Properties of Regular

Languages
COMP 455 – 002, Spring 2019

Jim Anderson (modified by Nathan Otterness) 1

What Languages Aren’t Regular?

The tool used for proving that a language

is not regular is the Pumping Lemma.

Jim Anderson (modified by Nathan Otterness) 2

The Pumping Lemma
Theorem 4.1 (the Pumping Lemma for regular languages):
Let 𝐿 be a regular language. Then there exists a constant
𝑛 (which depends on 𝐿) such that for every 𝑤 ∈ 𝐿, where
𝑤 ≥ 𝑛, there exists strings 𝑥, 𝑦, and 𝑧 such that
i. 𝑤 = 𝑥𝑦𝑧,
ii. 𝑥𝑦 ≤ 𝑛,
iii. 𝑦 ≥ 1, and
iv. for all 𝑘 ≥ 0, 𝑥𝑦 𝑘 𝑧 ∈ 𝐿.

Jim Anderson (modified by Nathan Otterness) 3

 Proof of the Pumping Lemma
Theorem 4.1: Let 𝐿 be a regular language. Then there exists a constant 𝑛 (which
depends on 𝐿) such that for every 𝑤 ∈ 𝐿, where 𝑤 ≥ 𝑛, there exists strings 𝑥, 𝑦, and 𝑧
such that (i) 𝑤 = 𝑥𝑦𝑧, (ii) 𝑥𝑦 ≤ 𝑛, (iii) 𝑦 ≥ 1, and (iv) for all 𝑘 ≥ 0, 𝑥𝑦 𝑘 𝑧 ∈ 𝐿.

 Since 𝐿 is regular, it is accepted by some DFA 𝑀.

 Let 𝑛 = the number of states in 𝑀.
 Pick any 𝑤 ∈ 𝐿, where 𝑤 > 𝑛.
 By the pigeonhole principle, 𝑀 must repeat a state
when processing the first 𝑛 symbols in 𝑤.

Jim Anderson (modified by Nathan Otterness) 4

 Proof of the Pumping Lemma
Theorem 4.1: Let 𝐿 be a regular language. Then there exists a constant 𝑛 (which
depends on 𝐿) such that for every 𝑤 ∈ 𝐿, where 𝑤 ≥ 𝑛, there exists strings 𝑥, 𝑦, and 𝑧
such that (i) 𝑤 = 𝑥𝑦𝑧, (ii) 𝑥𝑦 ≤ 𝑛, (iii) 𝑦 ≥ 1, and (iv) for all 𝑘 ≥ 0, 𝑥𝑦 𝑘 𝑧 ∈ 𝐿.

𝑀 must repeat a state, 𝑞, when processing the first 𝑛

symbols of input 𝑤.
Define strings 𝑥, 𝑦, and 𝑧 as in this figure:
𝑦

𝑥 𝑧
Start 𝑞0 𝑞

Jim Anderson (modified by Nathan Otterness) 5

 Proof of the Pumping Lemma
Theorem 4.1: Let 𝐿 be a regular language. Then there exists a constant 𝑛 (which
depends on 𝐿) such that for every 𝑤 ∈ 𝐿, where 𝑤 ≥ 𝑛, there exists strings 𝑥, 𝑦, and 𝑧
such that (i) 𝑤 = 𝑥𝑦𝑧, (ii) 𝑥𝑦 ≤ 𝑛, (iii) 𝑦 ≥ 1, and (iv) for all 𝑘 ≥ 0, 𝑥𝑦 𝑘 𝑧 ∈ 𝐿.
𝑦

