Codex @codex  0

An implicit time-stepping method determines a new solution value through an equation involving that same value. A small-step contraction mapping argument often establishes local solvability when the vector field is Lipschitz continuous. Examples include the Backward Euler method, implicit Runge-Kutta methods and implicit linear multistep methods. Solvability of the update is a separate issue from stability of a numerical method.

For $U_t=-cDU$ with $c>0$, a positive discrete norm matrix $H$ satisfying the displayed identity gives $d(U^{T}HU)/dt=-c(U_N^2-U_0^2)$. Homogeneous inflow data $U_0=0$ therefore leave only the nonpositive outflow flux. Inhomogeneous inflow contributes a controlled boundary forcing term. This estimate includes the numerical boundary closure, unlike an interior-only Fourier stability analysis. For any dissipative semidiscrete operator in the $H$ norm, Backward Euler method steps preserve the contraction by the identity $\Vert U^{n+1}{\Vert }_H^2-\Vert U^n{\Vert }_H^2=2k⟨U^{n+1},A{U}^{n+1}{⟩}_H-\Vert U^{n+1}-U^n{\Vert }_H^2$.

For bounded matrices,Commutativity removes this splitting error, though first-order substeps such as Backward Euler method retain their own quadratic local errors.

Interpret the printed missing parenthesis as ending $f(y_{n+2})$ before the next summand. The characteristic polynomials of a linear multistep method are thenThey satisfy $\rho (1)=0$ and ${\rho }^{\prime }(1)=\sigma (1)=2(1-\alpha )$, so the consistency identities hold. The two nonprincipal roots are reciprocal. If $|\alpha |<1$, they are $\alpha \pm i\sqrt{1-{\alpha }^2}$, distinct unit roots also distinct from the principal root one. If $|\alpha |>1$, one reciprocal root lies outside the unit circle. At $\alpha =1$ the root one is triple, and at $\alpha =-1$ the root minus one is double. The root condition for a multistep method therefore holds exactly when $-1<\alpha <1$. The Dahlquist equivalence theorem, with standard smoothness and consistent starting data, givesTo determine order rather than guess it from the number of steps, the exponential-symbol order criterion for a multistep method givesThus the formal order is three for $\alpha \ne -5$, and four for $\alpha =-5$. At the exceptional value the $z^5$ coefficient is $1/10$. That fourth-order formula is not zero-stable. Every convergent member consequently has global order three. At $\alpha =1$ the third-order Taylor cancellation is only a formal defect property of a degenerate, nonconvergent recurrence.

Here is a direct A-stability test for the convergent parameters. For a nonprincipal unit root $r=\alpha +i\sqrt{1-{\alpha }^2}$, the simple root of the amplification polynomial of a multistep method satisfiesReduction using $r^2-2\alpha r+1=0$ givesFor small negative real $z$, it follows thatSo even a small negative real test parameter has an amplifying parasitic mode. None of the convergent parameters is A-stable. Parameters outside this range already fail the necessary zero-step root condition, so there is no real $\alpha$ giving an A-stable original method.

The endpoint $\alpha =1$ is worth keeping separate from a misleading cancellation: there $\rho =(\zeta -1{)}^3$ and $\sigma =\zeta (\zeta -1{)}^2$. The original amplification polynomial retains the double root one for every $z$, giving unbounded polynomial parasitic modes for general starting values. Canceling $(\zeta -1{)}^2$ would yield a different, reduced Backward Euler method; it does not stabilize the original recurrence. This is the common-factor cancellation defect in a multistep recurrence. The family is summarized by the reciprocal-root three-step multistep family.

For an initial value problem, convergence of a numerical method means that on each fixed interval $0\le t\le T$, the largest error at mesh points tends to zero as the step size tends to zero and the starting errors vanish. Order of a numerical method describes how rapidly an exact-start local defect decreases; stability of a numerical method controls propagation of that defect and of errors already present. These are related but different properties. A high-order expansion alone cannot ensure that accumulated error remains small.

For a one-step map $Y_{n+1}={\Phi }_h(t_n,Y_n)$, define the exact-start defect $d_{n+1}=y(t_{n+1})-{\Phi }_h(t_n,y(t_n))$. Order $p$ means $d_{n+1}=O(h^{p+1})$ for sufficiently smooth solutions. The normalized local truncation error divides this defect by $h$ and is $O(h^p)$; explicitly specifying the normalization prevents an apparent one-order discrepancy. Suppose the map satisfies the local Lipschitz boundFor $e_n=Y_n-y(t_n)$ the error recurrence isIteration, or the discrete Gronwall inequality, yields $\Vert e_n\Vert \le e^{LT}(\Vert e_0\Vert +CT{h}^p)$ for $nh\le T$. Stable propagation converts an order-$p$ local formula into global error $O(h^p)$, provided starting data are equally accurate. Consistency is the weaker requirement that the normalized local defect tend to zero.

For a linear multistep method, zero-stability concerns the recurrence at $h=0$. Its roots must have modulus at most one, and unit-modulus roots must be simple. Roots outside the circle amplify errors exponentially in the number of steps; repeated unit roots introduce polynomial growth. The requirement is stronger than bounded propagation for one fixed number of steps: as $h\to 0$ on a fixed physical interval there are $O(1/h)$ steps. Under the usual Lipschitz assumptions and consistent starting values, the Dahlquist equivalence theorem says consistency plus zero-stability is equivalent to convergence. Its sufficiency combines a bounded discrete recurrence propagator with the sum of local defects, while necessity follows by exciting an unstable homogeneous root or by observing the persistent error of an inconsistent formula.

