Compact containment Created 2026-10-14 Updated 2026-11-05
A family of path laws satisfies Compact containment when for every finite horizon and every error tolerance there is a compact set which contains the whole path up to that horizon with at least the complementary Probability, uniformly over the family. In Euclidean state space this is the displayed condition. It prevents escape to distant regions where local coefficient approximations give no control. A nonnegative Lyapunov function with uniformly bounded generator growth often proves it by the optional stopping theorem.
Decay of a nonnegative uniformly continuous integrable function Created 2026-10-06 Updated 2026-10-18
A nonnegative function that is a uniformly continuous function on the interval and has a finite integral tends to zero. Otherwise uniform continuity gives disjoint intervals of a fixed width on which the function exceeds a fixed positive value, contradicting integrability. This supplies a useful decay argument for Lyapunov functions with merely continuous vector fields.
To make the fixed width explicit, suppose some occurs as a lower bound along a sequence with . Choose from uniform continuity for output tolerance . On every interval lying in , the function is greater than . Extract a subsequence whose centers are more than apart. Each disjoint interval contributes at least to the integral, so the first contribute at least . This contradicts finiteness of the integral. The same input width at every center is the essential step; continuity alone would allow progressively narrower peaks.
First Lyapunov theorem Created 2026-09-24 Updated 2026-09-29
A positive-definite Lyapunov function with nonpositive orbital derivative proves Lyapunov stability of the equilibrium. Indeed, fix a sufficiently small ball in the function's domain and put
Continuity and give a ball on which . Since cannot increase along a trajectory, one starting in cannot cross the sphere , on which . It therefore remains in for all forward time.
Fokker--Planck free-energy functional Created 2026-09-24 Updated 2026-11-05
For a normalized probability density and a dimensionless potential , define the Fokker--Planck free-energy functional
Assume is finite and positive, and write . Then . Nonnegativity of relative entropy proves that the unique normalized minimizer is , up to equality almost everywhere. The integrability assumption matters: a constant potential on the whole real line has no normalized density proportional to .
For a sufficiently regular solution of with symmetric positive-definite and no-flux boundaries, integration by parts gives
The free energy is thus a Lyapunov functional for the diffusion. Decrease alone is not a quantitative convergence rate; an inequality relating the energy gap to its dissipation, such as a logarithmic Sobolev inequality in the Gaussian case, supplies additional control.
Gradient flow Created 2026-09-24 Updated 2026-11-05
For a smooth real function on Euclidean space, its downward gradient flow is the ordinary differential equation . The chain rule gives
Thus is a Lyapunov function, and a nonstationary trajectory cannot return to its starting point. For , the solution is . The explicit Euler method gives , namely gradient descent. The continuous decrease does not guarantee decrease for every discrete step size; even this quadratic example requires for strict contraction.
Logistic predator-prey model Created 2026-10-05 Updated 2026-10-14
This Lotka-Volterra predator-prey model includes self-limitation of the prey through the term , with positive constants . A positive coexistence equilibrium point exists when , at , . The Lyapunov function
satisfies : its predator-prey cross terms cancel. Its sublevel sets are compact in the positive quadrant. The only invariant subset of is the coexistence equilibrium point, because staying on requires . The LaSalle invariance principle therefore proves convergence to coexistence for every positive initial state.
Lyapunov function Created 2026-09-24 Updated 2026-11-05
A Lyapunov function near an equilibrium point is a continuously differentiable scalar with , for nearby , and nonpositive orbital derivative
Its sublevel sets act as nested barriers: a trajectory cannot cross from a lower value to a higher one. The First Lyapunov theorem uses these barriers to prove Lyapunov stability without an explicit solution.
For , choose ; then away from zero and trajectories converge to the equilibrium. For the planar rotation , , the choice instead has . It proves stability, but trajectories remain on circles and do not converge.
Thus a nonpositive derivative alone does not establish asymptotic stability. Strict negativity, or an argument such as LaSalle's invariance principle controlling the invariant part of the zero-derivative set, supplies the missing information. A Lyapunov functional applies the same idea to a whole function-valued state.
A mechanical energy storage function assigns a nonnegative stored-energy quantity to the state of a mechanical model so that its increase never exceeds entering work. For a one-port passive mechanical network with entering power , the mechanical dissipation inequality is
This state-based accounting includes energy present before the measurement begins. A temporary outward power is compatible with release of stored energy.[ref]
For an ideal linear spring, is elastic potential energy. For an ideal inerter, is kinetic energy associated with terminal relative motion. An ideal dashpot need not store mechanical energy: its entering power is dissipated. If these elements share terminals, their state variables obey and
This equality proves the inequality for . In a larger network, summing element balances cancels internal transfers, leaving power through the external terminals and total dissipation.
