Local Time (Mathematics)¶
The selected level density of a continuous semimartingale's quadratic-variation-weighted occupation, tracked as time advances.
Core Idea¶
Local time here is the selected density, at a state level, of how a continuous real-valued semimartingale occupies levels as time advances. For a process X and fixed time t, form the measure μ_t(A) = ∫₀ᵗ 1_A(X_s)d⟨X⟩_s on sets A of state levels. Its density relative to Lebesgue measure dx is written L^x_t(X): ∫₀ᵗ f(X_s)d⟨X⟩_s = ∫ℝ f(x)L^x_t(X)dx for suitable f. The clock is quadratic variation. For standard Brownian motion W, d⟨W⟩=dt, so this is also an ordinary-time occupation density. Those clocks need not coincide for another semimartingale.[1][2]
A density equation fixes L^x_t only for almost every x. Statements at a particular level, especially a reflecting boundary, therefore use a specified pointwise version. Björk gives a version continuous in t and right-continuous with left limits in x; with that choice, local time at fixed x is nondecreasing and grows only at times when X is at x. The Tanaka formula connects the same selected local time to the correction at the kink of |X-a|. It is an important relation, not a requirement that every application begin from that formula.[1][3]
Structural Signature¶
Signature: continuous real semimartingale and state level → quadratic-variation-weighted occupation measure → density relative to dx → selected level value as time advances.[1][2]
- Path and level. X supplies a continuous trajectory and x identifies the state value at which the density is read. Changing x changes the question; comparing two path times for self-contact defines a different functional.[1]
- Occupation clock and measure. d⟨X⟩ weights visits to sets of levels. This is ordinary elapsed time for standard Brownian motion, but a continuous bounded-variation process has zero quadratic variation and hence zero local time in this convention.[1]
- Reference and density. Lebesgue measure dx on state levels supplies the reference against which μ_t has density L^x_t. A raw visit count or a regulating push is not by itself this density.[2]
- Selected pointwise version. The occupation integral alone permits changes on dx-null sets. A fixed-level claim uses the right-continuous-in-level, Tanaka-compatible version in the consulted source. Its increase at level x is supported on visits to x, even though the path need not spend positive ordinary elapsed time at exactly that singleton.[1][3]
What It Is Not¶
It is not an ordinary-dt occupation density for every continuous semimartingale. Brownian motion makes dt and d⟨W⟩ coincide; that special equality cannot be carried over without checking the process. A differentiable path can have a nonzero chronological occupation density while its quadratic-variation local time is zero.[1]
It is not any regulator or queue idle-time variable named “local time.” A boundary regulator may be related to local time after the state, driver, clock, boundary version and normalization are specified. In Björk's reflected Brownian worked example, X=|W| has selected boundary value L⁰_t(X)=2L⁰_t(W), while the Skorokhod regulator F equals L⁰_t(W) of the underlying W. His later theorem prints an incompatible unqualified equality F=L⁰_t(X); this entry does not use that printed equality as a general rule.[4]
It is also not polymer self-intersection local time. That construction concerns pairs of path times at which a path contacts itself, not the one-state-level density L^x_t defined above.