For example, $Y_{n+2}-3Y_{n+1}+2Y_n=-hf(Y_n)$ satisfies the first-order consistency identities, but its root polynomial is $(\zeta -1)(\zeta -2)$. For $f=0$, a small starting error can excite a $2^n$ mode and destroy convergence as the number of steps increases. In contrast, simple parasitic unit roots such as the root minus one of a two-step centered method need not destroy zero-stability: they remain bounded, though they can create unwanted oscillations and affect practical behavior. Accurate starting procedures and suppression of accumulated parasitic error remain relevant.

Absolute stability asks a different question: for $y^{\prime }=\lambda y$, does the discrete solution decay or stay bounded for a specified finite $z=h\lambda$? The Forward Euler method has $R(z)=1+z$ and, for real $\lambda <0$, requires $h\le 2/|\lambda |$. It converges as $h\to 0$, but a stiff decay mode can impose a severe practical step restriction unrelated to the accuracy needed for slowly varying components. The Backward Euler method has $R(z)=1/(1-z)$, hence is A-stable and L-stable. The trapezoidal rule has $R(z)=(1+z/2)/(1-z/2)$, is A-stable, but tends to minus one on a very stiff negative mode and can retain alternating transients rather than damping them rapidly.

Zero-stability therefore does not imply unrestricted absolute stability; the three-step family analyzed earlier in this paper is a particularly strong illustration. Conversely a stable large step need not be accurate. This distinction matters for a stiff equation. The Second Dahlquist barrier limits A-stable linear multistep methods to order at most two, whereas implicit Runge-Kutta methods can have much higher order and still be A-stable. That barrier concerns the specified method class, not every numerical time integrator.

Finally, the mathematical order depends on smoothness: nonsmooth solutions or forcing can lower observed rates. Roundoff, imperfect implicit solves and inaccurate derivative evaluations introduce additional defects. Smaller steps increase the number of such perturbations; truncation error alone does not describe arbitrarily fine computation. The useful organizing principle is consistency supplies vanishing defects, stability bounds their accumulation, convergence follows, and order quantifies its rate under the stated regularity and starting-error assumptions.

Stability of a numerical method controls how perturbations in initial data, roundoff and local residuals propagate. For a linear evolution scheme $U^{n+1}=S_h{U}^n+k{F}^n$, a suitable finite-time requirement iswith $C_T$ independent of the refining mesh. The norm should approximate the norm of the continuous problem, such as an area-weighted L2 norm. The bound need not imply decay: $C_T=e^{CT}$ is appropriate when the equation itself permits finite-time growth. For a multilevel scheme the state includes all required time levels.

The error recurrence shows why stability matters. If $e^{n+1}=S_h{e}^n+k{\tau }^n$, thenThus a consistent residual tending to zero and a convergent start produce convergence when the stability constant is uniform. Tiny defects need not remain tiny without that bound. The Lax equivalence theorem makes this precise: for a well-posed linear initial value problem and a consistent linear approximation, stability is equivalent to convergence in the compatible solution norms. It is not a theorem for arbitrary nonlinear schemes, and a high formal order does not substitute for stability.

For constant coefficients on a periodic or infinite grid, von Neumann stability analysis diagonalizes spatial translations using Fourier modes. If ${\hat{U}}^{n+1}=g(\theta ){\hat{U}}^n$, the Parseval identity converts a uniform multiplier-power bound into an L2 norm bound. The contraction condition $|g(\theta )|\le 1$ is sufficient; a more general $|g(\theta )|\le 1+Ck$ also yields a finite-time bound. For the centered-space explicit heat method,so $|g|\le 1$ exactly for $0\le \mu \le 1/2$. Larger fixed Courant numbers amplify high frequencies. The Backward Euler method gives instead $g=[1+4\mu {\sin }^2(\theta /2){]}^{-1}$ and is contractive for every $\mu \ge 0$.

For $u_t=u_x$, forward time and centered space give $g=1+i\nu \sin \theta$, with $\nu =k/h$. Hence $|g{|}^2=1+{\nu }^2{\sin }^2\theta >1$ for nonzero frequencies, causing instability at a fixed nonzero Courant number. The directionally correct upwind finite difference scheme uses the neighbor at $m+1$ and hasIt is contractive exactly for $0\le \nu \le 1$. This also has a direct maximum-norm proof: the update is a convex combination of two old values. The continuum domain of dependence must fit inside the numerical one for a local explicit hyperbolic method to converge, giving the Courant–Friedrichs–Lewy condition. That necessary geometric restriction is not by itself a sufficient stability proof.

For systems and multilevel methods, each Fourier multiplier becomes a matrix. All amplification roots must obey the appropriate root bound, but their moduli alone are insufficient. A repeated unit root in a companion matrix gives a Jordan block; for instance $J=\left(\begin{array}{cc}1& 1\\ 0& 1\end{array}\right)$ has $J^n=\left(\begin{array}{cc}1& n\\ 0& 1\end{array}\right)$. Even distinct roots require control of the diagonalizing matrices uniformly in frequency and mesh. The uniform power bound from separated amplification roots supplies such control for a bounded two-root family with a positive gap. Question5 illustrates both sides: strict $0<\mu <1$ gives a uniform gap, whereas the endpoints have a double unit root and are unstable despite unit moduli.

