Incompatibility of quantum measurements is a crucial concept of quantum information, addressing whether two or more quantum measurements can be performed on a quantum system simultaneously. It highlights the unique and non-classical behavior of quantum systems. This concept is fundamental to the nature of quantum mechanics and has practical applications in various quantum information processing tasks like quantum key distribution and quantum metrology.
History
Early ages
The concept of incompatibility of quantum measurements originated from Heisenberg's uncertainty principle, which states that certain pairs of physical quantities, like position and momentum, cannot be simultaneously measured with arbitrary precision. This principle laid the groundwork for understanding the limitations of measurements in quantum mechanics.
Mid-20th century
In the mid-20th century, researchers began to formalize the idea of compatibility of quantum measurements, and to explore conditions under which a set of measurements can be performed together on a single quantum system without disturbing each other. This was crucial for understanding how quantum systems behave under simultaneous observations.
Late-20th century
The study of incompatibility of quantum measurements gained significant attention with the rise of quantum information theory. Researchers realized that measurement incompatibility is not just a limitation but also a resource for various quantum information processing tasks. For example, it plays a crucial role in quantum cryptography, where the security of quantum key distribution protocols relies on the incompatibility of certain quantum measurements.
21st century
Modern research focuses on quantifying measurement incompatibility using various measures. Quite a number of approaches involve robustness-based measures, which assess how much noise can be added to a set of quantum measurements before they become compatible.
Definition
In quantum mechanics, two measurements,
M
1
{\displaystyle M_{1}}
and
M
2
{\displaystyle M_{2}}
, are called compatible if and only if there exists a third measurement
M
{\displaystyle M}
of which
M
1
,
M
2
{\displaystyle M_{1},M_{2}}
can be obtained as margins, or equivalently, from which
M
1
,
M
2
{\displaystyle M_{1},M_{2}}
can be simulated via classical post-processings. More precisely, let
M
1
:
A
1
→
B
(
H
)
{\displaystyle M_{1}:{\mathcal {A}}_{1}\to {\mathcal {B}}(\mathbb {H} )}
and
M
2
:
A
2
→
B
(
H
)
{\displaystyle M_{2}:{\mathcal {A}}_{2}\to {\mathcal {B}}(\mathbb {H} )}
be two positive operator-valued measures (POVMs), where
(
X
1
,
A
1
)
,
(
X
2
,
A
2
)
{\displaystyle (X_{1},{\mathcal {A}}_{1}),(X_{2},{\mathcal {A}}_{2})}
are two measurable spaces, and
B
(
H
)
{\displaystyle {\mathcal {B}}(\mathbb {H} )}
is the set of bounded linear operators on a Hilbert space
H
{\displaystyle \mathbb {H} }
.
Then
M
1
,
M
2
{\displaystyle M_{1},M_{2}}
are called compatible (jointly measurable) if and only if there is a POVM
M
:
A
1
⊗
A
2
→
B
(
H
)
{\displaystyle M:{\mathcal {A}}_{1}\otimes {\mathcal {A}}_{2}\to {\mathcal {B}}(\mathbb {H} )}
such that
M
(
A
1
×
X
2
)
=
M
1
(
A
1
)
,
{\displaystyle M(A_{1}\times X_{2})=M_{1}(A_{1}),}
M
(
X
1
×
A
2
)
=
M
2
(
A
2
)
,
{\displaystyle M(X_{1}\times A_{2})=M_{2}(A_{2}),}
for all
A
1
∈
A
1
{\displaystyle A_{1}\in {\mathcal {A}}_{1}}
and
A
2
∈
A
2
{\displaystyle A_{2}\in {\mathcal {A}}_{2}}
. Otherwise,
M
1
,
M
2
{\displaystyle M_{1},M_{2}}
are called incompatible. The definition of compatibility of a finite number of POVMs is similar.
