Birkhoffconst PDF
Birkhoffconst PDF
Birkhoffconst PDF
RALPH FREESE
Let K be a set of algebras of the same similarity type and let V = H SPK
be the variety generated by K. First consider the case K = {A} consists of
a single algebra. We form the direct product of the algebras A(v), v ∈ AX .
This is an X–labelled algebra under the labelling
Y
x 7→ x̄ ∈ A(v) where x̄v = v(x)
v∈AX
Since av = bv ,
showing ā ηu b̄. This proves one direction. For the other, we need to assume
ηv ≤ ηu and show that the map v(x) 7→ u(x) extends to a homomorphism
ϕ : A(v) → A(u). We leave this as an exercise.
In the Lyndon groupoid we can use this thinning to reduced the number
of coordinates from 117, 649 down to 22, 526.
We can extend Lemma 3 to test if ηu ∧ ηv ≤ ηw . This will be the case
if and only if there is a homomorphsim respecting X from the subagebra
of A(u) × A(v) generated by {(u(x), v(x)) : x ∈ X} onto A(w). However,
this is usually too expensive since it involves looking at all 3 element subsets
of AX .
Exercises
1. Let N5 be the 5 element nonmodular lattice. Using the thinning
process described above show that when |X| = 3, the 125 coordinates
NX
of FN5 (X) ≤ N5 5 are thinned down to 18 coordinates. And find
the sizes of the N5 (v) that remain.
2. Show that when |X| = 4 the proceedure thins the 625 coordinates
down to 132 and find the sizes of the N5 (v) that remain.
3. Show that when |X| ≥ 5 the N5 (v)’s that remain after thinning all
have size 4 or 5.
4 RALPH FREESE
2. Birkhoff Basis
In testing if there is a homomorphism from A(v) onto A(u) respecting
the labelling we first check if the map ϕ : v(x) 7→ u(x) is well defined. (This
can fail if, say, v(x) = v(y) but u(x) 6= u(y).) If ϕ is well defined, we try
to extend it to all of A(v). We build A(v) ≤ A by closing the image of v
under the operations in the usual way and try to extend ϕ as we go.
So if a and b have been generated and f is a binary operation, we form
e = f (a, b). If this element is not in the closure we have so far, we add it
and we extend ϕ by setting ϕ(e) = f (ϕ(a), ϕ(b)). On the other hand if e is
already in the closure so ϕ(e) is already defined, we check if the existing ϕ(e)
is equal to f (ϕ(a), ϕ(b)). If this fails, we stop: there is no homomorphism
A(v) → A(u) respecting the labelling. Otherwise we continue.
Let V = V Q (A) and let F := FV (X) be the Birkhoff construction as the
subalgebra of v∈AX A(v) generated by {x̄ : x ∈ X}, x̄v = v(x). Using ideas
similar to the last paragraph, when closing {x̄ : x ∈ X} we can associate to
each element a of the closure a term ta in the variables X such that tF a = a.
Namely the term of x̄ is x and if c first appears as f (a, b) then tc = f (ta , tb ).
Now suppose g is an r-ary operation symbol, a1 , . . . , ar ∈ F and b =
g F (a1 , . . . , ar ). Then the equation
(2) tb ≈ g(ta1 , . . . , tar )
is true in A. Indeed, a substitution of the variables X into A is an element
v ∈ AX . If x1 , . . . xk are the variables occuring in the equation, then by the
way the terms ta were defined, the relation
tF F F F
b (x1 , . . . , xk ) = g (ta1 (x1 , . . . , xk ), . . . , tar (x1 , . . . , xk ))
holds in F. Now just apply πv to show that (2) holds in A under this
substitution.
The set of all equations of the form (2) for all basic operations g and
all r-tuples a1 , . . . , ar , r the arity of g, is called the Birkhoff basis for the
equations of A in the variables X. Recall that we have associated with each
element a of F a term ta ) with variables in X. One can use the Birkhoff
basis to transform an arbitrary term in X, t(x1 , . . . , xk ), xi ∈ X into the
term ta , where a = tF (x1 , . . . , xk ). Then, if s and t are terms in X, we can
decide if s ≈ t holds in V by transforming both s and t as above and seeing
if the results are equal. The details are left as an exercise.
By Corollary 2 if A is finite then the Birkhoff basis is finite. So we have
the following corollary.
Corollary 5. If A is a finite algebra, then the k-variable equations of A
are finitely based, for each finite k.
As we mentioned before, there are finite algebras with only finitely many
basic operations that do not have a finite equational basis for their identities.
Finally we mention that all of the above could be done for a set K of
algebras in place of the single algebra A.
6 RALPH FREESE
References
[1] J. Berman and B. Wolk, Free lattices in some small varieties, Algebra Universalis 10
(1980), 269–289.
[2] Joel Berman, The structure of free algebras, Structural theory of automata, semigroups,
and universal algebra, NATO Sci. Ser. II Math. Phys. Chem., vol. 207, Springer,
Dordrecht, 2005, pp. 47–76.
[3] G. Birkhoff, On the structure of abstract algebras, Proc. Cambridge Phil. Soc. 31
(1935), 433–454.
[4] S. Burris and H. P. Sankappanavar, A course in universal algebra, Springer-Verlag,
New York, 1981.
[5] Ralph Freese, Computing congruences efficiently, Algebra Uni-
versalis 59 (2009), 337–343, Online manuscript available at:
http://www.math.hawaii.edu/∼ralph/papers.html.