Number Systems with Simplicity Hierarchies: A ... - Ohio University
Number Systems with Simplicity Hierarchies: A ... - Ohio University
Number Systems with Simplicity Hierarchies: A ... - Ohio University
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
THE JOURNAL OF SYMBOLIC LOGIC<br />
Volume 66, <strong>Number</strong> 3, Sept. 2001<br />
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES:<br />
A GENERALIZATION OF CONWAY'S THEORY OF SURREAL<br />
NUMBERS<br />
PHILIP EHRLICH<br />
Introduction. In his monograph On <strong>Number</strong>s and Games [7], J. H. Conway introduced<br />
a real-closed field containing the reals and the ordinals as well as a great<br />
many other numbers including -co, co/2, 1/co, wS and co - 7T to name only a few.<br />
Indeed, this particular real-closed field, which Conway calls No, is so remarkably<br />
inclusive that, subject to the proviso that numbers-construed here as members of<br />
ordered "number" fields be individually definable in terms of sets of von Neumann-<br />
Bernays-Godel set theory <strong>with</strong> Global Choice, henceforth NBG [cf. 21, Ch. 4], it may<br />
be said to contain "All <strong>Number</strong>s Great and Small." In this respect, No bears<br />
much the same relation to ordered fields that the system of real numbers bears<br />
to Archimedean ordered fields. This can be made precise by saying that whereas<br />
the ordered field of reals is (up to isomorphism) the unique homogeneous universal<br />
Archimedean orderedfield, No is (up to isomorphism) the unique homogeneous<br />
universal orderedfield [14]; also see [10], [12], [13]. 1<br />
However, in addition to its distinguished structure as an ordered field, No has<br />
a rich hierarchical structure that (implicitly) emerges from the recursive clauses in<br />
terms of which it is defined. This algebraico-tree-theoretic structure, or simplicity<br />
hierarchy, as we have called it [15], depends upon No's (implicit) structure as a<br />
lexicographically ordered binary tree and arises from the fact that the sums and<br />
products of any two members of the tree are the simplest possible elements of the tree<br />
consistent <strong>with</strong> No's structure as an ordered group and an ordered field, respectively,<br />
Received September 2, 1998; revised March 29, 2000.<br />
Portions of this paper were presented at the 1998 ASL Spring Meeting in Los Angeles, the 1998<br />
ASL Summer Meeting in Prague, the 1999 Mal'tsev Meeting in Novosibirsk, and the <strong>University</strong> of<br />
Notre Dame Mathematical Logic Seminar. Research supported by the National Science Foundation<br />
(Scholars Award # SBR 9602154) and <strong>Ohio</strong> <strong>University</strong>. The author wishes to express his thanks to these<br />
institutions for their support and to Lou van den Dries and the referee for suggesting helpful ways for<br />
streamlining and improving the exposition.<br />
1 For the purpose of this paper, an ordered field (Archimedean ordered field) A is said to be homogeneous<br />
universal if it is universal every ordered field (Archimedean ordered field) whose universe is<br />
a set or a proper class of NBG can be embedded in A and it is homogeneous every isomorphism<br />
between subfields of A whose universes are sets can be extended to an automorphism of A. Since model<br />
theorists frequently use the above italicized terms in more general senses, in the model-theoretic settings<br />
of [10], [12], [13] and [14] the terms absolutely homogeneous universal, absolutely universal, and absolutely<br />
homogeneous were respectively employed in their steads.<br />
(?) 2001, Association for Symbolic Logic<br />
0022-4812/01/6603-001 5/$3.80<br />
1231
1232 PHILIP EHRLICH<br />
it being understood that x is simpler than y just in case x is a predecessor of y in<br />
the tree.<br />
In [15], the just-described simplicity hierarchy was brought to the fore2 and made<br />
part of an algebraico-tree-theoretic definition of No. In the pages that follow, we<br />
