The two circular functions whose inverses are genuinely not standardised. Covers why the secant is awkward to invert at all — its range has a gap, so the inverse's domain does too — and then presents both competing restrictions in full: the trigonometry-friendly one that keeps the arccosine's interval, and the calculus-friendly one chosen to make the tangent of an arcsecant a single formula rather than a piecewise one. The same worked examples are answered under each, so the differences are visible rather than described.
Subject: Trigonometry · 68 slides · symbolic lesson
Open the interactive version of this deck
Title
Trigonometry · Section 10.6.1 and 10.6.2
The two inverse functions the textbooks cannot agree on
Objectives
Arcsine, arccosine, arctangent and arccotangent each have one restriction that everybody uses. Arcsecant and arccosecant do not. Two restrictions are in circulation, both defensible, and they give different answers to the same question.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 826-833 — subsections 10.6.1 and 10.6.2
Everything you did in the last lesson had one right answer. Arccosine of a half is pi over three, full stop, in every book ever printed. This lesson is the exception. Arcsecant and arccosecant are the two inverse circular functions the world never standardised, and a problem with a negative input genuinely has two defensible answers.
That sounds like a mess, and it is a small one, but it is a mess with a clean structure. The two conventions agree completely for inputs of one and above, they agree at the single input minus one, and they disagree everywhere else on the negative side. Once you know that, you know exactly when you have to ask which book you are in.
So the plan is: first see why secant is awkward to invert at all, then meet both conventions in full, evaluate the same four expressions under each and watch which answers move, and finish with the calculator work. You will finish able to read a formula for arcsecant and tell, from the formula alone, which convention its author had in mind.
Warm-up
Discussion prompt
Cosine is not one-to-one, so it has no inverse until its domain is restricted. From memory: which interval was cosine restricted to, why that one, and what the resulting arccosine's domain and range are?
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 819 — the restriction of cosine
Write your answer down before you reveal anything. The three things to recover are the interval, the reason for it, and the resulting domain and range, and the reason is the part most people skip.
The interval from zero to pi was not chosen because it looks tidy. It was chosen because cosine does two things on it at once: it never repeats a value, so the inverse is well defined, and it produces every value from minus one to one, so the inverse has no gaps in its domain. Both properties are needed, and on that interval both come for free from a single unbroken piece of the graph.
Hold on to the second property. Secant is about to fail it. There is no single unbroken piece of the secant graph that produces every value secant is capable of producing, and that one failure is the source of everything unusual in this lesson.
Section
Part 1
Concept
Figure (svg): Two panels. Left: the graph of secant over minus two pi to two pi, its U-shaped branches separated by vertical asymptotes, with the horizontal strip between minus one and one shaded and empty. Right: the graph of cosecant over the same window, with the same empty strip.
Every other circular function you have inverted sweeps through an unbroken set of values. These two do not: their outputs jump straight from minus one to one with nothing in between.
\[ \text{range of } \sec = (-\infty, -1] \cup [1, \infty) \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 826 — the graphs before subsection 10.6.1
Look at the shaded strip on both panels. It runs from minus one up to one, and neither curve ever enters it. Secant is one over cosine, and cosine never exceeds one in size, so its reciprocal is never smaller than one in size. Cosecant behaves the same way for the same reason.
Sine, cosine, tangent and cotangent all have ranges with no holes in them: the first two fill an interval, the last two fill the whole number line. These two are the odd ones out, and the hole is exactly the thing that will force a choice.
One more thing to notice while you are looking. Between any two neighbouring vertical asymptotes the curve is one-to-one, so any one branch could be inverted on its own. The problem is never the one-to-one part. The problem is coverage.
Concept
Inverting a function swaps its domain and its range. So whatever hole sits in the range of secant reappears as a hole in the domain of arcsecant.
\[ \text{domain of } \operatorname{arcsec} = \text{range of } \sec = (-\infty, -1] \cup [1, \infty) \]
\[ \{x : |x| \ge 1\} \]
the domain condition — An arcsecant or arccosecant accepts a number only when its size is at least one. Anything strictly between minus one and one is not an input at all.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 827 — the remark after the two restrictions
Inverting a function is bookkeeping: whatever the function takes in, the inverse gives out, and whatever the function gives out, the inverse takes in. So the range of secant, with its hole, becomes the domain of arcsecant, with the same hole.
That is genuinely unlike anything you have inverted so far. The domain of arccosine is an interval. The domain of arctangent is everything. The domain of arcsecant is two rays with a hole between them, and you have to check membership before you evaluate, every single time.
The set notation in the second line is the compact way to say it, and calculus books use it constantly, so it is worth reading fluently: the set of numbers whose size is at least one. Reading it as a condition on size rather than on sign will save you from the single most common domain error in this section.
Notation
Annotate
On: \( \{x : |x| \ge 1\} = (-\infty, -1] \cup [1, \infty) \)
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 828 — Theorem 10.28, domain
Take the annotations one at a time. The absolute value is doing all the work in the first form, and it is a statement about size only. Minus one hundred and seventeen passes; nine tenths does not. If you find yourself thinking that negative inputs are somehow suspect, you have confused this condition with a sign condition.
The second annotation matters for how you write proofs and domains later. This is a union of two rays, not an interval, so a phrase like between the two endpoints makes no sense here, and an argument that works its way continuously from one side to the other is not available.
The third annotation is worth memorising outright. Arcsecant of one is zero and arcsecant of minus one is pi, under both conventions, because the secant of zero is one and the secant of pi is minus one, and both of those angles survive every restriction anyone has proposed. Those two values are free marks on any test.
Picture it
Figure (svg): The secant graph from minus pi over two to five pi over two, with its three branches drawn in three colours: the first and third open upward and reach only the values at or above one, while the middle branch opens downward and reaches only the values at or below minus one.
Between two consecutive vertical asymptotes the secant graph is one-to-one, so any single branch could be inverted on its own. The trouble is what it reaches: an upward branch produces only values at or above one, and a downward branch only values at or below minus one.
So no single continuous piece of the graph is both one-to-one and onto the whole range. The restriction must take one piece from the top and one from the bottom, and there is more than one sensible way to pair them up.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 826 — why a piecewise approach is needed
This picture is the argument for everything that follows, so spend a moment on it. Three consecutive branches of secant are drawn in two colours. Each one, on its own, passes the horizontal line test, so each one has an inverse.
But look at what each reaches. The upward branches only ever produce values of one and above; the downward branch only ever produces values of minus one and below. So if you restricted secant to a single branch, the resulting inverse would accept only half the numbers it should accept.
That is why the restriction has to be assembled from two branches, one from the top and one from the bottom. And that is where the freedom enters: you can take the bottom half from the branch immediately to the left, or from the one immediately to the right, and both give a perfectly good one-to-one function onto the whole range. Two reasonable choices, no way to call one of them wrong.
Prediction
Predict first
One of these four numbers is not in the domain of arcsecant. Which one?
Correct: 0.75
Why: Three quarters has size less than one, so no angle has it as a secant, and the arcsecant of it is undefined. Size, not sign, is what the domain asks about: a very large negative number like minus one hundred seventeen is perfectly acceptable, and both endpoints are included.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 841-842 — Exercises 79 and 80
Commit to an answer before you reveal. The trap here is a sign trap: three of the four options look harmless and one of them looks harmless but is not, and the instinct that flags the large negative number is the instinct to unlearn.
Three quarters is the odd one out because its size is below one. Ask yourself what angle would have to exist: an angle whose secant is three quarters has cosine four thirds, and no angle has a cosine bigger than one. Nothing to return, so the expression is undefined.
Minus one hundred and seventeen, on the other hand, is a fine input. So is one, and so is minus one, because both endpoints are included. When you check a domain here, ask about size and ignore sign entirely.
Worked example
Decide whether each of these is a number, and say why.
\[ \sec(\operatorname{arcsec}(2)) \qquad \text{and} \qquad \sec(\operatorname{arcsec}(0.75)) \]
Read the cancellation rule with its condition
Why: The outside-in rule holds only when the input is a legal input.
\[ \sec(\operatorname{arcsec}(x)) = x \quad \text{provided } |x| \ge 1 \]
Test the first input against the condition
Why: Two has size two, which is at least one.
\[ |2| = 2 \ge 1 \;\Longrightarrow\; \sec(\operatorname{arcsec}(2)) = 2 \]
Test the second input
Why: Three quarters has size less than one, so the inner arcsecant is already undefined.
\[ |0.75| < 1 \;\Longrightarrow\; \operatorname{arcsec}(0.75) \text{ is undefined} \]
Check the failure is genuine, not a rounding problem
Why: If some angle had secant three quarters, its cosine would be four thirds, and no cosine exceeds one.
\[ \sec t = \tfrac34 \iff \cos t = \tfrac43 > 1, \text{ impossible} \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 842 — Exercises 77-81
The cancellation rule for going outside in looks unconditional, and people quote it that way, but the condition is there in Theorem 10.28 and it is the domain condition. The secant of an arcsecant of a number returns the number only when the arcsecant existed in the first place.
Notice how the work is structured. You do not try to evaluate and then discover a problem; you test the input against the condition first, and the test is one comparison. Building that habit now is what will keep the harder compositions later in the section clean.
The final step is worth doing even when you are sure. Turning the impossible claim into a statement about cosine, and seeing a cosine larger than one, is a concrete reason rather than an appeal to a rule. Reasons survive; rules get misremembered.
Tweak it
Parameter explorer
The curve is the secant of x minus the number a. An angle with secant a exists exactly when the curve reaches the horizontal axis. Slide a upward from zero and say when that first happens.
\[ y = \sec(x) - {a} \]
Drag the slider slowly from the bottom. The curve you are watching is the secant curve pulled down by the amount you set, so wherever it crosses the horizontal axis there is an angle whose secant is your target.
While your target sits below one, the two visible branches straddle the axis but never touch it: the left branch stays above and the right branch stays below, with the forbidden strip in between. There is no crossing, which is the picture of an undefined arcsecant.
The moment your target reaches one, the bottom of the left branch touches down, and from there upward every target is met exactly once on the piece shown. That is the whole domain condition seen as a crossing problem, and it is also a preview of why the restriction is built the way it is: you want each legal target hit once, no more and no less.
