PH340 Problem Set I Solutions

(1) Recall definition 5.4 in the Open Logic Text: If A and B are models for the same
language, then A is a substructure of B if |A| |B| and i) cA = cB for all constants
c L, ii) ha1 , . . . , an i P A if and only if ha1 , . . . , an i P B for all a1 , . . . , an |A| for all
predicates P L, and iii) (f (a1 , . . . , an ))A = (f (a1 , . . . , an ))B for all a1 , . . . , an |A| for
all functions f L. In this case we write A B.
We additionally say that A is an elementary substructure of B if A is a substructure of
B and also for all a1 , . . . , an |A|, A |= (a1 , . . . , an ) if and only if B |= (a1 , . . . , an ).
Let A B and suppose that and every for every formula (x1 , . . . , xn , y) and tuple of
objects a1 , . . . , an |A|, the following holds:

(?) if B |= y(a1 , . . . , an , y), then there exists b |A| such that B |= (a1 , . . . , an , b).

Show that A is an elementary substructure of B. (?) is called the Tarski-Vaught criterion.

Solution:
Suppose that A B and that A and B satisfy the Tarski-Vaught criterion (?). In order
to show that A is an elementary substructure of B, we must show that

(+) for all a1 , . . . , an |A|, A |= (a1 , . . . , an ) if and only if B |= (a1 , . . . , an )

We do this by induction on the structure of (x1 , . . . , xn ).

Note first that by a straightforward induction on terms we have that for all terms
t(x1 , . . . , xn ) and a1 , . . . , an |A|, tA (a1 , . . . , an ) = (t(a1 , . . . , an ))A = (t(a1 , . . . , an )B =
tB (a1 , . . . , an ). From this the base case for statements of the form (x1 , . . . , xn )
t1 (x1 , . . . , xn ) = t2 (x1 , . . . , xn ) follows immediately. If (x1 , . . . , xn ) P (t1 , . . . , tn )
(where t1 , . . . , tn may have x1 , . . . , xk free), we have that for all a1 , . . . , ak |A|, A |=
P (t1 (a1 , . . . , ak ), . . . , tn (a1 , . . . , ak )) iff ht1 (a1 , . . . , ak ), . . . , tn (a1 , . . . , ak )i P A iff
ht1 (a1 , . . . , ak ), . . . , tn (a1 , . . . , ak )i P B iff B |= P (t1 (a1 , . . . , ak ), . . . , tn (a1 , . . . , ak )).
Using this as the base case, we may now show that the property (+) may be extended to
arbitrary formulas. We take advantage of the fact that {, , } is a truth functionally
complete set of connectives and assume that a1 , . . . , an |A|:

- Suppose (+) holds for (x1 , . . . , xn ) . Then A |= (a1 , . . . , an ) iff A 6|= (a1 , . . . , an )
iff B 6|= (a1 , . . . , an ) iff B |= (a1 , . . . , an ).

- Suppose (+) holds for (x1 , . . . , xn ) and (x1 , . . . , xn ) . Then A |= (a1 , . . . , an )

(a1 , . . . , an ) iff A |= (a1 , . . . , an ) and A |= (a1 , . . . , an ) iff B |= (a1 , . . . , an )
and B |= (a1 , . . . , an ) iff B |= (a1 , . . . , an ) (a1 , . . . , an ).

- Suppose that (+) holds for (x1 , . . . , xn , y) and suppose that A |= y(a1 , . . . , an , y)
with a1 , . . . , an |A|. But then there is b |A| such that A |= (a1 , . . . , an , b).
And hence since |A| |B|, B |= (a1 , . . . , an , b) and thus B |= y(a1 , . . . , an , y)
 as well. Conversely, suppose that B |= y(a1 , . . . , an , y) with a1 , . . . , an |A|. It
then follows by (?) that there exists b |A| such that B |= (a1 , . . . , an , b). But
then A |= y(a1 , . . . , an , y) as well.