𝑥 𝑧
Start 𝑞0 𝑞

The second occurrence of state 𝑞 must occur within

the first 𝑛 symbols of 𝑤, so 𝑥𝑦 ≤ 𝑛. Also, 𝑦 ≥ 1.
It must be possible to repeat the “𝑦-loop” 0 or more
times, and the resulting string will still be accepted.
Therefore, 𝑥𝑦 𝑘 𝑧 ∈ 𝐿, for any 𝑘 ≥ 0.
Jim Anderson (modified by Nathan Otterness) 6
Using the Pumping Lemma
 To show that a language 𝐿 is not regular, show that
the conditions of the Pumping Lemma do not hold.
 Formally, the Pumping Lemma says:
∃𝑛
∷ ൬∀𝑤: 𝑤 ∈ 𝐿 ∧ 𝑤 ≥ 𝑛
∷ ቀ∃𝑥, 𝑦, 𝑧: 𝑥𝑦𝑧 = 𝑤 ∧ 𝑥𝑦 ≤ 𝑛 ∧ 𝑦 ≥ 1
∷ ∀𝑘: 𝑘 ≥ 0 ∷ 𝑥𝑦 𝑘 𝑧 ∈ 𝐿 ቁ൰

Jim Anderson (modified by Nathan Otterness) 7

Using the Pumping Lemma
 To show that a language 𝐿 is not regular, show that
the conditions of the Pumping Lemma do not hold.
 The negation of the Pumping Lemma is:
∀𝑛
∷ ൬∃𝑤: 𝑤 ∈ 𝐿 ∧ 𝑤 ≥ 𝑛
∷ ቀ∀𝑥, 𝑦, 𝑧: 𝑥𝑦𝑧 = 𝑤 ∧ 𝑥𝑦 ≤ 𝑛 ∧ 𝑦 ≥ 1
∷ ∃𝑘: 𝑘 ≥ 0 ∷ 𝑥𝑦 𝑘 𝑧 ∉ 𝐿 ቁ൰

Jim Anderson (modified by Nathan Otterness) 8

Using the Pumping Lemma
∀𝑛 No matter what 𝑛 is…

∷ ൬∃𝑤: 𝑤 ∈ 𝐿 ∧ 𝑤 ≥ 𝑛 We can find a string 𝑤 longer than 𝑛 in 𝐿…

Where no matter how 𝑤 is divided into

∷ ቀ∀𝑥, 𝑦, 𝑧: 𝑥𝑦𝑧 = 𝑤 ∧ 𝑥𝑦 ≤ 𝑛 ∧ 𝑦 ≥ 1 substrings 𝑥, 𝑦, and 𝑧 (with some constraints)…

∷ ∃𝑘: 𝑘 ≥ 0 ∷ 𝑥𝑦 𝑘 𝑧 ∉ 𝐿 ቁ൰ We can find a value 𝑘 for which 𝑥𝑦 𝑘 𝑧 is not in 𝐿.

Show that this statement is true for 𝐿 to show that 𝐿

is not a regular language!

Jim Anderson (modified by Nathan Otterness) 9

Using the Pumping Lemma: Example

𝑖 2
 Consider the language 0 , where 𝑖 ≥ 1.
 We want to prove that this language is not regular.
 We need to show that ∀𝑛 ∷ ൬∃𝑤: 𝑤 ∈ 𝐿 ∧ 𝑤 ≥𝑛 ∷