Counterexample
Discussion prompt
Arcsine and arccosine have no rival conventions at all. Name the property of sine and cosine that arcsecant lacks, and explain why that property removes the argument.
Write a real answer before you look. The question is not why arcsine is easier; it is what specific property removes the choice, and the answer is one word long before you expand it.
The property is coverage by a single piece. Cosine's range is one unbroken interval, and one unbroken piece of the cosine graph already runs across all of it, so the restriction is forced and there is nothing to argue about. Tangent is the same story on an infinite range.
Secant's range comes in two pieces, so the restriction needs two pieces, and the second one can be taken from more than one place. That is the entire origin of the dispute. If you can say that sentence cleanly, the rest of this lesson is bookkeeping.
Concept
The book is explicit that this is a genuine fork in the road, and it refuses to hide one side of it.
Stitz and Zeager present both in full, the trigonometry-friendly one first, and their exercise set states which one is in force whenever the answer depends on it. That habit is worth copying.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 826 — one approach simplifies the Trigonometry, the other the Calculus
Notice what the book does here, because it is unusual and it is honest. Rather than pick one convention and hide the other, Stitz and Zeager set out both in full, with their own theorem each, and then work the same example twice so you can see exactly where the answers part company.
The names are descriptive, not official. The first keeps the intervals cosine and sine already use, which makes the reciprocal identities hold everywhere and makes hand evaluation easy, so they call it trigonometry friendly. The second sacrifices that in order to remove a case split that shows up constantly in calculus, so they call it calculus friendly.
One practical habit to copy: the book's exercise set states which range is assumed whenever the answer depends on it, and it does so in both directions. When you write up an arcsecant problem with a negative input, put the same sentence at the top. It costs you one line and removes every ambiguity.
Section
Part 2
Concept
Figure (svg): Two panels. Left: secant kept only on zero to pi over two and pi over two to pi, two arcs with solid endpoints at the point zero comma one and pi comma minus one. Right: the reflected graph, arcsecant, rising from zero at x equals one toward the dashed line at pi over two, and coming down to pi at x equals minus one.
\[ f(x) = \sec(x) \text{ on } \left[0, \tfrac{\pi}{2}\right) \cup \left(\tfrac{\pi}{2}, \pi\right] \]
Take the interval from zero to pi, the one cosine kept, and punch out the single point where secant is undefined. What is left is two arcs: one climbing from one, one arriving at minus one. Together they hit every value in the range exactly once.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 827 — subsection 10.6.1
Read the left panel first. Start from the interval that cosine was restricted to, from zero to pi, and remove the single point in the middle where secant has no value. What survives is two arcs: one running upward from the height one, and one arriving from far below at the height minus one.
Check the coverage claim on the picture. The first arc produces every value of one and above, the second produces every value of minus one and below, and between them they produce everything in the range exactly once. That is what makes this a legal restriction.
Now the right panel. Reflecting across the diagonal swaps the two axes: the vertical asymptote at pi over two becomes a horizontal asymptote at the height pi over two, and the two solid endpoints become the endpoints of the inverse's graph. Notice that the outputs of this arcsecant run from zero to pi, exactly where arccosine's outputs run. That is not a coincidence, and the next few slides cash it in.
Concept
Figure (svg): Two panels. Left: cosecant kept only on minus pi over two to zero and zero to pi over two, with solid endpoints at minus pi over two comma minus one and pi over two comma one. Right: the reflected graph, arccosecant, running from minus pi over two at x equals minus one up toward the dashed horizontal axis and on to pi over two at x equals one.
\[ g(x) = \csc(x) \text{ on } \left[-\tfrac{\pi}{2}, 0\right) \cup \left(0, \tfrac{\pi}{2}\right] \]
Same move, one function along: keep sine's interval and punch out the point where cosecant blows up. Because the kept set is symmetric about the origin and cosecant is odd, the inverse comes out odd too.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 827 — subsection 10.6.1
Same construction, one function along. Sine was restricted to the interval from minus pi over two to pi over two, so cosecant is restricted there too, with zero punched out because cosecant blows up at zero.
The right panel has a feature the secant version did not: the graph is symmetric about the origin. That happens because the kept set of angles is itself symmetric about zero and cosecant is an odd function, so the inverse inherits the symmetry. The book lists this as an extra property at the bottom of Theorem 10.28, and it is the one property that the other convention destroys.
Watch the asymptote too. Large inputs of either sign push the output toward zero, from above on the right and from below on the left. Compare that with the arcsecant, whose asymptote sits at pi over two. Both functions flatten out for large inputs, because a large secant or cosecant means a nearly right angle or a nearly straight one, and the angle has almost nowhere left to move.
Notation
Annotate
On: \( \operatorname{arcsec}(x) = t \iff \sec(t) = x, \; t \in \left[0, \tfrac{\pi}{2}\right) \cup \left(\tfrac{\pi}{2}, \pi\right] \)
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 828 — Theorem 10.28
The double arrow is the important symbol on this slide. It says the statement runs both ways, which means arcsecant of a number equals a given angle only when two conditions hold together: the secant is right, and the angle is in the range. People check the first and forget the second, and forgetting the second is how a right-looking answer turns out to be wrong.
The second annotation lines this range up with arccosine's, and that is the whole reason this convention is comfortable. Positive inputs give first-quadrant angles, negative inputs give second-quadrant angles, exactly as with arccosine, so nothing new has to be remembered about signs.
The third annotation explains the hole. Because pi over two is missing from the range, no input maps there; instead the outputs crowd up toward it from both sides as inputs get large. If you ever produce pi over two as an arcsecant, you have made an arithmetic error, because the secant of pi over two does not exist.
Concept
Under this convention arcsecant is not a new function to learn. It is arccosine wearing a reciprocal.
\[ \sec(t) = x \iff \cos(t) = \frac1x \]
\[ \operatorname{arcsec}(x) = \arccos\left(\frac1x\right) \quad \text{provided } |x| \ge 1 \]
\[ \operatorname{arccsc}(x) = \arcsin\left(\frac1x\right) \quad \text{provided } |x| \ge 1 \]
The identity works because the kept interval for secant is the kept interval for cosine, so the angle the left side names and the angle the right side names are the same angle, sign and all.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 828 — Theorem 10.28, the reciprocal properties
This is the payoff of matching cosine's interval, and the derivation is one line. Saying the secant of an angle is a number is the same as saying its cosine is the reciprocal of that number, because secant is one over cosine. So the angle arcsecant wants and the angle arccosine wants are described by the same condition.
The reason that argument needs the matching interval is subtle but worth getting right. Two functions can be described by the same equation and still return different angles if they are allowed to search different sets. Here the search sets are the same, apart from one removed point, so the two functions genuinely return the same angle, including for negative inputs.
In practice this means you never learn a new evaluation skill. Take the reciprocal, hand it to the inverse you already know, and you are done. Keep the condition attached, though, because the very next convention keeps the identity only on half the domain, and the habit of quoting the condition is what will protect you there.
Picture it
Figure (svg): Two horizontal number lines joined by arrows. On the top line the two rays from one rightward and from minus one leftward are highlighted; on the bottom line only the segment from minus one to one is highlighted. Arrows carry 2 to 0.5, 5 to 0.2, minus 2 to minus 0.5 and minus 1 to minus 1.
Taking one over a number of size at least one always produces a number of size at most one. So the reciprocal of a legal arcsecant input is always a legal arccosine input, with nothing left over and nothing missing except zero.
Follow the arrows. A number of size at least one goes to a number of size at most one, and it goes there reversibly: every value from minus one to one, except zero, comes back from exactly one input. Two goes to one half, five goes to one fifth, and the further out you start the closer to zero you land.
So the two highlighted sets on the diagram are the domain of arcsecant on top and the domain of arccosine on the bottom, and taking the reciprocal carries one exactly onto the other, apart from the single point zero. That is why handing the reciprocal to arccosine never produces an illegal input.
The missing zero is the picture of the missing output. Nothing maps to zero, and zero is the cosine of pi over two, which is precisely the angle that had to be removed from the range. Two holes, in two different places, both coming from the same fact.
Worked example
Find the exact value.
\[ \operatorname{arcsec}(2) \]
Apply the reciprocal identity
Why: Two has size at least one, so the identity in Theorem 10.28 applies.
\[ \operatorname{arcsec}(2) = \arccos\left(\frac12\right) \]
Evaluate the arccosine
Why: Cosine is one half at pi over three, and that angle lies in arccosine's range.
\[ \arccos\left(\frac12\right) = \frac{\pi}{3} \]
Figure (svg): A unit circle with a ray at sixty degrees, the terminal point marked one half comma root three over two, a dashed vertical segment down to the horizontal axis at one half, and the angle arc labelled pi over three.
Check by going forward
Why: Secant of pi over three is one over its cosine, and pi over three is inside the kept interval, so the answer is legal as well as correct.
\[ \sec\left(\frac{\pi}{3}\right) = \frac{1}{1/2} = 2 \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 829 — Example 10.6.3, part 1(a)
This is the easiest kind of arcsecant problem, and the point of working it slowly is the structure rather than the answer. Check the domain, apply the identity, evaluate a familiar arccosine, verify.
The unit circle makes the reciprocal visible. The horizontal coordinate of the point at pi over three is one half, so the cosine is one half, so the secant, its reciprocal, is two. You are reading the same picture you used for arccosine, just taking the reciprocal of the number you read off.
Do the verification even here. Taking the secant of your answer and getting the original input back confirms the value; noting that pi over three is inside the kept interval confirms that it is the right one of the infinitely many angles with that secant. Both halves of that check will matter enormously three slides from now.
