Effects and Effect Algebras

A quantum effect is a bounded operator $E$ satisfying $0\leq E\leq I$. It represents one measurement event through the probability $\operatorname{Tr}(\rho E)$. Effects form an ordered interval with a partial addition law, called a Hilbert-space effect algebra. This structure extends projective yes–no questions without pretending that all quantum events obey Boolean set operations.

Helpful background. Positive operators supplies operator order and square roots; POVMs supplies the measurement interpretation.

Effects are exactly bounded probability tests

For an effect $E$ and density operator $\rho$,

$$0\leq\operatorname{Tr}(\rho E)\leq1.$$

Conversely, if a bounded operator gives a real value in $[0,1]$ in every normalized vector state, its quadratic values imply both $E\geq0$ and $I-E\geq0$. Thus positivity and the upper bound are not independent aesthetic requirements: they are exactly what makes all state probabilities valid.

The binary measurement associated with $E$ is $\{E,I-E\}$. The complement $E^\perp=I-E$ describes the alternative event in that binary experiment. It is unique. An effect specifies a probability rule, not a state update or a physical apparatus.

For example, $E=\tfrac12I$ gives probability one half in every state. Such an effect is a state-independent coin toss. It is a valid measurement event even though it reveals no information about the quantum input.

Partial addition and its meaning

Define

$$E\oplus F=E+F \quad\text{only when }E+F\leq I.$$

This sum models joining two distinct outcomes of one POVM. It is commutative and associative whenever the required sums exist, and every effect has the complement just described. If $E\oplus I$ exists, then $E=0$.

These operations satisfy the effect-algebra axioms. The order can be recovered from addition:

$$E\leq F \quad\Longleftrightarrow\quad F=E\oplus G \text{ for an effect }G.$$

Here $G=F-E$ is positive and bounded above by $I$. A state induces an additive probability assignment on every defined partial sum.

The condition $E+F\leq I$ must not be replaced by $EF=0$. Orthogonality is appropriate to disjoint projections. Nonprojective effects can have a defined sum while overlapping strongly as operators. For example, $\tfrac13I\oplus\tfrac13I=\tfrac23I$.

Nor does a formal sum of two effects from unrelated experiments automatically represent the union of their events. An operational interpretation needs a common measurement or a specified randomization.

Sharpness and the spectral interval

Within this effect algebra, an effect is sharp when it has no nonzero common lower bound with its complement. Precisely, if $0\leq G\leq E$ and $G\leq I-E$, then $G=0$.

Every projection $P$ is sharp. Indeed $G\leq P$ forces $G$ to be supported in $\operatorname{ran}P$, while $G\leq I-P$ forces support in its orthogonal complement.

Conversely, if $E$ is not a projection, functional calculus gives a nonzero positive operator

$$G=\min(E,I-E),$$

where the minimum is taken as a scalar function of the single operator $E$. Since $\min(x,1-x)$ vanishes only at $0,1$, $G\ne0$ whenever $E^2\ne E$. It is bounded above by both $E$ and $I-E$. Thus sharp effects are exactly projections.

This scalar functional-calculus minimum does not imply that arbitrary pairs of noncommuting effects possess an operator-order minimum. The effect interval is generally not a lattice. Consequently one cannot define all intersections and unions by copying classical probability notation.

Coexistence versus mutual exclusion

For two binary effects $E,F$, a four-outcome joint measurement exists exactly when there is an operator $G$ with

$$0\leq G\leq E,\quad G\leq F,\quad E+F-I\leq G.$$

The four effects are $G,E-G,F-G,I-E-F+G$. Their sum is $I$, and their marginals recover the two binary measurements. The operator $G$ represents their joint yes–yes event.

By comparison, $E\oplus F$ requires the special choice $G=0$, meaning the events are mutually exclusive. Joint measurability is therefore a broader notion than the existence of their partial sum. The full criterion and qubit examples appear in Joint Measurability.

Busch and collaborators develop the operational effect structure; Heinosaari and Ziman relate it to observables, compatibility, and state discrimination.

Exercises

1. Let $E=aI$ and $F=bI$, with $a,b\in[0,1]$. Find their partial-sum condition and determine whether the corresponding binary POVMs are jointly measurable.

Solution

The partial sum exists iff $a+b\leq1$. The binary measurements are jointly measurable for all $a,b$: choose $G=abI$, corresponding to two independent classical coins. Thus compatibility need not mean exclusivity.

2. Show that a projection $P$ and the effect $E=\tfrac12P$ are not equally sharp, despite having the same support.

Solution

If $P\ne0$, then $E^2=\tfrac14P\ne E$. Moreover $E\leq I-E$, so $E$ itself is a nonzero common lower bound. Support identifies possible states for an event; it does not determine its sharpness. This distinction matters for unsharp measurements.

References

• Busch, Paul, Pekka Lahti, Juha-Pekka Pellonpää, and Kari Ylinen, Quantum Measurement, Springer, 2016 — effects, sharpness, and measurement compatibility. Publisher.
• Heinosaari, Teiko, and Mário Ziman, The Mathematical Language of Quantum Theory: From Uncertainty to Entanglement, Cambridge University Press, 2012 — states, effects, and observables. Publisher.