(2) Show that Q = hQ, <i is an elementary substructure of R = hR, <i. [Hint: Apply
the Tarsk-Vaught criterion]

Solution: Note that since Q R and the less-than ordering of Q is just the restriction
of the less-than ordering of R, we have that Q R. In order to show that Q is an
elementary substructure of R it hence suffices to show that this pair of models satisfy
the Tarski-Vaught criterion.
To this end, let q1 , . . . , qn Q be such that R |= y(q 1 , . . . , q n , y). From this it follows
that there exists a real number r R such that R |= (q 1 , . . . , q n , r). We want to show
that there also exists a rational number q Q which can be chosen as the witness for the
existential quantifier in the structure R i.e. such that R |= (q 1 , . . . , q n , q). Informally,
we know that this is possible because a formula in the language L< can only assert facts
about either the equality of q1 , . . . , qn , r or there relative order. But since Q is a dense
linear ordering without end points, we can always find a rational q which bears the same
order and equality relations to q1 , . . . , qn as r.
To turn this idea into a proof, suppose that R |= y(q 1 , . . . , q n , y) and let r R be
such that R |= (q 1 , . . . , q n , r). If additionally r Q, then we are done. Otherwise, we
must consider three cases: 1) there exists i, j such that qi < r < qj (in which case we
can assume without loss of generality that qi and qj are the closest rationals with this
property); 2) r < qi for all 1 i n; and 3) qi < r for all 1 i n.
In case 1), let q be such that qi < q < r (we know that such a q exists because the
rationals are dense) and define a mapping h : R R as follows:



x if x qi
qqi (x q ) + q

if qi < x r
rqi i i
h(x) = qj q

qj r
(x r) + q if r < x < qj


x if qj x

It is easy to check that h(x) is a bijection and that for all x, y R, x < y iff h(x) < h(y).
Thus h(x) is an automorphism of R = hR, <i i.e. an isomorphism of R into itself. It is
also easy to see that for any formula (x1 , . . . , xk ) and r1 , . . . , rn R, R |= (r1 , . . . , rk )
iff R |= (h(r1 ), . . . , h(rk )). But now observe that h(r) = q Q and h(qi ) = qi for all
1 i n. It hence follows that since R |= (q 1 , . . . , q n , r), R |= (q 1 , . . . , q n , q) as well.
The cases 2) and 3) are similar.
(3) Construct LA terms, formulas, and sentences expressing the following:

i) x5 + 5x + 5 (i.e. write out the corresponding LA -term explicitly)

Solution: x00000 + 000000 x + 000000

ii) x is divisible by 3
 Solution: y(0000 y = x) =df DivT hree(x)

iii) x divides y with no remainder

Solution: z(z y = z) =df x|y

iv) the remainder when x is divided by y is z

Solution: qr(y q + r = x r < q) =df Rem(x, y, z)

v) all natural numbers x are such that they are either divisible by 3 with no remainder,
or with remainder 1, or with remainder 2
Solution: x(Rem(x, 3, 0) Rem(x, 3, 1) Rem(x, 3, 2))

vi) x is a prime number

Solution: 1 < x y(y|x (y = 1 y = x)) =df P rime(x)

vii) there are infinitely many prime numbers

Solution: xy(x < y P rime(y))

(4) Recall the theory Th(N) = { SentLA | N |= } i.e. the set of sentences true
in the standard model of arithmetic. Let M |= Th(N) be non-standard i.e. M 6= N.
(Recall such models exist by the Compactness Theorem.) Show that M has the following
properties:

i) The minimal closure of the element z = 0M is the smallest set I |M| containing
z and closed under =0M (i.e. the interpretation of the successor function in M).
Show that there exist a |M| which are not in I.
Solution: Observe that the minimal closure of z under in M is the set I =
{z, z , z , z , . . .} = {z... (n times) : n N}. Now consider the mapping h :
N |M| given by h(n) = z... (n times) whose range is I. If h(x) were surjective
(i.e. I = |M|), then it would be an isomorphism between N and M. (See the proof
of Theorem 5.28.) But since M 6' N, this cannot be the case. Hence there must
exist an element a |M| I.

ii) Show that <M is not a well-ordering.

