Exam 1

Quantum Software Master Class

Prof. Jacob Biamonte, Skolkovo Institute of Science and
Technology (Skoltech)

16 February

MA03266 — Blue 408 — Thursday and Friday 10:00 to 12:00

Solutions by Max B.1

1
[email protected]

February 4, 2019

Pauli operators
X X
1 = |0ih0| + |1ih1| = |aiha|, X = |0ih1| + |1ih0| = |1 − aiha|,
X X
Y = −i|0ih1| + i|1ih0| = −i (−1)a |aih1 − a|, Z = |0ih0| − |1ih1| = (−1)a |aiha|

for a ∈ {0, 1}.

Projectors on |0i , |1i states
def 1 def 1
P0 = |0ih0| = (1 + Z), P1 = |1ih1| = (1 − Z).
2 2

Problem 1. Show that P1 ⊗ U + P0 ⊗ 1 is unitary if and only if U is unitary.

Solution: Denote the if statement by → and the ’only if’ by ←. Prove → first: if U is unitary,
then W = P1 ⊗ U + P0 ⊗ 1 should also be. Take Hermitian conjugate W †

W † = P1† ⊗ U † + P0† ⊗ 1† = P1 ⊗ U −1 + P0 ⊗ 1

and consider a product

W † W = P1 ⊗ U −1 + P0 ⊗ 1



P1 ⊗ U + P0 ⊗ 1 =
| {z } | {z } | {z } | {z }
denote ã b a b

where one should remember that (A1 ⊗ B1 )(A2 ⊗ B2 ) = (A1 A2 ) ⊗ (B1 B2 ) so since P0 P1 = P1 P0 =
1 2 2
4 (1 − Z ) = 0 and P0,1 = P0,1 , one gets

= ãa + ba + ãb +bb = (P0 + P1 ) ⊗ 1 = 1 ⊗ 1.

| {z }
=0

1
Lets now prove ←. If W is unitary, the above expression for W † W equals 1 ⊗ 1, but one has U †
instead of U −1 in W †

W † W = P1 ⊗ U † + P0 ⊗ 1 P1 ⊗ U + P0 ⊗ 1 = 1 ⊗ 1



| {z } | {z } | {z } | {z }
ã b a b

= ãa + bb = P1 ⊗ (U † U ) + P0 ⊗ 1 = 1 ⊗ 1
and so
P1 ⊗ (U † U ) = 1 ⊗ 1 − P0 ⊗ 1 = (1 − P0 ) ⊗ 1 = P1 ⊗ 1
so finally U † U = 1.
Comment: Let us remind the reader where does the algebraic definition of unitary operators comes
from. The operator of interest W = P1 ⊗ U + P0 ⊗ 1 acts on (complex vector) space M = C2 × C2 ,
which is simply M ' C4 , so now a vector |vi ≡ vi in M is carrying one (complex) index. The fact
that an operator W (automorphism W : M → M is unitary is the same as (if and only if) it
preserves the scalar products which are introduced in the following way
X
hu|vi = ūi vi
i

where z̄ is complex conjugate of z. So if W acts on M (space of ket-s), its Hermitian conjugate

W † acts on dual space M∗ (of bra-s, or equivalently of scalar functions on M). Thus for W to be
unitary one needs to ensure
D  E X
W † uW v = Wji† ūj Wik vk = hu|vi

ijk

so the operator W † W = 1 acts trivially.

Problem 2. Show that A† B † = B † A† if and only if [A, B] = 0.

Solution: Denote the if statement by → and the only if statement by ←. Let’s make use of
the following fact: (AB)† = B † A† . If we denote element-wise complex conjugate of A by Ā, then
† †
(B † A† )ij = Bik Akj = B̄ki Ājk = Ājk B̄ki = Ajk Bki = (AB)Tij = (AB)†

So now to prove → just take a Hermitian conjugate of A† B † = B † A† to get BA = AB. Same works
for the ←. A philosophical interpretation of this fact is the following: taking Hermitian conjugate
of observables (Hermitian operators) reverts time for them. But this should not break the fact of
their mutual observability [A, B] = 0.

