## On complete Lie algebras

### 2003.5

Abstract
This paper gives two methods to construct complete Lie algebras. Both of the methods show that the derivation algebras of Lie algebras of certain types are complete Lie algebras. The first method is different from any known methods and gives a way to construct all solvable complete Lie algebras.
Key word: complete Lie algebra, pair of type $\mathbf{I}$  , pair of type ${\mathbf{I}}_{1}$  , nilpotent Lie algebra of type $\mathbf{I}$  , nilpotent Lie algebra of type ${\mathbf{I}}_{1}$  , full graph MR(2000) Subject Classification: 17B45, 17B55, 17B56

1 Introduction

A Lie algebra $g$  is called a complete Lie algebra if its center $C\left(g\right)$  is trivial and all of its derivations are inner, i.e. $Der\left(g\right)=ad\left(g\right)$  .
D.J.Meng and L.S.Zhu have obtained a lot of results for complete Lie algebras.
Their results are based on the following basic fact.
• Let $g$  be a Lie algebra, and let $b$  be its Cartan subalgebra. Assume the following conditions hold:
(1). $b$  is commutative.
(2). $g$  has a decomposition $g=b\oplus {\sum }_{\alpha \in \Delta }{g}_{\alpha },$  where, $\Delta \in b*\\left\{0\right\}$  , and ${g}_{\alpha }=\left\{x\in g|\left[h,x\right]=\alpha \left(h\right)x,h\in b\right\}$  .
(3). There exists a basis ${\alpha }_{1},...,{\alpha }_{l}$  , such that $dim{g}_{±{\alpha }_{j}}\le 1,$  and that $\left[{g}_{{\alpha }_{j}},{g}_{-{\alpha }_{j}}\right]\ne 0$  , if $-{\alpha }_{j}\in \Delta$  .
(4). $b$  and $\left\{{g}_{±{\alpha }_{j}}1\le j\le l\right\}$  generate $g$  .
Then $g$  is a complete Lie algebra.
Their methods are limited, by which one can't even get all solvable complete Lie algebras.
In this paper, we give a generalize of the above basic fact (i.e. Theorem 2). Due to this fact, one can construct more complete Lie algebras and get all solvable Lie algebras. So our method gives a new way for the study of complete Lie algebras.
Section 2 introduces two concepts, which are pair $\left(g,b\right)$  of type $\mathbf{I}$  and that of type ${\mathbf{I}}_{1}$  . The main result (i.e. Theorem 2) of this section is that, the derivation algebra $h:=Der\left({h}_{1}\right)$  of ${h}_{1}:=b{×}_{t}g$  with $\left(g,b\right)$  of type ${\mathbf{I}}_{1}$  , is a complete Lie algebra. Pair of type $\mathbf{I}$  is a special case of pair of type ${\mathbf{I}}_{1}$  . But our method is more natural for type $\mathbf{I}$  than it for type ${\mathbf{I}}_{1}$  , so we state the method in the case of type $\mathbf{I}$  firstly, then show it in fact remains effective in the case of type ${\mathbf{I}}_{1}$  . Applying our main result, we gain a result (i.e. Theorem 3) for solvable Lie algebras, which shows that there exists a one to one correspondence between solvable Lie algebras and nilpotent Lie algebras of type ${\mathbf{I}}_{1}$  . We leave the problem of how to calculate the derivation algebra of a given lie algebra for the future.
In section 3, at first, we give a result to calculate the derivation algebra of the full graph of Lie algebra of certain type. Then we apply it to Heisenberg algebra $g$  and show that $Der\left({C}^{n}g\right)$  are complete Lie algebras for all positive integers $n$  . Here, $Der\left({C}^{n}g\right)$  means $n$  -power full graph of $g$  .
There is no direct relation between section 2 and section 3. All of the Lie algebras are defined on any field, except that the Lie algebras mentioned in Theorem 3 are defined on an algebraic closed field.

2 Complete Lie algebras of type ${\mathbf{I}}_{1}$

2.1 Notations and concepts

Let $g$  be a Lie algebra, and $Der\left(g\right)$  be its derivation algebra. A Lie subalgebra $b$  of $Der\left(g\right)$  is called a (maximal) torus on $g$  if and only if $b$  is a (maximal) subalgebra commutative subalgebra whose elements are all semisimple.
Fix a torus $b$  , then $g$  can be decomposed into
 $\begin{array}{c}g=\oplus {g}_{\alpha }\end{array}$ (1)
with $\alpha \in {b}^{*}$  and ${g}_{\alpha }=\left\{g\in g:D\left(g\right)=〈D,\alpha 〉g,\forall D\in b\right\}$  .
We call the pair $\left(g,b\right)$  is of type $\mathbf{I}$  if all $\alpha$  in formula (1) are nonzero. We recall that in the case $g$  is nilpotent and ${b}_{i}\left(i=1,2\right)$  is maximal torus on $g$  , there is some $\theta \in Autg$  such that $\theta \left({b}_{1}\right)={b}_{2}$  . So $\left(g,{b}_{1}\right)$  is of type $I$  if and only if $\left(g,{b}_{2}\right)$  is of type $I$  . In this case, we shall say simply that $g$  is a nilpotent Lie algebra of type $I$  .
If there is a homomorphism from Lie algebra $s$  to $Der\left(g\right)$  , we can define a Lie algebra $s{×}_{t}g$  by:
 $\begin{array}{c}\left[\left({s}_{1},{g}_{1}\right)\left({s}_{2},{g}_{2}\right)\right]=\left(\left[{s}_{1},{s}_{2}\right],{s}_{1}\left({g}_{2}\right)-{s}_{2}\left({g}_{1}\right)+\left[{g}_{1},{g}_{2}\right]\right)\end{array}$ (2)
It is easy to check that $\left(s{×}_{t}g,\left[,\right]\right)$  is really a Lie algebra. In the case $s=Der\left(g\right)$  , we call $s{×}_{t}g$  the full graph of $g$  and write it by $C\left(g\right)$  .