For an original numerical check, let , and . Start at and , so . Maintain that relative velocity for : displacement reaches and stored energy becomes . Dissipation is , so input work is . Counting all entering work as newly stored energy would overestimate storage.
The mechanical available storage is the maximum net work that could be recovered under allowed future terminal motions; it is a related operational quantity, rather than automatically every form of internal physical energy. A model's storage function is a certificate that must respect its accessible states and port variables. With zero external power, a positive-definite storage function can supply a Lyapunov function, but nonincreasing energy alone does not establish that every motion decays to rest: a lossless spring–inerter oscillation keeps a constant nonzero stored energy.
Orbital derivative Created 2026-09-29 Updated 2026-11-05
The orbital derivative of a differentiable scalar along an autonomous differential equation is
The chain rule identifies it with . It measures change along a trajectory, rather than the size of the gradient alone.
For and the rotation field , it is even though the gradient is nonzero away from the origin. For the contracting field , the same function has orbital derivative . This directional distinction is what makes an orbital derivative useful in a Lyapunov function or a first integral.
For a unit-jump population chain with one-step probabilities and , its equilibrium density is . Speed time by and center and scale the population as . The accelerated conditional drift is , and the second moment is . Their compact-set limits give the displayed Ornstein-Uhlenbeck process. If , the invalid-probability threshold stays a fixed positive density distance above equilibrium. Before that threshold, the generator of is at most one because its drift contribution is nonpositive. The stopped Lyapunov function bound proves compact containment and makes the probability of hitting that threshold on a fixed accelerated horizon tend to zero. This supplies the global hypotheses of the diffusion approximation theorem.
Since the exponential factor is positive, equilibria must have and . Thus . At either point the Jacobian matrix is . At its eigenvalues are , so this is a saddle point. At they are : the linearization has a centre equilibrium.
The requested first-order approximation follows by putting , : and . The linearized system conserves , giving the approximate ellipses
However, the printed nonlinear system has a weak repelling focus at , rather than a centre. A centre in the linearization need not be a nonlinear centre. Here that distinction can be proved explicitly, without relying on the numerical drift in the figure.
Put , , and . The weak focus of an exponential polynomial planar system has the local positive function
Expanding the vector field through cubic order and differentiating gives
The cubic correction cancels the cubic terms in the derivative of , and the quartic correction leaves the positive radial term . Since is positive and the angular velocity stays near , small nonzero trajectories spiral outward. Equivalently the reversed system has a strict local Lyapunov function and an asymptotically stable equilibrium. In particular, a periodic orbit sufficiently near the equilibrium is impossible, because would have to increase strictly and return to its starting value. This proves the failure of the printed centre assertion while retaining its correct linearized ellipse calculation.
Figure 1.
The linearized ellipses hide the nonlinear outward spiral
. Original fourth-order Runge-Kutta integration of the printed vector field, compared with its linearization from the same initial point. The local polynomial calculation, rather than the plot, proves the weak instability. Details.
The Bendixson-Dulac theorem states that if is a continuously differentiable planar vector field on a simply connected open region , and a continuously differentiable Dulac function makes
have one weak sign throughout , without vanishing identically on any nonempty open subset, then there is no nonconstant periodic orbit contained in . To see why, a hypothetical simple periodic orbit bounds a region in . The divergence theorem gives the integral of this expression as the outward flux of across the orbit. The flux is zero because the vector field is tangent to its orbit, whereas the one-sign area integral is nonzero, a contradiction. The usual stricter version assumes a strictly positive or strictly negative expression everywhere.
For the printed polynomial system, the origin is an equilibrium and
Thus its linearization is a centre equilibrium. For the full system take . The divergence is
On the open disk this is nonpositive, and is negative wherever . Its zero set inside the disk is a line, so it cannot vanish on any open subset. The Bendixson-Dulac theorem therefore proves
This also exposes a terminology problem in the source: a nonlinear centre equilibrium has surrounding closed orbits, so the printed claim of a centre cannot hold in that sense. The imaginary eigenvalues establish only a centre of the linearized system. In fact the full equilibrium is an asymptotically stable weak focus, as the following direct check shows.
