This is the talk that I gave at Wolfram Tech Conference 2025. It presents an in-depth exploration of the Zassenhaus, Baker–Campbell–Hausdorff and Magnus formulas, which decompose exponentials of operator sums in noncommutative settings. We detail their derivations, applications and recent algorithmic formulations for high-degree expansions. Special attention is given to symbolic implementations in Wolfram Language, showcasing efficient computation using custom-built functions and algebraic structures. Applications to operator theory, Lie algebras and symbolic matrix analysis will be discussed.I believe new tools for performing non-commutative algebraic computations can open up new horizons in quantum computing, from simulating error correction to studying open quantum systems and noise. CITATION: Mads Bahrami, Expanding Exponentials in Noncommutative Algebras: The Zassenhaus and Baker–Campbell–Hausdorff Formulas, Wolfram Technology Conference 2025. https://www.wolfram.com/events/technology-conference/2025/Baker-Campbell-Hausdorff (BCH) series: =+Zassenhaus series: =Magnus series: =A(t)Y(t)⇒Y(t)=(t)
1
2
m
…
m
∑
j=1
j
∞
∑
n=2
n
m
∑
j=1
m
m
∏
j=1
j
∏
k=2
k
Y(t)
t
∞
∑
n=1
Ω
n
e
Initialization
Initialization
Adjoint action
Adjoint action
The adjoint action of on another operator is define as the commutation . A direct interesting result is the exponential adjoint action B=B with [B]=[A,B]
A
B
[A,B]
A
-A
ad
A
ad
A
In[]:=
ClearAll[A,B]adjointAction[A,B,1]
Out[]=
A**B-B**A
The -th iterated adjoint of is defined as the -fold nested commutator of with , denoted by [B]=[A,[A,…[A,B]…]]:
j
A
n
A
B
(j)
ad
A
In[]:=
adjointAction[A,B,2]==Commutator[A,Commutator[A,B]]
Out[]=
True
In[]:=
adjointAction[A,B,3]==Commutator[A,Commutator[A,Commutator[A,B]]]
Out[]=
True
We shall study some examples from quantum optics. To do so, let us first define a function to reduce noncommutative polynomials of annihilation and creation operators.
Examples
Examples
Bosonic commutation relation
,=1
a
†
a
Consider a. Compute 10 terms in the exponential adjoint:
λa
†
a
-λa
†
a
In[]:=
NCPolynomialReductionBosonicSumNCPolynomialReductionBosonicadjointActionλ**a,a,j,{λ}j!,{j,0,10},{λ}
†
a
Out[]=
a1-λ+-+-+-+-+
2
λ
2
3
λ
6
4
λ
24
5
λ
120
6
λ
720
7
λ
5040
8
λ
40320
9
λ
362880
10
λ
3628800
Verify the summation of 10 terms is the same as Taylor expansion of
-λ
In[]:=
NormalSeries,{λ,0,10}==CoefficientRules%,a,[[1,-1]]
-λ
†
a
Out[]=
True
So we managed to verify a=a
λa
†
a
-λa
†
a
-λ
Consider -ζa with . Compute 10 terms in the exponential adjoint:
1
2
*
ζ
2
a
2
†
a
--ζ
1
2
*
ζ
2
a
2
†
a
ζ=r
ϕ
In[]:=
NCPolynomialReductionBosonicSumNCPolynomialReductionBosonicadjointAction,a,jj!,{,ϕ},{j,0,10},{,ϕ}
a**a-**
-ϕ
ϕ
†
a
†
a
2
Out[]=
a1++++++++++
2
2
4
24
6
720
8
40320
10
3628800
ϕ
1
6
ϕ
3
1
120
ϕ
5
ϕ
7
5040
ϕ
9
362880
†
a
Verify -ζa=ua+v with and :
1
2
*
ζ
2
a
2
†
a
--ζ
1
2
*
ζ
2
a
2
†
a
†
a
u=cosh(r)
v=-sinh(r)
ϕ
In[]:=
NormalSeriesCosh[],Sinh[],{,0,10}==CoefficientRules%,a,[[All,-1]]
ϕ
†
a
Out[]=
True
Baker-Campbell-Hausdorff (BCH) series
Baker-Campbell-Hausdorff (BCH) series
The BCH series expresses the product of two or more exponentials of non-commuting operators as a single exponential: =+
1
2
m
…
m
∑
j=1
j
∞
∑
n=2
n
For two operators, , one gets the Dynkin’s formula: =+>0,…,+>0,(+)=m-1…!!…!!(X+Y).
