.

Sine rises and then falls on this interval, so it repeats every value except its peak. The cosine's interval is useless to it.

Conclude

Why: Two different inputs share an output, so the horizontal line test fails.

\[ \sin(x) \text{ is not one-to-one on } [0, \pi] \]

Move the interval and test again

Why: Shift the window down so that sine only ever rises.

\[ \text{on } \left[-\tfrac{\pi}{2}, \tfrac{\pi}{2}\right]: \quad \sin \text{ increases from } -1 \text{ to } 1 \]

Check that the new interval meets all three demands

Why: Strictly increasing gives one-to-one; the endpoints give sines of minus one and one, so the full range from minus one to one is still attained; and the sine curve is smooth throughout. All three hold, so this is the interval the book takes.

\[ \sin\left(-\tfrac{\pi}{2}\right) = -1, \qquad \sin\left(\tfrac{\pi}{2}\right) = 1 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 820 — restricting sine

It would be convenient if all six functions used the same interval, and this example shows why they cannot. The test is cheap: find two different inputs inside the interval that give the same output, and you are done.

Pi over six and five pi over six are a natural pair to try, because they are reflections of each other in the vertical line through the middle of the interval, and the sine curve is symmetric about that line. Both have a sine of one half, so the picture shows the dashed line cutting the curve twice.

The last step is the one to copy in your own work. It is not enough to reject an interval; you have to produce one that works and check all three demands on it. Sliding the window down so that it runs from minus a quarter turn to a quarter turn turns the hill into a steady climb, and a steadily climbing function can never repeat a value.

9. Which restrictions are one-to-one?

Sorting

Sort into buckets

Sort each function-and-interval pairing by whether the function is one-to-one there.

One-to-one: invertible here
cosine on [0, π]; sine on [−π/2, π/2]; tangent on (−π/2, π/2); cotangent on (0, π)
Repeats a value
cosine on [−π/2, π/2]; sine on [0, π]; cosine on [0, 2π]
yes
On each of these the function moves steadily in one direction across the whole interval: cosine falls from 1 to −1, sine rises from −1 to 1, tangent rises from very negative to very positive, cotangent falls the other way. A strictly monotone function can never repeat a value.
no
Each of these folds. Cosine is even, so on [−π/2, π/2] it gives the same value at t and at −t; sine climbs then falls on [0, π]; and cosine on a whole period repeats almost every value twice.

Sort these before checking, and for each one picture the shape of the graph over that stretch rather than trying to remember a rule.

The single test that settles every item is whether the function moves in one direction the whole way across. If it rises throughout, or falls throughout, it cannot repeat a value. If it turns round anywhere inside the interval, it repeats everything on one side of the turn.

The cosine on the interval from minus a quarter turn to a quarter turn is the one most people place wrongly. It looks symmetric and tidy, and that is exactly the problem: the symmetry means it takes the same value at each end, and at every matching pair in between.

10. A different book, a different interval

Prediction

Another textbook decides to restrict the cosine to the interval from pi to two pi instead, and defines its inverse cosine from that.

Predict first

Does that produce a valid inverse function, and would its outputs match the ones in this book?

  • No, the cosine is not one to one on that interval
  • No, that interval loses part of the range
  • Yes, it is a valid inverse, but its outputs differ from this book's arccosine
  • Yes, and it gives exactly the same outputs

Correct: Yes, it is a valid inverse, but its outputs differ from this book's arccosine

Why: On that interval the cosine rises steadily from minus one back to one, so it is one to one and still reaches every value in its range. All three demands hold, so the inverse exists. But the outputs live in that interval instead, so where this book returns pi over three, the other book would return five pi over three. The choice of interval is a convention, and conventions have to be agreed before answers can be compared.

Commit to one of the four options before revealing. Two of them say the construction fails, and if you are drawn to those, ask yourself precisely which of the three demands is supposed to be broken.

Neither is broken. On that interval the cosine climbs steadily from minus one back to one, so it is one-to-one and it still reaches every value it ever reaches. The construction works perfectly and produces a genuine inverse function.

What differs is the answers. This book's arccosine of one half is pi over three; the rival book's would be five pi over three, because that is the number in its chosen interval with the right cosine. Neither is wrong. They are answers to two slightly different questions, and that is why the convention has to be agreed before anyone can compare results. You will meet a real, live version of this disagreement in the next lesson, where two rival conventions for the arcsecant are both in current use.

11. Arccosine and arcsine

Section

Part 2

12. Reflect the restricted cosine in the line y equals x

Concept

Reversing a function swaps each input with its output, which on a graph swaps the two coordinates of every point. That is exactly a reflection in the line whose equation sets the two coordinates equal.

Figure (svg): A square plot showing the cosine curve dashed, the piece from zero to pi drawn thickly, the dashed line y equals x, and the arccosine curve as the mirror image of that thick piece, with the interval from zero to pi marked on the vertical axis.

Every point of the thick cosine piece has a twin on the arccosine with its coordinates swapped. The kept interval becomes the vertical bar: the range of arccosine.

\[ f(x) = \cos(x), \; 0 \le x \le \pi \qquad \xrightarrow{\;\text{reflect}\;} \qquad f^{-1}(x) = \arccos(x) \]

\[ \arccos: [-1, 1] \longrightarrow [0, \pi] \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 819 — the reflection across y = x, by Theorem 5.3

Take a moment with this picture, because five later slides depend on reading it fluently. The dashed wavy curve is the cosine you already know. The thick curve on it, running from the top left down to the right, is the only piece that is being kept. The other thick curve is the arccosine.

The relationship between the two thick curves is purely mechanical. An inverse swaps inputs for outputs, and a point on a graph is an input paired with its output, so inverting swaps the two coordinates of every point. Swapping coordinates is a reflection in the dashed diagonal line. Check it on a single point: the kept cosine passes through the point with first coordinate zero and second coordinate one; the arccosine passes through the point with those swapped.

The short vertical bar on the left is the part to carry away. It marks the stretch of the vertical axis that the arccosine occupies, and it has exactly the same length and position as the interval the cosine was restricted to along the horizontal axis. The restriction you chose becomes the range you are stuck with.

13. What the reflection actually swaps

Intuition

Read the picture twice: once left to right and once bottom to top. The interval the cosine was allowed to use along the horizontal axis reappears as the stretch of the vertical axis the arccosine can reach, and the values the cosine produced become the inputs the arccosine accepts.

restricted cosinearccosine
domain: 0 to πrange: 0 to π
range: −1 to 1domain: −1 to 1
starts high at (0, 1), fallsstarts high at (−1, π), falls
both endpoints includedboth endpoints included

That last row is worth pausing on. Cosine was restricted to a closed interval, ends and all, so the arccosine really does attain zero and pi. Ask for the arccosine of one and you get zero exactly, not something just above it.

Tables of this kind are easy to skim past, so read it as a set of claims about the picture on the previous slide and verify each one against it.

The first two rows are the swap itself, and they are the reason the arccosine refuses inputs bigger than one: the cosine never produced anything bigger than one, so the arccosine has nothing to give back for such a request. The third row says the two curves have the same shape, both falling, which they must, since reflecting cannot turn a falling curve into a rising one.

The last row is the one that gets forgotten. Because the cosine was restricted to a closed interval, ends included, the arccosine really does attain both ends of its range. The arccosine of one is zero exactly, and the arccosine of minus one is pi exactly. In two slides' time you will meet the arctangent, where this is no longer true, and the difference is not a detail.

14. Reading Theorem 10.26, the arccosine half

Notation

Annotate

On: \( \arccos(x) = t \iff 0 \le t \le \pi \;\text{and}\; \cos(t) = x \)

  • A single number, never a family of angles. Solving an equation gives infinitely many answers; applying this function gives one.
  • This condition is doing the choosing. Among all the angles whose cosine is x, exactly one lies here, and that one is the answer.
  • The defining equation. It also tells you that x must be between −1 and 1, which is the domain of arccosine.
  • Both directions are usable. To check an answer, feed it forward: take its cosine and confirm you get x back, and confirm it lies in the interval.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 820 — Theorem 10.26, properties of arccosine

This one line is the definition, and everything else about the arccosine is a consequence of it. Step through the annotations rather than reading it as a block.

The middle condition is the working part. Without it, the statement would just say that the cosine of the answer is the input, which infinitely many numbers satisfy. With it, exactly one of them qualifies, and the arccosine is a function again.

Make particular use of the fact that the statement runs both ways. That gives you a free check on every evaluation you will do for the rest of the lesson: take your answer, apply the cosine, confirm you get the input back, and confirm the answer is between zero and pi. If either half fails, your answer is wrong, and the second half is the half that usually fails.

15. Arcsine: the same move on a different interval

Concept

Sine is restricted to the interval from minus pi over two to pi over two, where it climbs steadily from minus one to one, and the reflection is taken in exactly the same line.

Figure (svg): A square plot showing the sine curve dashed, the piece from minus pi over two to pi over two drawn thickly, the dashed line y equals x, and the arcsine curve as its mirror image, with the interval from minus pi over two to pi over two marked on the vertical axis.

Sine keeps a symmetric interval, so arcsine is an odd function: reflect the picture through the origin and nothing moves.

\[ \arcsin(x) = t \iff -\tfrac{\pi}{2} \le t \le \tfrac{\pi}{2} \;\text{and}\; \sin(t) = x \]

Because that interval is symmetric about zero and sine is odd, arcsine inherits the symmetry: the arcsine of a negative number is minus the arcsine of its size. Arccosine has no such property, because its interval is not symmetric about zero.

\[ \arcsin(-x) = -\arcsin(x) \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 820 — Theorem 10.26, properties of arcsine

Everything on this slide is the previous construction with one substitution, so the effort should go into noticing what changed rather than relearning the method.

What changed is the interval, and because the new interval is symmetric about zero, the arcsine inherits a symmetry the arccosine does not have. Reflect the whole picture through the origin and it maps onto itself: that is what it means for the function to be odd, and in practice it means a minus sign can be pulled straight out through an arcsine.

Do not try to do the same with the arccosine. Its interval runs from zero to pi, entirely on one side of zero, so its outputs are never negative and no minus sign can ever come out. The arccosine of a negative number is an angle past the quarter turn, not a negative angle, and that asymmetry between the two functions accounts for a large share of sign errors in this chapter.

16. Where the word arc comes from

Picture it

Figure (svg): A unit circle with a thick arc drawn counter-clockwise from the point one, zero up to a terminal point in the second quadrant, a dashed vertical line dropping that point onto the horizontal axis, and the x-coordinate marked there.

Cosine turns an arc into an x-coordinate. Arccosine runs that backwards, turning an x-coordinate into an arc, and that is where the name comes from.

The cosine of a real number t is built by walking a distance t around the unit circle from the point one, zero and reading off the first coordinate where you land. So cosine turns an arc into a coordinate.

