Local Time (Mathematics)¶
The selected level density of a continuous semimartingale's quadratic-variation-weighted occupation, tracked as time advances.
Core Idea¶
Local time describes how a continuous random path occupies the levels of its state space as time passes. For a continuous real-valued semimartingale X, the chosen occupation clock is its quadratic variation d⟨X⟩. At time t, the clock-weighted visits to a set A of levels form a measure μ_t(A)=∫₀ᵗ 1_A(X_s)d⟨X⟩_s. Local time L^x_t(X) is a density of that measure relative to ordinary length dx on the level axis: ∫₀ᵗ f(X_s)d⟨X⟩_s=∫ℝ f(x)L^x_t(X)dx. For standard Brownian motion W, d⟨W⟩=dt, so the clock also equals elapsed time in that special case.[ref-824b41c670dd][ref-824b41c670dd-3]
The integral relation identifies a density only almost everywhere in x. To discuss a value at one exact level, especially a boundary, choose a pointwise version. Björk gives one continuous in t and right-continuous with left limits in x. At a fixed level x, that local time is nondecreasing and increases only when X visits x. Tanaka's formula relates it to the correction at a kink such as |X-a|; the formula is a useful relation rather than the whole definition.[ref-824b41c670dd][ref-824b41c670dd-2]
Scope of Application¶
This entry concerns the one-dimensional continuous-semimartingale local time with the quadratic-variation clock and a selected pointwise version. Brownian motion at an interior level is the simple case because the clock is dt. A reflected Brownian state X=|W| at boundary zero is a case where the boundary convention matters. In Björk's worked example, the selected state value L⁰_t(X) is twice L⁰_t(W) of the underlying Brownian path. The regulator F for an auxiliary Skorokhod input W̃=∫sgn(W)dW equals L⁰_t(W), so the state value is 2F. A later theorem in the same source prints a conflicting unqualified F=L⁰_t(X); the worked factor cannot be promoted into a generic regulator rule.[^ref-824b41c670dd-4]
Harrison and Reiman connect reflected Brownian motion to heavy-traffic open queue networks, but the inspected original abstract does not identify a queue regulator or idle-time variable with this scalar local time. It is application context only.[^ref-a8afa2559212]
Clarity¶
“Time spent at x” can be misleading. Local time is a density of occupation over nearby levels, with a selected value at x; growth there does not require a positive ordinary-time interval spent exactly at x. Four questions disambiguate a claim: Which path? Which state level? Which occupation clock and reference measure? Which pointwise version if a boundary value is used? Brownian dt and d⟨W⟩ coincide, but those clocks can differ for another process.[ref-824b41c670dd][ref-824b41c670dd-3]
Manages Complexity¶
The occupation-density equation summarizes many visits across levels in one family L^x_t. Instead of separately counting crossings or dwell intervals, it relates an additive occupation measure to dx. Keeping process, clock, level and version visible also stops Tanaka correction terms, boundary regulators and queue applications from being merged just because they share “local time” vocabulary.[ref-824b41c670dd-3][ref-824b41c670dd-4]
Abstract Reasoning¶
Start a proposed calculation by writing its occupation measure. If it uses dt, check whether dt=d⟨X⟩ for the named path. Next distinguish an integral over x from the selected value at a single x; changing a density on a dx-null set preserves the integral but can change a claimed boundary value. Finally check the normalization before comparing a local-time term with a regulator. For a continuous bounded-variation path, d⟨X⟩ is zero, so this semimartingale local time is zero even if a separate chronological occupation density is nonzero.[ref-824b41c670dd][ref-824b41c670dd-4]
Knowledge Transfer¶
