Classification of Simply Laced Dynkin Diagrams of Finite Type (Theorem # 4718)
Theorem
Let $\Gamma$ be a finite connected simple graph, and let $C_\Gamma$ be the symmetric matrix indexed by the vertices of $\Gamma$ and defined by
\begin{align*}
(C_\Gamma)_{vv} &= 2,\\
(C_\Gamma)_{vw} &= -1 \quad \text{if } v \text{ and } w \text{ are adjacent in } \Gamma,\\
(C_\Gamma)_{vw} &= 0 \quad \text{if } v \ne w \text{ and } v,w \text{ are not adjacent in } \Gamma.
\end{align*}
If $C_\Gamma$ is positive definite, then $\Gamma$ is isomorphic to exactly one of the Dynkin diagrams
\begin{align*}
A_\ell \quad (\ell \ge 1), \qquad D_\ell \quad (\ell \ge 4), \qquad E_6,\ E_7,\ E_8.
\end{align*}
Conversely, each of the graphs $A_\ell$ for $\ell \ge 1$, $D_\ell$ for $\ell \ge 4$, and $E_6,E_7,E_8$ occurs as the Dynkin diagram of a finite reduced crystallographic root system.
Algebra
Discussion
This theorem states that Gamma be a finite connected simple graph, and let C Gamma be the symmetric matrix indexed by the vertices of Gamma and defined by (C ) vv &= 2,\\ (C ) vw &= -1 v w ,\\ (C ) vw &= 0 v w v,w .. It is used in the structure and classification of finite-dimensional Lie algebras, especially in arguments involving Cartan data, root systems, Weyl groups, and Dynkin diagrams.
Proof
[proofplan]
We translate the finite simply laced condition into positive definiteness of the symmetric Cartan matrix $C_\Gamma$. Positive definiteness first excludes cycles and high-valence vertices, so $\Gamma$ is a tree whose vertices have degree at most $3$. If there is no degree-$3$ vertex, the tree is a path and hence is $A_\ell$; if there is a degree-$3$ vertex, positive definiteness excludes a second branch point and then a determinant computation for a three-armed tree gives the inequality $\frac1p+\frac1q+\frac1r>1$. Solving this elementary inequality gives exactly the $D$ and $E$ diagrams, and the converse is supplied by the standard coordinate root systems for types $A$, $D$, and $E$.
[/proofplan]
[step:Express positive definiteness through the graph quadratic form]
Let $V$ be the finite vertex set of $\Gamma$. For a function
\begin{align*}
x: V &\to \mathbb{R}\\
v &\mapsto x_v,
\end{align*}
define the quadratic form $Q_\Gamma: \mathbb{R}^V \to \mathbb{R}$ by
\begin{align*}
Q_\Gamma(x) = \sum_{v,w \in V} (C_\Gamma)_{vw} x_v x_w.
\end{align*}
Since $C_\Gamma$ is symmetric, the hypothesis says that $Q_\Gamma(x)>0$ for every nonzero $x \in \mathbb{R}^V$. Expanding the entries of $C_\Gamma$ gives
\begin{align*}
Q_\Gamma(x)
= 2\sum_{v \in V} x_v^2 - 2\sum_{\{v,w\} \in E(\Gamma)} x_v x_w,
\end{align*}
where $E(\Gamma)$ is the set of unordered edges of $\Gamma$.
If $\Gamma'$ is an induced subgraph of $\Gamma$ with vertex set $V' \subset V$, then $C_{\Gamma'}$ is the principal submatrix of $C_\Gamma$ indexed by $V'$. Hence $C_{\Gamma'}$ is positive definite, because every principal submatrix of a positive definite real symmetric matrix is positive definite. We shall use this repeatedly to rule out forbidden induced subgraphs.
[guided]
The graph determines the matrix completely in the simply laced case: diagonal entries are $2$, adjacent vertices contribute $-1$, and non-adjacent distinct vertices contribute $0$. Thus a vector $x \in \mathbb{R}^V$ is the same thing as assigning a real number $x_v$ to each vertex $v$ of the graph, and the matrix condition becomes the positivity of the quadratic form
\begin{align*}
Q_\Gamma(x)
= \sum_{v,w \in V} (C_\Gamma)_{vw}x_vx_w
= 2\sum_{v \in V} x_v^2 - 2\sum_{\{v,w\} \in E(\Gamma)} x_vx_w.