2.2 Complete Lie algebras of type $\mathbf{I}$

In this subsection, we assume $b$  is a torus on $g$  and $\left(g,b\right)$  is of type $I$  . Let $\tau$  be the commutator $b$  in $Der\left(g\right)$  , that is $\left\{D\in Der\left(g\right):D{D}_{1}-{D}_{1}D=0\forall {D}_{1}\in b\right\}$  We denote $b{×}_{t}g$  by ${h}_{1}$  and $\tau ×g$  by $h$  . Then ${h}_{1}$  , $h$  can be written respectively as ${h}_{1}=b\oplus {\sum }_{\alpha \in \Delta }{g}_{\alpha }$  $h=\tau \oplus {\sum }_{\alpha \in \Delta }{g}_{\alpha }$  where $\Delta \subset {b}^{*}\\left\{0\right\}$  and ${g}_{\alpha }=\left\{x\in g:\left[\left(b,0\right),\left(0,x\right)\right]=\alpha \left(b\right)\left(0,x\right)\forall b\in b\right\}$  .
We can identity $b$  with $\left\{\left(b,0\right):b\in b\right\}$  in ${h}_{1}$  , $\tau$  with $\left\{\left(t,0\right):t\in \tau \right\}$  in $h$  and $g$  with $\left\{\left(0,g\right):g\in g\right\}$  in ${h}_{1}$  and $h$  respectively.
Lemma 1. Let $g$  , $b$  , $\tau$  be as above.
(i) Assume $D\in Der\left(h\right)$  such that $D\left(b\right)\subseteq \tau$  Then $D=ad\left(t\right)$  for some $t\in \tau$  (ii) Assume $D\in Der\left({h}_{1}\right)$  such that $D\left(b\right)\subseteq b.$  Then $D|b=0$  and $D|g=t$  for some $t\in \tau$  .
• Proof. We shall prove (i). (ii) can be proved similarly.
We notice first that $\tau$  fixes every ${g}_{\alpha }$  , since $\tau$  commutes with $b$  .
For any $\alpha \in \Delta ,{e}_{\alpha }\in {g}_{\alpha }$  , let $D{e}_{\alpha }={t}_{1}+{\sum }_{\beta \in \Delta }{x}_{\beta }$  with ${t}_{1}\in \tau$  and ${x}_{\beta }\in {g}_{\beta }$  . From $D\left[b,{e}_{\alpha }\right]=\left[Db,{e}_{\alpha }\right]+\left[b,D{e}_{\alpha }\right]$  with $b\in b$  , we gain:
 $\begin{array}{c}\alpha \left(b\right)\left({t}_{1}+{\sum }_{\beta \in \Delta }{x}_{\beta }\right)=\left[Db,{e}_{\alpha }\right]+{\sum }_{\beta \in \Delta }\beta \left(b\right){x}_{\beta }\end{array}$ (3)
and notice that:
 $\begin{array}{c}\left[Db,{e}_{\alpha }\right]\in {g}_{\alpha }.\end{array}$ (4)
The above formulas (3) and (4) show that  $\begin{array}{c}\left[Db,{e}_{\alpha }\right]=0\forall \alpha \in \Delta ,b\in b\end{array}$ (5)
 $\begin{array}{c}\alpha \left(b\right){x}_{\beta }=\beta \left(b\right){x}_{\beta }\forall \beta \in \Delta ,b\in b\end{array}$ (6)
 $\begin{array}{c}\alpha \left(b\right){t}_{1}=0\forall b\in b.\end{array}$ (7)
From (5), we gain $Db=0\forall b\in b$  . From (6), we gain  $\begin{array}{c}{x}_{\beta }=0\forall \beta \ne \alpha .\end{array}$ (8)
From (7), we gain ${t}_{1}=0$  since $\alpha \ne 0$  . Thus for any $\alpha \in \Delta$  ,  $\begin{array}{c}D\left({g}_{\alpha }\right)\subseteq {g}_{\alpha }\end{array}$ (9)
Now (9) tells us that $D|g$  commutate with $ad\left(b\right)|g$  , therefore there exists a $t\in \tau$  , such that  $\begin{array}{c}D|g=ad\left(t\right)|g\end{array}$ (10)
Let ${D}_{1}=D-ad\left(t\right)$  , then ${D}_{1}|b\oplus g=0$  Let $t\in \tau$  . From  $\begin{array}{c}{D}_{1}\left[t,x\right]=\left[{D}_{1}t,x\right]+\left[t,{D}_{1}x\right]\forall x\in b\oplus g\end{array}$ (11)
we gain  $\begin{array}{c}\left[{D}_{1}t,x\right]=0\forall x\in b\oplus g\end{array}$ (12)
Let ${D}_{1}t={t}^{\prime }+{x}^{\prime }$  with ${t}^{\prime }\in \tau ,{x}^{\prime }\in g$  . Then  $\begin{array}{c}\left[{x}^{\prime },b\right]=0\forall b\in b\end{array}$ (13)
Therefore ${x}^{\prime }=0$  , and ${D}_{1}t\in \tau$  . Now  $\begin{array}{c}\left[{D}_{1}t,g\right]=0\forall g\in g\end{array}$ (14)
shows that ${D}_{1}t=0$  as a element in $Der\left(g\right)$  and therefore ${D}_{1}t=0$  .
We gain ${D}_{1}=0$  and $D=ad\left(t\right)$  for some $t\in \tau$  . This proves (i).
Lemma 2 Let $g,b,\tau ,{h}_{1}$  and $h$  be as above.
(i) Assume $D\in Der\left(h\right)$  . then there exist ${x}_{\alpha }\in {g}_{\alpha }$  , such that $\forall b\in b$  , we have $Db={b}^{\prime }+{\sum }_{\alpha \in \Delta }\alpha \left(b\right){x}_{\alpha }$  with ${b}^{\prime }\in \tau$  , and ${x}_{\alpha }$  depend not on $b$  but only on $D$  .
(ii). Assume $D\in Der\left({h}_{1}\right)$  . Then there exist ${x}_{\alpha }\in {g}_{\alpha }$  , such that $\forall b\in b$  , we have $Db={b}^{\prime }+{\sum }_{\alpha \in \Delta }\alpha \left(b\right){x}_{\alpha }$  with ${b}^{\prime }\in b$  , and ${x}_{\alpha }$  depend not on $b$  but only on $D$  .
• Proof. As in the proof of Lemma 1, we only prove (i). (ii) can be proved similarly.
Assume that  $\begin{array}{c}Db={b}^{\prime }+{\sum }_{\alpha \in \Delta }{y}_{\alpha }\left(b\right)\end{array}$ (15)
with ${b}^{\prime }\in \tau$  and ${y}_{\alpha }\left(b\right)\in {g}_{\alpha }$  . As  $\begin{array}{c}\left[b,{b}_{1}\right]=0\forall {b}_{1}\in b\end{array}$ (16)
we gain  $\begin{array}{c}-{\sum }_{\alpha \in \Delta }\alpha \left({b}_{1}\right){y}_{\alpha }\left(b\right)+{\sum }_{\alpha \in \Delta }\alpha \left(b\right){y}_{\alpha }\left({b}_{1}\right)=0.\end{array}$ (17)
