Intuition What this page is for
The parent note gave you the six formulas and how they're born. This page is the drill floor : we hunt down every kind of problem the topic can throw at you and solve one of each — signs, domains, chain-rule traps, degenerate inputs, blow-up points, a word problem, and an exam twist. By the end you should never meet a case you haven't already seen defeated.
Keep these two "engine" formulas in view; almost everything below is one of them plus the Chain Rule :
d x d arctan x = 1 + x 2 1 , d x d arcsin x = 1 − x 2 1 .
Common mistake The four traps we keep pointing at (named so you don't have to jump back)
Throughout this page we refer to four classic errors. Here they are, defined once:
Trap A — dropping the chain-rule factor. Writing d x d arctan ( 3 x ) = 1 + 9 x 2 1 instead of 1 + 9 x 2 3 . The formula is built for arctan ( x ) ; if the inside isn't just x you owe an extra u ′ .
Trap B — wrong sign on a "co-" function. Thinking d x d arccos x = + 1 − x 2 1 . arccos decreases , so its slope must be negative .
Trap C — dropping the ∣ x ∣ in arcsec/arccsc. Writing x x 2 − 1 1 . arcsec is increasing on both branches, so its derivative is positive everywhere; the ∣ x ∣ enforces that.
Trap D — ignoring the domain. Trying to compute d x d arcsin x at x = 2 , where 1 − x 2 isn't even real.
Before solving anything, let's list every distinct kind of situation this topic contains. Each row is a "cell". Every cell gets covered by at least one worked example below.
Cell
What makes it different
Example that hits it
C1 Plain formula, positive input
just quote the derivative, no chain
Ex 1
C2 Chain rule, linear inside
extra constant factor appears
Ex 2
C3 Chain rule, nonlinear inside
inside squared/shifted, watch the domain
Ex 3
C4 A "co-" function (sign trap)
must carry the minus sign
Ex 4
C5 arcsec/arccsc with $
x
$
C6 Negative input / sign of slope
check slope stays positive/negative as theory says
Ex 6
C7 Limiting / blow-up value (x → ± 1 )
derivative → ∞ , geometry of the edge
Ex 7
C8 Degenerate / out-of-domain input
recognise "no answer exists"
Ex 8
C9 Word problem (rate of change)
inverse trig models a real angle
Ex 9
C10 Exam twist (identity + product)
combine two rules, simplify cleverly
Ex 10
C11 arccot (the sixth function)
last of the six, its own minus sign
Ex 11
Let me plant one picture first so every "slope" claim below has something to point at.
Intuition Read the picture
The red curve is arcsin x . Notice it only lives between x = − 1 and x = 1 (dashed walls), it is always rising (slope always positive), and it gets vertical at the two edges — that is the 1 − x 2 → 0 in the denominator making the slope explode. Cells C6, C7, C8 are all just "which part of this red curve am I standing on?"
Worked example Differentiate
f ( x ) = arctan x and evaluate the slope at x = 1 .
Forecast: guess before reading — is the slope at x = 1 bigger or smaller than 1 ?
Step 1. Quote the engine formula. Why this step? No inside function other than x itself, so no Chain Rule correction is owed.
f ′ ( x ) = 1 + x 2 1 .
Step 2. Plug in x = 1 . Why this step? "Slope at a point" means evaluate the derivative there.
f ′ ( 1 ) = 1 + 1 2 1 = 2 1 .
Verify: arctan flattens out as x grows (it heads to π /2 ), so its slope should be less than 1 . We got 2 1 < 1 . ✓ Forecast confirmed.
Worked example Differentiate
g ( x ) = arcsin ( 5 x ) .
Forecast: will the answer be 1 − 25 x 2 1 or something with an extra number?
Step 1. Identify inside u = 5 x . Why this step? The formula d x d arcsin u = 1 − u 2 u ′ needs to know what u is and its derivative u ′ = 5 .
Step 2. Apply the formula, replacing x by u = 5 x inside. Why this step? The 1 − □ 2 always contains the whole inside thing squared.
g ′ ( x ) = 1 − ( 5 x ) 2 1 ⋅ 5 = 1 − 25 x 2 5 .
Verify: the domain is now ∣5 x ∣ < 1 , i.e. ∣ x ∣ < 5 1 — a narrower window than plain arcsin , which makes sense because 5 x reaches ± 1 five times faster. The extra factor 5 is the classic chain-rule term (Trap A ). ✓
Worked example Differentiate
h ( x ) = arctan ( x 2 + 1 ) .