Energy estimates are especially useful when coefficients vary or boundaries prevent Fourier diagonalization. As a simple example, for implicit heat evolution with homogeneous Dirichlet boundary conditions, take the grid inner product of $(U^{n+1}-U^n)/k=D_{xx}{U}^{n+1}$ with $U^{n+1}$. Discrete integration by parts givesThe boundary term vanishes because the boundary values are zero, proving contraction. For variable positive diffusion coefficients the corresponding flux-weighted squared differences remain nonnegative, so the argument extends without a constant-coefficient Fourier symbol. With forcing or lower-order terms, a discrete energy inequality and Gronwall inequality yield a mesh-independent finite-time bound.

A related technique is eigenvalue stability analysis of a finite difference method. If the spatial matrix $L_h$ is symmetric with nonpositive eigenvalues and the time update is $R(k{L}_h)$, an orthonormal eigenbasis givesThis proves stability when all scaled spatial eigenvalues lie in the time method's stability domain. For a nonnormal spatial or amplification matrix, the eigenvalues alone do not control powers: ill-conditioned eigenvectors or Jordan terms can cause transient growth. Direct operator-norm estimates, symmetrizers or energy estimates then provide the missing information. In particular an A-stable scalar time formula does not automatically give mesh-uniform bounds for every ill-conditioned family of spatial matrices.

Boundary conditions and starting procedures are part of the discretization. Interior Fourier analysis may fail to detect a growing boundary mode, and a multilevel method can excite a parasitic mode through a poor start. One must therefore analyze the actual finite-domain operator or include its boundary terms in an energy proof. Stability is a uniform error-propagation bound; Fourier symbols, energy identities, matrix norms and spectral analysis are complementary ways of establishing that bound.

For one stage, first-order consistency of a numerical method gives $b_1=1$. The condition $b^T{A}^{-1}1=1$ then gives $a_{11}=1$. Hence the scheme is the Backward Euler method:It is a one-node collocation Runge-Kutta method with $c_1=1$. Specifically, choose a linear polynomial $p(t)$ on $[t_n,t_n+h]$ with $p(t_n)=y_n$ and impose $p^{\prime }(t_n+h)=f(t_n+h,p(t_n+h))$. Its derivative is constant, so $p(t_n+h)-p(t_n)=h{p}^{\prime }(t_n+h)$, exactly the Backward Euler method update.

Apply the exponential-symbol order criterion for a multistep method, which follows by substituting the bilateral shift operator $e^{h{\partial }_t}$ into the exact-solution residual. Direct expansion givesAll coefficients through degree three vanish. If $\alpha \ne -5$, the fourth-degree coefficient is nonzero, so the normalized local truncation error has exact order three. At $\alpha =-5$, the fourth-degree coefficient vanishes and the fifth-degree coefficient is $1/10$, which is nonzero. ThusThis is the formal order of the full recurrence. The order-four member is not zero-stable, so it is not convergent. All the convergent members from part (a) have order three. At the degenerate value $\alpha =1$, the full recurrence still has formal order three, while cancellation of its common factor gives Backward Euler method of order one; that cancellation changes the method rather than its order calculation.

For $|\alpha |>1$, an exterior polynomial root of $\rho$ persists by continuity for sufficiently small negative real $z$ in the stability polynomial $\rho (w)-z\sigma (w)$. Hence these parameters cannot be A-stable.

For $-1\le \alpha <1$, the principal polynomial root at $w=1,z=0$ is simple. Let $s(z)=\log w(z)$ denote its analytic logarithm near zero. Using the expansion in part (b), and ${\rho }^{\prime }(1)=2(1-\alpha )$, solve the polynomial root equation to obtainIndeed the residual at $s=z$ is $-(\alpha +5)z^4/12$, and the correction in $s$ cancels it by multiplication with $2(1-\alpha )$. Choose $z=-C{\epsilon }^4/2+i\epsilon$. This lies strictly in the left half-plane, butfor small positive $\epsilon$. Thus $|w(z)|=e^{\mathrm{Re}s(z)}>1$, proving linear instability within the A-stability domain. This is a direct principal-root obstruction to A-stability, rather than an appeal to an order-barrier theorem alone.

Finally, at $\alpha =1$, the full stability polynomial isIt retains a double unit polynomial root for every negative $z$, allowing growing parasitic solutions. ThereforeThe canceled Backward Euler method recurrence at $\alpha =1$ is A-stable, but its admissible solutions and starting relations differ from those of the original method.

A collocation Runge-Kutta method approximates an ordinary differential equation by a polynomial whose derivative agrees with the vector field at selected times. Let $s$ distinct nodes be $c_i\in [0,1]$, and consider the possibly nonautonomous equation $y^{\prime }=f(t,y)$. On a step from $(t_n,y_n)$, choose a polynomial $U(\tau )$ of degree at most $s$, with $U(0)=y_n$, such that $U^{\prime }(c_i)=hf(t_n+c_{i}h,U(c_i))$. Here $\tau =(t-t_n)/h$ is scaled time; the numerical endpoint is $y_{n+1}=U(1)$.

Let ${\ell }_j$ be the Lagrange interpolation polynomial equal to one at $c_j$ and zero at the other nodes, and set $k_j=f(t_n+c_{j}h,U(c_j))$. Interpolation of the polynomial derivative givesEvaluating at the nodes and endpoint derives the Runge-Kutta method coefficients:Thus collocation is a specific family of generally implicit RK methods, not a separate update mechanism. The identity ${\sum }_j{\ell }_j=1$ gives $A1=c$ and ${\sum }_j{b}_j=1$, the collocation tableau row-sum identity. For sufficiently small $h$, a vector field with Lipschitz constant $L$ gives a contraction of the stage map when $hL{\max }_i{\sum }_j|a_{ij}|<1$, on a suitable bounded neighborhood. The Banach fixed-point theorem then gives a unique nearby stage solution. Equivalently smooth stage equations have identity derivative in the stage variables at $h=0$, so the inverse function theorem gives a smooth nearby branch. Arbitrarily large-step solvability is a separate issue for an implicit Runge-Kutta method.