That is  $\begin{array}{c}\alpha \left({b}_{1}\right){y}_{\alpha }\left(b\right)=\alpha \left(b\right){y}_{\alpha }\left({b}_{1}\right)\forall \alpha \in \Delta ,b,{b}_{1}\in b\end{array}$ (18)
If $\alpha \left(b\right)=0$  , take an ${b}_{1}\in b$  such that $\alpha \left({b}_{1}\right)\ne 0$  , then formula (18) implies ${y}_{\alpha }\left(b\right)=0$  . If ${y}_{\alpha }\left(b\right)=0\forall b\in b$  , we let ${x}_{\alpha }=0$  . Now assume that there exists an $b\in b$  such that ${y}_{\alpha }\left(b\right)\ne 0$  , then $\alpha \left(b\right)\ne 0$  . We take ${x}_{\alpha }=\frac{1}{\alpha \left(b\right)}{y}_{\alpha }\left(b\right).$  Then formula (18) shows that  $\begin{array}{c}{y}_{\alpha }\left({b}_{1}\right)=\frac{\alpha \left({b}_{1}\right)}{\alpha \left(b\right)}{y}_{\alpha }\left(b\right)=\alpha \left({b}_{1}\right){x}_{\alpha }.\end{array}$ (19)
We have now completed the proof of (i)
We can now state the main result of this section.
Theorem 1. Let $g$  be a Lie algebra, $b$  be a torus on $g$  such that $\left(g,b\right)$  is of type $\mathbf{I}$  . Let $\tau$  be the commutator of $b$  in $Der\left(g\right)$  , and $h=\tau {×}_{t}g$  , ${h}_{1}=b{×}_{t}g$  .
Then $h=Der\left({h}_{1}\right)$  and $h$  is a complete Lie algebra.
• Proof. (i). We prove first that $h=Der\left({h}_{1}\right)$  .
Let $D\in Der\left({h}_{1}\right)$  . Then lemma 2 (ii) tells us that $Db={b}^{\prime }+{\sum }_{\alpha \in \Delta }\alpha \left(b\right){x}_{\alpha }\forall b\in b$  Therefore $D+ad\left({\sum }_{\alpha \in \Delta }{x}_{\alpha }\right)\in Der\left({h}_{1}\right)$  is derivation on ${h}_{1}$  such that  $\begin{array}{c}\left(D+ad\left({\sum }_{\alpha \in \Delta }{x}_{\alpha }\right)\right)b={b}^{\prime }\in b\forall b\in b\end{array}$ (20)
Lemma 1(ii) shows that $D+ad\left({\sum }_{\alpha \in \Delta }\right)|b=0$  , and  $\begin{array}{c}D+ad\left({\sum }_{\alpha \in \Delta }{x}_{\alpha }\right)|g=t\end{array}$ (21)
for some $t\in \tau$  .
We define the action of $h$  on ${h}_{1}$  by  $\begin{array}{c}{D}_{\left({t}_{0},{g}_{0}\right)}\left(\left(b,g\right)\right)=\left(0,{t}_{0}\left(g\right)-b\left({g}_{0}\right)+\left[{g}_{0},g\right]\right)\end{array}$ (22)
It is easy to check that the action in this way gives a homomorphism from $h$  to $Der\left({h}_{1}\right)$  . And ${D}_{\left({t}_{0},{g}_{0}\right)}=0$  if and only if $\left({t}_{0},{g}_{0}\right)=0$  . Thus we can regard $h$  as a Lie subalgebra of $Der\left({h}_{1}\right)$  via (22). But (21) shows that every $D\in Der\left({h}_{1}\right)$  lies in $h$  . Therefore $Der\left({h}_{1}\right)=h$  .
(ii). We are now going to prove $h$  is complete.
First, we prove $C\left(h\right)=0$  . Assume $c\in C\left(h\right)$  , then $c\in \tau$  . From $\left[c,g\right]=0\forall g\in g$  we gain $c=0$  . So $C\left(h\right)=0$  .
Now, we prove $Der\left(h\right)=ad\left(h\right)$  . Let $D\in Der\left(h\right)$  , then lemma 2(i) tells us that $Db={b}^{\prime }+{\sum }_{\alpha \in \Delta }\alpha \left(b\right){x}_{\alpha }\forall b\in b$  Therefore  $\begin{array}{c}\left(D+ad\left({\sum }_{\alpha \in \Delta }{x}_{\alpha }\right)\right)b={b}^{\prime }\in \tau \forall b\in b\end{array}$ (20′)
Lemma 1(i) tells us that there exists some $t\in \tau$  such that  $\begin{array}{c}D+ad\left({\sum }_{\alpha \in \Delta }{x}_{\alpha }\right)=ad\left(t\right)\end{array}$ (21′)
So $D\in ad\left(h\right)$  and $Der\left(h\right)=ad\left(h\right)$  .
Therefore $h$  is a complete Lie algebra.
If we regard ${h}_{1}$  as a subalgebra of $h$  in the natural way, then (22) is just the Lie bracket in $h$  , and ${h}_{1}$  is an ideal of $h$  .
Now we apply the above theorem to solvable complete Lie algebra.
A nilpotent Lie algebra is called completable if it is the nilpotent radical of some solvable Lie algebra.
Let $n$  be a nilpotent Lie algebra of type $\mathbf{I}$  , $b$  be a maximal torus on $n$  , and $\tau$  be the commutator of $b$  in $Der\left(n\right)$  . Then $\tau {×}_{t}n$  is the derivation algebra of $b{×}_{t}n$  and it is complete. We recall that a algebraic Lie algebra (for example:
$Der\left(n\right)$  ) has the following decomposition:
$\mathcal{S}+\mathcal{A}+\mathcal{N}$  with $\mathcal{S}$  a levi subalgebra, $\mathcal{A}$  is a torus, $\left[\mathcal{S},\mathcal{A}\right]=0$  and $\mathcal{N}$  is the nilpotent radical.
So $\tau$  is a solvable Lie algebra and so is $\tau {×}_{t}n$  . We have the following proposition:
Proposition 1 Let $n$  be a nilpotent Lie algebra of type $\mathbf{I}$  , $b$  be a maximal torus on $n$  , and $\tau$  be the commutator of $b$  in $Der\left(n\right)$  . Then $\tau {×}_{t}n$  is the derivation algebra of the solvable Lie algebra $b{×}_{t}n$  , and $\tau {×}_{t}n$  is a solvable complete Lie algebra. And $n$  is completable if and only if $\tau =b$  .
Given a nilpotent Lie algebra of type $\mathbf{I}$  , we can obtain a solvable complete lie algebra. but there are some solvable complete Lie algebras that are not obtained in this way. Therefore we should enlarge the concept of nilpotent Lie algebra of type $I$  .

