The paper summarizes the contributions of the three philosophies of mathematics—logicism, intuitionism-constructivism (constructivism for short) and formalism and their rectification—which constitute the new foundations of mathematics. The critique of the traditional foundations of mathematics reveals a number of errors including inconsistency (contradiction or paradox) and undefined and vacuous concepts which fall under ambiguity. Critique of the real and complex number systems reveals similar defects all of which are responsible not only for the unsolved long standing problems of foundations but also of traditional mathematics such as the 379-year-old Fermat’s last theorem (FLT) and 274-year-old Goldbach’s conjecture. These two problems require rectification of these defects before they can be resolved. One of the major defects is the inconsistency of the field axioms of the real number system with the construction of a counterexample to the trichotomy axiom that proved it and the real number system false and at the same time not linearly ordered. Indeed, the rectification yields the new foundations of mathematics, constructivist real number system and complex vector plane the last mathematical space being the rectification of the complex real number system. FLT is resolved by a counterexample that proves it false and the Goldbach’s conjecture has been proved both in the constructivist real number system and the new real number system. The latter gives to two mathematical structures or tools—generalized integral and generalized physical fractal. The rectification of foundations yields the resolution of problem 1 and the solution of problem 6 of Hilbert’s 23 problems.
The paper summarizes the contributions of the three schools of thought or philosophies of mathematics―logi- cism [
Until today, mathematicians have not grasped the significance of Hilbert’s discovery to the extent that all textbooks in mathematics assume 1 = 0.99××× When this issue was raised in internet forums in 1997, especially, SciMath, it sparked a howl of protest and controversy with a whisk of name calling that lasted over a decade and spilled over into many websites.
The full critique-rectification of traditional mathematics and its foundations done in [
One of the field axioms of the real number system [
Given two rationals x, y we can tell if x < y or x > y. Even then, we cannot line up all the rationals on the real line under the ordering < due to the ambiguity of the infinite number of rationals between any two given rationals and this is due to the ambiguity of the concept infinity. However, we can proceed with the following scenario: start with a certain rational interval [A, B] with A < C < B, and find a nested sequence of rational intervals [An, Bn], with
Since the nonterminating decimal C is defined to be the limit of a sequence of rationals (from the left and from the right), we can choose the end points An, Bn, of intervals [An, Bn] as members of two sequences {An}, {Bn} where {An} is a monotonic increasing sequence and {Bn} a monotonic decreasing sequence of rationalssatisfying
The counterexample says that the real number system is not linearly ordered by the relation “<” and collapses traditional calculus. We summarize the requirements for error-, ambiguity- and contradiction-free mathematical space; the sources of these defects are laid out in [
1) Every concept is defined by the axioms. Although undefined concepts may be introduced initially in the construction of a mathematical space the choice of the axioms is not complete until every concept is defined. While the choice of axioms is arbitrary depending on what the mathematical space is intended for, once chosen the mathematical space becomes deductive, i.e., every theorem follows from the axioms.
2) Avoidance of vacuous concept such as, root of the equation x2 + 1 = 0, denoted by
3) Avoidance of infinity that cannot be contained by replacing its traditional definition with the concept inexhaustibility as its essential property so that an infinite set cannot be contained in a finite set and if one tries to put its elements in a finite set some element will be left out at every step. This property of infinite set invalidates the axiom of choice when applied to infinite set [
4) Avoidance of self-reference [
Requirement 1) invalidates proof of theorem involving concepts from two distinct mathematical spaces, e.g., Gödel’s incompleteness theorems [
By meeting all the requirements of the new foundations of mathematics stipulated in [
The new methodology―qualitative mathematics―of the new mathematics is the mathematical or qualitative model of rational thought [
We continue the critique of mathematics beyond [
The first constructivist mathematical spaces are the modern calculus of variations, optimal control theory and functional analysis built on generalized curves and surfaces discovered and developed by L. C. Young in a series of papers that started in the 1930s [
This paper, its extensions to other mathematical spaces [
Axiom 1. R*contains the elements 0, 1.
Axiom 2. The addition table (
Axiom 3. The multiplication table (
Axiom 1 says 0, 1 Î R*; they are the additive and multiplicative identities defined by
Since the figure define finite sums and products the laws of addition and multiplication of arithmetic as well as the laws of signs which follow from the automorphism between the positive and negative numbers can be verified from them and need not be taken as axioms. In either case, finite mathematical induction can be applied if needed. The system of Hindu-Arabic numerals quantitatively models the metric system of measurement; so does the scientific notation.
