SCHAUM’S OUTLINE OF THEORY AND PROBLEMS OF BOOLEAN ALGEBRA and SWITCHING CIRCUITS BY ELLIOTT MENDELSON, Ph.D. Professor of Mathematics Queens College City University of New York SCHAUM’S OUTLINE SERIES McGRAW-HILL BOOK COMPANY New York, St. Louis, San Francisco, London, Sydney, Toronto, Mexico, and PanamaPreface This book is devoted to two separate, but related, topics: (1) the synthesis and simplification of switching and logic circuits, and (2) the theory of Boolean algebras. Those people whose primary interest is in switching and logic circuits can read Chapter 4 immediately after a quick perusal of Chapter 1. We have confined our treatment of switching and logic circuits to combinational circuits, ie. circuits in which the outputs at a given time depend only on the present value of the inputs and not upon the previous values of the inputs. The extensive theory of sequential circuits, in which the outputs depend also upon the history of the inputs, may be pursued by the reader in Introduction to Switching Theory and Logical Design by F. J. Hill and G. R. Peterson (reference 34, page 202), Introduction to Switching and Automata Theory by M. A. Harrison (ref. 33, page 202), and other textbooks on switching theory. The treatment of Boolean algebras is somewhat deeper than in most elementary texts. It can serve as an introduction to graduate-level books such as Boolean Algebras by R. Sikorski (ref. 148, page 207) and Lectures on Boolean Algebras by P. R. Halmos (ref. 116, page 207). There is no prerequisite for the reading of this book. Each chapter begins with clear statements of pertinent definitions, principles and theorems together with illus- trative and other descriptive material. This is followed by graded sets of solved and supplementary problems. The solved problems serve to illustrate and amplify the theory, bring into sharp focus those fine points without which the student continually feels himself on unsafe ground, and provide the repetition of basic principles so vital to effective learning. A few problems which involve modern algebra or point-set topology are clearly labeled. The supplementary problems serve as a complete review of the material of each chapter. Difficult problems are identified by a superscript ° following the problem number. The extensive bibliography at the end of the book is divided into two parts, the first on Switching Circuits, Logie Circuits and Minimization, and the second on Boolean Algebras and Related Topics. It was designed for browsing. We have listed many articles and books not explicitly referred to in the body of the text in order to give the reader the opportunity to delve further into the literature on his own. Queens College ELLIOTT MENDELSON July 1970CONTENTS Chapter THE ALGEBRA OF LOGIC os 11 Truth-Funetional Operations .... 1 12 Connectives ........ 4 13. Statement Forms - 4 14 Parentheses 5 15 Truth Tables .. dee sees ease eens ee neneneeeee . 6 16 Tautologies and Contradictions . aU 1.7 Logical Implication and Equivalence .............+ . 9 1.8 Disjunctive Normal Form 2 1.9 Adequate Systems of Connective . 16 Chapter 2 THE ALGEBRA OF SETS 21 Sets. . 30 2.2 Equality and Inclusion of Sets. Subsets eeneee. . 8h 23 Null Set. Number of Subsets Prreeereeereeeeeteeneyy a1 24° Union a2 25 Intersection ......++ 33 26 Difference and Symmetric Difference au 2.7 Universal Set. Complement . vee BB 28 Derivations of Relations among Sets Perera 36 2.9 Propositional Logic and the Algebra of Sets .........ssseseeceeeseee 37 210 Ordered Pairs, Functions . eeennneneren 38 211 Finite, Infinite, Denumerable, and Countable Sets ..... rere 39 212 Fields of Sets. cee eee eer 39 2.13 Number of Elements in a Finite Set PER YEeONEOOO 40 Chapter 3 BOOLEAN ALGEBRAS 3.1 Operations 82 32 Axioms for a Boolean Algebra 52 3.3 Subalgebras .... 56 34 Partial Orders... _ ar 3.5 Boolean Expressions and Functions, Normal Forms . : 60 3.6 Isomorphisms . @ 3.7 Boolean Algebras and Propositional Logic tee 63.CONTENTS Chapter 4 SWITCHING CIRCUITS AND LOGIC CIRCUITS = 41 Switching Circuits. a 42 Simplification of Circuits 72 43° Bridge Circuits tees 44 Logie Circuits ...s.se0s+ : 6 45 The Binary Number System 1 4.0 Multiple Output Logie Circuits . 7 47 Minimization - 80 48 Don’t Care Conditions . 8 49 Minimal Disjunctive Normal Forms 4 410° Prime Implicants ...s.seseseese+ a 411 The Quine-McCluskey Method for Finding All Prime Implicants .......... 86 412 Prime Implicant Tables 90 413 Minimizing with Don’t Care Conditions . 98 414 The Consensus Method for Finding Prime Implicants . on 4.15 Finding Minimal Dnf's by the Consensus Method . se eteeeeee, ae 96 4.16 Karnaugh Maps ...... 98 417 Karnaugh Maps with Don’t Care Conditions . 108 418 Minimal Dat’s or Cat's 107 Chapter 5 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 51 Lattices .......ee eee 138 52 Atoms . eae en 76) 5.3 Symmetric Difference. Boolean Rings ........-..-0ceeesseereseneeenee 137 54 Alternative Axiomatizations ...... Peeper sees M0 55 Ideals ee ees us 5.6 Quotient Algebras .......-.- seeeeer eres, 146 5.7 The Boolean Representation Theorem seeeeeeeeeeees 148 5.8 Infinite Meets and Joins .......+.s00+0+0+ 149 59 Dustity a - 181 510 Infinite Distributivity .. 11 cesses 153 511 m-Completeness Appendix A ELIMINATION OF PARENTHESES . : Appendix B PARENTHESIS-FREE NOTATION .. 195 Appendix C THE AXIOM OF CHOICE IMPLIES ZORN'S LEMMA ..... 198 Appendix D A LATTICE-THEORETIC PROOF OF THE SCHRODER-BERNSTEIN THEOREM 200 IBLIOGRAPHY ...... sees 201 INDEX . 209 INDEX OF SYMBOLS AND ABBREVIATIONS ... oro 218Chapter 1 The Algebra of Logic 11 TRUTH-FUNCTIONAL OPERATIONS There are many ways of operating on propositions to form new propositions. We shall limit ourselves to those operations on propositions which are most relevant to mathematics and science, namely, to truth-functional operations. An operation is said to be truth- functional if the truth value (truth or falsity) of the resulting proposition is determined by the truth values of the propositions from which it is constructed. The investigation of truth-functional operations is called the propositional calculus, or, in old-fashioned terminology, the algebra of logic, although its subject matter forms only a small and atypically simple branch of modern mathematical logic. ion is the simplest common example of a truth-functional operation. If A is a Proposition, then its denial, not-A, is true when A is false and false when Ais true. We shall use a special sign “1 to stand for negation. Thus, 1A is the proposition which asserts the denial of A. The relation between the truth values of 1A and A can be made explicit by a diagram called a truth table. a [qa T|F Fir Fig. 1-1 In this truth table, the column under A gives the two possible truth values T (truth) and F (falsity) of A. Each entry in the column under 1A gives the truth value of 1A corresponding to the truth value of A in the same row. Conjunction Another truth-functional operation about which little discussion is necessary is con- junction. We shall use A & B to stand for the conjunction (A and B), The truth table for & is Fig.1-22 THE ALGEBRA OF LOGIC [CHAP.1 ‘There are four possible assignments of truth values to A and B, Hence there are four rows in the truth table, The only row in which A & B has the value T is the first row, where each of A and B is true. Disjunction ‘The use of the word “or” in English is ambiguous. Sometimes, “A or B’t means that at least one of A and B is true, but that both A and B may be true. This is the inclusive sage of “or”. Thus to explain someone's success one might say “he is very smart or he is very lucky”, and this clearly does not exclude the possibility that he is both smart and lucky. The inclusive usage of “or” is often rendered in legal documents by the expression “and/or”. ‘Sometimes the word “or” is used in an exclusive sense. For example, “Either I will go skating this afternoon or I will stay at home to study this afternoon” clearly means that I ill not both go skating and stay home to study this afternoon. Whether the exclusive usage is intended by the speaker or is merely inferred by the listener is often difficult to determine from the sentence itself. In any case, the ambiguity in usage of the word “or” is something that we cannot allow in a language intended for scientific applications. It is necessary to employ distinct symbols for the different meanings of “or”, and it turns out to be more convenient to introduce a special symbol for the inclusive usage, since this occurs more frequently in mathematical assertions.tt “Av B” shall stand for “A or B or both”. Thus in its truth table (Fig. 1-8) the only case where is false is the case where both A and B are false. The expression Av B will be called a disjunction (of A and B). Fig. 1-3 Conditionals In mathematics, expressions of the form “If A then B” occur so often that it is necessary to understand the corresponding truth-functional operation. It is obvious that, when A is T and B is F, “If A then B” must be F. But in natural languages (like English) there is no established usage in the other cases (when A is F, or when both A and B are T). In fact when the meanings of A and B are not related (such as in “If the price of milk is 25+ per quart, then high tide is at 8:00 P.M. today”), the expression “If A then B” is not regarded as having any meaning at all. tStrictly speaking, we should employ quotation marks whenever we are talking about an expression rather than using it However, this would sometimes get the reader lost in a sea of quotation marks, and we adopt instead the practice of omitting quotation marks whenever misunderstanding is improbable. ttIn some natural languages, there are different words for the inclusive and exclusive “or”, For example, in Latin, “vel” is used in the inclusive sense, while “aut” is used in the exclusive sense.CHAP. 1] THE ALGEBRA OF LOGIC 3 ‘Thus if we wish to regard “If A then B” as truth-functional (i.e. the truth value must be determined by those of A and B), we shall have to go beyond ordinary usage. To this end we first introduce + as a symbol for the new operation. Thus we shall write “A> B” instead of “If A then B”. A~B is called a conditional with antecedent A and consequent B. The truth table for > contains so far only one entry, in the third row. Fig.14 As a guideline for deciding how to fill in the rest of the truth table, we can turn to “If (€ & D) then C”, which seems to be a proposition which should always be true. When C is T and D is F, (C & D) is F. Thus the second line of our truth table should be T (since (C & D) is F, € is T, and (If (¢ & D) then C) is T). Likewise when € is F and D is F, (C&D) is F. Hence the fourth line should be T. Finally, when € is T and D is T, (C & D) is T, and the first line should be T. We arrive at the following truth table: Figs A~B is F when and only when A is T and B is F. To make the meaning of A-> B somewhat clearer, notice that A~> B and (1A) v B always have the same truth value. (Just consider each of the four possible assignments of truth values to A and 8.) Thus the intuitive meaning of A> ot-A or B”. This is precisely the meaning which is given to “If A then 8” in contemporary mathematics. A proposition A+ B is T whenever A is F, irrespective of the truth value of B. Notice also that A ~ B is automatically T whenever B is T, without regard to the truth value of A. In these two cases, one sometimes says that A~B is trivially true by virtue of the falsity of A or the truth of B. Example 1.1. The, propositions 2+2=5 + 11 and 2+2=5 + 1=1 are both trivially true, since 2+2=5 is false, Biconditional At this time we shall introduce a special symbol for just one more truth-functional operation: A if and only if B. Let A «+B stand for “A if and only if B”, where we under- stand the latter expression to mean that A and B have the same truth value (i.e. if A is T, so is B, and vice versa). This gives rise to the truth table:4 THE ALGEBRA OF LOGIC (CHAP. 1 Fig. 1-6 A proposition of the form A «> B is called a biconditional. Notice that AB always takes the same truth value as (A~ B) & (B~ A); this is reflected in the mathematical practice of deriving a biconditional A « B by proving A+B and B~ A separately. 12 CONNECTIVES Up to this point, we have selected five truth-functional operations and introduced special symbols for them: 1, &,v,>,«>. Of course if we limit ourselves only to two variables, then there are 2'=16 different truth-functional operations. With two variables, a truth table has four rows: Fig. 1-1 A truth-functional operation can have either T or F in each row. Hence there are 2+2-2-+2 possible binary truth-functional operations. Corresponding to any truth-functional operation (i.e. to any truth table) we can introduce a special symbol, called a connective, to indicate that operation. Thus the symbols 11, & v, >, @ are connectives. These five connectives will suffice for all practical purposes. Example 12. ‘The operation corresponding to the exclusive usage of “or” could be designated by a connective +, having as its truth table: Ate . T T F F Fig. 1-8 13 STATEMENT FORMS To study the properties of truth-functional operations we introduce the following notions. By a statement form (in the connectives “1, & v, >, «») we mean any expression built up from the statement letters A, B,C, ..., Ar Bi, Ci, ... by a finite number of applications of the connectives 1, &, v,+,«+. More precisely, an expression is a statement form if it can shown to be one by means of the following two rules:CHAP. 1] THE ALGEBRA OF LOGIC 5 (1) All statement letters (with or without positive integral subscripts) are statement forms. (2) If A and B are statement forms, so are (1A), (A&B), (Av 8B), (A~B), and (Ac+B).t Example 13. Examples of statement forms: @ (A> BvC&IAM; Gi) (WAS (B; (HH (14D) > (A> AD). Clearly we can talk about statement forms in any given set of connectives (instead of just 1, & v, >, «) by using the given connectives in clause (2) of the definition. 