\end{align*}
The factor $2$ in the edge term appears because the symmetric matrix contains both entries $(C_\Gamma)_{vw}$ and $(C_\Gamma)_{wv}$ for an unordered edge $\{v,w\}$.
We also need a heredity observation. If $\Gamma'$ is an induced subgraph with vertex set $V'$, then the matrix $C_{\Gamma'}$ is exactly the principal submatrix of $C_\Gamma$ obtained by restricting to rows and columns in $V'$. If $y \in \mathbb{R}^{V'}$ is nonzero, extend it to $x \in \mathbb{R}^V$ by setting $x_v=y_v$ for $v \in V'$ and $x_v=0$ for $v \notin V'$. Then
\begin{align*}
y^\top C_{\Gamma'}y = x^\top C_\Gamma x > 0.
\end{align*}
Therefore every induced subgraph of $\Gamma$ also has positive definite Cartan matrix. This lets us disprove possible shapes by finding one nonzero vector on a smaller subgraph for which the quadratic form is non-positive.
[/guided]
[/step]
[step:Exclude cycles and vertices of degree at least four]
Suppose first that $\Gamma$ contains a cycle with distinct vertices $v_1,\dots,v_m$, where $m \ge 3$, and edges $\{v_i,v_{i+1}\}$ for $1 \le i<m$ and $\{v_m,v_1\}$. Let $\Gamma'$ be the induced subgraph on these cycle vertices, and define $x \in \mathbb{R}^{V(\Gamma')}$ by $x_{v_i}=1$ for every $i$. The cycle edges alone contribute $m$ edges, and any additional edges in the induced subgraph contribute non-positive terms. In particular, for the cycle subgraph itself one has
\begin{align*}
2\sum_{i=1}^m x_{v_i}^2 - 2\sum_{i=1}^m x_{v_i}x_{v_{i+1}} = 2m - 2m = 0,
\end{align*}
where $v_{m+1}=v_1$. Thus the cycle subgraph has a nonzero vector of quadratic value $0$, contradicting positive definiteness of the corresponding principal submatrix. Hence $\Gamma$ has no cycle.
Now suppose that a vertex $v_0$ has four distinct neighbours $v_1,v_2,v_3,v_4$. Let $\Gamma''$ be the induced subgraph on $\{v_0,v_1,v_2,v_3,v_4\}$, and define $x \in \mathbb{R}^{V(\Gamma'')}$ by
\begin{align*}
x_{v_0}=2,\qquad x_{v_i}=1 \quad \text{for } 1 \le i \le 4.
\end{align*}
Since $\Gamma$ has no cycles, no two of $v_1,\dots,v_4$ are adjacent. Therefore
\begin{align*}
Q_{\Gamma''}(x)
&= 2\left(2^2+\sum_{i=1}^4 1^2\right)-2\sum_{i=1}^4 2\cdot 1\\
&=2(4+4)-16\\
&=0.
\end{align*}
This again contradicts positive definiteness. Therefore every vertex of $\Gamma$ has degree at most $3$.
[guided]
A positive definite simply laced diagram cannot contain a cycle. To see the obstruction, take a cycle with vertices $v_1,\dots,v_m$, where $m \ge 3$, and put the same value $1$ on every vertex of the cycle. On the cycle subgraph, the diagonal contribution is
\begin{align*}
2\sum_{i=1}^m 1^2 = 2m,
\end{align*}
while the edge contribution is
\begin{align*}
2\sum_{i=1}^m 1\cdot 1 = 2m.
\end{align*}
Thus the quadratic form has value $0$ on a nonzero vector. A positive definite matrix cannot have such a vector. Since positive definiteness passes to principal submatrices, this excludes cycles inside $\Gamma$.
Next we rule out degree at least $4$. Suppose $v_0$ has four distinct neighbours $v_1,v_2,v_3,v_4$. Because we have already excluded cycles, two of these neighbours cannot be adjacent: if $v_i$ and $v_j$ were adjacent, then $v_i-v_0-v_j-v_i$ would form a cycle. Now assign value $2$ to the central vertex and value $1$ to the four neighbours:
\begin{align*}
x_{v_0}=2,\qquad x_{v_i}=1 \quad \text{for } 1 \le i \le 4.
\end{align*}
The quadratic form on this five-vertex induced subgraph is
\begin{align*}
Q_{\Gamma''}(x)
&= 2\left(2^2+\sum_{i=1}^4 1^2\right)-2\sum_{i=1}^4 2\cdot 1\\
&= 2(4+4)-16\\
&=0.