The additive inverse of an integer x, denoted by −x, satisfies the equation,
The prefix “−” is called the negative sign and “−x” is called “negative or additive inverse of x; when a decimal has no prefix it is understood that its prefix is “+” and the decimal is positive. The difference between integers x and y, denoted by x−y, is defined as x + −y. This operation called subtraction is the inverse of addition. It is clear that the additive inverse of the additive inverse of a decimal x, i.e., − −x (this notation is confusing and we replace it by −(−x)) which is x. This notation is consistent with the distributivity of signs with respect to addition which can be checked from
To avoid confusion, we may write the product of two integers a and b as a(b) or ab. The multiplicative inverse of a nonzero integer x, denoted by 1/x, (called reciprocal of x) satisfies,
Provided x does not have a prime factor other than 2 and 5. The quotient of two integers x and y, denoted by x/y, where y has no prime factor other than 2 and 5, is a number z that satisfies the equation, x = yz. Note that if x, y are relatively prime and the divisor has a prime factor other than 2 and 5 the quotient is not defined being a nonterminating decimal. Thus, only terminating decimals are defined in R. When the quotient of x by y is defined we denote it by x/y.
In scientific notation we write an integer N as follows:
where the ajs,
where
A sequence of terminating decimals of the form,
where N is a nonterminating decimal, is the g-limit (g-lim) of its nth g-term,
as n ® ¥. This is called the g-nom of the nonterminating decimal (6); thus, the g-norm of a nonterminating decimal is the decimal itself. This, a nonterminating decimal has been defined for the first time and R* and every digit is computable. For example, the digits of p can be computed from its infinite series expansion [
We formally define an integer as the integral part of adecimal. The concept is no longer vacuous in R* since the decimals have been completely defined. The field axioms of R and Peano’s axioms [
In R the rationals coincide with the terminating decimals which are periodic. Since being nonterminating nonperiodic is not verifiable, the concept nonterminating nonperiodic decimal is vacuous and does not exist. Therefore, this concept is ill-defined, ambiguous. There are special nonterminating decimals which can be computed aside from p such as the natural logarithmic base e which can be computed from their series expansions; likewise, radicals can be computed [
The nth g-term of a nonterminating decimal repeats every preceding g-term so that if finite initial g-terms are deleted the g-terms and g-limit of the remaining g sequence are unaltered. Thus, a nonterminating decimal has many g-sequences belonging to the equivalence class of its g-limits.
Since addition and multiplication and their inverse operations, subtraction and division, are defined only on terminating decimals computing a nonterminating decimal can only approximate it by the nth g-term (n-trunca- tion) of g-term. The same approximation holds for the difference, product and quotient (if defined). Computation involving terminating decimals alone is exact unless the result is nonterminating. This scheme holds for any combination of binary operations. Thus, we have retained standard computation but avoided the ambiguity and contradictions of the real number system the only difference being that every concept is defined in R*. We have also avoided vacuous approximation because nonterminating decimals are g-limits of their g-sequences which exist and belong to R*. Moreover, we have contained the ambiguity of nonterminating decimals by approximating them by their nth g-terms. This is an example of containment of ambiguity which is admissible in a mathematical space. In fact, containment of ambiguity is a standard technique in approximation theory. Converting an ordinary sequence to a g-sequence is obvious.
As we raise n in (6), the tail digits of the nth g-term of any decimal recedes to the right indefinitely, i.e., it becomes steadily smaller until it is unidentifiable from the tail digits of the rest of the decimals. Although it tends to 0 in the standard norm it never reaches 0 in the g-norm since the tail digits are never all equal to 0; it is also not a decimal since the digits are not fixed nor is it a real number. It is a set-valued number belonging to R* called algebraic continuum (continuum for short) denoted by d*; it is distinct from the topological continuum defined in terms of open sets. R* is an extension of R (the terminating decimals) so that every real number is an element of R*. The continuum d* is a special element of R* not belonging to R.
We introduce the general exponent and exponential (base e) where the exponent is allowed to take values along nonterminating decimals. Again, the exponent is the g-limit of the g-sequence of terminating decimals so that it is approximated by the nth g-sequence at desired error just as a non-terminating decimal is approximated.