Scope of Application¶
This entry covers the one-dimensional, continuous-semimartingale construction under the d⟨X⟩ clock and a chosen pointwise version. Brownian local time at an interior level is the direct case: its quadratic-variation clock is ordinary time, and its fixed-level increase occurs on visits to that level. The reflected Brownian state |W| at zero is a boundary case that requires the pointwise version and factor convention to be named.[1][4]
Reflected Brownian processes also appear in queueing limits. Harrison and Reiman's original abstract connects reflected Brownian motion in an orthant to heavy-traffic open queue networks. That abstract does not prove that a multidimensional regulator or operational idle time equals this scalar level density. It provides application context, not a third positive instance of the same one-level formula.[5]
Clarity¶
“Time at a level” sounds like elapsed time spent at one exact state value. Local time instead reads a selected density of occupation over nearby levels. Its growth can be concentrated on visits to x even when singleton visits occupy no positive ordinary-time interval. Ask which process is indexed, which clock weights occupation, which reference measure defines the density, and which pointwise version fixes a boundary value. Each answer prevents a different ambiguity.[1][2]
The reflected example adds a naming test. W is the underlying Brownian path; X=|W| is the reflected state; W̃=∫sgn(W)dW is the auxiliary Skorokhod input; F is its regulator. The source's worked computation identifies F with L⁰(W), and its selected state boundary value with 2F. Calling all four objects “the local time of the queue” erases a factor and a state/driver distinction.[4]
Manages Complexity¶
The measure-density equation compresses many individual visits to many nearby levels into one level-indexed family L^x_t. The analyst need not enumerate crossings or dwell intervals. Process, clock, reference measure and pointwise convention are the small set of data needed to determine what a formula about occupation means. The Brownian case simplifies the clock to dt; the general continuous-semimartingale case keeps d⟨X⟩ visible.[1][2]
It also keeps neighboring constructions separate. Tanaka's kink correction, a reflected-boundary regulator, and a queue's heavy-traffic limit can be discussed together only after their respective identities are specified. A shared boundary vocabulary does not substitute for the process and normalization equations.[3][4][5]
Abstract Reasoning¶
To test a proposed local-time claim, first write the occupation measure and the reference measure. If the claimed integral uses dt, check whether d⟨X⟩=dt for the named process. Next ask whether the conclusion concerns an integrated density or its value at one x. The integrated relation tolerates changes on a dx-null set; the fixed-level conclusion requires the selected version. Only then compare a Tanaka correction or boundary regulator and its coefficient.[1][2][3]
This sequence diagnoses the bounded-variation near miss. Its chronological residence near a level may be nontrivial, but d⟨X⟩ is zero. The stated semimartingale occupation measure is then zero, so its local time in this convention is zero. The difference follows from the clock rather than a claim that the path never visited the level.[1]
Knowledge Transfer¶
Within stochastic calculus, the same process–clock–level–version test carries from Brownian motion at an interior point to a reflected Brownian state at its boundary. At the boundary, the density's pointwise convention becomes visible: the selected L⁰(|W|) is twice L⁰(W) in Björk's worked calculation, while the regulating term F is L⁰(W). The general habit of checking clocks and conventions transfers; that coefficient does not become a universal equation for reflected queues.[4]
Beyond this domain, Prime Measure supplies the prerequisite idea of additive size on sets. Here one occupation measure is related to a second, Lebesgue reference measure, by a density. Prime Measure travels to many settings; the named local-time construction additionally requires a continuous stochastic path, quadratic-variation clock, level index and selected pointwise version. This added structure is why Local Time remains a domain-specific entry rather than a renaming of Measure.[2]
Examples¶
Brownian motion at an interior level¶
Let W be standard Brownian motion and fix a∈ℝ. For any suitable test function f, occupation up to t can be computed either as ∫₀ᵗ f(W_s)ds or as ∫ℝ f(x)L^x_t(W)dx, because d⟨W⟩=dt. The selected L^a_t(W) grows only when W visits a. In Tanaka's formula, it also supplies the correction associated with the kink of |W-a| under Björk's normalization.[1][3][2]
Mapped back: W is the path and a the level; dt=d⟨W⟩ is the occupation clock; the weighted visit assignment is the occupation measure; dx is the reference; and Björk's continuous Brownian, pointwise L^a_t is the selected version. The equality of clocks belongs to W, not to every process.
Reflected Brownian state at its boundary¶
Set X=|W| and inspect level zero. Björk's worked calculation uses the auxiliary Brownian input W̃=∫sgn(W)dW and a regulating term F=L⁰_t(W). The selected reflected-state boundary local time obeys L⁰_t(X)=2L⁰_t(W)=2F. This illustrates why one must distinguish the state density from a push associated with an auxiliary input. The source's later contradictory theorem statement limits any extrapolation beyond this worked case.[4]
Mapped back: X is the path and zero the level; d⟨X⟩=dt is the occupation clock in this case; visits to nonnegative level sets define the occupation measure relative to dx; the right-continuous/Tanaka-compatible boundary value is the selected version. F belongs to the reflected representation and is not substituted for L⁰(X) without the factor.