Its inverse must run that backwards: give it a coordinate and it hands back the arc. That is the arc in arccosine, and it is also why the book refuses the notation with a minus one exponent, which is too easy to read as one over the cosine.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 819 — the origin of the prefix, and the notation warning

The name is not decoration; it is an accurate description of what the function does, and knowing that makes the definitions easier to reconstruct if you ever forget them.

Start at the rightmost point of the unit circle and walk counter-clockwise a distance equal to your input. The thick arc in the picture is that walk. Where you stop, read off the first coordinate: that is the cosine of your input. So the cosine is a machine that turns an arc length into a coordinate.

Run the machine backwards and you have a device that takes a coordinate and returns the arc that produced it, which is exactly the arccosine. This also explains the book's refusal to use the notation with a minus one written as an exponent. Everywhere else in trigonometry a small number after the function name means a power, so that notation invites you to read it as one over the cosine, which is the secant and a completely different function.

17. Example 10.6.1(a) and (b): two exact values

Worked example

Find each exact value from the definition rather than from a calculator.

\[ \text{(a) } \arccos\left(\tfrac12\right) \qquad \text{(b) } \arcsin\left(\tfrac{\sqrt2}{2}\right) \]

Restate (a) as a search for one angle

Why: The arccosine of one half is the single number in the interval from zero to pi whose cosine is one half.

\[ \text{find } t: \quad 0 \le t \le \pi \;\text{and}\; \cos(t) = \tfrac12 \]

Name the special angle

Why: Cosine is one half at sixty degrees, which is pi over three radians, and that number lies inside the interval.

\[ \cos\left(\tfrac{\pi}{3}\right) = \tfrac12, \qquad 0 \le \tfrac{\pi}{3} \le \pi \]

Write the answer to (a)

Why: One number, not a family.

\[ \arccos\left(\tfrac12\right) = \tfrac{\pi}{3} \]

Restate (b) the same way

Why: Arcsine chooses from the interval between minus pi over two and pi over two.

\[ \text{find } t: \quad -\tfrac{\pi}{2} \le t \le \tfrac{\pi}{2} \;\text{and}\; \sin(t) = \tfrac{\sqrt2}{2} \]

Name the special angle and answer (b)

Why: Sine is root two over two at forty-five degrees, and pi over four is inside the interval.

\[ \arcsin\left(\tfrac{\sqrt2}{2}\right) = \tfrac{\pi}{4} \]

Check both by going forwards

Why: Feed each answer back through the original function and compare with the input; both also lie in the right interval.

\[ \cos\left(\tfrac{\pi}{3}\right) = 0.5 \;\checkmark \qquad \sin\left(\tfrac{\pi}{4}\right) \approx 0.7071 = \tfrac{\sqrt2}{2} \;\checkmark \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 821 — Example 10.6.1, parts 1(a) and 1(b)

The most useful habit in this whole section is the one the first step models: before touching a special-angle table, restate the problem as a hunt for a number satisfying two conditions. It costs one line and it prevents nearly every error in the section.

With the problem restated, the special-angle knowledge you already have does the rest. The cosine is one half at pi over three, and pi over three is inside the interval, so the hunt is over. The same for the arcsine: root two over two is the sine of pi over four, and pi over four is inside the interval.

In these two the answers came out positive and the interval condition was satisfied without any thought, which is exactly why they are the warm-up. The next slide changes the sign of the input, and the interval condition starts doing real work.

18. Example 10.6.1(c) and (d): negative inputs

Worked example

Negative inputs are where the two ranges part company, so do these slowly.

\[ \text{(c) } \arccos\left(-\tfrac{\sqrt2}{2}\right) \qquad \text{(d) } \arcsin\left(-\tfrac12\right) \]

Locate (c) in the range

Why: The cosine is negative only in the upper-left part of the circle, so t is past pi over two but at most pi.

\[ \tfrac{\pi}{2} < t \le \pi \;\text{and}\; \cos(t) = -\tfrac{\sqrt2}{2} \]

Use the reference angle

Why: The reference angle for a cosine of root two over two is pi over four, and in that part of the circle the angle is pi minus the reference angle.

\[ t = \pi - \tfrac{\pi}{4} = \tfrac{3\pi}{4} \]

Locate (d) in the range

Why: Arcsine's range runs from minus pi over two to pi over two, and a negative sine puts t below zero.

\[ -\tfrac{\pi}{2} \le t < 0 \;\text{and}\; \sin(t) = -\tfrac12 \]

Use oddness

Why: The arcsine of one half is pi over six, and arcsine is odd, so the sign passes straight through.

\[ \arcsin\left(-\tfrac12\right) = -\arcsin\left(\tfrac12\right) = -\tfrac{\pi}{6} \]

Verify both numerically

Why: Take the cosine of the first answer and the sine of the second and compare with the inputs.

\[ \cos\left(\tfrac{3\pi}{4}\right) \approx -0.7071 = -\tfrac{\sqrt2}{2}, \qquad \sin\left(-\tfrac{\pi}{6}\right) = -0.5 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 821 — Example 10.6.1, parts 1(c) and 1(d)

Negative inputs are where the two functions genuinely part company, so slow down here. Ask, before anything else, which part of the range a negative input sends you to.

For the arccosine the range is entirely on the positive side, running from zero to pi, and the cosine is negative only on the second half of that stretch. So a negative input means an answer between a quarter turn and a half turn, and you get it by subtracting the reference angle from pi. Here the reference angle is pi over four, so the answer is three pi over four.

For the arcsine the range straddles zero, and the sine is negative on the lower half. So a negative input means a negative answer, and since the arcsine is odd you can simply pull the minus sign out. The contrast is worth saying explicitly: a negative input to the arccosine gives a large positive answer, while a negative input to the arcsine gives a negative answer. Mixing those up is the single commonest error in the exercise set.

19. Find the error: an answer outside the range

Error analysis

Annotate

On: \( \arcsin\left(-\tfrac12\right) = \tfrac{7\pi}{6} \)

  • Seven pi over six really does have a sine of minus one half, so this is not a sign slip or a reference-angle slip. The arithmetic is fine.
  • Seven pi over six is about 3.67, well above pi over two. Arcsine can only return numbers between minus pi over two and pi over two, so this number is not in its range at all.
  • The equation sin t = −1/2 was solved instead of the function being applied. Solving gives a family; the function must pick one member of it.
  • From that family, keep only the members inside the range. Here that is −π/6, and it is the answer.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 823 — the book's remarks on the commonest errors

Study the line before revealing the annotations, and resist the urge to check the arithmetic, because the arithmetic is correct.

That is what makes this error so durable. The angle offered really does have a sine of minus one half. Somebody who checks their work by taking the sine of their answer will confirm it and move on satisfied. The check has to include the second condition, the one about the interval, and that is the one this answer fails.

The underlying confusion is between solving and evaluating. The family of angles with a sine of minus one half is infinite, and the offered angle is a perfectly good member of it. But the arcsine is not asking for a member of the family; it is asking for the one member inside its range. Get into the habit of finishing every evaluation by checking the answer against the interval, out loud if necessary.

20. Match each input to its arccosine

Matching

Match the pairs

  • a. arccos(1/2)
  • b. arccos(−1/2)
  • c. arccos(√2/2)
  • d. arccos(−√3/2)
  • e. arccos(0)
  • p. π/4
  • q. π/3
  • r. π/2
  • s. 2π/3
  • t. 5π/6

Why: Every answer lies between zero and pi, so no answer is ever negative. A positive input gives an answer below pi over two and a negative input gives one above it, which halves the work before you even reach for a reference angle. The input zero sits exactly on the boundary and returns pi over two.

Do the matching before checking, and use the shortcut in the explanation rather than computing each one from scratch.

The shortcut halves the work. Every answer lies between zero and pi, so none is negative. A positive input means the answer is in the first half of that stretch, below a quarter turn; a negative input means the second half, above it. That splits five items into two groups before you touch a reference angle.

Then the reference angle finishes it. For a positive input the answer is the reference angle itself; for a negative input it is pi minus the reference angle. The input zero is the hinge between the two cases, and its answer is exactly the quarter turn.

21. Trap: reporting every angle instead of the one

Trap

The trap

A common way to answer the first worked example:

\[ \arccos\left(\tfrac12\right) = \pm\tfrac{\pi}{3} + 2\pi k \]

Wrong. That is the solution set of an equation, not the value of a function.

The fix

An arccosine is a number. The infinitely many angles are real and they matter when you are solving an equation, but the function has to choose, and it always chooses the one in its range.

\[ \arccos\left(\tfrac12\right) = \tfrac{\pi}{3} \]

Keep the two jobs apart: solve an equation and you list families; evaluate an inverse and you write one number.

This is the same confusion as the error-analysis slide, but written the other way round, and it is worth seeing both faces of it. There one wrong member of the family was handed in; here the whole family is.

Giving the whole family feels safe, and in a solving question it would be exactly right. But an arccosine is a number, in the same way that the square root of four is a number and not the pair plus and minus two. If you found yourself writing a family, you answered a question that was not asked.

A useful test when you are unsure which job you are doing: look at whether the question says solve or says find the exact value. Solve means list the families. Find the value means write one number and check it is in the range.

22. Push past the ends of the domain

Edge cases

\[ \text{(i) } \sin\left(\arcsin\left(\tfrac54\right)\right) \qquad \text{(ii) } \cos(\arccos(\pi)) \]

Discussion prompt

Exercises 61 and 66 ask for these two. The cancellation rule seems to answer both instantly. Decide each, and say what the rule's small print is protecting you from.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 842 — Exercises 61 and 66

Answer both before revealing. The temptation here is strong, because the cancellation rule seems to give the answer without any work at all.

The trap is that the rule is being applied from the outside in, when the expression has to be built from the inside out. There is no arcsine of five quarters, because the sine never produces a number bigger than one, so the inner expression does not name anything. An outer function applied to nothing is still nothing.

This is why the rules in Theorem 10.26 are always written with the words provided, and a condition. Those words are doing work. When you quote a rule in an exam answer, quote its condition too and show that it holds; a bare rule with no condition is an incomplete argument even when the answer is right.

23. The cancellation rules

Section

Part 3

24. One direction is free, the other is conditional

Concept

Inverses undo each other, but only where both are defined, and the two orders are not on the same footing at all.

\[ \cos(\arccos(x)) = x \quad \text{provided } -1 \le x \le 1 \]

\[ \arccos(\cos(x)) = x \quad \text{provided } 0 \le x \le \pi \]

The first condition is almost free: it just says the arccosine has to exist. The second is a real restriction, because the cosine happily accepts every real number, and most of them are outside the interval from zero to pi.

\[ \sin(\arcsin(x)) = x \;\; (-1 \le x \le 1), \qquad \arcsin(\sin(x)) = x \;\; \left(-\tfrac{\pi}{2} \le x \le \tfrac{\pi}{2}\right) \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 820 — Theorem 10.26, the cancellation clauses

Four statements, two of which carry a serious condition and two of which barely carry one at all. The asymmetry is the whole content of this part of the lesson, so it is worth being able to say which is which without looking.

