A-alpha stability 2026-10-14
A numerical time-stepping method is -stable if its absolute stability region contains the negative-axis sector , for . With a stability function , this requires throughout that sector. At , this is A-stability.
Adaptive Euler step doubling for exponential decay Created 2026-10-14 Updated 2026-10-14
Consider the nonstiff scalar initial value problem , , on . Its exact solution is the exponential decay . Use step-doubling error estimation with the explicit Euler method: from the same numerical starting value , form one coarse step and two half steps giving :
Accept when the absolute and relative error tolerances test passes, with . There is no Richardson extrapolation in the accepted value. Both approximations have order of a numerical method one, while this local error estimator has step exponent . Consequently the adaptive step-size controller uses a square root, not a cube root:
Start with . A rejected trial retains both and , caps at one, and retries. A zero estimate uses . Before every trial, cap at . Stop with a failure if floating-point arithmetic gives , or after 100000 trials. These rules specify the example completely; they are not a prescription for every ordinary differential equation solver.
Figure 1.
Adaptive two-half-step Euler decay: exact and numerical trajectories, accepted step sizes and rejected trials
. Details.
The first two trials, and , have and and are rejected. The next trial has and and is accepted. For positive decreasing numerical values, , so the local error estimator gives
When the relative term dominates, this suggests nearly constant accepted steps, approximately . When falls below the crossover size , the absolute term increasingly permits longer steps. The sharp final decrease in only lands on ; it does not indicate a sudden loss of smoothness.
The actual signed local defect from the same numerical input is
Here is Big O notation, obtained from Taylor's theorem for the exponential function. Thus agrees with the leading term as , but is not identical to . At , , the estimate is while the actual local defect is . Their closeness improves for the smaller accepted steps. Had we added using Richardson extrapolation, the accepted formula would instead be , of order of a numerical method two; the same raw difference would still scale as and would not estimate that improved value's leading defect.
Figure 2.
Estimated versus actual local Euler defects, rejected trials, and actual accumulated global error scaled by the local tolerance
. Details.
The global error obeys an exact recurrence relation involving the actual local defects:
Each previous global error contracts, yet the local defects keep adding. In the plotted run, 101 steps are accepted and two trials are rejected. Every accepted is at most approximately , while the largest is approximately . At , the numerical value is approximately , compared with the exact , giving absolute error approximately . The right-hand plot deliberately scales actual global error by the same used for local acceptance; it is not a new global error tolerance test.
Tightening both absolute and relative error tolerances by a factor ten gives 315 accepted steps and endpoint absolute error approximately ; tightening both by a factor one hundred gives 991 accepted steps and endpoint absolute error approximately . This is a convergence comparison for this particular initial value problem. The endpoint error decreases by about a factor three per tenfold tightening, rather than by ten: a first-order Euler method accumulates defects across many steps, while its local error estimator varies quadratically with . A variable-step global ODE error bound needs propagation information as well as accepted local tests.
The absolute stability boundary remains a separate constraint. On the Dahlquist test equation with , the accepted two-half-step formula has stability function . It strictly damps perturbations only for . Its local error estimator is , which can pass an absolute-tolerance test at tiny even outside this damping interval. In a stiff two-mode linear system, a fast mode can therefore constrain the step long after its visible amplitude has become small. An adaptive step-size controller does not turn an explicit Euler method into an A-stable method. Official solver documentation likewise distinguishes explicit Runge-Kutta methods for nonstiff problems from implicit Runge-Kutta methods and backward differentiation formulas for stiff equations.[ref]
Amplification polynomial of a multistep method Created 2026-10-05 Updated 2026-10-14
Applying a linear multistep method to the Dahlquist test equation with produces a constant-coefficient recurrence whose modes satisfy . Absolute stability requires the resulting roots of a polynomial to lie in the unit disk, with unit-modulus roots simple. Thus multistep stability depends on every recurrence mode, rather than a single one-step stability function.
A-stability of a symmetric two-stage implicit Runge-Kutta family Created 2026-09-24 Updated 2026-09-28
For the two-stage implicit Runge-Kutta method with
the stability function is
For ,
The denominator has no zero in the closed left half-plane, so the method is A-stable for every real .
For nonnegative degrees with , the Padé approximant to the exponential function has no poles in the left half-plane and satisfies for . Consequently a numerical method having that stability function is A-stable. In particular, the near-diagonal choice gives the Radau IIA method stability function, while corresponds to the Gauss collocation method. The theorem concerns these exponential approximants, not arbitrary rational approximations with the same degrees.
Classical fourth-order Runge-Kutta method Created 2026-10-05 Updated 2026-10-06
The classical four-stage explicit Runge-Kutta method uses
It satisfies the fourth-order conditions for a Runge-Kutta method. Its polynomial stability function excludes A-stability, while gives the imaginary-axis stability interval .
