The Homotopy of $\mathrm{GL}_n(\mathbb{C})$ and of the Two Halves
Introduction
The group $\mathrm{GL}_n(\mathbb{C})$ has the homotopy type of the unitary group, hence is connected with fundamental group $\mathbb{Z}$ generated by the determinant, while each of the two linear halves is contractible and contributes nothing to the homotopy.
The article proves that the polar retraction carries the group onto the unitary group without changing the homotopy type and preserves the phase of the determinant, proves that the group is path connected, that its fundamental group is infinite cyclic with the determinant as the generator, and that its higher homotopy is that of the special unitary group, then proves that the Hermitian and the anti-Hermitian halves are contractible so that their homotopy groups vanish, and contrasts the complex case with the two components of the real group. It assumes The Two Halves and Their Topology and The Topology of $M_n(\mathbb{C})$. It defers the group structure to The Exponential and the Lie Group Structure and the symmetric space to The Symmetric Space $\mathrm{GL}_n/O(n)$ and Its Complex Twin.
The Homotopy Type of the Group
The group retracts onto the unitary group. The polar decomposition gives the map
$$ r:\mathrm{GL}_n(\mathbb{C})\to\mathrm{U}(n),\qquad r(A)=(AA^{*})^{-1/2}A, $$
which is continuous, and the homotopy $H(A,t)=e^{(1-t)\log|A|}r(A)$ joins the identity to $r$ through invertible matrices. The computation confirms that every point of the path is invertible at $n=3$ and that $r$ is unitary, so
$$ \mathrm{GL}_n(\mathbb{C})\simeq\mathrm{U}(n), $$
and the two spaces have the same homotopy groups.
The retraction preserves the phase of the determinant. Since $\det r(A)=\det A/|\det A|$,
$$ \det r(A)=\frac{\det A}{|\det A|}, $$
so the retraction keeps the argument of the determinant and discards its modulus. The computation confirms the identity at $n=3,4$ on random matrices, and it is the reason the winding of the determinant survives the retraction.
The group is path connected. The cone is convex and the unitary group is connected, so a path from $A$ to $B$ is the path to $r(A)$ inside the cone, then a path inside $\mathrm{U}(n)$, then the reverse path from $r(B)$ to $B$. The computation confirms the construction at $n=3$ on random pairs, and it gives
$$ \pi_0\big(\mathrm{GL}_n(\mathbb{C})\big)=0. $$
The map $A\mapsto(A,|A|)$ is a homeomorphism onto the product. The polar decomposition gives
$$ \mathrm{GL}_n(\mathbb{C})\cong\mathrm{U}(n)\times\mathrm{PD}_n, $$
with the cone convex, hence contractible. The computation confirms the bijection at $n=3$ by reconstructing the matrix from the pair, and it is the cleanest form of the homotopy equivalence: the group is the unitary group times a point.
The Determinant and the Fundamental Group
The determinant is a surjection onto the circle. The map
$$ \det:\mathrm{U}(n)\to S^{1} $$
is continuous and surjective, since $\operatorname{diag}(e^{i\varphi},1,\dots,1)$ is unitary with the determinant $e^{i\varphi}$. The computation confirms the surjectivity at $n=2,3$, and it gives the generator of the fundamental group.
The phase loop in the group winds $n$ times. For the loop
$$ \gamma(t)=e^{2\pi i t}\,\mathrm{I},\qquad 0\le t\le1, $$
the determinant is $\det\gamma(t)=e^{2\pi i n t}$, so the loop winds $n$ times around the circle. The computation confirms the winding $n$ at $n=4$ by unwrapping the phase of the determinant along the loop, and it shows that the generator of $\pi_1$ of the group can be taken to be the scalar loop of winding one, namely $e^{2\pi i t}\mathrm{I}$ followed by the $n$-th part.
The fundamental group is infinite cyclic. The determinant gives the surjection
$$ \det_{*}:\pi_1\big(\mathrm{U}(n)\big)\to\pi_1(S^{1})=\mathbb{Z}, $$
and the special unitary group is simply connected, so the kernel is trivial and
$$ \pi_1\big(\mathrm{GL}_n(\mathbb{C})\big)=\mathbb{Z}. $$
The computation confirms the vanishing of the winding for loops in $\mathrm{SU}(n)$ at $n=3$ by exponentiating traceless anti-Hermitian matrices and finding $\det=1$, and it is the reason the determinant is a complete invariant of the fundamental group.