introduce a novel class of structures whose properties generalize those of No so<br />
construed and explore some of the relations that exist between No and this more<br />
general class of s-hierarchical ordered structures as we call them. In ? 1 we define a<br />
number of types of s-hierarchical ordered structures groups, fields, vector spacesas<br />
well as a corresponding type of s-hierarchical mapping, identify No as a complete<br />
s-hierarchical ordered group (s-hierarchical ordered field; s-hierarchical ordered<br />
vector space), and show that there is one and only one s-hierarchical mapping of<br />
an s-hierarchical ordered structure into No (or any complete s-hierarchical ordered<br />
structure, more generally). These mappings are found to be embeddings of their<br />
respective kinds whose images are initial subtrees of No, and this together <strong>with</strong><br />
the completeness of No enables us to characterize No, up to isomorphism, as<br />
the unique complete as well as the unique nonextensible and the unique universal,<br />
s-hierarchical ordered group (s-hierarchical ordered field; s-hierarchical ordered<br />
vector space). Following this, in ?2 and ?4 we turn our attention to uncovering<br />
the spectrum of s-hierarchical ordered structures. Given the nature of No alluded<br />
to above, this reduces to revealing the spectrum of initial substructures of No, i.e.,<br />
the subgroups, subfields, subspaces of No (considered as an s-hierarchical ordered<br />
algebraic structure) that are initial subtrees of No. Included among our findings are<br />
the following two results that were originally stated as conjectures by the author at<br />
the AMS special session on Surreal <strong>Number</strong>s in January of 1989.<br />
I. Every divisible ordered abelian group is isomorphic to an initial subgroup of No.<br />
II. Every real-closed orderedfield is isomorphic to an initial subfield of No.<br />
In ?3, as part of the groundwork for the proof of II, we provide novel proofs that each<br />
surreal number x can be represented by a unique formal sum which may be treated<br />
as a canonicalproper name of x and the closely related fact that No considered as<br />
an ordered field is isomorphic to the formal power series field JR (No)On.<br />
In ?5, we generalize and amplify Conway's theories of ordinals and omnific integers<br />
by showing that every nontrivial s-hierarchical ordered group (s-hierarchical<br />
ordered field) A contains a cofinal, canonical subsemigroup (subsemiring) On(A)<br />
the ordinalpart of A which in turn is contained in a discrete, canonical subgroup<br />
2More specifically, in [15], following a suggestion of Conway, the just-described simplicity hierarchy<br />
was brought to the fore and thereby freed from the ambiguity that befalls it in Conway's own treatment<br />
in [7]. The ambiguity arises because remarks made in [7] make it possible (if not more likely) to interpret<br />
"x is simpler than y" as x has an earlier birthday than y, rather than in the manner specified above as<br />
Conway had intended (Private Conversation: see [15, pp. 257-258: note 1]). For some purposes the<br />
ambiguity is of little consequence. For example, one may show that No has precisely one automorphism<br />
that preserves simplicity regardless of which one of the above two interpretations of the simpler than<br />
relation is adopted [4], [5], [15]. On the other hand, as the succeeding pages only begin to show, from<br />
the standpoint of exploring the internal structure of No, it the tree-theoretic interpretation that is the<br />
more revealing. For treatments of No in which "x is simpler than y" is interpreted as x has an earlier<br />
birthday than y, see, for example, [4], [5], and [6].
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1233<br />
(subring) Oz(A) of A the omnific integer part of A-in which for each x E A there<br />
is a z E Oz(A) such that z < x < z + e where e which is the simplest positive<br />
element of A is the least positive element of Oz(A). When A is a substructure<br />
