Intuition Why this page exists
The parent note taught the machine : touch a y , pay the toll y ′ ; then collect, factor, divide. But a machine is only as trustworthy as the number of situations you've watched it survive. Here we deliberately hunt down every awkward case — the point where the tangent goes vertical (denominator = 0 ), the point where the curve crosses itself , a spot where y ′ comes out undefined , a moving-in-time word problem, and an exam trap. If you've seen the technique bend without breaking in all of these, nothing on a test can surprise you.
We reuse only what the parent built: Chain Rule (the toll), Product Rule (for mixed x y ), and the recipe differentiate → toll every y → collect → factor → divide .
Definition Naming the whole equation
F ( x , y )
Throughout this page, when an implicit equation is written as "left side = right side," we quietly move everything to one side and call that combined expression F ( x , y ) . For a circle, x 2 + y 2 = 25 becomes F ( x , y ) = x 2 + y 2 − 25 . Then:
F x means "differentiate F treating only x as the variable, y frozen" — the partial derivative in x .
F y means the same with only y moving, x frozen.
These two numbers are exactly what the parent's shortcut d x d y = − F y F x uses. We only need them below to diagnose special points: where F y = 0 the denominator dies (vertical tangent), and where both F x = F y = 0 we get 0/0 (a crossing). That is precisely the condition the Implicit Function Theorem flags.
Every implicit-differentiation problem lands in one of these boxes. The whole page is a tour that hits each row at least once .
#
Case class
What makes it tricky
Covered by
A
Clean slope at a normal point
nothing — warm-up
Ex 1
B
Sign / quadrant sensitivity
same x , different y → opposite slopes
Ex 2
C
Denominator = 0 (vertical tangent )
y ′ blows up; geometry, not error
Ex 3
D
Self-crossing / degenerate point
two slopes at one point
Ex 4
E
Product and chain together
x y inside a transcendental
Ex 5
F
Second derivative y ′′
must substitute y ′ back
Ex 6
G
Limiting / horizontal tangent (y ′ = 0 )
numerator = 0
Ex 7
H
Real-world related rate (time t )
both vars depend on t
Ex 8
I
Exam twist — solve for a parameter
reverse the question
Ex 9
9 x 2 + 4 y 2 = 1 at ( a , b )
Statement. Find d x d y on this ellipse, then its value at ( 2 3 , 2 ) .
Forecast: This point is in the first quadrant (both coordinates positive) and on a downward-sloping part of the ellipse — so guess a negative slope.
Differentiate both sides w.r.t. x :
9 2 x + 4 2 y d x d y = 0.
Why this step? The constant 1 on the right differentiates to 0 ; the y 2 term hides a y , so the chain rule attaches d x d y .
Collect / divide:
d x d y = − 2 y /4 2 x /9 = − 9 y 4 x .
Why this step? Only one y ′ term exists, so "collect–factor–divide" is a single division.
Plug in x = 2 3 , y = 2 :
d x d y = − 9 ⋅ 2 4 ⋅ 2 3 = − 9 2 12/ 2 = − 9 ⋅ 2 12 = − 3 2 .
Verify: slope is negative ✓ (matches the forecast). Sanity check the point is on the curve: 9 ( 3/ 2 ) 2 + 4 ( 2 ) 2 = 9 9/2 + 4 2 = 2 1 + 2 1 = 1 ✓.
x 2 + y 2 = 25 at the two points with x = 3
Statement. The vertical line x = 3 pierces the circle at y = + 4 and y = − 4 . Find the slope at each .
Forecast: By symmetry the two slopes should be equal and opposite .
From the parent note, d x d y = − y x .
Why this step? 2 x + 2 y y ′ = 0 ⇒ y ′ = − x / y ; we reuse it rather than re-derive.
Upper point ( 3 , 4 ) : y ′ = − 4 3 .
Lower point ( 3 , − 4 ) : y ′ = − − 4 3 = + 4 3 .
Why this step? The sign of y flips the whole ratio — this is exactly why implicit differentiation covers both branches at once : one formula, two answers.
Look at the figure below. The blue curve is the circle; the dashed white line is x = 3 . It meets the circle at the yellow dot ( 3 , 4 ) (top branch) and the red dot ( 3 , − 4 ) (bottom branch). The short yellow line has slope − 4 3 (falling to the right), the red line slope + 4 3 (rising) — mirror images across the x -axis.