2.3 Solvable complete Lie algebras and nilpotent Lie algebras of type ${\mathbf{I}}_{1}$

Let $g$  be a Lie algebra, and $b$  be a torus on $g$  . Then $g$  can be decomposed into
 $\begin{array}{c}g={\oplus }_{\alpha \in \Delta }{g}_{\alpha }\end{array}$ (23)
with $\alpha \in {b}^{*}$  and ${g}_{\alpha }=\left\{g\in g:D\left(g\right)=g,\forall D\in b\right\}$  .
We say that the pair $\left(g,b\right)$  is of type ${\mathbf{I}}_{1}$  if ${g}_{0}\subseteq \left[g,g\right]$  and ${g}_{0}\cap C\left(g\right)=\left\{0\right\}$  . A nilpotent Lie algebra $n$  is called of type ${I}_{1}$  if the pair $\left(n,b\right)$  with maximal torus $b$  on $n$  is of type ${\mathbf{I}}_{1}$  .
We shall show that the method used in subsection 2 remains valid in the case of type ${\mathbf{I}}_{1}$  .
Let $\left(g,b\right)$  be of type ${\mathbf{I}}_{1}$  . Let $\tau$  be the commutator of $b$  in $Der\left(g\right)$  . We can extend $\tau$  to derivations on $b{×}_{t}g$  by setting $\tau |b=0$  . Let ${h}_{1}=b{×}_{t}g$  . But now $h$  is no longer $\tau {×}_{t}g$  , but the sum (not direct sum) of $\tau$  and $\left\{ad\left(g\right):g\in g\right\}$  in $Der\left(b{×}_{t}g\right)$  . We notice that $\tau \cap \left\{ad\left(g\right):g\in g\right\}=\left\{ad\left(g\right):g\in {g}_{0}\right\}$  and that $h=\tau {×}_{t}g$  if ${g}_{0}=\left\{0\right\}$  . In general
 $\begin{array}{c}h=\tau {×}_{t}g/\mathbf{r}\end{array}$ (24)
with $\mathbf{r}=\left\{\left(t,x\right):t=-ad\left(x\right)\in Der\left(b{×}_{t}g\right)\right\}$  .
Lemma 3. (i). Assume $D\in Der\left(h\right)$  such that $D\left(b\right)\subseteq \tau .$  Then $D=ad\left(t\right)$  for some $t\in \tau$  .
(ii). Assume $D\in Der\left({h}_{1}\right)$  such that $D\left(b\right)\subseteq b+{g}_{0}.$  Then $D=t$  for some $t\in \tau$  .
Lemma 4. (i). Assume $D\in Der\left(h\right)$  . Then there is some ${x}_{\alpha }\in {g}_{\alpha }$  for each $\alpha \ne 0$  , such that for any $b\in b$  , we have $Db={b}^{\prime }+{\sum }_{\alpha \in \Delta \\left\{0\right\}}\alpha \left(b\right){x}_{\alpha }$  with ${b}^{\prime }\in \tau$  , and ${x}_{\alpha }$  depends not on $b$  but only on $D$  .
(ii). Assume $D\in Der\left({h}_{1}\right)$  . Then there is some ${x}_{\alpha }\in {g}_{\alpha }$  for each $\alpha \ne 0$  , such that for any $b\in b$  , we have $Db={b}^{\prime }+{\sum }_{\alpha \in \Delta \\left\{0\right\}}\alpha \left(b\right){x}_{\alpha }$  with ${b}^{\prime }\in b+{g}_{0}$  and ${x}_{\alpha }$  depends not on $b$  but only on $D$  .
Theorem 2. Let $g$  be a Lie algebra, $b$  be a torus on $g$  such that $\left(g,b\right)$  is of type ${\mathbf{I}}_{1}$  . Let $\tau$  be the commutator of $b$  in $Der\left(g\right)$  , and ${h}_{1}=b{×}_{t}g$  . Let $h$  be the sum of $\tau$  and $\left\{ad\left(g\right):g\in g\right\}$  in $Der\left({h}_{1}\right)$  . Then $h=Der\left({h}_{1}\right)$  and $h$  is a complete Lie algebra.
• Proof. Lemma 3(ii) and Lemma 4(ii) imply that $h=Der\left({h}_{1}\right)$  .
$C\left({h}_{1}\right)=0$  implies that $C\left(h\right)=C\left(Der\left({h}_{1}\right)\right)=0$  Lemma 3(i) and lemma 4(i) imply that $Der\left(h\right)=ad\left(h\right)$  .
Lemma 4 can be proved in the same way as the proof of lemma 2.
We are going to prove lemma 3 in the similar way as the proof of lemma 1.
Proof of lemma 3:
• Proof. We prove only (i).
For any $\alpha \in \Delta \\left\{0\right\}$  , ${e}_{\alpha }\in {g}_{\alpha }$  , let  $\begin{array}{c}D{e}_{\alpha }={t}_{1}+{\sum }_{\beta \in \Delta \\left\{0\right\}}{x}_{\beta }\end{array}$ (25)
with ${t}_{1}\in \tau$  and ${x}_{\beta }\in {g}_{\beta }\left(\beta \in \Delta \\left\{0\right\}\right)$  .From  $\begin{array}{c}D\left[b,{e}_{\alpha }\right]=\left[Db,{e}_{\alpha }\right]+\left[b,D{e}_{\alpha }\right]\end{array}$ (26)
with $b\in b$  ,we obtain $\alpha \left(b\right)\left({t}_{1}+{\sum }_{\beta \in \Delta \\left\{0\right\}}{x}_{\beta }\right)=\left[Db,{e}_{\alpha }\right]+{\sum }_{\beta \in \Delta \\left\{0\right\}}\beta \left(b\right){x}_{\beta }.$  And notice that $\left[Db,{e}_{\alpha }\right]\in {g}_{\alpha }$  , so  $\begin{array}{c}\left[Db,{e}_{\alpha }\right]=0\forall \alpha \in \Delta \\left\{0\right\},b\in b\end{array}$ (27)
 $\begin{array}{c}\alpha \left(b\right){x}_{\beta }=\beta \left(b\right){x}_{\beta }\forall \beta \in \Delta \\left\{0\right\}\end{array}$ (28)
 $\begin{array}{c}\alpha \left(b\right){t}_{1}=0\forall b\in b\end{array}$ (29)
Now (28) implies ${x}_{\beta }=0\forall \beta \ne \alpha$  , (29) implies ${t}_{1}=0$  , so for any $\alpha \in \Delta \\left\{0\right\}$  , $D\left({g}_{\alpha }\right)\subseteq {g}_{\alpha }$  . Since ${g}_{0}\subseteq \left[g,g\right]$  , we have $D\left({g}_{0}\right)\subseteq {g}_{0}$  ,too. Therefore $D$  commutate with $ad\left(b\right)|g$  , and so there is a $t\in \tau$  such that $D|g=ad\left(t\right)|g$  .