For example, if a is a number,
Consider the sequence of decimals,
where δ is any of the decimals, 0.1, 0.2, 0.3, ×××, 0.9, a1, ×××, ak, and the digits are not all 0 simultaneously. We call the nonstandard sequence (7) d-sequence and its nth term the nth d-term. For fixed combination of δ and the ajs,
Like a nonterminating decimal, an element of d* is unaltered if finite d-terms are altered or deleted from its d-sequence. When δ = 1 and
obtained from the iterated difference,
excess remainder of 0.1;
Taking the g-limits of the extreme left side of (9) and recalling that the g-limit of a decimal is itself and denoting by dp the d-limit of the principal d-sequence on the rightmost side we have,
Since all the elements of d* share its properties then whenever we have a statement “an element d of d* has property P” we may write “d* has property P”, meaning, this statement is true of every element of d*. This applies to any equation involving an element of d*. Therefore, we have,
Like a decimal, we define d* as the d-norm of d*, i.e., the d-norm of d* is d*. We state our findings:
Theorem. The d-limit of the indefinitely receding to the right nth d-terms of d* is a continuum that coincides with the g-limits of the tail digits of the nonterminating decimals traced by them as the ajs vary along the digits.
If x < 1 and a nonzero decimal, there is no difference between (0.1)n and x(0.1)n as they become indistinguishably small, i.e., as n increases indefinitely. This is analogous to the sandwich theorem of calculus that says, lim(x/six) = 1, as x ® 0 [
A decimal integer has the form N.99×××; if x is an integer or decimal integer (having the form x = N.99×××, N = 0, 1, ×××) we have,
, (12)
The first line of (12) says that d* cannot be separated from an integer but the second and third lines say that d* can be separated from a decimal integer.
It follows that the g-closure of R, i.e., its closure in the g-norm, is R* which includes the additive inverses and well-defined multiplicative inverses and d*. We also include in R* the upper bounds of the divergent sequences of terminating decimals and integers (a sequence is divergent if the nth terms are unbounded as n increases indefinitely, e.g., the sequence 8, 88, ×××) called unbounded number u* which is countably infinite since the countable union of countable sequences is. We call the divergent sequence of an unbounded number u s-sequence. Like d* it is set-valued. We follow the same convention for u*: whenever we have a statement “u has property P for every element u of u*” we can simply say “u* has property P). Then u* satisfies for given x,
Neither d* nor u* is a decimal and their properties are solely determined by their sequences. Then d* and u* have the following dual or reciprocal properties and relationship:
Numbers like u*-u*, d*/d* and u*/u* are still indeterminate but indeterminacy is avoided by computation with the g- or d-terms or s-terms. Thus, we now have a well-defined arithmetic of infinitesimal and infinity where d* and u* are counterparts of each other in the constructivist real number system respectively.
We can check that associativity, commutativity and distributivity of multiplication with respect to addition follow from the axioms and need not be taken as axioms. We note that the rules of signs apply to the decimals but not d*.
The decimals are linearly ordered by the lexicographic ordering by “<” defined as follows: two elements of R*are equal if corresponding digits are equal. Let
Then,
and, if x is any decimal we have,
The trichotomy axiom follows from the lexicographic ordering of R*. This is the natural ordering sought by mathematicians among the real numbers but it does not hold there because there is no linear ordering of R according to the above counterexample.
(Dark number d* mathematically models the superstring, fundamental building block of matter [
Mathematical systems are better understood by bringing in the notion of dual systems as it introduces symmetry that may be useful. We consider divergent sequences, i.e., sequences whose terms become bigger and bigger and indistinguishable from each other, as the dual of convergent sequences. In this sense they also form an algebraic continuum. We look at d* as the dual of u* and R* that of the system of additive and multiplicative inverses. Thus, R* is a semi-field, the nonzero integers forming a semi-ring since some of them have no multiplicative inverses. Like d*, u* cannot be separated from the decimals, i.e., there is no boundary between either of them and the decimals and between finite and infinite. Thus, there is no boundary to cross between finite and infinite and that beyond a certain finite decimal everything else is infinite. This is what is meant by the expression u* + x = u* for any decimal x. Duality is also seen in this case: let l> 1 be a terminating decimal then the sequence ln,
The decimal integers are decimals of the form,
We note that 1 + 0.99… is not defined in R since 0.99××× is nonterminating but we can write 0.99××× = 1 − d* so that 1 + 0.99××× = 1 + 1 −d* = 2 −d* =1.99×××; we now define 1 + 0.99××× = 1.99××× or, in general, N ? d* = (N − 1).99××× Twin integers are pairs (N, (N − 1).99×××), N = 1, 2, ×××; their first and second components are isomorphic.
