Continuity

Motivation: why this matters in quant finance

Continuity is the property that says a function has "no gaps" — small changes in input produce small changes in output. In quant finance, continuity is both a modelling assumption and a mathematical requirement:

Brownian motion paths are continuous. Axiom (BM4) of Brownian motion states that almost every sample path $t \mapsto W_t(\omega)$ is continuous. This assumption is what makes continuous-time hedging possible: if the stock price moves continuously, you can adjust your hedge continuously without gaps. The entire Black-Scholes derivation rests on this — if the stock could jump, the $dW_t$ term cannot be perfectly cancelled, and the hedged portfolio is no longer locally riskless.
Pricing functions should be continuous. If the option price $v(S, t)$ were discontinuous in $S$ , a tiny move in the underlying would cause a finite jump in the option value with no opportunity to hedge. Continuity of $v$ is a necessary (though not sufficient) condition for the Greeks to be well-defined and for hedging to work.
Continuity is required for differentiability. A function must be continuous at a point before it can be differentiable there. Since the entire differentiation framework (and hence the Greeks) depends on derivatives existing, continuity is the prerequisite. Importantly, the converse is false: continuity does not imply differentiability. Brownian motion is the canonical counterexample — continuous everywhere, differentiable nowhere.
The intermediate value theorem guarantees solutions. Many calibration problems in quant finance amount to finding a root: "find the implied volatility $\sigma^*$ such that $v_{\text{BS}}(S, K, T, r, \sigma^*) = v_{\text{market}}$ ." The intermediate value theorem (a consequence of continuity) guarantees that such a root exists when the pricing function is continuous in $\sigma$ and the market price lies between the boundary values. Without continuity, existence of a solution is not guaranteed.

Definition: continuity at a point

A function

f

is continuous at $a$ if three conditions hold:

$f(a)$ is defined (the function has a value at $a$ ).
$\lim_{x \to a} f(x)$ exists (the limit from both sides converges).
$\lim_{x \to a} f(x) = f(a)$ (the limit equals the function value).

In a single line: $f$ is continuous at $a$ if and only if $\lim_{x \to a} f(x) = f(a)$ .

The epsilon-delta formulation is: for every

\varepsilon > 0

, there exists

\delta > 0

such that

|x - a| < \delta \implies |f(x) - f(a)| < \varepsilon

Note the difference from the limit definition: here we use

|x - a| < \delta

(not

0 < |x - a| < \delta

) because we include the point

a

itself, and the conclusion involves

f(a)

(not some external value

L

Intuition: You can draw the graph of a continuous function without lifting your pen. Small perturbations in input cause small perturbations in output — there are no sudden jumps.

Types of discontinuity

When continuity fails, it fails for one of the three conditions above. The type of failure determines the type of discontinuity.

Removable discontinuity

The limit $\lim_{x \to a} f(x) = L$ exists, but either $f(a)$ is undefined or $f(a) \neq L$ . The discontinuity can be "removed" by redefining $f(a) = L$ .

Example:

f(x) = \frac{x^2 - 4}{x - 2}

is undefined at

x = 2

, but

\lim_{x \to 2} f(x) = \lim_{x \to 2}(x + 2) = 4

. Defining

f(2) = 4

makes the function continuous.

Finance example: Some pricing formulas have removable singularities. The Black-Scholes vega at

T = 0

involves

\sqrt{T}

in the denominator, but the overall expression has a well-defined limit of zero as

T \to 0

for non-ATM options. Implementations must handle these carefully to avoid division by zero.

Jump discontinuity

Both one-sided limits exist but are unequal: $\lim_{x \to a^-} f(x) \neq \lim_{x \to a^+} f(x)$ .

Finance example — the digital (binary) option payoff:

H(S_T) = \begin{cases} 1 & \text{if } S_T \geq K \\ 0 & \text{if } S_T < K \end{cases}

At $S_T = K$ :

\lim_{S_T \to K^-} H = 0, \qquad \lim_{S_T \to K^+} H = 1

The payoff jumps from 0 to 1 at the strike. This discontinuity makes digital options notoriously difficult to hedge: the delta at $S = K$ near expiration is a spike (approximating a Dirac delta), which means the hedge ratio changes violently with small stock moves. The jump in the payoff propagates into extreme Greeks.