∀𝑥, 𝑦, 𝑧: 𝑥𝑦𝑧 = 𝑤 ∧ 𝑥𝑦 ≤ 𝑛 ∧ 𝑦 ≥ 1 ∷ ∃𝑘: 𝑘 ≥ 0 ∷ 𝑥𝑦 𝑘 𝑧 ∉ 𝐿 ൰ is

true for 𝐿.
 The proof is like a game with an adversary:
❖ The adversary makes the “∀” choices.
❖ We make the “∃” choices.
Jim Anderson (modified by Nathan Otterness) 10
Using the Pumping Lemma: Example
∀𝑛
𝑖 2 ∷ ൬∃𝑤: 𝑤 ∈ 𝐿 ∧ 𝑤 ≥ 𝑛
 𝐿= 0 |𝑖 ≥1
∷ ቀ∀𝑥, 𝑦, 𝑧: 𝑥𝑦𝑧 = 𝑤 ∧ 𝑥𝑦 ≤ 𝑛 ∧ 𝑦 ≥ 1
 Assume 𝐿 is regular (this will be a proof ∷ ∃𝑘: 𝑘 ≥ 0 ∷ 𝑥𝑦 𝑘 𝑧 ∉ 𝐿 ቁ൰
by contradiction).
 Select any 𝑛.
𝑛2
 Let 𝑤 = 0 .
 Select any 𝑥𝑦𝑧 where 𝑤 = 𝑥𝑦𝑧, 𝑥𝑦 ≤ 𝑛, and 𝑦 ≥ 1.
 This implies that 1 ≤ 𝑦 ≤ 𝑛.
 Let 𝑘 = 2. This means that 𝑥𝑦 𝑘 𝑧 has 𝑛2 + 𝑦 0s.
Jim Anderson (modified by Nathan Otterness) 11
Using the Pumping Lemma: Example

𝑖 2
 (Reminder) 𝐿 = 0 |𝑖 ≥1
 Let 𝑘 = 2. This means that 𝑥𝑦 𝑘 𝑧 has 𝑛2 + 𝑦 0s.
 𝑤 has 𝑛2 0s, and any entry in 𝐿 longer than 𝑤 must
have at least 𝑛 + 1 2 0s.
 𝑛2 < 𝑛2 + 1 ≤ 𝑛2 + 𝑦 ≤ 𝑛2 + 𝑛 < 𝑛2 + 2𝑛 + 1.

The length The length of The length The length of = 𝑛 + 1 2 , the length
of 𝑤 (in 𝐿) 𝑥𝑦 2 𝑧 if 𝑦 = 1. of 𝑥𝑦 2 𝑧 𝑥𝑦 2 𝑧 if 𝑦 = 𝑛. of the next string in 𝐿.

 So, 𝑥𝑦 𝑘 𝑧 ∉ 𝐿, contradicting the Pumping Lemma.

Jim Anderson (modified by Nathan Otterness) 12

Using the Pumping Lemma: 2nd Example

 Consider the language 𝐿 =

𝑥 | 𝑥 contains an equal number of 0s and 1s
 Is 𝐿 regular?

Jim Anderson (modified by Nathan Otterness) 13

Using the Pumping Lemma: 2nd Example

 𝐿 = 𝑥 | 𝑥 contains an equal number of 0s and 1s

 Define 𝑤 = 0𝑛 1𝑛 for an arbitrary 𝑛. (𝑤 ∈ 𝐿)
 No matter how 𝑤 is divided into 𝑥, 𝑦, and 𝑧, 𝑦 must
consist solely of 0s because 𝑥𝑦 ≤ 𝑛 and 𝑦 ≠ 𝜀.
 Therefore, 𝑥𝑦 0 𝑧 has fewer 0s than 1s and is not in 𝐿.
❖ In this case, 𝑘 = 0.
❖ This proof would also work with any 𝑘 ≥ 2,
because 𝑥𝑦 𝑘 𝑧 would have more 0s than 1s.

Jim Anderson (modified by Nathan Otterness) 14

Closure Properties
 A closure property of regular languages is a
property that, when applied to a regular language,
results in another regular language.
❖ Union and intersection are examples of closure
properties.
 We will demonstrate several useful closure
properties of regular languages.
 Closure properties can also be useful for proving
that languages aren’t regular.

Jim Anderson (modified by Nathan Otterness) 15

Closure under Union
Theorem 4.4: If 𝑀 and 𝑁 are regular languages, then
𝑀 ∪ 𝑁 is a regular language.

Proof: Say that 𝑅 and 𝑆 are regular expressions where

𝐿 𝑅 = 𝑀 and 𝐿 𝑆 = 𝑁. Construct a regular expression
𝑅 + 𝑆 . This matches 𝑀 ∪ 𝑁. Since a regular
expression exists for 𝑀 ∪ 𝑁, 𝑀 ∪ 𝑁 is a regular language.