Worked example
Find the exact value, staying with the trigonometry-friendly range.
\[ \operatorname{arccsc}(-2) \]
Apply the reciprocal identity
Why: Minus two has size two, so the identity applies to it.
\[ \operatorname{arccsc}(-2) = \arcsin\left(-\frac12\right) \]
Evaluate the arcsine
Why: Arcsine returns the angle nearest zero, and minus pi over six has sine minus one half.
\[ \arcsin\left(-\frac12\right) = -\frac{\pi}{6} \]
Confirm the answer is inside the range
Why: Minus pi over six lies in the lower half of the allowed set, which is exactly where negative inputs must land.
\[ -\frac{\pi}{2} \le -\frac{\pi}{6} < 0 \;\checkmark \]
Check by going forward
Why: The sine of minus pi over six is minus one half, so the cosecant is its reciprocal.
\[ \csc\left(-\frac{\pi}{6}\right) = \frac{1}{-1/2} = -2 \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 829 — Example 10.6.3, part 1(b)
A negative input, and under this convention it changes nothing about the method. The reciprocal identity holds for every legal input, so you take the reciprocal and hand it to arcsine exactly as you did for the positive case.
Arcsine returns the angle nearest zero with the required sine, which here is minus pi over six. Check that against the range: negative inputs must land in the lower half of the allowed set, from minus pi over two up to but not including zero, and minus pi over six is comfortably inside.
Remember this answer. In two parts of the lesson you will meet the same expression under the other convention and get seven pi over six instead, and the pair minus pi over six and seven pi over six is the single clearest illustration of what is at stake. Both angles have cosecant minus two; they differ by a full half turn plus a bit, and nothing but the declared range chooses between them.
Worked example
Find the exact value of an arcsecant applied to a secant.
\[ \operatorname{arcsec}\left(\sec\left(\frac{5\pi}{4}\right)\right) \]
Test the cancellation condition first
Why: Inside-out cancellation needs the inner angle to be in the kept interval. Five pi over four is bigger than pi, so it is not.
\[ \frac{5\pi}{4} \notin \left[0, \tfrac{\pi}{2}\right) \cup \left(\tfrac{\pi}{2}, \pi\right] \]
Work from the inside out instead
Why: Five pi over four is a third-quadrant angle with reference angle pi over four, where cosine is minus root two over two.
\[ \sec\left(\frac{5\pi}{4}\right) = \frac{1}{-\frac{\sqrt2}{2}} = -\sqrt2 \]
Now take the arcsecant of that number
Why: Minus root two has size greater than one, so the reciprocal identity applies.
\[ \operatorname{arcsec}(-\sqrt2) = \arccos\left(-\frac{\sqrt2}{2}\right) \]
Evaluate the arccosine
Why: Cosine is minus root two over two at three pi over four, which is in arccosine's range.
\[ \arccos\left(-\frac{\sqrt2}{2}\right) = \frac{3\pi}{4} \]
Check the answer against the input
Why: Three pi over four is not five pi over four, but the two angles are coterminal in secant: both have secant minus root two, and only one of them is in the range.
\[ \sec\left(\frac{3\pi}{4}\right) = -\sqrt2 = \sec\left(\frac{5\pi}{4}\right) \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 829 — Example 10.6.3, part 1(c)
The first step is the one to internalise: test the cancellation condition before you cancel. Five pi over four is bigger than pi, so it is not in the kept interval, and the inside-out rule simply does not apply. Writing five pi over four as the answer at this point is the standard mistake.
With the shortcut unavailable, you work from the inside out, which is always available. Five pi over four is a third-quadrant angle with reference angle pi over four, so its cosine is negative root two over two and its secant is minus root two. Now you have an ordinary arcsecant of a number, and the reciprocal identity finishes it.
The final check is worth reading twice. Three pi over four and five pi over four are different angles with the same secant, which is exactly the situation a restricted range exists to resolve. The inverse must pick one, and it picks the one inside its range. You are not losing information; you are obeying the definition of a function.
Trap
The outside-in rule is unconditional, so the inside-out one feels like it should be too:
\[ \operatorname{arcsec}\left(\sec\left(\frac{5\pi}{4}\right)\right) = \frac{5\pi}{4}? \]
Wrong under this convention. Five pi over four is not an output arcsecant can ever produce.
The inside-out rule carries a condition, and it is the range condition.
\[ \operatorname{arcsec}(\sec(x)) = x \]
\[ \text{only if } 0 \le x < \tfrac{\pi}{2} \text{ or } \tfrac{\pi}{2} < x \le \pi \]
Check the condition, and when it fails, evaluate the inside first and start again. The correct answer here is three pi over four.
The asymmetry here catches nearly everybody, so name it clearly. Going outside in, the secant of an arcsecant, only needs the input to be legal, and it returns the input unchanged. Going inside out, the arcsecant of a secant, needs the inner angle to be in the range, and when it is not the rule is simply false.
The way to remember which is which is to ask what has to be true for each expression to exist. For the outside-in version, the inner arcsecant must exist, so the condition is a domain condition. For the inside-out version, the answer must be something arcsecant can produce, so the condition is a range condition.
Whenever you see an inverse wrapped around its own function, stop and check the inner value against the range before doing anything else. If it fails, evaluate the inside and start again from an ordinary inverse problem, which is what the previous slide did.
Worked example
Find the exact value.
\[ \cot\left(\operatorname{arccsc}(-3)\right) \]
Name the inner angle
Why: Give the arccosecant a letter so the question becomes one about an ordinary angle.
\[ t = \operatorname{arccsc}(-3), \quad \csc(t) = -3 \]
Locate the angle from the sign
Why: The cosecant is negative, so the angle sits in the lower half of the range.
\[ -\frac{\pi}{2} \le t < 0 \]
Use the Pythagorean identity for cotangent
Why: This identity relates exactly the two functions in play.
\[ 1 + \cot^2(t) = \csc^2(t) \]
Substitute and solve
Why: Square minus three, subtract one, then take the root of both signs.
\[ 1 + \cot^2(t) = 9 \;\Longrightarrow\; \cot(t) = \pm\sqrt{8} = \pm 2\sqrt2 \]
Figure (svg): A unit circle with a ray into the fourth quadrant at about minus nineteen degrees; the height of its terminal point is minus one third and its horizontal coordinate is positive two root two over three, so the cotangent, the ratio of the two, is negative.
Choose the sign from the quadrant
Why: In the fourth quadrant the horizontal coordinate is positive and the height is negative, so their ratio is negative.
\[ \cot\left(\operatorname{arccsc}(-3)\right) = -2\sqrt2 \]
Check numerically
Why: The angle is about minus 0.3398 radians; its cosecant is minus three and its cotangent is about minus 2.8284, which is minus two root two.
\[ t \approx -0.3398, \quad \csc t = -3, \quad \cot t \approx -2.8284 \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 829 — Example 10.6.3, part 1(d)
The substitution in the first step is the technique that runs through this whole section: give the inverse expression a letter, and the problem turns into an ordinary identity problem about an angle. Everything you know about identities becomes available.
The second step is the one the identity cannot do for you. Squaring destroys sign information, so the Pythagorean identity can only ever tell you the size of the cotangent. The sign has to come from knowing where the angle is, and you know where it is because the cosecant is negative and the range has only one place for negative cosecants.
Look at the circle. The angle sits in the fourth quadrant, where the horizontal coordinate is positive and the height is negative. The cotangent is the first divided by the second, so it is negative, and you choose minus two root two. Every problem of this shape has that same three-part structure: identity for the size, quadrant for the sign, arithmetic for the check.
Fill the middle
Fill in the blanks
\operatorname-\frac{1}{2}(-2) = \arccos\left(\frac{2\pi}{3}\right) = ___
Why: The reciprocal of minus two is minus one half, and the angle in arccosine's range whose cosine is minus one half is two pi over three. This is Exercise 49, which states the trigonometry-friendly assumption explicitly.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 842 — Exercise 49
Two blanks, and the second one is the one that tests whether you are thinking about the range or just about cosines. The reciprocal of minus two is minus one half, which is straightforward. The angle is the part to be careful with.
Many angles have cosine minus one half. The one you want is the one in arccosine's range, from zero to pi, which is two pi over three. If you wrote four pi over three you have written a real angle with the right cosine, but not one this convention can return.
This is Exercise 49 from the section, and it is worth noticing that the exercise set announces its assumption in the instructions before the problem. When a question about a negative input does not tell you the convention, it is the question that is defective, not you.
Worked example
Rewrite as an algebraic expression in x, and state where the equivalence is valid.
\[ \tan(\operatorname{arcsec}(x)) \]
Substitute a letter for the inverse
Why: Naming the angle turns the composition into an ordinary identity problem.
\[ t = \operatorname{arcsec}(x), \quad \sec(t) = x, \quad t \in \left[0, \tfrac{\pi}{2}\right) \cup \left(\tfrac{\pi}{2}, \pi\right] \]
Choose the identity linking secant to tangent
Why: Only one Pythagorean identity contains both.
\[ 1 + \tan^2(t) = \sec^2(t) \]
Substitute and solve for the tangent
Why: Replace the secant by x and take the root of both signs.
\[ 1 + \tan^2(t) = x^2 \;\Longrightarrow\; \tan(t) = \pm\sqrt{x^2-1} \]
Split on the quadrant
Why: Tangent is not negative in the first quadrant and not positive in the second, so the range splits into two cases.
\[ \tan(t) = \begin{cases} \sqrt{x^2-1}, & 0 \le t < \tfrac{\pi}{2} \\ -\sqrt{x^2-1}, & \tfrac{\pi}{2} < t \le \pi \end{cases} \]
Translate the cases back into conditions on x
Why: Secant is at least one on the first piece and at most minus one on the second.
\[ \tan(\operatorname{arcsec}(x)) = \begin{cases} \sqrt{x^2-1}, & x \ge 1 \\ -\sqrt{x^2-1}, & x \le -1 \end{cases} \]
Check both branches at a number
Why: At five the formula gives root twenty-four, about 4.8990, and the tangent of arcsec five is 4.8990; at minus five the formula gives minus 4.8990 and so does the tangent.
\[ \tan(\operatorname{arcsec}(5)) \approx 4.8990, \qquad \tan(\operatorname{arcsec}(-5)) \approx -4.8990 \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 829 — Example 10.6.3, part 2(a)
This is the example the whole lesson turns on, so work it slowly. The first three steps are pure technique and will be identical under the other convention: name the angle, pick the Pythagorean identity that contains both functions, substitute and solve.
The split happens at the fourth step. The identity gave you plus or minus a square root, and the range is in two pieces where tangent has opposite signs: not negative on the first-quadrant piece, not positive on the second-quadrant piece. So the answer is genuinely a two-case formula, and the cases are inherited from the shape of the range.
