P versus NP Problem

Safi Khan, Theory of computation,

MS in Computer Sciences, Virtual University of Pakistan
[Software Engineering] Dated: 01-05-2016

Abstract:
There are still unsolvable problems in the computer science history known as P versus NP. The phenomena is that, is
the every given problem solve quickly by a computer program once it verified the solution of that given problem.
The matter posed in 1956 by Kurt Gödel while asking to John von Neumann in a written letter. The question was
arising is that, “whether a certain NP-complete problem could be solved in quadratic or linear time” [2]. In 1971
Stephen Cook one of the most senior computer scientist was introduced very first time the term, “P versus NP” in his
paper titled “The complexity of theorem proving procedure” [3]. That problem is considered as most important in
the computer science field [4], and prize for that solution is in US$1,000,000 announced by the Clay Mathematics
Institute.
The problem solved quickly by a computer program only when it solves the given problem in polynomial
time. That’s mean that there should be an algorithm exists that solve the given problem in polynomial time. If that
kind of an algorithm exists that solve the problem in polynomial time then that given problem is fall in “class P” or
some time just “P”. However there are some problems that do not fall in “class P” or in other words, not solve
“quickly” or “not in polynomial time” then it may be possible that the given answer by showing information, one it
can verify the answer “quickly” or “polynomial time”. That kind of problems falls in “class NP” because only the
answer of these questions can be verified in polynomial time. Where, NP stands for "nondeterministic polynomial
time".
The subset sum problem can be verified easily but it is very difficult to compute its answer. So that problem
can be checkable quickly (NP) but not quickly solvable (P). For example, there are a set of some integers and we
want to know is there any of the subset which has the sum of 0? In practically, if there is a set {-5, -10, -15, -20, -25,
50, 75, 100} and we quickly verify that there are two subset with the sum 0. First one is {-5, -10, -15, -20, 50}, and
the second one is {-5, -10, -15, -20, -25, 75}. When we add the integer of both these sets it will return the answer is
0. Both are quickly verified by adding some integers but there are not any known algorithms that can find these
kinds of subsets in polynomial time. But at the same time there may be an exponential time algorithm can be exists
if and only if when P =NP.
A problem that can be verified in polynomial time can also solved in polynomial time if P = NP. If it is not
then it termed as P ≠ NP, means that there are a problem that is in NP and it is easy to verify and hard to compute.
So that the given problem (subset sum problem) cannot be solved in polynomial time but its answer can be verified
in polynomial time. So the P versus NP is an important problem in history of computer science as well as in
computational theory, so the proof may be profound implication for cryptography, mathematics, artificial
intelligence, algorithm research, multimedia processing, economics, game theory, philosophy and many other
fields.

1 – Introduction: A computer model is required for such kind of

Computational theory defines the complexity analysis which analyzed the required time.
relation between the P and NP classes. The These types of models are deterministic and
resources those are required to solve the sequential. The deterministic models are those
computational problem are deals within theory are taking values as input and a present sate
of computation. These resources are considered and perform only one possible action that a
as time and space. Time specifies the steps to given computer might take. On the other hand,
solve a particular problem and space deals with the sequential models are those that can be
the memory required to solve a particular perform a single action one after the other.
problem.