,
+
,
−
,
{
+
,
−
}
}
→
B
(
C
2
)
{\displaystyle M:\{\varnothing ,+,-,\{+,-\}\}\to {\mathcal {B}}(\mathbb {C} ^{2})}
on a qubit it is possible to write
σ
→
]
,
{\displaystyle M_{\pm }={\cfrac {1}{2}}\left[(1\pm \alpha )\mathbf {I} \pm {\vec {m}}\cdot {\vec {\sigma }}\right],}
σ
→
=
m
x
σ
x
+
m
y
σ
y
+
m
z
σ
z
{\displaystyle {\vec {m}}\cdot {\vec {\sigma }}=m_{x}\sigma _{x}+m_{y}\sigma _{y}+m_{z}\sigma _{z}}
with
σ
x
,
σ
y
,
σ
z
{\displaystyle \sigma _{x},\sigma _{y},\sigma _{z}}
Pauli matrices.
Let
M
±
|
1
{\displaystyle M_{\pm |1}}
and
M
±
|
2
{\displaystyle M_{\pm |2}}
be two such POVMs with
α
1
=
α
2
=
0
{\displaystyle \alpha _{1}=\alpha _{2}=0}
,
m
→
1
=
(
μ
,
0
,
0
)
{\displaystyle {\vec {m}}_{1}=(\mu ,0,0)}
, and
m
→
2
=
(
0
,
0
,
μ
)
{\displaystyle {\vec {m}}_{2}=(0,0,\mu )}
, where
μ
∈
[
0
,
1
]
{\displaystyle \mu \in [0,1]}
. Then for
μ
∈
[
0
,
1
/
2
]
{\displaystyle \mu \in [0,1/{\sqrt {2}}]}
one can verify that the following POVM
M
+
,
+
:=
1
4
[
I
+
μ
(
σ
x
+
σ
z
)
]
,
{\displaystyle M_{+,+}:={\cfrac {1}{4}}\left[\mathbf {I} +\mu (\sigma _{x}+\sigma _{z})\right],}
M
−
,
+
:=
1
4
[
I
+
μ
(
−
σ
x
+
σ
z
)
]
,
{\displaystyle M_{-,+}:={\cfrac {1}{4}}\left[\mathbf {I} +\mu (-\sigma _{x}+\sigma _{z})\right],}
M
+
,
−
:=
1
4
[
I
+
μ
(
σ
x
−
σ
z
)
]
,
{\displaystyle M_{+,-}:={\cfrac {1}{4}}\left[\mathbf {I} +\mu (\sigma _{x}-\sigma _{z})\right],}
M
−
,
−
:=
1
4
[
I
+
μ
(
−
σ
x
−
σ
z
)
]
,
{\displaystyle M_{-,-}:={\cfrac {1}{4}}\left[\mathbf {I} +\mu (-\sigma _{x}-\sigma _{z})\right],}
is a joint POVM of
M
±
|
1
{\displaystyle M_{\pm |1}}
and
M
±
|
2
{\displaystyle M_{\pm |2}}
, despite the fact that
M
±
|
1
{\displaystyle M_{\pm |1}}
and
M
±
|
2
{\displaystyle M_{\pm |2}}
are noncommutative for
μ
∈
(
0
,
1
/
2
]
{\displaystyle \mu \in (0,1/{\sqrt {2}}]}
.
As for
μ
∈
(
1
/
2
,
1
]
{\displaystyle \mu \in (1/{\sqrt {2}},1]}
, it was shown that
M
±
|
1
,
M
±
|
2
{\displaystyle M_{\pm |1},M_{\pm |2}}
are incompatible.
Criteria for compatibility
Analytical criteria
Let
M
1
,
M
2
{\displaystyle M_{1},M_{2}}
be two two-outcome POVMs on a qubit. It was shown that
M
1
,
M
2
{\displaystyle M_{1},M_{2}}
are compatible if and only if
m
→
2
−
α
1
α
2
)
2
,
{\displaystyle (1-C_{1}^{2}-C_{2}^{2})\left(1-{\cfrac {\alpha _{1}^{2}}{C_{1}^{2}}}-{\cfrac {\alpha _{2}^{2}}{C_{2}^{2}}}\right)\leqslant \left({\vec {m}}_{1}\cdot {\vec {m}}_{2}-\alpha _{1}\alpha _{2}\right)^{2},}
where
C
i
=
1
2
[
(
1
+
α
i
)
2
−
|
m
→
i
|
2
+
(
1
−
α
i
)
2
−
|
m
→
i
|
2
]
,
{\displaystyle C_{i}={\cfrac {1}{2}}\left[{\sqrt {(1+\alpha _{i})^{2}-\left|{\vec {m}}_{i}\right|^{2}}}+{\sqrt {(1-\alpha _{i})^{2}-\left|{\vec {m}}_{i}\right|^{2}}}\right],}
for
i
=
1
,
2
{\displaystyle i=1,2}
.