The fifth step translates the cases from conditions on the angle to conditions on the input, which is what the question asked for. First-quadrant angles have secant at least one, second-quadrant angles have secant at most minus one, so the case boundary in x is exactly the hole in the domain. Keep this formula in view: the next part of the lesson exists to collapse it into one line.
Worked example
Rewrite as an algebraic expression in x, and state the domain of validity.
\[ \cos(\operatorname{arccsc}(4x)) \]
Substitute a letter
Why: The argument is four x, not x, so the domain restriction will come from four x.
\[ t = \operatorname{arccsc}(4x), \quad \csc(t) = 4x \]
Convert the cosecant to a sine
Why: Cosecant is the reciprocal of sine.
\[ \sin(t) = \frac{1}{4x} \]
Use the Pythagorean identity
Why: This is the identity that pairs sine with cosine.
\[ \cos^2(t) + \left(\frac{1}{4x}\right)^2 = 1 \]
Solve for the cosine
Why: Subtract, put over a common denominator, and take the root of both signs.
\[ \cos(t) = \pm\sqrt{\frac{16x^2-1}{16x^2}} = \pm\frac{\sqrt{16x^2-1}}{4|x|} \]
Choose the sign from the range
Why: The whole trigonometry-friendly range for arccosecant lies between minus pi over two and pi over two, where cosine is never negative.
\[ \cos(\operatorname{arccsc}(4x)) = \frac{\sqrt{16x^2-1}}{4|x|} \]
Find where the equivalence is valid
Why: The inner arccosecant needs its argument to have size at least one.
\[ |4x| \ge 1 \iff x \le -\frac14 \text{ or } x \ge \frac14 \]
Check at a negative value of x
Why: At x equal to minus one half the inner argument is minus two, the angle is minus pi over six, and its cosine is root three over two, about 0.8660; the formula gives root three over two as well, because the absolute value kills the sign.
\[ \cos(\operatorname{arccsc}(-2)) = \cos\left(-\tfrac{\pi}{6}\right) \approx 0.8660 = \frac{\sqrt{3}}{2} \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 829-830 — Example 10.6.3, part 2(b)
The new feature here is the four in front of the x. It changes nothing about the identity work, but it changes the domain: the condition applies to the whole argument of the arccosecant, which is four x, not to x. Solving the size condition for x gives quarters, not ones, and forgetting that is the usual slip.
The sign choice is easier than in the previous example, and it is worth seeing why. The entire trigonometry-friendly range for arccosecant lies between minus pi over two and pi over two, where cosine is never negative, so there is no case split at all. One range gives a case split for the tangent and no case split for the cosine; the shape of the range decides.
The absolute value in the denominator is not decoration. Because x may be negative, the square root of sixteen x squared is the size of four x, not four x itself, and writing four x there would give a negative value for a cosine that must be positive. The numerical check at the end is a negative value of x precisely so that you can see the absolute value earning its keep. One note on the printed text: the book's line just before this result drops the x from inside the square root, a typographical slip; the correct numerator is the root of sixteen x squared minus one throughout.
Matching
Match the pairs
Why: Every input here is positive, so take the reciprocal and hand it to arccosine or arcsine. One half gives pi over three, root two over two gives pi over four, and one half through arcsine gives pi over six. The two endpoint cases are worth memorising: arcsec of one is zero and arccsc of one is pi over two, because the secant of zero and the cosecant of pi over two are both one.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 841 — Exercises 33-40
Every input on the left is positive, which means every one of these is a place where the conventions agree and you can use the reciprocal shortcut without declaring anything. Take the reciprocal, hand it to arccosine or arcsine, read off the special angle.
Two of the five are worth committing to memory rather than recomputing. Arcsecant of one is zero, because the secant of zero is one; arccosecant of one is pi over two, because the cosecant of pi over two is one. They look like edge cases and they appear constantly, including as the endpoints on every graph in this lesson.
If you found yourself reaching for a calculator, slow down and go back to the unit circle. The reciprocals here are one half, root two over two and one half again, which are three of the values you already read off the circle in the last lesson. Nothing new is being asked; only the reciprocal step is new.
Error analysis
Annotate
On: \( \operatorname{arcsec}(2) = \frac{1}{\arccos(2)}? \)
This error is a genuine confusion about where the reciprocal lives, and it is worth slowing down on because the two directions really do differ. For the functions themselves, secant is the reciprocal of cosine, so the reciprocal applies to the output. For the inverse functions, the reciprocal applies to the input.
Work out why the direction flips. Arcsecant asks for an angle, and the arccosine of the same number would give a different angle, not a reciprocal one. Angles do not scale like values. What the two inverses share is the angle, so the relationship has to be arranged on the input side.
The clinching detail is the second annotation: the wrong expression is not merely a wrong number, it is not a number at all, because two is outside arccosine's domain. If you ever write down an arccosine of something bigger than one, that alone is enough to stop and reread the line.
Section
Part 3
Concept
Figure (svg): Two panels. Left: secant kept on zero to pi over two and on pi to three pi over two, the second piece now the branch that lies below minus one. Right: the reflected graph, an arcsecant whose right-hand branch is unchanged but whose left-hand branch now sits between pi and three pi over two.
\[ f(x) = \sec(x) \text{ on } \left[0, \tfrac{\pi}{2}\right) \cup \left[\pi, \tfrac{3\pi}{2}\right) \]
The first piece is unchanged. The second piece is the branch one half-turn further along, the one that starts at minus one and dives downward. It reaches exactly the same set of values as before, so the restricted function is still one-to-one and onto.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 830 — subsection 10.6.2
Compare the left panel here with the left panel two parts ago, and find the only difference. The first arc is identical. The second arc has moved one half-turn to the right, from the second quadrant to the third, and it is now a branch that starts at minus one and dives downward rather than one that climbs up to minus one.
Check that the restriction is still legal. The first piece produces every value of one and above, the second produces every value of minus one and below, so the coverage is exactly as good as before and the function is still one-to-one. Nothing about the mathematics is worse; it is a different valid choice.
In the right panel, the branch for positive inputs is unchanged, and the branch for negative inputs has lifted into the band between pi and three pi over two. Both asymptotes are drawn: outputs still crowd toward pi over two for large positive inputs, but now they crowd toward three pi over two for large negative ones.
Concept
Figure (svg): Two panels. Left: cosecant kept on zero to pi over two and on pi to three pi over two. Right: the reflected graph, an arccosecant whose right-hand branch runs down from pi over two toward zero and whose left-hand branch sits between pi and three pi over two.
\[ g(x) = \csc(x) \text{ on } \left(0, \tfrac{\pi}{2}\right] \cup \left(\pi, \tfrac{3\pi}{2}\right] \]
Both kept pieces now sit to the right of zero, which is what breaks the odd symmetry. This arccosecant returns only positive angles, so it is not an odd function, and the property listed at the bottom of Theorem 10.28 simply disappears from Theorem 10.29.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 830 — subsection 10.6.2
The cosecant version is the same half-turn shift, and it costs something visible. Both kept pieces now lie to the right of zero, so this arccosecant only ever returns positive angles, and the graph is no longer symmetric about the origin.
That is why Theorem 10.29 has no oddness clause at the bottom, while Theorem 10.28 does. If you are reading a reference that claims arccosecant is odd, you have learned something about which convention it uses before you have read another line.
Take a moment to be sure you can generate these two restrictions yourself rather than memorising four intervals. The recipe is: first piece is always the first-quadrant one, from zero to pi over two; second piece is either the second-quadrant one, which is the trigonometry-friendly choice, or the third-quadrant one, which is the calculus-friendly choice. The open and closed ends follow from where the function is undefined.
Notation
Annotate
On: \( \operatorname{arcsec}(x) = t \iff \sec(t) = x, \; t \in \left[0, \tfrac{\pi}{2}\right) \cup \left[\pi, \tfrac{3\pi}{2}\right) \)
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 831 — Theorem 10.29
Start by noticing what did not change. The domain is identical, still everything of size at least one, and the definition still runs both ways with the same two conditions. Only the set the angle is searched for in has moved.
The third annotation is the exception worth memorising. The bracket at pi is closed under this convention and the bracket at pi is also reachable under the other one, so arcsecant of minus one is pi in both books. It is the one negative input on which everybody agrees, and it makes a good sanity check when you are unsure which convention a source uses: if it disagrees at minus one, you have misread something.
The last annotation is the behaviour to carry into any graph you sketch. Large positive inputs still approach pi over two, but large negative inputs now approach three pi over two, from below. Sketching the two asymptotes first, before any curve, is the quickest way to get one of these graphs right.
Concept
This is the price of the new range. The shortcut that made the first convention pleasant only survives on half the domain.
\[ \operatorname{arcsec}(x) = \arccos\left(\frac1x\right) \quad \text{for } x \ge 1 \text{ only} \]
\[ \operatorname{arccsc}(x) = \arcsin\left(\frac1x\right) \quad \text{for } x \ge 1 \text{ only} \]
For a negative input the reciprocal identity returns a second-quadrant angle, and no second-quadrant angle is in this range. You have to go back to the definition and find the angle between pi and three pi over two instead.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 831 — Theorem 10.29, with its two footnotes
This is the cost of the new range, stated plainly, and the footnotes in the book exist to draw attention to it. The shortcut survives only for inputs of one and above.
Understand why rather than memorising the condition. The arccosine of a reciprocal always returns an angle between zero and pi. When the input is positive that angle is in the first quadrant, which is in the new range, so all is well. When the input is negative that angle is in the second quadrant, and the new range contains no second-quadrant angles at all, so the shortcut returns an answer that is simply not in the codomain.
What you do instead is go back to the definition, which is always available and never has conditions. Ask for the angle between pi and three pi over two with the required secant or cosecant, find the reference angle from the size of the input, and place it in the third quadrant. Two extra lines, and they are the same two lines every time.
Discrimination
Sort into buckets
Assume the calculus-friendly ranges. For which of these may you write the answer immediately as an arccosine or arcsine of the reciprocal?
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 831 — Theorem 10.29, footnotes b and c
Sort these by sign, because sign is the whole criterion here. Under the calculus-friendly range the reciprocal identity carries the condition input at least one, so every positive input is safe and every negative one is not.