Problem 3. Let P 2 = P and show that

e−iθP = 1 + P (e−iθ − 1) (1)

by the series expansion

∞
A
X Ak
e = (2)
k!
k=0

2
Solution: In the power series for the exponent
∞
−iθP
X (−iθ)k
e = Pk
k!
k=0

any power P k for k ≥ 1 equals P k = P by successive use of P 2 = P , so one gets

∞
−iθP
X (−iθ)k
e = |{z}
1 +P
k!
k=0 k=1
| {z }
=e−iθ −1

Problem 4. Let A2 = 1 and show that

e−iθA = 1 cos(θ) − iA sin(θ) (3)

Solution: If an exponent of operator exp(A) is defined as a formal power series, and the sine sin(A)
and the cosine cos(A) are also, then Euler’s identity holds (formally):

e−iθA = cos(θA) − i sin(θA)

where cos only has even powers of A that by A2 = 1 turn into 1, and sin has only odd powers
starting from A1 , so A can be taken out as a common multiple, ans A2 = 1 would turn all the
resting even powers to 1, so one gets the desired answer.

Problem
√ 5. (1) Show that√the corresponding eigenvectors of the Pauli matrices X, Y, Z are
2 |±i = |0i ± |1i and 2 |y± i = |0i ± i · |1i and |ji for j = 0, 1.

(2) Consider the operators X ⊗ X, Y ⊗ Y , Z ⊗ Z. Use the tensor product to define an

eigenbasis and calculate the eigenvalues for these same operators.

(?) Find an eigenbasis for these same operators which can not be factorized using the tensor
product. (In quantum theory, these basis states are called, entangled.)

Solution: See the solution of Problem 21 for the eigensystems of X and Y . As for Z, computational
basis is the eigenbasis for it already.
Now by definition of tensor product of operators and vector spaces

(A ⊗ B) |ui ⊗ |vi = (A |ui) ⊗ (B |vi)

one gets the eigenbasis of X ⊗ X

{|+i , |−i} ⊗ {|+i , |−i} = {|++i , |+−i , |−+i , |−−i}

of Y ⊗ Y
{|y+ i , |y− i} ⊗ {|y+ i , |y− i} = {|y+ y+ i , |y+ y− i , |y− y+ i , |y− y− i}
and of Z ⊗ Z accordingly, just combining the basis vectors. The span of these tensor products of
eigenspaces is clearly all the 22 -dimensional vector-space.

3
The non-decomposable basis for X ⊗ X, Y ⊗ Y and Z ⊗ Z will arise in Problem 23 and is called
Bell’s basis.
Φ = √1 |00i + |11i , Φ− = √1 |00i − |11i ,
 +
 

2 2
Ψ = √1 |01i + |10i , Ψ− = √1 |01i − |10i .
 +
 

2 2
And the eigenvalues are
|Φ+ i |Φ− i |Ψ+ i |Ψ− i
X ⊗X +1 −1 +1 −1
Y ⊗Y −1 +1 +1 −1
Z ⊗Z +1 +1 −1 −1
Notice that if one adds a row of (1, 1, 1, 1) to the table (eigenvalues of 1 ⊗ 1), the rank of the
resulting matrix is 4, which it should be.

Problem 6. Show that for Pa2 = Pa , Pa has eigenvalues 0, 1. Show that for Z 2 = 1, Z has
eigenvalues 0, 1. Consider det(Z − λ1) = λ2 − 1 = 0 and det(P − λ1) = λ(λ − 1) = 0. By
substituting λ 7→ Pa (or Z) and sending scalars c to c1, show explicitly that Pa and Z satisfy
their own characteristic equation.

Solution: First consider the (operator) identity Pa2 = Pa . On the eigen-subspace spanned by
eigenvector |λi it gives λ2 = λ a quadratic equation with only two roots λ1,2 = 0, 1. Same approach
works for Z 2 = 1 providing λ1,2 = −1, 1.
The fact that both Pa and Z satisfy their own characteristic equations follows directly from the
identities, but another explanation (if desired) is that Hamilton-Caley theorem for 2 × 2 matrices
states that any matrix A satisfies

A2 − Tr(A) A + det(A) 1 = 0

where Tr(Z) = −1 + 1 = 0, Tr(Pa ) = 0 + 1 = 1 and det(Z) = −1, det(Pa ) = 0.

. 
Problem 7. Express h∧ = ∆(P001 + P011 + P101 + P110 ) over the basis 1, Z, ⊗ . Note in
general that transitioning between these two bases will not increase locality.