Virtual University of Pakistan 1

A deterministic sequential machine solves those
decision problems that falls in class P, in
polynomial size of the input. Whereas the class
NP verified the solution of given decision
problems in polynomial time, or in other words
the solution of those problems in be found in
the polynomial time on a non-deterministic
machine [5]. Hence the P ⊆ NP and the
computational theory define the relationship
between these two classes. Is P = NP?
Figure 2: Euler diagram for P, NP, NP-complete, and
A poll of researcher was conduct in 2002 and NP-hard set of problems
also in 2012 to answer that question, in which
many researchers was participates and replied By the Cook-Levin theorem, the entire Boolean
various answer [6]. The table in figure 1 below satisfiability problem falls in the NP-complete
shows the actual figures. The researcher those class. So if a problem in NP then it can be
are not certain about the possibility of prove or mechanically transformed into Boolean
disprove, believe that the question may be satisfiability problem in the polynomial time.
independent of currently accepted axioms [7]. Hence, these Boolean satisfiability problems are
one of the NP-complete problems. The P = NP if
Year 2002 2012 and only if the NP-complete problem is in P, but
No. of Researchers 100 151 there are a large number of problems that are
Yes 9 12 (9%) in NP-complete, and there are not any known
No 61 126 (83%) solution or algorithm for them.
Unsure 22 8 (5%)
Impossible to prove 8 5 (3%) If we examine clearly only on the definition
or disprove then we do not found obviously that the
Figure 1: The pool of researcher in 2002 and 2012 problems of NP-complete are exist. However a
NP-complete problem can be formulated
2 – NP – Complete: trivially as: if we give a description of a Turing
The concept of NP-Complete is very useful in machine M that guaranteed to be halt the
order to find the solution of P = NP. All the NP- Turing machine in polynomial time, so the
problems which can be reduced in the question is that, is there any polynomial-size
polynomial time fall in the set of NP-complete input exists that M will accept[8]? Obviously,
problems, because the solution of these this is in class NP because on given an input one
problems can be verified in polynomial time. In it can simply check via a simulation that M
the set of NP-complete problem each problem accepts the input. It is also in NP-complete due
can be easily transform in any other problem, to the verifier of an instance for a problem that
hence all these problems are equally tough. is in NP, and one it can be encoded with
On the other hand, NP-hard problems are machine M in polynomial-time as an input and
equally hard then NP problems, and can be verified the solution. So that instance is based
reduced in polynomial time by NP problems. on valid input.
There is no need to equate the NP-hard
problems with NP because there is no need to The Boolean satisfiability problem is the very
verify them in polynomial time. Following figure first NP-complete problem to be proven.
2 shows the relationship between the P, NP, Actually this is the Cook-Levin theorem because
NP-complete, and NP-hard problems. it’s proof that the satisfiability is the NP-
complete. The technical details which it

Virtual University of Pakistan 2

contains about the Turing machines can be required for known problem, and this run time
relate to the NP definition. In that way if we can be more critical in case of more difficult
reduce the other problem in Boolean problems, i.e. undecidable problems which are
satisfiability problem than those problems are known as halting problems. The algorithms
also in NP-complete, i.e. subset sum problem. In cannot be solved these problems completely
that way a large class of related problems can because there are a particular input of each
be reduce to one another and treated as same algorithm that produce the wrong answer or do
problem. The following figure 3 shows the not produce right answer or it do not produce
complexity relationship between these classes. the conclusive answer. Some time it may be run
That figure shows that P ≠ NP, the problem forever and do not producing any right answer
within class NP are clearly outside both the at all.
classes of P and NP-complete, the Lander’s
theorem base on that assumption [1]. 4 – Problems in NP not known to be in P or NP-
complete:
Lander believes that if P ≠ NP then there are
some problems exists in NP that many not are
both in P and NP-complete [1]. These problems
known as the intermediate problems, the
example of NP-intermediate problems are;
discrete logarithm problems, isomorphism
Figure 3: complexity relationship between the problems and integer factorization problems.
classes. However there are some problems in NP that
are still not known if it falls in class P or NP-
3 – Harder problems: complete.
Yet it is unknown that whether P = NP, but the
problems outside of P are known. There are a The graph isomorphism problem which is
small number of problems that are in EXPTIME- known as computational problem basically are
complete. These kinds of problems are not used to determination weather two graphs are
operating on the normal input; its input isomorphic or not. That is the unsolved problem
requires the computational description. At the in the theory of complexity. That is no known
same time these problems are treated as P ≠ whether the graph isomorphic problems in P,
EXPTIME, hence they are outside of class P, NP-complete, or NP-intermediate. The scientist
because more than polynomial time is required sill not believes but they sure that the problem
to solve them. In fact, if we examine the in not in class NP-complete [11]. The hierarchy
theorem of time hierarchy then it can found of polynomial time is collapses to the second
that these problems cannot be solved in less level if the class NP-complete contains the
than exponential time significantly. Board graph isomorphic problem [12] [13]. Sine at any
games, i.e. chess (on the N x N board) [9] are finite level the hierarchy of polynomial does not
examples to find a perfect strategy [10]. collapse and at the same time the class NP-
complete do not contain the graph
On the other hand, the problems in Presburger isomorphism problems. For the graph that have
arithmetic statements requires more time to n vertices, Laszlo Babi and Eugene Lucks
finding the truth. Rabin and Fischer found in denotes the run time 2O(√nlog(n)) as an best
1974 that the decision algorithm to known the algorithm.
truth of Presburger statements requires 22^cn,
where c is some constant and the length of the Another computational problem that
Presburger statement denoted by the n. Hence, determines the prime factorization of a given
there are more than exponential run time is integer is known as integer factorization

