Gaussian Integrals

For a physicist Gaussian integral is a common integral which has a following form:
∞
1 2 +𝛽𝑥 𝛽 2 2𝜋
∫ 𝑒 −2𝛼𝑥 𝑑𝑥 = exp [ ]√
2𝛼 𝛼
−∞

In the integrand we have a quadratic form of 2x2 matrices present in power, which can be
generalized for any positively definite symmetric 𝑛 × 𝑛 matrix 𝐴. We know that such matrix
can be diagonalized with orthogonal transformations :

𝛼1
𝑇
𝐷 =𝑅 𝐴𝑅 =( ⋱ )
𝛼𝑛

Then integral can be written in a generalized form:

∞ ∞ ∞ ∞
1 1
− ⟨𝑥 𝐴 𝑥⟩
𝐼 = ∫ … ∫ 𝑑𝑥1 … 𝑑𝑥 𝑛 𝑒 2 = ∫ … ∫ 𝑑𝑦1 … 𝑑𝑦 𝑛 𝑒 −2⟨𝑦 𝐷 𝑦⟩
−∞ −∞ −∞ −∞

Where, 𝑦 = 𝑅 𝑇 𝑥 , therefore integral can be taken step by step:

∞ ∞
1
− 𝛼𝑦 (𝑦𝑖 )
2 2𝜋 2𝜋 (2𝜋)𝑛
𝐼 = ∫ 𝑑𝑦 … 𝑑𝑦 1 𝑛
∫ 𝑑𝑦 𝑒 𝑖 2 =√ ×…× √ = √
𝛼1 𝛼𝑛 ∆𝐴
−∞ −∞

This can be further generalized adding a linear term in exponent:

∞ ∞
1
⟨𝑥 𝐴 𝑥⟩+⟨𝐵 𝑥⟩
𝐼𝐿 = ∫ … ∫ 𝑑𝑥1 … 𝑑𝑥 𝑛 𝑒 −2
−∞ −∞

Now if we translate a variable in following way: 𝑦 = 𝑥 − 𝐴−1 𝐵, then:

∞ ∞
1
⟨𝑦 𝐴 𝑦⟩+⟨𝐵 𝐴−1 𝐵⟩
1 −1 𝐵⟩ (2𝜋)𝑛
𝐼𝐿 = ∫ … ∫ 𝑑𝑦 1 … 𝑑𝑦 𝑛 𝑒 −2 = 𝑒 −2⟨𝐵 𝐴 √
∆𝐴
−∞ −∞

In the following section, we will verify the convergence of Fresnel integrals:

∞ ∞
1
𝐼𝐹 = ∫ cos 𝑥 2 𝑑𝑥 = ∫ cos 𝑦 𝑑𝑦
2√𝑦
0 0

It’s clear that cosine is a bounded function, while lim √𝑦 = ∞ , therefore, if we split up
𝑦→∞
integral into Riemann’s sum, we will have a converging sum.
Now we will try to find connection between Gaussian function and delta function, for this
we will consider the following integral:
∞ 𝑥0 +𝜖
1 𝑥2 1 𝑥2
−
lim ∫ 𝑒 2𝛼 ⋅ 𝑓(𝑥)𝑑𝑥 = 𝑓(𝑥0 ) lim ∫ 𝑒 −2𝛼 𝑑𝑥 = 𝑓(𝑥0 )
𝛼→0 2√𝜋𝛼 𝛼→0 2√𝜋𝛼
−∞ 𝑥0 −𝜖

Therefore, we can treat functional as a Dirac delta, notice that if we replace 𝛼 → 𝛼𝑖 we get
the Lorentzian version of delta definition. In the next paragraph we will try to study time
evolution transition amplitudes, for that, we ought to define time evolution operator:
𝑖
̂
𝑈𝑡 = 𝑒 −ℏ𝑡𝐻
2
̂ = 𝑝̂ + 𝑉(𝑞̂) is a Hamiltonian operator. Here, aim is to analyze the matrix
where, 𝐻 2𝑚
elements of the evolution operator between the eigenstates of the coordinate operator
⟨𝑦|𝑈𝑡 |𝑥⟩
Which, for a free particle in potential-free scenario becomes:
̂2
𝑖 𝑝 ̂2
𝑖 𝑝
− 𝑡 − 𝑡
⟨𝑦|𝑒 ℏ 2𝑚 |𝑥⟩ = ⟨𝑦|𝑒 ℏ 2𝑚 |𝑝⟩ ⟨𝑝|𝑥⟩

