advanced30 min studyLesson 491 in Mathematics

Matrix Lie groups and smooth subgroups

Combine matrix symmetries with smooth structure and distinguish a closed subgroup from an arbitrary parametrized subgroup.

On this page

What you’ll understand

  • Check smooth group operations for matrix examples.
  • Use defining equations and regular values to identify subgroup dimensions.
  • State the closed subgroup theorem with the correct ambient topology.

Rotations can be composed and inverted, but they also vary smoothly with parameters. Lie groups keep both structures at once: a finite-dimensional smooth manifold and a group whose multiplication and inversion are smooth. A matrix representation provides concrete calculations without making every algebraic subgroup automatically a well-behaved submanifold. The topology matters. We will work with real finite-dimensional Lie groups and explicit matrix examples, separating closed embedded subgroups from the immersed subgroups that appear later in the correspondence with Lie algebras.

An open matrix domain already has smooth operations

The invertible real n-by-n matrices form an open subset of the vector space of all matrices, since nonzero determinant is an open condition. Matrix multiplication is polynomial in the entries. Inversion is smooth there because the adjugate formula divides polynomial entries by a nonvanishing determinant. Consequently GL-n is a Lie group of dimension n squared. Neither the singular matrices nor the ambient vector space with its usual matrix multiplication form a group of the same kind.

A subgroup inherits a candidate multiplication, but it still needs an appropriate manifold structure. One useful construction is a regular level set of smooth matrix constraints. Another is the closed subgroup theorem: a subgroup closed in a finite-dimensional Lie group has a unique embedded Lie subgroup structure compatible with the ambient one. This is stronger than merely observing that a parametrization is smooth, and its hypothesis is relative closedness in the ambient group rather than necessarily closedness in all matrices.

GLn(R)={A:det⁡A≠0},A−1=adj⁡(A)det⁡AGL_n(\mathbb R)=\{A:\det A\ne0\},\qquad A^{-1}=\frac{\operatorname{adj}(A)}{\det A}
The ambient matrix space has dimension n². Smooth inversion follows only on the open set where the denominator is nonzero.↔ Scroll wide equations horizontally.

Orthogonality gives equations with a dimension count

The orthogonal group is defined by transpose-A times A equal to the identity. The constraint map takes matrices into symmetric matrices. Its derivative at the identity sends an increment X to X plus its transpose, a surjection onto the symmetric matrices. The implicit function theorem therefore gives the expected local dimension n times n minus one divided by two. Translation in the group carries the same manifold description to every orthogonal matrix. This argument counts independent equations rather than subtracting all n squared entries as if they were independent.

Taking determinants shows that an orthogonal matrix has determinant plus or minus one. The special orthogonal subgroup selects the plus sign and has the same local dimension, because this discrete choice does not add a new local equation. In dimension two the plus component consists of rotations, while a reflection belongs to the other component. Matrix constraints therefore reveal both continuous degrees of freedom and discrete information that a tangent space alone cannot remember.

O(n)={A:ATA=I},d(ATA)I(X)=XT+X,dim⁡O(n)=n(n−1)2O(n)=\{A:A^{\mathsf T}A=I\},\qquad d(A^{\mathsf T}A)_I(X)=X^{\mathsf T}+X,\qquad \dim O(n)=\frac{n(n-1)}2
The derivative is surjective onto the n(n+1)/2-dimensional symmetric matrix space. This establishes the local dimension, with group translations extending the result globally.↔ Scroll wide equations horizontally.

Closedness and parametrization answer different questions

A closed matrix subgroup is closed relative to GL-n. For example, the positive scalars form a closed and open subgroup of the multiplicative group of nonzero real scalars, even though they are not closed in the whole real line. A dense one-parameter subgroup of a torus supplies a different phenomenon: its natural immersed Lie group topology can be finer than the subset topology inherited from the torus. An injective smooth group map need not be an embedding.

When checking a proposed matrix symmetry family, verify closure under multiplication, identity, and inverse before studying its geometry. A surface of matrices can look smooth while failing group closure, and a group can have multiple components with the same local equations. Later, tangent Lie algebras simplify calculations near the identity; they do not replace this initial topological bookkeeping. Distinguishing intrinsic parameters, ambient entries, and group components makes dimension claims both useful and verifiable.