The two conditions look superficially similar, both being intervals, but they mean different things. The first kind only says that the inner expression has to exist: if the input is outside the domain of the arccosine, there is nothing to cancel. The second kind says something much stronger, because the inner cosine is perfectly happy with any real number at all; most real numbers make the statement false even though every expression in it is defined.

So when you see a composition, your first question should be which of the two shapes it has. If the inverse is on the inside, you are almost certainly safe. If the inverse is on the outside, stop and check the angle.

25. Why the inside-out direction is the risky one

Intuition

Follow a number through each composition and the asymmetry stops being a rule to memorise.

Start with a number between minus one and one and apply arccosine first. The result is automatically inside the interval from zero to pi, because that is the only place arccosine ever lands, so the cosine that follows receives a perfectly legal input and hands the number straight back.

Now start with an angle and apply cosine first. The cosine forgets where the angle was: many angles share that value. When the arccosine then chooses, it chooses the representative from zero to pi, which is the angle you started with only if you happened to start there.

\[ \arccos(\cos(x)) = \text{the representative of } x \text{ in } [0, \pi] \]

Rules are easier to keep straight when you can rebuild them, so trace a number through each order rather than memorising which one has the condition.

Going one way, the arccosine runs first. Whatever it returns is guaranteed to be inside the interval from zero to pi, because that is the only place it ever lands. So when the cosine receives it, we are in the region where the two functions are genuine inverses, and the number comes straight back.

Going the other way, the cosine runs first, and the cosine is forgetful. Infinitely many angles produce the same value, and once the value is computed there is no trace of which one you gave it. The arccosine then has to choose, and it always makes the same choice: the representative from zero to pi. If your angle happened to be that representative you get it back; otherwise you get a stranger with the same cosine.

26. Reading the four cancellation statements

Notation

Annotate

On: \( \cos(\arccos(x)) = x \qquad \arccos(\cos(x)) = x \)

  • This one is safe whenever it makes sense at all. Any legal input to arccosine comes back unchanged, because the middle value always lands in the interval where cosine agrees.
  • This one has teeth. Cosine accepts every real number, and arccosine returns only numbers between zero and pi, so the two can only agree on that interval.
  • Before writing either line, look at what is on the inside and ask whether it is legal for that direction. For the second rule that means: is the angle between zero and pi?
  • Nothing is lost. Evaluate the inside, then apply the outer function to the number you get. That always works, and it is what the book does in Example 10.6.1(f).

Two statements side by side, and the annotations are mostly about the difference between them, so read them as a pair rather than one at a time.

The practical advice is in the third annotation. Before you write down a cancellation, look at what is inside and ask whether it is legal for that direction. That question takes two seconds and it is the difference between a complete argument and a guess that happens to be right.

The fourth annotation is the one that removes the fear. Failing the condition costs you nothing but a line of work. Evaluate the inside, get a number, then apply the outer function to that number. That route always works, whether the condition holds or not, so if you are ever unsure, take it.

27. What these compositions really look like

Picture it

Figure (svg): Two panels over minus two pi to two pi. Left: the graph of arccosine of cosine x, a triangular wave between zero and pi, with the dashed line y equals x touching it only between zero and pi. Right: the graph of arcsine of sine x, a zigzag between minus pi over two and pi over two, matching y equals x only on that interval.

Both compositions are defined for every real number, and both are almost never the identity. They agree with the input exactly on the interval the function was restricted to.

Each composition is a perfectly ordinary function of a real variable, and drawing it settles the matter. The straight dashed line is where the composition would equal its input; the thick highlighted piece is the only stretch where the graph actually sits on it.

Everywhere else the graph folds back on itself, always keeping its output inside the range. Outside the highlighted interval the composition still returns an answer, just not the one you put in.

These two graphs are worth more than any amount of talk about conditions, because they turn a rule with small print into a picture you can recall in one glance.

In each panel the faint diagonal is where the composition would equal its input, and the thick bright piece marks the only stretch where the graph actually lies along it. Everywhere else the graph turns and comes back, because it is not allowed to leave the range: the left graph is trapped between zero and pi, and the right one between minus a quarter turn and a quarter turn.

Notice that both compositions are defined for every real number. Nothing is undefined here, and no calculation fails. The composition just quietly returns a different number from the one you gave it, which is far more dangerous than an error message, because nothing warns you.

28. Example 10.6.1(e): inside the interval

Worked example

The book works this one twice, once by theorem and once from the definition, to show they agree.

\[ \arccos\left(\cos\left(\tfrac{\pi}{6}\right)\right) \]

Check the condition

Why: The rule needs the inner angle to lie between zero and pi, and pi over six is about 0.524.

\[ 0 \le \tfrac{\pi}{6} \le \pi \;\checkmark \]

Apply the rule

Why: The condition holds, so the composition returns its input.

\[ \arccos\left(\cos\left(\tfrac{\pi}{6}\right)\right) = \tfrac{\pi}{6} \]

Now redo it from the inside out

Why: Evaluate the cosine first and treat the result as an ordinary arccosine problem.

\[ \cos\left(\tfrac{\pi}{6}\right) = \tfrac{\sqrt3}{2} \;\Longrightarrow\; \arccos\left(\tfrac{\sqrt3}{2}\right) \]

Find the one angle in the range

Why: The number in the interval from zero to pi whose cosine is root three over two is pi over six.

\[ 0 \le \tfrac{\pi}{6} \le \pi, \quad \cos\left(\tfrac{\pi}{6}\right) = \tfrac{\sqrt3}{2} \;\Longrightarrow\; \arccos\left(\tfrac{\sqrt3}{2}\right) = \tfrac{\pi}{6} \]

Check that the two routes agree

Why: Both give pi over six, about 0.5236, which is the point: when the condition holds, the shortcut and the long way are the same calculation.

\[ \tfrac{\pi}{6} \approx 0.5236 \;=\; 0.5236 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 821 — Example 10.6.1, part 1(e)

The book deliberately solves this twice, and you should read it as two separate demonstrations rather than as one long solution.

The first route quotes the rule, having first checked the condition. That check is not padding: pi over six is about half a radian, comfortably inside the interval from zero to pi, and writing that line is what entitles you to the shortcut.

The second route ignores the rule entirely and just computes. The cosine of pi over six is root three over two, and then you hunt for the number in the interval whose cosine is that. You arrive at pi over six again. When the condition holds, the shortcut is not a different method; it is the same calculation with the middle removed.

29. Example 10.6.1(f): outside the interval

Worked example

The same composition, one angle further round the circle, and the answer changes.

\[ \arccos\left(\cos\left(\tfrac{11\pi}{6}\right)\right) \]

Check the condition

Why: Eleven pi over six is about 5.76, which is larger than pi, so the rule does not apply and cannot be used.

\[ \tfrac{11\pi}{6} \approx 5.76 > \pi \;\Longrightarrow\; \text{rule unavailable} \]

Evaluate the inside

Why: Eleven pi over six sits just below the horizontal axis on the right, where the cosine is positive and the reference angle is pi over six.

\[ \cos\left(\tfrac{11\pi}{6}\right) = \tfrac{\sqrt3}{2} \]

Now take the arccosine of that number

Why: This is the previous slide's last step, unchanged.

\[ \arccos\left(\tfrac{\sqrt3}{2}\right) = \tfrac{\pi}{6} \]

Write the answer

Why: The composition returns pi over six, not the eleven pi over six that went in.

\[ \arccos\left(\cos\left(\tfrac{11\pi}{6}\right)\right) = \tfrac{\pi}{6} \ne \tfrac{11\pi}{6} \]

Check against the graph and against the square-root analogy

Why: On the left panel of the earlier figure, the height above 5.76 is about 0.52, which is pi over six, not 5.76. It is the same phenomenon as the root of a square: the composition returns the representative from the restricted interval.

\[ \arccos(\cos(5.76)) \approx 0.5236, \qquad \sqrt{(-2)^2} = 2 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 822-823 — Example 10.6.1, part 1(f), and the remarks that follow

One angle further round the circle and the answer changes completely, which is the whole reason this example exists directly after the previous one.

The first step is the one that matters, and it is a rejection rather than a calculation. Eleven pi over six is nearly six radians, well past pi, so the rule is simply unavailable. Notice that nothing has gone wrong yet; you have merely established that a particular shortcut cannot be used.

From there the inside-out route runs exactly as before and delivers pi over six. Look at the last step, where the answer is compared against the graph and against the square root. Both comparisons say the same thing in different languages: the composition hands back the representative of your angle from the restricted interval, just as the square root of a square hands back the representative from the non-negative numbers.

30. Example 10.6.1(g): the safe direction, proved

Worked example

Why does the other order never need a condition beyond the domain? Do it without the theorem.

\[ \cos\left(\arccos\left(-\tfrac35\right)\right) \]

Name the inner value

Why: Give the arccosine a letter so there is something to reason about.

\[ t = \arccos\left(-\tfrac35\right) \]

Write down what that name means

Why: By the definition, t is in the interval from zero to pi and its cosine is the input.

\[ 0 \le t \le \pi \;\text{and}\; \cos(t) = -\tfrac35 \]

Substitute the name back into the question

Why: The expression asked for is just the cosine of t, which the previous line already gives.

\[ \cos\left(\arccos\left(-\tfrac35\right)\right) = \cos(t) = -\tfrac35 \]

Notice what was never needed

Why: Nowhere did the argument use the value minus three fifths; it only used that the value was a legal input to arccosine.

\[ \cos(\arccos(x)) = x \quad\text{for every } x \text{ in } [-1, 1] \]

Check the boundary of the claim

Why: At the ends of the domain it still holds, and just outside there is nothing to check because the inside is undefined.

\[ \cos(\arccos(1)) = \cos(0) = 1, \qquad \cos(\arccos(-1)) = \cos(\pi) = -1 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 822 — Example 10.6.1, part 1(g)

The book could have written one line here by quoting the rule. It does the work instead, and the reason is that the argument generalises to everything in the last part of this lesson.

The move is to give the inner expression a name. The moment you do, the definition hands you two facts about that name for free: an equation and an interval. In this problem the equation alone finishes the job, because the thing being asked for is precisely the left side of that equation.

The fourth step is the punchline. Read the argument again and notice that the particular value minus three fifths played no part. All that was used is that the value was a legal input to the arccosine. That is why this direction needs no condition beyond the domain, and it is a much better reason to believe the rule than having read it in a theorem.