Note: The proofs for concatenation and Kleene closure

are similar.
Jim Anderson (modified by Nathan Otterness) 16
Closure under Complementation
 If 𝐿 ⊆ Σ ∗ , then the complement of 𝐿, denoted 𝐿ത , is
Σ ∗ − 𝐿.
 Theorem 4.5: If 𝐿 is a regular language over Σ, then
𝐿ത is also a regular language.
 Proof sketch for Theorem 4.5:
1. Construct a DFA for 𝐿
2. This can be transformed into a DFA for 𝐿 ത by
making all accepting states non-accepting and
vice versa.
3. This can be proven correct by induction.
Jim Anderson (modified by Nathan Otterness) 17
Closure under Intersection
Theorem 4.8: If 𝑀 and 𝑁 are regular languages, then
𝑀 ∩ 𝑁 is a regular language.

ഥ ∪ 𝑁.
Proof: 𝑀 ∩ 𝑁 = 𝑀 ഥ

The book contains a more direct proof. The basic idea

is to construct a DFA with states labeled 𝑝, 𝑞 where
𝑝 tracks the state of a DFA for 𝑀 and 𝑞 tracks the
state of a DFA for 𝑁.
Jim Anderson (modified by Nathan Otterness) 18
Closure under Difference
 𝑀−𝑁 ≡ 𝑥|𝑥 ∈𝑀∧𝑥 ∉𝑁
 Theorem 4.10: If 𝑀 and 𝑁 are regular, then so is
𝑀 − 𝑁.
ഥ
 Proof: 𝑀 − 𝑁 = 𝑀 ∩ 𝑁.

Jim Anderson (modified by Nathan Otterness) 19

Closure under Reversal
 The reversal of 𝐿, written 𝐿𝑅 is 𝑥 | 𝑥 𝑅 ∈ 𝐿 (𝑥 𝑅 is
the string 𝑥 written backwards).
 Theorem 4.11: If 𝐿 is regular, then so is 𝐿𝑅 .

Jim Anderson (modified by Nathan Otterness) 20

Proof Sketch for Theorem 4.11
 Start with a DFA for 𝐿.
 Construct an 𝜀-NFA for 𝐿𝑅 as follows:
1. Reverse all of the transitions in the DFA
2. Make the DFA’s start state the only accepting
state.
3. Create a new start state with 𝜀-transitions to all
of the original accepting states.
 This can be proven correct by induction.
 (The book proves Theorem 4.11 by reasoning
about regular expressions.)
Jim Anderson (modified by Nathan Otterness) 21
Example: Reversal of a DFA

 A DFA for 𝐿:

 An NFA with 𝜀-
transitions for 𝐿𝑅 :

q′0

Jim Anderson (modified by Nathan Otterness) 22

Homomorphisms
 A homomorphism maps symbols in an alphabet Σ to
strings over a different alphabet, Δ.
 Example:
❖ Σ = 0, 1 , Δ = a, b
❖ ℎ 0 = ab, ℎ 1 = 𝜀
 Homomorphisms can be extended to strings:
❖ℎ 𝜀 =𝜀
❖ ℎ 𝑥𝑎 = ℎ 𝑥 ℎ 𝑎 , for string 𝑥 and symbol 𝑎
 Homomorphisms can be extended to languages:
❖ If 𝐿 is a language, ℎ 𝐿 = ‫ 𝐿∈𝑥ڂ‬ℎ 𝑥

Jim Anderson (modified by Nathan Otterness) 23

Homomorphism Example
 Σ = 0, 1 , Δ = a, b
 ℎ 0 = ab, ℎ 1 = 𝜀
 Homomorphism of a string:
❖ℎ 0011 = abab
 Homomorphism of a language:
❖ℎ 𝟏𝟎∗ 𝟏 = 𝐚𝐛 ∗

