# Dynamical system problem x(t+1) = Ax(t)

I have a problem that have been trying to solve but it's not going so good. I would like some guidelines on how to work myself around this problem:

Two neighboring countries spy on each other and adapt its defense budget against each other depending on the situation in the neighborhood. Let country A's annual defense budget be a(t) (in billion) and country B's b(t). The change in budget changes each year according to the relationship:

a (t + 1) = r a (t) + s B (t)

b (t + 1) = s a (t) + r b (t)

where r is a positive number a bit less than 1 and s is a small postive number a bit greater than 0. Suppose that r is greater than s, that is, r>s. The budget for the two countries can be described by the vector x (t) = (a b) (<--- matrix) where a = a(t) and b = b(t)

If a(0) = 30 and b(0) = 5, what will happen with both the countries defense budget in the long run?

Solve tasks a) to e) both on paper and in Mathematica.

a) Set up the system transition matrix A such that x(t + 1) = Ax(t)

So what I've done so far is to select r = 0.8 and s = 0.32 and defined it as a 2x2 matrix and I've also defined a column vector with a(0)=30 and b(0)=5, so what I have in mathematica right know is:

I don't feel like this is how I'm supposed to do. At least this doesn't feel like how it should turn out in the "long run".

I would be very happy for someone to help/guide me through this! Thank you!

EDIT: I solved the problem and would like to have some feedback on how it is, if it's messy to follow. I would be thankful! :)

2nd EDIT: In my second question I need to find the Eigenvalues, Eigenvectors and Eigenbase. I have calculated by hand and I've gotten the Eigenvalues to lamda=1 and lamda=0.8 (when using the matrix {0.9,0.1},{0.1,0.9}) and the v1 = {1,1} and v2 = {-1,1}..I've gotten the right answers for it also on mathematica:

MatrixA = {{0.9, 0.1}, {0.1, 0.9}};
EigenMatrix = {{lm, 0}, {0, lm]}};
detMatrix = MatrixA - EigenMatrix
{{0.9 - lm], 0.1}, {0.1, 0.9 - lm}}

KarEk = Det[detMatrix]
0.8 - 1.8lm + lm^2
LamdaVarden = Roots[KarEk == 0, lm]
lm == 0.8 || lm == 1.


However, when trying to solve the Eigenvectors I get:

Eigenvectors[{{0.9, 0.1}, {0.1, 0.9}}]
{{0.707107, 0.707107}, {-0.707107, 0.707107}}


This is wrong, what am I doing wrong?

• Take a look at RSolve and also MatrixPower for a start.
– gwr
May 2, 2016 at 11:21
• This is the same form as a matrix population model from biology. Search under that term if you need more info about its long term behavior, which will be determined by the dominant eigenvalue of A. May 2, 2016 at 14:48
• You cannot calculate with MatrixForm (just display result of a matrix calculation with it). May 2, 2016 at 20:13
• Post code blocks rather than screenshots. For your approach, recommend that you define a function: x[t_] := MatrixPower[{{0.8, 0.32}, {0.32, 0.8}}, t].{30, 5}] and then just Map x onto a range: x /@ Range[0, 10] This approach provides results but does not as easily identify the limits ("long run"). May 2, 2016 at 21:39
• The eigenvectors are correct... all eigenvector calculations are only right up to a constant multiple. Mathematica deals with this by choosing the vector that has unit norm... so {0.707107, 0.707107} points in the same direction as {1,1} and {-0.707107, 0.707107} points in the direction {-1,1}. May 3, 2016 at 21:29

x[t_] = {a[t], b[t]} /. RSolve[{
a[t + 1] == r*a[t] + s*b[t],
b[t + 1] == s*a[t] + r*b[t],
a[0] == 30, b[0] == 5}, {a[t], b[t]}, t][[1]]


Since 0 < s < r < 1 then 0 < r - s < 1 - s < 1 and for large t, (r -s)^t is approximately zero.