Within stochastic calculus, the process–clock–level–version test works for Brownian interior levels and for the boundary of |W|. The factor two in Björk's reflected example shows why the convention must travel with a formula. More generally, Prime Measure supplies the additive-size-on-sets prerequisite: the occupation measure has a density relative to Lebesgue measure. The named local time adds a stochastic process, quadratic-variation clock and selected pointwise value, so it stays domain-specific.[ref-824b41c670dd-3][ref-824b41c670dd-4]
Examples¶
Brownian interior level. Let W be standard Brownian motion and fix a. Since d⟨W⟩=dt, ∫₀ᵗ f(W_s)ds=∫ℝ f(x)L^x_t(W)dx. W is the path, a the chosen level, dt the occupation clock, dx the reference measure and L^a_t the selected level value. Its growth is supported on visits to a; under Björk's normalization it appears in Tanaka's formula for |W-a|.[ref-824b41c670dd][ref-824b41c670dd-2][^ref-824b41c670dd-3]
Reflected Brownian boundary. Set X=|W| and inspect level zero. The state remains a continuous semimartingale, and its quadratic-variation clock is dt. Björk's selected right-continuous boundary value is L⁰_t(X)=2L⁰_t(W). The regulator F in the auxiliary reflected representation equals L⁰_t(W), not the selected state value. Here X and zero fill the path and level roles, dx is the reference, and the chosen boundary version fixes the pointwise factor. The source's later contradictory theorem limits broader claims.[^ref-824b41c670dd-4]
Relationships to Other Abstractions¶
Current abstraction Local Time (Mathematics) Domain-specific
Parents (1) — more general patterns this builds on
-
Local Time (Mathematics) presupposes Measure Prime
The level density presupposes an occupation measure and a reference measure on state levels.
Hierarchy paths (2) — routes to 2 parentless roots
- Local Time (Mathematics) → Measure → Aggregation → Micro Macro Linkage
- Local Time (Mathematics) → Measure → Set and Membership
Neighborhood in Abstraction Space¶
Local Time (Mathematics) sits in a sparse region of the domain-specific corpus (87th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2551 abstractions)
Nearest neighbors
- Reflection principle (Wiener process) — 0.82
- Ornstein–Uhlenbeck Process — 0.82
- Doléans–Dade Exponential — 0.82
- Kinetic Exchange Models of Markets — 0.80
- Parabolic Hausdorff dimension — 0.80
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
An ordinary-time occupation density for an arbitrary path need not be this quadratic-variation local time. A queue regulator needs its own process, clock and normalization proof before it can be equated to a level density. Self-intersection local time counts contacts between pairs of path times instead of occupation at one fixed state level. An arbitrary a.e. density representative preserves the occupation integral but may give the wrong selected boundary value.[ref-824b41c670dd][ref-824b41c670dd-4][^ref-a8afa2559212]
References¶
[^ref-824b41c670dd]: Tomas Björk, The Pedestrian's Guide to Local Time, arXiv:1512.08912v1 (2015), author's preliminary proof-bearing notes, §§3.3 and 4.1, Definitions 3.1 and 4.1–4.3, Theorem 4.1, printed pp.9–10 and 13–15. [^ref-824b41c670dd-2]: Tomas Björk, The Pedestrian's Guide to Local Time, arXiv:1512.08912v1 (2015), author's preliminary proof-bearing notes, §4.3 Theorem 4.2, printed p.16. [^ref-824b41c670dd-3]: Tomas Björk, The Pedestrian's Guide to Local Time, arXiv:1512.08912v1 (2015), author's preliminary proof-bearing notes, §5 Theorem 5.2, printed p.24. [^ref-824b41c670dd-4]: Tomas Björk, The Pedestrian's Guide to Local Time, arXiv:1512.08912v1 (2015), author's preliminary proof-bearing notes, §§3.5 and 4.5 worked reflected-Brownian calculations, printed pp.11 and 17–18; §4.6 Theorem 4.4, printed p.19, contains the unresolved conflicting regulator equality. [^ref-a8afa2559212]: J. Michael Harrison and Martin I. Reiman, Reflected Brownian Motion on an Orthant, Annals of Probability 9(2), 302–308 (1981), original publisher metadata and abstract only; full article not independently inspected. DOI 10.1214/aop/1176994471.