Jim Anderson (modified by Nathan Otterness) 24

Closure under Homomorphism
Theorem 4.14: If 𝐿 is a regular language over Σ, and
ℎ: Σ → Δ∗ is a homomorphism, then ℎ 𝐿 is also a
regular language.
Proof:
 Let 𝐿 = 𝐿 𝑅 be the language defined by some
regular expression 𝑅.
 Replace each symbol 𝑎 in 𝑅 by ℎ 𝑎 . Call the
resulting regular expression ℎ 𝑅 .
 We will prove 𝐿 ℎ 𝑅 =ℎ 𝐿 .
Jim Anderson (modified by Nathan Otterness) 25
Proof that 𝐿 ℎ 𝑅 =ℎ 𝐿

 Let 𝐸 be a subexpression of 𝑅.
 Claim: 𝐿 ℎ 𝐸 =ℎ 𝐿 𝐸 .
 We will prove this by induction on the number of
operators in 𝐸.
 Base case: 𝐸 is 𝜀, ∅, or 𝐚, where 𝑎 ∈ Σ.
❖ The only interesting case is where 𝐸 is a single-
character regular expression, so ℎ 𝐿 𝐸 = ℎ 𝑎 .
❖ℎ 𝐸 is a regular expression for the same string
ℎ 𝑎 , so ℎ 𝐿 𝐸 = 𝐿 ℎ 𝐸 .
Jim Anderson (modified by Nathan Otterness) 26
Proof that 𝐿 ℎ 𝑅 = ℎ 𝐿 , continued

 Inductive step: 𝐸 has at least one operator, and

therefore has the form 𝐹 + 𝐺, 𝐹𝐺, or 𝐹 ∗ .
 Proof for +:
𝐿 ℎ 𝐸
=𝐿 ℎ 𝐹 +ℎ 𝐺 , by the definition of ℎ for reg. exps
=𝐿 ℎ 𝐹 ∪𝐿 ℎ 𝐺 , by the definition of the + operator.

ℎ 𝐿 𝐸
=ℎ 𝐿 𝐹 ∪𝐿 𝐺 , since 𝐿 𝐸 = 𝐿 𝐹 ∪ 𝐿 𝐺 .
=ℎ 𝐿 𝐹 ∪ℎ 𝐿 𝐺 , since ℎ is applied to individual strings.

Jim Anderson (modified by Nathan Otterness) 27

Proof that 𝐿 ℎ 𝑅 = ℎ 𝐿 ,continued

 We have shown the following properties are true for

the + operator:
❖ℎ 𝐿 𝐸 =ℎ 𝐿 𝐹 ∪ℎ 𝐿 𝐺
❖𝐿 ℎ 𝐸 =𝐿 ℎ 𝐹 ∪𝐿 ℎ 𝐺
 To conclude the proof for the + operator, we note
that ℎ 𝐿 𝐹 = 𝐿 ℎ 𝐹 and ℎ 𝐿 𝐺 = 𝐿 ℎ 𝐺 by
the inductive hypothesis. So, ℎ 𝐿 𝐸 = 𝐿 ℎ 𝐸 .
 The proofs for concatenation and Kleene closure are
similar.
Jim Anderson (modified by Nathan Otterness) 28
Inverse Homomorphisms
 If ℎ is a homomorphism from alphabet Σ to
alphabet Δ, and 𝐿 is a language in Δ, then ℎ−1 𝐿 is
the set of strings 𝑤 in Σ ∗ such that ℎ 𝑤 is in 𝐿.
ℎ
A language 𝐿 over Δ∗
ℎ

𝐿 may contain strings

that ℎ can’t map to.
ℎ

Strings over Σ Strings over Δ

Jim Anderson (modified by Nathan Otterness) 29
Inverse Homomorphism: Example
It can be difficult to determine ℎ−1 𝐿 !
Example:
 Σ = 0,1 , Δ = a, b .
 Let ℎ 0 = aa, ℎ 1 = aba.
 𝐿 = 𝐚𝐛 + 𝐛𝐚 ∗ 𝐚.
What is ℎ−1 𝐿 ?
What strings over 0, 1 can ℎ map to a string in 𝐿?

