A continuous adapted process is a local martingale if there are stopping times almost surely such that each stopped process is a martingale. In this definition is integrable. A continuous adapted process has finite variation if, almost surely, its paths have finite total variation of a function on each compact time interval:
The permitted fact that this total-variation process is continuous and adapted makes its level-hitting times stopping times.
Suppose , and put . Define
These times increase to infinity almost surely. The optional stopping theorem applied to the localized martingale shows that is a martingale; it is also bounded in absolute value by , with total variation at most . Fix and write . For a deterministic partition of an interval , orthogonality of square-integrable martingale increments gives
Pathwise the sum on the right is bounded by . For partitions with mesh tending to zero, continuity makes this converge to zero. It is bounded by , so dominated convergence gives . First use rational , then continuity, to obtain almost surely. Letting proves
This proves continuous finite-variation local martingale is constant without assuming integrability of the unstopped total variation. Continuity matters: compensated jump martingales can have finite variation without being constant.
The correct state in a general Gaussian forward-rate field is the whole current curve. In time-to-maturity coordinates , assume deterministic time-homogeneous volatility . Set . The Musiela forward-curve equation is
Under the appropriate maturity differentiability, or in its mild interpretation on a suitable function space, its solution is
For any future interval, the shifted current curve determines the deterministic part and future Brownian increments are independent of past information. This proves that the curve-valued process is time-homogeneous Markov. General observation-time-dependent deterministic volatility instead gives a time-inhomogeneous Markov curve. A low-dimensional state requires more: the maturity functions carrying the randomness must close under the shift evolution. Finitely many Brownian drivers by itself does not ensure that the short rate is Markov.
For a precise obstruction, take two independent stationary Ornstein-Uhlenbeck processes with distinct positive decay parameters , and let . Its centered covariance is with . A stationary scalar Gaussian Markov process must satisfy the covariance criterion for a stationary Gaussian Markov process:
Indeed Gaussian conditioning gives ; if the present is a Markov state, conditioning the future also on the earlier cannot change that mean. Multiplication by and taking expectations yields the equation. For the two-factor covariance the difference of its two sides is
So the short rate alone is not Markov, although the two-dimensional factor process is. Distinct exponentially decaying HJM volatility components realize such a multi-factor Gaussian curve model.
Stationarity is an additional condition, not a consequence of Gaussianity or time-homogeneity. A stationary Gaussian forward curve for the time-homogeneous model can be constructed when
is finite and the deterministic mean satisfies . Initialize with mean and covariance , independently of future noise. The mild solution has the same mean, since . Its covariance is
Thus its one-time Gaussian curve distribution is invariant, and the time-homogeneous transition law makes all time-shifted finite-dimensional distributions invariant as well. The initial field can explicitly be built from an independent noise on negative times using the same square-integrable kernels.
For example, with gives
where and the factors start in their stationary normal laws. This is a finite-dimensional Markov realization when the sum is finite. In contrast, a nonzero constant maturity volatility makes infinite, excluding a finite-variance stationary curve of this form. A deterministic initial curve is generally nonstationary even when a stationary distribution exists. These distinctions separate Gaussianity, Markov state sufficiency, and stationarity.
An interest-rate cap is the sum of caplets over successive reset and payment dates. Consider one accrual period with year fraction and strike . The reset rate is ; the unit-notional payment at is . At reset its value is
Hence the caplet is equivalent before reset to puts expiring at on the maturity- zero-coupon bond, with strike . This algebra accounts for payment at rather than incorrectly treating the payment date as .
Under the T-forward measure, use the bond-ratio process , which is a martingale. From the Gaussian HJM dynamics its relative volatility is . With deterministic volatilities its logarithm at reset is normal with integrated variance
For , write and . A normal integral gives and . Therefore the Gaussian caplet bond-put formula is
For , the value is . This is also obtained from the Gaussian bond-option formula by put-call parity. A cap price is the sum of its caplet prices, multiplied by the notional principal.