Let f be the mapping N → (N − 1). 99××× and extend it to the mapping d* → 0 even if d* is not a decimal; then we show that f is an isomorphism between the integers and decimal integers:
Thus, addition of decimal integers is the same as addition of integers. Next, we show that multiplication is also an isomorphism.
We have now established the isomorphism between the integers and the decimal integers (first established in [
(The second equation of (21) can be proved also by mathematical induction for a given n).
We exhibit other properties of 0.99××× Let K be an integer, M.99××× and N.99××× decimal integers. Then
To verify that 2(0.999×××) = 1.99×××, we note that (1.99×××)/2 = 0.99×××
Two decimals are adjacent if they differ by d*. Predecessor-successor pairs and twin integers are adjacent. For example, 74.5700××× and 74.5699××× are adjacent.
Since the decimals have the form N.a1a2×××an×××, N = 0, 1, 2, ×××, the digits are identifiable and, in fact, countably infinite and linearly ordered by lexicographic ordering. Therefore, they are discrete or digital and the adjacent pairs are also countably infinite. However, since their tail digits form a continuum, R* is a continuum with the decimals its countably infinite discrete subspace.
A decimal is called recurring 9 if its tail decimal digits are all equal to 9. For example, 4.3299… and 299.99××× are recurring 9s; so are the decimal integers. (In an isomorphism between two algebraic systems, their operations are interchangeable, i.e., they have the same algebraic structure but differ only in notation).
The recurring 9s have interesting properties. For instance, the difference between the integer N and its recurring 9, (N− 1).99…, is d*; such pairs are called adjacent because there is no decimal between them. In the lexicographic ordering the smaller of the pair of adjacent decimals is the predecessor and the larger the successor. The average between them is the predecessor. Thus, the average between 1 and 0.99××× is 0.99××× since (1.99×××)/2 = 0.99…; this is true of any recurring 9, say, 34.5799××× whose successor is 34.5800××× Conversely, the g-limit of the iterated or successive averages between a fixed decimal and another decimal of the same integral part is the predecessor of the former.
Since adjacent decimals differ by d* and there is no decimal between them, i.e., we cannot split d* into nonempty disjoint sets, we have another proof that d* is a continuum. The counterexample to the trichotomy axiom shows that an irrational number in the real number system cannot be expressed as limit of sequence of rationals since the closest it can get to it is some rational interval which still contains rationals whose relationship to it is unknown, another expression of the ambiguity of the concept irrational [
The g-sequence of a nonterminating decimal reaches its g-limit, digit by digit, an advantage in computation with the g-norm. Moreover, a nonterminating decimal is an infinite series of its digits:
We add the following results to the information we now have about the various subspaces of R* to provide a full picture of the structure of R*. The next theorem is a definitive result on the continuum R*; it does not hold in R.
Theorem. In the lexicographic ordering R* consists of adjacent predecessor-successor pairs (each joined by d*); hence, the g-closure R* of R is a continuum.
However, the decimals form countably infinite discrete subspace of R* since there is a scheme for labeling them by integers. (An integer is a decimal with 0 decimal digits) We can imagine them as forming a right triangle with one edge horizontal and the vertical one extending without bounds. The integral parts are lined up on the vertical edge and joined together by their branching digits between the hypotenuse and the horizontal that extend to d* which is adjacent to 0 (i.e., differs from 0 by d*) at the vertex of the horizontal edge.
Corollary. R* is non-Archimedean but Hausdorff in both the standard and the g-norm and the subspace R of decimals are countably infinite, hence, discrete but Archimedean and Hausdorff.
Clearly, R coincides with the set of terminating decimals. The following is a theorem in R* [
Theorem. Every real number is isolated from the rest.
This theorem, originally proved in R [
Theorem. The rationals and irrationals are separated, i.e., they are not dense in their union (the first indication of discreteness of the decimals).