Let ${D}_{1}=D-ad\left(t\right)$  , then ${D}_{1}\left(b\right)\subseteq \tau$  and ${D}_{1}|g=0$  .
For any $b\in b,x\in g,$  we have $\left[b,x\right]\in g$  , and so $0={D}_{1}\left[b,x\right]=\left[{D}_{1}b,x\right]+\left[b,{D}_{1}x\right]=\left[{D}_{1}b,x\right]$  Therefore ${D}_{1}b=0$  . Now let $t\in \tau$  , and ${D}_{1}t={t}^{\prime }+{x}^{\prime }$  with ${t}^{\prime }\in \tau$  , ${x}^{\prime }\in {\oplus }_{\alpha \in \Delta \\left\{0\right\}}{g}_{\alpha }$  . Then $\forall b\in b$  , $0={D}_{1}\left[t,b\right]=\left[{D}_{1}t,b\right]=\left[{x}^{\prime },b\right],$  so ${x}^{\prime }=0$  and ${D}_{1}t\in \tau$  . But $0={D}_{1}\left[t,x\right]=\left[{D}_{1}t,x\right]=\left({D}_{1}t\right)\left(x\right)\forall x\in g$  shows that ${D}_{1}t=0$  . We have now completed the proof of (i).
We mow apply theorem 2 to solvable complete Lie algebra.
Let $n$  be a nilpotent Lie algebra of type ${\mathbf{I}}_{1}$  , $b$  be a maximal torus on $n$  , and $\tau$  be the commutator of $b$  in $Der\left(n\right)$  . Let ${h}_{1},h$  be defined as above. Theorem 2 shows that $h=Der\left({h}_{1}\right)$  is a complete Lie algebra. It is also solvable.
Let $\mathcal{ℬ}$  be a solvable complete Lie algebra. Suppose that $\mathcal{ℬ}$  is decomposed into
 $\begin{array}{c}\mathcal{ℬ}=b{×}_{t}n=b{×}_{t}{\sum }_{\alpha \in \Delta }{n}_{\alpha }.\end{array}$ (*)
We define ${N}_{1}\left(\mathcal{ℬ}\right)$  to be the subalgebra generated by all ${n}_{\alpha }\left(\alpha \ne 0\right)$  , and ${B}_{1}\left(\mathcal{ℬ}\right)$  be the subalgebra $b{×}_{t}{N}_{1}\left(\mathcal{ℬ}\right)$  of $\mathcal{ℬ}$  . Of course ${N}_{1}\left(\mathcal{ℬ}\right)$  is of type ${\mathbf{I}}_{1}$  . ${N}_{1}\left(\mathcal{ℬ}\right)$  and ${B}_{1}\left(\mathcal{ℬ}\right)$  are ideals in $\mathcal{ℬ}$  . $C\left({B}_{1}\left(\mathcal{ℬ}\right)\right)=0$  .
It is easy to see that if $n$  is a nilpotent Lie algebra of type ${\mathbf{I}}_{1}$  with $b$  a maximal torus on it, we have that: ${N}_{1}\left(Der\left(b{×}_{t}n\right)\right)=n$  and ${B}_{1}\left(Der\left(b{×}_{t}n\right)\right)=b×n$  . Now let $\mathcal{ℬ}$  be be a solvable complete Lie algebra and be decomposed as in $\left(*\right)$  . We need to prove that $b$  is a maximal torus on ${N}_{1}$  and $\mathcal{ℬ}=Der\left({B}_{1}\left(\mathcal{ℬ}\right)\right)$  .
For this, we first show that $\mathcal{ℬ}$  is a subalgebra of $Der\left({B}_{1}\left(\mathcal{ℬ}\right)\right)$  , or the same thing:
$\forall x\in {n}_{0},ad\left(x\right)|{B}_{1}\left(\mathcal{ℬ}\right)\ne 0$  . If not, let $x$  be such a non-zero element with maximal index (let $n={n}^{\left(1\right)},$  and $\left[n,{n}^{\left(n-1\right)}\right]={n}^{\left(n\right)}$  ,then the elements in ${n}^{\left(n-1\right)}\{n}^{n}$  are said of index $n-1$  ). For any ${x}_{1}\in {n}_{0}$  , it is easy to see $ad\left(\left[x,{x}_{1}\right]\right)|{B}_{1}\left(\mathcal{ℬ}\right)=0$  and therefore $\left[x,{x}_{1}\right]=0\forall {x}_{1}\in {n}_{0}$  . Then $x\in C\left(\mathcal{ℬ}\right)$  which is a contradiction.
Now $\mathcal{ℬ}\subseteq Der\left({B}_{1}\left(\mathcal{ℬ}\right)\right)$  are two solvable complete Lie algebra. They are equal if and only if they have the same nilpotent radical. Let $\mathcal{N}$  be the nilpotent radical of $Der\left({B}_{1}\left(\mathcal{ℬ}\right)\right)$  . If $\mathcal{N}\ne n$  , there is a element $n$  of $N\n$  with maximal index. Then $\left[n,n\right]\subseteq n$  and induced a derivation on $\mathcal{ℬ}$  . Since $\mathcal{ℬ}$  is complete, there is a ${n}_{0}\in n$  such that $ad\left(n\right)|\mathcal{ℬ}=ad\left({n}_{0}\right)|\mathcal{ℬ}$  . And so $n={n}_{0}$  , which is a contradiction. We have now proved $\mathcal{ℬ}=Der\left({B}_{1}\left(\mathcal{ℬ}\right)\right)$  . Therefore there is a one to one correspondence between solvable complete Lie algebra and nilpotent Lie algebra of type ${\mathbf{I}}_{1}$  , and so every solvable Lie algebra can be obtained by the means stated above.
Now we have the following theorem:
Theorem 3. Let $n\left({n}_{1},{n}_{2}\right)$  be nilpotent Lie algebras of type ${\mathbf{I}}_{1}$  and $b\left({b}_{1},{b}_{2}\right)$  be the maximal torus (toruses) on $n\left({n}_{1},{n}_{2}$  respectively), then $Der\left(b{×}_{t}n\right)$  is a solvable complete Lie algebra. All solvable complete Lie algebras are obtained in such a way. Moreover, $Der\left({b}_{1}{×}_{t}{n}_{1}\right)\sim =Der\left({b}_{2}{×}_{t}{n}_{2}\right)$  if and only if ${n}_{1}\sim ={n}_{2}$  .
Therefore, to study solvable Lie algebras is to study nilpotent Lie algebras of type ${\mathbf{I}}_{1}$  .