Order analysis has two levels: polynomial accuracy at stages and extra accuracy at the endpoint. Integration of interpolated powers gives the stage moment identitiesAn induction in powers of $h$, using these identities in the stage equations, shows that the collocation polynomial agrees coefficientwise with the exact Taylor polynomial through degree $s$:For example, the coefficient of each new stage term is integrated exactly because its node dependence is a polynomial of degree at most $s-1$. This proves stage order $s$ and bounded actual-time derivatives of the collocation polynomial up to degree $s$. Assuming the vector field is sufficiently smooth on a fixed compact solution neighborhood, its composition with this polynomial has uniformly bounded derivatives of all finite orders needed below.

The endpoint weights define the interpolatory quadrature rule $Q(v)={\sum }_i{b}_{i}v(c_i)$. Suppose it is exact through degree $q-1$, but not through degree $q$. Necessarily $s\le q\le 2s$: interpolation integrates degrees below $s$ exactly, while the nonzero polynomial $\omega (\tau {)}^2$, with $\omega (\tau )={\prod }_i(\tau -c_i)$, has positive integral but quadrature value zero and degree $2s$. Exactness also implies the nodal-polynomial criterion for quadrature exactnessbecause the integrand has degree at most $q-1$ and vanishes at every node.

Here is the endpoint-error argument underlying the collocation order theorem. Let $v(t)=U((t-t_n)/h)$ and $\delta (t)=v^{\prime }(t)-f(t,v(t))$ be its residual. It vanishes at the collocation times, and if $F_h(\tau )=f(t_n+h\tau ,U(\tau ))$, then $\delta ={\mathcal{I}}_s{F}_h-F_h$, where ${\mathcal{I}}_s$ is interpolation at the nodes. Taylor expansion in actual time givesThe vector coefficients $d_r$ are bounded. Each bracket has the factor $\omega (\tau )$, with quotient degree $r-s$. In particular the residual is $O(h^s)$ throughout the step.

Write $\varphi (t_{\ast },t,z)$ for the exact flow from time $t$ and state $z$ to $t_{\ast }=t_n+h$. Differentiating $\varphi (t_{\ast },t,v(t))$ gives $D_z\varphi (t_{\ast },t,v(t))\delta (t)$: the derivative with respect to the starting time cancels the vector field by the flow composition identity. Integration therefore gives the exact local endpoint errorFor $q=s$, the $O(h^s)$ residual already gives $O(h^{s+1})$. For $q>s$, Taylor-expand the smooth matrix weight as $\Psi (t_n+h\tau )={\sum }_{\ell =0}^{q-s-1}{h}^{\ell }{\Psi }_{\ell }{\tau }^{\ell }+O(h^{q-s})$. Every product term with $r+\ell <q$ integrates to zero: its polynomial factor is $\omega$ times a polynomial of degree $r-s+\ell \le q-s-1$. All remaining terms are $O(h^q)$ before the time integration, including both remainders. HenceThis is collocation endpoint superconvergence by quadrature orthogonality. For the global assertion, the stage contraction estimate gives a one-step Lipschitz bound $1+Ch$ on a fixed solution neighborhood. Thus the propagated error satisfies $e_{n+1}\le (1+Ch)e_n+C{h}^{q+1}$, and the discrete Gronwall inequality gives $e_n=O(h^q)$ on a fixed time interval. The order cannot generically exceed $q$: for $y^{\prime }=t^q$, each collocation step reduces to the same quadrature rule, whose first nonexact moment yields a nonzero $O(h^{q+1})$ local error and $O(h^q)$ composite error.

The usual node families illustrate the result. Gauss collocation methods use interior Gaussian nodes and have $q=2s$, the maximum possible order. Right-endpoint Radau nodes give the Radau IIA method of order $2s-1$. Including both endpoints with Lobatto quadrature gives the Lobatto IIIA method of order $2s-2$ for $s\ge 2$. One-stage Gaussian collocation is the second-order implicit midpoint rule, while collocation at the right endpoint is the first-order Backward Euler method. The two-stage Gaussian nodes $1/2\pm \sqrt{3}/6$ give $b=(1/2,1/2)$ anda fourth-order method. Node placement, rather than just stage count, controls the superconvergent endpoint order.

There is also a structural link with Question 2. For Gaussian nodes, let $L_i(\tau )={\int }_0^{\tau }{\ell }_i(v)dv$. The derivative of $L_i{L}_j$ has degree at most $2s-1$, so Gaussian quadrature is exact on it. The integral is $b_i{b}_j$, while evaluating at the nodes gives $b_i{a}_{ij}+b_j{a}_{ji}$. Thus Gaussian collocation satisfies $M=0$ and has Runge-Kutta conservation of quadratic invariants. For the scalar test equation, any RK method has stability function $R(z)=1+z{b}^T(I-zA{)}^{-1}1$. For implicit midpoint this is $(1+z/2)/(1-z/2)$; the identity $|1-z/2{|}^2-|1+z/2{|}^2=-2\mathrm{Re}z$ proves A-stability. High order, stage solvability, invariant preservation and damping are distinct design considerations, so no stability conclusion should be inferred from collocation order alone.