Separate one rotation parameter from a component choice

Compare R=[[0,−1],[1,0]] and S=diag(1,−1). Verify group membership and derive the local dimension of O(2).

  1. Both satisfy AᵀA=I, so both are orthogonal and have inverse Aᵀ. Their determinants are det R=1 and det S=−1. Thus R lies in SO(2), while S is a reflection in the other orthogonal component.

  2. A real 2-by-2 matrix has four coordinates. The symmetric matrix constraint AᵀA−I has three independent coordinates. At the identity its derivative X↦X+Xᵀ is surjective, so the regular level set has local dimension four minus three, namely one.

  3. A general rotation has entries cos θ,−sin θ,sin θ,cos θ. The single angular parameter matches the local dimension. The determinant sign is discrete and does not create another tangent direction; it distinguishes components with the same dimension.

The result

Both matrices belong to a one-dimensional Lie group, but only the determinant-one component contains the identity and all planar rotations.

Explore the model

Interactive lab

Interactive lab · matrix

A determinant-one shear is not a rotation

Pause, Reset, then set diagonal scale a=1 and shear b=0 to inspect the identity. Repeat from Reset with both a=1 and b=1. Predict unchanged determinant one but a sheared unit square and unequal singular values. The second matrix lies in SL₂ but fails orthogonality. The instrument covers only its fixed triangular family and does not classify all matrix subgroups.

A determinant-one shear is not a rotationThe selected matrix is A = [[a,b],[0,1]]. Playback interpolates geometry between the identity and A; readouts always describe A. Its determinant is a and its singular values come from AᵀA. Positive a makes every supported matrix invertible. This two-dimensional family is not a general eigensolver or a learned network. Current readings: Area scale · det A: 1.4; Singular values · σmax, σmin: 1.597, 0.876; Condition number · κ₂(A): 1.823. All parameters are adjustable with the labeled controls below.Matrix A with first row 1.4, 0.6 and second row zero, one; display interpolates between identity and A while final determinant remains 1.4.−4−2024−2−1012x · equal geometric scaleyA = [ 1.4 0.6 ; 0 1 ]Display blend: 0% · final values below
Interpolated geometryOriginal unit squareSelected final transform
Area scale · det A
1.4
Singular values · σmax, σmin
1.597, 0.876
Condition number · κ₂(A)
1.823
t = 0 s

The selected matrix is A = [[a,b],[0,1]]. Playback interpolates geometry between the identity and A; readouts always describe A. Its determinant is a and its singular values come from AᵀA. Positive a makes every supported matrix invertible. This two-dimensional family is not a general eigensolver or a learned network. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.

Look forDeterminant and orthogonality impose different constraints; belonging to one matrix group does not imply belonging to another.

Check your understanding

What does the closed subgroup theorem require for a subgroup H of a finite-dimensional Lie group G?

Try it yourself

Show that SL₂(R), defined by det A=1, has dimension three. Why is the determinant constraint regular at every point of this set?

Take a moment to make a prediction or work it out on paper. Open a hint when you need one.

Give me a hint

At the identity the determinant derivative is trace. At a general invertible matrix use det(A)·tr(A⁻¹X).

Compare with the solution

The derivative at A is X↦det⁡(A)tr⁡(A−1X)X\mapsto\det(A)\operatorname{tr}(A^{-1}X). On det A=1 this is nonzero: choosing X=A gives trace I=2. Thus one is a regular value of determinant on GL₂, and its level set has dimension four minus one, namely three. Multiplication and inversion preserve determinant one, so it is a subgroup as well as an embedded submanifold.

Take it with you

  • Lie groups require compatible smooth multiplication and inversion.
  • Regular constraints count continuous dimensions, while components carry additional discrete data.
  • Closed subgroup means closed in the ambient Lie group, and immersed need not mean embedded.

Read further

Original sources and deeper explanations to follow at your own pace.

Optional progress, saved in this browser when available. Revisit any lesson at any time.

Lesson not marked complete.