Verify: at ( 3 , 4 ) the curve slopes down-right (top of circle, heading toward the right edge) → negative ✓; at ( 3 , − 4 ) it slopes up-right → positive ✓, and the magnitudes match ∣ − 3/4 ∣ = ∣3/4∣ ✓.
Common mistake "The circle only has one slope at
x = 3 "
Why it feels right: we're used to y = f ( x ) giving one output per x .
Fix: an implicit curve can stack several y -values above one x . You must name the point by both coordinates before asking for the slope.
x 2 + y 2 = 25 at ( 5 , 0 ) and ( − 5 , 0 )
Statement. Find the slope where the circle meets the x -axis.
Forecast: Picture the far-right point ( 5 , 0 ) : the circle is standing straight up there. A vertical line has no finite slope . So expect the formula to warn us.
Use d x d y = − y x . At ( 5 , 0 ) : denominator y = 0 .
Why this step? Substitution before thinking would give − 5/0 — undefined.
Interpret, don't panic: d x d y → ± ∞ means the tangent line is vertical (x = 5 ).
Why this step? A blown-up slope is geometry: Tangent and Normal Lines tells us the tangent is x = 5 and, usefully, the normal here is horizontal (y = 0 , slope 0 ).
Same at ( − 5 , 0 ) : vertical tangent x = − 5 .
Look at the figure below. The red solid line x = 5 hugs the circle's right edge — that's the vertical tangent at the yellow dot ( 5 , 0 ) ; the dashed red line is its mirror at x = − 5 . Notice the tangent touches but doesn't cross, exactly as a tangent should.
Verify: using the diagnostic from our definition, F ( x , y ) = x 2 + y 2 − 25 gives F y = 2 y , which is 0 exactly at y = 0 . That is where the denominator of − F x / F y dies, so the tangent is vertical — and it's precisely the spot the Implicit Function Theorem says y ( x ) fails to be a nice single-valued function. Matches the picture. ✓
Common mistake Writing "slope
= undefined, so no tangent"
Fix: there is a tangent — it's the vertical line. "Undefined slope" ≠ "no tangent"; it means infinite slope.
Worked example Lemniscate
( x 2 + y 2 ) 2 = 2 ( x 2 − y 2 ) — the crossing at the origin and the loop tips
Statement. This figure-eight passes through the origin , where it crosses itself. What are the slopes there? Also find the tangent behaviour at the loop tips on the x -axis.
Forecast: At a self-crossing the curve has two tangent directions. With F ( x , y ) = ( x 2 + y 2 ) 2 − 2 ( x 2 − y 2 ) , both F x and F y will vanish at the origin — the recipe gives 0/0 , the signal of a special point.
Differentiate ( x 2 + y 2 ) 2 = 2 ( x 2 − y 2 ) :
2 ( x 2 + y 2 ) ( 2 x + 2 y y ′ ) = 2 ( 2 x − 2 y y ′ ) .
Why this step? The left side is an outer square (chain rule) times the inner derivative 2 x + 2 y y ′ ; every y pays its toll.
Distribute both sides so each y ′ term is visible:
4 x ( x 2 + y 2 ) + 4 y ( x 2 + y 2 ) y ′ = 4 x − 4 y y ′ .
Why this step? Before you can collect y ′ terms you must expand the brackets so they're not hidden inside a product — otherwise you'd factor the wrong things.
Collect every y ′ term on the left, everything else on the right, then factor:
4 y ( x 2 + y 2 ) y ′ + 4 y y ′ = 4 x − 4 x ( x 2 + y 2 )
y ′ [ 4 y ( x 2 + y 2 ) + 4 y ] = 4 x [ 1 − ( x 2 + y 2 ) ] .
Why this step? This is the C ollect–F actor half of the recipe: gather the toll terms, pull y ′ out as a common factor.
Divide (the 4 's cancel):
d x d y = y [ ( x 2 + y 2 ) + 1 ] x [ 1 − ( x 2 + y 2 ) ] .
Why this step? D ivide — the final recipe step — leaves y ′ alone.
At the origin ( 0 , 0 ) : numerator = 0 , denominator = 0 → 0 0 .
Why this step? 0/0 is the algebraic fingerprint of a crossing or cusp — the single formula can't pick one slope because there are two distinct tangent directions.