In the Vasicek model, the instantaneous forward-rate volatility is . Thus
Indeed, , whose squared integral gives this expression without ambiguity. Gaussian short and instantaneous forward rates do not mean the reset floating rate is Gaussian: the latter is an inverse bond price minus one. The lognormal bond-ratio calculation, under the proper forward measure, is the reason for the normal-CDF pricing formula.
A Gaussian forward-rate field models the whole instantaneous forward rate curve. Let be a real separable Hilbert space, let be a cylindrical Brownian motion, and let deterministic satisfy the maturity-integrability and square-integrability conditions needed below. For a deterministic initial curve, specify under
Using an orthonormal basis, the noise is ; an infinite-dimensional cylindrical noise need not exist as an -valued random vector. Its covariation kernel across maturities is the positive semidefinite kernel . Thus arbitrary suitable maturity-correlation kernels can be represented by Hilbert-space factors, while finitely many factors give finite-rank kernels.
No arbitrage requires a drift restriction. Define the zero-coupon bond price and integrated volatility by
Differentiation of the moving lower endpoint and the Itô formula give
The discounted bond must have zero drift. Differentiating with respect to maturity yields the Heath-Jarrow-Morton model restriction
This is why forward-rate drift and covariance cannot be selected independently in an arbitrage-free random-field model. With deterministic volatilities square-integrable on the finite horizon, the discounted bond stochastic exponential is a true martingale, not only a local martingale.
The solution is explicitly Gaussian:
with mean given by the first two terms and covariance
The entire Gaussian random field can therefore be calibrated through its initial curve and maturity covariance structure. Bond prices are exponential functionals of these Gaussian rates; conditional bond ratios have lognormal laws under appropriate forward measures. Negative rates are possible. A large or infinite volatility space provides rich maturity correlations, but Gaussianity alone supplies neither a low-dimensional Markov state nor positivity of rates.
A Gaussian one-factor short-rate model is most naturally specified under a chosen risk-neutral measure , not by using a physical drift unaltered in a pricing formula. The Vasicek model has
Its explicit solution is . It is a Markov process, and its conditional mean and variance are and . Hence all finite-dimensional distributions are Gaussian. The model permits negative rates; Gaussianity does not enforce their positivity.
The unit zero-coupon bond price is . Write and . Stochastic integration of the explicit solution gives
This Gaussian variable has variance . Evaluating its exponential moment gives exponential-affine bond pricing
For example, , which gives the displayed expression by direct simplification. The term structure of interest rates is therefore analytically tractable, and bond-price changes are driven by the current short rate alone.
The Hull-White model replaces the drift by with deterministic . The same calculation gives
This retains the one-factor Gaussian and Markov structure while allowing an exact initial-curve fit through . If is the initial instantaneous forward rate and , differentiating the bond formula shows that
Thus calibration flexibility is separated from stochastic mean reversion. Time-dependent calibration generally destroys time stationarity even though the state remains Markov. In the time-homogeneous Vasicek case, a stationary initialization is independent of future increments; its covariance is . A fixed initial short rate does not give a stationary process.
We prove exhaustiveness, including existence of the positive density when neither form of arbitrage occurs. Use the subspace from part (a)(ii), and consider
This is a nonempty convex cone closed under positive scaling: supplies one element and guarantees weighted integrability even when itself has no first moment. Crucially, is relatively open in ; it need not be closed.
To prove openness at a point , put . The vectors , for bounded real measurable , span . Indeed, a vector orthogonal to all of them would satisfy for every such , forcing almost surely. Since , this says . Choose finitely many bounded for which form a basis of . For coefficients sufficiently small,
This gives a relative neighbourhood of . It also proves the openness step behind the strictly positive barycentre cone lemma without assuming a closed moment cone.
If , the second arbitrage form was already constructed. If , the separation of a point and an open convex set, applied inside the finite-dimensional space , gives a nonzero with . Positive scaling in the cone forces
We claim almost surely. Otherwise has positive probability, and the admissible positive weights give
contradicting the separation inequality. Since is nonzero, cannot vanish almost surely; thus it is strictly positive on an event of positive probability. This is the first arbitrage form.