Virtual University of Pakistan 3

problem. Like the decision problem this well in practice; however it have worst-case
problem also is decided that if input has factor time complexity exponentially but it known best
less than k. Like RSA algorithm, that factor is polynomial-time algorithms [16].
based on sever modern cryptographic system.
Although, still yet there are not any best known Second, some computation problems that do
integer factorization algorithm is exists. The not conforms the model of the Turing machine
class NP and co-NP contains the integer in which the P and NP are defined; these
factorization problem, sometime known as UP problems are as randomized algorithms as well
and co-UP [14]. The general number field sieve as quantum computation.
is known as best integer factorization algorithm
and it takes the expected time to factor an n-bit
integer.

O(exp ((64n/9 log(2))1/3 (log(nlog(2)))2/3 ))

Shor’s algorithm is also known as best quantum

algorithm to solve that problem, does not run in
polynomial time. With respect to non-quantum
complexity classes that problem does not sure
about the problem where it is lies. Figure 4: Time vs. problem size for knapsack
problem.
5 – Does P mean "easy"?
While reading the above discussion it is In the figure 4, the graph shows the 100
assumed that the problems those are fall in instances in ms which are using a 933 MHz
class P mean “easy”, and the problem those are Pentium III, which is as a state-of-art specialized
not fall in class P means as a “hard”, so these algorithm of time vs. problem size for knapsack
assumptions known as Cobham’s thesis. Some problems. The empirical algorithmic complexity
people or literatures known these assumptions which is suggests as quadratic fit for instances
accurate reasonably in complexity theory but with 50–10,000 variables is O((log(n))2).[15]
some time it is assume that it has some caveats.
First of all in practical that is not true always 6 – Reasons to believe P ≠ NP:
because a polynomial algorithm may have Many computer scientists, according to the
theoretically a constant factors or exponents polls held *6+*17+, believes the P ≠ NP. The
which are extremely large and thus most of the reason is that still now no one can find the
time it rendered as impractical. However, in polynomial time algorithm, after a decades
practice there may still some effective studying, i.e. studying more than three
approaches that can tackle the problem if it is in thousand NP-complete problems. These kinds
NP-complete and even if P ≠ NP. For NP- of algorithms were difficult to search before the
complete problems there are still many NP-complete concept. NP = co-NP and P = PH
algorithms that can be solve the many of real- were believed as false, so the P = NP result
world problem in the reasonable amount of startling then. The problems exist that are hard
time; such as traveling salesman problem, to solve but the solutions of these problems are
Boolean satisfiability problem and knapsack easy for the verification also argued intuitively
problem. For such kind of algorithms the to matches the real-world experience [18].
empirical average-case complexity (which is
based on time vs. problem size) can be very low According to the Scott Aaronson, MIT:
surprisingly. For instance the linear “If P = NP, then the world would be a profoundly
programming the simplex algorithm that works different place than we usually assume it to be.

Virtual University of Pakistan 4

There would be no special value in "creative no leads to an efficient methods. So both the
leaps," no fundamental gap between solving a positive and negative consequences arise due
problem and recognizing the solution once it's to these NP-complete problems which are
found.” fundamental in nature in many of fields.