Jim Anderson (modified by Nathan Otterness) 30

Inverse Homomorphism: Example
• Σ = 0,1 , Δ = a, b
• ℎ 0 = aa, ℎ 1 = aba
What is ℎ−1 𝐿 ? • 𝐿 = 𝐚𝐛 + 𝐛𝐚 ∗ 𝐚
 No string in 𝐿 begins with aa, so no string starting with 0
(over 0, 1 ) can be mapped to a string in 𝐿.
 Strings starting with 1 will always be mapped to strings that
start with a, because ℎ 1 = aba.
 Strings in 𝐿 that can start with a:
❖ aba 1 maps to this
❖ abab...
❖ abba… No string starting with 1 maps to these
❖a
 So, ℎ−1 𝐿 = 1 .
Jim Anderson (modified by Nathan Otterness) 31
Closure Under Inverse Homomorphism

Theorem 4.16: If 𝐿 is a regular language, then ℎ−1 𝐿 is

also regular.

Proof:
 Let 𝑀 = 𝑄, Δ, 𝛿, 𝑞0 , 𝐹 be a DFA accepting 𝐿.
 Let ℎ: Σ → Δ∗ .
 Define 𝑀′ = 𝑄, Σ, 𝛿 ′ , 𝑞0 , 𝐹 be a DFA accepting ℎ−1 𝐿 .

Jim Anderson (modified by Nathan Otterness) 32

Closure Under Inverse Homomorphism

𝑀′ = 𝑄, Σ, 𝛿 ′ , 𝑞0 , 𝐹
 On an input 𝑎 ∈ Σ, 𝑀′ simulates the behavior of 𝑀
on ℎ 𝑎 .
 Formally, 𝛿 ′ 𝑞, 𝑎 = 𝛿መ 𝑞, ℎ 𝑎 .

Symbol in Σ String over Δ

 Exercise: Prove by induction on 𝑥 that 𝑀′ accepts

𝑥 if and only if 𝑀 accepts h 𝑥 .

Jim Anderson (modified by Nathan Otterness) 33

Closure Under Inverse Homomorphism

Example. 𝐿 = 𝐚𝐛 + 𝐛𝐚 ∗ 𝐚, ℎ 0 = aa, ℎ 1 = aba.

𝑀: 𝑀′ :

𝐿 𝑀′ = 1

This can be a good strategy for figuring out ℎ−1 (𝐿).

Jim Anderson (modified by Nathan Otterness) 34
Decision Properties
 A property is a yes/no question about one or more
languages. Some examples:
❖ Is 𝐿 empty?
❖ Is 𝐿 finite?
❖ Are 𝐿1 and 𝐿2 equivalent?
 A property is a decision property for regular
languages if an algorithm exists that can answer the
question (for regular languages).

Jim Anderson (modified by Nathan Otterness) 35

Decision Properties
 The book focuses largely on efficiency issues when
discussing decision properties.
 Efficiency is more of a COMP 550 topic (not a
prerequisite), so these slides approach decision
properties slightly differently.
 We will consider the three properties from the
previous slide: emptiness, finiteness, and equivalence.

Jim Anderson (modified by Nathan Otterness) 36

Equivalence
 We want an algorithm that takes two languages, 𝐿1
and 𝐿2 , and determines if they are the same.
The algorithm:
1. Convert 𝐿1 and 𝐿2 to DFAs.
2. Convert 𝐿1 and 𝐿2 to minimal DFAs. (See next slide)
3. Determine if the minimal DFAs are the same.

Jim Anderson (modified by Nathan Otterness) 37

Minimal DFAs
 Section 4.4 of the textbook gives an algorithm that
takes a DFA and outputs a DFA that accepts the
same language, but has a minimal number of states.
 This minimal DFA is unique.
❖ This means that if two different DFAs define the
same language, both will be converted to the
same minimal DFA.
 We will be skipping this algorithm—it takes a long
time to explain and won’t be used later in the class.
You just need to know that it exists.
Jim Anderson (modified by Nathan Otterness) 38
Emptiness and Finiteness
Theorem:
The set of strings accepted by a DFA 𝑀 with 𝑛 states is:
 nonempty if and only if 𝑀 accepts a string of length
less than 𝑛;
 infinite if and only if 𝑀 accepts a string of length 𝑘,
where 𝑛 ≤ 𝑘 < 2𝑛.