31. Trap: cancelling without checking the condition

Trap

The trap

The line that costs the most marks in this section:

\[ \arcsin\left(\sin\left(\tfrac{3\pi}{4}\right)\right) = \tfrac{3\pi}{4} \]

Wrong. Three pi over four is about 2.36, larger than pi over two.

\[ \tfrac{3\pi}{4} > \tfrac{\pi}{2} \]

The fix

Work inside out instead. The sine of three pi over four is root two over two, and the arcsine of that is pi over four.

\[ \sin\left(\tfrac{3\pi}{4}\right) = \tfrac{\sqrt2}{2} \]

\[ \arcsin\left(\tfrac{\sqrt2}{2}\right) = \tfrac{\pi}{4} \]

Every answer an arcsine gives must be between minus pi over two and pi over two. If your answer is not, you have not checked the condition.

If you lose marks anywhere in this section, it will be here. The wrong line looks like an application of a rule you have just been taught, and it is written with confidence.

The defence is a single reflex: before cancelling an inverse against a function, look at the inner angle and compare it with the inverse's range. Three pi over four is a little over two and a third radians; the arcsine stops at about one and a half. It does not qualify, so the rule is off the table.

There is also a defence that works even if you forget to check. Every answer an arcsine gives must lie between minus a quarter turn and a quarter turn. Glance at your final answer; if it is outside, you have made this mistake, whatever route you took to get there.

32. Which of these cancel on sight?

Discrimination

Sort into buckets

Sort each composition by whether the cancellation rule applies straight away or you must evaluate the inside first.

Cancels immediately
cos(arccos(0.4)); arccos(cos(0.4)); sin(arcsin(−0.9)); arcsin(sin(−1))
Work it from the inside out
arccos(cos(5)); arcsin(sin(2))
now
Either the outer function is the trig one and the inner input is between −1 and 1 (items a and d), or the outer function is the inverse and the inner angle already lies in its range: 0.4 is inside [0, π], and −1 is inside [−π/2, π/2].
inside
In both of these the inner angle is outside the range of the inverse that follows. 5 is bigger than π ≈ 3.14, and 2 is bigger than π/2 ≈ 1.57, so the composition will fold the answer back and you must evaluate the inside first.

Sort these before checking, and work item by item with one question each: is the inverse on the inside or the outside, and if it is on the outside, is the inner angle in its range?

The items with the trig function on the outside are the easy ones. In both of them the number being fed to the inverse is comfortably between minus one and one, so the inverse exists and the cancellation is immediate.

The items with the inverse on the outside need the comparison. Numbers like 0.4 and minus one are small and sit inside the relevant range, so those cancel too. The numbers 5 and 2 are the ones to catch: five is bigger than pi, and two is bigger than a quarter turn, so both compositions will fold their answers back and must be worked from the inside.

33. Fill the missing step: an arcsine of a sine

Fill the middle

Fill in the blanks

\arcsin\left(\sin\left(\tfrac\tfrac{\sqrt{2}}{2}\tfrac{\pi}{4}\right)\right) = \arcsin\left(___\right) = ___

Why: Three pi over four is outside the arcsine's range, so no cancellation is allowed. Evaluate the sine first: three pi over four has reference angle pi over four and lies above the horizontal axis, so its sine is root two over two. The arcsine of that is pi over four, which is inside the range, as every arcsine answer must be.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 842 — Exercise 89

Fill both blanks before revealing, and notice that the structure of the line is itself the lesson: two steps, evaluate then invert, rather than one step of cancellation.

The first blank is ordinary special-angle work. Three pi over four sits in the upper left of the circle, where the sine is positive, and its reference angle is pi over four, so the sine is root two over two.

The second blank is where the range does its work. Among all the angles whose sine is root two over two, the arcsine must return the one between minus a quarter turn and a quarter turn, and that is pi over four. The angle you started with, three pi over four, is not in that interval, which is precisely why the answer is not the number you began with.

34. Break the claim that the two rules are alike

Counterexample

Discussion prompt

Someone writes both cancellation rules with no conditions at all, claiming they are symmetric. Give one number that breaks one of them, say which rule survives and why, and describe in one sentence exactly what the broken composition does return.

Write your counterexample before revealing. Producing a counterexample is a skill in its own right, and the useful discipline is to look for the cheapest one rather than the cleverest.

Any angle outside the interval from zero to pi breaks the inside-out rule, so there are infinitely many witnesses and you only have to name one. The book's own example works, but so does five, or seven, or minus one.

The second half of the question is the part that teaches. Saying which rule survives, and why, forces you to articulate the asymmetry: an arccosine can only produce numbers from the interval the cosine was restricted to, so the cosine that follows never sees anything illegal. Once you can say that sentence unprompted, you will never need to memorise which rule carries the condition.

35. Arctangent and arccotangent

Section

Part 4

36. Tangent, restricted to its fundamental cycle

Concept

Tangent already runs through every real value once on the open interval between minus pi over two and pi over two, so no choice is needed: that cycle is the restriction.

Figure (svg): A square plot with the tangent curve dashed on its outer branches, the middle branch drawn thickly between two dashed vertical asymptotes, the dashed line y equals x, and the arctangent curve rising between two dashed horizontal asymptotes.

Reflecting swaps the coordinates, so it swaps vertical asymptotes for horizontal ones. The arctangent can never reach plus or minus pi over two.

\[ \arctan(x) = t \iff -\tfrac{\pi}{2} < t < \tfrac{\pi}{2} \;\text{and}\; \tan(t) = x \]

The ends are open, and that is not a technicality. Tangent is undefined at plus and minus pi over two, so those two numbers are never outputs of the arctangent no matter how large the input.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 823 — restricting tangent to its fundamental cycle

The tangent makes life easy for once. It does not need a choice of interval imposed on it, because its fundamental cycle is already one-to-one and already covers every real value exactly once.

Read the picture the same way as the earlier ones: dashed outer branches discarded, thick middle branch kept, dashed diagonal, and the arctangent as the mirror image. The two vertical dashed lines are the asymptotes of the tangent and the two horizontal ones are the asymptotes of the arctangent, and each pair is the mirror image of the other.

The open ends matter. The tangent is not defined at the quarter turns, so those numbers were never in the restricted domain, so they are not in the arctangent's range. No matter how enormous the input, the arctangent's answer is strictly less than a quarter turn. There is no input that achieves it.

37. Asymptotes trade places

Intuition

Reflecting in the line where the coordinates are equal turns a vertical line into a horizontal one, so the tangent's two vertical asymptotes become the arctangent's two horizontal ones.

\[ x = \pm\tfrac{\pi}{2} \;\xrightarrow{\;\text{reflect}\;}\; y = \pm\tfrac{\pi}{2} \]

Read off the consequences. The arctangent accepts every real number, because the tangent produced every real number. It is bounded above and below, because the tangent was restricted to a bounded interval. And it never reaches its bounds, because the tangent never reached the ends of its interval.

\[ \lim\limits_{x \to \infty} \arctan(x) = \tfrac{\pi}{2}^{-}, \qquad \lim\limits_{x \to -\infty} \arctan(x) = -\tfrac{\pi}{2}^{+} \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 824 — Theorem 10.27, the end behaviour

This is one of those facts that is obvious once stated and easy to forget when it matters, so link it to the reflection rather than storing it separately. Reflecting in the diagonal turns vertical lines into horizontal ones, so of course the asymptotes swap orientation.

The three consequences in the middle paragraph are all worth reading off the same picture. The arctangent accepts every real number because the tangent produced every real number; it is bounded because the tangent's inputs were bounded; and it never reaches its bounds because the tangent never reached the ends of its interval.

That third consequence is the practically important one. If you ever produce a quarter turn as the value of an arctangent, something has gone wrong. The value is approached as the input grows, and the limit statements on the slide say so precisely, but it is never attained.

38. Cotangent needs only one branch

Concept

The cotangent's branch on the open interval from zero to pi already falls from arbitrarily large positive values to arbitrarily large negative ones, so that single branch covers the whole range and the book takes it.

Figure (svg): A square plot with the cotangent curve dashed on its outer branches, the branch between zero and pi drawn thickly between two dashed vertical asymptotes, the dashed line y equals x, and the arccotangent curve falling from a horizontal asymptote at pi to one at zero.

The cotangent branch between zero and pi already runs from plus infinity down to minus infinity, so one branch is enough, and its inverse decreases.

\[ \operatorname{arccot}(x) = t \iff 0 < t < \pi \;\text{and}\; \cot(t) = x \]

Its asymptotes were the vertical lines at zero and at pi, so the arccotangent has horizontal asymptotes at zero and at pi, and it decreases: large positive inputs give answers near zero, large negative inputs give answers near pi.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 823-824 — restricting cotangent to the interval from zero to pi

The cotangent is handled exactly like the tangent, and the only thing to notice is which branch is taken and which way it runs.

The branch between zero and pi is the one that contains no jump, and across it the cotangent falls from arbitrarily large positive values down through zero to arbitrarily large negative ones. That single branch therefore covers the entire range, and no other branch is needed.

Because the cotangent falls, so does its inverse. This is the one inverse in the set that decreases, and it catches people out: a large positive input gives an answer close to zero, and a large negative input gives an answer close to pi. If you ever have to reconstruct the arccotangent's behaviour, remember which way the branch runs and the rest follows.

39. Reading Theorem 10.27

Notation

Annotate

On: \( \arctan(x) = t \iff -\tfrac{\pi}{2} < t < \tfrac{\pi}{2} \;\text{and}\; \tan(t) = x \)

  • Both ends are open, unlike arccosine and arcsine. An answer of exactly pi over two is always wrong, however large the input.
  • There is no condition on the input at all. The tangent takes every real value, so the arctangent accepts every real number: arctan(1000) and arctan(−10⁶) are both perfectly ordinary.
  • This holds for every real x, with no small print, precisely because the domain of arctangent is everything.
  • This one carries the usual condition: x must lie strictly between minus pi over two and pi over two. Same asymmetry as before.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 824 — Theorem 10.27

The statement looks almost identical to the arcsine's, which is exactly the danger, so read the annotations for the differences rather than for the similarities.

The first difference is in the inequality signs. They are strict at both ends here, and the reason is not arbitrary: if a quarter turn were allowed as an answer, the second half of the statement would ask for the tangent of a quarter turn, which does not exist. The closed version of this theorem contradicts itself.

The second difference is that there is no condition on the input at all. Every real number has an arctangent, which is why the trig-outside cancellation for the arctangent is completely unconditional, unlike its arccosine and arcsine cousins. The inverse-outside cancellation, though, carries a condition just like all the others, and that is the pattern to carry forward.