of No, e is the surreal number 1 and the members of On(A) and Oz(A) are called<br />
ordinals and omnific integers, respectively Finally, in ?6 we specify directions for<br />
further research.<br />
Throughout the paper the underlying set theory is assumed to be NBG and as<br />
such by class we mean set or proper class, the latter of which, in virtue of the Axiom<br />
of Global Choice, always has the "cardinality" of the class On of all ordinals.<br />
Moreover, since the usual definition of a sequence is not a legitimate conception<br />
in NBG when proper classes are involved, we follow the standard practice of understanding<br />
by a "structure" whose universe A is a proper class and whose finitary<br />
relations R,, 0 < a < fi E On, on A are classes (which may be operations or<br />
distinguished elements treated as special relations) the class (A x {0}) U R where<br />
R = Uo
1234 PHILIP EHRLICH<br />
DEFINITION 1. A lexicographically ordered binary tree is an ordered binary tree<br />
(A,
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1235<br />
either z E Rs(x) 0 Ls(y) or z E Rs(y) 0 Ls(x) depending upon whether x
1236 PHILIP EHRLICH<br />
DEFINITION 2. (A, +,
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1237<br />
PROPOSITION 1. (i) 0 is the simplest element of (as well as the unique root in) an<br />
s-hierarchical ordered group A, i.e., 0 ={0 o 0}A; (ii) for each element x of an s-<br />
hierarchical ordered group A, -x = {-xR I XL}A; (iii) 1 is the simplest positive<br />
element of an s-hierarchical orderedfield A, i.e., 1 ={0 I}A.<br />
1.3. Complete s-hierarchical ordered algebraic structures. A binary tree will be<br />
said to be full if every element has two immediate successors and every chain of<br />
infinite limit length < On has an immediate successor. As Theorem 4 below makes<br />
clear, in the case of lexicographically ordered binary trees the concept of a full<br />
binary tree is intimately related to the following conception.<br />
DEFINITION 6. A lexicographically ordered binary tree (A,
1238 PHILIP EHRLICH<br />
XL, XR, YL, and yRrange over the members of Ls(x), Rs(x), Ls(Y), and Rs(Y), respectively.<br />
DEFINITION OF X + Y.<br />
DEFINITION OF-X.4<br />
DEFINITION OF Xy.<br />
x + y {X +LYy,X+ YL XR+ yX+yR}.<br />
-X = {xR I -X L}<br />
xy {xLy + xyLXLyLxRy + xyR R R y<br />
x y + xy -x YRXY + xyL _xRyL}.<br />
1.4. s-Hierarchical embeddings and complete s-hierarchical ordered algebraic structures.<br />
The present subsection culminates in a theorem that indicates that (No,
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1239<br />
PA' (f (x)). Conversely, if f is a mapping of the said kind and x E A, then<br />
f (Ls(x))
1240 PHILIP EHRLICH<br />
existence of f (x) so defined and thereby, in virtue of Lemma 2, the universal nature<br />
of A if A is an s-hierarchical ordered group or an s-hierarchical ordered field. For<br />
the case where A is an s-hierarchical ordered vector space, the universality of A is<br />
established by appealing to Lemma 2 in conjunction <strong>with</strong> the following result whose<br />
simple proof is left to the reader:<br />
PROPOSITION. Let K be an s-hierarchical orderedfield, f be the unique s-hierarchical<br />
mappingfrom K to No, and No be the multiplication in the s-hierarchical orderedfield<br />
No. (No, +,
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1241<br />
?2. The spectrum of s-hierarchical ordered vector spaces.<br />
2.1. s-Hierarchical ordered vector spaces. There are a great many ordered vector<br />
spaces that are not isomorphic to initial subspaces of No. The present subsection<br />
culminates <strong>with</strong> a result specifying necessary and sufficient conditions for those<br />
which are. To prepare the way for our proof of the said theorem, we first establish<br />
two preliminary results, the proof of the first of which makes use of the following<br />
ELEMENTARY OBSERVATION. If A is an initial subtree of a lexicographically ordered<br />
binary tree A' and x ={L R}A', where (L, R) is a partition of A, then A U {x} is<br />
