Lesson 59 — Differentiability and Smoothness
Differentiable implies continuous. Corner points, cusps, vertical tangents. C^k and C^∞ classes. Weierstrass function.
Used in: 2nd year advanced HS (calculus) · Japanese Equiv. Math III · German Klasse 12 Leistungskurs Equiv.
Differentiability at : The function is differentiable at if and only if this limit exists as a finite real number. If it exists, its value is the derivative — the slope of the tangent line to the graph at the point .
Rigorous notation, full derivation, hypotheses
Definitions and Theorems
Differentiability at a Point
"If exists, we say that is differentiable at . If is differentiable at every number in an open interval , then is differentiable on ." — OpenStax Calculus Volume 1, §3.2
Fundamental Theorem (Differentiability Implies Continuity)
"If is differentiable at , then is continuous at ." — Active Calculus, §1.7, Boelkins 2024 (Theorem 1.7.1)
Types of Non-Differentiability Points
The four main types of non-differentiable points. From left: corner (finite, distinct one-sided derivatives), cusp (opposite infinite one-sided derivatives), vertical tangent (derivative = +∞ from both sides), jump (function not continuous).
| Type | Example at | What Occurs |
|---|---|---|
| Corner | ||
| Cusp | ||
| Vertical Tangent | ||
| Jump Discontinuity | is not continuous | |
| Non-removable Oscillation | limit of quotient does not exist |
Hierarchy
Example: but not (Cauchy)
This function is and for all , but . Thus — it definitively separates the smooth and analytic classes.
Weierstrass Function
Solved Examples
Exercise list
40 exercises · 10 with worked solution (25%)
- Ex. 59.1
Let . Calculate the one-sided derivatives and from the definition. Conclude about differentiability at .
Show solution
Right-hand derivative: . Left-hand derivative: . Since , the two-sided derivative does not exist at : corner, not differentiable.Show step-by-step (with the why)
See the referenced source for the step-by-step walkthrough. - Ex. 59.2Answer key
Let . Calculate and . Is differentiable at ?
Show solution
: the argument of is shifted: corner at . and . Corner at : not differentiable. - Ex. 59.3
Let . Determine using the definition.
Show solution
. The extra factor smooths out the corner of .Show step-by-step (with the why)
- Apply the definition: . Why: .
- Cancel (remember: in the limit): .
- . Therefore . Differentiable.
Note: The extra factor multiplies the corner of and "smooths" it — a general pattern for .
- Ex. 59.4
Let . Calculate from the definition. What does the answer indicate geometrically?
Show solution
. As or : (always positive). Vertical tangent — not differentiable. Geometrically: the tangent line becomes increasingly vertical as it approaches . - Ex. 59.5Answer key
Let . Analyze differentiability at by calculating the one-sided derivatives from the definition.
Show solution
. For : . For : . Cusp pointing upwards — one-sided derivatives go to opposite infinities. Not differentiable. - Ex. 59.6
Let , x^2, & x \geq 0, -x^2, & x < 0. Calculate and . Is differentiable at ?
Show solution
For : , difference quotient . For : , difference quotient . Equal one-sided derivatives: .Show step-by-step (with the why)
- . Why: the branch for is .
- . Why: the branch for is .
- One-sided derivatives are equal: . Differentiable.
Note: Despite the graph having a "smooth" shape for this function, always check the one-sided derivatives at points where a function is defined piecewise.
- Ex. 59.7
Let for and . Check if is continuous at and if it is differentiable at .
Show solution
: as , so is continuous at 0. Differentiability: . This limit does not exist (oscillates between -1 and 1). Not differentiable at 0. - Ex. 59.8
Let for and . Show that using the Squeeze Theorem.
Show solution
. Since (Squeeze Theorem), the limit is . Differentiable, . - Ex. 59.9
Let (ReLU function). Calculate and . Is differentiable at ?
Show solution
: for , ; for , . Right-hand derivative: . Left-hand derivative: . Different: corner, not differentiable. - Ex. 59.10
Let . At what point does have a corner? Verify by calculating the one-sided derivatives at that point.
Show solution
: for , ; for , . At : and . Corner at . - Ex. 59.11
Let for and . Is it continuous at ? Is it differentiable at ?
Show solution
: and . Jump from -1 to 1 at : discontinuity. Thus not differentiable. - Ex. 59.12
Where is the floor function differentiable? Where is it not? Justify in each case.