Finance example — stock price gaps: Earnings announcements, central bank decisions, and market openings can cause stock prices to jump — moving from

S_{t^-}

S_t

with

S_t \neq S_{t^-}

. In a pure Brownian motion model, paths are continuous and such jumps cannot occur. This is a known limitation; jump-diffusion models (Merton, Kou) explicitly introduce jump discontinuities into the price process to capture this reality.

Essential discontinuity

The function oscillates or diverges near $a$ with no well-defined one-sided limits.

Example:

f(x) = \sin(1/x)

near

x = 0

oscillates infinitely often between

-1

and

1

; no limit exists. These are rare in finance models but relevant conceptually: a model whose output oscillates wildly with small parameter perturbations is poorly conditioned and effectively has an essential discontinuity from a numerical perspective.

Continuity on an interval and global continuity

A function is continuous on an interval

[a, b]

if it is continuous at every point of the open interval

(a, b)

and has the appropriate one-sided continuity at the endpoints:

\lim_{x \to a^+} f(x) = f(a), \qquad \lim_{x \to b^-} f(x) = f(b)

A function is continuous on $\mathbb{R}$ (or on its entire domain) if it is continuous at every point.

Key classes of continuous functions:

Polynomials: continuous everywhere.
Exponentials ( $e^x$ , $e^{-rT}$ ): continuous everywhere.
Logarithms ( $\ln x$ ): continuous on $(0, \infty)$ .
Compositions, sums, products, and quotients (where the denominator is nonzero) of continuous functions are continuous. This follows from the limit laws.
The standard normal CDF $\Phi(x)$ and density $\phi(x)$ : continuous everywhere.

The Black-Scholes pricing formula $C(S, t, \sigma, r, K, T)$ is a combination of exponentials, $\Phi$ , logarithms, and arithmetic — all continuous functions. Therefore $C$ is continuous in all its arguments (wherever they are well-defined), which is why calibration and root-finding algorithms work on it.

Theorems that require continuity

Several powerful theorems are available only when $f$ is continuous. Each has direct financial applications.

Intermediate Value Theorem (IVT)

If $f$ is continuous on $[a, b]$ and $c$ is any value between $f(a)$ and $f(b)$ , then there exists at least one $x^* \in (a, b)$ such that $f(x^*) = c$ .

Finance application — existence of implied volatility: The Black-Scholes price

C(\sigma)

is continuous in

\sigma

(0, \infty)

. As

\sigma \to 0^+

C \to \max(S - Ke^{-rT}, 0)

(the intrinsic value). As

\sigma \to \infty

C \to S

(the upper bound for a call). If the market price

C_{\text{mkt}}

lies strictly between these bounds, the IVT guarantees the existence of an implied volatility

\sigma^* > 0

such that

C(\sigma^*) = C_{\text{mkt}}

This is not a small point — it is the theoretical justification for the entire practice of quoting implied volatilities. Without continuity, you could not guarantee that a solution exists.

Extreme Value Theorem (EVT)

If $f$ is continuous on a closed, bounded interval $[a, b]$ , then $f$ attains its maximum and minimum on $[a, b]$ .

Finance application: When optimising a portfolio over a bounded parameter space — for example, finding the allocation weights

w \in [0, 1]^n

that maximise Sharpe ratio — the EVT guarantees that an optimum exists (provided the objective function is continuous). Without the closed-and-bounded condition, optima may not be attained (the function could approach a supremum without reaching it).

Uniform continuity

A function

f

is uniformly continuous on a set

D

if: for every

\varepsilon > 0

, there exists

\delta > 0

such that

|x - y| < \delta \implies |f(x) - f(y)| < \varepsilon \qquad \text{for all } x, y \in D

The difference from ordinary (pointwise) continuity is that

\delta

works for all points simultaneously, not just for a particular point

a

. By the Heine-Cantor theorem, any function continuous on a closed bounded interval is automatically uniformly continuous there.