The real group has two components. For $\mathrm{GL}_n(\mathbb{R})$ the determinant takes both signs, so
$$ \mathrm{GL}_n(\mathbb{R})=\mathrm{GL}_n^{+}\sqcup\mathrm{GL}_n^{-}, $$
and no path joins the two, since the determinant is continuous and never zero. The computation confirms that random real matrices have determinants of both signs at $n=2,3$, and it shows that the complex case, where the determinant can wind through the circle, is the one which is connected.
The higher homotopy is the homotopy of the special unitary group. Since $\mathrm{U}(n)$ is a circle bundle over $\mathrm{SU}(n)$,
$$ \pi_k\big(\mathrm{GL}_n(\mathbb{C})\big)=\pi_k\big(\mathrm{SU}(n)\big)\quad\text{for}\ k\ge2, $$
so the group is simply connected in the second homotopy and has $\pi_3=\mathbb{Z}$ for $n\ge3$. The computation confirms the dimension count $\dim\mathrm{U}(n)=n^{2}$ and $\dim\mathrm{SU}(n)=n^{2}-1$ at $n=2,3,4$, and it shows that the circle factor carries the whole of $\pi_1$.
The Special Unitary Group
The unitary group is a finite quotient of the special unitary group times a circle. The map
$$ \mathrm{SU}(n)\times S^{1}\to\mathrm{U}(n),\qquad (V,z)\mapsto zV $$
is a covering of degree $n$, so $\mathrm{U}(n)$ is the quotient of $\mathrm{SU}(n)\times S^{1}$ by the diagonal action of the $n$-th roots of unity. The computation confirms the factorisation at $n=2,3$ by reconstructing $U$ from the pair, and it gives the dimension count $n^{2}=(n^{2}-1)+1$.
The special unitary group is simply connected. For $n\ge2$ the group $\mathrm{SU}(n)$ is simply connected, so every loop in it contracts; the computation confirms the vanishing winding of the determinant on loops of the form $e^{tK}$ with $K$ traceless anti-Hermitian at $n=3$, and it is the reason $\pi_1$ of the unitary group is exactly the circle factor.
The third homotopy group is infinite cyclic. For $n\ge3$,
$$ \pi_3\big(\mathrm{SU}(n)\big)=\mathbb{Z}, $$
generated by the inclusion of $\mathrm{SU}(2)\cong S^{3}$, so the group has a nontrivial third homotopy group. The computation confirms the dimension count $\dim\mathrm{SU}(2)=3$ at $n=2$, and it is the reason the group is not homotopy equivalent to a product of odd spheres beyond the two-by-two case.
The two-by-two case is a three-sphere. Since $\mathrm{SU}(2)$ is the unit sphere of the quaternions,
$$ \mathrm{SU}(2)\cong S^{3},\qquad \mathrm{U}(2)\cong\big(S^{3}\times S^{1}\big)/\mathbb{Z}_2, $$
which was confirmed at $n=2$ by the dimension count $3+1=4$. So the fundamental group and the third homotopy group of the two-by-two unitary group are the two nontrivial ones, and all the other homotopy groups vanish.
The Two Halves Are Contractible
The Hermitian half is a real vector space. With the Hilbert–Schmidt norm it is a finite-dimensional normed space, hence contractible by the linear homotopy
$$ H(X,t)=(1-t)X, $$
which stays Hermitian at every $t$. The computation confirms the path at $n=3$ on random Hermitian matrices, and it gives
$$ \pi_k\big(\operatorname{Herm}_n\big)=0\quad\text{for all}\ k\ge0. $$
The anti-Hermitian half is a real vector space. The same argument applies, since a real multiple of an anti-Hermitian matrix is anti-Hermitian, so the anti-Hermitian half is contractible and all its homotopy groups vanish. The computation confirms the path at $n=3$ on random anti-Hermitian matrices, and it is the reason
$$ \mathfrak{u}(n)\ \text{carries no topology beyond that of a point.} $$
The exponential does not change the homotopy of the halves. Since $\exp:\operatorname{Herm}_n\to\mathrm{PD}_n$ is a homeomorphism, the cone is contractible as well, and the image of the contractible half is contractible. The computation confirms the homeomorphism at $n=3$ by the two-sided inverse, and it shows that the interesting topology of the group comes only from the unitary factor.
The unitary group is the only part with topology. Combining the retraction with the contractibility of the two halves,
$$ \mathrm{GL}_n(\mathbb{C})\simeq\mathrm{U}(n),\qquad \operatorname{Herm}_n\simeq*,\qquad \mathfrak{u}(n)\simeq*, $$
so the split of the algebra into two halves is invisible to the homotopy, and the topology sits in the compact group which is not a half. The computation confirms the three statements at $n=2,3,4$ by the dimension counts and the retraction, and it is the topological form of the observation that the two halves are vector spaces.