 x[t] /. (r - s)^t -> 0


i.e., a[t] ≈ b[t] for large t

Plot[
Evaluate[{
x[t] /. {r -> 0.8, s -> 0.32},
x[t] /. {r -> 0.68, s -> 0.32},
x[t] /. {r -> 0.6, s -> 0.32}}],
{t, -1, 20},
PlotStyle ->
(AbsoluteDashing /@ {{7, 3}, {7, 7}}),
Frame -> True, Axes -> False,
PlotLegends -> {
"a[t], r+s=1.12", "b[t], r+s=1.12",
"a[t], r+s=1", "b[t], r+s=1",
"a[t], r+s=0.92", "b[t], r+s=0.92"},
Epilog -> {Text["r + s = 1.12", {13, 100}],
Text["r + s = 1", {13, 35/2}, {0, -1.5}],
Text["r + s = 0.92", {13, 0}, {0, .5}]}]


In the limit,

Piecewise[
Assuming[{0 < r < 1, 0 < s < 1, s < r, #},
{Limit[x[t], t -> Infinity], #}] & /@
{r + s < 1, r + s == 1,
r + s > 1}]


x /@ Range[0, 10] /. {r -> 0.8, s -> 0.32} // Column


EDIT: For the generic solution

Format[a0] := Subscript[a, 0]
Format[b0] := Subscript[b, 0]

x1[t_] = {a[t], b[t]} /. RSolve[{
a[t + 1] == r*a[t] + s*b[t],
b[t + 1] == s*a[t] + r*b[t],
a[0] == a0, b[0] == b0},
{a[t], b[t]}, t][[1]] // Simplify


x2[t_] = MatrixPower[{{r, s}, {s, r}}, t].{a0, b0} // Simplify;


As J.M. pointed out in a comment, this can also be written as

x2[t_] = MatrixPower[{{r, s}, {s, r}}, t, {a0, b0}] // Simplify;


x1 and x2 are identical

x1[t] === x2[t]

(*  True  *)


Approximate solution for large values of t

xApprox[t_] = Simplify[x1[t] /. (r - s)^t -> 0]


a[t] and b[t] are approximately equal for large t (and identical in the limit)

Equal @@ xApprox[t]

(*  True  *)


In the limit,

Piecewise[
Assuming[
{0 < r < 1, 0 < s < 1, s < r, a0 > 0, b0 > 0, #},
{Limit[x1[t], t -> Infinity], #}] & /@
{r + s > 1, r + s == 1,
r + s < 1}]


• Hello, thank you for answering! :) I have gotten the same values, but I've used a different way. I would appreciate it if you could give an opinion on if it is good as an answer? I will post a screenshot on my question. May 2, 2016 at 17:57
• Thank you again for the detailed explanation! :) Really nice that you show everything so clearly! I'm a bit new when it's comes to Mathematica but I will try to study it closer! May 3, 2016 at 10:29
• MatrixPower[{{r, s}, {s, r}}, t].{a0, b0} can also be done as MatrixPower[{{r, s}, {s, r}}, t, {a0, b0}] (the action form). May 9, 2016 at 23:58
• @J.M. - Thanks, edited to include this form. May 10, 2016 at 2:23

The way to answer this generically is via eigenvalues. Set up your matrix and take its eigenvalues:

mat = {{r, s}, {s, r}};
Eigenvalues[mat]

{r - s, r + s}


If either r-s or r+s is greater than 1 (in magnitude) then the system will be unstable. If both are less than one, it will decay to zero. For your values r=0.8 and s=0.32, you have instability, which is why you see the numerical simulations diverge.

• Thank you for the reply! So you mean it's better if I for example use r=0.9 and s=0.1 because then my system won't be unstable? :) May 3, 2016 at 10:28
• "Better" might depend on context, but for stability (i..e, the states $a(t)$ and $b(t)$ converging to zero) you will need Abs[r+s]<1. You can see that happening in Bob Hanlon's simulations. May 3, 2016 at 13:10