One item deserves a second look after you sort. Arcsecant of minus one comes out as pi whether you use the shortcut or the definition, because pi is a shared endpoint. So the shortcut gives the right answer there. It still belongs in the second bucket, because the theorem does not license it and you would be relying on a coincidence rather than a rule.
If you sorted by whether the answer exists rather than by whether the shortcut is licensed, reread the question. All six expressions are perfectly well defined. The sort is about method, not about existence, and that distinction is exactly what a mathematical condition is for.
Worked example
Two values that the calculus-friendly range makes shorter, not longer.
\[ \operatorname{arcsec}(2) \qquad \text{and} \qquad \operatorname{arcsec}\left(\sec\left(\frac{5\pi}{4}\right)\right) \]
Evaluate the first with the reciprocal identity
Why: Two is at least one, so Theorem 10.29 permits the shortcut.
\[ \operatorname{arcsec}(2) = \arccos\left(\frac12\right) = \frac{\pi}{3} \]
Test the cancellation condition on the second
Why: Inside-out cancellation now needs the angle to lie between pi and three pi over two, and five pi over four does.
\[ \pi \le \frac{5\pi}{4} < \frac{3\pi}{2} \;\checkmark \]
Cancel directly
Why: The condition holds, so the two operations undo each other with no inside-out work at all.
\[ \operatorname{arcsec}\left(\sec\left(\frac{5\pi}{4}\right)\right) = \frac{5\pi}{4} \]
Check against the other convention's answer
Why: Both answers have secant minus root two, about minus 1.4142; the two angles differ by exactly pi over two, and each book keeps the one inside its own range.
\[ \sec\left(\frac{3\pi}{4}\right) = \sec\left(\frac{5\pi}{4}\right) = -\sqrt2 \approx -1.4142 \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 832 — Example 10.6.4, parts 1(a) and 1(c)
It would be easy to come away thinking the calculus-friendly convention is uniformly more work. These two parts are the corrective. The first value is identical to before, because the input is positive and the conventions agree there.
The second value is where this convention actually wins on effort. Five pi over four lies between pi and three pi over two, so it is inside this range, so the inside-out cancellation applies directly and the answer is five pi over four with no work at all. Under the other convention the same expression needed three steps of inside-out evaluation.
That is the honest summary of the trade. Each convention makes some angles cancel and others not, and which ones depends entirely on which half-turn the second piece was taken from. The final check makes the relationship concrete: the two answers differ by exactly pi over two, and both have secant minus root two.
Worked example
The same value as before, but the shortcut is now off limits.
\[ \operatorname{arccsc}(-2) \]
Test the shortcut condition
Why: Minus two is not at least one, so Theorem 10.29 does not license the arcsine of the reciprocal.
\[ -2 \not\ge 1 \;\Longrightarrow\; \text{no shortcut} \]
Return to the definition
Why: Ask for the angle in the range whose cosecant is minus two.
\[ t = \operatorname{arccsc}(-2), \quad \csc(t) = -2, \quad t \in \left(0, \tfrac{\pi}{2}\right] \cup \left(\pi, \tfrac{3\pi}{2}\right] \]
Convert to a sine and locate the quadrant
Why: A negative sine rules out the first piece, so the angle is in the third quadrant.
\[ \sin(t) = -\frac12, \quad \pi < t \le \frac{3\pi}{2} \]
Find the angle with that reference angle
Why: The reference angle for a sine of one half is pi over six, and the third-quadrant angle with that reference angle is pi plus pi over six.
\[ t = \pi + \frac{\pi}{6} = \frac{7\pi}{6} \]
Check by going forward
Why: The sine of seven pi over six is minus one half, so the cosecant is minus two, and seven pi over six is about 3.6652, comfortably inside the third-quadrant piece.
\[ \csc\left(\frac{7\pi}{6}\right) = -2, \quad \pi \approx 3.1416 < 3.6652 \le 4.7124 \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 832 — Example 10.6.4, part 1(b)
This is the same expression as Example 10.6.3 part 1(b), and the point is to feel the difference in method as well as in answer. There the reciprocal identity did the whole job in one line. Here the first step is to notice that the identity is barred.
The definition route is worth practising until it is automatic, because it is the only thing that always works. State the defining condition with the range attached, convert to a sine or cosine, use the sign to pick which piece of the range you are in, then build the angle from its reference angle.
The reference-angle step is the one to be careful with. You find the reference angle from the size of the value, which is one half here, giving pi over six, and then you place it by adding pi because you need a third-quadrant angle. The answer, seven pi over six, is not the negative of the other convention's answer; it is a genuinely different angle that happens to have the same cosecant.
Worked example
Repeat Example 10.6.3 1(d) under the new range and watch the answer flip.
\[ \cot\left(\operatorname{arccsc}(-3)\right) \]
Name the angle and locate it
Why: The cosecant is negative, and the only negative-cosecant piece of this range is the third-quadrant one.
\[ t = \operatorname{arccsc}(-3), \quad \pi < t \le \frac{3\pi}{2} \]
Use the same Pythagorean identity
Why: Nothing about the identity depends on the convention.
\[ 1 + \cot^2(t) = \csc^2(t) = 9 \;\Longrightarrow\; \cot(t) = \pm 2\sqrt2 \]
Figure (svg): A unit circle with a ray into the third quadrant at about one hundred ninety-nine degrees; the height of its terminal point is minus one third and its horizontal coordinate is minus two root two over three, so the cotangent, the ratio of the two, is positive.
Choose the sign from the new quadrant
Why: In the third quadrant both coordinates are negative, so the cotangent, their ratio, is positive.
\[ \cot\left(\operatorname{arccsc}(-3)\right) = 2\sqrt2 \]
Check numerically and against the other answer
Why: The angle is about 3.4814 radians, its cosecant is minus three and its cotangent about 2.8284; the trigonometry-friendly answer was minus 2.8284, the same size with the opposite sign.
\[ t \approx 3.4814, \quad \cot t \approx +2.8284 \quad\text{versus}\quad -2.8284 \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 832 — Example 10.6.4, part 1(d)
Set this side by side with the version from Part 2 and notice how little of the work changes. The substitution is the same, the Pythagorean identity is the same, and the size of the answer, two root two, is the same. Only the last step differs.
That is because the identity can only ever produce a size. Squaring threw the sign away, and the sign has to be recovered from the position of the angle. Under the earlier convention the angle was in the fourth quadrant, with a positive horizontal coordinate and a negative height, so the ratio was negative. Now it is in the third quadrant, with both coordinates negative, so the ratio is positive.
Two circles, two quadrants, two signs, one identity. If you ever find yourself trying to remember whether the answer is positive or negative, do not: draw the angle and read the two coordinates. The picture is faster than the memory and it never misleads you.
Worked example
Rewrite as an algebraic expression in x and state where it is valid, exactly as before.
\[ \tan(\operatorname{arcsec}(x)) \]
Substitute and write down the identity
Why: Everything up to here is word for word what the first convention gave.
\[ t = \operatorname{arcsec}(x), \quad 1 + \tan^2(t) = \sec^2(t) = x^2 \]
Solve for the tangent
Why: Two signs again, for the moment.
\[ \tan(t) = \pm\sqrt{x^2-1} \]
Read the sign off the range
Why: Tangent is not negative in the first quadrant and not negative in the third quadrant either, and those are the only two places this range goes.
\[ t \in \left[0, \tfrac{\pi}{2}\right) \cup \left[\pi, \tfrac{3\pi}{2}\right) \;\Longrightarrow\; \tan(t) \ge 0 \]
Write the single formula
Why: One rule for every legal input, with no reference to the sign of x.
\[ \tan(\operatorname{arcsec}(x)) = \sqrt{x^2-1}, \quad |x| \ge 1 \]
Check at a negative input
Why: At minus five the formula gives root twenty-four, about 4.8990; the angle is about 4.5110 radians, in the third quadrant, and its tangent is 4.8990. The other convention gave minus 4.8990.
\[ \tan(\operatorname{arcsec}(-5)) \approx 4.8990 \quad\text{versus}\quad -4.8990 \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 832 — Example 10.6.4, part 2(a)
Compare the first two steps with the trigonometry-friendly version and confirm for yourself that they are word for word identical. The identity does not know about conventions. Everything that follows from squaring is the same in both books.
The third step is where the convention pays off. This range is the first-quadrant piece together with the third-quadrant piece, and tangent is not negative in either of them, because in the first quadrant both coordinates are positive and in the third both are negative. So the plus sign is correct everywhere and the case split evaporates.
Look at what the answer now is: one square root, valid for every legal input, with no reference to the sign of x. That single line is the reason this convention exists, and it is the line that all the extra fiddling with negative inputs was bought with. The numerical check at minus five is there so that you can see the two conventions giving opposite signs at one concrete input.
Picture it
Figure (svg): Two panels of the tangent of an arcsecant against x. Left, trigonometry friendly: the right branch rises above the axis and the left branch falls below it, a two-piece formula. Right, calculus friendly: both branches rise above the axis in a single symmetric pair, one formula.
Set the two results side by side. On the left the graph has to be described in two pieces with two signs; on the right it is one square root for the whole domain. Every formula built on top of it inherits that difference, which is why the choice matters far beyond this section.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 826 — the two points of view
Put the two panels side by side and the argument becomes visual. On the left, the graph of the tangent of an arcsecant is a curve above the axis on one side and below it on the other; there is no single algebraic expression for it without a sign that depends on which side you are on.
On the right, the same composition under the other convention is symmetric about the vertical axis, so one square root describes all of it. If you go on to calculus, that is the difference between an integral you can write down and one that needs a case analysis every time.
Notice also that the two pictures are identical for positive inputs, which is the recurring theme of the lesson. Every difference between the conventions lives on the left half of these graphs, and every difference you will ever have to worry about comes from an input that is negative.