At the same time some of the other researchers Cryptography which relies some of the certain
believe that some researchers are problem being difficult, 3-SAT which have
overconfident about the P ≠ NP believing and constructive as well as efficient algorithm break
they should try to explore the proof for P = NP. down the cryptosystems such as;
Following statements were made in 2002 as [6]: • Public-key cryptography: that is the
foundation of most important security
According to the Moshe Y. Vardi, Rice application systems [19], i.e. financial
University: transaction on internet.
• Symmetric ciphers: i.e. AES or 3DES uses for
“The main argument in favor of P ≠ NP is the communication data encryption.
total lack of fundamental progress in the area of • Cryptographic hashing: it is a one way
exhaustive search. This is, in my opinion, a very function, i.e. find pre-image hashes to a
weak argument. The space of algorithms is very given value [21], sometime difficult to use
large and we are only at the beginning of its due to exponential time. Via reduction to
exploration. [...] The resolution of Fermat's Last SAT then finding pre-image can done in
Theorem also shows that very simple questions polynomial time if and only if P = NP [22].
may be settled only by very deep theories.”
There are some other positive consequences
According to the Anil Nerode, Cornell that may be enormous and rendering
University: intractable problems mathematically. For
example a lot of problems in operation research
“Being attached to a speculation is not a good are NP-complete, i.e. travelling salesman
guide to research planning. One should always problem, integer programming. Efficient
try both directions of every problem. Prejudice solution for these kinds of problems is the
has caused famous mathematicians to fail to logistically implicated. A lot of other problems
solve famous problems whose solution was like protein structure predication are also in NP-
opposite to their expectations, even though they complete [23]. If the solution of this problem
had developed all the methods required.” found efficiently than one it could be a
considerable advancement in the life of
7 – Consequences of solution: biotechnology as well as science.
There are some factors that may attract the
consequences of answer; it may be either that But at the same time such kind of changes may
enormously the resolution direction would be less significant than the methods that
the advance theory or it may have huge evolutionally efficient to solve the NP-complete
practical consequences as well. problems that would be cause in mathematics
itself. In the computational complexity the
7.1 – P = NP: Gödel thoughts the mechanical methods than
If we can solve important problems that belong can be solved any problem would be
the NP then in that case the proof of P = NP is considerable revolution mathematics [24] [25].
practically best consequences. If somehow the
polynomial bounds are very large in practice The greatest important consequences where a
that are not efficient or the proof is not to be machine rely is φ(n) ∼ k ⋅ n (or even ∼ k ⋅ n2).
constructive then in that case the proof possibly That’s mean it is the undecidability of

Virtual University of Pakistan 5

Entscheidungs problem, as the mathematician’s other instances that are selected randomly can
mental work concern only on Yes or No be solved efficiently. To deal with the question
questions and this is replaced completely by a of average-case complexity the Russell
machine. If we choose the natural number n is Impagliazzo describes the five hypothetical
so large, then one possibility is that the machine worlds that lead to the different possible
does not provide the result that is it makes no resolutions [29]. In this regard three ranges can
sense to think more about the problem. be defined as Algorithmica, Cryptomania,
Heuristica;
Stephen Cook says:
“It would transform mathematics by allowing a Algorithmica: In that case the P = NP and SAT
computer to find a formal proof of any theorem problems can be efficiently solved by all its
which has a proof of a reasonable length, since instances.
formal proofs can easily be recognized in
polynomial time. Example problems may well Cryptomania: In that case P ≠ NP and hard
include all of the CMI prize problems.” [26] instances of the problems that are generated
outside P is easy. Different possible distributions
Mathematicians as well as the researchers of difficulty are reflecting the three possible
spend their life and careers to prove the intermediate all over the instances of NP-hard
theorems. Some of the proofs took centuries to problems.
resolves, i.e. Fermat’s Last Theorem took near
about three centuries to prove. The reasonable Heuristica: The "world" where P ≠ NP but at the
size of the problem is main concern while same time all possible problems in NP are
guarantee to proof a theorem. tractable in the average case, in the paper “A
Princeton University workshop in 2009 studied
A researcher Donald Knuth believes that the P = the status of the five worlds” [30].
NP but is reserved about the impact of the
possible proof *27+, “I don't believe that the 8 – Results about difficulty of proof:
equality P = NP will turn out to be helpful even Is the P = NP? This is however million dollar
if it is proved, because such a proof will almost prize, a huge amount offered to researcher
surely be non-constructive”. solve this problem, some efforts are made and
it have led to various other new approaches.
7.2 – P ≠ NP: Some of the advance researches about the P
The computational benefits that P = NP =NP show that the techniques that has already
provides is reduced by the proof of P ≠ NP, but existing are not enough and powerful which can
at the same time it provide the computational answer our question, so some of the novel
advancement in the complexity theory as well techniques are required.
as guidance for future research. Formally it
shows that we cannot solve the common The problem is really difficult, so the all known
problems efficiently. At that point the partial proofing techniques essentially computational
solution becomes the attractive by the complexity theory can be categorize in one of
researchers. So the believe-ness of P ≠ NP by a the following class, however each of which is
lot of researchers is the result to focusing to failed to proof that P ≠ NP.
other similar problems [28].
8.1 – Relativizing proofs:
So the focusing the average-case complexity Suppose a scenario in which each algorithm
some of the harder problems in the class NP is made queries to a fixed subroutine called
due to that believe-ness of P ≠ NP. SAT requires oracle, and the running time of an algorithm in
the exponential time but on the other hand the not counted against the running time of oracle.