Loop tips. On the x -axis set y = 0 in the original equation: ( x 2 ) 2 = 2 x 2 ⇒ x 4 = 2 x 2 ⇒ x 2 = 2 ⇒ x = ± 2 . So the tips are ( ± 2 , 0 ) . Plug ( 2 , 0 ) into y ′ : numerator = 2 [ 1 − 2 ] = − 2 = 0 , but denominator = 0 ⋅ [ … ] = 0 → d x d y → ± ∞ , a vertical tangent at each tip.
Why this step? This distinguishes the two special kinds: at the tips only F y = 0 (vertical tangent, a clean number /0 ), while at the origin both vanish (0/0 , the crossing).
Look at the figure below. The blue figure-eight crosses itself at the red dot (origin) — you can see two curve branches passing through with different directions, which is why the slope there is 0/0 . The two yellow dots at ( ± 2 , 0 ) are the loop tips, where the curve turns straight up/down: vertical tangents.
Verify: the origin is on the curve: ( 0 ) 2 = 2 ( 0 ) ✓. Both F x = 4 x ( x 2 + y 2 ) − 4 x and F y = 4 y ( x 2 + y 2 ) + 4 y vanish at ( 0 , 0 ) ✓ — the 0/0 is genuine, not sloppy algebra. The tips satisfy x 4 = 2 x 2 at x = 2 ✓, with denominator y [( x 2 + y 2 ) + 1 ] = 0 there → vertical ✓.
e x y = x + y at ( 0 , 1 )
Statement. Find d x d y at the point ( 0 , 1 ) .
Forecast: e x y contains a product x y inside an exponential — expect a Chain Rule toll on the e and a Product Rule on x y , stacked.
Differentiate:
e x y ⋅ d x d ( x y ) = 1 + d x d y .
Why this step? d x d e u = e u u ′ (chain rule); the exponent u = x y is a product.
Expand the inner product d x d ( x y ) = y + x y ′ :
e x y ( y + x y ′ ) = 1 + y ′ .
Why this step? Product Rule on x y : derivative of x (which is 1 ) times y , plus x times the toll y ′ .
Distribute the e x y across the bracket so each y ′ term stands alone:
y e x y + x e x y y ′ = 1 + y ′ .
Why this step? You cannot collect y ′ terms while one of them is trapped inside a bracket multiplied by e x y — expand first.
Collect y ′ on one side, factor, divide:
x e x y y ′ − y ′ = 1 − y e x y ⇒ y ′ ( x e x y − 1 ) = 1 − y e x y
d x d y = x e x y − 1 1 − y e x y .
Why this step? The C ollect–F actor–D ivide finish: gather the two y ′ terms, pull y ′ out, divide.
At ( 0 , 1 ) : x y = 0 so e x y = 1 . Then
d x d y = 0 ⋅ 1 − 1 1 − 1 ⋅ 1 = − 1 0 = 0.
Verify: check the point is on the curve: e 0 = 1 and x + y = 0 + 1 = 1 ✓. Slope 0 means a horizontal tangent at ( 0 , 1 ) — a legitimate answer (this is a preview of Case G).
x 3 + y 3 = 6 x y — find y ′′ at ( 3 , 3 )
Statement. The parent found y ′ = y 2 − 2 x 2 y − x 2 and y ′ ( 3 , 3 ) = − 1 . Now find y ′′ there.
Forecast: We'll differentiate y ′ again; the danger is leaving a stray y ′ in the answer. Since y ′ ( 3 , 3 ) = − 1 , expect a clean number.
Start from the pre-solved line 3 x 2 + 3 y 2 y ′ = 6 y + 6 x y ′ . Differentiate again w.r.t. x :
6 x + 3 ( 2 y y ′ ⋅ y ′ + y 2 y ′′ ) = 6 y ′ + 6 ( y ′ + x y ′′ ) .
Why this step? d x d ( y 2 y ′ ) is a product: 2 y ( y ′ ) 2 + y 2 y ′′ . Every y still pays a toll.
Simplify the constants and gather like terms:
6 x + 6 y ( y ′ ) 2 + 3 y 2 y ′′ = 12 y ′ + 6 x y ′′ .
Why this step? We only multiplied out the 3 ( … ) and 6 ( … ) brackets — no new calculus, just arithmetic tidy-up so the y ′′ terms become visible and countable.
Collect the two y ′′ terms on the left, everything else on the right, then factor y ′′ out:
y ′′ ( 3 y 2 − 6 x ) = 12 y ′ − 6 x − 6 y ( y ′ ) 2
y ′′ = 3 y 2 − 6 x 12 y ′ − 6 x − 6 y ( y ′ ) 2 .