an initial subtree of A'.<br />
PROOF. If a C prA (x), then Ls(a) C Ls(x) < {x} < Rs(x) D Rs(a). But x -<br />
{L I R}A; and so, by Theorem 2, there is a z c L U R such that z > each x' c LS W<br />
and z < each x' c Rs(x), which <strong>with</strong> Theorem 2 implies a
1242 PHILIP EHRLICH<br />
obtain x = rb + a<br />
{L(x) I R(x) } where<br />
L(x) = rLb + (rbL -r LbL + a) rb + a , rR b + (rbR -r RbR + a) },<br />
R(x) {rLb + (rbR - rLbR + a) rb + aR, rRb + (rbR -r RbL + a) },<br />
and rL, rR, bL, bR, aL and aR are typical members of Ls(I.), Rs(I.), Ls(b), Rs(b), Ls(a)<br />
and Rs(a), respectively But since A is a vector space over K, (rbL - rLbL + a),<br />
(rbR- rRbR + a), (rbR- rLbR + a) and (rbR - rRbL + a) are all in A, since<br />
r, rL, rR c K and a, bL, bR C A; and consequently, rLb + (rbL -rLbL + a),<br />
r b+ (rbRrRbR + a), rL b + (rbR -rLbR + a) and rRb + (rbR rRbL a)<br />
are all in A< since each such sum is in one or another A(YE) where y < f,. Moreover,<br />
sums of the forms rb + a L and rb + a R are in A< since each such sum is in one or<br />
another A(p,,) where a < v; and as such L(x) U R(x) C A
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1243<br />
ordered class A we mean a pair of nonempty subclasses X and Y of A where X has<br />
no greatest member, Y has no least member, X U Y = A, and X < Y.<br />
DEFINITION 10. Let R = D U {{X I Y}: (X, Y) is a Dedekind gap in ED} where<br />
D = {x c No: PNo (x) < co}; as such, R is the set of all members of Uf
1244 PHILIP EHRLICH<br />
on these structures by identifying them <strong>with</strong> isomorphic copies of particular formal<br />
power series fields. One of the central components in our proof of the identification<br />
is the relation that exists between the formal power series fields in question and<br />
the complete s-hierarchical ordered field (No, +H, *H
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1245<br />
Lead (A) nA' ifA' is an initial subgroup of A; (vii) ifA' is an initial subgroup of A, then<br />
(Lead (A'),
1246 PHILIP EHRLICH<br />
DEFINITION 12. For each y c No, ogy = {, noLs(Y) | oy) }NO<br />
THEOREM 12. coy = ID-1 (y) = t0,n(L | 21nw(R} for each y C No and each<br />
ordered pair (L, R) such that y = {L I R}INo.<br />
Lsl(cl') = o{WX: x C Ls(y)} and Rsl(,Y) = fox x C Rs(y)}.<br />
And so, since (D-1 is s-hierarchical,<br />
PROOF. Plainly, coo = ID7o (0) = { 0}. Now suppose y c No - {0} and cox<br />
iD-1 (x) for all x c Ls(y) U Rs(y). Then, since by Theorems 10(v) and 11<br />
(D~~ 1y ~-1<br />
ONo ()<br />
~D'():xCL~~<br />
{No (X :t~)}|{No<br />
(- x<br />
()XC s(y)}}<br />
we have<br />
LeadNo<br />
-i<br />
CL<br />
(y) - { CD |Rs y) }Lead(No)<br />
And so, by Theorem 10 (iii) and the fact that {0} U {nwoLs(y) } is cofinal <strong>with</strong> L*y)<br />
and { 1oRs(y) } is coinitial <strong>with</strong> Rs*l,<br />
we have<br />
(D (y) {0, ncowLs(Y) |1 CoRs(y) } cy<br />
But, by Theorem 2, if y = {L IR}N, then L is cofinal <strong>with</strong> Ls(Y) and R is coinitial<br />
<strong>with</strong> Rs(Y), from which it follows that { 0, nL} is cofinal <strong>with</strong> {0, nwLs(y) } and<br />
{?w1n~)R} is coinitial <strong>with</strong><br />
{<br />
{0, nw L} < {WCoY} < { (OR R<br />
_)R5'(Y)}. But then coy = {0,nwOL (OR}N, since<br />
Since No is a vector space over R and every member of No - {0} is Archimedean<br />
equivalent to exactly one member of {coy: y C No} C No+, a familiar classical<br />
argument (cf. [19, Lemma 2.4]) leads to the following<br />
SIMPLE RESULT. If x c No - {0}, then there is a unique r C R - {O} and a unique<br />
y c No for which Ix - rcoy
Z<br />
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1247<br />
PROOF. If x c No - {0}, there is a unique ro c R - {0} and a unique yo C No for<br />
which Ix - (s0 + rocoy)I
1248 PHILIP EHRLICH<br />
simple consequence of Theorem 13 together <strong>with</strong> Definition 13 makes clear, the use<br />
of the summation and product signs are as appropriate as they are revealing.<br />
BASIC OBSERVATION. coy . r = rowy (i.e., = r No coY) for each y C No and each<br />
