Effect Algebras

Origin. Foulis and Bennett introduced effect algebras (1994) to generalize quantum logic and probability. Abstracts the structure of quantum effects (positive operators ≤ I). Connects quantum mechanics, fuzzy logic, and generalized probability. Part of the "quantum structures" tradition alongside orthomodular lattices.

Models. Unsharp measurements and partial operations. Quantum logic (orthomodular lattices) models sharp, yes/no measurements. Effect algebras model unsharp measurements: effects can be partially true. The partial sum a ⊕ b represents "a then b" when they're compatible. Models generalized probability beyond classical and quantum.

Formalism.

Effect algebra: (E, 0, 1, ⊕) where:

E is a set of effects
0 (zero effect), 1 (unit effect)
⊕ : E × E → E is a partial binary operation

Axioms:

Commutativity: if a ⊕ b defined, then b ⊕ a defined and equal
Associativity: if a ⊕ b and (a ⊕ b) ⊕ c defined, then b ⊕ c and a ⊕ (b ⊕ c) defined and equal
Zero: a ⊕ 0 = a for all a
Orthocomplement: for each a, unique a' with a ⊕ a' = 1
Zero-one: if a ⊕ 1 defined, then a = 0

Derived notions:

a' = orthocomplement (unique)
a ≤ b iff ∃c: a ⊕ c = b
a ⊥ b iff a ⊕ b defined (orthogonal/compatible)

Examples:

[0,1] interval: a ⊕ b = a + b when a + b ≤ 1
Orthomodular lattices: a ⊕ b = a ∨ b when a ⊥ b
Hilbert space effects: {E : 0 ≤ E ≤ I}, a ⊕ b = a + b when a + b ≤ I

MV-effect algebras: Totality is not the condition — it is degenerate. Axiom 5 says a ⊕ 1 defined implies a = 0, so a total ⊕ collapses E to {0}. The correct condition: an effect algebra is an MV-effect algebra iff it is lattice-ordered and has the Riesz decomposition property. Equivalently (Bennett and Foulis): lattice-ordered with a ∧ b = 0 implying a ⊥ b. The partial ⊕ then extends to the total truncated MV-sum, which is the Łukasiewicz strong disjunction — hence the connection to Łukasiewicz logic.

States: A state is s: E → [0,1] with s(1) = 1 and s(a ⊕ b) = s(a) + s(b). States assign probabilities respecting structure.

Sharp vs unsharp: a is sharp if a ∧ a' = 0 (completely determined). Quantum logic = sharp elements; effect algebras include unsharp.

Symbols.

Symbol	Unicode	Name	Meaning
⊕	U+2295	Circled plus / Partial sum	Compatible combination, defined only on orthogonal pairs
'	U+0027	Apostrophe / Orthosupplement	The unique a' with a ⊕ a' = 1
0	—	Zero	Null effect
1	—	Unit	Certain effect
≤	U+2264	Less-than or equal / Order	a ≤ b iff a ⊕ c = b for some c
⊥	U+22A5	Up tack / Orthogonality	a ⊥ b iff a ⊕ b is defined
∧	U+2227	Logical and / Meet	Greatest lower bound, when it exists
∨	U+2228	Logical or / Join	Least upper bound, when it exists
∃	U+2203	There exists	Witness in the definition of ≤
×	U+00D7	Multiplication sign / Product	Domain of the partial operation ⊕
E	—	Effects	The carrier; on a Hilbert space, operators with 0 ≤ E ≤ I
I	—	Identity operator	Upper bound of the Hilbert-space effects
s	—	State	Probability measure with s(a ⊕ b) = s(a) + s(b)
↦	U+21A6	Maps to	Assignment defining a state
ρ	U+03C1	Rho / Density operator	Represents a state by Busch's theorem
Tr	—	Trace	Tr(ρa) is the probability of effect a
[0,1]	—	Unit interval	The standard MV-effect algebra

Metatheory. Effect algebras subsume the sharp and unsharp cases at once: an effect algebra is an orthoalgebra iff a ⊕ a is defined only for a = 0, a lattice-ordered orthoalgebra is exactly an orthomodular lattice, and an effect algebra is an MV-effect algebra iff it is lattice-ordered with the Riesz decomposition property, in which case ⊕ extends to Łukasiewicz truncated addition. Ravindran (1996) proved that every effect algebra with the Riesz decomposition property is the interval [0, u] in an interpolation group; without that property the representation fails, so not every effect algebra is an interval effect algebra. States are the generalized probability measures, and Busch's theorem (2003) is the effect-algebra strengthening of Gleason: every state on the effects of a Hilbert space is a ↦ Tr(ρa) for a density operator ρ, and unlike Gleason's theorem it holds in dimension 2 as well, because the unsharp effects supply constraints the projections do not. Greechie (1971) constructed orthomodular lattices with no states at all, so state spaces may be empty.

Applies to. Quantum foundations (unsharp measurements). Generalized probability theory. Fuzzy sets and fuzzy logic. Quantum computing (effect semantics). Foundations of statistical mechanics. Philosophical foundations of uncertainty.

Limitations. Because ⊕ is partial, effect algebras are partial algebras and not a variety, so the equational methods that apply to ortholattices and MV-algebras are unavailable and there is no equational completeness theorem to prove. The motivating model is badly behaved as an order: by Kadison's antilattice theorem the self-adjoint operators form a lattice only in the commutative case, so the Hilbert-space effects are not lattice-ordered and hence not an MV-effect algebra, and meets and joins of effects generally do not exist. The sharp elements need not form a sub-effect algebra, so the sharp/unsharp split is not a retraction. There is no implication connective and no deduction theorem, so effect algebras support probability rather than inference. Composite systems are a problem: the tensor product of effect algebras need not exist and can be degenerate. And by Greechie's construction there are quantum structures on which no state is defined at all.