Thus,
̂2
𝑖 𝑝 𝑑𝑝 𝑖𝑡𝑝2
⟨𝑦|𝑒 −ℏ𝑡2𝑚 |𝑝⟩ ⟨𝑝|𝑥⟩ = ∫ exp (− ) exp(−𝑖𝑝(𝑦 − 𝑥)) =
2𝜋 2𝑚

𝑚 1 𝑚(𝑦−𝑥)2
=√ 𝑒 ℏ 2𝑡
2𝜋𝑖ℏ𝑡
To study evolution transition amplitude for linear potential, next step is to explore the
following operator identity:
2
𝑄̂ = 𝑒 −𝑏𝑞̂ 𝑒 𝑎𝑝̂ +𝑏𝑞̂
Let’s apply BCH formula (note, that following commutator terms are 0) :
1 1
[𝐴,𝐵]+ ([[𝐴,𝐵],𝐵]−[[𝐴,𝐵],𝐴])
𝑒 𝐴 𝑒 𝐵 = 𝑒 𝐴+𝐵+2 12

Since, [−𝑞̂, 𝑞̂] = 0, [𝑝̂ , 𝑝̂ ] = 0 and [−𝑏𝑞̂, 𝑎𝑝̂ 2 ] = −2𝑖ℏ𝑎𝑏𝑝̂

2 +1[−𝑏𝑞 1
̂,𝑎𝑝̂2 +𝑏𝑞̂]+ [−2𝑖ℏ𝑎𝑏𝑝̂,−𝑏𝑞̂] 2 −𝑖ℏ𝑎𝑏𝑝̂−1ℏ2 𝑎𝑏 2
𝑄̂ = 𝑒 𝑎𝑝̂ 2 12 = 𝑒 𝑎𝑝̂ 3

Thus, amplitude expression is of following form:

̂2 ̂2 𝑖
𝑖𝑡 𝑝
− ( +𝑚𝑔𝑞̂) 𝑖𝑡 𝑖 2 3 −
𝑖𝑡 𝑝
+ 𝑔𝑡 2 𝑝̂
⟨𝑦|𝑒 ℏ 2𝑚 |𝑥⟩ = exp (− 𝑚𝑔𝑦 − 𝑚𝑔 𝑡 ) ⟨𝑦|𝑒 ℏ 2𝑚 2ℏ |𝑥⟩
ℏ 6ℏ
We will simply use same unit operator to obtain the final result:
𝑖𝑡 𝑚 1 𝑚(𝑦−𝑥)2 𝑖 1 2
⟨𝑦|𝑒 − ℏ 𝐻1 |𝑥⟩ = √ 𝑒 2ℏ 𝑡 ⋅ 𝑒 −2ℏ 𝑚𝑔𝑡(𝑥+𝑦+12𝑔𝑡 )
2𝜋𝑖ℏ𝑡
If we take 𝑔 = 0 we simply reduce to initial case of free particle therefore amplitudes
coincide as well.
Finally, we have to move on the oscillator problem, for that we have to use relations
between the given operators
𝑖 2 𝑖 2 𝑖
𝑂̂+ = − 𝑝̂ , 𝑂̂− = − 𝑞̂ 𝑂̂ = − (𝑝̂ 𝑞̂ + 𝑞̂𝑝̂ )
2ℏ 2ℏ 4ℏ
and the matrices

0 1 0 0 1 1 0
𝑇+ = ( ), 𝑇− = ( ), 𝑇= ( )
0 0 1 0 2 0 −1
We can check that:

[𝑂̂, 𝑂̂± ] = ±𝑂± [𝑂̂− , 𝑂̂+ ] = −2𝑂̂

and,
[𝑇, 𝑇± ] = ±𝑇± [𝑇+ , 𝑇− ] = 2𝑇

Since they make a sl(2, ℝ) algebra, for any matrix 𝐴 present in that algebra
𝐴 ⋅ 𝐴 = ⟨𝐴 𝐴⟩ 𝐼
and
1
⟨𝐴 𝐴⟩ = Tr(𝐴 𝐴) = −𝑎𝑏 < 0
2
indicating elements are spacelike, therefore exponential representation would be
sin 𝛼
𝑒 𝐴 = cos 𝛼 𝐼 + 𝐴, 𝛼 = √−⟨𝐴 𝐴⟩ = √𝑎𝑏
𝛼
Or