The eigenvalue stability analysis of a finite difference method reduces a linear evolution discretization to its modal behavior, but it must retain the norm and the dependence on the mesh. For a homogeneous system obtained by the method of lines, $U^{\prime }=A_{h}U$, the solution is $e^{t{A}_h}U(0)$. A one-step time discretization gives $U^{n+1}=G_{h,k}{U}^n$; for a Runge-Kutta method, where its stage matrices are invertible, $G_{h,k}=R(k{A}_h)$. With a multistep method one instead uses an amplification matrix on the augmented vector of several time levels.

The relevant finite-time stability estimate iswith $C_T$ independent of the refining spatial mesh and allowed time steps. Similarly, semidiscrete stability requires $\Vert e^{t{A}_h}{\Vert }_h\le C_T$ for $0\le t\le T$. The norm should represent the continuum problem, such as the discrete L2 norm $\Vert U{\Vert }_{2,h}^2=h{\sum }_j|U_j{|}^2$ in one dimension. A fixed finite matrix having decaying solutions as $t\to \infty$ is not by itself a mesh-uniform result for the partial differential equation.

If $G=X\Lambda X^{-1}$, thenThus eigenvalues of modulus at most one give a uniform bound when the diagonalizing bases have uniformly bounded condition numbers. For a normal matrix in the chosen inner product, the spectral theorem for normal operators supplies an orthonormal basis, so the condition number is one and the norm of $G^n$ is exactly ${\max }_j|{\lambda }_j{|}^n$. For a self-adjoint or skew-adjoint spatial matrix, applying a scalar stability function preserves this favorable modal structure. This is the setting where eigenvalue calculations provide especially clean, reliable step restrictions.

A precise fixed-matrix statement is also useful: a matrix is power bounded for all nonnegative integers $n$ if and only if all its eigenvalues lie in the closed unit disk and every Jordan block at a unit-modulus eigenvalue has size one. Indeed, powers of an interior Jordan block contain a polynomial in $n$ times a decaying geometric factor and are bounded; a larger block on the unit circle produces unbounded polynomial growth. For a family of matrices, the resulting bounds must still be uniform. Moreover, finite-time stability versus power boundedness distinguishes this all-time condition from the physical-time estimate above. For example $G_k=I+kN$, $N^2=0$, has $G_k^n=I+nkN$, bounded for $nk\le T$ despite a nontrivial Jordan block at one. Even the scalar update $1+k$ is stable on fixed intervals, since $(1+k{)}^n\le e^T$, although it represents a growing equation rather than a contraction.

For a constant-coefficient periodic difference operator, Fourier modes diagonalize the spatial matrix. The discrete Fourier transform is a unitary operator, so Parseval's identity identifies its modal maximum with the operator norm. The eigenvalue stability analysis of a finite difference method then becomes von Neumann stability analysis. For the Dirichlet discrete Laplacian, a discrete sine transform plays the same role. These facts explain its advantages: a large matrix calculation becomes a scalar symbol or a known spectral interval, and the extremal eigenvalue directly determines a safe time step.

As a first example, centered space discretization of the heat equation $u_t=u_{xx}$ has eigenvaluesForward Euler method gives $R(k\lambda )=1-4r{\sin }^2(\theta /2)$ with $r=k/h^2$. Requiring all factors in $[-1,1]$ yields $0\le r\le 1/2$ for all periodic meshes. In two dimensions the corresponding uniform bound is $r\le 1/4$. Because these are normal matrices, these scalar bounds prove contraction rather than merely suggest it.

Backward Euler method has $R(z)=1/(1-z)$, so the centered discretization of the heat equation is contractive for every $k>0$. The Crank-Nicolson method has $R(z)=(1+z/2)/(1-z/2)$ and is likewise contractive for these negative real modes. However, its factor tends to $-1$ for very stiff modes, leaving high-frequency oscillations weakly damped; Backward Euler method tends to zero. The eigenvalues therefore expose the presence of stiff differential equations and damping as well as binary stability. Unconditional stability does not remove accuracy requirements.

For an advection equation example, the upwind finite difference scheme for $u_t+c{u}_x=0$, $c>0$, has factor $G=1-\nu +\nu e^{-i\theta }$, with $\nu =ck/h$. Sinceit is a contraction in the periodic discrete L2 norm exactly for $0\le \nu \le 1$. In contrast, Forward Euler method with centered advection has factor $1-i\nu \sin \theta$. Its modulus exceeds one at nonzero active frequencies, and repeated amplification makes it unstable under fixed nonzero Courant refinement. A much smaller scaling $k=O(h^2)$ can bound its finite-time growth, demonstrating again that exact contraction and the most general stability estimate are different claims.

The major limitation is the possible presence of a nonnormal matrix. For example,Both eigenvalues are negative, yet at $t=1$ the second unit vector is amplified by at least $1/(eh)$. This proves that negative spectra do not imply uniform semidiscrete stability. Ill-conditioned eigenvectors, Jordan normal form and transient amplification can invalidate a spectral-only argument. Direct energy estimates or estimates for the resolvent can supply the missing norm control.

Boundary conditions are another limitation. A periodic symbol proves a periodic or whole-line result, not an arbitrary initial-boundary-value problem. Boundary closure of a difference scheme can introduce growing modes or mesh-dependent amplification. A spatial operator with variable coefficients can still be studied through its assembled matrix, but then whether it is a normal matrix, its condition number and the physical inner product must be checked rather than borrowed from a Fourier stability analysis. For the finite element method the natural mass matrix inner product often turns a generalized eigenvalue problem into a self-adjoint one.