Show solution
At integers, has jumps: discontinuity, thus not differentiable at integers. Outside integers, is locally constant, so at those points. Differentiable on . - Ex. 59.13
Let . Calculate and using the limit definition.
Show solution
For : difference quotient at 0 is . For : . Thus . For : . Applying the definition of : . Thus . - Ex. 59.14Answer key
Let if and if . Determine if is differentiable at .
Show solution
The difference quotient at 0: . If : . If : . In both cases, the quotient goes to 0. Therefore . Differentiable at 0. - Ex. 59.15
Let . Calculate from the definition. Identify the type of non-differentiable point.
Show solution
. For : . Vertical tangent. Not differentiable. - Ex. 59.16
Let for and . Show that is differentiable at , but that . What is the maximum class of ?
Show solution
by definition (Squeeze Theorem, since ). For : . The term oscillates without limit as . Therefore and : differentiable but not .Show step-by-step (with the why)
- . Why: definition applied with .
- Squeeze Theorem: , and . Thus .
- For , product and chain rules: .
- does not exist — oscillates between -1 and 1. Thus .
Curiosity: This is the canonical example of differentiable-but-not-: a gap exists between the classes.
- Ex. 59.17
Find and such that , ax + b, & x \leq 1, x^2, & x > 1 is on .
Show solution
Continuity at 1: , so (I). Derivatives: and . Equating: (II). From (I): .Show step-by-step (with the why)
- Continuity at : , i.e., . Why: the two branches must match at the junction point.
- Left-hand derivative: , so .
- Right-hand derivative: , so .
- For : . Substituting: .
Mental shortcut: For piecewise linear+polynomial functions, the tangent line equation to the curve at the junction point gives the linear branch directly.
- Ex. 59.18
Find such that , cx + d, & x \leq 1, 3x^2 - 2, & x > 1 is at .
Show solution
Continuity at 1: . Derivatives: . So . - Ex. 59.19Answer key
Find such that , ax + b, & x \leq 0, \sin x, & x > 0 is at .
Show solution
Continuity at 0: . Derivatives: . Check: gives for , which is the tangent line to at . - Ex. 59.20
Where is not differentiable? Identify the type of point.
Show solution
: translation of to . At : . Vertical tangent at : not differentiable. - Ex. 59.21Answer key
Let , x^2, & x < 0, \sin x, & x \geq 0. Is at ? Is it at ?
Show solution
Continuity at 0: and . Continuous. Derivatives: and . Different: but not . Corner at 0. - Ex. 59.22
A cubic polynomial belongs to which class? Why is the answer not ?
Show solution
Every polynomial $p(x) = a_n x^n + \ldots + a_0$ is $C^\infty$: differentiating $n$ times gives a constant; beyond that, the derivative is zero — both are continuous. All derivatives exist and are continuous. - Ex. 59.23
Where does have corners? Calculate the one-sided derivatives at each point to confirm.
Show solution
: corner where , i.e., . At : and . Different: corner at . By symmetry, corner at as well. - Ex. 59.24Answer key
A cubic spline is formed by on and on . What conditions at guarantee the combined spline is ? List all equations.
Show solution
For a cubic spline to be : (1) (continuity of function), (2) (continuity of first derivative), (3) (continuity of second derivative). Three compatibility equations at the node . - Ex. 59.25Answer key
Let . At which points does have corners? Sketch the argument for the points .
Show solution
has corners where : at , . At : one-sided derivatives are and — opposites. Corners at . - Ex. 59.26
Let . Calculate for all and show that .
Show solution
. For : ... Actually, using the product rule for : ; for : . In both cases . Since , is continuous and . - Ex. 59.27
Analyze the differentiability of on all of . At which points does have corners?
Show solution
has corners at where changes sign (crosses zero). At all other points, and the function is locally , which is . Therefore: continuous everywhere (), but not differentiable at . - Ex. 59.28
Let . What is the regularity class of on ? Justify.
Show solution
Each term is . A finite sum of functions is . Therefore . It is not "just near 0" — it is on all of . - Ex. 59.29
The payoff of a European call option at expiration is . (a) Identify the non-differentiable point. (b) Calculate and . (c) What happens to the Greek Delta at this point?
Show solution
Call payoff: . Corner at : and . The Greek Delta jumps from 0 to 1 at . Not differentiable at . - Ex. 59.30
In machine learning, the ReLU activation function is . Why does the SGD algorithm work even though ReLU is not differentiable at ?