Moreover, let
M
1
,
M
2
,
M
3
{\displaystyle M_{1},M_{2},M_{3}}
be three two-outcome POVMs on a qubit with
α
1
=
α
2
=
α
3
=
0
{\displaystyle \alpha _{1}=\alpha _{2}=\alpha _{3}=0}
. Then it was proved that
M
1
,
M
2
,
M
3
{\displaystyle M_{1},M_{2},M_{3}}
are compatible if and only if
min
n
→
∈
R
3
∑
j
=
0
3
|
n
→
j
−
n
→
|
⩽
4
,
{\displaystyle \min _{{\vec {n}}\in \mathbb {R} ^{3}}\sum _{j=0}^{3}\left|{\vec {n}}_{j}-{\vec {n}}\right|\leqslant 4,}
where
n
→
0
=
m
→
1
+
m
→
2
+
m
→
3
{\displaystyle {\vec {n}}_{0}={\vec {m}}_{1}+{\vec {m}}_{2}+{\vec {m}}_{3}}
, and
n
→
j
=
2
m
→
j
−
n
→
0
{\displaystyle {\vec {n}}_{j}=2{\vec {m}}_{j}-{\vec {n}}_{0}}
for
j
=
1
,
2
,
3
{\displaystyle j=1,2,3}
.
Numerical criteria
For a finite number of POVMs
{
M
x
|
i
}
{\displaystyle \{M_{x|i}\}}
on a finite-dimensional quantum system, with a finite number of measurement outcomes (labeled by
x
{\displaystyle x}
), one can cast the following semidefinite program (SDP) to decide whether they are compatible or not:
g
i
v
e
n
{
M
x
|
i
}
,
{
p
(
x
|
i
,
λ
)
}
{\displaystyle \mathrm {given} \ \{M_{x|i}\},\{p(x|i,\lambda )\}}
max
{
G
λ
}
μ
{\displaystyle \max _{\{G_{\lambda }\}}\mu }
s
u
b
j
e
c
t
t
o
∑
λ
p
(
x
|
i
,
λ
)
G
λ
=
M
x
|
i
,
∀
x
,
i
,
{\displaystyle \mathrm {subject\ to} \sum _{\lambda }p(x|i,\lambda )G_{\lambda }=M_{x|i},\quad \forall x,i,}
G
λ
⩾
μ
I
,
∀
λ
a
n
d
∑
λ
G
λ
=
I
.
{\displaystyle G_{\lambda }\geqslant \mu \mathbf {I} ,\quad \forall \lambda \quad \mathrm {and} \quad \sum _{\lambda }G_{\lambda }=\mathbf {I} .}
where
{
p
(
x
|
i
,
λ
)
}
{\displaystyle \{p(x|i,\lambda )\}}
is a deterministic classical post-processing (i.e.,
0
{\displaystyle 0}
or
1
{\displaystyle 1}
valued conditional probability density). If for all
{
p
(
x
|
i
,
λ
)
}
{\displaystyle \{p(x|i,\lambda )\}}
the maximizer
{
G
λ
}
{\displaystyle \{G_{\lambda }\}}
leads to a negative
μ
{\displaystyle \mu }
, then
{
M
x
|
i
}
{\displaystyle \{M_{x|i}\}}
are incompatible. Otherwise, they are compatible.
Quantifiers of incompatibility
Incompatibility random robustness
Let
{
M
x
|
i
}
{\displaystyle \{M_{x|i}\}}
be a finite number of POVMs. Incompatibility noise robustness
R
n
{\displaystyle R_{\mathrm {n} }}
of
{
M
x
|
i
}
{\displaystyle \{M_{x|i}\}}
is defined by
R
n
(
{
M
x
|
i
}
)
=
inf
{
p
∈
[
0
,
1
]
|
{
(
1
−
p
)
M
x
|
i
+
p
I
/
|
x
|
}
are compatible
}
,
{\displaystyle R_{\mathrm {n} }(\{M_{x|i}\})=\inf\{p\in [0,1]|\{(1-p)M_{x|i}+p\mathrm {I} /|x|\}\ \ {\text{are compatible}}\},}
where
|
x
|
{\displaystyle |x|}
denotes the number of outcomes. This quantity is monotonic under quantum channels, i.e. it does not increase under pre-processing by a quantum channel.