Forecast: the inside is never 0 ; does the denominator 1 + ( inside ) 2 ever vanish?
Step 1. Inside u = x 2 + 1 , so u ′ = 2 x . Why this step? We must differentiate the inside separately — that's the whole Chain Rule .
Step 2. Slot into d x d arctan u = 1 + u 2 u ′ . Why this step? arctan 's formula has no square root, so no domain worry — only the fraction.
h ′ ( x ) = 1 + ( x 2 + 1 ) 2 2 x .
Step 3. Optionally expand ( x 2 + 1 ) 2 = x 4 + 2 x 2 + 1 , giving 1 + x 4 + 2 x 2 + 1 = x 4 + 2 x 2 + 2 . Why this step? Cleaner for later integration checks.
h ′ ( x ) = x 4 + 2 x 2 + 2 2 x .
Verify: the denominator x 4 + 2 x 2 + 2 is always ≥ 2 > 0 , so the derivative is defined for all real x — matching that arctan has domain all reals. At x = 0 , h ′ = 0 , and indeed arctan ( x 2 + 1 ) has a minimum there (inside bottoms out at x = 0 ). ✓
Worked example Differentiate
p ( x ) = arccos ( 2 x ) and find its slope at x = 0 .
Forecast: positive or negative slope? Remember arccos goes downhill .
Step 1. Recall d x d arccos u = − 1 − u 2 u ′ — note the minus . Why this step? arccos is a "co-" function; it decreases (bigger ratio → smaller angle), so its slope is negative everywhere (Trap B ).
Step 2. Inside u = 2 x , u ′ = 2 . Why this step? The chain rule supplies the factor u ′ = 2 on top.
p ′ ( x ) = − 1 − 4 x 2 2 .
Step 3. Evaluate at x = 0 . Why this step? "Slope at x = 0 " means substitute x = 0 into p ′ .
p ′ ( 0 ) = − 1 2 = − 2.
Verify: slope is negative , exactly as arccos's downhill shape demands. If you got + 2 , you dropped the co-minus. ✓
Worked example Differentiate
q ( x ) = arcsec ( x ) , evaluate at x = 2 and x = − 2 , then state and check the companion result for arccsc ( x ) .
Forecast: arcsec increases on both branches — so should the slope have the same sign at x = 2 and x = − 2 ? And do you expect arccsc's slope to be the opposite sign?
Step 1. Quote d x d arcsec x = ∣ x ∣ x 2 − 1 1 . Why the ∣ x ∣ ? Without it the algebra gives x x 2 − 1 1 , which would be negative on the left branch — but arcsec rises there too, so the derivative must stay positive . The ∣ x ∣ enforces that (Trap C ).
Step 2. At x = 2 substitute into the formula. Why this step? Direct evaluation of the derivative at the point.
q ′ ( 2 ) = ∣2∣ 4 − 1 1 = 2 3 1 = 6 3 .
Step 3. At x = − 2 substitute again. Why this step? To test the "same sign on both branches" claim, we must actually compute the left branch.
q ′ ( − 2 ) = ∣ − 2∣ ( − 2 ) 2 − 1 1 = 2 3 1 = 6 3 .
Step 4. The companion. Why this step? The cell is "arcsec /arccsc ", so we owe the csc case too. arccsc is the "co-" partner of arcsec, so by the parent's pattern its derivative is the negative of arcsec's:
d x d arccsc x = − ∣ x ∣ x 2 − 1 1 .
At x = 2 this gives − 6 3 , and at x = − 2 it also gives − 6 3 — a negative slope on both branches, because arccsc decreases everywhere.
Verify: arcsec came out equal and positive on both branches (+ 6 3 ≈ 0.2887 ), confirming why ∣ x ∣ (not x ) belongs there; arccsc came out equal and negative (− 6 3 ), the exact sign-flip the "co-flips the sign" rule predicts. ✓
Worked example Find the slope of
arcsin x at x = − 2 1 .
Forecast: the red curve in the figure rises everywhere — so the slope at a negative x should still be positive , right?
Step 1. Write down the derivative. Why this step? "Slope of arcsin x " is by definition its derivative, and we start from the engine formula.
f ′ ( x ) = 1 − x 2 1 .