Consequently absence of both forms forces , which supplies the required . Conversely either form excludes every such density by part (a). We have proved exactly one of arbitrage or a strictly positive pricing density holds. If , the same argument reduces to the elementary cases and , so no nondegeneracy assumption has been hidden.
Let and . This connection event is the union over of . The union bound therefore supplies a graph vertex with connection probability at least .
Every boundary graph vertex has some coordinate equal to or . By applying a permutation and a graph automorphisms, which preserve both the law and the boundary, choose such a graph vertex in the form while retaining its probability bound. Now is a reflection of . Translation and reflection invariance give
Positive association of random variables applied to the two connection events, whose intersection implies , proves the reflection comparison of random-cluster connections:
Also , since any path to meets the box boundary. Consequently
and hence
Here , so . Both outside logarithmic connection rates converge to by part (b). Squeezing gives the rate for every box size, and therefore for the requested even subsequence:
Write and . We use these standard facts about the infinite-volume wired random-cluster measure for : it is translation invariant and invariant under coordinate reflections/permutations; it has positive association of the random-cluster model; and every conditional edge-open probability is at least
The last bound follows because an edge either leaves its component count unchanged on opening, giving probability , or merges two components, giving . These are standard infinite-volume properties of the limit obtained from the finite wired laws by weak convergence of probability measures. Positive association of random variables applies to connection events by increasing approximation with finite-path events.
Opening the edges of the straight segment gives , so its logarithm is finite. The events and are increasing, and their intersection implies connection to . Thus
The sequence is a subadditive sequence, with . Part (a) proves the random-cluster inverse correlation length exists and is finite:
Since for every , exponentiating yields the requested bound
Set . It may be , but cannot be because is finite. Fix and write , . Repeated use of the subadditive sequence inequality gives
where . Dividing by and letting gives . Since this holds for every , the upper limit of a sequence is at most , while every ratio is at least . If is finite these bounds agree. If , choose with below any prescribed real number; the same upper-limit bound proves convergence to . This establishes the extended-real Fekete lemma:
A finite real limit of a sequence requires a lower linear bound. For example satisfies the premise but has normalized limit of a sequence , so that qualification cannot be dropped.
With the inner boundary specified in the PDF, the Ising boundary condition pins . Explicitly is conditioned on those Ising spins. Its partition function sums over unpinned interior Ising spins only; edges between two pinned sites contribute a constant that cancels on normalization.
Pinning coordinates preserves the FKG lattice condition, since meet in a lattice and join in a lattice still have the same pinned values. It also gives monotonicity in the pins. For any increasing observable , positive association of random variables with an unpinned Ising spin gives
Both conditional probabilities are positive. The same inequality holds after other Ising spins have been pinned. Changing pins one at a time proves stochastic domination of the measures as increases.
Now let and take an increasing observable depending on a fixed set inside the smaller box. Given the Ising spins on , the larger-box measure induces precisely the smaller-box interior law with those boundary values: nearest-neighbor interactions cannot see farther once this boundary is fixed. Its conditional law is therefore between the all-zero and all-one smaller-box laws. Averaging yields
These are bounded monotone sequences along any increasing box exhaustion.
For every finite set , the event that all Ising spins in are one is increasing, so its probability converges in both boundary sequences. Probabilities of arbitrary finite zero/one patterns are finite inclusion-exclusion combinations of these all-one probabilities, and hence converge too. The limits are consistent finite-dimensional distributions; they define measures on and give weak convergence of probability measures in the product topology. Thus both extremal infinite-volume Ising measures exist, with below in stochastic domination. This argument proves existence even when the two limits are different.