Solution: By part i), let a0 |M|I be non-standard. It thus follows that a0 6= 0M =
z. But now observe that N |= x(x 6= 0 y(y 0 = x)) i.e. the standard model
satisfies the sentence formalizing every number other than 0 has a predecessor.
Since M |= Th(N), this sentence is also true in M i.e. M |= x(x 6= 0 y(y 0 =
x)). There thus exists a1 |M| such that a1 = a0 . Since N |= xy(x0 = y x < y),
we also have M |= xy(x0 = y x < y) from which it follows that a1 a0 (where
=<M is the ordering relation of M). But note that a1 must also be nonstandard
(for if a1 were in I, then a0 = a1 would be in I as well). Repeating this procedure we
obtain a nonstandard a2 a1 , a nonstandard a3 a2 , . . . Together, these elements
form an infinite descending sequence . . . ai+1 ai . . . a3 a2 a1 a0 , thus
showing that <M is not a well-ordering.
iii) We have seen that every nonstandard model M |= Th(N) has an initial segment
order-isomorphic to the natural numbers i.e. a set I containing 0M and closed
under successor (in the sense of M) such that hI, <M N
I i is isomorphic to hN, < i
M M
(where <I is the restriction of < to I) . Show that there is no LA formula (x)
which defines I relative to M i.e. such that {a |M| : M |= (a)} = I. [Hint:
Observe that M satisfies the first-order induction schema and that if (x) defines I
in M, then M 6|= x(x).]
Solution: Suppose that (x) is an LA -formula such that {a |M| : M |= (a)} = I.
Note that since 0M = z I, M |= (0). Now consider an arbitrary b |M| such
that M |= (b). Since (x) defines I, it follows that b I, meaning that b = z... (n
0
times) for some n N. But then b = z... (n + 1 times) = (b )M is also an element
0
of I, from which it follows that M |= (b ). Since b was arbitrary, it follows that
M |= x((x) (x0 )).
Next observe that in the standard model N, all instances of the induction schema
are satisfied. So in particular, N |= ((0) x((x) (x0 ))) x(x). But since
M |= Th(N), we also have M |= ((0) x((x) (x0 ))) x(x). Since we
have just seen that both conjuncts of the antecedent of this conditional are true in
M, it thus follow that M |= x(x) as well. We have also seen, however, that since
M is nonstandard, there must exist a |M| I. But then M |= (a), contradicting
the fact that (x) defines I in M.

iv) Recall that n is an abbreviation for the LA -term 00...0 n-times e.g. 3 = 0000 . Prove
the Overspill Lemma for M i.e. if M |= (n) for arbitrarily large n N, then
M |= (a) for some non-standard element (i.e. a |M| N).
Solution: To say that M |= (n) for arbitrarily large n N is to say that M |=
xy(x < y (y)). Suppose that (x) is such a formula and that M |= (a) for
all nonstandard elements a. Consider the formula (x) = y(x < y (y)). It is
easy to see that under the two foregoing assumptions (x) defines the initial segment
I in M i.e. {a |M| : M |= (a)} = I. But this contradicts part iii).