r c R - {fO}, andfor each nonempty descending sequence (yO4f
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1249<br />
by their respective Conway names is inspired by a result due to Conway. Here<br />
and henceforth, for the sake of descriptive simplicity, we permit the concept of the<br />
Conway name of a surreal number to be naturally expanded so as to allow for the<br />
insertion and deletion of "dummy" terms <strong>with</strong> zeros for coefficients.<br />
THEOREM 16. (No,+H, H
1250 PHILIP EHRLICH<br />
x+HYis an approximation of (x +H y)+Hw. But then x +Hy
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1251<br />
PROOF. Let A be an initial subfield of No. To prove the "only if" portion of<br />
the theorem it suffices to show A is an approximation complete subfield of No<br />
(see Definition 14) where G = {y: coy E Lead (A)} is an initial subgroup of No<br />
for then plainly { E< rtY- c R (G)0,1 : w
1252 PHILIP EHRLICH<br />
({bu} U U6
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1253<br />
PROOF. Let A be an ordered field whose universe may be a proper class, K G, ,
1254 PHILIP EHRLICH<br />
The nature of the system of omnific integers of a nontrivial initial subgroup of<br />
No is greatly clarified by the following theorem, the nonroutine portion of which<br />
follows immediately from Conway's proof of Theorem 31 of [7].<br />
THEOREM 20. If A is a nontrivial initial subgroup of No, then Oz (A) is the subclass<br />
of all x c A such that<br />
x<br />
Zcow . ac,<br />
a
apt .<br />
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1255<br />
together <strong>with</strong> other familiar theorems concerning ordinals leads to the following<br />
well-known theorem due to Cantor: Every ordinal a < On has a unique Cantor<br />
Normal Form, i.e., a unique representation of the form<br />
E09c<br />
*oa c aa<br />
a
1256 PHILIP EHRLICH<br />
It is evident that every s-hierarchical ordered group is a-Archimedean for some<br />
a. Also evident is<br />
THEOREM 24. Let A be a nontrivial s-hierarchical ordered group and h:<br />
On (A) -> On be an s-hierarchical mapping. A is a-Archimedean if and only iffor<br />
each a E A there is an q Ec On (A) such that -a < a < q where h (q) < a. Moreover,<br />
if A is an s-hierarchical ordered field, then A is a-Archimedean if and only iffor all<br />
a,b E A+ where a > b there is an q < a such that qb > a; furthermore, A is<br />
Archimedean if and only if A is co-Archimedean.<br />
By appealing to Theorem 23 together <strong>with</strong> the definitions of +H and H as<br />
stated in Theorem 16, it is easy to see that an initial sequence {a,: a < Pl < On}<br />
of ordinals is closed under addition (multiplication) if and only if P = w0 for<br />
some ordinal (indecomposable ordinal) p < On. Accordingly, if for each ordinal<br />
(indecomposable ordinal) p < On we let SG (aLP) (SR (ai)) be the ordered class<br />
of all ordinals less than w0 <strong>with</strong> sums and products defined as in Theorem 16, then<br />
the SG (aiP)s (SR (0i)s) so defined are the only subsemigroups (subsemirings) of<br />
No that consist of initial sequences of ordinals of No.8 And this together <strong>with</strong><br />
Theorems 21, 23 and 24 and the fact that for each s-hierarchical ordered structure<br />
A there is one and only one s-hierarchical mapping hA: A -* No, hA being an<br />
s-hierarchical embedding, we have<br />
THEOREM 25. Every nontrivial s-hierarchical ordered group (s-hierarchical ordered<br />
field) A is w0-Archimedean for some nonzero ordinal (nonzero indecomposable ordinal)<br />
p < On; and if A is an w0-Archimedean s-hierarchical ordered group (shierarchical<br />
ordered field) then A contains a cofinal, canonical subsemigroup (subsemiring)<br />
On (A) called the ordinal part of A that is isomorphic to SG (00)<br />
(SR (0P)) and which consists of all a Ec A such that a = {L I 0}A for some<br />
L C A; On (A) in turn is contained in a discrete, canonical subgroup (subring)<br />
Oz (A) of A called the omnific integer part of A consisting of all x E A such that<br />