Work edge by edge. Let have two-bit restrictions to an edge. If these restrictions are comparable, taking their meet in a lattice and join in a lattice only reorders the two pairs, so their total equality indicators do not change. If they are incomparable, they are and , while their meet in a lattice and join in a lattice are and . In that case the sum of equality indicators increases from zero to two. Consequently
The partition functions cancel from the required ratio, leaving
This is the ferromagnetic Ising lattice inequality in binary-spin normalization, so the measure satisfies the FKG lattice condition and is positively associated. If , then : the conventional coupling is , which explains the factor two in the edge calculation.
Order configurations coordinatewise: when for every coordinate. An increasing event is an upward-closed set: and imply .
The finite-lattice FKG inequality states that a probability mass function satisfying the FKG lattice condition
is positively associated. Here meet in a lattice and join in a lattice are coordinatewise minimum and maximum. Thus for increasing observables ,
Taking indicator functions gives for all increasing events. The strictly positive version suffices for the Ising weights here; the theorem also holds for nonnegative log-supermodular weights of positive total mass.
Use and the positive return time . The walk has transitions , ; its transition matrix defines a reversible Markov chain, although the transition probabilities need not be symmetric.
The voltage equals one at , zero at , and is a harmonic function on a graph away from the terminals: . For this unit electric potential difference the electric current is
The first-step analysis therefore proves the return probability from effective resistance identity:
Finite irreducibility ensures eventual hitting of one of the terminals, so no missing escape event appears in this finite-network calculation.
For an infinite locally finite connected graph, exhaust it by finite sets and wire their exteriors to one sink. The event of exiting before the first return decreases to the event of never returning. Thus
with . In particular recurrence is equivalent to infinite effective resistance to infinity.
After deleting an edge, consider any infinite component and choose a root away from that edge's endpoints. Its vertex degree is unchanged. On every finite wired exhaustion the Rayleigh monotonicity principle makes the resistance no smaller; the limit of a sequence does too. A recurrent original network consequently leaves a recurrent component. Finite components give recurrent walks as well, with an isolated graph vertex viewed as an absorbing state. This proves that edge deletion preserves recurrence. The phrase “no less recurrent” refers to this recurrence comparison: at a deleted-edge endpoint the vertex degree factor also changes, so monotonicity of its first-return probability alone should not be inferred.
Orient each edge both ways. A unit flow from to is antisymmetric, , with divergence of a flow equal to at , at and zero elsewhere. Its energy is
where each unoriented edge is counted once. Put . The electrical unit current is , where the potential solves the Kirchhoff node law with those source and sink divergences. Because is a connected graph, this potential is unique up to an additive constant. The effective resistance is the electric potential difference needed for unit current,
Discrete summation by parts gives .
The Thomson principle says that this current minimizes energy among all unit flows. Indeed write any other unit flow as , where . Then
Consequently
This also proves uniqueness of the minimizing current.
If for every edge, then for every admissible flow. The admissible set is unchanged, so taking its minimum proves the Rayleigh monotonicity principle:
Deletion is the limiting operation : finite-energy flows must then carry zero current on that edge. If the terminals become disconnected, the effective resistance is infinite.
Put . Each successive crossing time , , is a stopping time, since in discrete time the event that a threshold is first met after the previous stopping time can be written as a finite union of events known at the candidate time. Use , with all later crossing times then infinite as well.
For the supermartingale assertion, build a process by finitely many switches. Begin with the constant supermartingale one. At switch to ; the new value at that time is at most one. At switch to the constant ; the old value there is . At switch to , whose value is at most , and at switch to the constant , whose value is at most the previous process. More generally, the switches at and are respectively from cash to and from that stock value to cash . Every switch is downward, including possible overshoots of either threshold.
Each candidate continuation is a supermartingale, being either constant or a fixed positive multiple of . Part (a) therefore proves by induction that the process obtained after the first switches is a supermartingale. It agrees with through . For fixed , its values are bounded in absolute value by a deterministic constant depending on times , which also verifies the required integrability rather than assuming it for an infinite sequence of pastings.