\end{align*}
Again this contradicts positive definiteness. Hence every vertex has degree at most $3$.
[/guided]
[/step]
[step:Reduce the classification to paths and three armed trees]
Since $\Gamma$ is connected and has no cycles, it is a tree. Since every vertex has degree at most $3$, either every vertex has degree at most $2$, or there is a vertex of degree $3$.
If every vertex has degree at most $2$, then a finite connected tree with maximum degree $2$ is a path. A path with $\ell$ vertices is the Dynkin diagram $A_\ell$.
It remains to consider the case where $\Gamma$ has a vertex of degree $3$. We first show that there can be only one such vertex. Suppose there are two distinct degree-$3$ vertices $a$ and $b$. Since $\Gamma$ is a tree, there is a unique simple path from $a$ to $b$; let this path have $m \ge 1$ edges. Choose two neighbours of $a$ not lying on this path and two neighbours of $b$ not lying on this path. Let $\Gamma'$ be the induced subgraph consisting of the path from $a$ to $b$ together with these four extra adjacent vertices.
Define $x \in \mathbb{R}^{V(\Gamma')}$ by assigning value $2$ to every vertex on the path from $a$ to $b$ and value $1$ to each of the four extra vertices. The path contains $m+1$ vertices and $m$ edges, and there are four extra edges from the branch vertices to the extra vertices. Hence
\begin{align*}
Q_{\Gamma'}(x)
&=2\left(4(m+1)+4\right)-2\left(4m+4\cdot 2\right)\\
&=8m+16-8m-16\\
&=0.
\end{align*}
This contradicts positive definiteness. Thus $\Gamma$ has exactly one degree-$3$ vertex in the branching case.
Therefore every branching diagram is a three-armed tree: there is a unique branch vertex $b$, and deleting $b$ leaves three path components. Let $p,q,r$ be the numbers of vertices in the three paths from $b$ to the corresponding terminal vertices, counted including $b$. Reorder them so that
\begin{align*}
2 \le p \le q \le r.
\end{align*}
[guided]
Once cycles are excluded, connectedness says that $\Gamma$ is a tree. A finite connected tree in which all degrees are at most $2$ must be a path: starting at one endpoint and walking along the unique unused edge lists all vertices in a single chain. Therefore this case gives exactly $A_\ell$, where $\ell$ is the number of vertices.
Now assume that some vertex has degree $3$. We need to know whether there can be more than one branch point. Suppose two distinct degree-$3$ vertices exist; call them $a$ and $b$. Since the graph is a tree, there is exactly one simple path from $a$ to $b$. Let this path have $m$ edges, so it has $m+1$ vertices. At $a$, one edge lies on the path toward $b$, so there are two other incident edges. At $b$, similarly, there are two other incident edges. Keep just the path and these four additional neighbouring vertices.
Now put value $2$ on every vertex of the path from $a$ to $b$ and value $1$ on the four extra leaves. The diagonal part of the quadratic form is
\begin{align*}
2\left(4(m+1)+4\right),
\end{align*}
because there are $m+1$ path vertices with square $4$ and four extra vertices with square $1$. The edge part consists of $m$ path edges, each with product $2\cdot2=4$, and four leaf edges, each with product $2\cdot1=2$. Therefore
\begin{align*}
Q_{\Gamma'}(x)
&=2\left(4(m+1)+4\right)-2\left(4m+4\cdot 2\right)\\
&=0.
\end{align*}
This nonzero vector contradicts positive definiteness. Hence a finite simply laced positive definite connected diagram has at most one branch point.
So the branching case has a single central vertex $b$ of degree $3$. Removing $b$ leaves three disjoint paths. We encode their lengths by integers $p,q,r$, where each number counts the vertices in the corresponding arm including the common branch vertex $b$. Thus each of $p,q,r$ is at least $2$, and after reordering we may assume
\begin{align*}
2 \le p \le q \le r.
\end{align*}
[/guided]
[/step]
[step:Compute the determinant of a three armed tree]
Let $T_{p,q,r}$ denote the three-armed tree with one branch vertex and arms of vertex-lengths $p,q,r$ counted including the branch vertex. We compute the determinant of its Cartan matrix.
For $n \ge 1$, let $A_n$ denote the path on $n$ vertices, and let $C_{A_n}$ be its Cartan matrix. Define $d_n=\det C_{A_n}$, with the convention $d_0=1$. Expanding the tridiagonal determinant along the last row gives
\begin{align*}
d_n = 2d_{n-1}-d_{n-2} \quad \text{for } n \ge 2,
\end{align*}
with $d_0=1$ and $d_1=2$. Hence $d_n=n+1$ for all $n \ge 0$.
