Differentiability
Continuity only asks whether a graph is unbroken. Differentiability asks a sharper question: at this point, does the graph have one well-defined, non-vertical tangent direction? A function can be perfectly continuous and still fail this test — is the standard counterexample, drawn without lifting the pen yet turning a sharp corner at . This section builds the formal definition of the derivative as a limit of a difference quotient, splits it into left-hand and right-hand versions, and uses that split to catch exactly where smooth-looking piecewise functions secretly break. It is the second pillar of the Continuity–Differentiability–Differentiation unit in ISC Class 12 Mathematics, and every later differentiation rule — chain rule, implicit, parametric, logarithmic — silently assumes the function you are differentiating passes this test first.
A function is differentiable at a point only when its left-hand and right-hand derivative limits both exist, are finite, and agree — a strictly stronger condition than mere continuity.
The core idea
1. The derivative as a limit. A function is differentiable at if the following limit exists and is finite:
where: is the derivative (instantaneous rate of change) at , and is a small increment added to . The quotient inside the limit is the slope of the chord joining to ; as that chord slope settles down to the slope of the tangent line, provided it settles to the same value however approaches zero.
2. Splitting the limit in two directions. Because a two-sided limit exists only when both one-sided limits exist and agree, differentiability is checked via the left-hand derivative (LHD) and right-hand derivative (RHD):
where: lets approach through negative values (approaching from the left) and through positive values (approaching from the right). is differentiable at exactly when and both are finite; that common value is then .
3. Differentiability is stronger than continuity. Every differentiable function is automatically continuous, but not every continuous function is differentiable. Symbolically,
and the converse is false. This one-directional arrow is worth memorising as a shape: differentiability sits nested strictly inside continuity, so a suspicious point (usually created by a modulus, a greatest-integer function, or a piecewise join) must be checked for continuity and for matching one-sided derivatives — passing one test says nothing about the other.
4. Reading it geometrically. A differentiable point has a unique, non-vertical tangent line — no sharp corner, no cusp, no vertical tangent. At , is continuous (both one-sided limits of the function equal ) but not differentiable, because the graph's left branch has slope and its right branch has slope : two different tangent directions meeting at a corner, so no single tangent exists there. A cusp is a related but distinct failure — the graph again comes to a sharp point, but now both one-sided slopes race off to the same infinity (both or both ) rather than settling to two different finite values; a vertical tangent is different again, where the slope tends to from both sides. This syllabus concentrates on the plain-corner case seen with modulus and piecewise functions, where the two one-sided derivatives are finite but simply unequal.
5. Why the order of checking matters. Since differentiability is strictly the stronger condition, it can never be verified in isolation — you always reach it through continuity, never around it. This is why nearly every exam question phrased as "test whether is differentiable at " secretly expects two answers: first, is continuous at at all, and only then, do the one-sided derivatives agree. Skipping the continuity check does not usually produce a wrong numerical answer when the function happens to be continuous, but it is an incomplete argument, and examiners award marks for stating it explicitly.
The goal of this derivation is to turn the single two-sided limit that defines into a practical two-part test — compare the left-hand and right-hand derivatives at a suspicious point — which is the single most-tested skill in exam problems involving modulus, greatest-integer, or piecewise definitions, and is exactly the tool used in both worked examples below.
Step 1. Start from the definition . A two-sided limit exists if and only if both one-sided limits exist and are equal, so this single limit is really a claim about two limits at once: one where shrinks to through negative values, and one where it shrinks through positive values.
Step 2. Define the left-hand derivative , using whichever formula for is valid for inputs just below , and simplify the quotient algebraically (factor, cancel, rationalise) before letting — never substitute into the raw quotient, since it is typically an indeterminate form before simplification.
Step 3. Define the right-hand derivative the same way, using whichever formula for is valid for inputs just above .
Step 4. Compare and . If they are equal and finite, the original two-sided limit exists, so is differentiable at and equals that common value. If they disagree — as with at , where and — the two-sided limit fails to exist, so is not differentiable at , and the graph has a corner, cusp, or vertical tangent there.
Step 5. Because continuity is a prerequisite, it is standard practice to check continuity at first, before computing either one-sided derivative — if continuity already fails, differentiability is ruled out immediately and no derivative limits need be computed at all.
Worked example
Problem. Find the values of and so that for and for is differentiable at .
Step 1 — Continuity condition first. Since differentiability requires continuity, set the two one-sided limits, and , equal: and . Equating gives , labelled equation (i).
Step 2 — Left-hand derivative at . For , , so on that branch, giving .
Step 3 — Right-hand derivative at . For , , so everywhere on that branch, giving .
Step 4 — Equate the derivatives. Differentiability demands , so .
Step 5 — Solve for . Substituting into equation (i): , so .
Final answer. , and at that join the function is both continuous and differentiable, with .
Competitive stretch
Test at using the same four-step method, purely to see a case that fails. Continuity holds (). But while ; since , is continuous at yet not differentiable there — the vertex of the V-shaped graph.
Common traps
- Assuming continuity automatically guarantees differentiability — always check both one-sided derivatives separately near a modulus, greatest-integer, or piecewise join.
- Computing only or only and stopping — a valid conclusion needs both, and they must be finite and equal.
- Substituting directly into the raw difference quotient instead of simplifying algebraically first, which usually produces a false instead of the correct limit.
- In piecewise problems, matching only the derivatives while forgetting the continuity condition — an pair that satisfies derivative-matching but not function-value-matching is not a valid final answer.
Ask the tutor: want to see why actually is differentiable at despite involving , unlike alone — work through the difference-quotient limit side by side with the case.
Why does differentiability imply continuity but not the other way round?
If is differentiable at , then , and as this tends to , forcing — that is continuity. But continuity only says the function values line up at a point; it says nothing about the slopes approaching from each side lining up too, which is the extra condition differentiability adds. A corner like satisfies the first but not the second.
Do I always need to check continuity before checking differentiability?
Yes, and it is the efficient order to work in. Continuity is a necessary condition for differentiability, so if a function fails to be continuous at a point, you can immediately conclude it is not differentiable there without computing any derivative limits at all. In piecewise problems where you are solving for unknown constants, checking continuity first also usually gives you one equation (matching function values) before you bring in the second equation from matching derivatives.
Is every function built from nice pieces automatically differentiable at the join, since each piece is smooth on its own?
No — this is a very common trap. Each piece being individually differentiable (a polynomial, say) tells you nothing about what happens exactly at the seam where the pieces meet. You must explicitly compute using the left piece and using the right piece and compare them, as in the worked example. Two smooth curves can still meet at a sharp corner if their slopes disagree at the join.