This theorem, proved in R [
Theorem. The largest and smallest elements of the open interval (0,1) are 0.99××× and d*, respectively [
Theorem. An even number greater than 2 is the sum of two primes
This is the 274-year-old conjecture in the real number system called Goldbach’s conjecture [
Corollary. Every integer or terminating decimal has dark component inseparable from it.
We highlight some of the major results in R*.
1) Every convergent sequence has a g-subsequence defining a decimal adjacent to its standard limit. If the g-limit of a sequence is terminating then it coincides with its standard limit.
2) It follows from 1) that the standard limit of a sequence of terminating decimals can be found by evaluating the g-limit of its g-subsequence which is adjacent to it. This is an alternative way of computing the limit of ordinary sequence.
3) In [
4) Given two decimals and their g-sequences and respective nth g-terms An, Bn we define the nth g-distance as the g-norm
5) Every sequence of elements of R* bounded below has a greatest lower bound and every sequence of elements of R* bounded above has a least upper bound. This retrieves in R* a flawed theorem in R. It follows that R* is well ordered. This is not so for R. Moreover, if the standard limit of a sequence of decimals exists then it is constructible since it is adjacent to the g-limit of some g-sequence which is a decimal.
We have identified some ambiguous concepts of R, defined some of them in R* and discarded those that cannot be fixed, e.g., the irrationals. We also identified some theorems in R that are false in R*, e.g., that the fractions are dense in the real number system. In fact, all theorems there that rely on the axiom of choice are false, e.g. the Heine-Borel theorem and existence of nonmeasurable set [
be extended to R* so that the resolution of FLT will be done in R*.
Although it is sufficient to resolve this 379-year-old FLT by a counterexample, we shall show that there are countably infinite counterexamples to it.
Given the impossibility of proving a negative proposition or statement we use the negation of FLT, i.e. the existence of solution of Fermat’s Equation (28). That would be a counterexample to FLT. We summarize the properties of the digit 9.
1) A finite string of 9s differs from its nearest power of 10 by 1, e.g., 10100 − 99×××9 = 1.
2) If N is an integer, then (0.99×××)N = 0.99××× and, naturally, both sides of this equation have the same g-se- quence. Therefore, for any integer N, ((0.99×××)10)N = (9.99×××)10N.
3)
Then the exact solutions of Fermat’s equation are given by the triples
for n = NT > 2. Moreover, for k = 1, 2, ×××, the triples (kx, ky, kz) also satisfy Fermat’s equation. They are the countably infinite counterexamples to FLT that prove the conjecture false [
The advantages of the g-norm over other norms are as follows:
a) It avoids some indeterminate forms.
b) Since the g-norm of a decimal is the decimal itself, computation with it yields the answer directly, digit by digit, and avoids intermediate approximations of standard computation. It means considerable reduction in computer time for large computation.
c) Since the standard limit is adjacent to the g-limit of some g-sequence, evaluating it reduces to finding some nonterminating decimal adjacent to it; the decimal is approximated by the appropriate nth g-term. Both the com- putation and approximation are precise. In fact, the exact margin of error is d*. This applies to the result of any computation: it is adjacent to some nonterminating decimal and the latter is found using the g-norm.
d) In iterated computation along successive refinements of sequence xj that tends to a as j ® ¥, the iteration is simplified by taking averages between the sequence of points xj and its g-limit.
e) Approximation by the nth g-term or n-truncation contains the ambiguity of a nonterminating decimal.
f) Calculation of distance between two decimals with the g-norm is direct, digit by digit. It involves no root or radical at all.
g) In general radicals in computation, e.g., taking root of a prime, is avoided, by using the nth g-term approximation or n-truncation to any desired margin of error where accuracy is measured by the number of digits of the result obtained.
The g-norm is the natural norm for computation since a) it puts rigor in computation in accordance with the new definition of the previously ill-defined nonterminating decimals in terms of the defined terminating decimals, b) the margin of error is precisely determined and c) the result of computation is obtained digit by digit and avoids intermediate unnecessary calculation proceeds directly to the result digit by digit.
Constructivism requires that proof of existence alone is not sufficient. In the case of a normed space like the setting (mathematical space) for the rectified calculus of variations and optimal control theory the appropriate norm is the Young Measure [
Edgar E. Escultura, (2016) The Constructivist Real Number System. Advances in Pure Mathematics,06,593-607. doi: 10.4236/apm.2016.69048