3 Some relation between complete Lie algebra and Heisenberg algebra

3.1 General result

In this section, we are going to study the full graphs of Lie algebras of certain kinds. Let $g$  be a Lie algebra. Suppose $\mathcal{S}$  is a complete Lie algebra and there is a Lie homomorphism $L$  from $\mathcal{S}$  to $Der\left(g\right)$  , which make $g$  a $\mathcal{S}$  -module. We can define a Lie algebra $h=\mathcal{S}{×}_{t}g$  .
We define ${Z}_{s}\left(g\right)$  to be:
 $\begin{array}{c}{Z}_{s}\left(g\right)=\left\{g\in C\left(g\right):s\left(g\right)=0\forall s\in \mathcal{S}\right\}\end{array}$ (30)
A simple calculate shows:
Lemma 5. The center of $h=\mathcal{S}{×}_{t}g$  is $\left\{\left(0,g\right):g\in {Z}_{s}\left(g\right)\right\}$  .
Let ${F}_{s}\left(g\right)$  be:
 $\begin{array}{c}{F}_{s}\left(g\right)=\left\{D\in Der\left(g\right):\left[D,{D}_{1}\right]\in ad\left(g\right)\forall {D}_{1}\in L\left(\mathcal{S}\right)\right\}\end{array}$ (31)
It is easy to see that ${F}_{s}\left(g\right)$  is a Lie subalgebra of $Der\left(g\right)$  and $ad\left(g\right)$  is a ideal of ${F}_{s}\left(g\right)$  . We can regard $Der\left(g\right)$  as a $\mathcal{S}$  -module in the nature way:
 $\begin{array}{c}s\left(D\right)=\left[L\left(s\right),D\right]\forall s\in \mathcal{S},D\in Der\left(g\right)\end{array}$ (32)
We can define a Lie algebra $\stackrel{~}{h}=\mathcal{S}{×}_{t}{F}_{s}\left(g\right)$  . It is easy to see that the Lie bracket in $\stackrel{~}{h}$  is:
 $\begin{array}{c}\left[\left({s}_{1},{D}_{1}\right),\left({s}_{2},{D}_{2}\right)\right]=\left(\left[{s}_{1},{s}_{2}\right],\left[L\left({s}_{1}\right),{D}_{2}\right]+\left[{D}_{1},L\left({s}_{2}\right)\right]+\left[{D}_{1},{D}_{2}\right]\right)\end{array}$ (33)
Now assume $C\left(g\right)=0$  , then $g$  is isomorphic to $ad\left(g\right)$  . We write $I$  for the inverse morphism of $ad$  which is a Lie homomorphism from $ad\left(g\right)$  to $g$  :
 $\begin{array}{c}I\left(ad\left(x\right)\right)=x\forall x\in g\end{array}$ (34)
Lemma 6. If $C\left(g\right)=0$  , then $Der\left(h\right)\supseteq \stackrel{~}{h}$  .
Furthermore, if one of the following conditions (a)(b)(c) is satisfied, then $Der\left(h\right)=\stackrel{~}{h}$  :
(a). $Der\left(g\right)\supseteq \mathcal{S}\supseteq ad\left(g\right)$  (b). $\left[g,g\right]=g$  (c). $g=b×n$  is a solvable Lie algebra with $\left(n,b\right)$  of type ${\mathbf{I}}_{1}$  , and $\mathcal{S}$  commutates with $b$  :
$\mathcal{S}⊑\left\{s\in Der\left(g\right):\forall b\in b,s\left(b\right)=0\right\}$
• Proof. For any $\left(s,D\right)\in \stackrel{~}{h}$  , define a derivation on $h$  by  $\begin{array}{c}{D}_{\left(s,D\right)}\left({s}_{1},g\right)=\left(\left[s,{s}_{1}\right],s\left(g\right)+D\left(g\right)+I\left(\left[D,L\left({s}_{1}\right)\right]\right)\end{array}$ (35)
A direct calculate shows that:
• (i). ${D}_{\left(s,D\right)}$  is a derivation.
• (ii). $\left[{D}_{\left({s}_{1},{D}_{1}\right)},{D}_{\left({s}_{2},{D}_{2}\right)}\right]={D}_{\left[\left({s}_{1},{D}_{1}\right)\left({s}_{2},{D}_{2}\right)\right]}$  .
• (iii). ${D}_{\left(s,D\right)}=0$  if and only if $s=0$  and $D=0$  .
So $\stackrel{~}{h}$  is a Lie subalgebra of $Der\left(h\right)$  .
Now we assume one of the conditions (a),(b) and (c) is satisfied. Let ${D}_{1}\in Der\left(h\right)$  . Write ${D}_{1}=\left({T}_{1}^{\prime },{T}_{2}\right)$  . Then ${T}_{1}^{\prime }|\mathcal{S}$  is a derivation on $\mathcal{S}$  . since $\mathcal{S}$  is complete, we may assume that ${T}_{1}^{\prime }|\mathcal{S}=ad\left(s\right)|\mathcal{S}$  with some $s\in \mathcal{S}$  . Let $D={D}_{1}-{D}_{\left(s,0\right)}$  , then $D=\left({T}_{1},{T}_{2}\right)$  with ${T}_{1}|\mathcal{S}=0$  . We are going to show that ${T}_{1}=0$  . By $D\left(\left[\left(0,{g}_{1}\right)\left(s,{g}_{2}\right)\right]\right)=\left[D\left(0,{g}_{1}\right),\left(s,{g}_{2}\right)\right]+\left[\left(0,{g}_{1}\right),D\left(s,{g}_{2}\right)\right]$  we obtain  $\begin{array}{c}-{T}_{1}\left(s\left({g}_{1}\right)\right)+{T}_{1}\left(\left[{g}_{1},{g}_{2}\right]\right)=\left[{T}_{1}\left({g}_{1}\right),s\right]\forall s\in \mathcal{S},{g}_{1},{g}_{2}\in g\end{array}$ (36)
Formula (36) shows that  $\begin{array}{c}{T}_{1}|\left[g,g\right]=0\end{array}$ (37)