Step 2. Substitute x = − 2 1 . Why this step? "Slope at x = − 2 1 " means evaluate the derivative there; we compute 1 − x 2 first so the square root is easy.
1 − x 2 = 1 − 4 1 = 4 3 ⟹ f ′ ( − 2 1 ) = 3/4 1 = 3 2 = 3 2 3 .
Verify: the answer is positive even though x < 0 — because x only appears squared in the formula, the sign of x can't flip the slope. That's the algebraic reason the red curve never turns downhill. Numerically 3 2 3 ≈ 1.1547 > 1 , and indeed the curve is steeper than 4 5 ∘ once you move away from the centre. ✓
Worked example What happens to
d x d arcsin x as x → 1 − ?
Forecast: finite number, or infinite?
Step 1. Look at the denominator 1 − x 2 as x → 1 − . Why this step? The whole behaviour of the slope is controlled by that square root shrinking.
1 − x 2 → 1 − 1 = 0 + ⟹ 1 − x 2 → 0 + .
Step 2. Divide a fixed numerator by that vanishing denominator. Why this step? A constant 1 over something heading to 0 + is exactly the recipe for a blow-up.
lim x → 1 − 1 − x 2 1 = + ∞.
Verify (geometry): in the picture below the red curve becomes vertical at the walls x = ± 1 . A vertical tangent = infinite slope. The [!recall] in the parent said it: "near a ratio of ± 1 the angle changes super fast." ✓
Intuition Why this figure matters
This second picture is the proof by eyeball of Ex 7. The black curve is the same arcsin x ; the red highlighted stretch near x = 1 shows the tangent line tilting up toward vertical . When a graph goes vertical, "rise over run" has a run heading to zero — that's the 1 − x 2 → 0 making the slope → + ∞ . Keep this image: it also explains why Ex 8 (just past the wall) has no slope at all — there's no curve out there to be tangent to.
d x d arcsin x at x = 2 .
Forecast: try to guess an answer — then check whether the point even exists.
Step 1. Test the domain first. Why this step? arcsin only accepts inputs with ∣ x ∣ ≤ 1 ; nothing else is defined (Inverse Functions and their Domains ).
Step 2. Substitute x = 2 into 1 − x 2 . Why this step? The formula's square root needs 1 − x 2 ≥ 0 ; checking its sign tells us whether a real answer can exist.
1 − x 2 = 1 − 4 = − 3 < 0 ⟹ 1 − x 2 = − 3 is not real .
Conclusion: the derivative does not exist at x = 2 — there's no point on the red curve there at all (it's outside the dashed walls). This is the correct answer , not a number (Trap D ).
Verify: the figures' red curve simply stops at x = 1 ; asking for its slope at x = 2 is like asking for the temperature of a colour. ✓
Worked example A person stands
10 m from a wall. A picture's bottom edge is at eye level; its top edge is at height x metres above eye level. The viewing angle to the top edge is θ = arctan ( 10 x ) . How fast does θ change (radians per metre) when x = 10 m ?
Forecast: at x = 10 the line of sight is at 4 5 ∘ ; do you expect the angle to change fast or slowly with height there?
Step 1. Differentiate θ with respect to x . Inside u = 10 x , so u ′ = 10 1 . Why this step? "How fast does θ change per metre of height" is exactly d x d θ .
d x d θ = 1 + ( 10 x ) 2 1 ⋅ 10 1 = 10 1 ⋅ 1 + 100 x 2 1 = 100 + x 2 10 .
Why the algebra? Multiply top and bottom by 100 to clear the inner fraction — cleaner to evaluate.
Step 2. Evaluate at x = 10 . Why this step? The question asks for the rate at that particular height , so we substitute x = 10 .
d x d θ x = 10 = 100 + 100 10 = 200 10 = 20 1 = 0.05 rad/m .
Verify (units + sense): units are radians per metre ✓. Value 0.05 rad/m ≈ 2. 9 ∘ per metre — a gentle change, which matches the picture: near 4 5 ∘ the triangle is "balanced" and adding height tips the angle only modestly. ✓
Worked example Differentiate
r ( x ) = x arctan x − 2 1 ln ( 1 + x 2 ) , and simplify.
Forecast: exam questions like this usually collapse to something tiny. Guess what r ′ ( x ) might be.