Solution: Let’s try to solve it in a manner a bit wiser than just substituting P0 = 21 (1 + Z) and
P1 = 12 (1 − Z) and opening the brackets to get a huge ton of terms. Notice that the expression
we’re working with
h = P001 + P011 + P101 + P110
is missing four terms (of 23 = 8 totally possible) to be a full tensor-power
 
X
h= Pijk  − (P100 + P000 + P111 + P010 )
i,j,k=0,1

where  
X
 Pijk  = (P0 + P1 ) ⊗ (P0 + P1 ) ⊗ (P0 + P1 ) = 1 ⊗ 1 ⊗ 1
i,j,k=0,1

4
and the rest gives
(P100 + P000 + P010 ) + P111 = (P10 + P00 + P01 ) ⊗ P0 + P111
= 1 ⊗ 1 ⊗ P0 + (−P110 + P111 ) = 1 ⊗ 1 ⊗ P0 + P1 ⊗ P1 ⊗ (P0 − P1 = Z)
| {z }
so we finally have
h = 1 ⊗ 1 ⊗ 1 − 1 ⊗ 1 ⊗ P0 + P1 ⊗ P1 ⊗ Z
so substituting the expressions for P0 and P1 here one has not as many terms, so this gets bearable.

Problem 8. (AND penalty). Derive a series of inequalities (equalities) and find a solu-
tion providing integer values for a, b, c, d such that h∧ in (4) has a zero eigenspace given as
span{|x1 x2 x3 i | x3 = x1 x2 } and all other eigenspaces are > 1.

h∧ = aP13 + bP12 ⊗ P13 + cP11 ⊗ P12 + dP11 ⊗ P13 . (4)

Solution: Here is the eigensystem of h∧

Eigenvector |100i |010i |111i |001i |011i |110i |101i |111i
Eigenvalues 0 0 0 a a+b c a+d a+b+c+d
where the eigenvectors are the computational basis itself since h∧ is diagonal in it. So the only
nonzero satisfying x3 = x1 x2 is |111i and thus the answer is that a + b + c + d = 0 and all other
expressions in the table are ≥ 1.

Problem 9. Develop a two-body penalty function that performs the copy operation. In other
words, develop the penalty function such that the low-energy subspace is in span{|000i , |111i}.

Solution: Our function should penalize any of the three spins being different from the rest, so
making it two-body interactions (or clauses including two variables only) one arrives to three terms
each making the two interacting spins being equal, so
X
E(x1 , x2 , x3 ) = − (OR(xi , xj ) − AND(xi , xj ))
3 pairs

Problem 10. Show that

− x1 x2 x3 = min z(2 − x1 − x2 − x3 ) (5)

z∈B

and that
− x1 x2 x3 = min z(−x1 + x2 + x3 ) − x1 x2 − x1 x3 + x1 (6)
z∈B

Solution: Let us consider all possible options for x1 , x2 , x3 and z Boolean values, keeping in mind
that both sides of the expression
−x1 x2 x3 = min z(−x1 + x2 + x3 )
z∈B
| {z }
call this RHS

are symmetric with respect to (x1 , x2 , x3 ), so one only cares about the the total number of, say,
1 values, but not their order. See the table

5
 # of 1s among x1,2,3 −x1 x2 x3 value RHS(z = 0) RHS(z = 1) RHS
0 0 0 2 0
1 0 0 1 0
2 0 0 0 0
3 -1 0 -1 -1
And to prove the second identity

−x1 x2 x3 = min z(−x1 + x2 + x3 ) − x1 x2 − x1 x3 + x1

z∈B

it is helpful to notice that its RHS (it was for right-hand-side) is symmetric with respect to x2 ↔ x3 ,
and to take x1 out of brackets for the not-in-min terms to get x1 (1 − x2 − x3 ) so the corresponding
table gives
x1 value # of 1s among x2,3 −x1 x2 x3 value RHS(z = 0) RHS(z = 1) RHS
0 0 0 0 0 0
0 1 0 0 1 0
0 2 0 0 2 0
1 0 0 1 0 0
1 1 0 0 0 0
1 2 -1 -1 0 -1

Problem 11. Provide a value assignment of x1 , x2 , x3 , x4 such that f (x1 , x2 , x3 , x4 ) = 1 and

hence show that f is satisfiable.

