Pumping Lemma for Context-Free Languages

CSCI 3130 Formal Languages and Automata Theory

Siu On CHAN
Fall 2018
Chinese University of Hong Kong

1/18
 regular
context-free

L1 = {an bn | n > 0}
L2 = {z | z has the same number of a’s and b’s}
L3 = {an bn cn | n > 0}
L4 = {zz R | z ∈ {a, b}∗ }
L5 = {zz | z ∈ {a, b}∗ }

These languages are not regular

Are they context-free?

2/18
An attempt

L3 = {an bn cn | n > 0}

Let’s try to design a CFG or PDA

S → aBc | ε read a / push x

read b / pop x
B → ???
???

3/18
 Suppose we could construct some CFG G for L3
S ⇒ CC
⇒ SBC
e.g. ⇒ SCSC
S → CC | BC | a ⇒ SSBSC
B → CS | b ⇒ SSBBCC
C → SB | c ⇒ aSBBCC
⇒ aaBBCC
How does a long
⇒ aabBCC
derivation look like?
⇒ aabbCC
⇒ aabbcC
⇒ aabbcc

4/18
Repetition in long derivations

If a derivation is long enough, some variable must appear twice on

the same root-to-leave path in a parse tree
S ⇒ CC
⇒ SBC
⇒ SCSC S
⇒ SSBSC C C
⇒ SSBBCC S B c
⇒ aSBBCC
a C S
⇒ aaBBCC
S B B C
⇒ aabBCC
⇒ aabbCC a b b c

⇒ aabbcC
⇒ aabbcc
5/18
Pumping example

Then we can “cut and paste” part of parse tree

S
S C C
C C S B c

S B c a C S
a C S S B B C
S B B C a C S b c
a b b c S B B C

aabbcc a b b c

aaabbcbcc

6/18
Pumping example

We can repeat this many times

aabbcc ⇒ aaabbcbcc ⇒ aaaabbcbcbcc ⇒ . . .
⇒ (a)i ab(bc)i c

Every sufficiently large derivation will have a middle part that can be
repeated indefinitely

7/18
Pumping in general

u
u A y

u A y v

A y v A x u
v A x v A y

A x v A x w
w A x v uwy
uvwxy w A x
2 2
uv wx y w
uv wx 3 y
3

8/18
Example

L3 = {an bn cn | n > 0}

If L3 has a context-free grammar G, then for any sufficiently long

s ∈ L(G)

s can be split into s = uvwxy such that L(G) also contains uv 2 wx 2 y,

uv 3 wx 3 y, …

What happens if s = am bm cm

No matter how it is split, uv 2 wx 2 y ∈

/ L3

9/18
Pumping lemma for context-free languages

For every context-free language L

There exists a number m such that for every long string s in L
(|s| > m), we can write s = uvwxy where
1. |vwx| 6 m
2. |vx| > 1
3. For every i > 0, the string uv i wx i y is in L

At least
one is
non-empty

u v w x y

6m

10/18
Pumping lemma for context-free languages

To prove L is not context-free, it is enough to show that

For every m there is a long string s ∈ L, |s| > m, such that for every
way of writing s = uvwxy where

1. |vwx| 6 m
2. |vx| > 1

there is i > 0 such that uv i wx i y is not in L

11/18
Using the pumping lemma

L3 = {an bn cn | n > 0}

1. for every m
2. there is s = am bm cm (at least m symbols)
3. no matter how the pumping lemma splits s into uvwxy
(|vwx| 6 m, |vx| > 1)
4. uv 2 wx 2 y ∈
/ L3 (but why?)

12/18
Using the pumping lemma

Case 1: v or x contains two kinds of symbols

aa aabb
| {z } bbcccc
v
2 2
Then uv wx y ∈
/ L3 because the pattern is wrong

Case 2: v and x both contain (at most) one kind of symbol

aaa |{z}
a b |{z}
bb bcccc
v x
2 2
Then uv wx y does not have the same number of a’s, b’s and c’s

Conclusion: uv 2 wx 2 y ∈
/ L3

13/18
 Which is context-free?

L1 = {an bn | n > 0} 3
L2 = {z | z has the same number of a’s and b’s} 3
n n n
L3 = {a b c | n > 0} 7
R
L4 = {zz | z ∈ {a, b} }∗
3
L5 = {zz | z ∈ {a, b} } ∗

14/18
Example

L5 = {zz | z ∈ {a, b}∗ }

1. for every m
2. there is s = am bam b (at least m symbols)
3. no matter how the pumping lemma splits s into uvwxy
(|vwx| 6 m, |vx| > 1)
4. Is uv 2 wx 2 y ∈
/ L5 ?

15/18
Example

L5 = {zz | z ∈ {a, b}∗ }

1. for every m
2. there is s = am bam b (at least m symbols)
3. no matter how the pumping lemma splits s into uvwxy
(|vwx| 6 m, |vx| > 1)
4. Is uv 2 wx 2 y ∈
/ L5 ?

aaa |{z}
a aba |{z}
a aaab
v x

15/18
Example

L5 = {zz | z ∈ {a, b}∗ }

1. for every m
2. there is s = am bm am bm (at least m symbols)
3. no matter how the pumping lemma splits s into uvwxy
(|vwx| 6 m, |vx| > 1)
4. Is uv i wx i y ∈
/ L5 for some i?

Recall that |vwx| 6 m

16/18
Example

Three cases

Case 1 aaa aabbb

| {z } bbaaaaabbbbb
vwx
vwx is in the first half of am bm am bm

Case 2 aaaaabb bbbaa

| {z } aaabbbbb
vwx
vwx is in the middle part of am bm am bm

Case 3 aaaaabbbbbaaa aabbb

| {z } bb
vwx
vwx is in the second half of am bm am bm

17/18
Example

Apply pumping lemma with i = 0

Case 1 aaa aabbb

| {z } bbaaaaabbbbb
vwx
uwy becomes aj bk am bm , where j < m or k < m

Case 2 aaaaabb bbbaa

| {z } aaabbbbb
vwx
uwy becomes am bj ak bm , where j < m or k < m

Case 3 aaaaabbbbbaaa aabbb

| {z } bb
vwx
uwy becomes am bm aj bk , where j < m or k < m

Not of the form zz

This covers all cases, so L5 is not context-free
18/18