Stopping a discrete supermartingale preserves the property. Indeed its stopped increment is , with an -measurable indicator; integrability at any fixed time follows from a finite sum of the absolute values at earlier times. Hence
This is the multiplicative upcrossing supermartingale construction. The printed case uses , while was only defined from time zero. With the natural pre-start convention , it is the constant process one and the assertion also holds for . Without such an extension, that single displayed stopped process is undefined.
For the probability bound, the actual initial value is
because exactly when . For an integer , stop after the th completed upcrossing, at . The nonnegative stopped supermartingale satisfies
since on the event in the left side. Letting , and identifying , proves the Dubins upcrossing inequality:
Neither finiteness of nor uniform integrability is needed. The bound is for positive integers ; extending it to would generally be false, because its left side would be one while can be less than one.
Interpret the comparison at on ; values at an infinite stopping time are not required. The stopping time events make adapted, and gives integrability. On the next value is . On it is unless , and on that latter event switching only decreases the value. Consequently the pointwise inequality
holds even though need not be known at time . Both indicators displayed here are -measurable. Taking conditional expectations and using the two supermartingale inequalities therefore gives
Thus This proves pasting supermartingales with downward jumps without incorrectly treating the future switching event as predictable.
For the finite probability distribution, put . These are independent fair zero-one variables, and
The binary integer is uniform on . Hence
These are the midpoints of equal subintervals of . For every bounded continuous , the Riemann sums give
Thus in the sense of weak convergence of probability measures.
For the almost-sure and assertions, the Dyadic Rademacher series converges absolutely on every sign sequence, since . Its tail satisfies , proving almost sure convergence and convergence in L2. More precisely, zero means and independence eliminate the cross terms:
This identity follows first for finite tails and then by dominated convergence. Bounded continuous test functions and bounded convergence theorem identify the probability distribution of with the weak limit above, namely the uniform distribution on .
For the requested product identity, each Rademacher random variable has characteristic function . The characteristic function of a sum of independent variables therefore yields
By part (a), or directly by bounded convergence theorem, this converges to the characteristic function of a uniform distribution, . Consequently the dyadic cosine product is
Both sides are one at zero. At zeros of the sinc function the identity is understood as the limit of the finite products, so no logarithm or division by a vanishing factor is needed.
For probability measures on a metric space, weak convergence of probability measures means for every bounded continuous real function . On the real line it is equivalently convergence of the cumulative distribution functions at every continuity point of the limiting cumulative distribution function. Bounded continuous upper and lower approximations to interval indicators give one direction; a partition at continuity points, uniform approximation on a compact interval, and control of the tails give the reverse direction.
For a probability measure on , its characteristic function is . The Lévy continuity theorem states: implies for every real ; conversely, if these characteristic functions have a pointwise limit which is continuous at zero, there is a unique probability measure with characteristic function , and . In particular, for a specified limiting probability measure, weak convergence is equivalent to pointwise convergence of its characteristic functions.
The forward implication follows by testing cosine and sine. The dominated convergence theorem shows that every characteristic function is continuous at zero and has value one there. For the converse, first establish tightness of probability measures. By Fubini's theorem, for ,
The integrand on the right is at least when . This proves the characteristic-function tightness bound
The integrals converge for each by dominated convergence, and continuity of at zero makes the limit as small as desired by choosing small. Thus all sufficiently late measures have a common tail bound; the finitely many earlier measures also have small tails once the radius is increased. This is uniform tightness of probability measures.
We give the required compactness argument explicitly. For any subsequence, choose a further diagonal subsequence whose cumulative distribution functions converge at every rational , to . Define
This function is nondecreasing and right-continuous: given a rational with close to , the same bound holds for when . Tightness gives and , so is a cumulative distribution function of a probability measure . For rationals , . The limits bracket and ; at a continuity point of they coincide. Thus at all such points. To obtain weak convergence, choose a compact interval with small tails and partition it using continuity points of . A bounded continuous function is uniformly continuous there, so step-function approximations and convergence of interval probabilities prove convergence of its integral. This proves Helly selection for distribution functions and supplies a weak subsequential limit .
