Filter löschen
Filter löschen

Eigenvectors not changing with constant parameter

1 Ansicht (letzte 30 Tage)
SHUBHAM PATEL
SHUBHAM PATEL am 23 Apr. 2022
Bearbeitet: Bruno Luong am 24 Apr. 2022
Dear All,
I am trying to calculate the eigenvectors(V) using eig() function. My 2x2 matrix(M) contains a constant parameter 'a' in it. But I see that changing a does not change my eigenvectors. Generally the eigevectors and eigenvalues change with the matrix elements. Here my eigenvalues are varying but not the eigenvectors. Could someone figure out the issue?
sx = [0 1; 1 0];
sy = [0 -1i; 1i 0];
a = 0.18851786;
kx = -0.5:0.1:0.5;
ky = kx;
for i = 1:length(kx)
for j = 1:length(ky)
M = a.*(sx.*ky(i)-sy.*kx(j));
[V,D] = eig(M);
V
end
end

Melden Sie sich an, um diese Frage zu beantworten.

Akzeptierte Antwort

Bruno Luong
Bruno Luong am 23 Apr. 2022
Bearbeitet: Bruno Luong am 23 Apr. 2022
"Could someone figure out the issue?"
But there is no issue beside thet fact that you expect something that not going to happen.
if V and diagonal D the eigen decomposition of A1
A1*V = V*D
then for any constant a
(a*A1)*V = V*(a*D)
Meaning V and a*D (still diagonal) are eigen decomposition of Aa := a*A1.
So A1 and Aa respective eigen decomposition can have the same V (eigen vectors, MATALB always normalized them to have norm(V(:,k),2)=1 for all k) but eigen values are proportional to a.
  5 Kommentare
3 ältere Kommentare anzeigen
Bruno Luong
Bruno Luong am 23 Apr. 2022
Bearbeitet: Bruno Luong am 24 Apr. 2022
This is just a coincidence in YOUR parametrization.
Your matrix M has the same eigen vectors as this matrix
A := M / (kx(i).^2+ky(j).^2) % see your original question
A = I + K(a);
with
K = [0 conj(z);
z 0]
z(a) := a*(kx(i) + 1i*ky(j)) / (kx(i).^2+ky(j)).^2 % but it doesn't matter, any complex still works
So
A = [1 conj(z);
z 1];
This (Hermitian) matrix A has two eigen values and (unormalized) eigen vectors
d1 = 1 - abs(z); V1 = [conj(z); -abs(z)];
d2 = 1 + abs(z); V2 = [conj(z); +abs(z)];
So when you scale z up and down by a (a non zero real value) the vector V1 and V2 remains the same after normalization (by MATLAB.
V1 = [conj(z); -abs(z)] / sqrt(2*abs(z)^2) = [conj(z)/abs(z), -1] / sqrt(2)
V2 = [conj(z); abs(z)] / sqrt(2*abs(z)^2) = [conj(z)/abs(z), +1] / sqrt(2)
Quod erat demonstrandum
SHUBHAM PATEL
SHUBHAM PATEL am 24 Apr. 2022
Thank you @Bruno Luong for such a detailed explaination. I understood the gist.

Melden Sie sich an, um zu kommentieren.

Weitere Antworten (1)

Walter Roberson
Walter Roberson am 23 Apr. 2022
eigenvectors are geometrically directions. When you scale a matrix by a nonzero constant, the direction does not change.
  1 Kommentar
SHUBHAM PATEL
SHUBHAM PATEL am 23 Apr. 2022
Thanks W. Roberson,
I understood your point that if we multiply the matrix by a constant it does not change the directions.
But I am seeing that if I include the diagonal term also and multiply the off-diagonal term only with a constant, even then the vectors are not changing, although the eigen values are changing.
Could you comment on that?
I am writing the code again including the diagonal term.
sx = [0 1; 1 0];
sy = [0 -1i; 1i 0];
a = 0.18851786;
kx = -0.5:0.1:0.5;
ky = kx;
for i = 1:length(kx)
for j = 1:length(ky)
M = (kx(i).^2+*ky(j).^2)*I + a.*(sx.*ky(i)-sy.*kx(j));
[V,D] = eig(M);
V
end
end

Melden Sie sich an, um zu kommentieren.

Kategorien

Mehr zu Linear Algebra finden Sie in Help Center und File Exchange

Tags

Produkte


Version

R2022a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by