Structural Tensions¶
The consulted construction does not establish a necessary two-objective tradeoff in the identity of local time. Brownian dt and d⟨W⟩ are equal descriptions of one clock in that case, while the general semimartingale definition fixes d⟨X⟩. The choice between an integrated density claim and a selected boundary value is a specification requirement, not an irreducible pressure to be optimized. These distinctions are recorded in the boundary and reasoning tests rather than presented as invented tensions.[1][2]
Structural–Framed Character¶
This entry sits toward the structural end of the spectrum: it is an exact stochastic-calculus object once process, clock and normalization are given. Its defining equation contains no moral or policy evaluation, and the object does not originate in an institution or require a human procedure to occur. Terms such as “time spent” and “reflection” can travel informally, but importing the named local-time formula to another domain requires proving an occupation-density relation with the relevant clock and pointwise convention; resemblance is not recognition of the same object.[1][2]
The portable skeleton is the parent Measure relation—additive assignment on level sets and density against a reference measure. That skeleton appears beyond stochastic calculus, but the selected Brownian/semimartingale clock, support-on-level property, and boundary normalization do not automatically travel with it. Its character: structurally defined within a specified stochastic setting, with domain-bound assumptions that keep the named entry from becoming a general Prime.
Structural Core vs. Domain Accent¶
The structural core is a density relating an occupation measure on state levels to Lebesgue measure, hence the strict presupposition of Prime Measure. The domain accent is constitutive here: a continuous real semimartingale, quadratic-variation clock, fixed state level, time-indexed additive functional and selected pointwise version. Remove those and one may still have a measure or density, but no longer this local time.[1][2]
That is also the Prime boundary. Measure has many substrates and preserves its additive-set-function identity; Local Time's identity depends on stochastic-path hypotheses and a convention-sensitive boundary value. The method of checking clock and version can inform other domains, but a transferable checking habit does not establish that the named mathematical object itself has the substrate independence of its Prime parent.
Instantiates / Related Primes¶
This entry presupposes Measure.
Prime Measure — strict prerequisite. The fixed-t occupation assignment gives nonnegative additive size to level sets, and dx gives the reference size for a density. Local time is the density value, not itself a measure on subsets, so this is a composition/presupposes edge rather than kind_of.[2]
Prime Stochastic Process — related carrier, pending separate edge test. X is a stochastic process, but this entry's reviewed strict relation is the measure-density prerequisite; a process alone does not give a local time without clock, occupation measure and version. Prime Measurement is not a parent because a defined density is not an observer's empirical measuring act. Prime Accumulation captures the nondecreasing t profile but does not supply the level-density identity; monotonicity alone fails the full signature.[1]
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.For fixed t, A ↦ ∫₀ᵗ 1_A(X_s)d⟨X⟩_s is a nonnegative additive measure on level sets, and Lebesgue measure dx is the reference relative to which its density L^x_t is defined. Removing either measure destroys the occupation-density relation. Local time adds a specified stochastic path, clock, level and selected pointwise version, so it presupposes Measure rather than being a kind of Measure.
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¶
Chronological occupation density may differ from quadratic-variation local time outside the Brownian special case. A queue regulator may be linked to a boundary local time only with a proved state/driver and normalization relation; Harrison–Reiman's abstract does not supply a generic equality. Self-intersection local time concerns pairs of path times rather than a single fixed state level. An arbitrary a.e. representative gives the same integrated density but may change a boundary point value, so it cannot replace the selected version when L⁰ is claimed.[1][4][5]
References¶
[1] 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. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r
[2] 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. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l
[3] 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. registry ↩a ↩b ↩c ↩d ↩e
[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. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g
[5] 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. registry ↩a ↩b ↩c