40. Example 10.6.2(a) and (b): two exact values

Worked example

Two more evaluations, now with the tangent and cotangent ranges in play.

\[ \text{(a) } \arctan(\sqrt3) \qquad \text{(b) } \operatorname{arccot}(-\sqrt3) \]

Restate (a)

Why: Find the number strictly between minus pi over two and pi over two whose tangent is root three.

\[ \text{find } t: \quad -\tfrac{\pi}{2} < t < \tfrac{\pi}{2} \;\text{and}\; \tan(t) = \sqrt3 \]

Name the special angle

Why: Tangent is root three at sixty degrees, and pi over three is about 1.047, comfortably inside the interval.

\[ \arctan(\sqrt3) = \tfrac{\pi}{3} \]

Restate (b)

Why: Arccotangent chooses from the open interval from zero to pi, and a negative cotangent puts the answer past pi over two.

\[ \text{find } t: \quad \tfrac{\pi}{2} < t < \pi \;\text{and}\; \cot(t) = -\sqrt3 \]

Use the reference angle

Why: A cotangent of size root three comes from a reference angle of pi over six, and in that part of the circle the angle is pi minus the reference angle.

\[ t = \pi - \tfrac{\pi}{6} = \tfrac{5\pi}{6} \;\Longrightarrow\; \operatorname{arccot}(-\sqrt3) = \tfrac{5\pi}{6} \]

Check both by computing the ratios

Why: The cotangent of five pi over six is its cosine over its sine, and the tangent of pi over three is its sine over its cosine.

\[ \cot\left(\tfrac{5\pi}{6}\right) = \frac{-\sqrt3/2}{1/2} = -\sqrt3, \qquad \tan\left(\tfrac{\pi}{3}\right) = \frac{\sqrt3/2}{1/2} = \sqrt3 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 825 — Example 10.6.2, parts 1(a) and 1(b)

Same routine as the earlier evaluations, so the interest is entirely in the second part, where the range is one you have only just met.

For the arctangent, the interval is the open one between the two quarter turns, and pi over three sits comfortably inside it, so part one is quick. Notice that the interval condition was still written down; it costs a line and it is what makes the answer defensible.

For the arccotangent, the range runs from zero to pi, exactly like the arccosine's, so a negative input pushes you past the quarter turn. The reference-angle move is the same as it was for the arccosine of a negative number: find the reference angle from the size of the input, then subtract it from pi. If you find yourself writing a negative answer for an arccotangent, you have used the arctangent's range by mistake.

41. Example 10.6.2(c): cancelling with no condition at all

Worked example

The book does this one the long way on purpose, to show that nothing is being hidden.

\[ \cot(\operatorname{arccot}(-5)) \]

Name the inner value

Why: Set a letter equal to the arccotangent.

\[ t = \operatorname{arccot}(-5) \]

Write down what the name means

Why: By the definition, t lies strictly between zero and pi and its cotangent is minus five.

\[ 0 < t < \pi \;\text{and}\; \cot(t) = -5 \]

Substitute back

Why: The expression asked for is the cotangent of t, which the line above already gives.

\[ \cot(\operatorname{arccot}(-5)) = \cot(t) = -5 \]

See why there was no condition to check

Why: The arccotangent accepts every real number, so there is no input that could have failed.

\[ \cot(\operatorname{arccot}(x)) = x \quad\text{for every real } x \]

Check with an input the other way round

Why: The reverse composition does carry a condition, and minus five is outside the interval from zero to pi, so it folds back.

\[ \operatorname{arccot}(\cot(-5)) \approx 1.2832 \ne -5 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 825 — Example 10.6.2, part 1(c)

The book works this the long way even though one line would do, and the reason is that the long way explains why the one line is legal.

The name-it move is the same as in the arccosine version: set a letter equal to the inverse and read off the two facts the definition gives you. Once again the equation alone finishes it, because the thing asked for is the left side of that equation.

The check at the end is the part to look at hardest. The reverse composition, with the arccotangent on the outside, gives about 1.28 rather than minus five, because minus five is nowhere near the interval from zero to pi. Two compositions built from the same two functions, one of them unconditional and the other wrong by a wide margin: that is the asymmetry again, in the last of the four functions.

42. The four inverse functions on one table

Comparison

Comparison matrix

functiondomainrangeare the ends attained?
arccos x-1 to 10 to piyes, both
arcsin x-1 to 1-pi/2 to pi/2yes, both
arctan xall real numbers-pi/2 to pi/2no, neither
arccot xall real numbers0 to pino, neither

Fill the blanks before checking. This table is the single most useful thing to be able to reproduce from memory in this lesson, and the fastest way to learn it is to write it out rather than read it.

Two patterns hold it together. The two functions built from cosine and cotangent both use the interval from zero to pi; the two built from sine and tangent both use the symmetric interval about zero. And the two built from the bounded functions, cosine and sine, take inputs only between minus one and one, while the two built from the unbounded functions accept everything.

The last column is the one that gets learned last and tested first. The closed intervals belong to the functions whose parents were defined at the ends; the open ones belong to the functions whose parents blow up there.

43. The four ranges drawn on one axis

Picture it

Figure (svg): Four horizontal number lines from minus pi to pi, one per inverse function, each with its range drawn as a thick segment; arccosine and arcsine have solid endpoints, arctangent and arccotangent have hollow ones.

Arccosine and arcsine reach their endpoints, because cosine and sine were restricted to closed intervals. Arctangent and arccotangent never do, because tangent and cotangent blow up there.

Two ranges start at zero and run to pi; two are centred on zero and run from minus pi over two to pi over two. Which pair you are in is decided by the function, and getting it wrong is the single most common source of a wrong sign.

The endpoints follow the same logic. Cosine and sine were restricted to closed intervals and are defined at the ends, so those ends are attained. Tangent and cotangent blow up at their ends, so those ends are approached and never reached.

All four ranges on one axis, drawn to the same scale, so that the comparisons the table made in words can be checked by eye.

The vertical alignment is the point. The arccosine and arccotangent bars occupy the same stretch of the axis, and the arcsine and arctangent bars occupy the same stretch as each other. Two shapes, not four, which cuts what you have to remember in half.

The endpoints carry the rest of the information. A solid dot means that value really is attained, and a hollow one means the bar gets arbitrarily close and stops short. If you ever write an answer that lands on a hollow dot, you have made an error somewhere earlier.

44. Steepen the arctangent and watch the band

Tweak it

Parameter explorer

Slide the number a from small to large. The curve gets steeper and steeper near the middle. Does it ever leave the horizontal band it lives in, and what does that band have to do with the range of arctangent?

\[ y = \arctan(a x), \quad a = {a} \]

  • a — from 0.2 to 12: steepness a

Move the slider slowly from its smallest setting upward and watch what changes and what does not. The steepness through the middle changes dramatically; the height the curve settles at does not change at all.

That is the range, made visible. Multiplying the input by a large number makes the arctangent shoot up almost immediately, but it still only ever climbs to the same invisible ceiling, and it still only ever approaches it. The ceiling and floor sit at the two quarter turns, and no setting of the slider ever reaches them.

Push the number as high as it goes and the curve looks almost like a step, flat at one height for negative inputs and flat at another for positive ones. Even there, at every single input, the value is strictly between the two levels. That is what an open interval looks like as a picture.

45. Find the error: closed ends on an open range

Error analysis

Annotate

On: \( \arctan(x) = t \iff -\tfrac{\pi}{2} \le t \le \tfrac{\pi}{2} \;\text{and}\; \tan(t) = x \)

  • They should be strict. Copying arcsine's statement across is the natural slip, because arcsine really does use the same two numbers as its range endpoints.
  • With the closed version, t = π/2 becomes a legal answer, and the second half of the statement then asks for tan(π/2), which does not exist. The statement contradicts itself.
  • A student who believes the closed version will happily write arctan of a huge number as π/2. It is close, but it is never equal, and on an exact-value question it is wrong.
  • Strict on both sides. Arctangent maps all of the real line onto the open interval, getting arbitrarily close to each end and reaching neither.

Read the statement before revealing and try to say precisely what is wrong with it, because the error is a single character repeated twice and it is easy to slide past.

The inequality signs should be strict. The slip is a natural one, since the arcsine's statement uses the very same two numbers as its endpoints and does include them. Copying that across is almost reflexive.

The second annotation gives the argument that settles it for good. If the quarter turn were an allowed answer, the statement would go on to demand the tangent of a quarter turn, which does not exist. A definition that asks for something impossible is not a definition, so the closed version cannot be right. In practice the symptom is a student writing a quarter turn as the exact value of the arctangent of some very large number, which is close but never correct.

46. Compositions as algebraic expressions

Section

Part 5

47. Name the inverse, and it becomes a triangle problem

Concept

Every remaining problem in this lesson has the same opening move. Whatever the inverse function is, give its value a letter. The letter comes with two pieces of information attached, and between them they decide everything.

Figure (svg): Two right triangles sharing an angle t at the origin. On the left the horizontal leg points right and is labelled x; on the right it points left and is labelled the absolute value of x, with a dashed ray showing that t is then obtuse.

The hypotenuse is 1 because the cosine of t is x over 1. Only the horizontal leg changes sign with x; the vertical leg is a length, so it stays positive.
  1. The equation: the original function of the letter equals the input.
  2. The interval: the letter lies in that inverse's range, which fixes the signs.

\[ t = \arccos(x) \;\Longrightarrow\; \cos(t) = x \;\text{and}\; 0 \le t \le \pi \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 822 — the substitution the book uses throughout Example 10.6.1 part 2

Everything left in this lesson uses one opening move, and this slide is that move. Do not skip past it; if you can perform it fluently, the remaining worked examples are mostly bookkeeping.

Naming the inner value converts a strange-looking composition into a familiar situation: an angle, something known about it, and something wanted. The two facts that come free with the name are the equation and the interval, and they have different jobs. The equation gives you a number to substitute into an identity. The interval decides every sign you will have to choose.

The two triangles in the picture are the same triangle in two positions. When the input is positive the angle is acute and the horizontal leg points right; when it is negative the angle is obtuse and the horizontal leg points left. In both positions the vertical leg points up, which is the drawn version of the statement that the sine of the angle is not negative. That single observation is what picks the sign of the square root two slides from now.

48. Example 10.6.1(h): the sine of an arccosine

Worked example

Now the outer function is not the inverse of the inner one, so nothing cancels.

\[ \sin\left(\arccos\left(-\tfrac35\right)\right) \]

Name the inner value and write what it knows

Why: Set the letter equal to the arccosine and record both facts.

\[ t = \arccos\left(-\tfrac35\right) \;\Longrightarrow\; \cos(t) = -\tfrac35, \quad 0 \le t \le \pi \]

Narrow the interval using the sign

Why: The cosine is negative, which rules out the first half of the range.

\[ \cos(t) < 0 \;\Longrightarrow\; \tfrac{\pi}{2} < t < \pi \]

Figure (svg): A unit circle with a ray into the second quadrant ending at the point minus three fifths, four fifths, a right triangle dropped from that point to the horizontal axis, and the angle t marked at the origin as obtuse.

A negative cosine and a range of zero to pi force the point into the second quadrant, where the second coordinate, the sine, is positive.