Order the vertices of $T_{p,q,r}$ so that the branch vertex comes first and the three arms without the branch vertex follow. The block corresponding to the three deleted arms is
\begin{align*}
B = C_{A_{p-1}} \oplus C_{A_{q-1}} \oplus C_{A_{r-1}}.
\end{align*}
The branch vertex has diagonal entry $2$ and is adjacent to the first vertex in each arm. By the Schur complement formula,
\begin{align*}
\det C_{T_{p,q,r}}
= \det(B)\left(2-u^\top B^{-1}u\right),
\end{align*}
where $u$ is the column vector whose nonzero entries are $-1$ in the first coordinate of each arm block.
For the path matrix $C_{A_n}$, [Cramer's rule](/theorems/3305) gives
\begin{align*}
(C_{A_n}^{-1})_{11}=\frac{\det C_{A_{n-1}}}{\det C_{A_n}}=\frac{n}{n+1}.
\end{align*}
Applying this with $n=p-1,q-1,r-1$, we obtain
\begin{align*}
u^\top B^{-1}u
= \frac{p-1}{p}+\frac{q-1}{q}+\frac{r-1}{r}.
\end{align*}
Also
\begin{align*}
\det(B)=pqr.
\end{align*}
Therefore
\begin{align*}
\det C_{T_{p,q,r}}
&=pqr\left(2-\frac{p-1}{p}-\frac{q-1}{q}-\frac{r-1}{r}\right)\\
&=pqr\left(\frac1p+\frac1q+\frac1r-1\right).
\end{align*}
Since $C_{T_{p,q,r}}$ is positive definite, its determinant is positive. Hence
\begin{align*}
\frac1p+\frac1q+\frac1r>1.
\end{align*}
[guided]
The classification now depends on a numerical restriction on the three arm lengths. Let $T_{p,q,r}$ be the three-armed tree with arm lengths $p,q,r$, counted including the common branch vertex. We compute the determinant of its Cartan matrix and use the fact that a positive definite matrix has positive determinant.
First consider a path $A_n$ with $n$ vertices. Its Cartan matrix is tridiagonal, with diagonal entries $2$ and adjacent off-diagonal entries $-1$. Let
\begin{align*}
d_n=\det C_{A_n}, \qquad d_0=1.
\end{align*}
Expanding the tridiagonal determinant along the last row gives
\begin{align*}
d_n=2d_{n-1}-d_{n-2} \quad \text{for } n \ge 2,
\end{align*}
with $d_1=2$. The sequence $d_n=n+1$ satisfies the same recurrence and the same initial conditions, so
\begin{align*}
\det C_{A_n}=n+1.
\end{align*}
Now remove the branch vertex from $T_{p,q,r}$. The remaining graph is the disjoint union of three paths with $p-1$, $q-1$, and $r-1$ vertices. Thus the corresponding block of the Cartan matrix is
\begin{align*}
B=C_{A_{p-1}}\oplus C_{A_{q-1}}\oplus C_{A_{r-1}}.
\end{align*}
The branch vertex contributes a scalar diagonal block $2$. It is connected to the first vertex in each arm, so the off-diagonal vector $u$ has exactly three nonzero entries, each equal to $-1$, one in the first coordinate of each path block. The Schur complement formula gives
\begin{align*}
\det C_{T_{p,q,r}}
= \det(B)\left(2-u^\top B^{-1}u\right).
\end{align*}
We need the first diagonal entry of the inverse of a path Cartan matrix. By Cramer's rule, deleting the first row and first column of $C_{A_n}$ leaves $C_{A_{n-1}}$, so
\begin{align*}
(C_{A_n}^{-1})_{11}
=\frac{\det C_{A_{n-1}}}{\det C_{A_n}}
=\frac{n}{n+1}.
\end{align*}
Using this for the three arm blocks gives
\begin{align*}
u^\top B^{-1}u
= \frac{p-1}{p}+\frac{q-1}{q}+\frac{r-1}{r}.
\end{align*}
Since
\begin{align*}
\det(B)=pqr,
\end{align*}
we obtain
\begin{align*}
\det C_{T_{p,q,r}}
&=pqr\left(2-\frac{p-1}{p}-\frac{q-1}{q}-\frac{r-1}{r}\right)\\
&=pqr\left(\frac1p+\frac1q+\frac1r-1\right).