Show solution
ReLU is not differentiable at 0, but the set has Lebesgue measure zero. SGD uses subgradients (typically 0 or 1 at 0 by implementation convention). In practice, the stochastic gradient descent algorithm rarely hits exactly 0 with floating-point precision; and when it does, any subgradient from the interval suffices to ensure convergence. - Ex. 59.31Answer key
In structural engineering, a knotted elastic cable has continuous displacement but a jump in slope at the knot. (a) What is the regularity class of ? (b) What does the corner in the graph of represent physically?
Show solution
Cable with a knot: displacement is continuous (no physical tearing — no point on the cable breaks), but (proportional to internal tension) has a jump at the knot. Thus but . The corner in the displacement graph corresponds to the jump in tension. - Ex. 59.32
A natural cubic spline on with a node at imposes what regularity conditions? What is the resulting class? Why and not ?
Show solution
Natural cubic spline imposes continuity of , , and at interior nodes. At the boundary nodes, (natural spline condition — "no curvature at the ends"). Thus on the entire interval. - Ex. 59.33
For a wave equation with a discontinuous initial condition (step function), what regularity is expected for the solution ? Why does a solution not exist?
Show solution
A discontinuous initial condition (step function) for the wave equation: the solution is but has derivative discontinuities propagating along characteristics. The classical solution (which requires ) does not exist; a weak formulation (distributional solution) is needed. - Ex. 59.34
Is the converse of "differentiable continuous" true? What is the simplest counterexample?
Show solution
The converse is false. The standard counterexample: $f(x) = |x|$ is continuous everywhere on but not differentiable at $x = 0$. In fact, the Weierstrass function is continuous everywhere on and is not differentiable at *any* point. - Ex. 59.35
Let . What is the maximum class of ? Calculate for all to justify.
Show solution
. For : . For : . In both cases: . . Therefore is continuous and . - Ex. 59.36
The Cantor function satisfies: continuous on , , , and almost everywhere. Why does the Fundamental Theorem of Calculus not apply?
Show solution
The Cantor function is continuous, non-decreasing, , , and almost everywhere (on the set complement of the Cantor set, measure 1). FTC fails because . FTC with Riemann integral requires absolute continuity of — the Cantor function is not absolutely continuous. - Ex. 59.37
Does a continuous function exist that is not differentiable at any point? Describe the main construction.
Show solution
Yes. The Weierstrass function (with and ) is continuous on all of (uniformly convergent series, dominated by geometric series ) and not differentiable at any point. Weierstrass, 1872. Hardy improved the criterion to in 1916. - Ex. 59.38Answer key
Let for and . Show that and that for all . What does this imply about the Taylor series of at ?
Show solution
See the referenced source for the detailed solution. - Ex. 59.39
Prove: if is differentiable at , then is continuous at .
Show solution
Let be differentiable at . For , write . As , the first factor tends to (finite by hypothesis) and the second tends to . Thus the product tends to , proving . Therefore is continuous at .Show step-by-step (with the why)
- Algebraic identity for : . Why: multiplying and dividing by is valid when .
- By hypothesis of differentiability: , which is finite.
- Trivially: .
- Limit of product = product of limits: .
- Therefore : definition of continuity at .
Note: The key is factoring the increment — a simple algebraic trick with profound consequences. Memorize this proof; it reappears in multivariable calculus.
- Ex. 59.40Answer key
Prove that if , then is Lipschitz on . (Hint: use the Mean Value Theorem and the fact that is bounded on a compact set.)
Show solution
If , then is continuous on the compact set , hence bounded: such that for all . By the Mean Value Theorem: for any , there exists between them with . Therefore . Thus is Lipschitz with constant .Show step-by-step (with the why)
- implies is continuous on a compact set. Why: compact + continuous implies bounded (Weierstrass Theorem for functions).
- such that for all .
- By MVT: for some between and .
- . Therefore Lipschitz.
Mnemonic: implies Lipschitz implies — the full hierarchy flowing downwards.
Sources
- Active Calculus — Boelkins, 2024 · §1.7 "Limits, Continuity, and Differentiability" · CC-BY-NC-SA. Primary source.
- Calculus Volume 1 — OpenStax, 2016 · §3.2 "The Derivative as a Function" · CC-BY-NC-SA.
- APEX Calculus — Hartman et al., 2024 · v5 · §2.1 "Instantaneous Rates of Change: The Derivative" · CC-BY-NC.