Virtual University of Pakistan 6

Most of the classical proofs can be apply These are some massive barriers which leads to
uniformly with the oracles regardless of what computer scientists that P vs NP may be
the oracle does. These proofs are known as independent in nature of standard axiom
relativizing. With the respect of oracle, in 1975 system, i.e. ZFC which cannot be prove or
some researchers namely, Baker, Gill and disprove. By examining these independent
Solovay proof that P =NP, while P ≠ NP for other result, one it conclude that there are no
oracles [31]. These relativizing proofs can prove polynomial-time algorithm exists for the NP-
only uniformly true statements regard to oracle, complete problems, so the construction of the
and that way it shows that the relativizing proof is failed, i.e. ZFC, or if the polynomial-time
techniques cannot be solve the P =NP. algorithm exists for NP-complete problem exists
but it is impossible to prove [33]. Using
8.2 – Natural proofs: technique of sort one it can be shown that the
For the circuit complexity of lower bounds, a problem is undecidable even with the weaker
general class of proving techniques is defined by assumption with extending the Peano axioms
the Alexander Razborove and Steven Rudich in (PA) for the integer arithmetic, then for every
1993, which is known as natural proof. This NP problem the nearly-polynomial-time
circuit complexity approach was well known to algorithms exist [34]. However it is believes that
resolve the P = NP, since previously all known there are not efficient algorithms for the entire
circuit techniques of lower bounds were problem in NP leads to the phenomena that
natural. Both of the researcher shows that the independence proof using those techniques
no method of natural proof can be distinguish cannot be possible. This shows that using those
between P and NP, if one-way functions exist. know techniques to prove the independence for
However, it never been proven that formally it PA or ZFC are not easier than proving that the
exist, however one it believe that they do by efficient algorithms exists for all NP problems.
most of the mathematicians. At the same time
the proof or disproof of their existence is much 9 – Claimed solutions:
stronger then quantification of P relative to NP. Some of the researcher claimed the solution of
So it is unlikely that these natural proofs can unsolved problem of P vs NP [35]. A
alone resolve the P =NP. comprehensive list is held by Gerhard J.
Woeginger [36]. Vinay Deolalikar and Palo Alto
8.3 – Algebrizing proofs: also claimed the proof in August 2010 that P ≠
Some result of Baker, Gill and Solovay shows NP and they got heavy attention on the internet
that newly techniques which are non- and press [37]. Publicly this proof has been
relativizing can be successfully used to prove reviewed by some academics [38][39] and an
that IP = PSPACE. Scott Aaronson and Avi expert Neil Immerman pointed out some of the
Wigderson in 2008 shows that some of the fatal error in that proof [40].
technical tool which used to proof of IP =
PSPACE known as arithmetization was Deolalikar work in detail on that proof in
insufficient also to resolve P = NP [32]. September 2010, [41] and the other studies
show by some of the theoretical computer
All of these barriers are another reason that scientists that the proof is not correct nor have
why the NP-complete problems are very useful. any significant advancement to understand the
If the demonstration of the polynomial-time problem [42]. In 2013 the assessment is
algorithm for these NP-complete problems than prompted by The New Yorker article to call that
one it could solve P = NP problem, including the proof attempt thoroughly discredited [43].
above result.

