05.08.2013 Views

Survey 1979: Equational Logic - Department of Mathematics ...

Survey 1979: Equational Logic - Department of Mathematics ...

Survey 1979: Equational Logic - Department of Mathematics ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

WALTER TAYLOR 25<br />

greatest example <strong>of</strong> an undecidable set is given by GOdel's incompleteness theorem<br />

[op. cit.], about which we will say no more. Here we will discuss decidability<br />

properties <strong>of</strong> equations. When we say that a property P <strong>of</strong> finite sets E <strong>of</strong> equations is<br />

(un)decidable, we mean that<br />

is (un)decidable.<br />

w0=<br />

We may first ask when a theory E itself is a decidable set <strong>of</strong> equations, or, as it is<br />

frequently put, "the word problem for free E-algebras is solvable." (See the discussion<br />

<strong>of</strong> word problems below.) There are two common methods for showing that E is a<br />

decidable equational theory, the first being to find a recursive procedure to convert<br />

every term o to a unique ("normal form") term o' with (o = o') E 2 and such that if o<br />

and r are distinct normal forms then (o = r) 2. (The decision procedure then<br />

reduces to comparison <strong>of</strong> normal forms - and conversely, a decision procedure for E<br />

obviously implies the existence <strong>of</strong> normal forms.) E.g. every group term reduces<br />

uniquely to either 1 or<br />

.nl .n2 n k<br />

Xl x --' Xik,<br />

where xil : xi2 :'" : Xik. Several <strong>of</strong> the best known equational theories are<br />

decidable, as may be seen similarly. See e.g. Margaris [284] for implicative<br />

semilattices, following work <strong>of</strong> McKay and Diego. Eq(2,+) is decidable (in fact its full<br />

first order theory is decidable, by Ehrenfeucht and Bfichi [70])- a simple method<br />

given by Selman and Zimbarg-Sobrinho [unpublished] is closely related to<br />

Karn<strong>of</strong>sky's identities 8.12 above. Martin [ 286] gave a decision procedure for (2,4-,.)<br />

with normal forms (cf. 9.20 above). Richardson [377] gave normal forms for<br />

(co, l,x+y,xy,xY) (cf. Problem 2 in õ8). Finally, we remark that the Birkh<strong>of</strong>f-Witt<br />

theorem yields a procedure for finding normal forms for (free) Lie algebras and rings,<br />

as observed by P. Hall [175]. See Bergman [43] for some detailed methods for<br />

finding normal forms, mainly in ring theory; also see [ 158] and [248].<br />

Notice that representing free algebras uniquely via terms (as we did for FK(X ) in<br />

õ2) really requires a normal form. Often a normal form is required for finding the<br />

cardinality <strong>of</strong> Fv(X), a topic we will come to in 14.5 below. For some other results<br />

related to normal forms, see Hule [194].

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!