Maths WorkshopMaths problems with full solutions
Exercise 1461Topology › Topology of Mn(ℝ) › Applications and the case n = 2L3Added 10 Nov 2026

The nonzero nilpotents of M_2(R) form two components

#M_2(R)#nilpotent matrices#connected components#complex coordinates

Statement

Identify R2\mathbb{R}^2 with C\mathbb{C} and write A∈M2(R)A\in M_2(\mathbb{R}) as Ax=zx+wxˉAx=zx+w\bar x, with z=(a+d)+i(c−b)2z=\frac{(a+d)+i(c-b)}2, w=(a−d)+i(b+c)2w=\frac{(a-d)+i(b+c)}2 for A=(abcd)A=\begin{pmatrix}a&b\\c&d\end{pmatrix} (so tr⁡A=2Re⁡z\operatorname{tr}A=2\operatorname{Re}z, det⁡A=∣z∣2−∣w∣2\det A=|z|^2-|w|^2). Show that the set N∗\mathcal N^* of nonzero nilpotent matrices of M2(R)M_2(\mathbb{R}) has exactly two connected components, each homeomorphic to S1×(0,∞)S^1\times(0,\infty).

Hint

A 2×22\times2 matrix is nilpotent iff tr⁡A=det⁡A=0\operatorname{tr}A=\det A=0. Translate this into z=iqz=iq and ∣w∣=∣q∣|w|=|q|.

Show solutionHide solution

Solution

Idea. The two equations tr⁡=det⁡=0\operatorname{tr}=\det=0 leave only a "cone" in R4\mathbb{R}^4; removing the vertex 00 separates the cone into two nappes, distinguished by the sign of Im⁡z\operatorname{Im}z.

Step 1: nilpotent iff tr⁡=det⁡=0\operatorname{tr}=\det=0. If AA is nilpotent, all its eigenvalues vanish, so χA(t)=t2\chi_A(t)=t^2, i.e. tr⁡A=det⁡A=0\operatorname{tr}A=\det A=0. Conversely, if tr⁡A=det⁡A=0\operatorname{tr}A=\det A=0, Cayley-Hamilton gives A2=0A^2=0.

Step 2: coordinates. The two equations become Re⁡z=0\operatorname{Re}z=0 and ∣w∣2=∣z∣2|w|^2=|z|^2. Write z=iqz=iq with q∈Rq\in\mathbb{R}; then ∣w∣=∣q∣|w|=|q|. The matrix is nonzero iff (z,w)≠(0,0)(z,w)\ne(0,0) iff q≠0q\ne0 (if q=0q=0 then z=0z=0 and ∣w∣=0|w|=0).

Step 3: structure. So
N∗≅{(q,w):q∈R∖{0}, ∣w∣=∣q∣}.\mathcal N^*\cong\{(q,w):q\in\mathbb{R}\setminus\{0\},\ |w|=|q|\}.
The map (q,w)↦(q, w/∣w∣)∈(R∖{0})×S1(q,w)\mapsto\bigl(q,\ w/|w|\bigr)\in(\mathbb{R}\setminus\{0\})\times S^1 is continuous with continuous inverse (q,u)↦(q,∣q∣u)(q,u)\mapsto(q,|q|u). Thus
N∗≅(R∖{0})×S1=((−∞,0)×S1)⊔((0,∞)×S1).\mathcal N^*\cong(\mathbb{R}\setminus\{0\})\times S^1=\bigl((-\infty,0)\times S^1\bigr)\sqcup\bigl((0,\infty)\times S^1\bigr).

Step 4: components. Each piece is a product of connected spaces, hence connected. The sign of q=Im⁡zq=\operatorname{Im}z is a continuous function on N∗\mathcal N^* taking the two values ±\pm, so the two pieces are open and closed in N∗\mathcal N^*. Hence they are exactly the connected components.

Check. E=(0100)E=\begin{pmatrix}0&1\\0&0\end{pmatrix} has a=d=c=0a=d=c=0, b=1b=1, so z=−i2z=-\frac i2, w=i2w=\frac i2: q=−12<0q=-\frac12<0 and ∣w∣=12=∣q∣|w|=\frac12=|q|. The matrix ET=(0010)E^T=\begin{pmatrix}0&0\\1&0\end{pmatrix} has z=i2z=\frac i2, w=i2w=\frac i2: q=+12q=+\frac12, on the other component.

Remark. Including 00, the set {A2=0}={∣w∣=∣Im⁡z∣, Re⁡z=0}\{A^2=0\}=\{|w|=|\operatorname{Im}z|,\ \operatorname{Re}z=0\} is a double cone over a circle in a 3-dimensional subspace; it is connected (through its vertex 00), and removing the vertex disconnects it. For n≥3n\ge3 the situation is different: the nonzero nilpotent cone of Mn(R)M_n(\mathbb{R}) is connected (it is star-shaped about 00 and 00 can be approached through other orbits).

Lesson. A cone minus its vertex can be disconnected; a continuous sign function (here sign⁡Im⁡z\operatorname{sign}\operatorname{Im}z) is the tool to prove disconnection.