CSE303 - Introduction to the Theory of Computation

Sample Solutions for Exercises on

Context-Free Languages and Pushdown Automata

1. Exercise 3.1.4
Let Σ be the alphabet {a, b, (, ), ⊘, ∪, ⋆ }. The grammar G = (V, Σ, R, S)
generates all strings that are regular expressions over {a, b}, where

V = Σ ∪ {S} and
R = {S → ⊘,
S → a,
S → b,
S → (SS),
S → (S ∪ S),
S → S ⋆ }.

2. Exercise 3.1.9
(a) A suitable grammar is one with start symbol S and rules

S → aSb
S → aS
S → e

(b) A suitable grammar is one with start symbol A1 and rules

A1 → e A2 → e A3 → e
A1 → aA1 d A2 → aA2 c A3 → bA3 d
A1 → A2 A2 → A4 A3 → A4
A1 → A3 A4 → e
A1 → A4 A4 → bA4 c

(d) A suitable grammar is one with start symbol S and rules

S → ab T → aa T → aT b
S → aT b T → ab T → bT a
S → bT b T → aT a T → bT b
 (f) A suitable grammar is one with start symbol S and rules

S → aaSb
S → aSb
S → Sb
S → e

3. Exercise 3.3.1
Let M be the given pushdown automaton.
(a) We first trace all possible sequences of transitions of M on input
aba.

State Unread input Stack Comments

s aba e
s ba a
s a aa
s e aaa Accepts not
s aba e
s ba a
s a aa
f e aa Accepts not
s aba e
f ba e Accepts not

(b) Part (a) shows that the string aba is not accepted by M . In a
similar way one can show that the strings aa and abb are not accepted
by M either.
The string baa, on the other hand, is accepted, as the following deriva-
tion shows:

State Unread input Stack Comments

s baa e
s aa a
f a a
f e e Accepts

In a similar way one can show that the strings bab and baaaa are
elements of L(M ).
 (c) The automaton M accepts all strings of odd length with an a in
the middle, i.e., L(M ) = {uav ∈ {a, b}∗ : |u| = |v|}.
Note that in state s the automaton M may read a symbol, a or b,
and push one a onto the stack while remaining in state s (“phase 1”).
Similarly, in state f the automaton M may read a symbol, a or b,
and pop one a from the stack while remaining in state f (“phase 2”).
In addition, M may change from s to f when reading an a, without
changing the stack. The automaton can only end up with an empty
stack if the number of symbols pushed in phase 1 is the same as the
number of symbols popped in phase 2. This is only possible if the
input string is of odd length and M changes from s to f when reading
the middle symbol of the input, which consequently must be an a.

4. Exercise 3.3.2
(c) The language {w ∈ {a, b}∗ : w = w R } is accepted by the pushdown
automaton M = ({s, f }, Σ, Γ, ∆, s, {f }), where Σ = Γ = {a, b} and

∆ = {((s, a, e), (s, a)),

((s, b, e), (s, b)),
((s, e, e), (f, e)),
((s, a, e), (f, e)),
((s, b, e), (f, e)),
((f, a, a), (f, e)),
((f, b, b), (f, e))}.

(d) The language {w ∈ {a, b}∗ : w has twice as many b’s as a’s} is ac-
cepted by the pushdown automaton M = ({s, q, p, f }, Σ, Γ, ∆, s, {f }),
where Σ = {a, b}, Γ = {a, b, c} and

∆ = {((s, e, e), (q, c)),

((q, a, e), (q, bb)),
((q, b, b), (q, e)),
((q, b, c), (p, bc)),
((r, a, bb), (r, e),
((r, a, bc), (q, bc),
((r, a, c), (q, bbc),
((r, b, e), (q, b)),
 ((q, e, c), (f, e)),
((r, e, c), (f, e))}.

The symbol c is used to mark the bottom of the stack. The marker
is put on the stack initially, when M changes from state s to state q
(and before any input sybols have been read). It can be removed only
when M changes to a final state. A nonempty stack will be of the
form bi c, where i ≥ 0.
During the processing of the input the automaton is either in state q or
state r. If it is in state q, then 2na ≥ nb , where nσ denotes the number
of occurrences of the symbol σ that have been read. The stack height
in that case will be 2na − nb + 1. If it is in state r, then 2na ≤ nb , and
the stack height will be nb − 2na + 1.