Dahlquist test equation Created 2026-09-28 Updated 2026-11-05
The Dahlquist test equation is the scalar linear initial-value problem . A one-step method reduces it to , thereby defining its stability function .
Euclidean logarithmic norm Created 2026-10-06 Updated 2026-10-07
The Euclidean logarithmic norm is the largest eigenvalue of a matrix's Hermitian part. Differentiating along gives . It is the least exponent for such a bound with prefactor one, as a first-order expansion at zero shows. It equals the spectral abscissa for a normal matrix, but need not do so otherwise. A scalar rational stability function does not generally inherit a bound by its value at this real number.
Euler method Created 2026-10-14 Updated 2026-10-20
The Euler method approximates an initial value problem , , by evaluating the derivative at the current point and moving along that tangent line for a time . With , its displayed update needs one evaluation of per step. This simple rule turns local slope information into a whole approximate trajectory. MIT's course notes derive it by dropping the second-order Taylor remainder.[ref]
Take , , whose exact solution is the exponential function . The Euler method gives
At , the first values are . At , five steps give , compared with . The result decays but is too small: absolute stability alone does not ensure accuracy. At , the first values are , and at five steps give , while the exact solution is about .
Figure 1.
Euler approximations to exponential decay on the same finite time interval
. Details.
For the Dahlquist test equation , the Euler method has stability function with . Its absolute stability region is , the closed disk centred at with radius one. For the present real decay problem and , bounded numerical values require ; strict decay requires . At the values alternate without decay. These conditions follow directly from the amplification factor , matching MIT's criterion for .[ref]
The Euler method is first order: with bounded second derivative and suitable Lipschitz continuity, its global error is on a fixed interval. In this example the logarithm gives for small positive . At , this means . It explains both the downward bias and why halving the step roughly halves the error once is sufficiently small.
Fourth-order two-stage Gauss collocation method Created 2026-10-07 Updated 2026-10-14
The two-stage Gauss--Legendre Runge-Kutta method uses nodes , weights and coefficients , , . The Butcher order conditions give order four. Its stability function is the displayed rational function, which is A-stable. The algebraic stability matrix is zero, so positive weights also give nonlinear contractivity for dissipative vector fields.
Lobatto IIIA method Created 2026-10-05 Updated 2026-10-14
A Lobatto IIIA method collocates at the Lobatto quadrature nodes, including both endpoints of the time interval. Its three-stage version has
The Butcher order conditions give order four. Its stability function is
which is A-stable but not L-stable. It is not algebraically stable: the first diagonal entry of the defining matrix is . This distinguishes linear A-stability from nonlinear algebraic stability.
L-stability Created 2026-09-28 Updated 2026-10-06
An A-stable one-step method is L-stable when its stability function also satisfies as within the left half-plane. It damps modes whose decay time is far shorter than the numerical step.
Maximal-order Runge-Kutta methods are A-stable Created 2026-10-14 Updated 2026-10-14
An -stage Runge-Kutta method has stability function , a rational function with numerator and denominator degrees at most . Order forces , so uniqueness of the diagonal Padé approximant identifies with . That approximant has no poles in the closed left half-plane and modulus at most one there. Hence every such maximal-order method is A-stable. Since the Padé approximant numerator and denominator are coprime and both have degree , the original stage determinant cannot have hidden canceled poles in that half-plane.
Suppose an -stage Runge-Kutta method has order , an invertible matrix , and . Its stability function is . The adjugate matrix formula gives a rational function with denominator degree and numerator degree at most ; its zero limit at infinity lowers the latter degree to at most . The order of a numerical method gives , identifying the Padé approximant. Uniqueness follows because the cross-multiplied difference of two candidates has degree at most but a zero of order at least , so is identically zero. The A-stability of near-diagonal exponential Padé approximants then proves A-stability, and the zero limit proves L-stability. These conditions identify the stability function; they do not alone claim uniqueness of all entries of the Runge-Kutta method tableau.
Let be the coefficients of the -stage Runge-Kutta method. On the Dahlquist test equation , its stability function is
Both polynomials have degree at most , with denominator normalized to one at zero. Order implies . A rational function with these degree bounds and this accuracy is uniquely the diagonal Padé approximant . To see uniqueness directly, if and have the required accuracy, then is a polynomial of degree at most vanishing to order at least at zero, hence identically zero.
The diagonal exponential Padé approximant can be written
Use the permitted Padé property that the zeros of lie strictly in the right half-plane and numerator and denominator are coprime. Thus has no poles in the closed left half-plane. Its real coefficients give , so . Also at infinity. Applying the maximum modulus principle on left half-disks and letting their radii increase proves for . There are no hidden stage poles there either: the irreducible Padé denominator has degree , exhausting the possible degree of , so the latter cannot contain an additional canceled factor. Every such maximal-order Runge-Kutta method is A-stable. This is maximal-order Runge-Kutta methods are A-stable; it does not assert L-stability, since the stability function does not tend to zero at infinity.