Virtual University of Pakistan 7

10 – Logical characterizations: hand, if the answer is “no” the algorithm runs
Some of the certain classes which contain the forever. If P = NP then the algorithm is
logical statements to measure the descriptive enormously impractical. The algorithm try to
complexity to restated the problem of P = NP. other programs first at least in 2b – 1 if a short
For that all languages with the finite structures program that can be solve the SUBSET-SUM in
is considered in a linear order relation. Then all the polynomial time if it is b bits long.
the classes which fall in the language P could be // Algorithm that accepts the NP-complete
express in 1st order logic with the some addition language SUBSET-SUM. This is a polynomial-
of some suitable combinatory of fixed-point. time algorithm if and only if P = NP.
This combinational order allows the recursive "Polynomial-time" means it returns "yes" in
functions definition effectively. At least one polynomial time when the answer should be
function or predicates contain by the signature "yes", and runs forever when it is "no"
which additionally distinguished the order // Input: S = a finite set of integers
relation. These kinds of functions take // Output: "yes" if any subset of S adds up to 0.
polynomial space amount to store the finite // Runs forever with no output otherwise.
structures, characterizes as P. // Note: "Program number P" is the program
obtained by
The languages that are logically fall in 2nd order // writing the integer P in binary, then
in class NP. The functions, relations and subset // considering that string of bits to be a
by the universal quantification are however // program. Every possible program can be
excluded in that 2nd order logic. As well as that // generated this way, though most do nothing
2nd order logic is related to all the languages // because of syntax errors.
that are in polynomial hierarchy (PH). So the
question arises is that, is P which is a subset of FOR N = 1...∞
NP can be reformulated as; is the 2nd order logic FOR P = 1...N
existentially able to describe the languages that Run program number P for N steps with input
the 1st order logic cannot with the fixed point? S
[44] That language is based on finite linearly IF the program outputs a list of distinct
ordered structures with nontrivial signature. In integers
the previous characterization the word AND the integers are all in S
“existential” can be dropped, sine P = NP ↔ P = AND the integers sum to 0
PH as we establish that NP = co-NP implies that THEN
NP = PH. OUTPUT "yes" and HALT

11 – Polynomial time algorithms: 12 – Formal Definition:

For the NP-complete problem there is no The formal definition contains the following
polynomial time algorithm exists. However if P points.
= NP then the polynomial time algorithm for
NP-complete may be exists with some 12.1 – P and NP:
enormous constant which make the algorithm A decision problem takes string ‘w’ as an input
impractical. Levin represent the following from an alphabet Σ and output the Boolean
algorithm which accepts the NP-complete value, i.e. “yes” or “no”. If there are a TM, i.e.
SUBSET-SUM language correctly without any an algorithm that takes the input string of
citation in the polynomial time if and only if P = length n and produce the answer correctly at
NP. most cnk steps (k and c both are constant and
independent of the input string), then one can
The algorithm runs in polynomial time if it is guess that the problem can be solved in the
accepting, i.e. answer is “yes”. On the other polynomial time that fall in class P (set of the

Virtual University of Pakistan 8

languages which can be decided by a 12.1.1 – Example:
deterministic polynomial time TM. For example let;
COMPOSITE  x  N | x  pq for integers p,q > 1
P = { L : L  L  M  for some deterministic
polynomial-time Turing machine M}
R  x, y   N  N |1  y  
x and y divides x
The question is that if x is composite is same as
Where L  M   w   *: M accepts w
if x is member of composite, where COMPOSITE
∈ NP by the definition of the verifier satisfying
Hence, a deterministic polynomial-time TM is a the definition narrated above. This can be done
deterministic TM M that satisfies the only if the natural numbers can be identifying
conditions; via binary representations. However via [45]
1. M halts on all input w [46] COMPOSITE can also be happened in P.
2. k  n Such that TM  n   O  nk  ; here O refers
to big O notation and, 12.2 – NP-completeness:
TM  n   max tM  w : W   *,| w | n There are many ways for describing the
concepts of NP-completeness.
tM  w = number of steps M takes to halt on
Let L be a language over a finite alphabet 
input w The following two conditions should be satisfied
to prove that L in NP-complete.
NP can be defined traditionally by 1. L  NP
nondeterministic TM. The major concepts of 2. L' in NP is polynomial time reducible to L as
certificate and verifier can be used to define the
L'  p L if and only if following two
NP in modern way. Verifiers in form of set of
languages contain the finite alphabet is used to conditions satisfied.
defined NP formally which runs in polynomial a. There exists f :  *   * such that for all
time. The notion of verifier is defined as: w in  * we have:
f :  *   * w  L'  f  w  L 
Let L be a language over a finite alphabet, Σ. b. There exists a polynomial-time TM that
halts with f  w on its tape on any input w
L ∈ NP if, and only if, there exist a binary
relation R   *  * and a positive integer k
13 – Conclusion:
such that the following two conditions are
The entire problems that falls in NP-
satisfied:
completeness have some order parameter (at
1. For all x   *, x  L  y   * such that
least one), and some time hard to solve due to
 x, y   R and  x, y   R  
y O x
k
those critical values that are near around to the
order parameter. Separation of a region from
2. The language LR  x # y :  x, y   R over
one to other is performed by those critical
 # is decidable by a TM in values such as over and under constrained
polynomial time. regions of the problem space. In that situation
the transition in phase occurs 0 to 1 due to
LR is the verifier for L of the TM, where y is the change in solution probability. However the
certificate such that  x, y   R . In that case the phase transition does not arise in P problems
verifier does not have the polynomial time, but unless it arises to bounded N, so it has bounded
verifier should be run in polynomial time for the cost.
L to be in NP.