Apply the Pythagorean Identity

Why: Substitute the known cosine and solve for the sine.

\[ \left(-\tfrac35\right)^2 + \sin^2(t) = 1 \;\Longrightarrow\; \sin^2(t) = 1 - \tfrac{9}{25} = \tfrac{16}{25} \]

Take the root and choose the sign from the interval, not from habit

Why: In the second quadrant the sine is positive, so the positive root is the right one.

\[ \sin(t) = \pm\tfrac45 \;\Longrightarrow\; \sin(t) = \tfrac45 \]

Check numerically

Why: The arccosine of minus three fifths is about 2.2143 radians; take its sine and compare with four fifths.

\[ \arccos(-0.6) \approx 2.2143, \quad \sin(2.2143) \approx 0.8 = \tfrac45 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 822 — Example 10.6.1, part 1(h)

This is the first problem where the outer function is not the inverse of the inner one, so nothing cancels and the naming move has to do the work.

Follow the two facts. The equation gives a cosine, and the Pythagorean Identity converts that into a sine up to a sign. The interval, narrowed by the sign of the cosine, then chooses between the two signs. Both facts were needed, and neither on its own would have finished the problem.

The picture shows the narrowing happening. A negative cosine and a range running from zero to pi leave only the upper left of the circle, and every point there has a positive second coordinate. So the sine is positive, and the answer is four fifths rather than minus four fifths. Get into the habit of narrowing the interval before you take any square root, not after.

49. Example 10.6.2(d): the sine of an arctangent

Worked example

The same routine, now starting from a tangent, and with a negative answer.

\[ \sin\left(\arctan\left(-\tfrac34\right)\right) \]

Name the inner value and write what it knows

Why: Arctangent's range is the open interval between minus pi over two and pi over two.

\[ t = \arctan\left(-\tfrac34\right) \;\Longrightarrow\; \tan(t) = -\tfrac34, \quad -\tfrac{\pi}{2} < t < \tfrac{\pi}{2} \]

Narrow the interval using the sign

Why: A negative tangent on that interval forces t below zero.

\[ \tan(t) < 0 \;\Longrightarrow\; -\tfrac{\pi}{2} < t < 0 \]

Pick an identity that connects tangent to sine

Why: The book uses the identity relating cotangent and cosecant, which are the reciprocals of the two functions in play.

\[ 1 + \cot^2(t) = \csc^2(t), \qquad \cot(t) = \frac{1}{\tan(t)} = -\tfrac43 \]

Substitute and solve

Why: Square the cotangent, add one, and take the root.

\[ 1 + \left(-\tfrac43\right)^2 = \csc^2(t) = \tfrac{25}{9} \;\Longrightarrow\; \csc(t) = \pm\tfrac53 \]

Choose the sign from the interval

Why: Below zero and above minus pi over two the sine is negative, so the cosecant is negative too, and the sine is its reciprocal.

\[ \csc(t) = -\tfrac53 \;\Longrightarrow\; \sin(t) = -\tfrac35 \]

Check numerically

Why: The arctangent of minus three quarters is about minus 0.6435 radians, and its sine is minus 0.6.

\[ \arctan(-0.75) \approx -0.6435, \quad \sin(-0.6435) \approx -0.6 = -\tfrac35 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 825 — Example 10.6.2, part 1(d)

Same routine, new starting function, and this time the answer is negative, so the sign choice is not decoration.

The identity chosen here may look like an odd pick. The book uses the one relating cotangent and cosecant because those are the reciprocals of the two functions actually in play, the tangent you know and the sine you want, and because it is valid everywhere on the arctangent's range. Choosing an identity that is valid on the whole surviving interval is a genuine step, not a formality.

The narrowing works exactly as before. A negative tangent, on an interval that straddles zero, forces the angle below zero, and below zero the sine is negative. The final check at minus 0.6435 radians is the kind of numerical confirmation worth doing whenever an exact answer has a sign you had to argue for.

50. Before reaching for any identity

Step zero

\[ \tan(\arccos(x)) \]

Discussion prompt

You are about to set t equal to the arccosine of x. Write down everything t knows before you use a single identity, and then say which values of t must be thrown out and why.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 822 — the opening of Example 10.6.1, part 2(a)

Write your answer before revealing, and take the instruction literally: everything the letter knows, before a single identity is used.

Two facts come from the substitution itself, and they should be automatic by now. The third piece of information is the one people miss, and it comes not from the inverse but from the function on the outside. The tangent does not exist at a quarter turn, so that value of the letter has to be removed before any identity is applied.

Removing it has a consequence that pays off immediately. With the quarter turn gone, the letter lies in the first or the second quadrant, and in both of those the sine is not negative. That is the fact that will choose the sign of the square root later, and having it in hand before you start is the difference between a clean solution and a coin flip.

51. Example 10.6.1(2a): the tangent of an arccosine

Worked example

Rewrite as an algebraic expression in x, and state where the equivalence is valid.

\[ \tan(\arccos(x)) \]

Substitute and record the interval

Why: The previous slide did this work: t is in the first or second quadrant, and pi over two is excluded.

\[ t = \arccos(x): \quad \cos(t) = x, \quad 0 \le t < \tfrac{\pi}{2} \;\text{ or }\; \tfrac{\pi}{2} < t \le \pi \]

Choose the quotient identity

Why: Tangent is sine over cosine, and the cosine is already known.

\[ \tan(t) = \frac{\sin(t)}{\cos(t)} \]

Get the sine from the Pythagorean Identity

Why: Substitute the known cosine and solve.

\[ x^2 + \sin^2(t) = 1 \;\Longrightarrow\; \sin(t) = \pm\sqrt{1 - x^2} \]

Choose the sign from the interval

Why: In the first and second quadrants the sine is never negative, so the positive root is the right one.

\[ \sin(t) = \sqrt{1 - x^2} \]

Assemble the expression

Why: Divide the sine by the cosine.

\[ \tan(\arccos(x)) = \frac{\sqrt{1 - x^2}}{x} \]

State where it is valid

Why: Two exclusions: the input must be a legal input to arccosine, and the value that produced the forbidden pi over two must go.

\[ x \in [-1, 0) \cup (0, 1], \qquad \text{since } \cos\left(\tfrac{\pi}{2}\right) = 0 \]

Check at two sample values

Why: Take a positive x and a negative x and compare both sides numerically.

\[ x = 0.6: \quad \tan(\arccos(0.6)) \approx 1.3333 = \tfrac{0.8}{0.6} \]

\[ x = -0.8: \quad \tan(\arccos(-0.8)) \approx -0.75 = \tfrac{0.6}{-0.8} \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 822 — Example 10.6.1, part 2(a)

This is the model solution for the whole of the last part of the lesson, so read it once for the method and once for the bookkeeping.

The method is four moves: name, narrow, identity, sign. The identity here is the quotient one, chosen because the cosine is already known and the tangent is the sine over the cosine, so only the sine is missing. The Pythagorean Identity supplies it up to a sign, and the interval settles the sign.

The bookkeeping is the last two steps and it is where marks are usually lost. There are two separate exclusions, arriving from two separate places. The domain of the arccosine keeps the input between minus one and one; the tangent's failure at a quarter turn removes the input zero, because zero is the cosine of a quarter turn. Write each exclusion down as soon as you meet it, and the final interval assembles itself.

52. Why the interval of validity is not optional

Intuition

The finished expression usually makes sense for numbers the original composition never accepted, and that gap is where marks disappear.

\[ \frac{\sqrt{1-x^2}}{x} \text{ needs } |x| \le 1; \qquad 1 - 2x^2 \text{ needs nothing at all} \]

The second of those is the answer to the next worked example. Read as an ordinary polynomial it is defined for every real number, but it is only equal to the composition it came from where that composition exists.

The book's own comparison is the cleanest way to hold this. The expression x is defined for every real number, yet the statement that the square of the square root of x equals x is only true for x that are not negative. An equivalence is only as wide as the narrower of the two sides.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 823 — the remarks following Example 10.6.1

This slide exists because the interval feels like an afterthought and is not. The finished expression almost always makes sense for inputs the original composition never accepted.

Look at the two expressions on the slide. One of them wears its restriction on its sleeve: a square root of one minus a square complains loudly as soon as the input exceeds one in size. The other is an innocent polynomial, defined for every real number you could name, and it gives no warning at all. That is the dangerous case, and it is the one in the next worked example.

The book's comparison is the cleanest way to hold this, so borrow it. The expression x is defined for every real number, and yet the statement that squaring the square root of x returns x is only true when x is not negative. Two sides of an equation are only equal where both sides exist, and the answer to the question is the narrower of the two.

53. Example 10.6.1(2b): the cosine of twice an arcsine

Worked example

A double angle inside, so a double angle identity is the tool.

\[ \cos(2\arcsin(x)) \]

Substitute and record the interval

Why: Set the letter equal to the arcsine and write both facts.

\[ t = \arcsin(x): \quad \sin(t) = x, \quad -\tfrac{\pi}{2} \le t \le \tfrac{\pi}{2} \]

Ask whether anything must be excluded

Why: The cosine of twice t is defined for every t, so unlike the previous example nothing is thrown out here.

\[ \cos(2t) \text{ is defined for all } t \;\Longrightarrow\; \text{no extra exclusions} \]

Choose the right form of the double angle identity

Why: Three forms are available; take the one written entirely in terms of the sine, since the sine is what is known.

\[ \cos(2t) = \cos^2(t) - \sin^2(t) = 2\cos^2(t) - 1 = 1 - 2\sin^2(t) \]

Substitute

Why: Replace the sine of t by x.

\[ \cos(2\arcsin(x)) = 1 - 2x^2 \]

State where it is valid

Why: The only restriction comes from the substitution itself: arcsine only accepts inputs between minus one and one.

\[ x \in [-1, 1] \]

Check at a sample value and at an illegal one

Why: At three fifths both sides agree; at x equal to two the right-hand side gives a number while the left-hand side does not exist, which is exactly why the interval has to be stated.

\[ x = 0.6: \quad \cos(2\arcsin(0.6)) \approx 0.28 = 1 - 2(0.36) \]

\[ x = 2: \quad \arcsin(2) \text{ does not exist} \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 822-823 — Example 10.6.1, part 2(b), and the remarks

A double angle on the inside means a double angle identity, and the only real decision is which of its three forms to use.

Choose the form written entirely in terms of the function you know. Here the substitution gave a sine, so the form with only a sine in it turns the problem into a single substitution with no further work. Choosing one of the other two forms is not wrong, but it costs an extra conversion.

The second step is the one to imitate. Before doing any algebra, ask whether the outer function can fail anywhere. Here it cannot, because the cosine is defined for every number, so nothing extra is excluded and the interval comes entirely from the inner arcsine. Compare that with the previous example, where the outer tangent did fail somewhere and cost an extra exclusion. Asking the question is the habit; the answer varies.