v) Show that there exist nonstandard elements a, b |M| satisfying the formulas you
constructed in Problem 3 parts ii) and vi) expressing that x is divisible 3 and is
prime. Is there also a nonstandard element c |M| satisfying both of these formulas
simultaneously?
Observe that N |= xy(x < y P rime(y)) and N |= xy(x < y DivT hree(y))
i.e. the standard model satisfies the the statements there are arbitrarily large
primes and there are arbitrarily large numbers divisible by 3. Since M |= Th(N),
these statements are true in M as well. Hence by the Overspill Lemma, there are
nonstandard a, b |M| such that M |= P rime(a) and M |= DivT hree(b).
Note that in the standard model, the only number that satisfies P rime(x) and
DivT hree(x) simultaneously is 3 itself (which is standard). In particular, N |=
x(3 < x (P rime(x) DivT hree(x)). So again since M |= Th(N), we have
M |= x(3 < x (P rime(x) DivT hree(x)). Thus there is thus no nonstandard
a |M| satisfying P rime(x) and DivT hree(x) (although of course the standard
M
element 3 = z satisfies both formulas in M).
(5) Show that the following functions and relations are primitive recursive by proving
suitable primitive recursive definitions. In so doing, you may rely on functions or relations
shown to be primitive recursive in lecture or in the textbooks (but you must be explicit
about which functions you are using and how they figure in your definitions).

i) const5 (x) = 5, for all x N

Solution: const5 (x) is primitive recursive since it may be define in terms of the
successor and zero functions and composition. In particular, we have const5 (x) =
s(s(s(s(z(x))) where z(x) = 0 and s(x)is the successor function.

ii) 5x Solution: Note that 5x = x+(x+(x+(x+x))) and we have seen that the addition
function x + y has a primitive recursive definition add(x, y). We may hence define
a function T woT imes(x) by composition as add(x, x). (Formally this is given by
composing the binary function add(x, y) the unary projection functionP01 (x0 ) = x0
which is just the identity function twice so that T woT imes(x) = add(P01 (x), P01 (x)).)
We may define T hreeT imes(x) = add(x, T woT imes(x)), F ourT imes(x) =
add(x, T hreeT imes(x)) and F iveT imes(x) = add(x, F ourT imes(x)). Or expanding
the definitions

F iveT imes(x) = add(x, (add(x, add(x, add(x, x)))))

iii) x5 Solution: Note that x5 = x (x (x (x x))) and we have seen that the
addition function x y has a primitive recursive definition mult(x, y). As above we
can define

F if thP ower(x) = mult(x, (mult(x, mult(x, mult(x, x)))))

iv) x5 + 5x + 5 Solution: Combining the above, x5 + 5x + 5 can be defined by composing

add(x, y), const5 (x), F ivT imes(x) and F if thP ower(x) as

add(F if thP ower(x), add(F iveT imes(x), Const5 (x)))

v) DivisibleByT hree(x) iff x is divisible by 3

Solution: Recall that we have seen that relation x|y (i.e. x divides y) is primitive
recursive because it is definable by bounded quantification as x|y if and only if
z y(mult(x, y) = z). DivisibleByT hree(x) is hence defined by const3 (x)|y i.e
by z x(mult(const3 (x), z) = x).

vi) Let div(x, y) = bx/yc i.e., rounded division, where you disregard everything after
the decimal point and when y = 0 we stipulate div(x, y) = 0. Show div(x, y) is
primitive recursive in two ways: a) by giving an explicit definition of div(x, y) using
primitive recursion and composition (you will have to employ an auxiliary function
you cannot use recursion on the arguments x or y themselves); b) by using bounded
minimization.
Solution:
 a) It would be natural to define div(x, 0) = 0 and div(x, y) = div(xy, y) + 1.
However this is not a proper primitive recursive definition as the value of div(x, y)
is given as a function of xy and y instead x and y 1. An alternative definition
makes use of the two auxiliary functions f (u, x, y, z) and CountU p(x, y, z) defined
as
(
u+y if x y z
f (u, x, y, z) =
u otherwise

and CountU p(x, y, 0) = 0, CountU p(x, y, z + 1) = f (CountU p(x, y, z), x, y, z). We

may then define div(x, y) = CountU p(x, y, x).
b) div(x, y) can be defined by bounded minimization and definition by cases as
follows:
(
min z x(mult(y, z) x) if y 6= 0
div(x, y) =
0 otherwise

vii) mod(x, y) = the remainder when y is divided by x

Solution: mod(x, y) = x(mult(div(x, y), y))

viii) OneOrThreeModFour(x) iff the remainder when x is divided by 4 is 1 or 3

Solution: OneOrThreeModFour(x) is defined by the condition mod(x, 4) = 1
mod(x, 4) = 3. Thus the characteristic function of this relation is given by

max(= (mod(x, 4), 1), = (mod(x, 4), 3))

where = (x, y) is the characteristic function of the equality relation.