Virtual University of Pakistan 9

For the particular problem process I have [7] William I. Gasarch. "The Second P=?NP poll"
provide the conjectures as empirical evidence. I (PDF). SIGACT News 74.
also show that one hard problem in one space [8] Scott Aaronson. "PHYS771 Lecture 6: P, NP, and
can be map to another hard problem in the Friends". Retrieved 27 August 2007.
[9] Aviezri Fraenkel and D. Lichtenstein (1981).
other space. In that way preserve the phase
"Computing a perfect strategy for n×n chess
boundary under the mapping. I have also shown requires time exponential in n". J. Comb. Th. A
some of the cases where P and NP classes have (31): 199–214.
the distinction and also show that how the P [10] David Eppstein. "Computational Complexity of
classes was excluded from the critical region. Games and Puzzles"
[11] Arvind, Vikraman; Kurur, Piyush P. (2006).
In that cases where the conjectures are true, "Graph isomorphism is in SPP". Information and
then an NP problem turn into the P problem Computation 204 (5): 835–852.
adding some restrictions excluding the order [12] Schöning, Uwe. "Graph isomorphism is in the
parameter of the critical values. low hierarchy". Proceedings of the 4th Annual
Symposium on Theoretical Aspects of Computer
Science 1987: 114–124.
Note that only the reduced problems carry [13] Schöning, Uwe (1988). "Graph isomorphism is in
these results explaining why not these results the low hierarchy". Journal of Computer and
previously noticed particularly. So there are a System Sciences 37: 312–323.
lot of outstanding question arises, such that: [14] Lance Fortnow. Computational Complexity Blog:
1. What happened with NP-hard problems? Complexity Class of the Week: Factoring. 13
2. Is there any situation where the hard September 2002.
problem occurs in the non-critical region? [15] Pisinger, D. 2003. "Where are the hard knapsack
3. Do the other types of problems, i.e. games, problems?" Technical Report 2003/08,
optimization problems, etc. have the same Department of Computer Science, University of
Copenhagen, Copenhagen, Denmark
properties?
[16] Gondzio, Jacek; Terlaky, Tamás (1996). "3 A
computational view of interior point methods".
14 – References: In J. E. Beasley. Advances in linear and integer
[1] R. E. Ladner "On the structure of polynomial programming. Oxford Lecture Series in
time reducibility," Journal of the ACM, 22, pp. Mathematics and its Applications 4. New York:
151–171, 1975. Corollary 1.1. Oxford University Press. pp. 103–144. MR
[2] Hartmanis, Juris. "Gödel, von Neumann, and the 1438311. Postscript file at website of Gondzio
P = NP problem" (PDF). Bulletin of the European and at McMaster University website of Terlaky
Association for Theoretical Computer Science [17] Rosenberger, Jack (May 2012). "P vs. NP poll
38: 101–107. results". Communications of the ACM 55 (5): 10.
[3] Cook, Stephen (1971). "The complexity of [18] Scott Aaronson. "Reasons to believe". point 9.
theorem proving procedures". Proceedings of [19] See Horie, S. and Watanabe, O.; Watanabe
the Third Annual ACM Symposium on Theory of (1997). "Hard instance generation for SAT".
Computing. pp. 151–158. Algorithms and Computation. Lecture Notes in
[4] Fortnow, Lance (2009). "The status of the P Computer Science (Springer) 1350: 22–31.
versus NP problem" (PDF). Communications of arXiv:cs/9809117. Bibcode:1998cs........9117H.
the ACM 52 (9): 78–86. doi:10.1007/3-540-63890-3_4. ISBN 978-3-540-
doi:10.1145/1562164.1562186. 63890-2. for a reduction of factoring to SAT. A
[5] Sipser, Michael: Introduction to the Theory of 512 bit factoring problem (8400 MIPS-years
Computation, Second Edition, International when factored) translates to a SAT problem of
Edition, page 270. Thomson Course Technology, 63,652 variables and 406,860 clauses.
2006. Definition 7.19 and Theorem 7.20. [20] See, for example, Massacci, F. and Marraro, L.
[6] William I. Gasarch (June 2002). "The P=?NP (2000). "Logical cryptanalysis as a SAT problem".
poll." (PDF). SIGACT News 33 (2): 34–47. Journal of Automated Reasoning (Springer) 24
doi:10.1145/1052796.1052804. Retrieved 29 (1): 165–203. doi:10.1023/A:1006326723002.
December 2008. CiteSeerX: 10.1.1.104.962. in which an instance