54. Example 10.6.2(2a): the tangent of twice an arctangent

Worked example

Here the exclusions come from the outer function, and finding them is most of the work.

\[ \tan(2\arctan(x)) \]

Substitute and record the interval

Why: Arctangent's range is the open interval between minus pi over two and pi over two.

\[ t = \arctan(x): \quad \tan(t) = x, \quad -\tfrac{\pi}{2} < t < \tfrac{\pi}{2} \]

Find where the outer tangent fails

Why: The tangent is undefined when its argument is pi over two plus a whole number of pi, and the argument here is twice t.

\[ 2t = \tfrac{\pi}{2} + \pi k \;\Longrightarrow\; t = \tfrac{\pi}{4} + \tfrac{\pi}{2}k, \quad k \in \mathbb{Z} \]

Keep only the members inside the range

Why: Of that whole family, only two lie strictly between minus pi over two and pi over two.

\[ t = \pm\tfrac{\pi}{4} \]

Apply the double angle identity for tangent

Why: It is valid at every remaining value of t, which is worth checking before using it.

\[ \tan(2t) = \frac{2\tan(t)}{1 - \tan^2(t)} = \frac{2x}{1 - x^2} \]

Translate the excluded angles into excluded inputs

Why: The substitution says x is the tangent of t, so apply the tangent to the two forbidden values.

\[ x = \tan\left(\pm\tfrac{\pi}{4}\right) = \pm 1 \]

State where it is valid

Why: Every real number except one and minus one.

\[ x \in (-\infty, -1) \cup (-1, 1) \cup (1, \infty) \]

Check at a sample value and at an excluded one

Why: At one half both sides agree; at x equal to one the formula divides by zero, and indeed twice the arctangent of one is pi over two, where the tangent does not exist.

\[ x = \tfrac12: \; \tan(2\arctan(0.5)) \approx 1.3333 = \frac{1}{0.75}; \qquad x = 1: \; 2\arctan(1) = \tfrac{\pi}{2} \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 825-826 — Example 10.6.2, part 2(a)

In this one the exclusions are most of the work, so plan for that and do not be surprised when the identity turns out to be the quick part.

The tangent fails at a quarter turn and at every half turn after that, and here its argument is twice the letter, so the failures happen at half those places: the eighth turn, and every quarter turn after that. That is an infinite family, but the letter is confined to the arctangent's range, and only two members of the family fit inside it.

The final translation step is the one that is easy to forget. The exclusions were found as values of the letter, which is an angle, but the answer has to be stated in terms of the input. The substitution says the input is the tangent of the letter, so apply the tangent to each excluded angle. The two forbidden angles give inputs of one and minus one, and those are what the answer must exclude.

55. Example 10.6.2(2b): the cosine of an arccotangent

Worked example

The last of the book's four, and the one where the triangle picture pays off most.

\[ \cos(\operatorname{arccot}(2x)) \]

Substitute and record the interval

Why: Arccotangent's range is the open interval from zero to pi.

\[ t = \operatorname{arccot}(2x): \quad \cot(t) = 2x, \quad 0 < t < \pi \]

Ask whether anything must be excluded

Why: The cosine is defined everywhere, so there are no extra exclusions and the interval stays whole.

\[ \cos(t) \text{ is defined for all } t \;\Longrightarrow\; \text{no extra exclusions} \]

Figure (svg): A right triangle with the horizontal leg labelled two x, the vertical leg labelled one, the hypotenuse labelled the square root of four x squared plus one, and the angle t at the left-hand vertex.

The cotangent of t is the adjacent leg over the opposite leg, so putting 2x against 1 reproduces exactly the equation t satisfies.

Choose a route from cotangent to cosine

Why: Cotangent is cosine over sine, so getting the sine is enough.

\[ \cot(t) = \frac{\cos(t)}{\sin(t)} \;\Longrightarrow\; \cos(t) = \cot(t)\sin(t) \]

Get the sine from the Pythagorean Identity

Why: The identity relating cotangent and cosecant holds throughout the interval from zero to pi.

\[ 1 + \cot^2(t) = \csc^2(t) \;\Longrightarrow\; 1 + 4x^2 = \csc^2(t) \]

Choose the sign from the interval

Why: Between zero and pi the sine is positive, so the cosecant is positive and the sine is the reciprocal of the positive root.

\[ \csc(t) = \sqrt{4x^2+1} \;\Longrightarrow\; \sin(t) = \frac{1}{\sqrt{4x^2+1}} \]

Assemble the expression and state where it is valid

Why: Multiply the cotangent by the sine; arccotangent accepts every real number and nothing was excluded.

\[ \cos(\operatorname{arccot}(2x)) = \frac{2x}{\sqrt{4x^2+1}}, \qquad x \in (-\infty, \infty) \]

Check at a positive and a negative value

Why: At one half the angle is pi over four; at minus one and a half the formula must come out negative, and it does.

\[ x = \tfrac12: \quad \cos(\operatorname{arccot}(1)) \approx 0.7071 = \tfrac{1}{\sqrt2} \]

\[ x = -\tfrac32: \quad \cos(\operatorname{arccot}(-3)) \approx -0.9487 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 826 — Example 10.6.2, part 2(b)

The last of the book's four compositions, and the one where the triangle picture earns its place, because the route from a cotangent to a cosine is not a one-step identity.

The strategy is stated before any algebra: the cotangent is the cosine over the sine, so if you can get hold of the sine you can multiply and be done. The Pythagorean Identity connecting cotangent and cosecant is the tool, and it is valid throughout the interval from zero to pi, which is worth confirming before using it.

The sign choice is easier here than in the earlier examples, and for a reason worth noticing. The arccotangent's range is the whole open interval from zero to pi, which lies entirely in the upper half of the circle, so the sine is positive no matter what the input is. No narrowing was needed at all. The picture is drawn for a positive input, but the algebra never assumed that, which is why the final answer is valid for every real number.

56. One from the exercises: the sine of an arctangent

Worked example

This one is worth doing because its answer has no exclusions at all, and because it turns up constantly later.

\[ \sin(\arctan(x)) \]

Substitute and record the interval

Why: The usual opening move.

\[ t = \arctan(x): \quad \tan(t) = x, \quad -\tfrac{\pi}{2} < t < \tfrac{\pi}{2} \]

Relate the tangent to the secant

Why: The identity connecting them holds on the whole open interval.

\[ 1 + \tan^2(t) = \sec^2(t) \;\Longrightarrow\; \sec^2(t) = 1 + x^2 \]

Choose the sign from the interval

Why: Strictly between minus pi over two and pi over two the cosine is positive, so the secant is positive.

\[ \sec(t) = \sqrt{1+x^2} \;\Longrightarrow\; \cos(t) = \frac{1}{\sqrt{1+x^2}} \]

Recover the sine

Why: Sine is tangent times cosine.

\[ \sin(t) = \tan(t)\cos(t) = \frac{x}{\sqrt{1+x^2}} \]

State where it is valid

Why: Arctangent accepts everything and the denominator is never zero, so the equivalence holds everywhere.

\[ \sin(\arctan(x)) = \frac{x}{\sqrt{1+x^2}}, \qquad x \in (-\infty, \infty) \]

Check at a sample value and at the extremes

Why: At two the two sides agree, and as x grows the expression tends to one, which matches the arctangent approaching pi over two, whose sine is one.

\[ x = 2: \; \sin(\arctan(2)) \approx 0.8944 = \tfrac{2}{\sqrt5}; \qquad \frac{x}{\sqrt{1+x^2}} \to 1 \]

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 841 — the pattern of Exercises 107 to 118, with the homework's own seed

This one is not in the book's worked examples but its pattern fills the exercise set, and the answer turns up so often in later courses that it is worth committing to memory.

The identity connecting tangent and secant is the natural choice, because the secant is the reciprocal of the cosine and the sine is the tangent times the cosine. Two steps and you are there. As always, the interval picks the sign: on the arctangent's range the cosine is positive, so the secant is the positive root.

The reason there are no exclusions is worth spelling out. The arctangent accepts every real number, the outer sine is defined everywhere, and the denominator in the answer is a square root of something at least one, so it can never be zero. All three of the usual sources of an exclusion come up empty, which is unusual and makes this a good answer to have at hand. The final check is a limiting one: as the input grows the expression tends to one, which matches the arctangent climbing toward a quarter turn, whose sine is one.

57. Trap: handing in the expression with no interval

Trap

The trap

A complete-looking answer to the second worked example of the set:

\[ \cos(2\arcsin(x)) = 1 - 2x^2 \]

Half the answer. The question asked for the expression and the interval.

\[ x = 2: \; 1 - 2(4) = -7 \]

But the left-hand side does not exist at two.

The fix

The polynomial on the right is defined everywhere. The composition on the left is only defined where the inner inverse is, so the two are equal only there.

\[ \cos(2\arcsin(x)) = 1 - 2x^2 \]

\[ \text{valid for } x \in [-1, 1] \]

Get into the habit of writing the interval down at the moment you make the substitution, not at the end. It comes from two places: the domain of the inner inverse, and any value the outer function cannot handle.

The wrong answer here is not wrong, it is incomplete, and incomplete answers are harder to spot in your own work than wrong ones.

The demonstration is the fastest possible. Put the input two into the polynomial and it happily returns minus seven. Put two into the composition and there is nothing there at all, because two has no arcsine. A statement that one thing equals another cannot be true where only one of them exists.

The practical advice is in the last line. Write the interval at the moment you make the substitution, while the reason for it is in front of you, rather than trying to reconstruct it at the end. It comes from exactly two places: the domain of the inner inverse, and any value the outer function cannot handle. Check both, every time.

58. Order the steps of a composition problem

Ranking

Put in order

Put the steps of a rewrite-as-an-algebraic-expression problem into the order that works.

  1. Name the inner value: let t be the inverse function applied to x
  2. Write down the equation t satisfies and the interval t lies in
  3. Throw out any t the outer function cannot handle
  4. Narrow the interval using the sign of the given value
  5. Use an identity that is valid on the surviving interval
  6. Choose the sign of the root from the interval, not from habit
  7. Translate the thrown-out values of t back into values of x

Why: Naming the inner value comes first because nothing can be said until there is something to talk about. The interval arrives with the name and is what makes every later choice. Exclusions are found before any identity is used, because which identities are legal depends on which values of the letter survive. The sign of a root is settled by the interval, never by habit. And the exclusions are translated back into the variable last, because only then do you know what to translate.

Put these in order before checking, and for each step ask what information the previous one supplied that this one needs.

Two orderings are genuinely tempting and both are wrong. Using an identity before finding the exclusions is the first: which identities are legal depends on which values of the letter survive, so the exclusions come first. Translating the exclusions back into the variable early is the second: until the identity is done you do not know what the final expression looks like, and it is cleaner to keep the exclusions as angles until the end.