14 PARENTHESES ‘The need for parentheses in writing statement forms seems obvious. An expression such as Av B&C might mean either ((Av B) &C) or (Av (B&C)), and these two statement forms are not, in any sense, equivalent. While parentheses are necessary, there are many cases in which some parentheses may be conveniently and unambiguously omitted. For that purpose, we adopt the following conventions for omission of parentheses. (1) Every statement form other than a statement letter has an outer pair of parentheses. We may omit this outer pair without any danger of ambiguity. Thus instead of (Av B) & (10), we write (Av B) & (10). (2) We omit the pair of parentheses around a denial (1A). Thus instead of (1.4) v C, we write 14 vC, This cannot be confused with 1(AvC), since the parentheses will not be dropped from the latter. As another example consider (A&B) v (1(1(1))). This becomes (A&B) v 111B. (8) For any binary connective, we adopt the principle of association to the left. For example, A & B & C will stand for (A&B) & C, and A> B~ C will stand for (A+B) >. Example 14. Applying (1)-(8) above, the statement forms in the column on the left below are reduced to the equiva- lent expressions on the right. (AM &Q) v (Ay) WWALOY 1A (Av (1B) & (C&(7A)) (Av 18) 8 (C&A) (Av (18) & 0) & (14) (AV 1B) ECHTA (14) + 8+ (WAV Oo) 1A + (B> WAV Q) More far-reaching conventions for omitting parentheses are presented in Appendix A. In addition, Appendix B contains a method of rewriting statement forms so that no parentheses are required at all. ‘An even more rigorous definition is: 8 is a statement form if and only if there is a finite sequence Ay ---+ ‘A, such that (1) &, is | (2 if 1A).6 THE ALGEBRA OF LOGIC [cHaP.1 15 TRUTH TABLES Every statement form A defines a truth-function: for every assignment of truth values to the statement letters in A, we can calculate the corresponding truth value of A itself. ‘This calculation can be exhibited by means of a truth table. Example 15. ‘The statement form (1AVvB) © A has the truth table (AVR) OA was ay Fig. 1.9 Each row corresponds to an assignment of truth values to the statement letters, The columns give the corresponding truth values for the statement forms occurring in the step-by-step construction of the given statement form. Example 16. ‘The statement form (Av (B&C)) +B has the truth table Av (B&O) +B Fig.1-10 ‘When there are three statement letters, notice that the truth table has eight rows. In general, when there are n statement letters, there are 2* rows in the truth table, since there are two possibilities, T or F, for each statement letter. Abbreviated Truth Tables By the principal connective of a statement form (other than a statement letter), we mean the last connective used in the construction of the statement form. For example, (Av B)->C has + as its principal connective, A v (B+ C) has v as its principal connective, and (4 vB) has 71 as its principal connective. ‘There is a way of abbreviating truth tables so as to make the computations shorter. We just write down the given statement form once, and, instead of devoting a separate column to each statement form forming a part of the given statement form, we write the truth value of every such part under the principal connective of that part. Example 17. Abbreviated truth table for (1A vB) <>A. We begin with Fig. 1-11. Notice that each occurrence of a statement letter requires a repetition of the truth assignment for that letter.CHAP. 1] THE ALGEBRA OF LocIc 7 O4vVBR eA TOT ror F TF T FPR F Fig. 1-11 ‘Then the negation is handled: GAvVBoA FT T T Tr oF FT F oT Tr F oF followed by the disjunction GAVB OA FITT TRTT F FTFF T TRTF F and, finally, the biconditional (G4vpyea PITT TT TRFTT FF FTFF FT TFETF FF Of course our use of four separate diagrams was only for the sake of illustration. In practice all the work can be carried out in one diagram. 1.6 TAUTOLOGIES AND CONTRADICTIONS A statement form A is said to be a tautology if it takes the value T for all assignments of truth values to its statement letters. Clearly A is a tautology if and only if the column under A in its truth table contains only T’s. Example 18. AA is 2 tautology. AlAnA Tl T Flot Fig. 112 Example 19. Av 7A is a tautology.8 THE ALGEBRA OF LoGIC [CHAP.1 Example 110. (Av B) + (Bv A) is a tautology. (Av B)e (BVA) T T F Tr r T F T Fig. 1-14 Example Lil. [A & (BV C)] + [(A&B)v(A&O)] is a tautology.t x 4 c & &C | A&B)ViA&C) | A T T T T F F F T T F T T F F F T T T F T F F F T T F F T F F F T Fig. 115 ‘Theorem 11. If K is a tautology, and statement forms A,B,C, ... are substituted for the statement letters A, B,C, ... of K (the same statement form replacing all occurrences of a statement letter), then the resulting statement form K® is a tautology. Example 1.12. (Av B) + (BVA) is a tautology. Replace A by (BvC) and simultaneously replace B by A. The new statement form [(Bv C)v A] [Av (Bv C)] is a tautology. Proof of Theorem 1.1. K determines a truth-funetion /(A, B,C, ...) which always takes the value T no matter what the truth values of A, B,C,... maybe. Let 9:, gs, gs, ... be the truth-functions determined by A,B,C,.... Then the truth-function determined by K* must have the form f? = f(91(...), 92(---), 9s(---), ---), and, since f always takes the value T, f* also always takes the value T. A contradiction is a statement form which always takes the value F. Hence A is a contradiction if and only if 1A is a tautology, and A is a tautology if and only if 1A is a contradiction. Example 118, A & 7A is a contradiction, Fig. 1-16 tin writing this statement form, we have replaced some parentheses by brackets to improve legibility. For the same purpose, we also shall use Laces.CHAP. 1] THE ALGEBRA OF LOGIC 9 Example 114 A © 74 is a contradiction. Aji | aor tl/F F Fit F Fig.117 Example 115. (Av B)& 1A & 1B is a contradiction, (AvB)& 74 | (AvB)& TA & 1B aasale er ] Fig. 1-18 1.7 LOGICAL IMPLICATION AND EQUIVALENCE _We say that a statement form A logically implies statement form B if and only if every assignment of truth values making A true also makes B true, Example 1.16. A logically implies A. Example 1.17. A logically implies Av B. For, whenever A is true, Av B also must be true. Example 118. A & B logically implies A. Theorem 1.2, A logically implies B if and only if A~ B is a tautology. Proof. A logically implies B if and only if, whenever A is true, B must also be true. Therefore A logically implies B if and only if it is never the case that A is true and B is false. But the latter assertion means that A ~ B is never false, ie. that A ~ B is a tautology. > Since we can effectively determine by a truth table whether a given statement form is a tautology, Theorem 1.2 provides us with an effective procedure for checking whether A logically implies B. ‘Example 119. Show that (A +B) ~ A logically implies A. Proof, Fig, 1-19 shows that ((A +B) > A) +A is a tautology. ? sao wa F T F Fig. 1.19 Statement forms A and B are called logically equivalent if and only if A and B always take the same truth value for any truth assignment to the statement letters. Clearly this means that A and B have the same entries in the last column of their truth tables.10 THE ALGEBRA OF Logic (CHAP. 1 Example 120, A.B is logically equivalent to (A +B) &(B~ A), Fig. 1-20 Theorem 13. A and B are logically equivalent if and only if A« B is a tautology. Proof. A+B is T when and only when A and B have the same truth value. Hence A. Bis a tautology (ie. always takes the value T) if and only if A and B always have the same truth value (i.e. are logically equivalent). > Example 121. A+ (B+ ©) is logically equivalent to (A &B) >C. Proof. [A 7 (BC)] + [(A &B) ~ C] is a tautology as shown in Fig. 1-21. A ABO) | A&B | (A&B) | [ABO OA B+} T T F T T T F T T v F T T v F T Fig. 1-21 Corollary 14. If A and B are logically equivalent and we replace statement letters in A and B by statement forms (all occurrences of the same statement letter being replaced in both A and B by the same statement form), then the resulting statement forms are also logically equivalent. Proof. This is a direct consequence of Theorems 1.3 and 1.1. Example 1.22. A~(B~O) and (A&B)+C are logically equivalent, Hence so sre (Cv A)~(B~(AvB)) and ((Cv A) & B) + (Av B) (and, in general, so are A~+(8~C) and (A&B)+C for any statement forms A8,0). Theorem 15 (Replacement). If B and € are logically equivalent and if, within a statement form A, we replace one or more occurrences of B by C, then the resulting statement form A® is logically equivalent to A. Proof. In the calculation of the truth values of A and A*, the distinction between B and C is unimportant, since B and always take the same truth value. > Example 1.23. Let A be (AVB)~C, Since Av B is logically equivalent to Bv A, A is logically equivalent to (BvAy>c.CHAP. 1] THE ALGEBRA OF LOGIC cbt The following examples of logically equivalent pairs of statement forms will be extremely useful in the rest of this book, for the purpose of finding, for a given statement form, logically equivalent statement forms which are simpler or have a particularly revealing structure. We leave verification of their logical equivalence as an exercise. Example 124, 771A and A (Law of Double Negation) Example 125’ (a) A&A and A @) AvA and A (dempotence) Example 1.28. (a) A&B and BGA () AvB and BVA ee Example 127, (a) (A&B) &C and A& (8&C) (Assoeiativity) @) (AvB)VE and Av (Bve) As a result of the associative laws, we can leave out parentheses in conjunctions or dis- junctions, if we do not distinguish between logically equivalent statement forms. For example, A v B v Cv D stands for ((A vB) v C) v D, but the statement forms (A v (Bv C)) v D, Av ((BvC)vD), (AvB) v (CvD) and Av (Bv (Cv D)) are logically equivalent to it. Terminology: In A,v Av --- v As, the statement forms A, are called disjuncts, while in A & Ac & ... & Ay the statement forms A; are called conjuncts, Example 128. De Morgan’s Laws. (® YAv®) and 10878 @) AGS) and WAV 18 Example 129. Distributive Laws (or Factoring-out Laws). (a) A&(BvC) and (A&B) v (A&C) (0) Av (B&C) and (Av 8) & (AVC) Notice that there is a distributive law in arithmetic: a+(b +c) = the other distributive law, a +(b-¢) = (a+b)-(a+e) is false. (Take Example 130, Absorption Laws. @ Av (A&B) and A (©) A&(AVB) and A (a+b) + (are); but 1) () (@) (A&B) v 1B and Av 1B @) (Ava) & 1B and A&W (II) If T is a tautology and F is a contradiction, @ (T&A) and A) (F&A) and F (®) (fvA) and T (@ (FVA) and A We shall often have occasion to use the logical equivalence between (A & 1B) v B and Av B, and-between (Av 1B) & B and A&B. We shall justify this by reference to Example 1,80(II), since it amounts to substituting 1B for B in Example 1.30(II) and then using Example 1.24, Example 1.81, A+B and 18+ 1A —(Contrapositive) Example 1:32. Elimination of conditionals, (@) Ave and TAVE (6) A+B and (AG 1B)12 THE ALGEBRA OF LoGIC Example 138 Elimination of biconditionals. (@) Ae and (AGB) v (TAL 1B) () AB and (TAVB) & (TBVA) Examples 1.82 and 1.88 enable us to transform any given statement form into a logica=> equivalent statement form which contains neither > nor «. 18 DISJUNCTIVE NORMAL FORM By literals we mean the statement letters A,B,C,... and the denials of stateme== letters 14, 1B, 1C, By a fundamental conjunction we mean either (i) a literal <= (i) a conjunction of two or more literals no two of which involve the same statement letter. For instance, 42, 1B, A&B, 1A:&A&C are fundamental conjunctions, while 1714. A&B&A, BEA&C& 7B are not fundamental conjunctions. One fundamental conjunction A is said to be included in another B if all the literals c: A are also literals of Bt, For example, A & B is included in A & B, B& (110) is included in (1C) & B, B is included in A & B, and 1 & A is included in A & B& 1C, while B is not included in A & 1B. A statement form A is said to be in disjunctive normal form (dnf) if either (i) A is 2 fundamental conjunction, or (ii) A is a disjunction of two or more fundamental conjunctiozs. of which none is included in another. Example 134. The following statement forms are in dnf. (@) B () OVC () Av (IBEO) (@ (A&7B)v (14 & BRO) (2) B&IA)Y (14 & BED) Vv Av (B&CH& 1D). Example 135. The following statement forms are not in dnf. (e) C&C @) (CVD) &A (0) (C&A 1B) v (ICRA) (C& 1B), Replacing statement forms by logically equivalent ones, we can transform a statemen= form into one in disjunctive normal form. Example 136. (AVB)& (Av Cv 1B) Av (B&(Cv 1B) Distributive Laws, Example 1.29) Av B&Q (Absorption Law, Example 1.80(110)) Example 137, WA vC)V (A+B) WAV O)v (TAB) (Example 1.32(a)) (VAG TOV TAY B (Example 1.28(2), De Morgan's Laws) qAVB (Example 1.30(10)) ‘tMore precisely, if all literals of A which do not occur within another literal of A are also literals of B ‘which do not oceur within another literal of 8. Thus B& A is not included in C&A & 1B, and 1B&A is not included in B& A.CHAP. 1] THE ALGEBRA OF LOGIC 13 Example 138. (A & 1B) + (Bv A) [(4& 1B) & BVA) v [1(4&7B)& BVA Example 1.38(a) Bramples 1270), Eeamples | Example 1.392) (A&BEA) v [IAB (1BE A) Exmpie2s | Example 1.01) (esa) vy (1A 1B)& 14] Example 1.25(0) (ew) [14 & 73] This is in disjunctive normal form. However, it is logically equivalent to (Av 1A) & 1B (Example 1.29(0)) 2 (Example 1.30(111@)) ‘This example shows that there are two logically equivalent statement forms, both of which are in disjunctive normal form. The fact illustrated in Examples 1.36-1.38 is codified in the following proposition. Theorem 1.6. Every statement form which is not a contradiction is logically equivalent to a statement form in disjunctive normal form. Proof. By Examples 1.82 and 1.88 we may find a logically equivalent statement form in the connectives 1, & v, and then, by De Morgan’s Laws (Example 1.28) we can move the negation signs inward so that negation signs apply only to statement letters. Thus we may confine our attention to statement forms built up from literals by means of & and v. The proof proceeds by induction on the number n of the connectives & and v in the given state- ment form A. If n=0, A isa literal, and every literal is already in dnf. Assume now that A contains k of the connectives & and v, and that the theorem is true for all natural numbers 1” < k. Case 1: Ais BvC. By inductive hypothesis, B and © are logically equivalent to state- ment forms B* and G*, respectively, in dnf. Hence A is logically equivalent to B* v C+. Now if any disjuncts D, of B* or of C* are included in any other disjuncts D: of B* or of C*, then we drop the disjuncts D; (by Example 1.30(Ia)). The resulting statement form is in dnf and is logically equivalent to A. Case 2: AisB&C. By inductive hypothesis, B and € are logically equivalent to state- ment forms B* and C?, respectively, in dnf. Hence A is logically equivalent to B* & C*. Let us assume that Bf is (B,v --- v B,) and C# is (Cv --- vG,), where the B’s and Gs are fundamental conjunctions, and r= 1,1. Then B* & C* is (Biv & (Civ ---vG,) which by a Distributive Law (Example 1.29(b)) is logically equivalent to (Biv s+ vB) & Cvs v (Biv ++ vB) & C] and, again by a Distributive Law (Example 1.29(a)), each (B,v --- vB,) & C, is logically equivalent to (B:&C,) v --- v (Br&C,). ‘Thus we obtain the disjunction of all Bi & €;, where 1u THE ALGEBRA OF LOGIC [omar.1 Remark (1) on Theorem 1.6. A statement form in dnf cannot be a contradiction. For, if b: & ... & is one of its disjunction, where each L is a literal, then we assign to the statement letter appearing in L the value T if L: is the statement letter itself, and the value F if Ly is the denial of the statement letter. This assignment of truth values makes each L, true and hence L, & ... & lk true, and therefore the whole disjunction must be true (since one of its disjuncts is true). Thus the disjunction cannot be @ contradiction. Example 1.39. In (A&1B&C)v(1A& 1BEO, if we make A true, B false and C true, then the first disjunct A& 1B& Cis true and, alternatively, if we make A false, B false and C true, then the second disjunet is true). Remark (2) on Theorem 1.6. From the proof it is clear that the logically equivalent statement form in dnf may be chosen so that its statement letters already occur in the given statement form, i.e. no new statement letters are introduced. ‘There is a special type of dnf which will be very useful. A statement form A in dnf is said to be in full disjunctive normal form (with respect to the statement letters $i, . .., Si) if (i) any statement letter in A is one of the letters S,, (ii) each disjunct in A contains all the letters Si, .. Si, and Example 140, ‘The statement forma (A & B& 10) v (1A&B&C) v (A & 1B& IC) and 1B&A & 7C are in full dis- junetive normal form (with respect to A,B,C), However, (A &B)v(1A&B&C) and 1AV(A& 1BE IC) are not in full disjunctive normal form with respect to A, B, C. Example 141. ‘The statement form 1B is in full dnf with respect to B, but not with respect to A and B. The state. ment form (A & 1B) v (A&B) is in full dnf with respect to A and B, but not with respect to any other collection of letters. Theorem 1.7. Every non-contradictory statement form A containing $;,...,S, as its statement letters is logically