and that  $\begin{array}{c}-{T}_{1}\left(s\left(g\right)\right)=\left[{T}_{1}\left(g\right),s\right]\end{array}$ (38)
The condition (b) shows that ${T}_{1}=0$  . Now assume (a). Formula (37) and (38) show that  $\begin{array}{c}\left[{T}_{1}\left(g\right),s\right]=0\forall s\in ad\left(g\right),g\in g,\end{array}$ (39)
since we have assumed $ad\left(g\right)\subseteq \mathcal{S}$  . Notice the fact that $\left[D,ad\left(g\right)\right]=ad\left(D\left(g\right)\right)\forall D\in Der\left(g\right),g\in g.$  We have ${T}_{1}\left(g\right)\left({g}_{1}\right)=0\forall g,{g}_{1}\in g.$  So ${T}_{1}\left(g\right)=0$  for any $g\in g$  .
Now we assume condition (c). Condition (c) implies that $g=b+\left[g,g\right]$  .
Formula (38) shows that $\left[{T}_{1}\left(b\right),s\right]=0\forall b\in b,s\in \mathcal{S}$  Since $C\left(\mathcal{S}\right)=0$  , we gain ${T}_{1}\left(b\right)=0$  for any $b$  in $b$  . Since ${T}_{1}|\left[g,g\right]=0$  ,the proof of ${T}_{1}=0$  is completed.
Now we have $D={T}_{2}$  , and  $\begin{array}{c}{T}_{2}\left(\left[\left({s}_{1},0\right),\left({s}_{2},0\right)\right]\right)=\left[\left({s}_{1},0\right),{T}_{2}\left({s}_{2},0\right)\right]+\left[{T}_{2}\left({s}_{1},0\right),\left({s}_{2},0\right)\right]\end{array}$ (40.1)
 $\begin{array}{c}{T}_{2}\left(\left[\left(s,0\right),\left(0,x\right)\right]\right)=\left[\left(s,0\right),{T}_{2}\left(0,x\right)\right]+\left[{T}_{2}\left(s,0\right),\left(0,x\right)\right]\end{array}$ (40.2)
 $\begin{array}{c}{T}_{2}\left(\left[\left(0,{x}_{1}\right),\left(0,{x}_{2}\right)\right]\right)=\left[\left(0,{x}_{1}\right),{T}_{2}\left(0,{x}_{2}\right)\right]+\left[{T}_{2}\left(0,{x}_{1}\right),\left(0,{x}_{2}\right)\right]\end{array}$ (40.3)
Formula (40.3) shows that ${T}_{2}{|}_{\left(0,g\right)}\in Der\left(g\right)$  and we may assume  $\begin{array}{c}{T}_{2}{|}_{\left(0,g\right)}=D.\end{array}$ (41)
Then (40.2) shows that  $\begin{array}{c}ad\left({T}_{2}\left(s\right)\right)=\left[D,L\left(s\right)\right]\forall s\in \mathcal{S}\end{array}$ (42.1)
or that  $\begin{array}{c}I\left(\left[D,L\left(s\right)\right]\right)={T}_{2}\left(s\right)\forall s\in \mathcal{S}\end{array}$ (42.2)
And so $D\in {F}_{s}\left(g\right)$  . Now formula (41), (42.1) show that $D={D}_{\left(0,D\right)}\in \stackrel{~}{h}$  . So ${D}_{1}\in \stackrel{~}{h}$  , and $Der\left(h\right)=\stackrel{~}{h}$  , which completes the proof of lemma 6.
Lemma 5 and lemma 6 imply:
Theorem 4. Let $g$  be a Lie algebra with trivial center, and assume that $Der\left(g\right)$  is a complete lie algebra. Let $F\left(g\right)$  be the subalgebra of $Der\left(g\right)$  such that $F\left(g\right)=\left\{D\in Der\left(g\right):\left[D,{D}_{1}\right]\in ad\left(g\right)\forall {D}_{1}\in Der\left(g\right)\right\}$  Then the derivation algebra of the full graph $C\left(g\right)=Der\left(g{×}_{t}g\right)$  of $g$  is $Der\left(g\right){×}_{t}F\left(g\right)$  , and the center of $C\left(g\right)$  is 0. $C\left(g\right)$  is a complete Lie algebra if and only if $F\left(g\right)=g$  .
We define ${C}^{1}\left(g\right)=C\left(g\right)$  and ${C}^{n}\left(g\right)=C\left({C}^{n-1}\left(g\right)\right)$  by induction.
Corollary 1. Let $g$  be a complete Lie algebra, then it full graph is a complete Lie algebra.
In this case, $C\left(g\right)\sim =g\oplus g$  Corollary 2. Assume that $g$  is a non-complete Lie algebra with trivial center, and assume that $Der\left(g\right)$  is a complete Lie algebra such that $\left[Der\left(g\right),Der\left(g\right)\right]\subseteq ad\left(g\right)$  , then we have the following conclusions:
(i). $Der\left({C}^{n}\left(g\right)\right)={C}^{n}\left(Der\left(g\right)\right)$  is complete but ${C}^{n}\left(g\right)$  is not.
(ii). The center of ${C}^{n}\left(g\right)$  is trivial.
(iii). $\left[Der\left({C}^{n}\left(g\right)\right),Der\left({C}^{n}\left(g\right)\right)\right]\subseteq ad\left({C}^{n}\left(g\right)\right)$  .
(iv). Dimension formula: $dim\left(Der\left({C}^{n}\left(g\right)\right)\right)-dim\left({C}^{n}\left(g\right)\right)=dim\left(Der\left(g\right)\right)-dim\left(g\right)$  .
It is easy to prove corollary 2 by applying theorem 4 and by induction.