(This approach is different from the book.)

Jim Anderson (modified by Nathan Otterness) 39

Proof of “Nonempty” Claim
“…nonempty if and only if 𝑀 accepts a string of length
less than 𝑛”
 “If”: Obvious
 “Only if”: Let 𝑤 be the length of the shortest string
accepted by 𝑀.
❖ If 𝑤 < 𝑛 we’re done.
❖ If 𝑤 ≥ 𝑛, then by the Pumping Lemma 𝑤 = 𝑥𝑦𝑧,
and 𝑥𝑧 ∈ 𝐿 𝑀 . Contradiction. We claimed that 𝑤 was the
shortest string accepted by 𝑀.

Jim Anderson (modified by Nathan Otterness) 40

Proof of “Infinite” Claim
“…infinite if and only if 𝑀 accepts a string of length
𝑘, where 𝑛 ≤ 𝑘 < 2𝑛”
 “If”: If 𝑤 ∈ 𝐿 𝑀 and 𝑛 ≤ 𝑛 < 2𝑛, then 𝐿 𝑀 is
infinite by the Pumping Lemma.
 “Only if”: Assume 𝐿 𝑀 is infinite.
❖ Then, there exists some 𝑤 where 𝑤 ≥ 𝑛. If
such a string has a length under 2𝑛, we’re done.
❖ Otherwise…

Jim Anderson (modified by Nathan Otterness) 41

Proof of “Infinite” Claim, continued
“…infinite if and only if 𝑀 accepts a string of length 𝑘, where
𝑛 ≤ 𝑘 < 2𝑛”
…otherwise, all strings with a length greater than 𝑛 are
longer than 2𝑛. Assume this is the case.
Let 𝑤 be the shortest such string. By the Pumping Lemma,
𝑤 = 𝑥𝑦𝑧, 𝑦 ≥ 1, and 𝑥𝑧 ∈ 𝐿 𝑀 .
 If 𝑥𝑧 ≥ 2𝑛, we’ve contradicted the assumption that 𝑤 is
the shortest string longer than 2𝑛.
 If 𝑥𝑧 < 2𝑛, then 𝑛 ≤ 𝑥𝑧 < 2𝑛. This contradicts the
assumption that the shortest string longer than 𝑛 is longer
than 2𝑛.

Jim Anderson (modified by Nathan Otterness) 42

Decision Algorithms
 Algorithm for “nonemptiness”:
❖ See if any string with length at most 𝑛 is in 𝐿 𝑀 .
❖ Can be done using breadth- or depth-first search
to find paths from the start state to a final state.
 Algorithm for “infiniteness”:
❖ See if any string with length 𝑘 , where 𝑛 ≤ 𝑘 < 2𝑛
is in 𝐿 𝑀 .
❖ More efficient to check for “reachable cycles”.
❖ Can use depth-first search, but it’s less efficient.
Jim Anderson (modified by Nathan Otterness) 43
Another Equivalence Algorithm
We can now build another algorithm for testing
equivalence:
 Let 𝐿1 = 𝐿 𝑀1 and 𝐿2 = 𝐿 𝑀2
❖ 𝑀1 and 𝑀2 are DFAs
 Create 𝑀3 , where 𝐿 𝑀3 = 𝐿1 ∩ 𝐿2 ∪ 𝐿1 ∩ 𝐿2
❖ Note that 𝐿 𝑀3 is nonempty if and only if
𝐿1 ≠ 𝐿2 .
 Test whether 𝐿3 is empty using the
“nonemptiness” algorithm.
Jim Anderson (modified by Nathan Otterness) 44