The original PDF has a sign defect in the lower-left stage coefficient. Put . With the printed negative sign, the row sums and the relevant Butcher order conditions are
Thus the claimed universal second order is false. For example gives, on , two stages equal to and the update . This is the first-order Forward Euler method. For the literal tableau, order at least two holds exactly when or , subject to distinct allowed nodes.
For completeness the literal tableau's A-stability can also be classified algebraically. Define
The stability function obtained from is
If , the denominator has no zero in the closed left half-plane: its reciprocal roots are the stage-matrix eigenvalues, whose positive sum and nonnegative product give positive real parts for nonzero eigenvalues. On the imaginary axis,
Consequently for the literal tableau with uniquely solvable stages throughout the left half-plane, the complete parameter criterion is
These explicit inequalities in are necessary by the small and large imaginary parameters, and sufficient by the maximum modulus principle, including the bound at infinity. For the last inequality forces , leaving the stable linear-fractional case. For there is a negative real stage pole. If one defines stability solely through an analytically continued scalar stability function, the only additional possibilities are
Indeed these are precisely the cases where the left pole cancels and is an A-stable theta method. The original stage equations are nevertheless singular at , so they do not give uniquely determined stages there. This distinction prevents a canceled rational factor from concealing an ill-defined stage solve.
The natural intended repair is to replace just the lower-left coefficient by . This makes the method the collocation Runge-Kutta method on the two nodes. To derive the repaired entries, take
Integration gives the other three printed entries and the weights unchanged, with the positive lower-left entry. Since and , it follows that , and . Hence the corrected family is always of order at least two.
For the corrected family let , and . Its stability function has the same displayed rational form, now with . The denominator again has no left-half-plane zero, and on the imaginary axis the squared-modulus difference is . Because , the necessary and sufficient inequality is , or
For sufficiency, on the imaginary axis and at infinity; apply the maximum modulus principle on growing left half-discs to get the same bound inside. If , the condition leaves the endpoint trapezoidal rule and its linear denominator. If , every nonzero imaginary argument gives . This is the two-node collocation A-stability criterion, explicitly distinguished from the literal defective tableau.
Introduce one stage with , and update . This is the implicit midpoint rule as a one-stage Runge-Kutta method, with Butcher tableau
For a smooth vector field, write . Expanding its stage equation gives
which matches the exact Taylor expansion through degree two. The method therefore has order at least two. On the Dahlquist test equation , its stability function is
The third coefficient differs from the exponential's , so the order is exactly two.
There is no pole in , and
Thus throughout the closed left half-plane:
This conclusion is A-stability, not strong damping of arbitrarily stiff modes: along the negative real axis.
Apply the Runge-Kutta method to the Dahlquist test equation and put . Its stability function is
Since is an invertible matrix, has polynomial degree . The adjugate matrix formula gives , but the additional condition gives
so actually . The specified order of a numerical method implies near zero. Therefore the Padé identification of a Radau stability function makes the Padé approximant to the exponential function: its numerator and denominator have the required degrees, and its Taylor series agrees through degree .
In the permitted A-stability of near-diagonal exponential Padé approximants, take , . Then , so the scheme is A-stable. The same stability function also tends to zero at infinity, giving L-stability.
The nodes are . Their Lagrange interpolation basis on the normalized step is
Integrating these polynomials from zero to each node gives . At the row vanishes; at the integrals are ; at they are . The endpoint integrals also give . These are exactly the coefficients in the original PDF.
Indeed, the degree-three polynomial determined by and satisfies . Integrating it to the nodes gives precisely the stage equations, and integrating to the endpoint gives the update. This proves it is a collocation Runge-Kutta method, specifically the three-stage Lobatto IIIA method.
For an explicit order proof, the collocation integrals imply , and , with powers componentwise. The endpoint weights are Simpson weights, so for . Therefore the eight Butcher order conditions through order four are satisfied:
For example the last expression is , because . These conditions are obtained by matching all elementary differential terms through degree four in the exact and numerical Taylor expansions, not just by testing a scalar linear equation. Nonautonomous equations are included by adjoining .
To exclude higher order, the scalar stability function computed below satisfies . The local defect is therefore nonzero at degree five even on the scalar test equation. Hence .
Applying the stage equations to gives the stability function
The poles are , both in the right half-plane. To determine the whole domain rather than only its boundary, multiply numerator and denominator by 12 and put . Direct expansion gives
Thus exactly when , with no pole in that set. Consequently
On the imaginary axis , and in the open left half-plane it is less than one. Its limit along the negative real axis is one, so A-stability here does not imply stiff decay or L-stability.