A quartic Lyapunov correction for a weak focus makes the nonlinear damping visible. Put and consider the Lyapunov function
It is positive definite sufficiently close to zero. Differentiating along the polynomial vector field gives
For clarity, the quartic terms come from in the derivative of , and from applying to the quartic correction. Their sum is ; all other terms have degree at least six. Therefore in a sufficiently small punctured neighbourhood, proving asymptotic stability by Second Lyapunov theorem. Finally
so nonconstant nearby orbits turn clockwise as they converge. The equilibrium is a weak focus despite the centre in its linearization. This interpretation resolves both the local dynamics and the absence of closed orbits without changing the printed coefficients.
At an equilibrium point, and , so . With , the two equilibrium points and their Jacobian matrix are
At , . The trace-determinant stability criterion therefore gives an unstable saddle equilibrium. At the origin, the eigenvalues are : its linearization is a center equilibrium. The linearization stability theorem does not classify this nonhyperbolic equilibrium of the nonlinear system.
Nonlinear classification. In fact the answer depends on :
Here is an explicit derivation, rather than an inference from the imaginary eigenvalues. When , set . Near the origin , and the first integral
has
Its Hessian matrix at the origin is the identity. Small level curves are closed ovals on which the vector field is nonzero, proving the center equilibrium classification and Lyapunov stability.
For , a higher-order Lyapunov function calculation settles the issue. Put , , , , and . Then
Write and introduce the differential operators
Successively cancel the terms of degrees three, four and five in the derivative of , using these homogeneous polynomials:
Direct differentiation verifies , and . The remaining degree-six term has circular average
For example, the only nonzero averages of degree-six monomials needed here are and . Since in polar coordinates, the zero-average part can be cancelled by a degree-six homogeneous polynomial satisfying
Indeed its restriction to the unit circle is obtained by integrating a zero-average trigonometric polynomial; its Fourier modes have orders , so multiplication by gives a polynomial. Consequently the positive Lyapunov function satisfies
For this is strictly negative near the origin, giving asymptotic stability. For it is strictly positive, and in a sufficiently small neighbourhood; every nonzero starting point must eventually leave that neighbourhood. Finally , so the nearby trajectories rotate: these are weak foci, with the stability stated above. Merely calling the nonlinear origin a center for every would miss this distinction.
The term represents prey reproduction and limits growth by available food, giving carrying capacity in the absence of hunters. The loss represents encounters between prey and hunters. For hunters, is prey-supported growth and is mortality; their per-capita growth changes sign at prey population . The coordinate axes are invariant, and positive initial populations remain positive.
For , the physical equilibrium points are the origin, , and when . The Jacobian matrix is
At its eigenvalues are and , so the origin is always a saddle equilibrium, attracting along the hunter-only axis and repelling along the prey-only axis. At the eigenvalues are
It is a saddle equilibrium for and an asymptotically stable node for .
For , the coexistence Jacobian matrix is
Its trace is negative and determinant positive. Its discriminant is negative, so is an asymptotically stable focus. Interior trajectories approach it in damped, counterclockwise oscillations in the plane. Both populations ultimately coexist.
To justify the global interior fate for this logistic predator-prey model, let and use the Lyapunov function
It is nonnegative and its sublevel sets are compact inside the positive quadrant. Differentiation gives the explicit cancellation
The only invariant subset on which this vanishes is : remaining at requires . LaSalle's invariance principle therefore proves that every strictly positive initial state converges to this coexistence point.
For , the formal coexistence point has negative and is outside the physical state space. The only physical equilibria are the saddle equilibrium at the origin and the stable prey-only node. Since , positive prey has . Eventually with a fixed margin, so hunter population decays exponentially. When is sufficiently small, comparison with gives for every . Thus and for positive initial prey: hunters die out and prey reaches carrying capacity.
On the boundary, hunter-only initial data approach the origin; prey-only positive data approach in either parameter range. The sketches show representative values in the two requested ranges. Dashed and dotted lines are the prey and hunter nullclines; filled dots are attractors and open dots are saddle equilibria.
Figure 1.
Predator-prey phase portraits showing stable coexistence for b=0.25 and hunter extinction for b=1.5
. Details.
Write . The admissible equilibrium points are , when , and when . The Jacobian matrix is
At the eigenvalues are and ; at they are and . At a positive coexistence equilibrium point,
It is therefore asymptotically stable. It is a node for , a focus for the reverse inequality, and has repeated real eigenvalues at equality. The node–focus change does not alter the attracting equilibrium point's topology.
There are three open physical parameter regions:
  • : extinction. Only is admissible and it is a sink.
  • , : predator extinction. is a saddle, and is a sink.
  • : coexistence. and are saddles on the boundary, while is a sink.