Worked example
The same composition as before, now under the calculus-friendly range.
\[ \cos(\operatorname{arccsc}(4x)) \]
Repeat the algebra
Why: Nothing before the sign choice depends on the convention.
\[ \sin(t) = \frac{1}{4x} \;\Longrightarrow\; \cos(t) = \pm\frac{\sqrt{16x^2-1}}{4|x|} \]
Split on the two pieces of the range
Why: Cosine is not negative in the first quadrant and not positive in the third.
\[ \cos(t) = \begin{cases} +\frac{\sqrt{16x^2-1}}{4|x|}, & 0 < t \le \tfrac{\pi}{2} \\ -\frac{\sqrt{16x^2-1}}{4|x|}, & \pi < t \le \tfrac{3\pi}{2} \end{cases} \]
Translate each case into a condition on x
Why: The first piece has cosecant at least one, the second has cosecant at most minus one.
\[ 0 < t \le \tfrac{\pi}{2} \iff x \ge \tfrac14, \qquad \pi < t \le \tfrac{3\pi}{2} \iff x \le -\tfrac14 \]
Resolve the absolute value in each case
Why: On the first piece the absolute value of x is x; on the second it is minus x, and that minus cancels the leading minus.
\[ \frac{\sqrt{16x^2-1}}{4x} \quad\text{and}\quad -\frac{\sqrt{16x^2-1}}{4(-x)} = \frac{\sqrt{16x^2-1}}{4x} \]
Write the single formula
Why: The two cases produced the same expression, so the case split collapses.
\[ \cos(\operatorname{arccsc}(4x)) = \frac{\sqrt{16x^2-1}}{4x}, \quad x \le -\tfrac14 \text{ or } x \ge \tfrac14 \]
Check at a negative value of x
Why: At x equal to minus one half the angle is seven pi over six, whose cosine is about minus 0.8660; the formula gives root three over two divided by minus two, which is also minus 0.8660.
\[ \cos\left(\frac{7\pi}{6}\right) \approx -0.8660 = \frac{\sqrt{3}}{-2} \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 832-833 — Example 10.6.4, part 2(b)
This one is a little more intricate than the tangent example, because the case split does not vanish immediately. The range still has two pieces where cosine has opposite signs, so you do get two cases.
The magic is in the fifth step. On the first piece x is positive, so the size of x is x. On the second piece x is negative, so the size of x is minus x, and that minus cancels against the minus already in front of the fraction. Two cases in, one expression out.
So the absolute value that was mandatory under the trigonometry-friendly convention is not merely optional here, it is wrong: it would give a positive cosine for third-quadrant angles, where the cosine is negative. Check that on the numerical line at the end. At x equal to minus one half the true cosine is about minus 0.8660, and only the version without the absolute value produces it.
Concept
Differentiating each version of arcsecant shows the whole point of the fork in one line. Both are obtained by differentiating the reciprocal identity, and they differ by exactly the sign the case split introduced.
\[ \text{trig friendly:} \quad \frac{d}{dx}\,\operatorname{arcsec}(x) = \frac{1}{|x|\sqrt{x^2-1}} \]
\[ \text{calc friendly:} \quad \frac{d}{dx}\,\operatorname{arcsec}(x) = \frac{1}{x\sqrt{x^2-1}} \]
An absolute value inside a derivative is a nuisance: it blocks a clean antiderivative and forces a case split in every integral that uses it. That single character is what the second convention was designed to remove, and it is the fastest way to tell which convention a calculus book has adopted.
This slide is a preview rather than a required result, but it is the reason the argument has survived for a century, so it is worth seeing. Differentiate the reciprocal identity in either convention and you get the same expression apart from one character.
The absolute value is a serious nuisance in an integral. An antiderivative containing it has to be handled in two cases, and every substitution that produces an arcsecant inherits the split. Remove it and a whole family of integrals becomes a single line.
It is also the most reliable way to identify a book's convention in about two seconds. Find the derivative of arcsecant in its formula table. An absolute value means trigonometry friendly, no absolute value means calculus friendly, and you now know which values that book's negative inputs will produce without reading a word of its prose.
Trap
Working in a calculus-friendly book, out of habit:
\[ \operatorname{arcsec}(-2) = \arccos\left(-\tfrac12\right) = \tfrac{2\pi}{3}? \]
Wrong here. Two pi over three is not in this range at all.
The identity's footnote says for x at least one only. With a negative input, go to the definition and find the third-quadrant angle with that secant.
\[ \sec(t) = -2, \quad \pi \le t < \tfrac{3\pi}{2} \]
\[ \operatorname{arcsec}(-2) = \frac{4\pi}{3} \]
Both answers have cosine minus one half; only one of them is inside the range this book declared.
This is the error the footnotes in Theorem 10.29 are written to prevent, and it is easy to commit because the shortcut is so comfortable. You do the reciprocal, you do the arccosine, you write down a real angle with the correct secant, and nothing in the arithmetic complains.
What complains is the range. Two pi over three is a second-quadrant angle, and this convention's arcsecant never returns a second-quadrant angle, so the answer is outside the function's codomain. The verification habit catches it immediately: check the value, then check the range, and the second check fails.
The repair is mechanical. Find the reference angle, which is pi over three here, and place it in the third quadrant by adding pi, giving four pi over three. Both candidate answers have cosine minus one half; only one of them is inside the declared range.
Section
Part 4
Concept
Figure (svg): A horizontal axis from minus six to six. The stretch from minus six to minus one is a thick red bar labelled conventions disagree, the stretch from minus one to one is an empty grey gap labelled not in the domain, and the stretch from one to six is a thick green bar labelled conventions agree.
On the positive half of the domain the two conventions keep the same piece of the graph, so they return the same angle, always. The first restricted piece is the interval from zero to pi over two in both theorems.
\[ x \ge 1 \;\Longrightarrow\; \operatorname{arcsec}(x) = \arccos\left(\frac1x\right) \text{ under both conventions} \]
The instant an input goes negative, the answer depends on the book. There is one exception worth knowing: both send minus one to pi, because pi is the shared endpoint of the two second pieces.
This strip is the map of the whole dispute, and it is smaller than it looks. The green stretch on the right is complete agreement: both conventions restrict secant to the same first piece, so both return the same first-quadrant angle, and no declaration is needed.
The grey gap in the middle is not a disagreement at all, it is the hole in the domain. Neither convention has anything to say there because there is nothing to say.
Only the red stretch on the left is contested, and even there the two answers are closely related: one is the reflection of the other about the vertical line at pi. There is one exception, the point minus one, where the two second pieces share an endpoint and both conventions return pi. When you are working a problem, the practical rule is short: positive input, just answer; negative input, declare first.
Picture it
Figure (svg): One pair of axes carrying both versions of arcsecant. For x at least one the two versions lie on top of each other in a single curve rising toward pi over two. For x at most minus one there are two separate curves: one descending to pi at minus one, the other running between pi and three pi over two.
The right-hand branch is drawn once because both conventions produce it. On the left there are genuinely two different curves. Read any negative input off both and you get two angles whose secants agree and whose values differ by more than a right angle.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 827-830 — the two reflected graphs
This is the single picture to carry out of the lesson. The thick curve on the right is drawn once because both conventions produce exactly it. The two thinner curves on the left are the whole disagreement.
Trace one input across. Pick a value around minus four on the horizontal axis and go up: you meet the lower curve at an angle a bit above pi over two, and the upper curve at an angle a bit below three pi over two. Both angles have secant minus four. Neither is wrong.
Notice the geometric relationship between the two left branches. They are mirror images in the horizontal line at pi, which is exactly what you would expect: one convention's second piece is the other's reflected through the point pi on the angle axis. That relationship is also the conversion rule if you ever have to move an answer from one book to the other.
Picture it
Figure (svg): One pair of axes carrying both versions of arccosecant. For x at least one the two curves coincide, running down from pi over two toward zero. For x at most minus one one version lies below the axis near minus pi over two while the other sits high, between pi and three pi over two.
Here the split is even more visible: one version answers below the horizontal axis and the other high above it. The dot at minus one shows the disagreement at its cleanest, minus pi over two against three pi over two, two angles a full turn plus a half apart in appearance and identical in cosecant.
The cosecant version of the same picture, and the split here is even more dramatic to the eye, because one convention answers below the horizontal axis and the other answers high above it.
The two dots at minus one are worth staring at. One convention says minus pi over two, the other says three pi over two. Those two angles are two full right angles plus a half turn apart on the page, and they are the same direction on the circle: the cosecant of both is minus one. All that separates them is which set of angles the definition was told to search.
Also notice which curve keeps the symmetry. The lower one is symmetric through the origin, so that arccosecant is odd; the upper one is not, so that arccosecant is not. If a problem uses the oddness of arccosecant anywhere in its solution, it has silently committed to the trigonometry-friendly convention.
Comparison
Comparison matrix
| question | trigonometry friendly | calculus friendly |
|---|---|---|
| arcsec(2) | pi/3 | pi/3 |
| arcsec(-2) | 2pi/3 | 4pi/3 |
| arccsc(-2) | -pi/6 | 7pi/6 |
| arcsec(-1) | pi | pi |
| is arccsc odd? | yes | no |
| tan(arcsec x), x at most -1 | -sqrt(x^2-1) | sqrt(x^2-1) |
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 828-831 — Theorems 10.28 and 10.29 compared
Fill the blanks from the two theorems rather than from memory of the examples, and you will find that four of the six rows write themselves. The first row is a positive input, so the two columns must match. The fourth row is the shared endpoint, so those must match too.
The two rows in the middle are the ones that force you to think about where negative inputs go. Second quadrant for the left column, third quadrant for the right, and the arithmetic follows from the reference angle.
The last two rows are the structural consequences rather than single values. Oddness survives only when the range is symmetric about zero, which only the left column is. And the tangent formula needs a case split only when the range contains a quadrant where tangent is negative, which again only the left column does. Both differences come from the same source.
Edge cases
Parameter explorer
The curve is arcsecant. The parameter w morphs the branch for negative inputs from the trigonometry-friendly position at zero to the calculus-friendly position at one. Watch the two halves and say which one moves.
\[ y = \operatorname{arcsec}(x), \quad w = {w} \]
Put the slider at zero first and look at the shape: the branch on the right rises toward a level, and the branch on the left sits below a level that is roughly twice as high. That is the trigonometry-friendly arcsecant.