equivalent to a statement form in full dnf (with respect to Si, ..., Sr). Proof. A is known by Theorem 1.6 to be logically equivalent to a statement form B in nf, and the statement letters of B already occur in A. Now assume that some statement letter S; is missing from a disjunct D, of B. However, D; is logically equivalent to D,& (Siv 1S) (by Example 1.80(IIla)), which in turn is logically equivalent to (D,&S,) v (D, & 1S) (by Example 1.29(a)). Hence we replace Dy by (D;&S.) v (D, & 1S). In this way we can introduce the letters S:, . .., Ss into any of the disjuncts from which they are missing. The final result is in full dnf with respect to $:, ..., St. > Example 142. (A& 1B) v Bv(1A&1B& 10) is in dnf, but not in full dnf with respect to A,B,C. We obtain a logically equivalent full dnf as follows: (A&IBEO) v (AB IBENC) v By (TAB BENC) (A& BEC) v (AE TBE NO v (BEA) v (BETA) v (1A & BEI) (A& B&O) v (AL TBE AO v (BRALO) v (BEAL IO) Vv (BE TAR) V (BEAK NO) v (TA&IBEAC)CHAP. 1) THE ALGEBRA OF LOGIC 15 In general, the method indicated in Theorem 1.7 can be summarized in the following way. If letters $,, ..., $j, are missing from a disjunct Dj, we add as conjuncts to D, all of the 2 possible combinations of S;,, ..., S;, or their denials. For example, to obtain a state- ment form in full dnf (with respect to A,B,C) logically equivalent to 1B, we construct (1B& A&C) v (1BEAEK NC) v (1B&E TALC) v (BE TALE NO) 1.9 ADEQUATE SYSTEMS OF CONNECTIVES As we have remarked earlier, every statement form determines a truth function, and this truth function may be exhibited by means of a truth table. The converse problem suggests itself: For any given truth function, is there a statement form determining it? There are 2° truth functions of n variables. For, there are 2* truth assignments to the » variables, and, to each of these assignments, the truth function can associate the value T or the value F. Example 143. ‘The four truth functions of one variable are Example 1.44, ‘The sixteen truth functions of two variables are & tTilr| Fle T F T T F/T) T|F T F T F T|F F/T T P T F rFlelrlgr tT FP F F AB | ASB | Wem | BoA | 14a | vw | WBA | 1423) T T F T F F F F 1 F T F F T T F F F r T F Tr F T T Tv F tT Tr T F F Fig. 123 Theorem 18. Every truth function is determined by a statement form in the connectives 1, &v. Proof. The given truth function f(z1, ..., 2s) can be exhibited as a “truth table”:16 THE ALGEBRA OF LOGIC [CHAP.1 Fig. 1-24 There are 2* rows in the table. In each row, the last column indicates the corresponding value f(z1,...,2s). In constructing an appropriate statement form, we shall associate the letters A;, ...,4n with the variables 21, ..., 24. Case 1: The last column contains only F's. Then the statement form (A1& 1A) v -+- v (Aa &14,) determines f. (Of course, any contradiction also determines f.) Case 2: There are some T’s in the last column. For 1 Remark on Theorem 1.8. If the given truth function is not always F (Case 2), the statement form A constructed in the proof is in full disjunctive normal form. This gives us a way of constructing a full dnf logically equivalent to a given non-contradictory state- ment form ©, Just write down the truth table for © and then construct the corresponding statement form A as in the proof of Theorem 1.8. Example 145. Given the truth funetion Fig. 1-25 Deis 1A& Ag Dy is Ay& Ay Dyin 1A,&1Ay Hence (14, BAD v (A, & TAQ) v (14, & 14) Aetermines the given truth funetion,CHAP. 1} THE ALGEBRA OF LOGIC 17 Example 146, Given the truth function | 2 | ty | olen 2) T T/T T Fir} 7 T T F/T T Fleir F {tl} P F rit] Fr F T/|F F P rlelr T Fig. 1.26 ‘The statement form having g as its truth function is (Ay & Az & Ay) v (1A, &Ag& AS) v (Ap ADGA) ¥ (14, & 14, AD) Example 147. To find a statement form in full dnf logically equivalent to (Av B) & (Av Cv 1B), construct the latter's truth table: A (Av B)& (Av Cv 1B) T F T F F T T T T F T P T F T F F PF T T T F T F Fig.1-27 ‘Then the method of Theorem 1.8 yields (A&B&C) v (1AGBE&O) v (AR TBRC) V (ABBE NO) v (AK IBE NC) By an adequate system of connectives we mean a collection B of connectives such that every truth function is determined by a statement form in the connectives of B. Thus ‘Theorem 1.8 asserts that (1, &, v} is an adequate system of connectives. Corollary 1.9. Each of the following is an adequate system of connectives: (a) (7, &}, (0) (1,v}, ) (1,7) Proof. (a) By Theorem 1.8, (1, &,v} is adequate. But, replacing any statement form Av B by the Iogically equivalent statement form 1(1A &1B), we obtain for any statement form iri (1, &, v} a logically equivalent statement form in (7, &). () We proceed as in (a), but here we replace every A& B by 1(1Av 18 (©) We can replace every Av B by the logically equivalent (1A) > B. > There are two binary connectives such that each of them alone forms an adequate system. Let | be the connective corresponding to the truth-functional operation of alternative denial, given by the truth table18 THE ALGEBRA OF LOGIC [cHAP. 1 F Fig. 1-28 A|B means “not both A and B”. The connective | is called the Sheffer stroke. {|) is adequate, since 1A is logically equivalent to A|A, and Av B is logically equivalent to (AA) | (8/8). Let | be the connective corresponding to the truth-functional operation of joint denial, given by the truth table Fig. 1-29 A | Bis read “neither A nor B”. {|} is adequate, since 1A is logically equivalent to AJA, and A& B is logically equivalent to (AJA) | (B1B). Theorem 1.10. The only one-element adequate systems of binary connectives are {|} and {1}. Proof. Let g(z,y) be the truth function of a binary connective # forming an adequate system. Clearly, g(T,T)=F. For, if 9(T,T) were T, then any statement form in # alone would always take the value T when its statement letters all took the value T, and no such statement form could determine the negation operation. For the same reason (reversing the roles of T and F), o(F,F)=T. The situation at this stage is given by Fig. 1-30. Fig. 1-30 Case 1. The second row is F and the third row is T. Then A # B is logically equivalent to 1B, and all the statement forms in # alone using the letters A and B would be logically equivalent to one of A, B, 14, 1B. Then (#) would not be adequate. Case 2. The second row is T and the third row is F. This is handled in exactly the same way as Case 1, since A #B would be logically equivalent to 1A. Case 3. The second and third rows are F. Then # is |. Case 4. The second and third rows are T. Then # is |. >CHAP. 1) THE ALGEBRA OF LOGIC 19 LL 12, 13. Solved Problems Reduce the following sentences to statement forms. (a) A necessary condition for x to be prime is that 2 is odd or z= 2. (b) A sufficient condition for f to be continuous is that f is differentiable. (c) ‘A necessary and sufficient condition for Jones to be elected is that Jones wins 75 votes. (@) Grass will grow only if enough moisture is available. (¢) It is raining but the sun is still shining. (f) He will die today unless medical aid is obtained. (g) If taxes are increased or government spending decreases, then inflation will not occur this year. Solution: (@) P+ (OvD), where P is “z is prime”, O is “x is odd”, and D is “z (0) DC, where D is “fis differentiable” and Cis () BV, where E is “Jones will be elected” and V is “Jones will win 75 votes” (@ GM, where G is “grass will grow”, and M is “enough moisture is available”. (© R&S, where Ris “it is raining”, and S is “the sun is still (Note that “but” indicates conjunction, usually with an element of surprise.) () D> M (or, equivalently, IMD), where D is “he will die today”, and M is “medical aid is obtained”. () TvG+ 11, where 7 is “taxes are increased”, G is “government spending decreases”, and 7 is “inflation will occur this year”. is continuous”. Eliminate as many parentheses as possible from: (@) ([(AvB)>(10)] v [((1B)& OQ) &B)} (0) (A &(1(1B))] © [Bo (Cv B)}} (©) (Bo (CvB) @ (A&(1(1B))] Solution: (@) (Av B) +10) v (1B&CeB] (®) [A& 1B] [BO (Cv) () BO CvB) © (A& 1B) Write the truth tables for (2) (Av 1B) (C&A), (b) (A+ 1B) v (B> A). Solution: (@ A B (Av 1B) (C&A) T F T F F Tr T T T F T F T F F F F Tr T T F F T F20 THE ALGEBRA OF LOGIC [CHAP.1 @ A (Ae 1B)v BA) T T F T T T F T Write abbreviated truth tables for (@) (A>B)>B)v 1A, (b) (14&1B)> (BOO). Solution: (e) (A+B)*B)v 1A (+) (1A & 1B)+ (BOC TIT TT TFT FIFFT TTTT FITTT TTF TRPFFTTTTT TRPTF TFT PrETR TFET FIP PP TTF TRTTP PFFT FIFFTTTFF TFEFFFTTTFF PrFTF TFTF TRTTF TRTP Show that the following are tautologies. (a) (A@(A& 1A) 14, (0) (A+B) +0) ((C+4)+(D>A)), (©) (A+B) > ((B>C)> (A> 0). Sota (a) (b) Instead of using a truth table, we show that the statement form cannot be F, Assume that some assignment makes it F. ‘Then ((A~ B)-+ C) is T while ((C>A)->(D~A)) is F. Since the latter is F, CA is T but DA is F. Since the latter is F, D is T and A is F, Since C+ Ain T and A is F, C must also be F. Since ((A +B) ~>C) is T and C is F, it follows that A~BisF. But this is impossible, since A is F. (©) As in (8), we shall show that the statement form is a tautology by proving that the assumption that it is ever F leads to a contradiction, Assume that some assignment makes it F. Then AB is T, while (B+ C)->(A~C) is F, Since the latter is F, B->C is T and AC is F. Since the latter is F, A is T and C is F. Since BC is T and C is F, it follows that B is F. Sinee A +B is T and B is F, we know that A is F, contradicting the fact that A is T. Show that the following are contradictions. (0) (Av B) & (Av 1B) & (TAvB)& (1Av 1B), (&) (A&O)v (BETO) © [(14&C)v (1BE NC). Solntis (@) Any truth ansignment to A and B makes one of the conjuncts false. (®) Let A stand for the statement form.CHAP. 1] THE ALGEBRA OF LOGIC 21 La 18. (A&C) V (BEIGE) | (AED V(1BEI0) Bas Basa ms st td | Prove that if A and A~ B are tautologies, then so is B. Solutio ‘More generally, any truth assignment making both A and A-B true must also make B true. For, if 8 were false, then, since A is true, A+B would be false by virtue of the truth table for ~. Prove: (a) If A logically implies B and B logically implies C, then A logically implies C. () If A is logically equivalent to B and B is logically equivalent to C, then A is logically equivalent to C. (0) A is logically equivalent to A. (d) A logically implies A, (e) If A is logically equivalent to B, then B is logically equivalent to A. Solution: (@) To show that A logically implies C, we shall show that, whenever A is T, C must also be T. peers T. Since A logically implies 8, B must be T. Since B logically implies C, C must (®) If A and B always take the same truth valu then A and € always take the same truth vah (©) A always takes the same truth value as A. (@ follows from (e). (© This is clear from the definition of logical equivalence. and B and © always take the same truth value, For each of the following, find a logically equivalent statement form in disjunctive normal form: (@) (AvB)&(1BVO), (bt) TAv (B10). Solution: (@) (Av B) &(ABVO) () IA V(B+10) (A& 1B) v (A&C) v (BE 7B) v (BEC) TA v (IBV 10) (A&B) v (A&C) v (BEC) qAv By IC22 110. 1a THE ALGEBRA OF LOGIC [CHAP.1 For each of the following, find a logically equivalent statement form in full disjunctive normal form (with respect to all the variables occurring in the statement form): (@) (A&1B)v (A&O) (0) B>(Av10) (0) (AvB) eo TC (@ (A+B)>(B+C)> (AO) Solution: (2) (A&7B)V (A&C) (A& B&O) v (A& BEC) (ARBEC) ® AvBerc (Av B)& 10) v (AV B)&O) (AB IOV B&AD)v (1A 4 BEC) (A&BE AC) v (AK TBE IC) Vv (A&BE IC) V (TAKBE IC) v (14 & 1BEC) (A&BE AC) v (A&BE AC) v (TA&BL NC) v (1A & BEC) (e) B>(Av 1c) ABV (AV 1O) aBvaviac (@) WA>B)v (B+C) + (A +0) TIA 1B) v (B>0)v (AO) (A&AB)y (TWBE IDV AVE) (A&1B)vV B&IQV AVE (AE ABEO)V (AG IBEIO) V (ABBE IOV (1AGBE NC) V (TAR BEOQY (1A& BAO) (1A 8 BE NC) v (A&BEC) Two statement forms A and B in full dnf (with respect to the same statement letters) are logically equivalent if and only if they are essentially the same (ie. they contain the same fundamental conjunctions except possibly for a change in the order of the conjunets in each conjunction). & ... &B, (where each ‘Under the truth assign- ¢ literals 8, and assigns F to a statement letter if its denial is one of the literala B, By & ... & 8, is T, and hence A is also, But every other essentially different fundamental conjunction is F, and therefore B must be F. Thus A and B ‘eould not be logically equivalent. By a fundamental disjunction we mean either (i) a literal or (ii) a disjunction of two or more literals no two of which involve the same statement letter. One fundamental disjunction A is said to be included in another B if all the literals of A are also literals of B. A statement form A is in conjunctive normal form (cnf) if either (i) A is a fundamental disjunction or (ii) A is a conjunction of two or more fundamental dis- junctions of which none is included in another. A statement form A in enf is said to be in full enf (with respect to the statement letters Si, ..., S:) if and only if every conjunct of A contains all the letters Si, ..., Sx. (2) Which of the following are in enf? Which are in full enf? () (AvBv1Q)&(Av 1B) (iii) (AvB) & (BV 1B) (il) (Av By 10) & (AvB) (iv) 14CHAP. 1] THE ALGEBRA OF LOGIC 23 (®) The denial of a statement form A in (full) dnf is logically equivalent to a statement form B in (full) enf obtained by exchanging & and v and by changing each literal to its opposite (i.e. omitting the negation sign if it is present or adding it if it is absent). (Example: 1((A & 1B &C) v (14&1B&C)) is logically equivalent to (TAvBy 10) & (AvBv 10)) (©) Any non-tautologous statement form A is logically equivalent to a statement form in full enf (with respect to all statement letters in A). (@ For each of the following, find a logically equivalent statement form in enf (and one in full enf). (i) (A+ 1B) &(Av(BEC), (i) (A&B) v (1A & 1B), (iii) A & (BV 10). (@) Given a truth table for a truth function (not always taking the value T), construct a statement form in full enf determining the given truth function. Solution: (a) (i) In enf, but not in full enf. (ii) Not in cnf, since one conjunct is included in the other. (iii) Not in enf, since By 1. not a fundamental disjunction. (iv) In full nf, (0) This follows by several applications of De Morgan’s Laws (Example 1.28(a)). (© Astume A is non-tautologous. Then TA is not a contradiction, and, by Theorem 1.7, 1A is logically equivalent to a statement form in full dnf (with respect to all the statement letters in A), Hence by part (6), 17 is logically equivalent to a statement form in full cnf. But A is logically equivalent to 171A. @ 0) (A>78)& (Av B&Q) (VAv 1B) & (Av B)&(AvC)) (TAv 1B) &AVB)&IAVO) (ent) (Av By 0) & (1A Vv IBv 10) & (Av By ©) & (AV Bv 10) & (Av IBV C) (i) (4&B)v (14 1B) (Av 7A) & (Av 1B) & (By 14) & (BV 1B) (Av 1B) &(Bv 14) (fullenfy (i) A (BV 10) (A > (By 10) & (By 10) + A) (Av Bv 10 & (BY TO)V A) (Av Bv 10) & (IB& QV A) (1A v By 10) & (ABV A) & (CVA) (TAV BV 10) &(AVAB)E(AVC) (ent) (AV BV 10) & (Av BVO) & (Av TBv 10)& (AvBVC) —(fullenf) (©) Use the same procedure as in the proof of Theorem 1.8, except that we use only the rows ending in F (rather than T), we exchange & and v throughout, and we replace each literal by its opposite, Example. (full enf) F T T T F F T T o Answer: (Way By 10 & (1AV BVO & (Av BVO)24 1.13, 1 115. 