f = (x1 ∨ x3 ∨ x4 ) ∧ (x2 ∨ x3 ∨ x4 ) ∧ (x1 ∨ x2 ∨ x3 ) (7)

Solution: To find these just type

SatisfiabilityInstances[f[x1, x2, x3, x4], {x1, x2, x3, x4}, 16]
(where f (x1 , x2 , x3 , x4 ) is our Boolean function) in your Mathematica to look for all possible satis-
fying combinations among 24 = 16 of those to get the possible options and then type
TeXForm[Transpose[Boole[%]]]
to get the following table:
x1 1 1 1 1 1 0 0 0 0 0
x2 1 1 1 0 0 1 1 0 0 0
x3 1 1 0 1 1 1 0 1 0 0
x4 1 0 0 1 0 1 0 1 1 0

Problem 12. (Eigenvalues of thermal states ?). It can be shown that

P
e−β j Zj O1 X  
ρβ = P = 1 − (−1)b tanh(β) |bij hb|j (8)
tr{e−β j Zj
} j
2
b∈{0,1}

derive a formula for the eigenvalues of a state in terms of the Boolean variables in the general

6
 bit string |x1 , x2 , . . . , xl i.

Solution: First, since [Zi , Zj ] = δij , where

(
1, i = j
δij =
6 j
0, i =

the exponent of sum can be decomposed into a product of exponents

 X  Y
exp − β Zj = exp (−βZj ) = exp (−βZ)⊗l
j j

and it holds that tr (A ⊗ B) = (trA) · (B), so one shows indeed that

 ⊗l   −β ⊗l
exp(−βZ) 1 e 0
ρβ = =
tr exp(−βZ) 2 cosh β 0 e+β

And the eigenvalue ρβ |x1 , x2 , . . . , xl i = λ |x1 , x2 , . . . , xl i reads

Y1 X  
λ= 1 − (−1)b tanh β δb,xj
2
j b∈{0,}

Problem 13. Physical systems at thermal equilibrium are said to be described by a Gibbs
state
−βH
def e
ρβ = (9)
Z
P
where the Hamiltonian H = i Ci is a sum over projectors onto 3SAT clauses and the partition
function is
Z = tr{e−βH }
def
(10)
Let |ii denote (possibly degenerate) lowest eigenstates of H. We label these possibly degen-
erate states by letting i range from 1 upto d. Call λmin the lowest eigenvector of H.
Show that
d
1 X D  −βH E
lim ie (i) = 1 (11)
β→∞ Z
i=1

and hence establish that in infinite time, sampling a thermal system can solve SAT instances
with probability one.
P
Solution: Since 3SAT Hamiltonian H = i Ci is classical (and thus is a diagonal matrix), the
state X
e−βH = e−βEj |ji
j

is a (from a quantum perspective) a pure state (just a vector), the scalar products give
d D 
X E Xd X
ie−βH = e−βEj hi|ji = d · e−βEmin

i=1 i=1 j

7
since the eigenvectors are chosen orthonormal hi|ji = δij . And the limit of interest is
 
d · e−βH  X
lim = 1+ e−β(Ej −Emin )  = 1
β→∞ Z
Ej >Emin

. The physical intuition behind this is that evolution (Hamiltonian being the generator of the
evolution) of a physical system in imaginary time t → it ≡ β (aka Wick rotation) takes it to the
ground state (or ground states), but for so the system should have interaction within. For 3SAT on
a connected graph this holds true.

Problem 14. Show that the complex numbers C form a complex Euclidean space.

Solution: By definition of addition on complex numbers, they form a vector space, so one needs
to ensure that the metric d : C × C → R satisfies metric axioms: non-negativity, identity on
indiscernibles, symmetry and triangle inequality. Standard metric, induced by norm in complex
numbers, called modulus p
z = a + ib → |z| = a2 + b2
is given by
d(z1 , z2 ) = |z1 − z2 |
which gives for z1 = x1 + iy1 , z2 = x2 + iy2
p
d(z1 , z2 ) = (x1 − x2 )2 + (y1 − y2 )2

which is standard Euclidean metric on a plane (2-dimensional Euclidean space) that is known to
satisfy these axioms for 3000 years or so.