In stochastic calculus, uniform continuity ensures that approximating sums converge uniformly to integrals, which strengthens convergence results. The sample paths of Brownian motion on $[0, T]$ are uniformly continuous (being continuous on a closed interval), which is used when constructing Itô integrals.

The relationship between continuity and differentiability

This relationship is asymmetric and fundamental:

\text{Differentiable at } a \implies \text{Continuous at } a

\text{Continuous at } a \;\not\!\!\!\implies \text{Differentiable at } a

Differentiability implies continuity (proof sketch)

If $f$ is differentiable at $a$ , then:

\lim_{x \to a} [f(x) - f(a)] = \lim_{x \to a} \frac{f(x) - f(a)}{x - a} \cdot (x - a) = f'(a) \cdot 0 = 0

So $\lim_{x \to a} f(x) = f(a)$ , which is continuity.

Continuity does not imply differentiability (counterexamples)

The absolute value function

f(x) = |x|

is continuous at

x = 0

but not differentiable there — the left-derivative is

-1

and the right-derivative is

+1

. This is exactly the kink in the payoff

(S - K)^+

S = K

Brownian motion is the extreme case: continuous everywhere, differentiable nowhere (almost surely). The paths are so jagged that no tangent line exists at any point. This is the fundamental reason that the ordinary chain rule fails for stochastic processes and must be replaced by Itô's Lemma.

The implication for quant finance is precise: continuity of price paths is assumed and is sufficient for the Itô machinery to work. But differentiability of price paths is not assumed and does not hold — which is exactly why Itô calculus is needed instead of ordinary calculus.

Continuity in multiple dimensions

For a function $f: \mathbb{R}^n \to \mathbb{R}$ , continuity at a point $\mathbf{a}$ means:

\lim_{\mathbf{x} \to \mathbf{a}} f(\mathbf{x}) = f(\mathbf{a})

where

\mathbf{x} \to \mathbf{a}

means

\|\mathbf{x} - \mathbf{a}\| \to 0

along any path.

The option pricing function

v(S, t)

is a function of two variables. Continuity of

v

in both arguments jointly means that small simultaneous changes in the stock price and time produce small changes in the option value — there are no cliffs or gaps in the pricing surface. This is what makes the Taylor expansion

dv = v_S\,dS + v_t\,dt + \cdots

valid: the expansion assumes the function is smooth (and hence continuous) in a neighbourhood of the current point.

A subtlety: $v$ may be continuous in $S$ for fixed $t$ and continuous in $t$ for fixed $S$ (separately continuous in each variable) but not jointly continuous. Joint continuity is the stronger and correct requirement. For Black-Scholes pricing functions, joint continuity holds everywhere except possibly at the boundary $t = T$ , $S = K$ (where the payoff has a kink).

Examples and applications

Example 1: verifying continuity of the Black-Scholes call price

The Black-Scholes call price is:

C(S) = S\Phi(d_1) - Ke^{-rT}\Phi(d_2)

where $d_1 = \frac{\ln(S/K) + (r + \sigma^2/2)T}{\sigma\sqrt{T}}$ and $d_2 = d_1 - \sigma\sqrt{T}$ .

Each component is continuous: $S$ is continuous in $S$ (trivially), $\ln(S/K)$ is continuous for $S > 0$ , $\Phi$ is continuous everywhere, and $e^{-rT}$ is a constant. Products and sums of continuous functions are continuous. Therefore $C(S)$ is continuous for all $S > 0$ .

At the boundary $S \to 0^+$ : $d_1 \to -\infty$ , so $\Phi(d_1) \to 0$ and $C \to 0$ . At $S \to \infty$ : $d_1 \to +\infty$ , so $\Phi(d_1) \to 1$ and $C \to S - Ke^{-rT} \to \infty$ .