1.16. THE ALGEBRA OF LOGIC {CHAP. 1 Find a statement form in 11, &, v determining the truth function f(A, B,C): Solution: (A&B&C) v (ABBE) Find a statement form in the Sheffer stroke | alone and one in | alone logically equiv- alent to the statement form A & 1B. Solution: For the Sheffer stroke, A&B W1AvB) (A A)vB) ‘(A | A) | (41 Ay] | BBY (AA) {CA 1D] | BLY) | ((l4 | A) 14 [ 4) | (1B) For $, A&B, AG&(BIB), (ALA)S (B4B)4 (B48) Show that {+,v) is not an adequate system of connectives. Solution: If Aisa statement form in ~,v, then A takes the value T when the statement letters are T (sinee T+T=T and TvT=T). Hence negation is not determined by any statement form in Prove that {1, +») is not an adequate system of connectives. Solution: ‘The eight truth functions in the following diagram are the only ones determined by statement forms in 1, «>, For, if we apply 71 to any of them or if we apply + to any two of them we obtain another of them. Alternative solution, We shall show that the truth function determined by 2 statement form in 7, # takes T an even number of times. This is clearly true for statement letters, and, when it holds for A, it must hold for TA. It remains to show that, if it holds for A and B, it also holds for AGH, Let n be the number of rows in the truth table. 1 is even (since 1 is of the form 2, where = 1), Let j and [ be the number of T's of A and B respectively. Let m be the number of T's of ‘AG, and let « be the number of T's of TA& 1B, Then j+!—m=n—e; hence j+—n=m—s, Since j,1,m are even, it follows that m—a is even, i. m and s have the same parity (both odd or both even). Hence m+2 is even. But m+ is the number of T's of A¢>B.CHAP. 1] THE ALGEBRA OF LOGIC 25 Lit. 1.18. Determine whether the following arguments are correct by representing the sen- tences as statement forms and checking to see whether the conjunction of the assump- tions logically implies the conclusion. (a) Either Arlen is lying or Brewster was in Mexico in April or Crawford was not a blackmailer. If Brewster was not in Mexico in April, then either Arlen is telling the truth or Crawford was a blackmailer. Hence Brewster must have been in Mexico in April. (®) If the budget is not cut, then a necessary and sufficient condition for prices to remain stable is that taxes will be raised. Taxes will be raised only if the budget not cut. If prices remain stable, then taxes will not be raised. Hence taxes will not be raised. Solu (@) Assumptions: Av BV IC, 1B (1AvO. Conclusion: B. Does (A v By 10) & (1B + (1A C)) logically imply B? In other words, is [(A v Bv 10) & (1B +(1Av Oy] +B a tautology? Let us try to find a truth assignment making this statement form false. ‘Then the antecedent (A vB v ‘1C) & (1B ~ (1A v C)) must be T and the consequent B must be F. Hence Av Bv 1C is T and 1B~ (AVC) is T. Since Bis F, 1B is T; and ‘therefore since 1B (1A. C) is T, it follows that 1AvC is T. Since AvBv 1C is T but Bie F, it follows that Av 1C is T. It is clear now that, if we take A to be T, C to be T, and B tobe F, then the statement form is F. Therefore the conclusion is not implied by the premises, (®) Assumptions: 1B- (PR), R->1B, P+ 4K. Conclusion: 1B. Does (1B + (P<>R)) & (R> 1B) & (P+ 1R) logically imply 127 Let us try to find a ‘truth assignment making the former true and the latter false. Then 1B ->(P¢>R) is T, RB BisT, and P+ 1K isT. Since Ris F, Ris T. But R- 1B is T, and therefore 1B is T. Hence by the trath of 1B + (P-<>R), (P<>R) is T. Since R is T, P must be T. ‘Then since P+ 12 is T, 12 is T, which is impossible. Therefore the argument is correct. (a) and (6) can be solved by writing down the truth tables, but the method used above is usually faster, Are the following assumptions consistent? f will be continuous (D) if g is bounded (C) or his linear (E). g is bounded and h is integrable (B) if and only if h is bounded (A) or f is not continuous. If g is bounded, then is unbounded. If g is unbounded or h is not integrable, then A is linear and f is not continuous. Solution: Assumptions: (Cv E)-*D, (C&B) (AV 1D), C+ 7A, (1Cv 1B) +(E& 7D). Assume that these are all T. Case 1. Cis T. Then A is F. Since CvE is T, Dis T. Therefore Av 1D is F. Hence C&BisF. Hence Bis F,and 1Cv 1B isT. Thus 2 & 1D is T, and D is F, which is impossible. Case 2. Cis F, Thus 1C is T, and therefore so is 1Cv 1B. It follows that EB & 1D is T, and therefore E is T and D is F. Since Cv is T, then D is T, which is a contradiction. Hence the assumptions are inconsistent. ‘This and similar problems ean also be solved by writing out the complete truth table (which, in this case, has sixteen rows)26 1.19. 1.20. Lat, THE ALGEBRA OF LOGIC [CHAP.1 If A is a statement form in 1, & v, and At results from A by interchanging & and-v and replacing all statement letters by their denials, show that A* is logically equivalent to 1A. Solution: Apply De Morgan's Laws (Example 1.28) to 1 until no denials of conjunctions or disjunctions remain. The result is A*, Let A and B be statement forms in 1, & v. By the dual A‘ of A we mean the siate- ment form obtained from A by interchanging & and v. Notice that (A‘)= A. (a) Show that A is a tautology if and only if 1(A‘) is a tautology. (b) Prove that, if A~B is a tautology (ie. A logically implies B), then B¢> A‘ is a tautology (i.e. B logically implies A‘), (Example: Since A & B~ A is a tautology, soisA+AvB.) (0) Prove: Ao B is a tautology (ie. A is logically equivalent to B) if and only if A¢ «+ B¢is also a tautology (ie. At is logically equivalent to 8). (Example: Since 1(A.v B) is logically equivalent to 14 & 1B, it follows that “1(A & B) is logically equivalent to 14 v 1B.) Solution: (a) By Problem 1.19, 1(A¢) is logically equivalent to (A¢)*, where (A‘)* is obtained from A‘ by exchanging & and v, and replacing all their statement letters by their denials. But then (A®)* is obtained from A by replacing all statement letters by their denials; and therefore if A tautology, so is (A®)* (by Theorem 1.1); and conversely if (A4)* is a tautology, so is A. (In this case we substitute for each statement letter its denial and then again use Theorem 1.1 plus the Law of Double Negation (Example 1.24).) () Assume AB is a tautology. Then 1AV = is a tautology, and, by part (2), 1((7Av B)4) is a tautology. But 1((1Av By!) is 1(1A¢& B4), which is logically equivalent to Bt + Ad, (©) Assume A<+B is a tautology. Then A~B and 8+A are tautologies. By part (b), B+ Ad and A‘ Bé are also tautologies. Hence A <> B¢ is a tautology. Conversely, if A¢ > Bt is a tautology, then, by what we have just proved, (A¢)# «> (8#)¢ is a tautology. But (A‘)¢ is A, and Supplementary Problems Write the following sentences as statement forms. (e) A depression will occur if government spending does not inerease, and inflation will result only if government spending increases. (®) Jones will lose his job unless he makes good on the deficit, although Jones is the cousin of the boss's wife, (@) Either f is discontinuous or if f is nonlinear, then f is differentiable. Assume that the truth values of A, 8, € are T, P, F. Compute the truth values of (a) (A— 18) (AVC) ER), (6) (A (AB) v (AC).CHAP. 1] THE ALGEBRA OF LOGIC 27 12s, 125, 1.38, 12. 1.28, 129, 1.80. 131. 132, If AB is T, what can be deduced about the truth values of (@) (AVO)+(BvC), (8) (A&C) (B&O), (0) (TALB) (AVE In each of the following cases, what further truth values can be deduced from those already given? (@) Av (A+B), (6) WA&B)(1Av 7B), (0) (TAVB) 2 (A+ 10). F T F ‘Which of the following are statement forms? For the statement forms, determine the principal ‘connectives. (@) (Av (1B) + A) & (14), @) (A > B) > A) > A) vB), (0) (Av B)v C) # (1B) Eliminate as many parentheses as possible from () KA v (1B) v C} & [Av (109B))}} & (14) @) (Av B) v (Cv (BY) ©) (A >(1BvEy) > (A) > BY), ‘Write truth tables and abbreviated truth tables for the statement forms of Exereise 1.26, Determine which of the following are tautologies, which are contradictions, and which are neither. () (A+B) > 1B =A) & Ae) () (A+B) +B) +B (© (A+B) (C+D) + [B> (WDA) (C7 AY @ AP BOAeBea)) () B& WAVE) O (Av BvC) [(((A +B) + B)+C)>C} (0) (A+B) +A) (B>(B+ A) () (A+ @e1B) +74 ‘Show that A is logically equivalent to B if and only if A logically implies 8 and B logically implies A. Show that a statement form logically equi logically equivalent to a contradiction is Give an example to show that, if A logically implies 8, then B does not necessarily logically imply A. ically equivalent to E, those for D. Of the following pairs D and &, find those pairs for which which D logically implies E, and those for which E logically impl > . (@) € «> (AKC)v (B& IC) B+ (e+a) ) Av Bec) (AvB) o (Ave) © save) 1 @ Av (B&O) ave © ace Bera o Aewec) acnec28 1.33, 134, 1.35. 1.36. 1st. 1.38, 139, 140, a. 142, THE ALGEBRA OF LOGIC (CHAP. 1 Prove: A,v Ay v --- v A, is T if and only if at least one of the A's is T; and A, & A, & ... BA is T if and only if all of the A/s are T. Prove generalizations of DeMorgan’s Laws and the Distributive Laws, in the sense tha: che following pairs are logically equivalent: @ Way vay) TA, & a TA @) TAe... &aD TA Tay (© A& MV va) (A&B) Vv (AGB) @ AVG&...&3) (vB) &... &(AVB,) ‘Which of the following are fundamental conjunctions? (@) A&B&NA, (0) B&AA, (0) B&CHALC. Which of the following are in disjunctive normal form? (@) (A&B)V (TAB) (A410) @) B&IA)v (A&B) V (1A GB) (© A&W @ A&B) (ALB) v A&BEO) (2) (A&B) v (A& 1B) v (TA &B) v (14 & 1B) Find a statement form in disjunctive normal form logically equivalent to: (2) (A+B) & (BVO), (6) (AVIBVOC, (0) (Av By 10) & (BV Which of the following are in full dnf (with respect to all the variables occurring in the statement form)? For those not in full dnf, find a logically equivalent statement form in full dnf. (@) (A&B) (1A 1B) @ sBvAseo (0) (A&BEC)y (ABBE NO) v (AB 1BEO) @ (AeBve ‘Which of the following are in conjunctive normal form? In full enf? For any not in full enf (with respect to all its statement letters), find a logically equivalent statement form tn full enf. @ AvB () (AvB)& (Av 1B) & (AV BVO) © A&B @ (AvB)& (Av 1B) (© (A&B) (148 1B) ‘Which statement forms are in both dnf and enf? For each of the following truth tables (a,) (8), (c), construct a corresponding statement form in full dnf and one in full enf. Sawa ayaa Bsa og aaa A statement form A is a tautology if and only if it is logically equivalent to a statement form in fall dnf having 2* disjuncts, where x is the number of statement letters in A.CHAP. 1] THE ALGEBRA OF LOGIC 29 14, , 16. 146, ar, 148, 149, 150, 11. 1522 1.53.0 1.54.0 ind statement form in 7, &,v logically equivalent to the statement forms (a) (A > (BY C) +14, () (A+B) B) >A, Find a statement form in the Sheffer stroke | and a statement form in } logically equivalent to the statement forms (a) (A v 1B) & (A ~+C), (b) 1(A & 7B). Show that A+B is logically equivalent to 1(A<>B). (Cf. Example 1.2 on page 4.) Find a statement form logically equivalent to the denial of (4 v (B&0)) v (1A &(BvC)). Show that {>,-+,4,v)} is not an adequate system of connectives, Show that {7,+) is not an adequate system of connectives, For which binary connectives (] does {7, (]} form an adequate system of connectives? Determine whether the following arguments are correct. (a) If April is rainy, then flowers will bloom in May and mosquitoes will thrive in June. If mos quitoes thrive in June, then malaria will increase in July. If flowers bloom in May, there will be a lot of honey in September. If April is not rainy, then the lawns will be brown this summer. Hence either there will be a lot of honey in September and malaria will increase in July, or the lawns will be brown this summer. (®) If f is continuous, then g or h is differentiable. If g is not differentiable, then / is not continuous and f is hounded. A sufficient condition for g or A to be differentiable is that f be bounded. Hence ¢ is differentiable. (©) If an orange precipitate forms, then either sodium or potassium is present. If sodium is not present, then iron is present. If iron is present and an orange precipitate forms, then potassium is not present, Hence sodium is present. Check the consistency of the following sets of assumptions. (@) If the roof needs repair or the house has to be painted, then either the house will be sold or no vacation will be taken this summer. The house will be sold if and only if the roof needs repair and a vacation will be taken this summer. If the house has to be painted, then the house will not be sold or the roof does not need repair. ither a vacation will be taken this summer, or the house has to be painted and the house will be sold. (0) Either devaluation will occur, or, if exports do not decrease, then price controls will be imposed. If devaluation does not occur, then exports will decrease. If price controls are imposed, then exports will not decrease. A computer (called Farfel) has been built to answer any yes-or-no question, but it has been pro- grammed either to answer all questions truthfully or to give incorrect answers to all questions. If we wish to find out whether Fermat’s Last Theorem is true, what question should we put to the computer? (Hint: Let A stand for “Fermat’s Last Theorem is true” and let 8 stand for “Farfel answers all questions truthfully”. Construct a statement form A such that, if “A?” is put as a question to Farfel, then the answer will be ‘“Yes” if and only if A is true.) Find the dusls of atatement forms in 1, &, v which are logically equivalent to A+B and AB, and extend Problems 1.19-1.20 to statement forms in 1, &, V, +, ©. Prove that a statement form A whose only connective is <> is a tautology if and only if every state- ‘ment letter occurs in A an even number of times. (Hint: Problem 1.32(¢, f).)Chapter 2 The Algebra of Sets 21 SETS By a set we mean any collection of objects.t For example, we may speak of the set of all living Americans, the set of all letters of the English alphabet, or the set of all real numbers less than 4. In most cases, sets will be defined by means of a characteristic property of the objects belonging to the set. In the examples above, we used the properties of being a living American, a letter of the English alphabet, or a real number less than 4. Notation: For a given property P(z), let (x: P(z)} denote the set of all objects x such that P(z) is true. Example 2.1. ‘The set of all real roots of the equation x*—2z?—3=0 is denoted by {z: zis areal number & xt—22?~3 = 0) Sometimes we shall define a set merely by listing its elements within braces: {a,,c,...,h}. In particular, (6) is the set having } as its only member. Such a set (b) is called a singleton. The set (b,c) contains b and c as its only members, and, if b xc, then (8, c) is called an unordered pair. Notice that (b,c) = {c, b). Example 22. ‘The net of integers strictly between 1 and 5 is equal to {2, 8, 4}. ‘Example 23. ‘The set of all real roots of the equation 2‘—2c?—3 = 0 is equal to the set (VS, —V3). We shall extend this method of denoting sets by listing a few elements of the set, fol- lowed by dots, in such a way as to indicate the characteristic property of the elements of the set. Example 24. {1, 2, 3, 4, } is intended to represent the eet of positive integers. {1, set of squares of positive integers. {Washington, Adams, Jefferson, Madi: of the United States. Definition: An object # belonging to a set A is said to be a member or element of A. We shall write x € A to indicate that is a member of A. The denial of 2 € A will be written 2 € A. Example 25. 6 € (x: eisaneven integer), 1 @ {e: 2 is an even integer} 16, 25, ..., n?, ...} is the ) is the set of Presidents ‘Synonyms for set are totality, family, and class. 