Problem 15. (Z2 symmetry). Consider the tunable two-body Ising Hamiltonian acting on n
spins as X
H= Jij Zi Zj (12)
i<j
Nn
Using the identity that XZX = −Z or otherwise, for X̃ = l=0 Xl show that
h i
X̃, H = 0 (13)

and hence establish that the definition of |0i, |1i is entirely arbitrary with respect to H.

Solution: This is indeed a more formal expression of the total-flip-symmetry argument. First of all
notice that the operators in H act as Z on i-th and j-th spins and as identity on others, so formally
speaking
i−1 n
! !
O O
Zi = 1 ⊗Z ⊗ 1
k=1 k=i+1

and so the ij-th term of the Hamiltonian in the [X̃, H] expression gives (omit Jij )

X̃Zi Zj − Zi Zj X̃

8
which we now multiply by X̃ keeping in mind that X 2 = −1, and so get
X̃Zi Zj X̃ + Zi Zj
and now taking X̃ through the 1-s of not i-th or j-th spins gives −1 acting on them and Zi and Zj
turn into −Zi and Zj , so one gets
−Zi Zj + Zi Zj = 0

Problem 16. (Number partitioning). Given a set of N positive numbers S = n1 , . . . , nN , is

there a partition of this set of numbers into two disjoint subsets R and SR, such that the sum
of the elements in both sets is the same?
Let ni (i = 1, . . . , N = |S|) describe the numbers in set S. Show that
!2
X
H= ni si ≥0 (14)
i

vanishes if and only if such a disjoint partition exists. Here si is a spin variable ∈ ±1.
Solution: This problem is one of many NP-problems, embedding of which is discussed in the well-
recommended article arxiv.org/abs/1302.5843. The explanation is quite clear: if such a partition
exists, there exists an assignment (degenerate by a flip si → −si of all spins) si that takes H to
zero, and since the values of H are nonnegative, this is the lowest-energy state. If this expression
never vanishes, (by this we are proving the ”only if” part, the
vanishes ← exists
since if A ← B, then ¬A → ¬B) then the sum in brackets
X
ni s i
i

is never zero (for any of 2N assignments), which means there is no such assignment. The Hamiltonian
is a square of this sum not only to satisfy non-negativity, but also to be some quadratic (Ising-form)
expression in si .

Problem 17. Using the result from Problem 7, let H∧ (Z1 , Z2 , Z3 ) be the penalty Hamiltonian
applying the AND gate. Show that NAND can be recovered by calculating H∧ (Z1 , Z2 , X3 Z3 X3 ).
Solution: Have a look at the truth tables of AND and NAND
X1 X2 X3 X1 X2 X3
0 0 0 0 0 1
AND: 0 1 0 NAND: 0 1 1
1 0 0 1 0 1
1 1 1 1 1 0
this is to make sure that NAND is really the negation of AND. Now if spin number 3 is the ’answer’-
bit, one simply needs to flip it. If a state vector is acted on with an operator Z, to flip one needs
to make a linear transform
Z → X −1 ZX
since X is a linear transform of the state space corresponding to a spin flip. As far as X 2 = 1, X −1
can be chosen to simply be X.

9
 Problem 18. Let H = H † , H 2 = 1. Show that each such H gives rise to a self-adjoint
projector—and conversely—hence establish a bijection between self-adjoint unitary maps and
projectors.

Solution: Simply take one of two options

1
P± = (1 ± H)
2
so projector property
1
1
P±2 = 1 ± 2H + H 2 = (2 ± 2H) = P±
4 4
and the ”self-adjointness”:
1 1
P±† = (1 + H)† = (1 + H † ) = P±
2 2
since H † = H.

Problem 19. Building on the results from Problem 18, let M be the invertible map taking
projectors to self-adjoint unitaries. Show that for orthogonal projectors Pi , Pj , with corre-
sponding unitary maps M (Pi ), M (Pj ) the composition (?) of self-adjoint unitaries is given by
the negated sum of projectors under the image of M viz. M (Pi ) ? M (Pj ) = −M (Pi + Pj ) and
hence show that [M (Pi ), M (Pj )] = 0 taken with respect to the defined ? product.

Solution: So the map can be chosen from two possible options

M± (P ) = ±(2P − 1)

Now whichever option is chosen, the composition defined in the following way

M (Pi ) ? M (Pj ) = −(2(Pi + Pj ) − 1)

is clearly symmetric with respect to Pi , Pj , so the commutator

M (Pi ) ? M (Pj ) − M (Pj ) ? M (Pi ) = 0

equals zero.