Step 1. Split r into two pieces and differentiate the first with the product rule . Why this step? x arctan x is a product of two functions of x (x and arctan x ), so d x d ( uv ) = u ′ v + u v ′ applies with u = x , v = arctan x .
d x d ( x arctan x ) = u ′ 1 ⋅ arctan x + x ⋅ v ′ 1 + x 2 1 = arctan x + 1 + x 2 x .
Why v ′ = 1 + x 2 1 ? That's just the engine formula for arctan .
Step 2. Differentiate the second piece − 2 1 ln ( 1 + x 2 ) with the logarithm–chain rule . Why this step? ln of something needs d x d ln ( w ) = w w ′ ; here the inside is w = 1 + x 2 with w ′ = 2 x .
d x d ( − 2 1 ln ( 1 + x 2 ) ) = − 2 1 ⋅ 1 + x 2 2 x = − 1 + x 2 x .
Why the − 2 1 survives out front? It's a constant multiplier, so it just rides along unchanged.
Step 3. Add the two derivatives and watch the cancellation. Why this step? r ′ is the sum of the pieces from Steps 1 and 2; the promised simplification is that the two 1 + x 2 x terms are equal and opposite.
r ′ ( x ) = ( arctan x + 1 + x 2 x ) + ( − 1 + x 2 x ) = arctan x + 1 + x 2 x − 1 + x 2 x = arctan x .
Verify: the extra fractions annihilate, leaving the clean r ′ ( x ) = arctan x . This is exactly why r ( x ) is the antiderivative of arctan x — a favourite Integration by Inverse Trig result read backwards. Sanity check at x = 0 : r ′ ( 0 ) = arctan 0 = 0 , and indeed the messy expression also gives 0 + 0 − 0 = 0 . ✓ Forecast (something tiny) confirmed.
Worked example Differentiate
s ( x ) = arccot ( 4 x ) and find its slope at x = 0 .
Forecast: arccot is the "co-" partner of arctan, so its base derivative carries a minus . Positive or negative slope at x = 0 ?
Step 1. Recall the sixth engine formula d x d arccot x = − 1 + x 2 1 . Why the minus? arccot decreases everywhere (bigger ratio → smaller angle), so its slope is negative — the "co-flips the sign" rule again (Trap B in disguise).
Step 2. Inside u = 4 x , u ′ = 4 . Why this step? The Chain Rule supplies the factor u ′ = 4 on top, and the 1 + □ 2 takes the whole inside squared.
s ′ ( x ) = − 1 + ( 4 x ) 2 1 ⋅ 4 = − 1 + 16 x 2 4 .
Step 3. Evaluate at x = 0 . Why this step? "Slope at x = 0 " means substitute x = 0 into s ′ .
s ′ ( 0 ) = − 1 + 0 4 = − 4.
Verify: the slope is negative (− 4 ), exactly as a decreasing "co-" function demands; if you'd used + 1 + x 2 1 you'd have got the wrong sign. Domain is all reals (no square root), matching that arccot accepts every x . ✓
Recall Which cell was which?
Plain formula ::: Ex 1 (C1)
Chain rule with a linear inside gives an extra constant factor ::: Ex 2 (C2)
Nonlinear inside, domain stays all reals for arctan ::: Ex 3 (C3)
"co-" function needs the minus sign ::: Ex 4 (C4)
arcsec has ∣ x ∣ , same positive slope on both branches; arccsc is its negative ::: Ex 5 (C5)
Slope of arcsin at a negative x is still positive (x appears squared) ::: Ex 6 (C6)
Derivative → + ∞ as x → 1 − (vertical tangent) ::: Ex 7 (C7)
Out-of-domain input has NO derivative, not a number ::: Ex 8 (C8)
Word problem: d x d θ = 100 + x 2 10 , equals 0.05 rad/m at x = 10 ::: Ex 9 (C9)
d x d ( x arctan x − 2 1 ln ( 1 + x 2 ) ) = arctan x ::: Ex 10 (C10)
d x d arccot ( 4 x ) = − 1 + 16 x 2 4 , slope − 4 at x = 0 ::: Ex 11 (C11)
Mnemonic Three questions before you differentiate
"Domain? Inside? Sign?" — (1) Is the input allowed (Trap D , C8)? (2) Is there an inside function needing the Chain Rule (Trap A , C2, C3, C11)? (3) Is it a "co-" function that owes a minus (Trap B , C4, C11)?
Multiply by inside derivative