The step most often missed altogether is the one about choosing the sign of the root from the interval. Habit says take the positive root. Habit is right roughly half the time, and the half where it is wrong is exactly the half where the input is negative.

59. Putting it together

Section

Part 6

60. Pattern: anything with an arc in it

Pattern

Figure (svg): A flow diagram with three branches: a trig function outside an inverse cancels whenever the input is in the domain; an inverse outside a trig function cancels only when the inner angle is in the range, otherwise you work inside out; and mismatched pairs are handled by naming the inner value t.

Three shapes, three routines. Only the middle one carries a condition, and it is the one that costs marks.
  1. An inverse on its own. Find the one number in that inverse's range with the right value. Check the input is in the domain first.
  2. Trig outside, inverse inside. It cancels, provided the input is legal for the inverse.
  3. Inverse outside, trig inside. It cancels only if the inner angle already lies in the inverse's range. Otherwise evaluate the inside and start again.
  4. Anything mismatched. Name the inner value, write its equation and its interval, narrow the interval by the sign, then use one identity.
  5. Always state the interval of validity when the answer is an expression in x rather than a number.

This is the routine for every expression in the section, arranged by shape. Identify the shape first, because the shape decides the work.

The first three lines are the evaluation cases, and only the third has a condition. Learn the list in that order, cheapest first: a bare inverse is a hunt in one interval, a trig function around an inverse cancels, and an inverse around a trig function cancels only if you check the angle.

The fourth line is the general machine, and it also handles the first three if you ever forget them. Naming the inner value and writing down its equation and its interval never fails; it is only slower. The last line is the one to write on your hand: if the answer is an expression rather than a number, it is not finished until the interval is stated.

61. Check: choosing from the range

Check

Check your understanding

Which number is the arcsine of negative one half?

  • A. −π/6 (correct)
  • B. 7π/6
  • C. 11π/6
  • D. −5π/6

Answer: A

Why: Every one of these four angles has a sine of minus one half, so the sine value cannot decide between them. The range does: arcsine returns only numbers between minus pi over two and pi over two, and negative pi over six is the only one of the four inside that interval.

Why B tempts people
Its sine is right, but seven pi over six is about 3.67, far above pi over two, so it is outside the range of arcsine.
Why C tempts people
Also has the right sine, but eleven pi over six is about 5.76, again outside the range; this is the angle a unit-circle sweep would land on last.
Why D tempts people
Its sine is minus one half and it is negative, which makes it tempting, but minus five pi over six is about −2.62, below minus pi over two.

All four options have the right sine, so testing the sine value will not separate them. The question is really about which interval the arcsine is allowed to answer from.

Two of the wrong answers are large positive angles, which are the ones a sweep round the unit circle produces. The third is negative, which makes it feel like an arcsine answer, but it is beyond minus a quarter turn and so still outside the range. If you picked it, the fix is to check the size of the angle against the quarter turn, not just its sign.

62. Check: when cancellation is allowed

Check

Check your understanding

Which of these compositions cannot be simplified by cancelling straight away?

  • A. cos(arccos(0.3))
  • B. arccos(cos(2))
  • C. arccos(cos(4)) (correct)
  • D. tan(arctan(9))

Answer: C

Why: For the inverse-outside form the inner angle must lie in the inverse's range. Four is larger than pi, which is about 3.14, so it is outside the interval from zero to pi and the rule does not apply; you must evaluate the cosine first and then take the arccosine, which returns two pi minus four.

Why A tempts people
The trig function is outside and the input 0.3 is between minus one and one, so this cancels with no further thought.
Why B tempts people
Two is less than pi, so the inner angle is already inside the range of arccosine and the rule applies directly.
Why D tempts people
Arctangent accepts every real number, so the trig-outside form here has no condition at all to check.

Two of the options have the trig function on the outside, and neither of those can be the answer: the inverse-inside direction never needs more than the input being legal, and in both cases it is.

That leaves the two with the inverse on the outside, and they are separated by one comparison. Two is smaller than pi, so that inner angle is already inside the arccosine's range and the rule applies. Four is larger than pi, so it is not, and the composition has to be computed from the inside, where it returns two pi minus four rather than four.

63. Check: the interval of validity

Check

Check your understanding

When the tangent of an arccosine is rewritten as an algebraic expression in x, which inputs have to be excluded?

  • A. nothing: the expression works for every x
  • B. only x = 0
  • C. x = 0, and everything outside −1 to 1 (correct)
  • D. only x = 1 and x = −1

Answer: C

Why: Two exclusions arrive from two different places. The domain of arccosine keeps x between minus one and one, and the tangent is undefined at pi over two, whose cosine is zero, so x equal to zero must also go. The valid set is everything from minus one up to but not including zero, together with everything above zero up to one.

Why A tempts people
The expression with a square root in it does happen to be defined outside the interval in some cases, but the composition it came from is not, so the equivalence fails there.
Why B tempts people
Zero does have to go, but forgetting the domain of the inner arccosine leaves the answer claiming values of x that never had an arccosine at all.
Why D tempts people
Plus and minus one are perfectly fine here: their arccosines are zero and pi, and the tangent exists at both. Those two are excluded in the arctangent problem, not this one.

There are two exclusions here, from two different causes, and every wrong option misses one of them.

The first cause is the inner inverse: the arccosine only accepts inputs between minus one and one, so everything outside that never had a composition to begin with. The second cause is the outer tangent, which fails at a quarter turn, and the input whose arccosine is a quarter turn is zero.

The option naming plus and minus one is worth a second look, because those are excluded in the twice-an-arctangent problem and it is easy to carry the memory across. Here they are perfectly fine: their arccosines are zero and pi, and the tangent exists at both.

64. Say the asymmetry in your own words

Explain it to yourself

Discussion prompt

In two or three sentences, and without using the words range or domain, explain to someone who has not met this lesson why taking the cosine of an arccosine always gives you back your number, while taking the arccosine of a cosine often does not.

Write your sentences before revealing, and take the constraint seriously: without the technical words you have to describe the mechanism rather than quote the rule, and that is the point of the exercise.

The mechanism is about information. One order never destroys any, because the inverse hands over an angle from a narrow band and the trig function simply reports its value. The other order destroys information at the first step, because the trig function gives the same value to many angles and keeps no record of which one it was given.

If you can say that unprompted, you will never again need to remember which of the two cancellation rules carries a condition. You will be able to work it out in the five seconds it takes to trace a number through.

65. Why programs carry a second arctangent

Real world

Figure (svg): A coordinate plane with one arrow to the point minus three, four in the second quadrant and another to the point three, minus four in the fourth quadrant, both lying on the same dashed straight line through the origin.

Both points give four over minus three when you divide. One arctangent has one output, so it cannot possibly distinguish them.

Discussion prompt

A robot's position relative to its base is given as a pair of numbers, and the heading is wanted as an angle. A programmer computes it as the arctangent of the second number divided by the first. Explain, using the picture, exactly which positions this handles wrongly, and why nearly every programming language also supplies a two-argument arctangent.

OpenStax Algebra and Trigonometry 2e, §8.3 Inverse Trigonometric Functions §8.3 — the same inverse functions in OpenStax

Answer before revealing. This is the most common real-world consequence of the restricted range, and it is responsible for a great many bugs in graphics, navigation and robotics code.

The picture shows the failure exactly. Two positions on opposite sides of the origin lie on the same straight line through it, so dividing the second coordinate by the first gives the same number for both. The division has thrown away which side you were on, and the arctangent, which can only answer from the right-hand half of the plane, has no way to recover it.

The two-argument version is handed both numbers separately, so the signs survive and it can return an angle anywhere in a full turn. The size of the error from using the one-argument version is not small or subtle: it is exactly half a turn, for every position in the left-hand half of the plane. The ladder problem in the homework is the opposite case, where the angle really is confined to the first quadrant and the ordinary inverse is perfectly safe.

66. Exit ticket

Exit ticket

\[ \text{(i) } \sin\left(\arccos\left(-\tfrac45\right)\right) \qquad \text{(ii) } \arcsin\left(\sin\left(\tfrac{5\pi}{6}\right)\right) \]

Discussion prompt

Evaluate both exactly. For the second one, say first whether the cancellation rule applies and why, before you write anything down.

Do both before revealing, and for the second one force yourself to write the verdict on the cancellation rule before you write any working.

The first is the naming routine, with a three-four-five triangle behind it. A negative cosine puts the angle in the upper left of the circle, where the sine is positive, so the square root is taken positive and the answer is three fifths. If you wrote minus three fifths, you took the root without narrowing the interval first.

The second is the cancellation trap, dressed in a different angle. Five pi over six is about 2.6, comfortably past the quarter turn, so no cancellation is allowed and the answer is not the angle you started with. Work inside out, get one half, and the arcsine of one half is pi over six.

67. Recap

Recap

inversedomainrangethe cancellation with a condition
arccos x−1 to 1, ends included0 to π, ends includedarccos(cos x) = x only for x from 0 to π
arcsin x−1 to 1, ends included−π/2 to π/2, ends includedarcsin(sin x) = x only for x from −π/2 to π/2
arctan xevery real number−π/2 to π/2, ends excludedarctan(tan x) = x only for x strictly inside
arccot xevery real number0 to π, ends excludedarccot(cot x) = x only for x strictly inside

One idea runs through all of it. The restriction is not a footnote; it is the definition. It sets the range, it decides which of infinitely many angles is the answer, it is the condition on the inside-out cancellation rule, and it is what fixes the sign of every square root in a composition.

Next, Lesson 10.6b inverts the last two functions, secant and cosecant, where no single unbroken interval will do the job and two rival conventions are in use.

Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 819-826 — Theorems 10.26 and 10.27, Examples 10.6.1 and 10.6.2

Four functions, and the table is the whole factual content of the lesson. Two of them answer from zero to pi and two from minus a quarter turn to a quarter turn; two accept only inputs between minus one and one and two accept everything; two include their endpoints and two do not.

Underneath the table is one idea, and it is worth stating in full because it explains everything the table records. The restriction is not a technicality that had to be dealt with before the real material could begin. It is the definition. It fixes the range, it decides which of infinitely many candidate angles is the answer, it is the condition attached to the inside-out cancellation rule, and it is what chooses the sign of every square root in a composition problem.

One thing this lesson did not have to face is a function for which no single unbroken interval will do. The secant and the cosecant are exactly that, and the next lesson has to make a harder choice, with two rival conventions still in circulation as a result.

Sources

  1. Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions — Stitz & Zeager, College Trigonometry, Version 3 Corrected Edition, July 4 2013, pp. 819-827
  2. OpenStax Algebra and Trigonometry 2e, §8.3 Inverse Trigonometric Functions

Want this taught 1-on-1? Alexander tutors Trigonometry — $55/session, free consultation.

Book on Wyzant · Text (657) 465-8108