Incompatibility weight
Incompatibility weight
W
{\displaystyle W}
of POVMs
{
M
x
|
i
}
{\displaystyle \{M_{x|i}\}}
is defined by
W
(
{
M
x
|
i
}
)
=
inf
{
q
⩾
0
:
{
M
x
|
i
−
q
N
x
|
i
1
−
q
}
are compatible
}
,
{\displaystyle W(\{M_{x|i}\})=\inf \left\{q\geqslant 0:\left\{{\cfrac {M_{x|i}-qN_{x|i}}{1-q}}\right\}\ \ {\text{are compatible}}\right\},}
where
{
N
x
|
i
}
{\displaystyle \{N_{x|i}\}}
is any set of POVMs.
This quantity is monotonic under pre-processing by a quantum channel and classical post-processing.
Incompatibility robustness
Incompatibility robustness
R
{\displaystyle R}
of POVMs
{
M
x
|
i
}
{\displaystyle \{M_{x|i}\}}
is defined by
R
(
{
M
x
|
i
}
)
=
inf
{
r
⩾
0
:
{
M
x
|
i
+
r
N
x
|
i
1
+
r
}
are compatible
}
,
{\displaystyle R(\{M_{x|i}\})=\inf \left\{r\geqslant 0:\left\{{\cfrac {M_{x|i}+rN_{x|i}}{1+r}}\right\}\ \ {\text{are compatible}}\right\},}
where
{
N
x
|
i
}
{\displaystyle \{N_{x|i}\}}
is any set of POVMs. Similar to incompatibility weight, this quantity is monotonic under pre-processing by a quantum channel and classical post-processing.
All these quantifiers are based on the convex distance of incompatible POVMs to the set of compatible ones under the addition of different types of noise. Specifically, all these quantifiers can be evaluated numerically, as they fall under the framework of SDPs. For instance, it was shown that incompatibility robustness can be cast as the following SDP:
g
i
v
e
n
{
M
x
|
i
}
{\displaystyle \mathrm {given} \ \{M_{x|i}\}}
min
(
{
G
λ
}
,
{
p
(
x
|
i
,
λ
)
}
)
1
d
∑
λ
t
r
(
G
λ
)
{\displaystyle \min _{\left(\{G_{\lambda }\},\{p(x|i,\lambda )\}\right)}{\cfrac {1}{d}}\sum _{\lambda }\mathrm {tr} (G_{\lambda })}
s
u
b
j
e
c
t
t
o
∑
λ
p
(
x
|
i
,
λ
)
G
λ
⩾
M
x
|
i
,
∀
x
,
i
,
{\displaystyle \mathrm {subject\ to} \sum _{\lambda }p(x|i,\lambda )G_{\lambda }\geqslant M_{x|i},\quad \forall x,i,}
G
λ
⩾
0
,
∀
λ
a
n
d
∑
λ
G
λ
=
I
d
[
∑
λ
t
r
(
G
λ
)
]
,
{\displaystyle G_{\lambda }\geqslant 0,\quad \forall \lambda \quad \mathrm {and} \quad \sum _{\lambda }G_{\lambda }={\cfrac {\mathbf {I} }{d}}\left[\sum _{\lambda }\mathrm {tr} (G_{\lambda })\right],}
where
{
p
(
x
|
i
,
λ
)
}
{\displaystyle \{p(x|i,\lambda )\}}
is a deterministic classical post-processing, and
d
{\displaystyle d}
is the dimension of the Hilbert space on which
{
M
x
|
i
}
{\displaystyle \{M_{x|i}\}}
are performed. If the minimizer is
(
{
G
λ
′
}
,
{
p
(
x
|
i
,
λ
)
′
}
)
{\displaystyle \left(\{G_{\lambda }'\},\{p(x|i,\lambda )'\}\right)}
, then
R
(
{
M
x
|
i
}
)
=
1
d
∑
λ
t
r
(
G
λ
′
)
−
1
{\displaystyle R(\{M_{x|i}\})={\cfrac {1}{d}}\sum _{\lambda }\mathrm {tr} (G_{\lambda }')-1}
.