380CHAP. 2] THE ALGEBRA OF SETS 31 22 EQUALITY AND INCLUSION OF SETS. SUBSETS ‘The sets A and B are equal when and only when A and B have the same members. Equality of A and B is designated in the usual way by A=B, and denial of this equality by A*B. Example 26. {ei st =1 and aise real number} = (e: 2 =1 or #=-1} We say that A is a subset of B if and only if every member of A is also a member of B. We write A C B as an abbreviation for: A is a subset of B. Sometimes, instead of saying that A is a subset of B, one says that A is included in B. The denial of A C B is written AGB. Example 27. {1,8} ¢ (42,3,6 {a} ¢ {ead (12,4) ¢ (2,5) Obvious properties of the inclusion relation are Inel(i) ACA (Refiexivity), Incl(ii) If ACB and BCC, then ACC (Transitivity). Inel (iii) A=B if andonly if (ACB & BCA). It is convenient to introduce a special sign for the relation of proper inclusion. We shall use A C Bas an abbreviation for ACB & A*B. Thus AC B if and only if every member of A is a member of B but there is a member of B which is not a member of A. If ACB, we say that A is a proper subset of B. Hence the only subset of B which is not a proper subset of B is B itself. The denial of A C B is written A ¢ B. Some basic properties of proper inclusion are: PI) AGA. PI(ii) If ACB&BCC, then ACC. PUiii) If AC B& BCO, then ACC. Pliv) If ACB, then BGA. Example 28. 1,3} ¢ (1.23: 0,8} ¢ (3 (4) € 3) 23 NULL SET. NUMBER OF SUBSETS ‘Whenever P(z) is a property satisfied by no objects at all, then {z: P(z)} is a set having no members. For example, {2:2} isa set with no members. We shall use Q to denote a set with no members. The set @ is called the null set or empty set. There is precisely one null set, since any two null sets would contain the same members (namely, none at all) and therefore must be equal. The null set is included in every set: @ CA for all A. Example 29. ‘The only subset of @ is @ itself. Example 2.10, ‘The subsets of {2) are Q and {=}. Thus a singleton has two subsets. Example 2.11. If «oy, the subsets of the unordered pair (x, y) are @, (=), {v} and {x,y}. Thus a two-element set has four subsets.82 THE ALGEBRA OF SETS [CHAP. 2 Example 212, If 2, y and s are three distinet objects, then the subsets of (z,y,2) are 9, (2), {y), (2), (ey), {2,2} {v2} and (,y,2). ‘Thus there are elght subsets of a three-element set. Let P(A) denote the set of all subsets of A. Then P(A) = (B:BCA}. Examples 2.9-2.12 suggest the following result. ‘Theorem 2.1. For any non-negative integer n, if a set A has n elements, then the set P(A) of all subsets of A has 2* elements. Firat proof: The result is clear when »=0 (Example 2.9). Assume a set A has elements, where % >0. In choosing an arbitrary subset C of A, there are two possibilities for each element 2 of A: #€C or x€C. Whether one element = is in the subset C is independent of whether any other element y is in C. Hence there are 2" ways of choosing a subset of A. Second proof: By induction on n. The case for n=0 is clear (Example 2.9). Assume that the result is true for n=, and assume that A is a set with k+1 elements, ie. A = (G1, ...,dy,dx+1}. We must prove that A has 2**1 subsets, Let B= {a, ..., a). Since B has k elements, then by inductive hypothesis B has 2* subsets. Every subset C of B can be thought of as determining two distinct subsets of A, ie. C itself and C together with the element as+:. In addition, every subset D of A is determined in this way by pre- cisely one subset C of B, ie. C is obtained by removing ax+1 from D (where, if a+: € D, then C is identical with D). Thus the number of subsets of A is twice the number of subsets of B. But since B has 2* subsets, A has 2**? subsets. > 24 UNION Given sets A and B, their union AUB consists of all elements of A or B or both. Thus AUB = (2: z€Av2€B). Remember that v stands for the inclusive “or”, ie. for any sentences A, B, Av B means A or B or both. Example 213. (1, 2,3) U (1,8, 4.6) = (1,28, 46) fe) U (= {0 0) {0,2, 4, 6,8, ...} U {1,8,5,7,9,...) = {0,1,2,8, 4,5, - If we represent the elements of A and B by points within two circles, then their union consists of all points lying within either of the two circles (see Fig. 2-1). The union operation on sets has the () obvious properties: Ui) AUA =A (Idempotence) Uli) AUB = BUA (Commutativity) Fig. 24 Uli) AUD =A Uliv) (AUB)UC = AU(BUC) (Associativity) Uw) AUB =B ifandonlyif ACB Uwi) ACAUB & BCAUBCHAP. 2] THE ALGEBRA OF SETS 88 25 INTERSECTION Given sets A and B, their intersection ANB consists of all objects which are in both Aand B. Thus, ANB = (2: 2€A & 2€B) Example 2.14, {1, 2,4) 0 (1, 8,4) = (1, 4) , 3,5) 0 @ 4,6) = 9 (1, 8,5).9 (0,2) = {0,1, 2) 1 {0, 3) = {0} Pictorially, we can visualize the intersection as consisting of the shaded area of Fig. 2-2. Two sets A and B such that ANB=@ are said () to be disjoint. ‘The following properties of the intersection opera- tion are evident. Fig. 22 Int(i) AMA = A (Idempotence). Int(ii) ANB = BNA (Commutativity). Int (iii) AND = g. Int(iv) (ANB)NC = AN(BNC) (Associ Int(v) ANB=A ifandonlyif ACB. Int(vi) ANBCA & ANBCB. The associative laws for unions and intersections allow us to omit parentheses in writing unions or intersections of three or more sets. Thus we write ANBNC to stand for either (ANB)NC or AN(BNC), since these sets are equal. Similarly ANBNCND has a unique meaning, since any of the five ways of inserting parentheses yields the same result. Important relations between unions and intersections are given by the distributive laws: Dist (i) AM(BUC) = (ANB)U(ANO), Dist (ii) AU(BNC) = (AUB)N (AUC). Property Dist (i) can be verified directly from the definitions by logical manipulations. mus. ANn(BUC) = (2: t€A & rE BUC) = (2: 264 & (2EBvzE0)} = (t: (6A & 2EB)v (ZEA & ZEO)) = (%: Z€ANBy cE ANC} = (ANB)U(ANG) We also can check Dist (i) pictorially. In Fig. 2-8 below, we have vertical lines for BUC and horizontal lines for A. Hence A N (BUC) is represented by the cross-hatched area, In Fig. 2-4 below, the vertical lines indicate ANB and the horizontal lines AMC. The com- bined area represents (AMB)U(ANC) and is seen to be identical with the cross-hatched area of Fig. 2-3. Dist (ii) may be handled in a similar manner (see Problem 2.3).Bd THE ALGEBRA OF SETS [cHaP.2 Fig. 23 Fig. 24 Diagrams as shown in Fig. 2-8 and 2-4 are called Venn diagrams. They are useful for verifying identities involving operations on sets, but should not be considered tools of rigorous mathematical proof. Similar pictorial methods can be given for four or more sets (see [112}t and J. F. Randolph, American Mathematical Monthly, 1965, pp. 117-121), but this does not seem fruitful enough to warrant consideration here. Example 215. Show that An(AUB) = A. By U(vi), ACAUB, Hence, by Int (v), AN(AUB) = Example 216. Show that AU(ANB) = A. AU(ANB) = (AVA) (AUB) = AN(AUB) = A ‘The first equality is justified by Dist (ii), the second by U(i), and the third by Example 2.15. The distributive laws have the obvious generalizations: Dist (i’) AM (BiUB.U---UB,) = (ANB) U(ANB) U- Dist i”) AU(BiNB:N---9B,) = (AUB) N(AUB)N- These can be proved directly, using mathematical induction. U(ANBy) n(AUB,) 26 DIFFERENCE AND SYMMETRIC DIFFERENCE By the difference B~ A of B and A we mean the set of all those objects in B which are not in A (the shaded area of Fig. 2-5). Thus, B~A = (a: 2€B& ZEA) Clearly, Di) a B~B=9 Di) B~p=B Dili) @~A=9 Pia. 25 Div) (A~B)~C = A~ (BUC) = (A~O)~B () The symmetric difference AAB of sets A and B is (A~B)U(B~A) (the shaded area of Fig. 2-6). Fig. 2-6 makes it clear that AAB = (AUB) ~ (ANB). Fig. 2.6 t Throughout this book numbers in brackets refer to Bibliography, pages 201-208.CHAP. 2] THE ALGEBRA OF SETS 35 The following properties are easily verified. SD) AsA=@ SD(ii) AaB = BaA SDiiii) Aag=A c {0,1,4,5). ‘Then BCA, CGA, ANC 2,3), C~B={4,5), ALB = {5}, ASC 27 UNIVERSAL SET. COMPLEMENT We shall often find it useful to confine our attention to the subsets of some given set X. In such a case, X is called the universal set or the universe (of discourse). The union, intersection, difference, and symmetric difference of subsets of X are again subsets of X. When we restrict ourselves to subsets of X, there is still another operation which can be introduced. If ACX, then the complement A of A is defined to be X~A. Thus, A = (2: €X & EA}. Whenever we use complements, it is assumed that we are dealing only with subsets of some fized universe X. The following assertions are easily verified. ci) Az=aA Ci) AUB = acat —_ De Morgan’s Laws Ci) ANB = AUB Le Civ) ANA =O Civili) ACB if andonly if BCA Cw) AUA=xX Clix) A = B ifandonlyit A = B Cwi) @=X C(x) A~B= ANB Cwi) T= Oxi) AaB = (AnB)u(AnB) From C(x) and C(xi) we see that difference and symmetric difference are dispensable in the presence of union, intersection and complement. Example 2.18, Let us check C(ii) using definitions and logical transformations. AUB = (2: 2€X &2€AUB) = (2: 2EX & 1eeAv zed} = (e: rEX & (EAGER) = (2: (ZEX& LEA) & (2EX& ZEB) = (i EX & ee Aln (ze: cEX&rEBY = An ‘We also may use Venn diagrams to verify the validity of C(ii), Compare Fig. 2-7 and 2-8, An B is the cross-hatched area. Fig. 2836 THE ALGEBRA OF SETS [cHAP. 2 De Morgan’s Laws C( have the obvious generalizations; Cli’) A;UA:0 = Andin:::nAe Ci) AN Asn A\UA.U+++U Ay Example 2.19. ACB if and only if ANB=9. hatched area is ANB. To say that thi saying that A is entirely within B. Logical proof: A = AnX = An(BUB) = (ANB)U(AnB) Hence if ANB=Q, then A= ANB; therefore, by Int (v), ACB. On the other hand, if ACB, then by Int(v), A= ANB and therefore AnB = (AnB)nB = An(Bnd) = And = 9 Fig. 29 28 DERIVATIONS OF RELATIONS AMONG SETS We have seen two ways of verifying propositions about sets: by means of analogous logical laws, or by pictorial methods (usually Venn diagrams). The first method is the only rigorous one, but the use of diagrams is sometimes quicker and more intuitive. Example 220. Prove A ~ (BUC) = (A~B)N(A~C). This is clear from Figs. 2-10 and 2-11. XY Cross-hatched: (4 ~ B)n (A ~C) Unshaded: A ~ (BUC) Fig.210 Fig. 21 ‘More rigorously, A~(BUQ) = {2: 2€4 & 2€(BUO)} fe: EA & (2@B& 2€0)} {e: (2EA &2@B) & (ZEA EQ) = (2: c€A & eEB) NO (x: 2EA & cEC} (A~B)nA~O) Example 221. Prove: (AUB)NB = A ifandonly if AnB=9. 2-12 below, the cross-hatched part represents (AUB) 1B and lies entirely within A. The rest, Tens-shaped intersection ANB. Hence to say that (AUB) NB is identical with A is equiva- lent to saying that ANB =.CHAP. 2} THE ALGEBRA OF SETS 37 Logical proof: (AvB)nB = (AnB)U(BnB) by Dist (ii)) = (AnB)ug (by Ciiv)) = anb (by UGii)) Hence (AUB)N B =A if and only if ANB =A. But AnB=A if and only if ACB (Int(v)), which holds if and 5 only if AnB=Q (by Example 2.19 and C(i). Fig. 22 Example 222. Simplity AnBu (Bn). AnBu(Bng = (AvB)u(Bnc) iby CUiii)) (duBvene (by ct) AvBveng) (by Viv) = AuB (by Example 2.16) In simplifying expressions, we make frequent use of De Morgan's Laws C(ii) and C(iii) for distributing complement bars over smaller expressions, C(i) for eliminating double complements, Examples 2.15 and 2.16, and the distributive laws Dist (i) and Dist (ii). Example 223. Simplify (A UBUC)n (AnBn@)Né. (AuBUO) Nn (ANBNE) NE = (AUBUQA(AUBUO) NE (De Morgan) = [AvBuan Aus) ne (Associativity of n) = [(And)u (Bug ae (Dist (i) = [Bug|né = Bnd ulend = (Bndvo (ctw), UU) = Bné 29 PROPOSITIONAL LOGIC AND THE ALGEBRA OF SETS Every truth-functional operation determines a corresponding operation on sets. For example, denial determines complementation: A = (x: I(z€A)}; conjunction determines the intersection operation: ANB = {z: z€A & EB); and disjunction determines the union operation: AUB = (x: s€Av Z€B). In general, if # is a connective corre- sponding to a truth function f(z:, ....2), then we define a corresponding operation @ on sets by @(Ax,...,4s) = (2: #(@EAs,...,2€A,)}. Thus the set-theoretic operation of symmetric difference corresponds to the exclusive usage of “or”. Example 2.24. The operation of alternative denial determines the set-theoretic operation AA B, while joint denial determines the operation An B, In general, a uniform way of determining the set-theoretic operation corresponding to a given truth function is to express the latter in terms of 1, &, v, and then replace 1, & v by ~, A, U respectively, The statement letters need not be replaced since they can serve as set variables in the new expression.38 THE ALGEBRA OF SETS {CHAP. 2 210 ORDERED PAIRS. FUNCTIONS If xy, then {z,y) was called the unordered pair of x and We say “unordered” because {x,y} ={v,z). Let us define an ordered pair (z,y), which is determined by x and v, in,that order. By this we mean that the following proposition holds: if (x,y) = (0), then 2=u and y=v. Theorem 22. (x,y) = ({2), (z,y}} is an adequate definition of an ordered pair. Proof. Assume (x,y) = (u,v). We must prove that z=u and y=v. We have {(2}, (zy) = ({u}, (4, 0}} (24) Since (x) is a member of the left side of equation (2.1), it must also be a member of the right side. Hence {2} = {4} or = {2} = {u,o} Therefore +=u or z=u=v. In either case, r=u, Now by (2.1), {ey} = (uy or = {zy} = {u,v} {u,v}, then (z,y)=(2,v) since z=u, Hence y=z or y=0. If y=z, {y,v} and y=v. In all cases, y=v. If {x,y} {u,v}, then (x,y) = {ue} u, By (21), {u,v} = (z} or = (u,v) = (2,9) a. Therefore y= If ey then {y) and so = md Since (u,v) * {x,y}, (u,v) = {2} and so u= Let us recall the definition of a function. A function f from A into B is a way of associating an element of B to each element of A. The phrase “way of associating” may be replaced by a more precise notion: (1) f is a set of ordered pairs such that, if (x,y) €f and (x,2)€f, then y (2) for every « in A there exists some y in B such that (z,y)€f. (Such an object y must be unique, by virtue of (1); it is denoted in the standard way by f(z).) We say that f is a function from A onto B if f is a function from A into B and every element of B is a value f(x) for some z in A. Example 225. ‘The function f such that f(z) = 2? for every 2 in the set A of all integers is a function from A into (but not onto) A. On the other hand, f is a function from A onto the set B of all squares. A function f is said to be one-one if it assigns different values to different arguments, ie. f(z) = f(y) implies z= y. Example 226. ‘The function f in Example 2.25 is not one-one, since f(—n) =? = f(n) for all integers n. On the other hand, the function g such that g(z) = 2 for all non-negative integers = is a one-one function, since w= of implies «=~ for all non-negative integers u and v. A one-one function from A onto B is called a one-one correspondence between A and B. For example, the function h such that A(z) = z+1 for all odd integers z is a one-one correspondence between the set of all odd integers and the set of all even integers.CHAP. 2 THE ALGEBRA OF SETS 39 211 FINITE, INFINITE, DENUMERABLE, AND COUNTABLE SETS A finite set is a set which is either empty or can be enumerated by the positive integers from 1 up to some integer m. More precisely, A is finite if there is a positive integer n such that there is a one-one correspondence between A and the set of all positive integers less than m. (When n= 1, A must be the empty set.) For example, to justify the assertion that the set of fingers on a hand is finite we set up the correspondence It is clear that a subset of a finite set is finite (and hence that the intersection of any set with a finite set is finite). Also obvious is the fact that the union of two finite sets is finite. A set is said to be infinite if it is not finite, Examples are the set of positive integers, the set of rational numbers, and the set of real numbers. Clearly any set containing an infinite set must also be infinite, and therefore the union of an infinite set with any other set is infinite. However, the intersection of two infinite sets need not be infinite. For example, the set of even integers and the set of odd integers have an empty intersection. A set A is said to be denumerable (or countably infinite) if and only if A can be enumer- ated by the set P.of all positive integers, i.e. if there is a one-one correspondence between Pand A. Example 227. (1) The set of positive even integers is denumerable, Here the one-one correspondence is given by Ain) = 2n. (2) The set of all integers is denumerable. Here the enumeration is given by 0,1, —1, 2-2, 3, wR ifmiseven + The oneone correspondence is g(n) = = yeaa Clearly the union of a finite set and a denumerable set is also denumerable. (Just enumerate the finite set first and continue with the enumeration of the denumerable set, omitting repetitions.) The union of two denumerable sets is again denumerable. (For, if A= {a1,02,...) and B= {bi,b:,...), then AUB= {ay,bi,d2,bs,...