Skip to content

Characteristic Polynomial for a Difference Equation


Definition

For the homogeneous linear difference equation of order kk with constant coefficients and a0≠0a_0\ne0,

xt+k+ak−1xt+k−1+⋯+a1xt+1+a0xt=0,x_{t+k}+a_{k-1}x_{t+k-1}+\cdots +a_1x_{t+1}+a_0x_t=0,

the characteristic polynomial is

p(r)=rk+ak−1rk−1+⋯+a1r+a0,p(r)=r^k+a_{k-1}r^{k-1}+\cdots+a_1r+a_0,

and p(r)=0p(r)=0 is the characteristic equation. Its roots r1,…,rkr_1,\ldots,r_k (real or complex, counted with multiplicity) are the characteristic roots.


Interpretation

Substituting the trial solution xt=rtx_t=r^t with r≠0r\ne0 gives rt p(r)=0r^t\,p(r)=0. So the geometric sequence rtr^t solves the equation exactly when rr is a characteristic root, and the problem of solving a recurrence reduces to finding the roots of one polynomial.

Left, the roots e^(i pi/3) and e^(-i pi/3) of r^2 - r + 1 on the unit circle at angle pi/3; right, the solution 1, 1, 0, -1, -1, 0, 1, ... as stems repeating with period 6.

The figure takes the complex-roots example xt+2−xt+1+xt=0x_{t+2}-x_{t+1}+x_t=0. The roots of p(r)=r2−r+1p(r)=r^2-r+1 lie on the unit circle at the angles ±π/3\pm\pi/3 (left). Because ∣r∣=1|r|=1 the solutions neither grow nor shrink, and because each step turns by π/3\pi/3 they return after 66 steps: the solution from x0=x1=1x_0=x_1=1 (right) is 1,1,0,−1,−1,01,1,0,-1,-1,0 repeated.

The polynomial is read off the equation by replacing xt+jx_{t+j} with rjr^j. Its coefficients are those of the equation written with every term on the left; with xt+2=xt+1+xtx_{t+2}=x_{t+1}+x_t one must first rewrite xt+2−xt+1−xt=0x_{t+2}-x_{t+1}-x_t=0 to get p(r)=r2−r−1p(r)=r^2-r-1.

The name is not a coincidence. The equation is equivalent to yt+1=Cyt\boldsymbol{y}_{t+1}=C\boldsymbol{y}_t for the companion matrix CC of a higher-order equation, and p(λ)=det⁡(λI−C)p(\lambda)=\det(\lambda I-C), which is the characteristic polynomial det⁡(C−λI)\det(C-\lambda I) of that matrix up to the factor (−1)k(-1)^k. Characteristic roots are therefore the eigenvalues of CC.


Key Results

  • rtr^t is a solution if and only if p(r)=0p(r)=0; since a0≠0a_0\ne0, 00 is never a root.
  • A root rr of multiplicity mm gives the mm solutions rt, t rt, …, tm−1rtr^t,\,t\,r^t,\,\ldots,\,t^{m-1}r^t; together over all roots these form a complex basis of solutions for complex-valued sequences.
  • Real coefficients make nonreal roots appear in conjugate pairs ρe±iθ\rho \mathrm{e}^{\pm\mathrm{i}\theta}; for a real basis, replace each pair of multiplicity mm by tjρtcos⁡(θt)t^j\rho^t\cos(\theta t) and tjρtsin⁡(θt)t^j\rho^t\sin(\theta t) for j=0,…,m−1j=0,\ldots,m-1.
  • Vieta: the roots sum to −ak−1-a_{k-1} and multiply to (−1)ka0(-1)^k a_0.
  • p(1)=1+ak−1+⋯+a0p(1)=1+a_{k-1}+\cdots+a_0; if p(1)≠0p(1)\ne0, the equation with constant right-hand side bb has the equilibrium x∗=b/p(1)x^*=b/p(1).

Examples

Distinct roots. xt+2−3xt+1+2xt=0x_{t+2}-3x_{t+1}+2x_t=0 has p(r)=r2−3r+2=(r−1)(r−2)p(r)=r^2-3r+2=(r-1)(r-2), so xt=c1+c2 2tx_t=c_1+c_2\,2^t.

Repeated root. xt+2−4xt+1+4xt=0x_{t+2}-4x_{t+1}+4x_t=0 has p(r)=(r−2)2p(r)=(r-2)^2. Besides 2t2^t, t 2tt\,2^t is a solution: (t+2)2t+2−4(t+1)2t+1+4t 2t=2t(4t+8−8t−8+4t)=0(t+2)2^{t+2}-4(t+1)2^{t+1}+4t\,2^t=2^t(4t+8-8t-8+4t)=0. Hence xt=(c1+c2t)2tx_t=(c_1+c_2t)2^t.

Complex roots. xt+2−xt+1+xt=0x_{t+2}-x_{t+1}+x_t=0 has p(r)=r2−r+1p(r)=r^2-r+1 with roots 12±32i=e±iπ/3\tfrac{1}{2}\pm\tfrac{\sqrt{3}}{2}\mathrm{i}=\mathrm{e}^{\pm\mathrm{i}\pi/3}. So xt=c1cos⁡(πt/3)+c2sin⁡(πt/3)x_t=c_1\cos(\pi t/3)+c_2\sin(\pi t/3), a solution of period 66; from x0=x1=1x_0=x_1=1 the recursion gives 1,1,0,−1,−1,0,1,…1,1,0,-1,-1,0,1,\ldots

Fibonacci. p(r)=r2−r−1p(r)=r^2-r-1 has roots φ=1+52\varphi=\tfrac{1+\sqrt{5}}{2} and ψ=1−52\psi=\tfrac{1-\sqrt{5}}{2}; with x0=0x_0=0, x1=1x_1=1 this gives xt=(φt−ψt)/5x_t=(\varphi^t-\psi^t)/\sqrt{5}.


Prerequisites



Sources

Aliases: characteristic equation of a difference equation, auxiliary equation of a recurrence, characteristic polynomial of a recurrence Keywords: trial solution x_t = r^t, characteristic roots, repeated root t r^t, companion matrix, p(r) = r^k + a_{k-1} r^{k-1} + ... + a_0 Symbols: lambda, phi, pi, psi, rho, square root, theta

Last updated: 27 October 2026 11:52