x = {x - e I x + e}A where e is the simplest positive element of A; in addition, for<br />
each x E A there is a z E Oz (A) such that z < x < z + e.9<br />
?6. Concluding remarks. While laying the groundwork, the present paper merely<br />
takes a first step in the development of a general theory of s-hierarchical ordered algebraic<br />
structures. We conclude by mentioning a few directions for further research<br />
that the author (and, hopefully, others) will turn to in the not too distant future. To<br />
begin <strong>with</strong>, in addition to shedding further light on the nature of the s-hierarchical<br />
8As the reader will notice, for each ordinal A, SG (co9') is essentially the ordered semigroup of all<br />
ordinals (written in Cantor normal form) less than cow <strong>with</strong> order defined by first differences and<br />
addition defined a la Hessenberg [cf. 27], and for each indecomposable ordinal A, SR (cow') is essentially<br />
the ordered semiring that results from supplementing SG (co9') <strong>with</strong> a Hessenberg product [cf. 27].<br />
Unlike the SG (co9')s, thus construed, which have been discussed in the literature [26], the SR (co9')s<br />
appear to be new (except for the cases where the cow s are regular initial numbers [27]).<br />
9The omnific integer part of an s-hierarchical ordered field is an integer part in the sense of Mourgues<br />
and Ressayre [23], i.e., a discrete subring I of an ordered field A in which for each x E A there is a z E I<br />
such that z < x < z + 1. However, while every s-hierarchical ordered field has an integer part, there are<br />
ordered fields having integer parts that do not admit relational extensions to an s-hierarchical ordered<br />
field each Hahn field JR (G) where G does not admit a relational extension to an s-hierarchical ordered<br />
group is an example of such a system.
NUMBER SYSTEMS WITH SIMPLICITY HIERARCHIES 1257<br />
ordered algebraic structures we have investigated thus far, it remains to investigate<br />
their natural generalizations the s-hierarchical ordered algebraic structures that<br />
are defined by replacing the references to ordered groups, ordered fields and ordered<br />
K-vector spaces in Definitions 2-4 <strong>with</strong> references to ordered semigroups, ordered<br />
rings, and ordered K-modules, respectively. While some of our results such as<br />
Lemma 2 and Theorem 5 as well as their proofs carry over to these more general<br />
classes of s-hierarchical ordered structures, their isolation and analysis remain wide<br />
open. Similar remarks apply to s-hierarchical analogs of ordered semirings and ordered<br />
K-semimodules, which also deserve attention. In addition, unlike the ordered<br />
fields of reals and surreals which admit (up to isomorphism) only one relational<br />
extension to an s-hierarchical structure, many s-hierarchable structures admit relational<br />
extensions to nonisomorphic s-hierarchical structures and an analysis of the<br />
classes of such alternative extensions also remains to be provided. Finally, while we<br />
have devoted the bulk of our attention to s-hierarchical ordered structures per se,<br />
detailed analyses of the hierarchical relations that exist between their respective elements<br />
also remain to be provided including a detailed analysis of the tree-theoretic<br />
relations that exist between surreal numbers denoted by their respective Conway<br />
names.<br />
REFERENCES<br />
[1] N. ALLING, On the existence of real closedfields that are d -sets of power N, Transactions of the<br />
American Mathematical Society, vol. 103 (1962), pp. 341-352.<br />
[2] , Conway'sfield of surreal numbers, Transactions of the American Mathematical Society,<br />
vol. 287 (1985), pp. 365-386.<br />
[3] , Foundations of analysis over surreal numberfields, North-Holland Publishing Co., Amsterdam,<br />
1987.<br />
[4] N. ALLING and P. EHRLICH, An alternative construction of Conway's surreal numbers, La Societt<br />