}, where in the latter enumeration we omit any repeated objects.) If we remove a finite number of elements from a denumerable set, the remaining set is still denumerable. A set is said to be countable if and only if it is either finite or denumerable. Obviously, any subset B of a countable set A is also countable, (For, in an enumeration of A, we omit all objects which are not in B. The resulting enumeration of B does or does not terminate. If it does, B is finite. If it does not, B is denumerable.) The union of two countable sets is ‘ountable set. This follows from what has been said above about finite and denumerable sets, 212 FIELDS OF SETS By a field of sets on X we mean a non-empty collection F of subsets of X such that, for any members A and B in F, the sets AUB, ANB, and A are also in ¥. Another way of expressing’ this is to say that is closed under the operations of union, intersection and complementation, Since AUB=ANB and ANB=AUB, it suffices to verify closure under complementation and either union or intersection.40 THE ALGEBRA OF SETS [cHAP.2 Examples of fields of sets are: (1) the set P(X) of all subsets of X; (2) the set of all finite subsets of X and their complements; (8) {9,X). Notice that any field ¥ of subsets of X must contain both @ and X. For, if B € F, then BEF and hence Q@=BNBEF. Therefore X= PEF. 213 NUMBER OF ELEMENTS IN A FINITE SET Let #(A) stand for the number of elements in a finite set A. Clearly #(ALUAs) = #(Ax) + #(As) — #(A1 Aa) For three sets, we have #(AUA2UAs) = #(A,) + #(As) + #(As) — #(A1Ay) — #(A1 As) — #(A2As) + #(A:NA:N Ad) and, for four sets, #(ALUA2UASUA,) = #(41) + #(As) + #(As) + #(Ad) — #(A1M As) — #(A1 Aa) — #(AINAd — #(AaM Ad) — #(A2NAY — #(ASNAD) + #(A1MALNAs) + #(ALM ARM Ag) + #(AIN ASN Ag) + #(ABNASNAD = #(AIN ALN AINAd) ‘The general formula for n sets should be clear from the examples for n = 2,3, 4, Example 228. In a two-party election district consisting of 135 voters, 67 people voted for at least one Democrat and 84 people voted for at least one Republican. How many people voted for candidates of both parties? #BnD) = #(R) + 4) — #RUD) = 84+67-195 = 16 Here R is-the set of people who voted for at least one Republican and D the set of people who voted for ‘at least one Democratic candidate. Hence RUD is the set of all voters and RMD is the set of all people ‘who split their ballots. Example 229, ‘A government committee reported that, among the students using marijuana, LSD or heroin at a certain university, 90% used marijuana, 6% used LSD and 7% heroin, while 4% took marijuana and LSD, 8% marijuana and heroin, 2% heroin and LSD, and 1% took all three. Are the committee's figures ‘consistent? Note that, if there are m students taking at least one of the drugs, and if H is = set of students, then the percentage in H is #(H)/n. Hence if we let A, B, C be the sets of students taking marijuana, LSD and heroin respectively, and we divide the equation for #(AUBUC) by m to obtain the percentages, %(AUBUC) = (A) + %4B) + %(C) — %(ANB) — %ANC) — BNC) + %ANBNC) 100 = 94+6+7-4- 5-241 = 93 ‘which is impossible. Hence the figures are not consistent.CHAP, 2] THE ALGEBRA OF SETS 41 Solved Problems 21. Show that the cancellation law if AUB = AUC thn B=C is false by giving a counterexample. Solution: A B=9. 22, Show that parentheses are necessary for writing expressions involving more than one of the operations and U. Solution: Consider ANBUC. This is either AN(BUC) or (ANB)UC. But these two sets are not necessarily equal. Take A=Q and B=C*Q, Then AN(BUC) =, but (ANB)UC = gue =6. 23. Prove the distributive law Dist (i), page 88: AU(BNC) = (AUB)N(AUC). Solution: Logical Proof. AU(BN® = (e: 264 v #€(BNO)} = {e: c€Av EB & 2€0) = (2: (FEA Vv SEB) & (CEA V ZEO} = (2: 2A v ZEB) n (z: ZEA v 2EC} = (AUB)N(AUG) Pictorial Proof. In Fig. 2-18, the vertical lines indicate BC and the shaded area is A. In Fig, 2-14, the vertical lines indicate AUB, the horizontal lines AUC, and the cross-hatched area (AUB) 1 (AUC) is identical with the marked area of Fig. 2-13. Fig. 213 Fig. 2:14 24, Prove the generalized distributive law Dist (ii), page 34: AU(BiN-++MB,) = (AUB) +++ N(AUBy) Solution: For #=1, the assertion is obvious and the case m=2 is the distributive law Dist (if). ‘Now using mathematical induction, we assume the result true for n=. Then for n=k+1, AU (B+ BNByas) = AU (BB) Bes) = [AU (Bn +BY] [AU Beal (by Distii)) (AUB) Nn --- M(AUBY] A AUB,s:) (by the inductive hypothesis) = (AUBy) 9 ++ 9 (AUB) A (AUBus)42 THE ALGEBRA OF SETS (CHAP. 2 DIFFERENCE AND SYMMETRIC DIFFERENCE 25. Using a Venn diagram, determine whether the following conditions are compatible. @) AnB=9 iii) (CNA)~B =O (i) (CNB)~A =D (iv) (CNA) U(CNB)U(ANB) * @ Solution: In Fig. 2-15, (iv) says that BUF UGUH is non-empty. (i) says that BUF is empty. Hence GH is non-empty. (iii) says that G is empty and (ii) says that H is empty. Hence G UH-is empty. Therefore conditions (j)-(iv) are inconsistent. 4 B 4 B (2 (- (52) (XN ‘c ‘c Fig, 215 Fig. 2.16 26. Show that 4a (BaC) = (AaB) aC. Solution: In Fig. 216, AAB=EuHUGUJ and C=HUIUIUK, and so (AAB)AC =EU GuIUK. BAC = FUGUHUK and A = EUFUHUI, and 50 4A (BAC) = GUKUIUE, Thus (A4B) aC = Aa (BAC). Observe that 2 A x 7 —— — a (AaB) ac = (anBdunBndu nBnG vu Anno) A logical derivation of this result is rather tedious and is left to the reader. It is easiest to prove by showing (AAB)AC C AA(BAC) and AA(BAC) C (AAB)AC. 2.7. Show that AB = @ if and only if A=B. Solution: AAB=Q ifandonlyif (A~B)U(B~A) = 9, if and only if A~B=Q and B~A ifand only if ACB and BCA, and only if A= B. Note: By C(xi), page 85, this result ean be restated as A=B ifandomyif (AnB)u(dnB) = 9 28. Prove the cancellation law: If AAB=AaC, then B=C. Solution: Assume ASB=AAC. Then ASAAB=AAAAC (parentheses can be omitted by virtue of Problem 2.6). Since ASA=Q, we obtain: PAB=QAC. But QAD=D for any D. Hence B=C.CHAP. 2| THE ALGEBRA OF SETS 43 29. Prove the distributive law: AM(BAC) = (ANB) A(ANC). Solution: An(Bac) = An(B~Q)U(C~B) = ANB~O)uAn(C~B) = (ANB) ~ C)U (ANC) ~ B) = (ANB) ~ (And) u (Ane) ~ (ANB) = (ANB) a (Ang The problem can also be handled by means of » Venn diagram, as in Problem 2.6. 2410. Prove C(iii): ANB=AUB, logically and pictorially. Solution: Logical Proof. ANB = {a: 2€X & 2€(AnB)} {z: 2€X & (z€A v 2€B)} (a: (@EX & 2@A)v (EX & 2EB)) = (@: 2€X & eeA}u ie: ceX & reB} = AUB Pictorial Proof. See Fig, 2-17 and 2-13, A: vertical, 5: horizontal Unshaded region: 705 Marked region: AUB Fig.2.17 Fig. 2:18 ACB if and only if Bc A. Recall that A and B are subsets of some fixed universe X. Then ACB if and only if, for any z in X, if EA, then z€B, if and only if, for any x in X, if eB, then 2€A,t it and only if, for any = in X, if z€B, then 2A, if and only if BCA. 2.12, Using mathematical induction prove the generalized De Morgan Law C(ii AiN---0A, = A,UA.U---UA, Solution: It is obvious for » ‘Then for n=k+1, aa The case n=2 is simply C(iii). Assume the result true for "=k. Antanas ey CU) (by inductive hypothesis) UA) Uns VALU Aga ‘We have used here the logical law of contraposition: P+ Q is logically equivalent to 1Q-> 1P.44 213, 244, 2.15. 2.16, THE ALGEBRA OF SETS ez Prove: AUB = AnB and ANB = AuB. Solution: By De Morgan's law C(ii), AUB =AnB. Henee XOB=AnB. put 7 Likewise, by De Morgan's law C(ii), 29 =AuB, Henee AnB = ANB = AvB. Prove: (1) (AUB)NA = BNA (8) (ANB)UB=B (2) (AnB)UA = BUA (4) (AUB)NB = B Solution: @) (4uB)NA = (And)u (Bnd) = QuiBnd) = BNA. @) (AnB)ud = (AvA)n (Bud) = Xn(Bud) = Bud. @) (ANB)UBCBUB = B. Also, BC(ANB)UB. Hence (AnB)UB = (4) (AUB)OB = (ANB)U(BNB) = (ANB)UB = B by (3). (@) Show that the four ellipses in the diagram below form an approprizs= diagram for four sets. D 4 B (®) Using the diagram of part (a), what conclusion can you draw from the = assumptions? (ij) Cc (BND) U(DNB). (ii) Everything in both A and C is either in both B and D or in neither B (iii) Everything in both B and C is either A or D, (iv) Everything in both C and D is either in A or B. Solution (a) Show that the fifteen regions of the diagram cover all possible cases: ANBnCnD, AnBn€nD, AnBacnD, ANBNCND, A~B-C-L. AoBnénB, AnBacnD, AnBnend, AnBoénD, A-B- rd. AnBncob, AnBoenB, AnBnenD, AnBncnb, A~B-O-5. (0) C=O. Algebra of Sets and Algebra of Logic. Given a statement form in 1, & be the expression obtained from € by substituting —, A, U for “1, & v res: Example: S(AvB)& 10) = (AUB)NO (e) Prove: A is logically equivalent to B if and only if S(A) = S(B) holds for a= === (where the statement letters of a statement form € are interpreted in S.C a= == variables ranging over all subsets of a fixed universe).CHAP. 2} THE ALGEBRA OF SETS 45 217. (6) Prove that A logically implies 8 if and only if S(A) C S(B) holds for all sets. Solution: (a) If we replace each statement letter t in A by z€t, then the resulting sentence is equivalent to ESA) (since z€ W,NW; if and only if 2eW, & zeWs; 2€W,UW, if and only if ZEW, v € Ws; and 2€W if and only if 1(2€W)). Hence if A is logically equivalent to 8, then «€S(A) if and only if s€S(B), which implies that S(A) = S(8). Conversely, assume that A is not logically equivalent to 8. In general if we are given a truth assignment to the state. ‘ment letters in an arbitrary statement form C, and if we replace statement letters which are T by (O) and statement letters which are F by Q, then, under this substitution of sets for statement letters, S(C) = (Q} if © is T under the given assignment, and S(C)=9 if © is F under the given assignment. This holds because, under the correspondence associating {9} with T and Q with F, the trath-functional operations correspond to the set-theoretic operations (where sets are restricted to subsets of the universe {9}). wT=F @=9 wWwe=T D = T&éT=T {9} 0 {9} = {0} T&F =F&T=F W119 = ONO =9 FaF=F OnO=9 TvT=TvVF=FvT=T (GUM) = UO = UD} = FVF=F DUD=09 Since A is not logically equivalent to B, there is a truth assignment making one of them T and the other F, say A is T and 8 is F. ‘Then under the substitution of {9} for the true state- ment letters and of @ for the false statement letters, S(A)={Q} and S(®)=Q. Hence ‘S(A) = S(8) does not always hold. Remark: Lurking behind this rather long-winded discussion is what in mathematics is called an “isomorphism” between the structures ATF}, 1,6 v) and (DD, Note that we also have shown that an equation S(A) = S(B) holds for all sets if and only if it holds in the domain {(9}, 0} of all subsets of (9). a, uy (®) A logically implies B if and only if A&B is logically equivalent to A. By part (a), the latter holds if and only if S(A&B)=S(A) always holds. But S(A&8) = S(A)NS(B), and ‘S(A)NS(B) = S(A) if and only if S(A)CS(B). Define ordered n-tuples (for x = 8) by induction: Tn-1), Bad (21, %2), 3), 20). Prove that if sy Ba = tne (sy Way oy Ba) = Ke By Aaa, te), ts) and (a1, a2, 2,24) (ti, Ua, oe ey Un), then 21 =u, Z2= us, . Thus (#1, 22, 23) (G1, aay... De) Solution: ‘We alrendy have proved this result for n= 2. Now assume it is true for n=k™2, and we shall prove it must then hold for k+1. We have, by assumption, ay Bay oe i ga) = (My May oy My Mead Hence by definition, Ce ay oon EDs esd = ay May oe tds Mead) May «oo ty)s But the By the result for n= 2, we conclude 2qs= esi and (224 ee te latter equation, by virtue of the inductive hypothesis, implies =;46 THE ALGEBRA OF SETS (CHAP. 2 FINITE, INFINITE, DENUMERABLE, AND COUNTABLE SETS 218, 2.19. 2.20, 2.21. Prove that the set W of all ordered pairs of non-negative integers is denumerable. Solution: Arrange W in the following infinite array: 0, 0) (0,1) (0, 2) ( 0, 4) ane Plea ae ee ae an an” an. a aa 2, 0) 2,1) (2, 2) 2, 8) 4 47 7 0) aD (8,2) 8, 3) \ 7 a1) ee . 4, 0) \ Enumerate the ordered pairs as indicated by the arrows, going up each diagonal from left to right. Notice that the pair (i,j) appears in the [((i+ ji ++ 10/2) + G+ 1)]th place in the enumeration. ‘This can be seen ag follows: all pairs in the same diagonal have the same sum. Adding up all pairs in diagonals preceding the one containing (i, j), we obtain tebe t Gta There are j pairs in the same (i, i) and preceding (i, 3). onal Prove that the set of all non-negative rational numbers is denumerable. Solution: Every non-negative rational number corresponds to a fraction m/n, where (i) m and m are non-negative integers, (ii) n%0, and (iii) m and » have no common integral factors other than #1. ‘We can associate the ordered pair (m,n) with m/n, and use the enumeration given in Problem 2.18, merely omitting those pairs (m,n) which do not satisty conditions (i)-(ii), ‘The set A of all real roots of all nonzero polynomials with integral coefficients (such roots are called real algebraic numbers) is denumerable. Solution: Any nonzero polynomial has only a finite number of roots. First list the finite set of all real Toots of all polynomials of degree at most one whose coefficients are in magnitude “1 (i.e. whose coefficients are either 0,1, or 1), ‘Then list the finite set of all real roots of all polynomials of degree 2 whose coefficients are in magnitude <2, ete. In general, at the nth step we list the finite set of all real roots of all polynomials of degree =n whose coefficients are in magnitude , Determine whether each of the following sets of ordered pairs is a function. () (@v): wand y are human beings and z is the father of y). ) (z,y): and y are human beings and y is the father of =}. (ii) ((e,y): = and y are real numbers and 2?+ y? = 1}, (iv) (ey: @=1 & y=2) v (2-1 & y=0)). For each of the following functions f from the set of integers I into I, determine whether f is function from I onto I and also whether / is one-one. eth e-1 ifs isodd (i) fl) =-« (iv) fle) = 22-3045 (i) Me) = Bet Git) f(@) = { Prove: The set of all rational numbers is denumerable. Prove: The set of all irrational real numbers is not countable. By a left-open interval of the set R of real numbers, we mean either an interval (ab) = @: a<2=d) or an infinite interval of the form (we) = (2: ace) or (eb) = (2: 2B} Let F be the collection of sets of real numbers consisting of @, R, and all unions of a finite number of left-open intervals. Show that F is a field of seta.CHAP. 2] THE ALGEBRA OF SETS 51 269, 270, an. 2m. 213, 214, (1) For finite sets A, B, C, D, derive the formula #(AUBUCUD) = #(A) + #(B) + (0) + #—D) = #(ANB)— #(A NC) — #(4nD) - #(BNO) — #~BND) — END) + #(AMBNO) + #(ANBND) + #ANEND) + #(BNCND) = #(AnBncnD) (2) Generalize the result of (1) to any finite seta Ay, ..., Ay. In an advertising survey of a hundred coffee and tea drinkers, 70 people were found to drink coffee at times, and 30 people drank both tea and coffee, How many people sometimes drank tea? Among Americans taking vacations last year, 90% took vacations in the summer, 65% in the winter, 10% in the spring, 7% in the autumn, 55% in the winter and summer, 8% in the spring and summer, 6% in the autumn and summer, 4% in the winter and spring, 4% in the winter and autumn, 8% in the spring and autumn, 3% in the summer, winter and spring, 3% in the summer, winter and autumn, 2% in the summer, autumn and spring, and 2% in the winter, spring and autumn. ‘What percentage took vacations during every season? If A and B both contain n elements, prove that A —B and B— A have an equal number of elements. Show that this is no longer the ease when A and B are infinite. (a) Show that the maximum number of sets obtainable from A and B by applying the union and difference operations is eight. (b) Show that the maximum number of sets obtainable from A, B, C by applying the union and difference operations is 128 = 2". (Hint: How many regions appear in the Venn diagram for A,B,C?) (c) Generalize (4) to the case of m sets Ay, ..