Constructing a large tridiagonal matrix with alternating signs

Question

What is the best way to construct a large tridiagonal matrix, in the following form (notably with alternating signs)?

$\mathbf M = \begin{pmatrix} 0 & a & 0 & 0 & 0 & \cdots \\ a & 0 & -a & 0 & 0&\cdots \\ 0 & -a & 0 & a & 0 &\cdots \\ 0 & 0 & a & 0 & -a &\cdots \\ \vdots & \vdots & \vdots & \vdots & \vdots & \ddots \end{pmatrix}$

I was using code like that shown below to produce constant sign diagonals of $a$ above and below the main diagonal for an $n\times n$ matrix.

DiagonalMatrix[Array[a &, n - 1], -1]


DiagonalMatrix[Array[a &, n - 1], 1] 

Any help is appreciated.

Answer 1

Here is a way using Band:

SparseArray[{Band[{1, 2}, {4, 5}] -> {a, -a}, 
          Band[{2, 1}, {5, 4}] -> {a, -a}}, {5, 5}] // MatrixForm

$$\left( \begin{array}{ccccc} 0 & a & 0 & 0 & 0 \\ a & 0 & -a & 0 & 0 \\ 0 & -a & 0 & a & 0 \\ 0 & 0 & a & 0 & -a \\ 0 & 0 & 0 & -a & 0 \\ \end{array} \right)$$

To make it easier to modify the size and the pattern of elements on the off-diagonals, I'd suggest something like this (incorporating J.M.'s suggestion):

With[{n = 7, pattern = {a, -a}},
 SparseArray[{Band[{1, 2}, {-2, -1}] -> pattern, 
   Band[{2, 1}, {-1, -2}] -> pattern}, {n, n}]]

The negative indices are counted from the end, so that we don't need to use the dimension n in the Band specification. The reason I specify the beginning and end of each band is that only in this case will a cyclic repetition of pattern occur, as desired. If the end specification is omitted, pattern will only be used to fill the first few entries of the Band.

Answer 2

Here are two alternatives to Jens's procedure.

With Band[]:

With[{n = 7, a = 1}, 
     SparseArray[{Band[{1, 2}] -> #, Band[{2, 1}] -> #}] & @
     PadRight[{}, n - 1, {a, -a}]]

Without Band[]:

With[{n = 7, a = 1}, 
     SparseArray[{{j_, k_} /; Abs[j - k] == 1 && EvenQ[Max[j, k]] -> a,
                  {j_, k_} /; Abs[j - k] == 1 && OddQ[Max[j, k]] -> -a}, {n, n}]]

Both should yield

$$\begin{pmatrix} 0 & 1 & 0 & 0 & 0 & 0 & 0 \\ 1 & 0 & -1 & 0 & 0 & 0 & 0 \\ 0 & -1 & 0 & 1 & 0 & 0 & 0 \\ 0 & 0 & 1 & 0 & -1 & 0 & 0 \\ 0 & 0 & 0 & -1 & 0 & 1 & 0 \\ 0 & 0 & 0 & 0 & 1 & 0 & -1 \\ 0 & 0 & 0 & 0 & 0 & -1 & 0 \\ \end{pmatrix}$$

Answer 3

You can do it with DiagonalMatrix easily as well:

n = 7; 
DiagonalMatrix[-a*(-1)^Range[n - 1], -1] + DiagonalMatrix[-a*(-1)^Range[n - 1], 1]

Answer 4

EDIT

fun[n_, a_] := SparseArray[{{i_, j_} /; j == i + 1 :> a (-1)^(i - 1),
   {i_, j_} /; j == i - 1 :> a (-1)^(i)}, {n, n}]

e.g.

Grid[Partition[(fun[#, a] // MatrixForm) & /@ Range[2, 10], 3]]

See comments from J.M. nicer first condition: {i_, j_} /; j == i + 1 :> a (-1)^j