royale du Canada. Comptes Rendus Mathimatiques de l'Academie des Sciences. (Mathematical Reports)<br />
VIII, vol. 8 (1986), no. 4, pp. 241-246.<br />
[5] , An abstract characterization of a full class of surreal numbers, La Societt royale du<br />
Canada. Comptes Rendus Mathimatiques de l'Academie des Sciences (Mathematical Reports) VIII,<br />
(1986), pp. 303-308.<br />
[6] , Foundations of analysis over surreal numberfields, North-Holland Publishing Co., Amsterdam,<br />
1987, Sections 4.02 and 4.03.<br />
[7] J. H. CONWAY, On numbers and games, Academic Press, 1976.<br />
[8] H. G. DALES and W H. WOODIN, Super-real orderedfields, Clarendon Press, Oxford, 1996.<br />
[9] F DRAKE, Set theory; an introduction to large cardinals, North-Holland Publishing Co., Amsterdam,<br />
1974.<br />
[10] P. EHRLICH, The absolute arithmetic and geometric continua, PSA 1986 (Lansing, MI) (Arthur<br />
Fine and Peter Machamer, editors), vol. 2, Philosophy of Science Association, 1987, pp. 237-247.<br />
[11] I An alternative construction of Conway's orderedfield No, Algebra Universalis, vol. 25<br />
(1988), pp. 7-16, errata, Algebra Universalis, vol. 25, (1988), page 233.<br />
[12] ' Absolutely saturated models, Fundamenta Mathematicae, vol. 133 (1989), pp. 39-46.<br />
[13] ' Universally extending continua, Abstracts of papers presented vol. 10, p. 15, American<br />
Mathematical Society, January 1989.<br />
[14] , Universally extending arithmetic continua, Le labyrinth du continu: Colloque de Cerisy<br />
(H. Sinaceur and J. M. Salanskis, editors), Springer-Verlag, Paris, France, 1992, pp. 168-177.<br />
[15] ' All numbers great and small, Real numbers, generalizations of the reals, and theories of<br />
continua, Kluwer Academic Publishers, 1994, pp. 239-258.<br />
[16] H. GAIFMAN, Operations on relational structures, functors and classes. I, Proceedings of the Tarski<br />
symposium (Leon Henkin et al., editor), American Mathematical Society, Providence, Rhode Island,
1258 PHILIP EHRLICH<br />
1974, pp.21-39.<br />
[17] H. GONSHOR, An,-introduction to the theory of surreal numbers, Cambridge <strong>University</strong> Press,<br />
1986.<br />
[18] H. HAHN, Uber die nichtarchimedischen Grdssensysteme, Sitzungsberichte der Kaiserlichen<br />
Akademie der Wissenschaften, Mathematisch - Naturwissenschaftliche Klasse (Wien), vol. 116, 1907,<br />
(Abteilung Ila), pp. 601-655.<br />
[19] M. HAUSNER and J. G. WENDEL, Ordered vector spaces, Proceedings of the American Mathematical<br />
Society, vol. 3, 1952, pp. 977-982.<br />
[20] A. LEVI, Basic set theory, Springer-Verlag, Berlin, 1979.<br />
[21] E. MENDELSON, Introduction to mathematical logic, third ed., Wadsworth & Brooks/Cole Advanced<br />
Books & Software, Monterey, California, 1987.<br />
[22] M. H. MOURGUES and J.-P. RESSAYRE, A transfinite version of Puiseux's theorem <strong>with</strong> applications<br />
to real closedfields, Logic Colloquium'90, ASL Summer Meeting in Helsinki (J. Oikkonen and J. Vddnden,<br />
editors), Springer-Verlag, Berlin, 1991, pp. 250-258.<br />
[23] , Every real closedfield has an integer part, this JOURNAL, vol. 58 (1993), pp. 641-647.<br />
[24] S. PRIESS-CRAMPE, Angeordnete Strukturen, Gruppen, Kiirper, projektive Ebenen, Springer-Verlag,<br />
Berlin, 1983.<br />
[25] F. J. RAYNER, Algebraically closedfields analogous tofields of Puiseux series, The Journal of the<br />
London Mathematical Society, vol. 8 (1974), no. 2, pp. 504-506.<br />
[26] N. SANKARAN and R. VENKATARAMAN, A generalization of the ordered group of integers, Mathematishe<br />
Zeitschrift, vol. 79 (1962), pp. 21-3 1.<br />
[27] R. SIKORSKI, On an ordered algebraic field, Comptes rendus des stances de la classe 111, sciences<br />
mathimatiques et physiques, vol. 41, La Soci&t6 des Sciences et des Lettres de Varsovie, 1948, pp. 69-96.<br />
[28] P. UCSNAY, Wohlgeordnete Untermengen in totalgeordneten Gruppen mit einer Anwendung auf<br />
Potenzreihenkorper, Mathematishe Annalen, vol. 160 (1965), pp. 161-170.<br />
DEPARTMENT<br />
OHIO UNIVERSITY<br />
OF PHILOSOPHY<br />
ATHENS, OH 45701, USA<br />
E-mail: ehrlich~ohiou.edu