Incompatibility and quantum information processing
Measurement incompatibility is intrinsically tied to the nonclassical features of quantum correlations. In fact, in many scenarios, it becomes evident that incompatible measurements are necessary to exhibit nonclassical correlations. Since such correlations are essential for tasks like quantum key distribution or quantum metrology, these connections underscore the resource aspect of measurement incompatibility.
Bell nonlocality
Bell nonlocality is a phenomenon where the correlations between quantum measurements on entangled quantum states cannot be explained by any local hidden variable theory. This was first demonstrated by Bell through Bell's Theorem, which showed that certain predictions of quantum mechanics are incompatible with the principle of locality (an idea that objects are only directly influenced by their immediate surroundings).
If the POVMs on Alice's side are compatible, then they will never lead to Bell nonlocality. However, the converse is not true. It was shown that there exist three binary measurements on a qubit, pairwise compatible but globally incompatible, yet not leading to any Bell nonlocality.
Quantum steering
Quantum steering is a phenomenon where one party (Alice) can influence the state of a distant party's (Bob's) quantum system through local measurements on her own quantum system. This was first introduced by Schrödinger and is a form of quantum nonlocality. It demonstrates that the state of Bob's quantum system can be "steered" into different states depending on the local measurements performed by Alice.
For quantum steering to occur, the local measurements performed by Alice must be incompatible. This is because steering relies on the ability to create different outcomes in Bob's system based on Alice's measurements, which is only possible if those measurements are not compatible. Further, it was shown that if the following POVMs
{
σ
i
−
1
/
2
σ
x
|
i
σ
i
−
1
/
2
}
{\displaystyle \left\{\sigma _{i}^{-1/2}\sigma _{x|i}\sigma _{i}^{-1/2}\right\}}
are incompatible, then a state assemblage
{
σ
x
|
i
}
{\displaystyle \{\sigma _{x|i}\}}
is steerable, where
σ
i
=
∑
x
σ
x
|
i
{\displaystyle \sigma _{i}=\sum _{x}\sigma _{x|i}}
.
Quantum contextuality
Quantum contextuality refers to the idea that the outcome of a quantum measurement cannot be explained by any pre-existing value that is independent of the measurement context. In other words, the result of a quantum measurement depends on which other measurements are being performed simultaneously. This concept challenges classical notions of reality, where it is assumed that properties of a system exist independently of measurement.
It was proved that any set of compatible POVMs leads to preparation noncontextual correlations for all input quantum states. Conversely, the existence of a preparation noncontextual model for all input states implies compatibility of the involved POVMs.
See also
References
Further reading
Alter, Orly; Yamamoto, Yoshihisa (2001). Quantum Measurement of a Single System. New York: Wiley. doi:10.1002/9783527617128. ISBN 978-0-471-28308-9.
Busch, Paul; Lahti, Pekka; Pellonpää, Juha-Pekka; Ylinen, Kari (2016). Quantum measurement. Vol. 23. Springer. pp. 261–262. Bibcode:2016qume.book.....B. doi:10.1007/978-3-319-43389-9. ISBN 978-3-319-43387-5.
Braginsky, Vladimir B.; Khalili, Farid Ya. (1992). Quantum Measurement. Cambridge University Press. ISBN 978-0-521-41928-4.
Greenstein, George S.; Zajonc, Arthur G. (2006). The Quantum Challenge: Modern Research On The Foundations Of Quantum Mechanics (2nd ed.). Jones & Bartlett Learning. ISBN 978-0-7637-2470-2.
Jordan, Andrew N.; Siddiqi, Irfan A. (2024). Quantum Measurement: Theory and Practice. Cambridge University Press. ISBN 978-1-009-10006-9.
Wheeler, John A.; Zurek, Wojciech H., eds. (1983). Quantum Theory and Measurement. Princeton University Press. ISBN 978-0-691-08316-2.