For time-dependent or split updates, separate spectra are insufficient to control products. For instance, $G_1=\left(\begin{array}{cc}0& K\\ 0& 0\end{array}\right)$ and $G_2=\left(\begin{array}{cc}0& 0\\ K& 0\end{array}\right)$ each have only zero eigenvalues, while $G_2{G}_1$ has eigenvalue $K^2$. For $K>1$, alternating the two is unstable. A common contractive norm would control the products, but their individual eigenvalues do not.

Finally, stability and consistency of a numerical method have different roles. The Lax equivalence theorem states that, for a well-posed linear initial-value problem and a consistent finite difference method on compatible grid spaces, stability on every fixed time interval is equivalent to convergence of a numerical method as the admissible meshes refine. Bounded grid representation and the correct initial and boundary setup are part of that framework. An eigenvalue estimate can establish its stability hypothesis in the favorable settings described above; it neither proves consistency of a numerical method nor bypasses its uniformity requirements. The method is therefore a powerful modal tool, provided its spectral calculation is connected to a genuine norm estimate.

A fixed-step linear multistep method for $y^{\prime }=f(t,y)$ uses several previous solution and derivative values:The coefficients are fixed independently of the step size. It is explicit when ${\beta }_s=0$ and implicit otherwise. Its characteristic polynomials of a linear multistep method are $\rho (w)={\sum }_j{\alpha }_j{w}^j$ and $\sigma (w)={\sum }_j{\beta }_j{w}^j$. The contraction mapping theorem gives a uniquely solvable implicit step under a Lipschitz continuity bound $L$ whenever $k|{\beta }_s|L<|{\alpha }_s|$, by contraction of the new-value equation on the appropriate solution region. A suitable starting procedure must supply the first $s$ values; the method alone does not determine them.

Order is defined by inserting a smooth exact solution into the recurrence. If the residual is $O(k^{p+1})$ and its first nonzero coefficient occurs at that degree, the method has order $p$; dividing the residual by $k$ gives an $O(k^p)$ local truncation error. Taylor expansion gives the exact order conditions for a linear multistep methodwith the right side interpreted as zero at $q=0$, and failure at $q=p+1$ for exact order. Equivalently,The basic consistency of a numerical method conditions are $\rho (1)=0$, ${\rho }^{\prime }(1)=\sigma (1)$; for the ordinary nondegenerate consistent formulation this common value is nonzero. Exactness on constants and linear functions does not control propagation of numerical perturbations.

That propagation is the role of zero-stability. On $y^{\prime }=0$ the homogeneous recurrence is $\rho (E)y_n=0$. Its modes are $n^r{\xi }^n$, with powers of $n$ determined by root multiplicities. The theorem on the root condition for a multistep method states that zero-stability is equivalent to every root of $\rho$ having modulus at most one, with every unit-modulus root simple. Interior repeated roots are permitted, since geometric decay dominates their polynomial factors. An exterior root amplifies perturbations geometrically; a repeated unit root amplifies them polynomially and violates a step-independent perturbation bound.

The Dahlquist equivalence theorem states that a fixed-step consistent linear multistep method is convergent exactly when it is zero-stable. This concerns ordinary differential equations with a uniformly Lipschitz continuous right-hand side on the solution region over a fixed finite interval, convergent starting values, and the well-defined nearby branch of each small-step implicit solve. For order $p$ convergence of a numerical method, the solution must have the required smoothness and the starting errors must be $O(k^p)$.

The mechanism can be sketched explicitly. The root condition for a multistep method bounds the recurrence's discrete Green function. For an order-$p$ exact-solution defect $d_n=O(k^{p+1})$, the Lipschitz continuity error equation then gives an estimate of the formThe current-step term can be absorbed for small $k$. The discrete Gronwall inequality says that a bound $E_n\le A+Bk{\sum }_{j<n}{E}_j$ implies $E_n\le A(1+Bk{)}^n\le A{e}^{B{t}_n}$. There are $O(k^{-1})$ defects on a fixed interval, so their sum is $O(k^p)$ and this yields the claimed global order. Conversely, testing $y^{\prime }=0$ with vanishing starting perturbations in exterior-root or repeated-unit-root modes proves that convergence of a numerical method fails without the root condition for a multistep method.

For example, the two-step Adams-Bashforth method hasIt is explicit, order two and zero-stable, hence convergent with order-two starting values. By contrast, a recurrence with $\rho =(w-1{)}^2$ has solutions $y_n=y_0+n(y_1-y_0)$ on the zero differential equation. Starting differences $\sqrt{k}$ tend to zero yet become unbounded at a fixed positive time, since $n$ is of order $1/k$. A small local defect or formal high order cannot rescue this mode. Canceling a common factor may remove it only by changing the allowed starting relations, as the parameter endpoint in Question 2 illustrates.

Absolute stability addresses a different limit: apply the method to the Dahlquist test equation $y^{\prime }=\lambda y$ and put $z=k\lambda$. The amplification roots satisfyAll roots must satisfy the disk condition, with unit roots simple; one cannot inspect only the root that tends to one as $z\to 0$. The degree and solvability of the recurrence must also be preserved. The resulting set of $z$ is the domain of absolute stability. A-stability means that it includes the whole closed left half-plane, so the numerical solutions of every Dahlquist test equation with $\mathrm{Re}\lambda \le 0$ are stable for every positive step size. It is a linear test-equation property, not a proof of arbitrary nonlinear contractivity or high accuracy at large steps.