\end{align*}
Positive definiteness implies that this determinant is positive. Because $pqr>0$, the required numerical restriction is
\begin{align*}
\frac1p+\frac1q+\frac1r>1.
\end{align*}
[/guided]
[/step]
[step:Solve the branch length inequality]
Let $2 \le p \le q \le r$ be integers satisfying
\begin{align*}
\frac1p+\frac1q+\frac1r>1.
\end{align*}
If $p \ge 3$, then $q \ge 3$ and $r \ge 3$, so
\begin{align*}
\frac1p+\frac1q+\frac1r \le 1,
\end{align*}
a contradiction. Hence $p=2$.
If $q \ge 4$, then $r \ge q \ge 4$, and therefore
\begin{align*}
\frac1p+\frac1q+\frac1r
\le \frac12+\frac14+\frac14
=1,
\end{align*}
again a contradiction. Thus $q=2$ or $q=3$.
If $q=2$, then $r \ge 2$ is arbitrary. The corresponding three-armed tree has two arms of length $2$ and one arm of length $r$; it has $r+2$ vertices and is the diagram $D_{r+2}$. Since $r \ge 2$, this gives $D_\ell$ for $\ell \ge 4$.
If $q=3$, then
\begin{align*}
\frac12+\frac13+\frac1r>1
\end{align*}
is equivalent to $r<6$. Since $r \ge 3$, this gives
\begin{align*}
r=3,\quad r=4,\quad r=5.
\end{align*}
The corresponding diagrams are respectively $E_6$, $E_7$, and $E_8$.
Combining this with the path case, every connected simply laced positive definite diagram is one of
\begin{align*}
A_\ell \quad (\ell \ge 1), \qquad D_\ell \quad (\ell \ge 4), \qquad E_6,\ E_7,\ E_8.
\end{align*}
[/step]
[step:Realize the classical diagrams by coordinate root systems]
For $\ell \ge 1$, let $(e_1,\dots,e_{\ell+1})$ be the standard orthonormal basis of $\mathbb{R}^{\ell+1}$, and define
\begin{align*}
\Phi(A_\ell)=\{e_i-e_j : 1 \le i \ne j \le \ell+1\}.
\end{align*}
This is a finite reduced crystallographic root system in the hyperplane $\sum_{i=1}^{\ell+1} x_i=0$. The simple roots
\begin{align*}
\alpha_i=e_i-e_{i+1}, \qquad 1 \le i \le \ell,
\end{align*}
satisfy $(\alpha_i,\alpha_i)=2$, $(\alpha_i,\alpha_{i+1})=-1$, and $(\alpha_i,\alpha_j)=0$ when $|i-j|>1$. Hence their Dynkin diagram is $A_\ell$.
For $\ell \ge 4$, let $(e_1,\dots,e_\ell)$ be the standard orthonormal basis of $\mathbb{R}^\ell$, and define
\begin{align*}
\Phi(D_\ell)=\{\pm e_i \pm e_j : 1 \le i<j \le \ell\}.
\end{align*}
This is a finite reduced crystallographic root system. Choose simple roots
\begin{align*}
\alpha_i &= e_i-e_{i+1}, \qquad 1 \le i \le \ell-1,\\
\alpha_\ell &= e_{\ell-1}+e_\ell.
\end{align*}
A direct inner-product computation gives $(\alpha_i,\alpha_i)=2$ for all $i$, edges along the chain $\alpha_1,\dots,\alpha_{\ell-1}$, and the additional edge between $\alpha_{\ell-2}$ and $\alpha_\ell$, with all other non-diagonal inner products equal to $0$. Hence the Dynkin diagram is $D_\ell$.
[guided]
The converse starts with the standard coordinate root systems. For type $A_\ell$, the roots are all differences $e_i-e_j$ inside the hyperplane of $\mathbb{R}^{\ell+1}$ where the coordinate sum is zero:
\begin{align*}
\Phi(A_\ell)=\{e_i-e_j : 1 \le i \ne j \le \ell+1\}.
\end{align*}
The usual simple roots are consecutive differences,
\begin{align*}
\alpha_i=e_i-e_{i+1}, \qquad 1 \le i \le \ell.
\end{align*}
Each has squared length $2$. Adjacent simple roots satisfy
\begin{align*}
(\alpha_i,\alpha_{i+1})
=(e_i-e_{i+1},e_{i+1}-e_{i+2})
=-1,
\end{align*}
and non-adjacent simple roots have inner product $0$. Since the simply laced Dynkin diagram has an edge exactly when the inner product is $-1$, this gives the path $A_\ell$.