3.2 Apply to Heisenberg algebras

In this section, let $g$  be a Heisenberg algebra: $\left[g,g\right]=C\left(g\right)$  and $dim\left(C\left(g\right)\right)=1$  .
Let $0\ne c\in C\left(g\right)$  . Then there is an anti-symmetric quadratic form $\psi$  on $g$  of rank $dim\left(g\right)-1$  such that:
 $\begin{array}{c}\left[x,y\right]=\psi \left(x,y\right)c\forall x,y\in g\end{array}$ (43)
Then we have $dim\left(g\right)-1=2n$  . We may choose a basis $\left\{{x}_{1},...,{x}_{n};{y}_{1},...,{y}_{n};c\right\}$  such that for any $1\le i,j\le n,x\in g$  ,
 $\begin{array}{c}\psi \left({x}_{i},{x}_{j}\right)=\psi \left({y}_{i},{y}_{j}\right)==\psi \left(c,x\right)=0\end{array}$ (44.1)
 $\begin{array}{c}\psi \left({x}_{i},{y}_{j}\right)={\delta }_{ij}\end{array}$ (44.2)
or
 $\begin{array}{c}\left[{x}_{i},{x}_{j}\right]=\left[{y}_{i},{y}_{j}\right]=\left[c,x\right]=0,\left[{x}_{i},{y}_{j}\right]={\delta }_{ij}c\end{array}$ (44.3)
Let $\Psi$  be the matrix representing $\psi$  under the basis $\left\{{x}_{1},...,{x}_{n};{y}_{1},...,{y}_{n};c\right\}$  .
Then $\Psi =\left(\begin{array}{ccc}0& {I}_{n}& 0\\ -{I}_{n}& 0& 0\\ 0& 0& 0\end{array}\right)$  In [4,Meng], the following propositions are stated:
Proposition 2. $Der\left(g\right)$  is simple complete Lie algebra, and has the following decomposition:
$Der\left(g\right)=\mathcal{S}{×}_{t}\left(b{×}_{t}n\right),$  Here, $\mathcal{S}\sim =sp\left(2n,\mathbb{C}\right),n=ad\left(g\right)\sim =g/\mathbb{C}c$  is a irreducible $\mathcal{S}$  -module and $dim\left(b\right)=1$  .
Proposition 3. The full graph $C\left(g\right)$  of $g$  is not a complete Lie algebra, but its center is trivial. Proposition 4. The derivation algebra $Der\left(C\left(g\right)\right)$  of $C\left(g\right)$  is a simple Lie algebra. And $Der\left(C\left(g\right)\right)=ad\left(C\left(g\right)\right)+\mathbb{C}{D}_{0}$  with ${D}_{0}$  given by ${D}_{0}=\left\{\begin{array}{cc}0,& ifx\in Der\left(g\right)\\ 2x,& ifx\in \mathbb{C}c\\ 2x-ad\left(x\right),& ifx\in V=<{x}_{1},...,{x}_{n},{y}_{1},...,{y}_{n}>.\end{array}$  Proposition 4 shows that $\left[Der\left(C\left(g\right)\right),Der\left(C\left(g\right)\right)\right]\subseteq ad\left(C\left(g\right)\right)$  Applying corollary 2 of theorem 4, we gain:
Theorem 5. Let $g$  be a Heisenberg algebra, then for any positive integer $n$  , (i). The center of $C\left(g\right)$  is trivial.
(ii). $Der\left({C}^{n}\left(g\right)\right)={C}^{n-1}\left(Der\left(C\left(g\right)\right)\right)$  is complete but ${C}^{n}\left(g\right)$  is not.
(iii). $\left[Der\left({C}^{n}\left(g\right)\right),Der\left({C}^{n}\left(g\right)\right)\right]\subseteq ad\left({C}^{n}\left(g\right)\right)$  .
(iv). Dimension formula: $dim\left(Der\left({C}^{n}\left(g\right)\right)\right)-dim\left({C}^{n}\left(g\right)\right)=1$  .
References

1. Meng D.J., Some results on Complete Lie algebras, Comm.Algebrra, 22(1994): 5457-5508.
2. Meng D.J., Complete Lie Algebras and Heisenberg Algebras, Comm.Algebra, 22(1994): 5509-5524.
3. Meng D.J, Zhu L.s., Solvable complete Lie Algebra I, Comm. Algebra 24(1996): 4181-4197.
4. Meng D.J, Zhu L.S., Complete Lie Algebra (in Chinese),Beijing: Science Press, 2001.
5. Sato T. The derivations of Lie algebra, Tohoku, Math J., 23(1971): 21-36.
6. Zhu L.S., Meng D.J., One kind of complete Lie algebras, Algebras, Groups and Geometries, 17(1999): 57-72.