Basic Almost-Sure Properties of the Brownian Zero Set

The Brownian Zero Set

A Brownian path can visit the same level at many different times. The set of times at which it is exactly zero is therefore a natural object to study, but two different questions must be kept separate: is this set closed, and how much time does it occupy? Continuity answers the first question path by path. For the second, the Gaussian distribution at each fixed positive time combines with integration over time to show that the zero set has zero Lebesgue measure almost surely.

Work on a probability space \((\Omega,\mathcal{F},\mathbb{P})\) carrying a standard Brownian motion \((B_t)_{t\ge0}\) with continuous paths almost surely and \(B_0=0\). Its state space is \(\mathbb{R}\) and its time index is continuous, \(t\ge0\). Define the zero set of a path by \(Z(\omega)=\{t\ge0:B_t(\omega)=0\}\). For \(T>0\), write \(Z_T(\omega)=Z(\omega)\cap[0,T]\), and let \(\operatorname{Leb}\) denote Lebesgue measure on the time axis.

Continuity Makes the Zero Set Closed

Proof. Fix a path \(\omega\) that is continuous, and take any sequence \((t_m)_{m\ge1}\) in \(Z(\omega)\) that converges to \(t\in[0,\infty)\). By the definition of \(Z\), \(B_{t_m}(\omega)=0\) for every \(m\). Continuity at \(t\) gives \(B_t(\omega)=\lim_{m\to\infty}B_{t_m}(\omega)=0\), so \(t\in Z(\omega)\). Thus \(Z(\omega)\) is closed. Its intersection \(Z_T(\omega)\) with the closed bounded interval \([0,T]\) is closed and bounded in \(\mathbb{R}\), hence compact. This also covers the empty-set case, although here \(0\in Z(\omega)\) because \(B_0=0\). \(\square\)

Fixed Times Have Zero Probability of Being a Zero

For each fixed \(t>0\), the Brownian marginal distribution is \(B_t\sim N(0,t)\), with density \(x\mapsto (2\pi t)^{-1/2}\exp(-x^2/(2t))\). A density assigns probability zero to a singleton: the integral of the density over the one-point set \(\{0\}\), which has Lebesgue measure zero, is zero. Therefore \(\mathbb{P}(B_t=0)=0\) for each fixed \(t>0\). At \(t=0\), by contrast, \(\mathbb{P}(B_0=0)=1\); this single exceptional time has Lebesgue measure zero.

Expected Measure of the Zero Set

For a fixed \(T>0\), define the random variable \(L_T=\operatorname{Leb}(Z_T)\). Equivalently, \(L_T=\int_0^T\mathbf{1}_{\{B_t=0\}}\,dt\), where \(\mathbf{1}_{\{B_t=0\}}\) is the indicator of the event that \(B_t=0\). The Brownian process has a jointly measurable version, so the indicator is measurable in \((\omega,t)\). We use Tonelli's theorem in its standard nonnegative form: for a jointly measurable nonnegative function, the expectation of its time integral equals the time integral of its expectation. Tonelli's theorem is a standard measure-theoretic result; its proof is not needed for the Brownian calculation here.

Proof. For finite \(T>0\), Tonelli's theorem and the fixed-time probabilities give \[ \mathbb{E}[L_T] =\mathbb{E}\left[\int_0^T\mathbf{1}_{\{B_t=0\}}\,dt\right] =\int_0^T\mathbb{P}(B_t=0)\,dt. \] For every \(t>0\), the integrand is zero. At \(t=0\), it is one, but changing an integrand at a single time does not change its Lebesgue integral. Hence \(\mathbb{E}[L_T]=0\). Since \(L_T\ge0\), this implies \(L_T=0\) almost surely: if \(\mathbb{P}(L_T>0)>0\), then for some integer \(m\ge1\), \(\mathbb{P}(L_T\ge1/m)>0\), which would give \(\mathbb{E}[L_T]\ge \mathbb{P}(L_T\ge1/m)/m>0\), a contradiction. This proves the finite-horizon claim.

For the full time axis, apply the finite-horizon result to each positive integer \(T=n\). Each event \(\{\operatorname{Leb}(Z\cap[0,n])=0\}\) has probability one. Their countable intersection also has probability one, because the union of their probability-zero complements has probability zero. On that intersection, \(Z=\bigcup_{n\ge1}(Z\cap[0,n])\) has measure zero by countable subadditivity. \(\square\)

A Quantitative Check Using a Narrow Band

The same density calculation can give an upper bound for the amount of time a Brownian path spends near zero on an interval that stays away from time zero. This is a useful check on the role of the density: on such an interval its maximum is finite, and the probability of lying in a narrow band is at most the band width times that maximum.

What Zero Measure Does—and Does Not—Say

The result is an almost-sure statement about Lebesgue measure in time. It does not say that the zero set is empty: \(0\in Z\) for every Brownian path under the convention \(B_0=0\). Nor does zero measure mean that the set is finite. Closedness and zero measure are compatible with a set containing infinitely many points. These conclusions also differ from a statement at a single fixed time: \(\mathbb{P}(B_t=0)=0\) for \(t>0\) is a probability statement about one chosen time, whereas \(\operatorname{Leb}(Z\cap[0,T])=0\) is a pathwise statement about the measure of all zero times in an interval.

Common pitfall. A mistaken argument is: “For every \(t\in[1,2]\), \(\mathbb{P}(B_t=0)=0\); therefore the probability that there is any zero in \([1,2]\) is zero.” This treats an uncountable union of probability-zero events as though it were a countable union. Countable subadditivity only guarantees that a countable union of null events is null; it gives no such conclusion for an uncountable union. The correct integration argument instead computes the expected Lebesgue measure of the zero times and proves that this nonnegative random variable equals zero almost surely. A measure-zero set can still be nonempty, so that conclusion does not establish the mistaken claim.

The two main facts use different inputs. Closedness follows directly from path continuity: limits of zeros remain zeros. Zero Lebesgue measure uses the Gaussian distribution at every fixed positive time and Tonelli's theorem to pass from those probabilities to expected time spent exactly at zero. Keeping the pathwise and probabilistic parts distinct prevents confusing “closed,” “measure zero,” and “no zeros.”