The axes are invariant. On , decays exponentially; on , prey obey the logistic differential equation. Positivity follows because each right-hand side has its population as a factor. Also is bounded by logistic comparison, while satisfies , so all forward trajectories are bounded. The Dulac function has divergence in the interior, excluding periodic orbits by the Bendixson-Dulac criterion.
In the extinction region, , and then is eventually negative. In the prey-only region, the upper bound implies predator decay and convergence of interior prey to . In the coexistence region the Lyapunov function
has . Its positive-quadrant sublevel sets are compact, and the only invariant subset with also has . The LaSalle invariance principle proves convergence of every interior trajectory to . These facts justify the complete qualitative phase portraits, not just their local arrows.
At , and meet in a transcritical bifurcation; positive prey decay algebraically. At with , and exchange stability. On its centre manifold, gives , so the physically positive side is still attracting. There is no interior Hopf bifurcation since the coexistence trace never vanishes.
Figure 1.
Three continuous predator–prey phase portraits and their parameter regions
. Details.
A fixed point is Lyapunov stable if for every there is such that solutions starting within of stay within for all forward times. It has quasi-asymptotic stability if some neighborhood of consists of initial points whose solutions converge to . Asymptotic stability requires both properties.
Choose the Lyapunov function
It vanishes only at the origin, is positive elsewhere, and has compact sublevel sets. The vector field is polynomial, hence locally Lipschitz. Along a solution,
A small sublevel set is forward invariant and can be fitted inside any prescribed neighborhood, proving Lyapunov stability; boundedness also prevents finite-time escape.
For attraction, a bounded forward solution has a nonempty compact limit set, and decreases to a limit. Every point in that limit set has that same value. Continuous dependence of solutions makes the limit set forward invariant, so along every trajectory in it. Such a trajectory must have at all times. But then , forcing . Thus the only possible limit point is the origin, and every solution converges there. This verifies the invariant-set condition behind the LaSalle invariance principle, rather than incorrectly treating as strictly negative at every nonzero point. The origin is therefore asymptotically stable, indeed globally attracting.
For the symmetric zero-diagonal weights and asynchronous rule just specified, update only unit . In the energy difference, terms involving neither nor its symmetric partner cancel. Thus asynchronous Hopfield energy descent gives
If the unit was already aligned, it does not change and . If it flips into alignment, , so
At our tie rule retains the old activity. The network has only configurations. Strict descent on every actual change prevents revisiting a configuration and permits only finitely many changes. With a fair schedule every misaligned unit is eventually updated, so the terminal configuration is a fixed point. At that point each possible single-bit flip has energy cost . The network reaches a minimum against single-unit changes, not necessarily the global minimum. Its energy is a Lyapunov function for these asynchronous dynamics.
The qualifications are essential to the printed claim. Synchronous updates can cycle: for two units with , the state alternates with , keeping energy . Symmetry alone therefore does not prove convergence under simultaneous updates.
Nor does the valid descent proof guarantee a global minimum. Consider four units with and all four cross-group weights equal to , symmetrically and with zero diagonal. The state has aligned fields and energy , so it is a strict single-flip local minimum. The all-positive state has energy . This explicit counterexample shows why local minima may trap retrieval even though the energy always descends under the stipulated rule.
Choose in the proposed Lyapunov function. The cubic terms cancel and direct differentiation gives
For this is negative definite, while is positive definite and radially unbounded. Finite-dimensional equivalence of these positive quadratic forms gives for some . Thus , every solution remains bounded and exists for all forward times, and every solution tends to zero. Therefore the origin is globally asymptotically stable, indeed exponentially stable.
A random access network shares one communication channel among an unbounded collection of possible transmitters. Each packet occupies one slot; exactly one attempt succeeds, an idle slot carries nothing, and a collision delivers none of its attempted packets. Assume perfect slot synchronization, immediate idle/success/collision feedback, no channel errors, and independent random choices by packets. The important quantities are throughput, the size of the retransmission backlog, and delay. A good one-slot success probability alone does not establish stability of the backlog.
For a precise infinite-population slotted ALOHA model, let be independent Poisson random variables of mean , representing new packets which attempt immediately in slot . Let be the backlog before those arrivals. Each old packet independently retries with a constant probability , so conditional on , its retry count is , independent of . If records a successful slot, then
This includes a new packet in the backlog precisely when it has not succeeded; an old packet is removed precisely when it succeeds. The total backlog is a Markov chain in this constant-probability model. Its idle and success probabilities are
with the second term in interpreted as zero. A success may be either one new packet and no old attempt, or one old attempt and no new packet.