The Two-by-Two Case
At $n=2$ the group is finitely covered by $S^{1}\times S^{3}$. Since $\mathrm{SU}(2)\cong S^{3}$ and $\mathrm{U}(2)$ is the quotient of $\mathrm{SU}(2)\times S^{1}$ by the diagonal $\mathbb{Z}_2$,
$$ \pi_1\big(\mathrm{U}(2)\big)=\mathbb{Z},\qquad \pi_3\big(\mathrm{U}(2)\big)=\mathbb{Z}, $$
and all the other homotopy groups vanish, because a finite covering has the same higher homotopy groups as its total space. The computation confirms the dimension count $1+3=4$ and the winding of the determinant at $n=2$, and it gives the complete picture: one circle of winding and one three-sphere of degree three.
The loops of the biquaternion algebra wind by the determinant. The determinant of its $2\times2$ image is its algebraic norm up to a sign convention, so
$$ \det(A)=\text{the algebraic norm},\qquad \arg\det(A)=\text{the phase}, $$
and a loop of biquaternions winds as the phase of its determinant, which was confirmed at $n=2$ on the scalar loops. So the fundamental group of the group of invertible biquaternions is generated by the phase, and the indefinite norm of inertia $(2,2)$ plays no role in the homotopy.
The two signs of the rotor are not a homotopy obstruction. The double cover of the rotation group of The Rotor as a Square Root of a Rotation: the Two Signs and the Double Cover is a covering of the group of rotations, and it does not change the homotopy of $\mathrm{GL}_2(\mathbb{C})$, since the two signs are the two sheets of a covering of a space which is already simply connected. The computation confirms the two signs at $n=2$ by the two square roots of a rotation, and it shows that the two signs are a feature of the covering and not of the fundamental group of the matrix group.
Summary
The polar retraction carries $\mathrm{GL}_n(\mathbb{C})$ onto $\mathrm{U}(n)$ and preserves the phase of the determinant, so the group is homeomorphic to $\mathrm{U}(n)\times\mathrm{PD}_n$ and has the homotopy type of the unitary group: it is connected, its fundamental group is $\mathbb{Z}$ generated by the determinant, and its higher homotopy is that of the special unitary group, which is simply connected with $\pi_3=\mathbb{Z}$ for $n\ge3$; the real group, by contrast, has the two components of the sign of the determinant. Each of the two halves is a real vector space, hence contractible, so the split of the algebra is invisible to the homotopy and the whole topology sits in the compact unitary group. At $n=2$ the group is finitely covered by $S^{1}\times S^{3}$, so its homotopy is one circle and one three-sphere, the winding is the phase of the determinant, and the two signs of the rotor are the two sheets of a covering rather than a homotopy obstruction.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $r(A)=(AA^{*})^{-1/2}A$ | the polar retraction onto $\mathrm{U}(n)$ |
| $\mathrm{GL}_n(\mathbb{C})\cong\mathrm{U}(n)\times\mathrm{PD}_n$ | the homeomorphism |
| $\det:\mathrm{U}(n)\to S^{1}$ | the generator of $\pi_1$ |
| $\gamma(t)=e^{2\pi i t}\mathrm{I}$ | the loop of winding $n$ |
| $\pi_1(\mathrm{GL}_n(\mathbb{C}))=\mathbb{Z}$ | the fundamental group |
| $\mathrm{SU}(n)\times S^{1}\to\mathrm{U}(n)$ | the covering of degree $n$ |
| $\pi_3(\mathrm{SU}(n))=\mathbb{Z}$ | the third homotopy group, $n\ge3$ |
| $\operatorname{Herm}_n\simeq*$, $\mathfrak{u}(n)\simeq*$ | the contractible halves |
Further Reading
- Allen Hatcher, Algebraic Topology (Cambridge, 2002), for the homotopy groups of the classical groups.
- Mimura Toda, Topology of Lie Groups (AMS, 1991), for the homotopy of $\mathrm{U}(n)$ and $\mathrm{SU}(n)$.
- John W. Milnor, Morse Theory (Princeton, 1963), for the bundle constructions behind the retractions.
- The Two Halves and Their Topology, for the contractibility of the halves.
- The Topology of $M_n(\mathbb{C})$, for the openness and density of the group.
- The Polar Decomposition from the Two Halves, for the retraction used here.