Why this step? Same C ollect–F actor–D ivide finish as always, but now the "unknown" we solve for is y ′′ instead of y ′ ; the y ′ terms are just known coefficients.
Substitute x = 3 , y = 3 , y ′ = − 1 :
y ′′ = 3 ( 9 ) − 6 ( 3 ) 12 ( − 1 ) − 6 ( 3 ) − 6 ( 3 ) ( 1 ) = 27 − 18 − 12 − 18 − 18 = 9 − 48 = − 3 16 .
Why this step? This is the classic trap — you must plug in the value of y ′ , not leave the symbol.
Verify: y ′′ = − 3 16 < 0 → the curve is concave down at ( 3 , 3 ) , consistent with a tangent of slope − 1 sitting at a local high part of the loop. Concrete arithmetic check: numerator = 12 ( − 1 ) − 6 ( 3 ) − 6 ( 3 ) ( − 1 ) 2 = − 12 − 18 − 18 = − 48 ; denominator = 3 ( 9 ) − 6 ( 3 ) = 27 − 18 = 9 ; so y ′′ = − 48/9 = − 16/3 ≈ − 5.33 ✓ (also machine-checked in VERIFY). ✓
Worked example Where is the folium's tangent horizontal? Solve
x 3 + y 3 = 6 x y for the point.
Statement. Find the point in the first quadrant where d x d y = 0 .
Forecast: y ′ = 0 needs the numerator 2 y − x 2 = 0 , i.e. y = 2 x 2 . Feed that back into the curve.
Set numerator zero: y = 2 x 2 .
Why this step? A fraction is 0 only when its top is 0 (and bottom isn't).
Substitute into x 3 + y 3 = 6 x y :
x 3 + ( 2 x 2 ) 3 = 6 x ⋅ 2 x 2 ⇒ x 3 + 8 x 6 = 3 x 3 .
Why this step? Replacing y by 2 x 2 turns the two-variable curve into one equation in x alone, which we can actually solve.
Move all terms to one side and simplify. Subtract x 3 from both sides:
8 x 6 = 2 x 3 .
Why this step? 3 x 3 − x 3 = 2 x 3 ; grouping the x 3 terms isolates the higher power for the next step.
Since we want the first-quadrant point, x = 0 , so divide both sides by x 3 :
8 x 3 = 2 ⇒ x 3 = 16 ⇒ x = 1 6 1/3 = 2 4/3 .
Why this step? Dividing by x 3 is legal because x = 0 (the origin isn't the point we want); it drops the polynomial from degree 6 to a clean cubic.
Back-substitute to get y :
y = 2 x 2 = 2 ( 2 4/3 ) 2 = 2 1 2 8/3 = 2 8/3 − 1 = 2 5/3 .
Why this step? The horizontal-tangent point needs both coordinates; y comes straight from the condition y = 2 x 2 we imposed in step 1.
So the tangent is horizontal at ( 2 4/3 , 2 5/3 ) ≈ ( 2.52 , 3.17 ) .
Verify: denominator y 2 − 2 x = 2 10/3 − 2 ⋅ 2 4/3 = 2 10/3 − 2 7/3 = 0 ✓ (a genuine horizontal tangent, not 0/0 ). Numerator 2 y − x 2 = 2 ⋅ 2 5/3 − 2 8/3 = 2 8/3 − 2 8/3 = 0 ✓. On-curve check: x 3 + y 3 = 16 + 2 5 = 16 + 32 = 48 and 6 x y = 6 ⋅ 2 4/3 ⋅ 2 5/3 = 6 ⋅ 2 3 = 48 ✓.
Worked example The sliding ladder
Statement. A 10 m ladder leans on a wall. Its foot slides out at d t d x = 1 m/s . When the foot is x = 6 m from the wall, how fast is the top sliding down ?
Forecast: As the base rushes out, the top drops — expect d t d y < 0 . Near the wall the top barely moves; far out it plummets. At x = 6 (so y = 8 , a 6 -8 -10 triangle) expect a moderate value.
Constraint: x 2 + y 2 = 1 0 2 = 100 (Pythagoras — the ladder length is fixed).
Why this step? This is our implicit equation; x and y are both functions of t , so we differentiate w.r.t. t , a Related Rates move.
