What Is Forcing

?
W H A T
I S . . .
Forcing?
Thomas Jech
What is forcing? Forcing is a remarkably powerful

technique for the construction of models of set
theory. It was invented in 1963 by Paul Cohen1,
who used it to prove the independence of the
Continuum Hypothesis. He constructed a model
of set theory in which the Continuum Hypothesis
(CH) fails, thus showing that CH is not provable
from the axioms of set theory.
What is the Continuum Hypothesis? In 1873
Georg Cantor proved that the continuum is uncountable: that there exists no mapping of the set
N of all integers onto the set R of all real numbers.
Since R contains N, we have 20 > 0 , where 20
and 0 are the cardinalities of R and N, respectively. A question arises whether 20 is equal to
the cardinal 1 , the immediate successor of 0 .
Cantors conjecture that 20 = 1 is the celebrated continuum hypothesis made famous by David
Hilbert, who put it on the top of his list of major
open problems in the year 1900. Cohens solution
of Cantors problem does not prove 20 = 1 , nor
does it prove 20 1 . The answer is that CH is
undecidable.
What does it mean that a conjecture is undecidable? In mathematics, the accepted standard of
establishing truth is to give a proof of a theorem.
Thus one verifies a given conjecture by finding a
proof of the conjecture, or one refutes it by finding
a proof of its negation. But it is not necessary that
every conjecture can be so decided: it may be the
Thomas Jech is professor emeritus of mathematics at the
Pennsylvania State University, currently at the Institute
of Mathematics of the Czech Academy of Sciences. His
email address is [email protected].
1
Editors Note: Paul Cohen died in March 2008, and an
article containing reminiscences about him is in preparation for an upcoming issue of the Notices.
692
case that there exists no proof of the conjecture

and no proof of its negation.
To make this vague discussion more precise we
will first elaborate on the concepts of theorem and
proof.
What are theorems and proofs? It is a useful fact that every mathematical statement can
be expressed in the language of set theory. All
mathematical objects can be regarded as sets, and
relations between them can be reduced to expressions that use only the relation . It is not essential
how it is done, but it can be done: For instance,
integers are certain finite sets, rational numbers
are pairs of integers, real numbers are identified
with Dedekind cuts in the rationals, functions
are some sets of pairs, etc. Moreover all selfevident truths used in mathematical proofs can
be formally derived from the axioms of set theory.
The accepted system of axioms of set theory is
ZFC, the Zermelo-Fraenkel axioms plus the axiom
of choice. As a consequence every mathematical
theorem can be formulated and proved from the
axioms of ZFC.
When we consider a well formulated mathematical statement (say, the Riemann Hypothesis)
there is a priori no guarantee that there exists a
proof of the statement or a proof of its negation.
Does ZFC decide every statement? In other words,
is ZFC complete? It turns out that not only is
ZFC not complete, but it cannot be replaced by a
complete system of axioms. This was Gdels 1931
discovery known as the Incompleteness Theorem.
What is Gdels Incompleteness Theorem? In
short, every system of axioms that is (i) recursive
and (ii) sufficiently expressive is incomplete. Sufficiently expressive means that it includes the
Notices of the AMS
Volume 55, Number 6
self-evident truths about integers, and recursive means roughly that a computer program can
decide whether a statement is an axiom or not.
(ZFC is one such system, Peanos system of axioms
for arithmetic is another such system, etc.)
It is of course one thing to know that, by
Gdels theorem, undecidable statements exist,
and another to show that a particular conjecture
is undecidable. One way to show that some statement is unprovable from given axioms is to find a
model.
What is a model ? A model of a theory interprets
the language of the theory in such a way that the
axioms of the theory are true in the model. Then
all theorems of the theory are true in the model,
and if a given statement is false in the model, then
it cannot be proved from the axioms of the theory.
A well known example is a model of non-Euclidean
geometry.
A model of set theory is a collection M of sets
with the property that the axioms of ZFC are satisfied under the interpretation that sets are only
the sets belonging to M. We say that M satisfies
ZFC. If, for instance, M also satisfies the negation
of CH, then CH cannot be provable in ZFC.
Here we mention another result of Gdel, from
1938: the consistency of CH. Gdel constructed a
model of ZFC, the constructible universe L, that
satisfies CH. The model L is basically the minimal
possible collection of sets that satisfies the axioms
of ZFC. Since CH is true in L, it follows that CH
cannot be refuted in ZFC. In other words, CH is
consistent.
Cohens accomplishment was that he found a
method for constructing other models of ZFC. The
idea is to start with a given model M (the ground
model ) and extend it by adjoining an object G, a
sort of imaginary set. The resulting model M[G] is
more or less a minimal possible collection of sets
that includes M, contains G, and most importantly, also satisfies ZFC. Cohen showed how to find
(or imagine) the set G so that CH fails in M[G].
Thus CH is unprovable in ZFC, and, because CH is
also consistent, it is independent, or undecidable.
A consequence of Gdels theorem about L is
that one cannot prove that there exists a set outside the minimal model L, so we have to pull G
out of thin air. The genius of Cohen was to introduce so-called forcing conditions that give partial
information about G and then to assume that G
is a generic set. A generic set decides which forcing conditions are considered true. With Cohens
definition of forcing and generic sets it is possible
to assume that, for any ground model M and any
given set P of forcing conditions in the model M,
a generic set exists. Moreover, if G is generic (for
P over M), then M[G] is a model of set theory.
To illustrate the method of forcing, let us
consider the simplest possible example, and let
June/July 2008
us stipulate that G should be a set of integers. As forcing conditions we consider finite