-» Ay (0) Prove: (ANBNC)U(ANBND)u (AnCnD) U(BncnD) = (AUB)n (AUC) (AUD) n (BUC) A (BUD) N (CUD). () Prove: (AMB)U (ANC) U (AND) U (BNO) u (BND) U (END) = (AUBUOQ)0(AUBUD)N (AUCUD) n (BUCUD). (©? Prove the following generalization of (a), (b) and Problem 2.41: Given » sets Ay, «++, Awe Let # =n. Show that the union U of all intersections of k of the sets Ay, ..., Ay is equal to the intersection I of all unions of n—k+41 of the sets Ay, ...,Ay. (Note: In Problem a1, 3 in(®), n= 2) Hint:" Prove UCI and ICU.Chapter 3 Boolean Algebras 3.1 OPERATIONS An n-ary operation on a set Y is defined to be any function f which, to each ntuple (Yi, ++ +sYa) of elements ¥:,...,¥n in Y, assigns an element f(y:,...,¥s) in ¥. A more traditional way of asserting that f is an n-ary operation on Y is to say that Y is closed under the funetion f. Example Addition, multiplication and subtraction are binary operations on the set of integers. (We use “binary” instead of “2-ary”.) The function f such that /(z)= 2-1 for every integer x is a singulary operation on the set of all integers. (We use “singulary” instead of “1-ary”.) Example 32. ‘The subtraction function 2—y is not a binary operation on the set of non-negative integers, because the value =—y ia not always a non-negative integer. The division function z/y is not » binary operation an the set of positive integers. (Why?) 32 AXIOMS FOR A BOOLEAN ALGEBRA By a Boolean algebra we mean a set B together with two binary operations » and v on B, a singulary operation ’ on B, and two specific elements 0 and 1 of B such that the follow- ing axioms hold. (1) Forany zandyinB, zvy = yve (2) ForanyzandyinB, ay = yaz (8) For any 2, y, in B, x a(yvz) = (eay)v (waz) (4) For any 2, y,zinB, vy (ya2z) = (zvy)a (eve) (8) ForanyzinB, zv0 (6) For any zinB, zal = 2. (1) ForanyzinB, xv’ = 1. (8) ForanyzinB, zaz’ = 0. (9) O* 1. A Boolean algebra will be designated by a sextuple (B, »,v,’,0,1). Sometimes one refers to the set B as a Boolean algebra, but this is just a loose misuse of language. } Commutative Laws } Distributive Laws a Example 33. (@) The two-clement Boolean algebra By = KD.(9)},0, U,~ 9.49) where B= (9, (9): A = the ordinary set-theoretic intersection operation N; v = the ordinary set-theoretic union operation U; ’ = the ordinary set-theoretic operation of complementation: oO ; and 1={@}. In Chapter 2, we have verified properties (1)-(9). Of course, we first must note that 1, U and — are operations on (0, (9)). 52CHAP. 3} BOOLEAN ALGEBRAS 53 (®) The Boolean algebra of all subsets of = non-empty set A, under the usual operations of inter- section, union, and complementation, and with Q and A as the distinguished elements 0 and 1: (2(A), 1, U,~, 9, A). When there is no danger of confusion, we shall refer to this Boolean algebra simply as P(A). Part (a) is a special case of (b) when A = {Q). (Notice that we have omitted the case where A =9; in this case, 0= = A = 1, violating Axiom (9).) Example 34. (This example should be omitted by those not familiar with elementary number theory.) Let B be the set of all positive integers which are integral divisors of 70, Thus, B = {1,2,5,7,10,14, 35, 70} For any 2 and y in B, let xy be the greatest common divisor of # and y, let zvy be the least common multiple of and y, and let 2’ = 70/z. (For example, 5414=1, Sv 14= 70, 10035 =, 1038 = 70, ¥=14, 10'=7) ‘Then (B,A,v,',1,70) is a Boolean algebra. Verification of Axioms (1)-(9) uses ele- mentary properties of greatest common divisor and least common multiple, Example 35. It seems evident that a set of sentences closed under the operations of conjunction, disjunction and negation should form Boolean algebra. However, this is not quite so. For example, A&B and B&A are not equal, but only logically equivalent. ‘Thus we should have to replace the equality sign = in Axioms (1)-(9) by the relation of logical equivalence. In addition, 0 could be any sentence of the form A& 7A, and 1 could be any sentence of the form Av A. If we wish to retain the equality sign = with its usual meaning (ie. identity), then we may proceed as follows. By the statement bundle (A) determined by a statement form A, we mean the set of all statement forms whi logically equivalent to A. ‘Then, it is clear that: (i) [A] =(B} if and only if A is logically equivalent to B; (ii) if [A] (8), then {A]N [8] = 9. If K, and X; are statement bundles, it is obvious that if A, and B, are statement forms in Hy and Ay and B, are statement forms in K,, then A, & A, is logically equivalent to B, & Ba, A, v Az is logically equivalent to By v 8, and 1A, is logically equivalent to 18,, Therefore if we take an arbitrary statement form C, from K, and an arbitrary statement form C, trom K,, then we may define K, & Ky to be [C; & Ga}, Ky v Ky to be (C, v Gy], and Kj to be [16,}. If B is taken to be the set of all statement bundles, 0 is taken to be [A & 7A], and 1 is taken to be [Av 1A}, then (B,&,v,’,0,1) is a Boolean algebra. Verification of Axioms (1)-(9) reduces to well-known properties of the algebra of logic. (For to check Axiom (1), we consider any statement bundles XK, and K,, and we take any statement snd Cp in K, and K, respectively, Then Kyv Ky = (Cv G3) and K,v K, =(C,vC,]. But [6,v G4] = [C,V 6], since C,v G, is logically equivalent to €,v C;,) Terminology: =, y is called the meet of x and y. vy iscalled the join of x and y. 2! ig called the complement of 2. 0 is called the zero element. 1 is called the unit element. If it is necessary to distinguish the meet, join, complement, zero element and unit element of a Boolean algebra @ from those of another Boolean algebra, we shall add the subscript B: Ag, vg)’, Og, ly. Unless something is said to the contrary, we shall assume in what follows that B =(B,,,v,0,1) is an arbitrary Boolean algebra. Theorem 3.1. Uniqueness of the complement: If zvy=1 and zay=0, then y= 2’. Proof. First, y = yv0 by Axiom (5) yv(erz’) by Axiom (8) (v2) »(yv 2’) by Axiom (4) = (nvy) (yy 2’) by Axiom (1) la(yve’) by hypothesis (yvayal by Axiom (2) yee’ by Axiom (6)54 BOOLEAN ALGEBRAS ICHAP. 8 Second, 2 = z’v0 by Axiom (5) = av (tay) by hypothesis = (#’v2)a(e’vy) by Axiom (4) = (tv2’)a(z’vy) by Axiom (1) = 1a(e'vy) by Axiom (7) = (evyal by Axiom (2) =evy by Axiom (6) =yve by Axiom (1) =y by the first part above > Corollary 82, For any « in B, (2’)’=z. (Notation: We shall denote (2’)’ by 2”, ((e’))’ by 2”, ete.) Proof. First, 2’ vz = 2v 2" by Axiom (1) =1 by Axiom (7) az! by Axiom (2) 0 by Axiom (8) Hence by Theorem 3,1, taking to be 2’ and y to be z, we obtain = Second, 2’ »z Theorem 33. Idempotence: For any z in B, @) zaz=z, (i) eve=e Proof. @ 2=201 by Axiom (6) (i) z= 2v0 by Axiom (5) = 2a(eve’) by Axiom (7) zv(eaa’) by Axiom (8) = (ena) v (eaz’) by Axiom (8) (eva) (av) by Axiom (4) = (eaz)v0 by Axiom (8) (eva)al by Axiom (7) Saat by Axiom (6) Save by Axiom (6) > Definition: By the dual of a proposition concerning a Boolean algebra B, we mean the proposition obtained by substituting v for », » for v, 0 for 1, and 1 for 0, by exchanging «and v, and exchanging 0 and 1. Example 35. The dual of 2A (vv dual of eve’=1 is eae’ AW (Az) is ev WAZ) = levy) A(ev2), and vice versa. The nd vice versa. It is obvious that if B is the dual of A, then A is the dual of B. Theorem 34. Duality Principle (Proof-theoretic version): If a proposition A is derivable from Axioms (1)-(9), then the dual of A is also derivable from Axioms (1)-(9). Proof. The dual of each of Axioms (1)-(9) is again an axiom: (1) and (2) are duals of each other, and so are the pairs (3)-(4), (5)-(6), and (7)-(8). (9) is its own dual. Thus if in & proof of A we replace every proposition by its dual, the result is again a proof (since axioms are replaced by axioms), but this new proof is now a proof of the dual of A.>CHAP. 3} BOOLEAN ALGEBRAS 85 ‘The proof of Theorem 8.8 is twice as long as it need be. Since zvz =< is the dual of Zax=Z, it would have sufficed to prove +x = and then cite the Duality Principle to obtain zvz=z. As a matter of fact, the proof of the Duality Principle is illustrated in the proof of Theorem 3.8: the proof of (ii) is obtained by taking the duals of the proposi- tions in the proof of (i). Theorem 35. For all x, y,z in B, @) 200 =0 (i) ozvl=1 (ii) ea(evy) = (iv) ev @ay) = () [yaw = ene & yaw = 2ar]ry ae : } Absorption Laws (Wi) ev (va) = vv } Associative Laws (vii) = (eayyaz (viii ae } De Morgan's Laws (ix) avy ) (ayy (xi) @vyy (xii) Oo aa (ii) o = 1 (iv) = 0 (vl ea(a’vy) = aay (xvi) ev (e’ay) = avy Proof. From now on, we usually will not cite the particular axioms or theorems being used in a proof. @) BAO = A VO = (end) v (enw) = (ena) v(ZA0) = ta(av0) = eae =0 (ii) is the dual of (i). (iil) 2 (ev) = (@v0)a (zv9) = evry) = 2v0 = 2. (iv) is the dual of (iii). (v) Assume yar = 2nz & yar = ean’. Then y= yal = aleve) = (yrz)y yar’) (eaz)v (ean) = za(eve) = eal ae (vi) We shall use (v), replacing y by v (yv2) and z by (xvy)v 2. Thus to apply (v) we must show () @vyva)az = (avy)ve)ae and (0) (ev (yva))az = (evave)az’ Also, To prove (a): (zv(yvz)) aa = za(ev(yvz)) = x by (fi (avv)v a) az = ea((eva)va) [ea @vy)] v [ene] ev (eaz)_ by (iii) = by (iv)56 BOOLEAX A=SEBRAS [omar. 3 ‘Thus (evQyve)e = = = = ((evy)vz)az To prove (0): (zv(yv2))az’ = 2+ =~ (uve) ev (et a(yve)) B. et) va) = 2 a(yva) Also, (avy) va)az’ v8) = (24 (ey) v (@ a2) Paw) y (22) - [x’az] . org) = as (yva) ‘Thus (ev Qva)) ae = + ve) = ((Rvy)v2)nz (vii) is the dual of (vi). (viii) To prove (2vy)' = 2’ ay’, we use sie 2 <2 (Transitivity) (P03) (s=y & y<2)>2 = y (Anti-symmetry) Proof. (P01) zaz=z. (PO2) Assume xy = 2 and yas =y. Then Bae = (wayne = ZA(ynz) = aay =o (P08) Asoume zy = 2 and yne=y. Then eS eayrtyar=y » In general, a binary relation R on a set A is any subset of A A, ie. any set of ordered pairs (u,») such that w€A & v@A. For example, the relation of fatherhood on the set of human beings is the set of all ordered pairs (z,y) such that 2 and y are people and = is the father of y. In accordance with tradition, one often writes 2Ry instead of (x,y) € R. A binary relation R on a set A satisfying the analogues of (PO 2) - (PO 8), (2) (Ry & yRz) > 2Rz, (3) (Ry & yR2) > x=, is called a partial order on A. A partial order on a set A is said to be reflexive if and only if 2Rz holds for all 2 in A, while F is said to be irreflezive on A if and only if zRz is false for all in A. For example, the ordinary relation = on the set of integers is reflexive while the relation < on the set of integers is irreflexive. In Theorem 3.7 we have seen that the binary relation = on a Boolean algebra @ is a reflexive partial order. If 2 2K 85 BOOLEAN EXPRESSIONS AND FUNCTIONS. NORMAL FORMS By a Boolean expression we mean any expression built up from the variables z, y, 2, 21, Yu, 21, 22, Ya, 22, ... by applying the operations »,v,’ a finite number of times. In other words, all variables are Boolean expressions, and if + and ¢ are Boolean expressions, so are (rae), (eve) and (*/)t. Example 314. The following are Boolean expressions: (VD) AG YAY DAW UA VAN) VIA) A), MOY) A (209) We shall use the same conventions for omitting parentheses as were used for statement forms in Chapter 1 (ef. page 5). For this purpose, the symbols a, v, are to correspond to &, v, 1 Example 315, Using the conventions for omitting parentheses, we can write the Boolean expressions of Example 3.14 as follows: (eva) A2n, (WAZ)Y (AW, WA EVE) WARE AM WA (EAD) Given a Boolean algebra B=(B,.,v,’,0,1) and a Boolean expression +(11,...,) having its variables among 1, ..., tu, we can determine a corresponding Boolean function +%(u, ...,%): for each k-tuple (b1, ..., bs) of elements of B, +%(bi, ...,bs) is the element of B obtained by assigning the values b;,..., bs to tn, ..., ux respectively, and interpreting the symbols a, v,’ to mean the corresponding operations in @. (In order to make the corre- sponding function unique, we always shall list the variables us, ...,w in the order in which they occur in the list 2, y, Z, 21, Yi, £1, 2s, Ye %,-... For example, yv 2’ determines the function f(x,y) =yv 2’; thus f(1,0)=0 and f(0,1)=1) Example 316. The Boolean expression vy’ determines the following function f(z, y) with respect to the two-element Boolean algebra Bp. 0, 70,0) 10,0) = 1, (0,1) 4 fay =1 Notice that, if b:,...,ds are in (0,1} and +(t,...,t%) is a Boolean expression, then +5(b1, ...,bs) is also in {0,1}, since (0,1) is closed under », v and ’. Observe also that different Boolean expressions may determine the same Boolean func- tion. For example, 2 (yv2) and (ty) v (zz) always determine the same Boolean functions. such that ris ‘tMore precisely, ¢ is a Boolean expression if and only if there is a finite sequence (eyAre) oF x is «, and, if 12%, then either r, is a variable or there exist j,k Now we shall present a normal form theorem for Boolean algebras which is a gen- eralization of the disjunctive normal theorem for propositional logic (Theorem 1.6). The following notation will be convenient: if i=1 For any expression 7, Y expression if i=0 The symbol , with appropriate indices, will be used to indicate repeated use of v. In particular, = ofa) stands for o(0) v o(1), while SS ofa,,q,) stands for o(0,0) v o(0,1) v ape a i aso of=0 Theorem 3.11. (Disjunctive Normal Form) For any Boolean expression +(u,, ...,t), the equation Bo Ztlepar eve) amt aapen so ng mo azo is derivable from the axioms for Boolean algebra (and therefore the corre- sponding equation, with replaced by , holds in any Boolean algebra 8). CC Proof. See Problem 8.8, page 67.> Example 817. When k= 1, Theorem 3.11 reads tu) = Bela) aus = (r(0) Aw’) v (oC) Au) When k= 2, we obtain Huta) = 3, B, eoned auth aap = (r(0,0)A uy A wg) v (r(0, 1) A my A te) v (r(L, 0) A my Amp) V (r(2, 1) A my A 2) Exsmple 3.18. When + is 2 y, Theorem 8.11 states EV y = [OVO Az ay] v [Av 0) azayl] v [Ov DAZAy]y [Av Nazay) = [Onz’ay]y [taza y [Laz ag] y [Azay] = [envy fray] y feng) Exsmple 3.19. ‘The representation of (xv y) 4 (x’ vy’) in disjunctive normal form is fev Ate Vy) = Ons avdv Az ayy azay)y Onzary) = @ Aw (za)62 BOOLEAN ALGEBRAS (CHAP. 3 Corollary 3.12 Let +(tu,...,u:) and o(t,...,us) be Boolean expressions, and let B be some Boolean algebra. If the Boolean functions :(u,...,1u) and o7(us,...,ua) are equal, then: (a) the equation +(u:,...,t%) = o(t,...,us) is provable from the axioms for Boolean algebras; (0) + = oC for all Boolean algebras C. Proof. (a) Since (0,1) CB, +9 (bs, ...,bs) = 08 (01,...,d:) whenever bi, ...,bx € (0,1). But +?(bs,...,d4) = 0(b1,-..,0s) holds if and only if the corresponding equation x(01, ..