The function smoothly transitions from near-zero (deep OTM) to approximately $S - Ke^{-rT}$ (deep ITM), with no jumps.

Example 2: discontinuity at expiration

Consider the value of a European call at time $t$ very close to expiry $T$ . For $t < T$ , the option price $v(S, t)$ is smooth in $S$ (the Black-Scholes formula is analytic). But at $t = T$ , the value is the payoff $(S - K)^+$ , which has a kink at $S = K$ .

This means that

v(S, t)

is not smooth as a joint function of

(S, t)

at the point

(K, T)

. The delta

\Delta = v_S

transitions from the smooth

\Phi(d_1)

curve (for

t < T

) to a step function

\mathbf{1}_{S > K}

t = T

. The gamma

\Gamma = v_{SS}

becomes infinite at

(K, T)

— it approaches a Dirac delta. This blow-up of Greeks near expiration is a direct consequence of the payoff's non-differentiability and is one of the practical challenges of options market-making.

Example 3: continuity of Brownian sample paths

A standard Brownian motion has continuous paths by definition (axiom BM4). This continuity is what makes the stochastic integral

\int_0^T f(t)\,dW_t

well-defined as an

L^2

limit of Riemann-type sums. If

W_t

had jumps, the convergence theory would be fundamentally different (you would need a theory of integration against jump processes, which exists but is more complex).

Continuity of

W_t

also implies continuity of

S_t = S_0\exp((\mu - \sigma^2/2)t + \sigma W_t)

— the geometric Brownian motion stock price model. Since the exponential function is continuous and the composition of continuous functions is continuous,

S_t

is continuous whenever

W_t

is. No gaps, no jumps, no teleportation of prices. This is an idealisation — real markets do gap — but it is the modelling assumption that makes the Black-Scholes derivation work.

Common confusions and pitfalls

"Continuous means smooth." No. Continuous means "no jumps." Smooth means "infinitely differentiable." A continuous function can have kinks (like

|x|

), corners (like

(S-K)^+

), or be nowhere differentiable (like Brownian motion). Smoothness is a much stronger condition than continuity.

"If a function is defined at every point, it's continuous." No. The step function

\mathbf{1}_{x \geq 0}

is defined everywhere but discontinuous at

x = 0

. Definition and continuity are separate properties.

"Brownian motion is continuous, so we can differentiate it." This is the single most important misconception in stochastic calculus. Continuity is necessary for differentiability but not sufficient. Brownian paths are continuous everywhere and differentiable nowhere. The machinery of Itô's Lemma exists precisely because continuity alone does not give you derivatives.

"The IVT gives uniqueness." The IVT only guarantees existence of a root, not uniqueness. For implied volatility, uniqueness follows from the separate fact that the Black-Scholes price is strictly increasing in

\sigma

(monotonicity), not from continuity alone.

Where this goes next

Continuity is the bridge between limits and differentiation. It is:

The condition that makes the derivative possible (though not guaranteed).
The modelling assumption that makes Brownian motion paths tractable.
The hypothesis behind the IVT, which guarantees existence of implied volatilities, calibration solutions, and optimal hedge ratios.

The next step is differentiation, where the limit definition of the derivative builds directly on the concepts from this page and the limits page. The chain rule, product rule, and quotient rule are all proved using limit manipulations, and their stochastic extensions via Itô's Lemma rely on the continuity of Brownian paths to ensure convergence of the relevant sums.

In the probability branch of the vault, the analogue of continuity for stochastic processes is càdlàg (right-continuous with left limits) — the standard regularity condition for semimartingales. Brownian motion is the special case where the paths are fully continuous (not just right-continuous). For jump-diffusion models, the weaker càdlàg condition is the appropriate framework. See Brownian Motion for the continuous case.

References

Stewart, J. (2008). Single Variable Calculus: Early Transcendentals (6th ed.). Thomson Brooks/Cole. Ch. 2 Section 2.5 (Continuity) and related statements of the Intermediate and Extreme Value Theorems.