X+Y
m
m
∑
n=1
n-1
(-1)
n
∑
p
1
q
1
p
n
q
n
n
∑
j=1
p
j
q
j
p
1
ad
X
q
1
ad
Y
p
n
ad
X
q
n
ad
Y
p
1
q
1
p
n
q
n
Show a few terms of =
A
B
A+B+
∑
n=2
n
In[]:=
ClearAll[A,B];TableForm[Table[{n,TraditionalForm@BakerCampbellHausdorffTerms[{A,B},n,"CommutatorForm"->True]},{n,2,4}],TableHeadings->{None,{"n",""}},TableAlignments->Left]
n
Out[]//TableForm=
n | n |
2 | [A,B] 2 |
3 | - [A,[A,B]] 12 [A,[B,A]] 6 [B,[A,B]] 6 [B,[B,A]] 12 |
4 | [A,[A,[A,B]]] 16 [A,[A,[B,A]]] 16 [A,[B,[A,B]]] 24 [A,[B,[B,A]]] 16 [B,[A,[A,B]]] 16 [B,[A,[B,A]]] 12 [B,[B,[A,B]]] 16 [B,[B,[B,A]]] 16 |
Most quantum operators belong to small Lie algebras (Heisenberg–Weyl, SU(2), SU(1,1)), where the commutator chains truncate or close, allowing the BCH formula to yield clean, closed-form exponential identities.
Normal ordering and Weyl relations
Normal ordering and Weyl relations
We will explore some examples involving photonic operators, where the single-mode photon field satisfies the Bosonic commutation relation .
,=1
a
†
a
Annihilation and creation operator:
Annihilation and creation operator:
Consider the expression: = Compute the order-2 term in BCH series:
μ
†
a
ν
a
μ+ν+
†
a
a
∑
n
n
In[]:=
BakerCampbellHausdorffTermsμa,ν,2
†
a
Out[]=
1
2
†
a
1
2
†
a
Reduce/simplify it:
In[]:=
NCPolynomialReductionBosonic[%,{μ,ν}]
Out[]=
μν
2
Note the result is a c-number only.
Compute the order-3 term of BCH series:
In[]:=
BakerCampbellHausdorffTermsμa,ν,3
†
a
Out[]=
1
12
†
a
1
6
†
a
1
12
†
a
†
a
1
12
†
a
1
6
†
a
†
a
1
12
†
a
†
a
Reduce it:
In[]:=
NCPolynomialReductionBosonic[%,{μ,ν}]
Out[]=
0
Likewise, the other terms disappear as well. Thus one finds: =
μ
†
a
ν
a
μ+ν+μν/2
†
a
a
Canonical position and momentum
Canonical position and momentum
Consider canonical position and momentum operators: =(+) and =(-)
q
1
2
a
†
a
p
1
2
a
†
a
Define canonical position and momentum operators:
In[]:=
q=(a+))
†
a
2
;p=(a-†
a
2
;Verify
,=
q
p
In[]:=
NCPolynomialReductionBosonic@Commutator[q,p]
Out[]=
Now let’s try to compute .
α
q
β
p
Compute the second term in the BCH series:
In[]:=
NCPolynomialReductionBosonic[BakerCampbellHausdorffTerms[{αq,βp},2],{α,β}]
Out[]=
-αβ
1
2
Show that the higher-order terms are zero:
In[]:=
Table[NCPolynomialReductionBosonic[BakerCampbellHausdorffTerms[{αq,βp},n],{α,β}],{n,3,6}]
Out[]=
{0,0,0,0}
Thus it means =
α
q
β
p
-αβ/2
α+β
q
p)
Displacement operator
Displacement operator
Displacement operator is defined as . Show that
(α)=
α-a
†
a
*
α
(α)(β)=(α+β)
α-β
*
β
*
α
2
In[]:=
NCPolynomialReductionBosonicBakerCampbellHausdorffTermsα-Conjugate[α]a,β-Conjugate[β]a,2,{α,β}
†
a
†
a
Out[]=
-βConjugate[α]+αConjugate[β]
1
2
1
2
Note that the result is a c-number.
Show that the higher-order terms are zero:
SU(2) spin rotations
SU(2) spin rotations
Compute the second term of the BCH series:
Numerical example
Numerical example
Note that in the previous examples, all operators were symbolic. One can use the same thing with numerical operators, too.
Zassenhaus series
Zassenhaus series
Displacement operator:
Displacement operator:
Compute the reduced form of the order-2 term in Zassenhaus series:
Note the result is a c-number, which implies higher order terms are zero:
Compute the second term in Zassenhaus series:
Compute a few terms (3rd and more) and show they are zero:
SU(2) Trotter error
SU(2) Trotter error
SU(2) Gauss/Wei–Norman decomposition
SU(2) Gauss/Wei–Norman decomposition
Putting them together, one finds:
Define noncommutative variables:
Putting above results together and simplify it, one finds:
So the same for the final result we obtained above:
Equate coefficients:
Solve them:
Plug the result back into the solution:
Magnus series
Magnus series
Show three terms of Magnus series (without time integration):
Examples
Examples
For each term, find the coefficient for Pauli matrices (keep in mind that we will have only linear terms):
Now let's repeat the same process, but this time assigning explicit values to Pauli matrics and doing the commutation/dot calculations:
As expected, the coefficients in front of Pauli matrices are the same the ones we obtained from non - commutative algebraic calculations.
Calculate the unitary operator:
CITE THIS NOTEBOOK
CITE THIS NOTEBOOK
Exponentials in Noncommutative Algebras Zassenhaus, BCH and Magnus
by Mohammad Bahrami
Wolfram Community, STAFF PICKS, December 2, 2025
https://community.wolfram.com/groups/-/m/t/3584499
by Mohammad Bahrami
Wolfram Community, STAFF PICKS, December 2, 2025
https://community.wolfram.com/groups/-/m/t/3584499