1 𝑏
𝑒 𝑎𝑇+−𝑏𝑇− = cos 𝛼 𝐼 + sin 𝛼 ( 𝑇+ − 𝛽𝑇− ) , where 𝛽 = √
𝛽 𝑎

Now, let’s observe the following identity with keeping in mind that ⟨𝑇± 𝑇± ⟩ = 0:

ℒ̂ = 𝑒 −𝑥𝑇− 𝑒 𝑦𝑇+ 𝑒 −𝑥𝑇− = (𝐼 − 𝑥𝑇− )(𝐼 + 𝑦𝑇+ )(𝐼 − 𝑥𝑇− ) =

= (𝐼 − 𝑥𝑇− )(𝐼 − 𝑥𝑇− + 𝑦𝑇+ − 𝑦𝑇+ 𝑥𝑇− ) =
 1 0 1 + 𝑥𝑦 𝑦 1 − 𝑥𝑦 𝑦
=( )( )=( )
−𝑥 1 −𝑥 1 −2𝑥 + 𝑥 2 𝑦 1 − 𝑥𝑦
Ultimately resulting in null trace matrix. Therefore, it can be broken apart in terms of
𝑇+ , 𝑇− and Identity matrices:

ℒ̂ = (1 − 𝑥𝑦)𝐼 + 𝑦𝑇+ − 𝑥(2 − 𝑥𝑦)𝑇−

𝛼 sin 𝛼
If 𝑥 = 𝛽 tan ( 2 ) , 𝑦 = ,then we have an equivalence:
𝛽

sin 𝛼
𝑒 −𝑥𝑇− 𝑒 𝑦𝑇+ 𝑒 −𝑥𝑇− = (1 − 𝑥𝑦)𝐼 + 𝑇 − 𝑥(2 − 𝑥𝑦)𝑇− =
𝛽 +
𝛼 sin 𝛼 sin 𝛼 𝛼 𝛼 sin 𝛼
= (1 − 𝛽 tan )𝐼 + 𝑇+ − 𝛽 tan ( ) (2 − 𝛽 tan ) 𝑇− =
2 𝛽 𝛽 2 2 𝛽
𝛼 sin 𝛼 𝛼
= (1 − 2 sin2 ) 𝐼 + 𝑇+ − 𝛽 tan ( ) (2 − 2 sin2 𝛼) 𝑇− =
2 𝛽 2
sin 𝛼
= cos 𝛼 𝐼 + 𝑇 − 𝛽 sin 𝛼 𝑇− = 𝑒 𝑎𝑇+−𝑏𝑇−
𝛽 +
This means we can break up operator in 3 parts:
̂2
𝑖𝑡 𝑝 𝑖𝑚𝜔 𝜔𝑡 𝑖 𝑖𝑚𝜔 𝜔𝑡
− ( +𝑚𝜔 2 𝑞̂ 2 ) tan( )𝑞̂ 2 2 tan( )𝑞̂ 2
𝑒 2ℏ 𝑚 = 𝑒− 2ℏ 2 𝑒 −2ℏ𝑚𝜔 sin(𝜔𝑡)𝑝̂ 𝑒 − 2ℏ 2

Let’s fix parameters 𝑎 and 𝑏 such that:

𝑎 = 𝑡⁄𝑚 and 𝑏 = 𝑚𝜔2 𝑡
This also fixes other parameters introduced in formulas above:

𝛼 = √𝑎𝑏 = 𝜔𝑡, 𝛽 = √𝑏⁄𝑎 = 𝑚𝜔

Then,
𝛼 𝜔𝑡
𝑥 = 𝛽 tan ( ) = 𝑚𝜔 tan ( )
2 2
sin 𝛼 sin(𝜔𝑡)
𝑦= =
𝛽 𝑚𝜔
Thus, matrix element can be further reduced to
̂2
𝑖𝑡 𝑝 𝑖𝑚𝜔 𝜔𝑡 𝑖 sin(𝜔𝑡) 2
− ( +𝑚𝜔 2 𝑞̂ 2 ) tan( )(𝑦 2 +𝑥 2 )
⟨𝑦|𝑒 2ℏ 𝑚 |𝑥⟩ = 𝑒 − 2ℏ 2 ⟨𝑦|𝑒 −2ℏ 𝑚𝜔
𝑝̂
|𝑥⟩
Which can lead to transition amplitude exactly in the same manner as in the last two cases.