Differentiate w.r.t. t :
2 x d t d x + 2 y d t d y = 0 ⇒ d t d y = − y x d t d x .
Why this step? Same toll as before, but the "outer variable" is t , so each term gets d t d .
At x = 6 : y = 100 − 36 = 8 . So
d t d y = − 8 6 ( 1 ) = − 4 3 = − 0.75 m/s .
Look at the figure below. The blue segment is the ladder, fixed at 10 m . The green dot is the foot moving right (green arrow, + 1 m/s ); the red dot is the top moving down (red arrow, − 0.75 m/s ). The right triangle has legs x = 6 and y = 8 .
Verify: sign is negative (top descends) ✓. Units: m m ⋅ s m = s m ✓. Limiting check: as x → 10 , y → 0 , d t d y = − y x → − ∞ — the top slams down infinitely fast at the instant the ladder goes flat, matching physical intuition. ✓
Worked example Reverse-engineer a constant
Statement. A curve x 2 + k x y + y 2 = c is said to have a horizontal tangent at the point ( 1 , 2 ) . Find the constant k (and check whether the stated curve is even consistent).
Forecast: "Horizontal tangent at ( 1 , 2 ) " means y ′ = 0 there; that's one equation for k . The point must also lie on the curve; that's a second equation that pins down c . Watch for a trap where the two facts don't agree.
Differentiate x 2 + k x y + y 2 = c implicitly:
2 x + k ( y + x y ′ ) + 2 y y ′ = 0.
Why this step? k x y is a product of x and y times constant k → product rule; the y 2 pays its toll; c is constant so its derivative is 0 .
Impose y ′ = 0 at ( 1 , 2 ) :
2 ( 1 ) + k ( 2 + ( 1 ) ( 0 ) ) + 2 ( 2 ) ( 0 ) = 0 ⇒ 2 + 2 k = 0 ⇒ k = − 1.
Why this step? Setting y ′ = 0 kills every y ′ term, leaving a single linear equation in k .
Now find the consistent c . Put k = − 1 and the point ( 1 , 2 ) into the curve:
1 2 + ( − 1 ) ( 1 ) ( 2 ) + 2 2 = 1 − 2 + 4 = 3.
So the point lies on the curve only if c = 3 .
Why this step? A parameter problem has two unknowns hiding (k and the constant); the slope condition gives one, the on-curve condition gives the other.
Conclusion / resolution. With k = − 1 and c = 3 , the curve x 2 − x y + y 2 = 3 passes through ( 1 , 2 ) and has a horizontal tangent there — everything is consistent, so the answer is k = − 1 . Exam trap: if the problem had fixed the right-hand side at, say, c = 7 , then no value of k could satisfy both conditions at once (the point simply isn't on x 2 + k x y + y 2 = 7 for the k that flattens the tangent), and the correct exam answer would be "no such k exists — the data are inconsistent. " Always test the on-curve condition before reporting k .
Verify: with k = − 1 , curve value at ( 1 , 2 ) is 3 ✓. Slope there: 2 + ( − 1 ) ( 2 ) + [( − 1 ) ( 1 ) + 2 ( 2 )] y ′ = 0 ⇒ 0 + 3 y ′ = 0 ⇒ y ′ = 0 ✓ — genuinely horizontal.
Recall One-line summary of the matrix
Case ::: Signal / fix
A clean point ::: just divide
B two y 's over one x ::: name the point by both coords; sign of y flips slope
C denominator = 0 (only F y = 0 ) ::: vertical tangent, not an error
D numerator and denominator = 0 (0/0 , both F x = F y = 0 ) ::: self-crossing/cusp — two slopes
F second derivative ::: substitute y ′ back before you finish
G numerator = 0 ::: horizontal tangent, solve top = 0
H related rate ::: differentiate w.r.t. t ; check units and limits
Mnemonic Read the fraction, not just the number
After you get y ′ = bottom top : top = 0 → flat, bottom = 0 → vertical, both = 0 → crossing. Every special case is written on the face of the fraction.
Chain Rule — the toll paid in every example.
Product Rule — Ex 5, Ex 9 (x y terms).
Related Rates — Ex 8's sliding ladder.
Tangent and Normal Lines — Ex 1–3, 7 read off slopes.
Implicit Function Theorem — explains why Cases C & D are where F y = 0 (and where both F x = F y = 0 ).
Derivatives of Inverse Functions · Logarithmic Differentiation — sibling applications of the same machine.