Virtual University of Pakistan 10

 of DES is encoded as a SAT problem with 10336 [35] John Markoff (8 October 2009). "Prizes Aside,
variables and 61935 clauses. A 3DES problem the P-NP Puzzler Has Consequences". The New
instance would be about 3 times this size. York Times.
[21] Find a messageM that when hashed by the [36] Gerhard J. Woeginger. "The P-versus-NP page".
function H() gives a digest h, or H(M)=h Retrieved 25 May 2014.
[22] De, Debapratim and Kumarasubramanian, [37] Markoff, John (16 August 2010). "Step 1: Post
Abishek and Venkatesan, Ramarathnam (2007). Elusive Proof. Step 2: Watch Fireworks.". The
"Inversion attacks on secure hash functions New York Times. Retrieved 20 September 2010.
using SAT solvers". Springer. pp. 377–382. [38] Polymath Project wiki. "Deolalikar's P vs NP
[23] Berger B, Leighton T (1998). "Protein folding in paper"
the hydrophobic-hydrophilic (HP) model is NP- [39] Science News, "Crowdsourcing peer review"
complete". J. Comput. Biol. 5 (1): 27–40. [40] Dick Lipton (12 August 2010). "Fatal Flaws in
doi:10.1089/cmb.1998.5.27. PMID 9541869. Deolalikar's Proof?"
[24] History of this letter and its translation from [41] Dick Lipton (15 September 2010). "An Update
Michael Sipser. "The History and Status of the P on Vinay Deolalikar's Proof". Retrieved 31
versus NP question" December 2010
[25] David S. Johnson. "A Brief History of NP- [42] Gödel’s Lost Letter and P=NP, Update on
Completeness, 1954–2012" (PDF). From pages Deolalikar’s Proof that P≠NP
359–376 of Optimization Stories, M. Grötschel [43] Alexander Nazaryan (2 May 2013). "A Most
(editor), a special issue of ¨ Documenta Profound Math Problem". Retrieved 1 May 2014
Mathematica, published in August 2012 and [44] Elvira Mayordomo. "P versus NP" Monografías
distributed to attendees at the 21st de la Real Academia de Ciencias de Zaragoza 26:
International Symposium on Mathematical 57–68 (2004)
Programming in Berlin. [45] M. Agrawal, N. Kayal, N. Saxena. "Primes is in P"
[26] Cook, Stephen (April 2000). "The P versus NP (PDF). Retrieved 29 December 2008
Problem" (PDF). Clay Mathematics Institute. [46] AKS primality test
Retrieved 18 October 2006. [47] Geere, Duncan. "'Travelling Salesman'
[27] Knuth, Donald E. (May 20, 2014). "Twenty movie considers the repercussions if P
Questions for Donald Knuth". informit.com. equals NP". Wired. Retrieved 26 April 2012
InformIT. Retrieved 20 July 2014.
[28] L. R. Foulds (October 1983). "The Heuristic
Problem-Solving Approach". Journal of the
Operational Research Society 34 (10): 927–934.
doi:10.2307/2580891. JSTOR 2580891
[29] R. Impagliazzo, "A personal view of average-case
complexity," sct, pp.134, 10th Annual Structure
in Complexity Theory Conference (SCT'95), 1995
[30] "Tentative program for the workshop on
"Complexity and Cryptography: Status of
Impagliazzo's Worlds"". Archived from the
original on 2013-11-15.
[31] T. P. Baker, J. Gill, R. Solovay. Relativizations of
the P =? NP Question. SIAM Journal on
Computing, 4(4): 431–442 (1975)
[32] S. Aaronson and A. Wigderson (2008).
Algebrization: A New Barrier in Complexity
Theory (PDF). Proceedings of ACM STOC'2008.
pp. 731–740. doi:10.1145/1374376.1374481.
[33] Aaronson, Scott. "Is P Versus NP Formally
Independent?"
[34] Ben-David, Shai; Halevi, Shai (1992). "On the
independence of P versus NP". Technical Report
714. Technion.

Virtual University of Pakistan 11