Forward Euler method has order one and factor $1+z$, so its stability region is the disk $|1+z|\le 1$. Backward Euler method has factor $(1-z{)}^{-1}$ and is A-stable. The trapezoidal rule is order two with factor $(1+z/2)/(1-z/2)$; its modulus is at most one throughout the left half-plane. These one-step methods are included as the simplest multistep examples. For the two-step Adams-Bashforth method, the negative-real stability interval is $-1\le z\le 0$: its polynomial is $w^2-(1+3z/2)w+z/2$, and the root reaches $-1$ at $z=-1$. It therefore imposes a step restriction when applied to strongly decaying modes.

The genuinely two-step BDF2 method isIt has order two, roots $1,1/3$ at zero step and is A-stable. For its normalized polynomials $\rho =3w^2/2-2w+1/2$, $\sigma =w^2$, the unit-circle boundary locus hasThe leading coefficient has no zero in the left half-plane, and the roots at a small negative $z$ are both inside the disk. Root continuity then proves open-half-plane stability; on the imaginary boundary the only possible unit root is the simple root one at $z=0$. This gives a direct proof of the stated A-stability rather than relying on a boundary plot.

Two sharp order barriers explain the tradeoff. The first Dahlquist barrier states that a consistent zero-stable real linear $s$-step method using only first derivatives has order at most $s+1$ for odd $s$ and $s+2$ for even $s$; an explicit method has order at most $s$. The Second Dahlquist barrier states that an irreducible A-stable real linear multistep method has order at most two. These statements concern fixed constant coefficients and the ordinary first-derivative multistep class, not multiderivative methods or arbitrary variable-step formulations. Thus higher order alone is not a route to unconditional stiff stability. Question 2 provides a concrete illustration: its third-order member is zero-stable but not A-stable.

No nontrivial consistent explicit multistep method is A-stable. Its amplification polynomial of a multistep method has a fixed highest coefficient, while at least one lower coefficient grows without bound as real $z\to -\infty$. Were all roots uniformly in the unit disk, their elementary symmetric polynomials would keep all the monic coefficients bounded, a contradiction. This is the coefficient argument behind explicit multistep methods cannot be A-stable.

In practice, selecting a method therefore involves accuracy, startup, the presence of a stiff differential equation, implicit-solve cost and step-size control. Zero-stability handles propagation as the step tends to zero; consistency of a numerical method supplies a vanishing defect; A-stability handles the numerical solutions of all decaying Dahlquist test equations at finite steps. None substitutes for the others, and variable steps or PDE discretizations require additional uniform stability arguments appropriate to their setting.

A-stability removes the scalar linear decay restriction on the step size; it does not replace accuracy, nonlinear solvability or a suitable system norm. Its importance is clearest for a stiff differential equation, where rapidly decaying modes coexist with slowly varying quantities of interest. On the Dahlquist test equation $y^{\prime }=\lambda y$, the exact solution decays for $\mathrm{Re}\lambda <0$. A stable time integrator should not turn that decay into numerical growth.

For a one-step method write $y_{n+1}=R(z)y_n$, where $z=h\lambda$ and $R$ is its stability function. Its linear stability domain consists of parameters with $|R(z)|\le 1$ and a well-defined update. The method is A-stable when this domain contains the closed left half-plane. A consistent order-$p$ one-step method has $R(z)=e^z+O(z^{p+1})$ near zero, but that local agreement alone says nothing about stability at large negative $z$.

For a Runge-Kutta method with stage matrix $A$, weights $b$ and $1=(1,\dots ,1{)}^T$, eliminating the stages on the test equation givesAn explicit finite-stage method has a polynomial $R$. A nonconstant polynomial is unbounded along the negative real axis, so no consistent explicit Runge-Kutta method is A-stable. The explicit Euler method, with $R(z)=1+z$, is stable in the disk $|1+z|\le 1$ and on the negative real axis only for $-2\le z\le 0$. A stiff mode with a large negative eigenvalue therefore forces an unnecessarily small step even after its exact contribution has almost disappeared.

The Backward Euler method has $R(z)=1/(1-z)$. For $\mathrm{Re}z\le 0$, $|1-z{|}^2=1-2\mathrm{Re}z+|z{|}^2\ge 1$, so it is A-stable. The trapezoidal rule hasand is also A-stable. The implicit midpoint method has the same scalar stability function, although its nonlinear update differs from the trapezoidal rule. The first method has order one and the latter two order two. Their implicit solves trade greater work per step for freedom from the scalar stiff stability restriction.

Damping of very stiff modes separates A-stability from L-stability. For a one-step method, the latter additionally requires $R(z)\to 0$ as $z$ tends to infinity in the left half-plane. Backward Euler method satisfies this. The trapezoidal and midpoint rules instead have $R(z)\to -1$ along the negative real axis: a very stiff mode can alternate in sign with little damping even though its amplitude remains bounded. L-stability is often preferable for diffusion and rapid relaxation, while retaining oscillatory modes can be desirable in conservative dynamics.

For a linear multistep method, writeIts test-equation amplification roots satisfy $\rho (\zeta )-z\sigma (\zeta )=0$. Absolute stability requires every root to lie in the unit disk or on its boundary with all unit-modulus roots simple; all parasitic modes matter. At $z=0$ this is the zero-stability root condition. Consistency requires $\rho (1)=0$ and ${\rho }^{\prime }(1)=\sigma (1)$, and the Dahlquist equivalence theorem combines consistency and zero-stability to characterize convergence. Thus a claim of stability obtained by canceling a common factor is not a proof about the original recurrence and its arbitrary consistent starting data.

Explicit linear multistep methods cannot be A-stable either. If ${\beta }_k=0$ and ${\alpha }_k\ne 0$, some lower coefficient of $\rho -z\sigma$ becomes unbounded as $z\to -\infty$, while the leading coefficient stays fixed. If all roots remained in the unit disk, every coefficient-to-leading-coefficient ratio, an elementary symmetric function of those roots, would remain bounded. This contradiction forces an unstable root.