5. Exercise 3.5.1
(a) The language L = {am bn : m 6= n} is the union of the two lan-
guages {am bn : m > n} and {am bn : m < n}, both of which are
context-free (cf. Exercise 3.1.9 (a) for a very similar problem). Hence
L is also context-free.
(c) The language L = {am bn cpdq : n = q or m ≤ p or m + n = p + q}
is the union of three languages, {am bn cpdq : n = q}, {am bn cp dq : m ≤
p}, and {am bn cpdq : m + n = p + q}.
To show that {am bn cpdq : n = q} is context-free one can design a
corresponding context-free grammar, using similar ideas as for the
solution of Exercise 3.5.14 (a) below.
A suitable grammar for {am bn cp dq : m ≤ p} consists of start symbol
S and rules,

S → TD
T → aT c
T → Tc
T → B
B → bB
B → e
D → dD
D → e
 The set {am bn cp dq : m + n = p + q} was shown to be context-free in
Exercise 3.1.9 (b).
(e) The language L = {w ∈ {a, b}∗ : w = w R } is the union of the
three languages {ww R : w ∈ {a, b}∗}, {waw R : w ∈ {a, b}∗}, and
{wbw R : w ∈ {a, b}∗}, all of which are context-free (cf. Exercises 3.1.3
(a) and (b)). Hence L is also context-free.

6. Exercise 3.5.2
(b) We prove the assertion by contradiction. Suppose the set L =
2
{an : n ≥ 0} is context-free. By the version of the Pumping Theorem
discussed in class, there exists an integer K > 0, such that every string
w ∈ L of length at least K can be divided into five parts, w = uvxyz,
where |vy| > 0, |vxy| ≤ K, and for each i ≥ 0, uv i xy i z ∈ L.
2
Let K be such an integer for L and let w be the string aK . We
have w ∈ L and |w| > K. Suppose w is written as uvxyz, where
|vy| > 0 and |vxy| ≤ K, and take the string w ′ = uv 2 xy 2 z. Then
K 2 < |w ′ | ≤ K 2 + K = K(K + 1) < (K + 1)2 . Thus, w ′ 6∈ L, which
contradicts the condition that for each i ≥ 0, uv i xy i z ∈ L. We may
infer that the set L is not context-free.
(d) See Example 3.5.4 for a solution.

7. Exercise 3.5.14
(a) The language L = {am bn cp : m = n or n = p or m = p} is the
union of three sets L1 , L2 , and L3 , where L1 is the set {am bn cp | m =
n}, L2 the set {am bn cp | n = p}, and L3 the set {am bn cp | m = p}.
The set L1 can be generated by a grammar with start symbol S and
rules

S → BC
B → aBb
B → e
C → cC
C → e

Similar context-free grammars can be designed for the sets L2 and L3 .

Thus, L is the union of context-free languages, and hence is context-
free.
(b) The language {am bn cp : m 6= n or n 6= p or m 6= p} is context-free
as it is the union of three context-free sets, L1 = {am bn cp | m 6= n},
L2 = {am bn cp | n 6= p}, and L3 = {am bn cp | m 6= p}.
To see that these languages are context-free, note that L1 can be
generated by a context-free grammar with start symbol S and rules

S → TC
T → aT b
T → A
T → B
A → aA
A → a
B → bB
B → b
C → cC
C → e

A similar grammar can be designed for L2 , whereas a suitable grammar

for L3 uses start symbol S and rules

S → aSc
S → AB
S → BC
A → aA
A → a
B → bB
B → e
C → cC
C → c

(c) The set {am bn cp : m = n and n = p and m = p} is the same as

the set {ai bici : i ≥ 0}, which is not context-free; see Example 3.5.2.
(d) The specification of a set as consisting of all strings in {a, b, c}∗
that do not contain equal numbers of occurrences of a, b, and c is
slightly ambiguous.
Denoting by nσ (w) the number of occurrences of a symbol σ in a string
w, the intended set is probably

{w ∈ {a, b}∗ : na (w) 6= nb (w) or na (w) 6= nc (w) or nb (w) 6= nc (w)}.

This set is the union of three sets (cf. part (b) of this exercise), all of
which are recognized by pushdown automata and hence are context-
free.
But the description may also be interpreted as denoting the set

{w ∈ {a, b}∗ : na (w) 6= nb (w) and na (w) 6= nc (w) and nb (w) 6= nc (w)},

which is not context-free. This can be proved by applying a stronger

version of the Pumping Theorem known as “Ogden’s Lemma,” see the
references for Chapter 3. For a specific solution see, for example, p. 130
in Introduction to Automata Theory, Languages, and Computation by
John Hopcroft and Jeffrey Ullman.