ix) The Fibonacci function f ib(x) defined by the following recursive definition:
(
1 if x 1
f ib(x) =
f ib(x 1) + f ib(x 2) if x > 1

Solution: Unlike a standard primitive recursive definition, the value of f ib(x) (for x > 1)
is determined not just in terms of f ib(x 1), but in terms of the two proceeding values
f ib(x 1), f ib(x 2). In order to see that this function it is primitive we may use the
sequence coding function hx, yi of two arguments (i.e. hx, yi = 2x+1 3y+1 ) together with
the first and second element functions ()0 and ()1 which are defined such that (hx, yi)0 =
x and (hx, yi)1 = y. Using these, we may define an auxiliary function unction g(x) so
that that g(0) = hf ib(0), f ib(1)i, g(1) = hf ib(1), f ib(2)i, g(2) = hf ib(2), f ib(3)i, . . .. Such
a function can be defined by primitive recursion as follows: g(0) = h1, 1i, g(x + 1) =
h(g(x))1 , (g(x))0 + (g(x))1 i. Now define f ib(x) = (g(x))0 .

(6) Relative to the scheme for coding finite sequences of natural numbers described in
section 6.7 of The Open Logic Text, do the following:
 i) Calculate the number s coding h3, 1, 4, 1, 5i.
Solution: s = 24 32 55 72 116

ii) Calculate the numbers s1 , . . . , s5 coding h3i, h3, 0i, h3, 0, 0i, h0, 3, 0, 0, 3i and h0, 3, 0, 0, 3, 0i.
Solution: s1 = 24 , s2 = 24 31 , s3 = 24 31 51 , s4 = 21 34 51 71 114 ,
s5 = 21 34 51 71 114 131

iii) Find the sequence coded by s = 3074610000.

Solution: s = 24 31 54 71 114 . So s codes h3, 0, 3, 0, 3i.

(7) Relative to the system of Godel number described chapter 8 of the Open Logic Text,
calculate the Godel numbers of the following LA -term and formulas: 0, 00 , 000 , x1 , x2 , 0 +
x1 , (0 + x1 )0 , (0 + x1 )0 x2 , 0 < 000 and x1 x2 (x2 = x1 + x1 ). You may leave your answer
in prime power notation.

Solution
Since 0 = c0 , #(0) = h2, 0i = 23 31 .
Since 00 abbreviates 0 (0) and 0 is f01 (x), #(0 (0)) = h#(0 ), #((), #(0), #())i =
hh3, 1, 0i, h0, 7i, h3, 1, 0i, h0, 7i, h2, 0i, h0, 8i, h0, 8ii = h24 32 51 , 21 38 , 24 32 51 , 2
4 2 1 1 8 4 2 1 8 3 1
38 , 23 31 , 21 39 , 21 39 i = 22 3 5 +1 32 3 +1 52 3 5 +1 723 +1 112 3 +1
1 9 1 9
132 3 +1 172 3 +1 .
#(x1 ) = h1, 1i = 22 32
#(x1 ) = h1, 2i = 22 33
0 + x1 abbreviates +(0, x1 ) where + is f12 (x, y).
So #(+(0, x1 )) = hh3, 2, 0i, h0, 7i, h2, 0i, h0, 9i, h1, 1i, h0, 8ii = h24 33 51 , 21 38 , 23
4 3 1 1 8 3 1 1 10 2 2
31 , 2 310 , 22 32 , 21 39 i = 22 3 5 +1 32 3 +1 52 3 +1 72 3 +1 112 3 +1
1 9
132 3 +1
The others parts are similar.