Now push the slider slowly to one. Watch the right branch: it does not move at all, at any setting, because the conventions agree for inputs of one and above. All the motion is on the left, where the branch lifts into the upper band and flips over as it goes, so the end that was high becomes the end that is low.
The flip is the part worth noticing rather than the lift. Under one convention the arcsecant is increasing on the negative side; under the other it is decreasing there. That reversal of direction is exactly the sign difference you saw in the two derivative formulas, and this is what it looks like as a shape.
Worked example
Evaluate under both conventions and compare.
\[ \tan(\operatorname{arcsec}(5)) \]
Locate the angle under each convention
Why: Five is positive, so both conventions use the first kept piece, the first quadrant.
\[ t = \arccos\left(\tfrac15\right) \approx 1.3694 \text{ in both} \]
Apply the identity
Why: One plus tangent squared is secant squared, and the secant is five.
\[ 1 + \tan^2(t) = 25 \;\Longrightarrow\; \tan(t) = \pm\sqrt{24} \]
Choose the sign
Why: A first-quadrant angle has a positive tangent, under either convention.
\[ \tan(\operatorname{arcsec}(5)) = \sqrt{24} = 2\sqrt6 \]
Check numerically
Why: Two root six is about 4.8990, and the tangent of 1.3694 radians is 4.8990. Both formulas from Part 2 and Part 3 give the same value here, because x is at least one.
\[ 2\sqrt6 \approx 4.8990 = \tan(1.3694) \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 842 — Exercises 77-81, same pattern
Nothing in this example depends on the convention, and that is the point of working it just before the one that follows. A positive input sends both conventions to the first quadrant, so both find the same angle and both take the tangent of the same angle.
The technique is the same three-part structure you have now used four times: identity for the size, quadrant for the sign, numerical check at the end. In the first quadrant tangent is positive, so the plus sign is correct and there is nothing to argue about.
Keep the exact value and the decimal side by side. Two root six is the answer you write, and 4.8990 is the answer you check against. On the next slide only one of those two numbers changes, and being able to say in advance which one is a good test of whether you have understood the lesson.
Worked example
Change the sign of the input and repeat, under both conventions.
\[ \tan(\operatorname{arcsec}(-5)) \]
Find the angle the trigonometry-friendly range gives
Why: The reciprocal identity applies for every legal input there.
\[ t_1 = \arccos\left(-\tfrac15\right) \approx 1.7722, \text{ second quadrant} \]
Find the angle the calculus-friendly range gives
Why: The third-quadrant angle with the same secant is a half-turn further on.
\[ t_2 = 2\pi - 1.7722 \approx 4.5110, \text{ third quadrant} \]
Take the tangent of each
Why: The size is the same both times, because the identity only ever gave the size.
\[ \tan(t_1) \approx -4.8990, \qquad \tan(t_2) \approx +4.8990 \]
Read the two general formulas
Why: This is Example 10.6.3 2(a) against Example 10.6.4 2(a), at one input.
\[ -\sqrt{x^2-1} \;\text{versus}\; \sqrt{x^2-1} \quad \text{at } x = -5 \]
Check that both angles really have secant minus five
Why: The cosine of 1.7722 and the cosine of 4.5110 are both minus one fifth, so both angles are legitimate answers to the question; only the declared range picks between them.
\[ \sec(1.7722) = \sec(4.5110) = -5 \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 829-832 — the two versions of part 2(a)
Predict before you read the steps: which changes when the input goes negative, the size of the answer or its sign? The size comes from the Pythagorean identity, which never saw a sign, so the size cannot change. Only the sign can.
The two angles are worth comparing directly. One is about 1.7722 radians, in the second quadrant, where tangent is negative. The other is about 4.5110 radians, in the third quadrant, where tangent is positive. Both have secant exactly minus five, which the final check confirms, so both are honest answers to the question what angle has secant minus five.
This is the whole lesson in one example. Two books, one input, two answers, both correct, differing by a sign that a reader who does not know the convention has no way to predict. It is also why the last check is written the way it is: verifying that both angles have the right secant proves that the disagreement is about the range and nothing else.
Sorting
Sort into buckets
Sort each expression by whether the two conventions return the same value.
Sort these by asking one question of each: does evaluating it ever require you to reach the second piece of the range with a genuinely negative input? If not, the conventions agree.
Two of the same-bucket items are subtle. Arcsecant of minus one is the shared endpoint, so it is pi either way even though the input is negative. And the secant of an arcsecant of minus nine returns minus nine under both conventions, because the outside-in cancellation gives you the input back regardless of which angle the inverse picked in the middle. The disagreement is real but invisible, because it cancels.
That second case is worth holding on to. A convention difference can sit inside a calculation and never surface, if every use of the angle is through a function that cannot tell the two candidates apart. Secant cannot. Tangent and cotangent can, which is why those are the compositions that keep landing in the other bucket.
Error analysis
Annotate
On: \( \tan(\operatorname{arcsec}(-3)) = \sqrt{9-1} = 2\sqrt2, \quad \operatorname{arcsec}(-3) = \arccos\left(-\tfrac13\right) \)
This is the most expensive error in the section because neither half of it looks wrong. The formula on the left is a correct theorem. The formula on the right is a correct theorem. They are correct in different books.
Work out the contradiction numerically for yourself before reading the annotations. The arccosine of minus one third is about 1.9106 radians, and the tangent of that is about minus 2.8284. But the first half of the line claims the answer is plus 2.8284. One line, two values, no arithmetic mistake anywhere.
The defence is procedural rather than mathematical: write the convention at the top of the page, and from then on use only that theorem's properties. If you find yourself reaching for a property and cannot remember which theorem it came from, look it up rather than guessing. In this section, a property from the wrong theorem is worse than no property at all.
Real world
Discussion prompt
You open an unfamiliar calculus reference and find the derivative of arcsecant printed in its table of formulas. Without reading a word of the surrounding text, how can you tell which convention the book assumes, and what would go wrong if you mixed its table with a trigonometry-friendly evaluation?
Write your answer before revealing. The question is practical, and you will actually face it: sooner or later you will use a table of derivatives or a computer algebra system whose convention nobody told you.
The absolute value is the tell, and the reason it is a tell is the shape you saw on the slider slide. The trigonometry-friendly arcsecant increases on both halves of its domain, so its derivative is positive on both, which forces the absolute value into the denominator. The calculus-friendly one decreases for negative inputs, so its derivative must be negative there, and that is exactly what dropping the absolute value achieves.
The consequence of mixing them is the worst kind of error: silent and sign-sized. Values from one source and slopes from another would agree perfectly for every positive input, so your testing would pass, and then every negative input would come out with the wrong sign. If you are writing code that uses an arcsecant, test it at a negative input, and test it against the definition rather than against another library.
Section
Part 5
Concept
Almost every calculator offers only three inverse circular functions, labelled with the raised minus-one notation. The other three have to be built from them.
| want | press | valid when |
|---|---|---|
| arcsec(x) | arccos of 1/x | always, if the trigonometry-friendly range is in force |
| arccsc(x) | arcsin of 1/x | always, if the trigonometry-friendly range is in force |
| arcsec(x) | arccos of 1/x | only for x at least 1 under the calculus-friendly range |
| arccot(x) | arctan of 1/x | only for x positive (Theorem 10.27) |
Notice the pattern: the reciprocal trick is safe for positive inputs in every case. It is negative inputs, again, that need care.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 833 — subsection 10.6.3, opening
Calculators have three inverse buttons because arcsine, arccosine and arctangent are the three everyone agrees about. The other three have to be assembled, and the table is the assembly instruction.
Read down the last column and one pattern jumps out. Every entry is safe for positive inputs and every restriction concerns negative ones. That is the same pattern you have been meeting all lesson, now in a practical setting: a positive input never forces you to think, a negative one always does.
The arccotangent row is included because it belongs to the same family of workarounds, even though its convention was settled back in the previous lesson. For a negative input there you take a detour through the reference angle, and the next lesson works that case in full. Here, concentrate on the two rows that are yours: arcsecant and arccosecant.
Worked example
Approximate, in radian mode.
\[ \operatorname{arcsec}(5) \]
Check which property is available
Why: Five is at least one, so both Theorem 10.28 and Theorem 10.29 offer the same reciprocal identity.
\[ 5 \ge 1 \;\Longrightarrow\; \operatorname{arcsec}(5) = \arccos\left(\frac15\right) \]
Evaluate the reciprocal
Why: One fifth is 0.2 exactly.
\[ \frac15 = 0.2 \]
Press the arccosine key
Why: In radian mode, not degree mode.
\[ \arccos(0.2) \approx 1.3694 \]
Check the size of the answer
Why: The answer should be a first-quadrant angle just under pi over two, about 1.5708, because a large secant means a small cosine. It is, and the secant of 1.3694 is 5.0000.
\[ 1.3694 < \frac{\pi}{2} \approx 1.5708, \quad \sec(1.3694) \approx 5.0000 \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 833 — Example 10.6.5, part 1(b)
The reason this one is easy is the first step, and the first step is a check rather than a calculation. Five is at least one, so both theorems offer the identity, and you do not even have to know which convention is in force.
Make sure your calculator is in radian mode. An arcsecant reported as 78.46 is not a wrong calculation; it is the same angle in degrees, and in a section full of exact multiples of pi, degrees will cause you far more trouble than they solve.
The final step is the estimate that should accompany every calculator answer. A secant of five means a cosine of one fifth, which is small, so the angle must be close to a right angle but a little under it. You expect something just below 1.5708, and you get 1.3694. If you had got 0.2014 instead you would know at once that you had pressed arcsine rather than arccosine.
Worked example
Approximate under each convention, and notice that the calculator work is the same either way.
\[ \operatorname{arccsc}\left(-\frac32\right) \]
Take the reciprocal
Why: The reciprocal of minus three halves is minus two thirds.
\[ \frac{1}{-3/2} = -\frac23 \approx -0.6667 \]
Press arcsine for the trigonometry-friendly answer
Why: Theorem 10.28 allows the identity for every legal input.
\[ \operatorname{arccsc}\left(-\tfrac32\right) = \arcsin\left(-\tfrac23\right) \approx -0.7297 \]
Convert to the calculus-friendly answer
Why: That range wants the third-quadrant angle with the same sine, which is a half-turn past the reference angle.
\[ t = \pi + \arcsin\left(\tfrac23\right) \approx 3.1416 + 0.7297 = 3.8713 \]
Check both angles
Why: The sine of minus 0.7297 and the sine of 3.8713 are both about minus 0.6667, so both have cosecant minus 1.5000, and each lies in its own range.
\[ \csc(-0.7297) = \csc(3.8713) \approx -1.5000 \;\checkmark \]
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 833 — Example 10.6.5, part 1(d)
The useful discovery here is that the calculator work is the same under both conventions. You press exactly the same keys. What differs is what you do with the number afterwards.