For type $D_\ell$, the roots are
\begin{align*}
\Phi(D_\ell)=\{\pm e_i \pm e_j : 1 \le i<j \le \ell\}.
\end{align*}
Take
\begin{align*}
\alpha_i &= e_i-e_{i+1}, \qquad 1 \le i \le \ell-1,\\
\alpha_\ell &= e_{\ell-1}+e_\ell.
\end{align*}
The roots $\alpha_1,\dots,\alpha_{\ell-1}$ form a chain. The last root $\alpha_\ell$ is orthogonal to $\alpha_i$ for $i<\ell-2$, and
\begin{align*}
(\alpha_{\ell-2},\alpha_\ell)
=(e_{\ell-2}-e_{\ell-1},e_{\ell-1}+e_\ell)
=-1.
\end{align*}
Also
\begin{align*}
(\alpha_{\ell-1},\alpha_\ell)
=(e_{\ell-1}-e_\ell,e_{\ell-1}+e_\ell)
=0.
\end{align*}
Thus the branch occurs at $\alpha_{\ell-2}$, giving exactly $D_\ell$.
[/guided]
[/step]
[step:Realize the exceptional diagrams inside the standard $E_8$ root system]
Let $(e_1,\dots,e_8)$ be the standard orthonormal basis of $\mathbb{R}^8$, and define
\begin{align*}
\Phi(E_8)
&=
\{\pm e_i \pm e_j : 1 \le i<j \le 8\}\\
&\qquad \cup
\left\{
\frac12\sum_{i=1}^8 \varepsilon_i e_i :
\varepsilon_i \in \{\pm 1\},\ \prod_{i=1}^8 \varepsilon_i=1
\right\}.
\end{align*}
This is the standard finite reduced crystallographic root system of type $E_8$. Define roots
\begin{align*}
\alpha_1&=\frac12(e_1-e_2-e_3-e_4-e_5-e_6-e_7+e_8),\\
\alpha_2&=e_1+e_2,\\
\alpha_3&=e_2-e_1,\\
\alpha_4&=e_3-e_2,\\
\alpha_5&=e_4-e_3,\\
\alpha_6&=e_5-e_4,\\
\alpha_7&=e_6-e_5,\\
\alpha_8&=e_7-e_6.
\end{align*}
Each $\alpha_i$ belongs to $\Phi(E_8)$ and has squared length $2$. Direct computation gives the nonzero off-diagonal inner products
\begin{align*}
(\alpha_1,\alpha_3)&=-1,\\
(\alpha_2,\alpha_4)&=-1,\\
(\alpha_3,\alpha_4)&=-1,\\
(\alpha_4,\alpha_5)&=-1,\\
(\alpha_5,\alpha_6)&=-1,\\
(\alpha_6,\alpha_7)&=-1,\\
(\alpha_7,\alpha_8)&=-1,
\end{align*}
and all other off-diagonal inner products are $0$. Therefore the Dynkin diagram of $\alpha_1,\dots,\alpha_8$ is $E_8$.
The subsets $\{\alpha_1,\dots,\alpha_7\}$ and $\{\alpha_1,\dots,\alpha_6\}$ have the induced diagrams $E_7$ and $E_6$, respectively. The root subsystems generated by these subsets inside $\Phi(E_8)$ are finite, reduced, and crystallographic because they are closed subsystems of the finite reduced crystallographic root system $\Phi(E_8)$. Hence $E_6$, $E_7$, and $E_8$ are all realized by finite reduced crystallographic root systems.
Together with the coordinate realizations of $A_\ell$ and $D_\ell$, this proves the converse and completes the classification.
[/step]
Prerequisites (0/3 completed)
Prerequisites Graph
Interactive dependency map showing how this theorem builds on foundational concepts
Loading dependency graph...
Theorem
Definition
Current
Requires
Theorems
Definitions & Concepts
Explore Further
Connectedness
Definition
Determinant
Definition
Cramer's Rule
Theorem #3305
Cayley-Hamilton Theorem
Linear Algebra
Completeness of Irreducible Characters
Representation Theory
Cauchy-Schwarz Inequality
Linear Algebra
Dimension Formula For Subspace Sums
Linear Algebra
Sign Homomorphism
Group Theory
Reduction by a Non-Zero-Divisor
Algebra
Sylow's Theorems
Group Theory
Class Sums Form a Basis for the Center
Representation Theory
Algebra
Area