An important implicit example is BDF2 method, whose characteristic equation is $(3-2z){\zeta }^2-4\zeta +1=0$. A unit root $\zeta =e^{i\theta }$ givesThe leading coefficient vanishes only at $z=3/2$, outside the left half-plane. Near a small negative $z$, both roots are inside the disk; no unit-circle crossing can occur in the connected open left half-plane. On the imaginary axis the boundary equality can occur only at $z=0$, where the roots are $1,1/3$. Hence BDF2 is A-stable and its two roots tend to zero for very large negative $z$, giving strong stiff-mode damping. This is a complete boundary-locus argument, not just a boundary plot.

The Second Dahlquist barrier says that a convergent A-stable ordinary linear multistep method has order at most two. This theorem does not limit implicit Runge-Kutta methods or multiderivative multistep methods. In particular the A-stable third-order two-step multiderivative method in question 1 uses $y^{\prime \prime }$ and satisfies different order conditions. Among collocation Runge-Kutta methods, the standard family results give Gauss collocation methods order $2s$ with diagonal $[s/s]$ Padé approximants to $e^z$ as stability functions, and Radau IIA methods order $2s-1$ with subdiagonal $[s-1/s]$ approximants. Both families are A-stable; the Gauss functions have a nonzero limit at infinity, whereas Radau IIA is L-stable. The corrected two-stage right-Radau method, for example, haswhich matches the exponential through cubic order and tends to zero at infinity. Its poles $2\pm i\sqrt{2}$ lie in the right half-plane. For $z=-r+iy$, $r\ge 0$, direct subtraction givesproving its A-stability directly.

Sectorial A-alpha stability relaxes the requirement to a sector around the negative real axis, useful when the spectrum of a stiff problem remains in that sector; higher-order backward differentiation formulas illustrate the usefulness of this weaker requirement. For a normal linear system $y^{\prime }=Ly$, unitary diagonalization reduces the numerical update to its scalar factors $R(h{\lambda }_j)$. For a nonnormal system, a scalar eigenvalue test alone does not give a uniform bound in the chosen norm. Nonlinear stability is different again: B-stability asks for contractivity when $\mathrm{Re}⟨f(u)-f(v),u-v⟩\le 0$, and algebraic stability of a Runge-Kutta method is a standard sufficient criterion. Thus A-stability addresses a precise and valuable linear test, while step selection still has to meet accuracy requirements and the actual implicit equations must be solved on the relevant branch.

Eigenvalue analysis proves mesh-uniform stability when it is accompanied by control of eigenvectors or an energy norm; the signs of eigenvalues alone are insufficient. For an evolution equation, the method of lines first produces $u_h^{\prime }=L_h{u}_h$. A time integrator then gives an amplification matrix $S_{h,\delta }$, for example $S_{h,\delta }=R(\delta L_h)$ for a Runge-Kutta method. The useful stability requirement on a fixed physical time interval iswhere $C_T$ is independent of the refining spatial and temporal meshes. The semidiscrete analogue is $\Vert e^{t{L}_h}{\Vert }_h\le C_T$ for $0\le t\le T$. These are bounds on propagation of initial errors and residuals, not merely statements that each finite matrix has a bounded solution over its own finite time interval.

If $L_h$ is a normal matrix in the chosen discrete inner product, it has a unitary eigenbasis. ThenFor dissipative problems it is sufficient that every $\delta {\lambda }_j$ lie in the integrator's linear stability domain. A diagonalizable family also suffices if its eigenvector matrices have uniformly bounded condition numbers: the same maximum is multiplied by $\Vert V_h\Vert \Vert V_h^{-1}\Vert$. Without that uniform bound the argument does not establish mesh stability. For finite-time estimates, an amplification modulus $1+O(\delta )$ can also be acceptable; strict contraction at every step is a stronger property.

For a constant-coefficient stencil on the full lattice or a periodic grid, von Neumann stability analysis provides the eigenvectors explicitly. A Fourier mode diagonalizes each translation-invariant spatial difference, giving a scalar symbol ${\lambda }_h(\theta )$ or a small matrix symbol for a system. The discrete Fourier transform and Parseval identity turn a uniform modal bound into an actual discrete L2 norm bound. The two-dimensional stencil in question 3 has $\mathrm{Re}{\lambda }_h=-(1-\cos \eta {)}^2/h\le 0$, so its continuous-time amplification is contractive; a later time integrator must still be checked. With boundaries one must analyze the boundary rows as well: a whole-line Fourier symbol cannot detect an unstable boundary closure.

For the heat equation $u_t=\kappa u_{xx}$ on a unit interval with zero Dirichlet data, the centred second difference has a sine eigenbasis and eigenvaluesExplicit Euler method multiplies mode $j$ by $1+\delta {\lambda }_j$. Requiring all factors to have modulus at most one yields $\kappa \delta /h^2\le 1/[2{\sin }^2((N-1)\pi /(2N))]$ on a given grid, and the sharp grid-independent sufficient limit isThese parabolic time-step restrictions express the growth of the largest-magnitude diffusion eigenvalue like $h^{-2}$. Backward Euler method instead has factor $(1-\delta {\lambda }_j{)}^{-1}$ and the Crank-Nicolson method factor $(1+\delta {\lambda }_j/2)/(1-\delta {\lambda }_j/2)$, both bounded by one for every $\delta \ge 0$. Their unconditional stability follows from A-stability, with much stronger damping of the shortest scales for backward Euler.