Problem 20. (Rank-1 projectors ?). Show that a non-trivial operator P is a Schmidt rank-1
projector if and only if it can be written as |ψihψ|.

Solution: Again, start with the (if) → statement. So if P is a rank-1 projector, it can be SVD-
decomposed to σ|uihv| with σ just a number, since its rank-1. So projector property gives

σ 2 |ui hv|ui hv| = hv|ui σ 2 |uihv| = σ|uihv|

so since projector’s only allowed nonzero eigenvalue is σ = 1 that gives

hu|vi = 1

and thus |ui = |vi and thus (denoting it |Ψi) P = |ψihψ|. The ← obviously holds true P 2 = P
holds for P = |ψihψ|.

10
 Problem 21. By using Dirac notation (i) find eigenvalues and eigenvectors of Y, Z. (ii) Show
that
[σi , σj ] = 2iijk σk
where  is the Levi-Civita symbol, σ1 = X, σ2 = Y and σ3 = Z. (iii) Show that iσ1 σ2 σ3 = 1.

Solution: Using Dirac’s notation, one gets

h0|1i = h1|0i = 0, h0|0i = h1|1i = 1

So the action of the X and Y operators is quite clear:

( (
|0i → |1i |0i → +i |1i
X: Y :
|1i → |0i |1i → −i |0i

Noting this structure, it is straightforward to try a symmetric and an antisymmetric combinations

for X eigenvectors

X(|0i + |1i) = |1i + |0i , X(|0i − |1i) = −(|0i − |1i)

with the corresponding eigenvalues being ±1. Now for Y one gets in a similar fashion

Y (i |0i + |1i) = − |1i − i |0i , Y (−i |0i + |1i) = |1i − i |0i

and the corresponding eigenvalues are ∓1.

Problem 22. Provide iff statements defining conditions on the aij terms for the operator
X
aij |iihj| (15)
i,j∈{0,1}

to be (i) Hermitian and (ii) a projector.

Solution: From linear algebra we know that for A to be Hermitian it should satisfy A† = ĀT = A,
so āij = aji . Projector property A2 = A gives
X
aik akj = ai0 a0i + ai1 a1i = aij
k

Problem 23. Let

Φ = √1 |00i + |11i , Φ = √1 |00i − |11i ,

 +
 −

2 2
 + 1

Ψ = √ |01i + |10i ,
 − 1
Ψ = √ |01i − |10i .


2 2

Show that @ |ψ1 i , |ψ2 i ∈ C2 for the Bell state to be expressed as |Φ± i = |ψ1 ψ2 i as well as |Ψ± i.
The states |Φ± i are therefore called entangled.

11
Solution: One possible (quite straightforward way) to show this is a contrario: introduce such two
vectors  +  +
φ ⊗ φ = φ+ |0i + φ+ |1i ⊗ φ+ |0i + φ+ |1i = Φ+


 
1 2 10 11 20 21

to get a system of four equations (on four complex variables)


+ +
√
φ10 φ20 = 1/ 2


φ+ φ+ = 0

11 20
+ +
 φ 10 φ21 = 0
φ+ φ+ = 1/√2



11 21

that can’t be possibly satisfied. Same works for all other Bell states.

Problem 24. (Bell basis). Using the right hand side equations,
X X
σ0 = |aiha|, σ1 = |1 − aiha|,
X X
σ2 = i (−1)a |aih1 − a|, σ3 = (−1)a |aiha|.

for a ∈ {0, 1}. Show that

 
1 X 1 X
√  hl| ⊗ hl| σi ⊗ 1 = √ (hl| σi ) ⊗ hl| (16)
2 l∈{0,1} 2 l∈{0,1}

defines the Bell states (which defines an orthonormal basis in C2 ⊗ C2 . Here σi indexes the
Pauli matrices.

Solution: Let’s deal with the term in brackets hl| σi first

hl| σi σ0 σ1 σ2 σ3
h0| h0| h1| −i h1| + h0|
h1| h1| h0| +i h0| − h1|
to see the correspondence
Φ+ Ψ+ Ψ− Φ−
σ0 σ1 σ2 σ3

where the (global phase) multiple of i for σ2 is irrelevant.

12