Under the trigonometry-friendly range the arcsine output is the answer, because that range and arcsine's range coincide. Under the calculus-friendly range you take the reference angle, which is the size of that same output, and add pi to place it in the third quadrant. One extra addition.
The check at the end is the important habit. Both angles have cosecant minus 1.5000, so both are real answers to the question; the verification that separates them is the range test, not the value test. Get into the habit of writing both halves of the check, because on a negative input the value half can never catch a convention slip.
Estimation
\[ \operatorname{arcsec}(1000) \]
Predict first
Without a calculator, which is closest to the arcsecant of one thousand, under either convention?
Correct: Close to 1.5708
Why: A huge secant means a cosine near zero, and the angle whose cosine is near zero from above is just under pi over two, about 1.5708. Exactly: the arccosine of one thousandth is about 1.5698. The line at pi over two is the horizontal asymptote both conventions share on the positive side, so the outputs crowd up against it and never reach it.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, p. 828 — Theorem 10.28, the asymptotic behaviour
Answer from the shape of the graph rather than from arithmetic. A secant of a thousand needs a cosine of a thousandth, which is nearly zero, and the angle whose cosine is nearly zero, approached from the positive side, is just under a right angle.
The precise value is about 1.5698, against pi over two at about 1.5708. So the answer is within one hundredth of the asymptote, and it will never reach it however large the input grows. That is what a horizontal asymptote means on an inverse graph: an angle the function approaches and never attains.
This estimate is convention-independent, which is the other thing worth noticing. A thousand is positive, so both conventions send it to the same first-quadrant angle and both share the asymptote at pi over two. All the asymptotic disagreement is on the negative side, where one convention heads for pi over two and the other for three pi over two.
Section
Part 6
Pattern
Figure (svg): A flow diagram: check the size of the input first, and reject anything smaller than one. If the input is positive, both conventions give arccos of the reciprocal. If it is negative, branch: the trigonometry-friendly route still takes arccos of the reciprocal, while the calculus-friendly route finds the angle between pi and three pi over two with that secant.
Read the diagram as two cheap questions followed by one expensive one. The size test can end the problem immediately and costs one comparison. The sign test costs nothing and settles whether you need to think about conventions at all.
Only the bottom-right corner of the diagram is genuinely contested, and you reach it only when the input is negative. That is a useful thing to know before an exam: most arcsecant questions you will be set never touch the disputed part of the theory.
The last line of the list is the one people skip. Verification here has two halves. Taking the secant or cosecant of your answer checks the value, and confirming the angle lies in the declared range checks that you picked the right angle among the infinitely many with that value. On a negative input, only the second half can catch a convention error, so it is the half that earns its keep.
Ranking
Put in order
Put a complete calculus-friendly evaluation of arcsec of a negative number in order.
Why: The domain check comes first because it can end the problem immediately. Ruling out the shortcut before using the definition is what stops the most common error. The verification at the end has two halves, the value and the range, and the range half is the one that catches a convention slip.
Order these by asking what each step would cost you if it came later. The domain test goes first because it can end the problem in one comparison, and doing identity work on an undefined expression is pure waste.
The second step is the one that is easy to leave out entirely, and leaving it out is the standard error. Explicitly noting that the reciprocal identity is barred, before you write anything else, is what stops your hand from writing it out of habit.
The final step is deliberately phrased with two parts, the secant and the range. On a negative input the secant check alone always passes, because the wrong-convention angle has exactly the same secant as the right one. Only the range check distinguishes them, which is why it has to be written down rather than assumed.
Check
Check your understanding
Which of these expressions is undefined?
Answer: C
Why: Arcsecant accepts only numbers of size at least one, and three quarters fails that. If an angle had secant three quarters, its cosine would be four thirds, which is impossible. The domain condition is about size, never about sign.
This checks one thing only: whether you read the domain condition as a statement about size. Three of the four options have size at least one, and one does not.
If you were tempted by the very large negative number, notice that its size is one hundred and seventeen. Sign has no bearing on the domain of these two functions; the two conventions argue about what to return for negative inputs, never about whether to accept them.
The two endpoint options are both defined, and both are worth knowing cold. Arccosecant of one is pi over two under every convention. Arccosecant of minus one is minus pi over two or three pi over two depending on the book, which makes it a disagreement, not an exclusion.
Check
Check your understanding
Under the trigonometry-friendly convention, what is the arcsecant of minus two?
Answer: B
Why: Use the reciprocal identity: the arcsecant of minus two is the arccosine of minus one half, which is two pi over three. That angle is in the second quadrant, which is where this convention sends every negative input.
The question names the convention, which is your first clue about how to answer it: when a question bothers to state the convention, the input is going to be negative and the answer will differ between books.
Under this convention the reciprocal identity holds for every legal input, so the work is one line. The reciprocal of minus two is minus one half, and the angle in arccosine's range with that cosine is two pi over three.
Of the two wrong numerical options, the second is more interesting than the first. Four pi over three is a real angle with secant minus two; it is simply the other convention's answer. Being able to recognise a distractor as the right answer to a different question is a good sign that the distinction has landed.
Check
Check your understanding
A reference states that the tangent of the arcsecant of x equals the square root of x squared minus one for every x in the domain, with no case split. Which convention is that reference using, and how do you know?
Answer: B
Why: The calculus-friendly range is the first quadrant together with the third, and tangent is not negative in either, so one square root covers the whole domain. Removing that case split is precisely why the convention was invented.
This is the reading skill the lesson was built to give you, applied to a formula rather than to a graph. You are shown a result and asked to infer the range it must have come from.
Reason from the sign of the tangent. A single square root with no sign in front means the tangent is never negative on the range, and the only way for that to happen is for both pieces of the range to be quadrants where tangent is not negative. The first and third quadrants qualify; the second does not.
The last distractor is the one to make sure you have ruled out. The size of the tangent really does come from the identity and really is convention-independent, which makes the claim sound plausible. But the sign comes from the quadrant, the quadrant comes from the range, and the range is exactly what the convention chooses.
Explain it to yourself
Discussion prompt
In three or four sentences, explain to someone who has just met arccosine why arcsecant has two rival definitions and arccosine has none. Use the word range at least twice, and finish with the one sentence you would want printed at the top of every arcsecant problem.
Write the explanation out in full before you look at the sample answer. Being able to say something correctly is a much stricter test than recognising it, and this particular explanation is the one that compresses the whole lesson into four sentences.
Two things should appear in your version. First, the structural cause: one unbroken range means one forced restriction, a range in two pieces means a choice. Second, the fairness point: both choices give a genuine one-to-one function onto the full range, so this is a convention rather than a correctness question, and calling one of them wrong is a misunderstanding.
The closing sentence is not decoration. Get into the habit of writing a line like it at the top of any arcsecant or arccosecant problem with a negative input, in homework and on exams. It takes a few seconds, it tells the reader you know the issue exists, and it makes your answer checkable.
Exit ticket
\[ \operatorname{arcsec}(-\sqrt2) \]
Discussion prompt
Give the exact value under both conventions, say which one agrees with the arccosine of minus root two over two, and verify each answer by taking a secant.
Stitz & Zeager, College Trigonometry, Ch. 10 Foundations of Trigonometry — §10.6 The Inverse Trigonometric Functions §10.6, pp. 841-842 — Exercises 42 and 50
Do this one on paper before revealing, and do all three parts: both values, the comparison with the arccosine, and a verification of each.
The trigonometry-friendly value is the arccosine of minus root two over two, which is three pi over four, and it agrees with the arccosine because that convention keeps the reciprocal identity on the whole domain. The calculus-friendly value is the third-quadrant angle with the same secant, which is five pi over four. The two differ by a right angle, and both have secant minus root two.
Notice which exercises these are. The book asks for exactly this value twice, in two separate blocks of exercises, and each block states its assumption in the instructions. That is not padding; it is the book demonstrating the habit it wants you to adopt. If you can reproduce both answers and say which is which, you have the lesson.
Recap
| trigonometry friendly | calculus friendly | |
|---|---|---|
| range of arcsec | 0 to π/2, then π/2 to π | 0 to π/2, then π to 3π/2 |
| range of arccsc | −π/2 to 0, then 0 to π/2 | 0 to π/2, then π to 3π/2 |
| arcsec(x) = arccos(1/x) | for every legal x | only for x at least 1 |
| tan(arcsec x) | two cases, with a sign | one square root |
| is arccsc odd? | yes | no |
| derivative of arcsec | 1/(|x|√(x² − 1)) | 1/(x√(x² − 1)) |
\[ \text{domain, both: } \{x : |x| \ge 1\}; \quad \text{they agree for every } x \ge 1 \]
Next, Section 10.6.3 puts all six inverses on a calculator, finds the domain and range of transformed inverse functions, and uses them to solve equations whose values are not special angles.
OpenStax Algebra and Trigonometry 2e, §8.3 Inverse Trigonometric Functions §8.3 — the same material in OpenStax Algebra and Trigonometry
Read the table one row at a time and ask yourself, for each, whether you could have generated it from the two ranges alone. That is the test of whether you have learned the structure rather than six facts: every row below the first two is a consequence of the first two.
The single line under the table is the one to carry with you. The domain is identical, and the answers agree for every input of one and above. Everything contested in this lesson lives on the negative side of a domain that most problems never visit.
One last practical note. When you meet an arcsecant in a later course, look for the derivative in the reference you are using, or evaluate the function at minus two and see what comes back. Two seconds of checking, and you know which of these two tables applies for the rest of the problem.
Want this taught 1-on-1? Alexander tutors Trigonometry — $55/session, free consultation.