Its characteristic function is by the already proved forward implication. We also prove uniqueness rather than assume it. If two probability measures have the same characteristic function, convolve each with a centered normal distribution of variance . Their densities are, by the Gaussian integral and Fubini's theorem,
and likewise for . The integrable Gaussian factor justifies exchanging the integrals, so these densities agree. For every bounded continuous , letting in , with a standard normal random variable, gives by bounded convergence theorem. Hence the two measures have equal integrals against all these test functions, and approximating interval indicators shows . This proves the uniqueness theorem for characteristic functions.
Every subsequence therefore has a further subsequence weakly converging to the same unique . If the original sequence failed weak convergence, some bounded continuous test function would have a subsequence of integrals staying a fixed distance from its integral under , contradicting that further convergence. This completes both directions of the theorem. Continuity at zero is essential: the laws have pointwise limiting characteristic function zero away from zero and one at zero, and are not tight.
Fix and choose the positive root
Use the Exponential martingale for Brownian motion . Before the linear-boundary stopping time , path continuity gives . At the boundary there is equality. Consequently
The optional stopping theorem at the bounded stopping time yields . Separate the two events:
The last term is at most and tends to zero. The first term converges by the monotone convergence theorem. Taking therefore proves
This bounded-stopping argument does not assume in advance that is finite. For , letting makes the right side tend to one, so it also proves . Setting gives the Brownian first-passage Laplace transform
The zero-parameter value follows from the just-proved almost-sure finiteness.

Pinned article: Introduction to the OurBigBook Project

Welcome to the OurBigBook Project! Our goal is to create the perfect publishing platform for STEM subjects, and get university-level students to write the best free STEM tutorials ever.
Everyone is welcome to create an account and play with the site: ourbigbook.com/go/register. We belive that students themselves can write amazing tutorials, but teachers are welcome too. You can write about anything you want, it doesn't have to be STEM or even educational. Silly test content is very welcome and you won't be penalized in any way. Just keep it legal!
We have two killer features:
  1. topics: topics group articles by different users with the same title, e.g. here is the topic for the "Fundamental Theorem of Calculus" ourbigbook.com/go/topic/fundamental-theorem-of-calculus
    Articles of different users are sorted by upvote within each article page. This feature is a bit like:
    • a Wikipedia where each user can have their own version of each article
    • a Q&A website like Stack Overflow, where multiple people can give their views on a given topic, and the best ones are sorted by upvote. Except you don't need to wait for someone to ask first, and any topic goes, no matter how narrow or broad
    This feature makes it possible for readers to find better explanations of any topic created by other writers. And it allows writers to create an explanation in a place that readers might actually find it.
    Figure 1.
    Screenshot of the "Derivative" topic page
    . View it live at: ourbigbook.com/go/topic/derivative
  2. local editing: you can store all your personal knowledge base content locally in a plaintext markup format that can be edited locally and published either:
    This way you can be sure that even if OurBigBook.com were to go down one day (which we have no plans to do as it is quite cheap to host!), your content will still be perfectly readable as a static site.
    Figure 5. . You can also edit articles on the Web editor without installing anything locally.
    Video 3.
    Edit locally and publish demo
    . Source. This shows editing OurBigBook Markup and publishing it using the Visual Studio Code extension.
  3. https://raw.githubusercontent.com/ourbigbook/ourbigbook-media/master/feature/x/hilbert-space-arrow.png
  4. Infinitely deep tables of contents:
    Figure 6.
    Dynamic article tree with infinitely deep table of contents
    .
    Descendant pages can also show up as toplevel e.g.: ourbigbook.com/cirosantilli/chordate-subclade
All our software is open source and hosted at: github.com/ourbigbook/ourbigbook
Further documentation can be found at: docs.ourbigbook.com
Feel free to reach our to us for any help or suggestions: docs.ourbigbook.com/#contact