The fluid approximation of slotted ALOHA replaces the binomial retry count by a Poisson count when there are many packets with small retry probability. If is total attempt intensity, the approximate success probability is . Differentiation gives
This is the limiting ALOHA throughput bound, not an exact upper bound for every finite packet population: one isolated packet which certainly transmits succeeds with probability one. In a fluid model, when the equation has two roots. The root below one is locally stable for constant , while the root above one is unstable, because the drift derivative is . These approximations explain a low-backlog operating region and a congestion threshold; they do not imply a stationary law for the infinite-population chain.
Indeed, for fixed , the exact success probability tends to zero as , so the exact drift tends to . One can prove instability, rather than relying only on this positive drift. Fix and use . Since is zero or one,
The first term is strictly less than one and the second tends to zero. Hence is a nonnegative supermartingale outside a sufficiently large finite set, where is the first entrance into that set, say . The optional sampling theorem for a supermartingale, first at bounded stopping times, gives
This is an exponential-supermartingale escape bound. The chain is irreducible: a no-arrival slot with exactly one old retry can reduce any positive state by one, and Poisson arrivals can reach arbitrarily high states. Therefore the escape probability proves transience and excludes a positive recurrent Markov chain, for every and fixed . A locally stable fluid operating point can consequently coexist with eventual stochastic congestion. At , once at least two packets are backlogged, they collide forever, so constant unit retry probability also fails.
A state-dependent retry rule removes this fixed- congestion mechanism. As an ideal benchmark with exact backlog information and known arrival intensity , use
This is backlog-aware ALOHA with arrival-adjusted attempt probabilities. Its total attempt intensity approaches one: the mean old-attempt count is and the new-attempt count has mean . Substitution in the exact success formula yields
Thus the drift tends to . For the Lyapunov function , the drift is bounded above by a negative constant outside a finite set, and the expected one-step backlog is finite at every state. Foster theorem therefore gives positive recurrence. This proves a genuine stochastic stability result for this specified, informed controller. Real transmitters generally observe feedback rather than the exact backlog; estimating the backlog and using a probability of order its reciprocal motivates adaptive ALOHA, but an estimate's dynamics must be included in the state and analyzed before claiming the same stability result.
A different nonconstant rule uses a packet's own collision history. In geometric-attempt exponential backoff, a packet at stage attempts with probability , where . A collision advances every attempted packet by one stage, a success removes its packet, and new arrivals enter stage zero. For definiteness, in this version arrivals join at the end of a slot and first become eligible in the next slot. The full Markov chain state is the sequence of stage populations; the total backlog alone is insufficient. Conditional on that state,
These formulas follow from independent attempts and selecting which one packet succeeds. They reduce approximately to for many small probabilities with . The retry probabilities decrease after failure, preventing repeated aggressive attempts by the same packets. Binary exponential backoff instead chooses uniform random counters from expanding finite windows; that counter model requires residual counters as well as collision stages in its state and is not identical to geometric waiting.
Backoff also exposes a distinction between eventual delivery and acceptable delay. In the simplifying model where each attempt independently fails with a fixed probability , the probability of reaching stage is and its mean waiting time is . The mean delay under independent exponential-backoff failures is therefore
Although a packet succeeds almost surely whenever , its mean delay can be infinite. In an actual random access network, failures are correlated with the other stage populations, so this fixed- calculation is a diagnostic approximation, not a proof of network stability. These models show why random-access analysis must track the retry rule, feedback information and full state, and assess throughput, stability and delay separately.
A Lyapunov function near an equilibrium is a continuously differentiable function with , for nonzero nearby , and along trajectories there. Strict negativity away from the equilibrium gives local asymptotic stability. Lyapunov stability itself means that every prescribed small neighborhood contains a smaller neighborhood whose forward trajectories never leave the prescribed one.
For the proposed quadratic form, direct differentiation gives
When , the quadratic part is negative definite, so it dominates the cubic remainder in a sufficiently small ball; is then a strict Lyapunov function. If , take to make the leading derivative positive. At , that same line gives , positive on its negative half. Values fail positive definiteness of . Therefore
For the full derivative factors, with no small-amplitude truncation:
Every closed sublevel set lies in , is forward invariant, and has strictly decreasing away from the origin. Compactness and the Lyapunov stability argument therefore show that the ellipse belongs to the origin's basin of attraction. Boundary points with other than also have negative derivative and enter the ellipse. The exception is another fixed point, so the largest such sublevel ellipse cannot be included wholesale in the basin. This is a basin estimate from a Lyapunov sublevel set, not a claim that no additional points outside the ellipse converge to the origin.