sets of expressions a G and a G where a
ranges over the set of all integers. (Therefore
{1 G, 2 G, 3 G, 4 G} is a condition that
forces G {1, 2, 3, 4} = {1, 3, 4}.) The genericity of
G guarantees that the conditions in G are mutually compatible, and, more importantly, that every
statement of the forcing language is decided one
way or the other. We shall not define here what
generic means precisely, but let me point out
one important feature. Let us identify the above
generic set of integers G with the set G of forcing
conditions that describe the initial segments of
G. Genericity implies that for any statement A of
the forcing language, if every condition can be
extended to a condition that forces A, then some
condition p in G forces A (and so A is true in
M[G]). For instance, if S is a set of integers and
S is in the ground model M, then every condition
p of the kind described above can be extended
to a condition that forces G S: simply add the
expression a G for some a S (or a G for
some a S) where a is an integer not mentioned
in p. It follows that the resulting set of integers G
is not in M, no matter what the generic set is.
Cohen showed how to construct the set of forcing conditions so that CH fails in the resulting
generic extension M[G]. Soon after Cohens discovery his method was applied to other statements
of set theory. The method is extremely versatile:
every partially ordered set P can be taken as the
set of forcing conditions, and when G P is a
generic set then the model M[G] is a model of
ZFC. Moreover, properties of the model M[G] can
be deduced from the structure of P . In practice,
the forcing P can be constructed with the independence result in mind; forcing conditions usually
approximate the desired generic object G.
In the 45 years since Cohens discovery, literally hundreds of applications of forcing have been
discovered, giving a better picture of the universe
of sets by producing examples of statements that
are undecidable from the conventional axioms of
mathematics.
A word of caution: If after reading this you
entertain the idea that perhaps the Riemann Hypothesis could be solved by forcing, forget it. That
conjecture belongs to a class of statements that,
by virtue of their logical structure, are absolute
for forcing extension: such a statement is true in
the generic extension M[G] if and only if it is true
in the ground model M. This is the content of
Shoenfields Absoluteness Theorem.
And what is Shoenfields Absoluteness Theorem?
Well, that is for someone else to explain, some
other time.
Notices of the AMS
693

What Is Forcing

Uploaded by

Copyright:

Available Formats

What Is Forcing

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

What Is Forcing

Uploaded by

Copyright:

Available Formats

?

What is forcing? Forcing is a remarkably powerful

case that there exists no proof of the conjecture

Notices of the AMS

Volume 55, Number 6

us stipulate that G should be a set of integers. As forcing conditions we consider finite

Notices of the AMS

You might also like