-,0x) = o(bs, ...sbs) can be proved from the axioms for Boolean algebras. For, the equations 0.0 =0, 0,1=1,0=0, 1,1=1, etc,, are all derivable from these axioms, and the values +%(b,, ..., bs) and (bs, ..., bs) are computable from these equations. Hence by Theorem 8.11, r(tts,...,te) = o(us,...,t) is derivable from the axioms. (2) is an immediate consequence of (2). > The remarkable thing about Corollary 3.12 is that, if an equation holds for one Boolean algebra (in particular, if it holds for the two-element Boolean algebra 2), then it holds for all Boolean algebras. To mathematicians it probably would not have been surprising if we had only asserted that, if an equation holds for all Boolean algebras, then it is provable from the axioms. This latter assertion follows, in fact, from the completeness theorem for first-order logic (see Corollary 2.15(a), page 68 of [185]). 36 ISOMORPHISMS A function @ is called an isomorphism from a Boolean algebra B = (B, 4g, Vg,’3, 0g, 1g) into a Boolean algebra C = (C, Acs Ves 'C Og» Ie) if and only if (@) © is a one-one function from B into C, (6) for any z, y in B, WG Ag Y) = (2) r¢ H(y) O(tvq y) = Hz) ve By) (22) = (9(z))'c Such a function © is called an isomorphism from onto ( if, in addition, ® is a function from B onto C. Theorem 3.13. Let 6 be an isomorphism from a Boolean algebra B into (respectively, onto) a Boolean algebra ( (with the notation given above). Then (@) (0g) = 0¢ and (1g) = 1c. (®) It is not necessary to assume that (Zv— ¥) = (2) ve Oy) forallz,yinB Alternatively, we could omit the assumption that HG rgy) = (2) rc Hy) (©) If @ is an isomorphism from C into (respectively, onto) a Boolean algebra D = (D,Am;Vps'?,0p,1), then the composite mappingt 04 is an isomorphism from B into (respectively, onto) D. ¥The composite mapping (or composition) €°% is the function defined on the domain B of # such that (@°¢)(2) = e(e(z)) for each z in B. The inverse @-! is the function whose domain is the range @[8] of @ there, 918] = {efe): 26 B)) and auch that, for any y in #15), (@°!(y) isthe unigue = in soch that y= y.CHAP. 3] BOOLEAN ALGEBRAS 63 (d) The inverse mapping #1 is an isomorphism from the subalgebra of C determined by ©[B) onto 8, and, in particular, if © is onto (, then ©"! is an isomorphism from ( onto B. Proof. (a) (0g) = (tg 22) = 9(z) rc (2's) = (2) ng (0a = 0 lq) = 20,7) = (@0a)ye = 0F (b) P(vq y) = O((2's rng y'8)'8) = (8(2'8 rng ¥'®))'C = (28) re B(y'2 JC = ((@(z)VC re (W)C JE = F(z) ve HY) (©) First, @0@ is one-one. (If zy, then (z)~o(y) and therefore @(8(z)) 0(6(y)),) Second, (806)(z"2) = 0(9(2's)) = o((0(2\yC) = (0(0(2)))"> = ((@ea)(z))' Lastly, (Gc4)(trgy) = O(O(t rg v)) = O(8(z) »¢ O(y)) = 0(8(2)) av O(8(¥)) = (0°4)(2) av (O° S)(y) @ Assume z, w€[B]. Then z= (2) and w= o(y) for some and y in B. Hence “W2) and y=4"\(w), First, if zw, then zy (for, if =y, then = (2) = %(y) =w). Thus ®-! is one-one. Second, #(zv~ y) = (2) vc Oy) =z ve w. Hence Seve w) = tvey = (2) ve ow). Third, o(2’2) = (@(2)’c = zc. Hence ezc) = x's = (@-"2)'a.p We say that B is isomorphic with ( if and only if there is an isomorphism from @ onto C. From Theorem 8.18(d,c) it follows that, if B is isomorphic with (, then C is isomorphic with 8, and if, in addition, C is isomorphic with D, then B is isomorphic with D. Isomorphic Boolean algebras have, in a certain sense, the same Boolean structure. More precisely, this means that any property (formulated in the language of Boolean algebras) holding for one Boolean algebra also holds for any isomorphic Boolean algebra.t Example 320. Consider the two-element subalgebra C= (0g,1g) of any Boolean algebra S. Let D be the Boolean algebra whose elements are the integers 0 and 1 and whose operations are: (aq) ordinary multipli tion, Le, OAD 0= OAD 1L=1AQ0=0, In (va): addition modulo 2, i. 0V_ 0 =0, 0Vp1=1vp_0=1, lvpl= (0): the function 1—z, ie. O'D = and 1’ =0, ‘Then the function # on {0g,1g} such that #(0g)=0 and (1g) =1 is an isomorphism of C onto D. 3.7 BOOLEAN ALGEBRAS AND PROPOSITIONAL LOGIC A statement form A and a Boolean expression + are said to correspond if r arises from A by replacing 1, &,v by ’,a,v (respectively) and by replacing the statement letters A,B,C, Ai, Bs, Cr, ... bY 2,4, 2,21, Ys, 21, -.. (Fespectively), tFor a rigorous formulation of this assertion, see page 90 of [135].64 BOOLEAN ALGEBRAS [cHaP. 3 Example 321. AV (B& C) corresponds to £v (y Az’). 1(14.&(B,vA)) corresponds to (2 0 (y1v 2))'. ‘The statement form corresponding to a Boolean expression r will be denoted SF(:). ‘Theorem 3.14. The equation += holds for all Boolean algebras if and only if SF(z) is logically equivalent to SF(o). Hence we have a decision procedure to deter- mine whether += holds for all Boolean algebras. Proof. By Corollary 3.12, +=0 holds for all Boolean algebras if and only if :=0 holds for the two-element Boolean algebra C = ({F,T), &e.m, ven, #7, F,T), where the operations &¢.7), Vir.r)» 0-7 have the obvious meanings given by the usual truth tables, ‘These are given in detail in Problem 2.16, page 44. It is clear that r=o holds for C if and only if SF(+) and SF(c) always take the same truth values, (For, an assignment of truth values, T or F, to the statement letters in SF(z) and SF(c) corresponds to substitution of the same truth values for the corresponding variables in + and ¢.) > Example 322. Consider the equation =A (yvz) = (eny)v (zz), The corresponding statement forms are A & (BVOC) and (A&B) v (A&C). To check that these statement forms are logically equivalent, we substitute T and F for A,B, C in all possible ways and verify that the outcomes are the same. For example, it A is F, B is T, and C is F, then A & (BvC) and (A&B) v (A&C) both are F. The com- putation we make to determine this is essentially the same as the one we make to see that = A (yV2) = F = (eAg)v (cAz) when z is F, y is T, and zis F. (Namely, FA(TVF) = FAT =F and (PAT) v (PAP) = PvP =F) Solved Problems 8.1. In a Boolean algebra, let x ~ y be defined as zy’. Prove: (@) evy = zv(Qy~2) (0) e~(e~y) = way (c) A non-empty subset A determines a subalgebra if and only if A is closed under ~ and’. @ # =1~2 (@) tye a~y=0 (ie ry & tay =0) () <0 6 2=0 (9) tay=0 @ emyaa (A) ta (y~2) = (tay) ~ (erz) (® Does zv (y~z) = (zvy)~ (zvz) hold? Solution: @) ev ~2) = ev WANE) = eve) Aleve) = evNAL = avy () 2~(e~y) = BAleWwy = EA ANY = EAevy") = BAC) = (eae) Vv (ery) = OV (ey) = AYCHAP. 3] BOOLEAN ALGEBRAS 65 (©) If A is a subalgebra and if and y are in A, then s~y = say’ EA, Convereely, if A is closed under ~ and ’, and if 2 and y are in A, then z~yEA, and therefore zay = =~(e~y)EA. Since A is closed under » and’, A is a subalgebra. (@ t~2 = tae! (2) This is Theorem 8.5(xii) () e~2 = eax =0. Now use part (¢) with y=0. @) e~y= eo anyae © say’ = 0 (Theorem 3.5(xii)) zay=0 (hy & A ~2) = AAs) = eAyAz. On the other hand, (zAW) ~ (eas) = (AW A(EAZ = AWA (eve) vACealeve)) = yAlerz) = cayad @) No 1v(i~1)=1v0 However, (1v1)~(1v1) =1~1=0. 8.2. In our axiom system for Boolean algebras, prove that Axiom (9), 07 1, is equivalent (in the presence of the other axioms) to the assertion that the Boolean algebra contains more than one element. Solution: Clearly, if 01, then there is more than one clement. Conversely, assume 0=1. Then for any z, 2 =2A1=2A0=0. (Notice that in proving results about Boolean algebras we have not used Axiom (9).) ‘Thus every element is equal to 0, and the Boolean algebra contains just one element. 33. Let n be an integer greater than 1. Let B be the set of positivéintegers which are divisors of n. If z and y are in B, define 2’ =n/z, zay = the greatest common divisor (ged) of 2 and y, zvy = least common multiple (lem) of z and y. (This is a generalization of Example 3.5.) Show that (B, a, v,’,1,7) is a Boolean algebra if and only if n is square-free ( n is not divisible by any square greater than 1). Solution: Remember that the zero Og and unit 1g of the algebra are the integers 1 and 1 respectively. Axioms (1)-(6) and (9) represent simple properties of integers and of greatest common divisors and least common multiples (ef., for example, [129]). However, Axioms (7) and (8) hold if and only if, for all z in B, = and n/z have no factors in common (other than 1), and this condition is equivalent to m being square-free. (Example: if m=60, which is not square-free, 6’=10 and 6v6’= em (6,10) = 80 # 60 =19, 606’ = ged (6,10) =2 #1 = 0g.) SUBALGEBRAS 34. In the Boolean algebra of all divisors of 70 (see Example 3.4), find all subalgebras. Solution: We mast find all subsets A of (1,2,5,7, 10,14,85, 70) closed under © and ’. Remember that = Ay = gtd(e,y) and 2 = 70/2. Ay = (1,70) = (0g, 1g) A; = {1, 2,35, 70} (1, 5, 14, 70} {2, 7,10, 70} (1,2, 6, 7, 10, 14, 35, 70)35. BOOLEAN ALGEBRAS (CHAP. 3 (@) Given a subset D of a Boolean algebra @, show that the intersection of all sub- algebras of B containing D as a subset is itself a subalgebra of B (called the sub- algebra generated by D). () What is the subalgebra generated by the empty set 0? (c) If D={b}, what is the subalgebra generated by D? Solution: (@) Let C be the intersection of all subalgebras containing D. Clearly, if z and y are in C, then Ay and 2’ are in all subalgebras containing D and hence also are in C. subalgebra containing Q as a subset and is contained in all other subalgebras. 1g) is the subalgebra generated by 9. (©) {0g, 1g, 6, b’} is a subalgebra containing {b} and is contained in every subalgebra contain- ing {b}. Therefore (0g, lg, b, 8’) in the aubalgebra generated by (8). BOOLEAN EXPRESSIONS AND FUNCTIONS. NORMAL FORMS 36, a7. If D is a subset of a Boolean algebra , show that the subalgebra ( generated by D consists of the set C of all values obtained by substituting elements of D for the variables in all Boolean functions. ‘Solution: Every sudll value, being obtained from el must belong 0 every sub- algebra containing D. On the other hand, es clearly forms » subalgebra containing D, Hence C is the intersection of all subalgebras containing D. Prove Theorem 8.10: Given a Boolean expression +(u), which may contain other variables w,...,% 8 well as u. Then the equation au) = [x(0) aw’) v [r(1) au] is derivable from the axioms for Boolean algebras. Soluti We shall use induetion on the number m of occurrences of A,v,’ ins. If m either u or 1, (for some i). If + isu, then (0) = 0 and r(l) =1. “Thus au) =u = Aw] y [Lau] = [r(0) au] v [r(t) an] If ris wm then +(0)=r(1)=% Hence ta) = me = AWW YM) = OyAW) v (u AM) = (10) AW] v [rf1) Aw] Now let m> 0 and assume that the result is true for all expressions with fewer than m occur- ences of A,V,‘. Case 1. +(x) = [o(u)]’. Now, by inductive hypothesis, atu) = [o(0) Aw] v fot) a] {elu))’ = ({o(0) nw] v fot) nul)’ [ot0) Aw} A fo(1) au}! = [o(0)' vw} 0 oC)’ vw] [r(0) vu} A [r(0) vw] {H(0) 0 (0) v [H(0) Axe] v [r(d) ax] v [ual] [r€0) 0 r(0)] v fr(0) 0 w'] v [rC1) an] [(+(0) 4 (1) A (ev w)] v [r0) A w'] v [r(1) A we] = [HlO) a efd) a] v [r(0) a (1) Aw] v [r(0) 0 w'] v [oC 0 a] {lo{O) » H(A) 0.0) v (C1) Ay] v [Ur(0) A #(2) 0.0) v (10) A] = [eC Au] v [r(0) Aw} then + is Hence ru) ‘tThe intersection of a collection of sets is the set af all objerts belonging to every set in the collection.CHAP. 3) BOOLEAN ALGEBRAS 67 3.9. Case 2 r(u) = o(u) v plu). ‘Then the inductive hypothesis holds for ¢ and p. Hence rl) = oft) v ott) = [CoO Aw) v (oft) Aud] v [(o(0) 0 #) v (O(1) Au) (0(0) Au’) v (9(0) Aw’)! v [Co(1) 4 u) v (A) A u)] (e(0)v p(0)) aw] v {le(2) v o(4)) 4 x) (0) Aw] v [o(1) AW) Case 3. r(u) = o(u) p(w). This is similar to Case 2 and is left to the reader. Prove Theorem 8.11: For any Boolean expression r(u:, 1a 1 (tly 2%) = D> 2 2, [regs egy + + yy) AMEE A UREA ==> A Uf is derivable from the axioms for Boolean algebras (and therefore the corresponding equation for :® holds for any Boolean algebra B). Solution: We shall use induction on k, The case &=1 is an immediate consequence of Theorem 3.10. Now assume that the result holds for k and we shall prove it for an expression r(0ty, Way .--»y-+1)- By Theorem 3.10, HOitas oon Meer) = L(y thay rte er) Ami] Y [FCLy gs «oo the ead Aa) But, by inductive hypothesis, : 1 31 By [Oren cvanend uf os nae] BS, [rsa es cranasd Augen ors a upes] (0, tay «25M 4) and Otay eerie) = Hence 0 ty May s+ Mie) {(Bev Bip ten cumannatinss ante) af ere ate {3 oa 3 bie, sansa wpen ces natal) «ai Ctr eta) Ah A UE A oo a) B, [bey snared amy nage a o> nate) pansy) AWB A ui A oo A uShsH) Show that in any Boolean algebra @ there are 2" different Boolean functions of n variables. Solution: By the Disjunctive Normal Form Theorem (Theorem 3,11), the equation 1 2, fe pubic a a] r(uy, is derivable, Hence the function determined by + depends only on the 2* values r(ay, ...,a,), where each a; is either 0 or 1. Each such value is 0 or 1. Hence there are 2* different Boolean functions.68 3.10. Su. 3.12, BOOLEAN ALGEBRAS (CHAP. 3 If D is a finite subset of a Boolean algebra B, show that the subalgebra C generated by D is also finite. Solution: Let D have m elements. By Problem 3.6, the elements of C are the values obtained by substi- tuting elements of D for the variables in all Boolean functions +®. Clearly, we may confine our attention to functions +? of at most x variables, since variables for which the same element of D is substituted may be identified. By Problem 3.9, there are 2" such functions. Hence since each of the n variables may be replaced by any of the elements of D, we obtain at most 2*"-n* possible elements in C. If + and ¢ are Boolean expressions such that t(C1,.--10n) = 1 > afer, for all elements ¢1, . . ., ca of some Boolean algebra (, then r B Ar 2 Lo, Boon eee e > B83 ~— > Fig. 4-1 Fig. 4-2 Example 42, In the switching cireuit of Fig, 4-3 current can flow between the points z and y if and only if A&B is true. The two switches are said to be in series. This case may be generalized to the case of any finite number 2 > Rey of switches connected in series. The condition for cur- rent flow through the circuit of Fig. 4-4 is A,&Ag& Ay&... Ay, Fig. 43 Fig. 44 Example 43, In the switching eireuit of Fig. 4-5 below, current can flow if and only if (A&C) v (Av B) is true, u2 SWITCHING CIRCUITS AND LOGIC CIRCUITS [cnar.4 ———n_—__ rn» By SS Fig. 45 Example 4.3 shows that we may combine switches in parallel and in series in the same circuit. Such a circuit is called a series-parallel switching circuit. More precisely, if A is any sentence, then o——A\—o is a series-parallel switching circuit, and if S, Si,...,Sn are series-parallel switching circuits, we may form a new series-parallel switching circuit by replacing any switch in S by either kk 4a Clearly, a condition for flow of current through a series-parallel switching circuit can be written down by means of conjunctions and disjunctions, starting from the expressions representing the closure of the individual switches. In Example 43, this condition was (A&C) v (TAvB). 42 SIMPLIFICATION OF CIRCUITS ‘The condition for flow of current through the cir- cuit of Example 4.3 is (A &C)v (1AvB). The latter c statement form is logically equivalent to the statement form ((4&C)v 1A)vB, which in turn is logically ,__| > equivalent to Cv 1AvB. Hence the circuit of Fig. 45 may be replaced by the circuit of Fig. 4-6. The circuit of Fig. 4-6 is clearly a simplification ® of that of Fig. 4-5, since it involves fewer switches. Fig. 46 Example 44. A condition for current fiow through the circuit of Fig. 4-7 is (A &B& 10) v (3C& 1A). However, this is logically equivalent to 1C & ((A & B) v 1A), which in turn is logically equivalent to 1C & (Bv 7A). Hence an equivalent, but simpler, circuit is that of Fig. 4-8. (The two circuits are equivatent in the sense that one allows passage of current if and only if the other does.) RI = 10 I Fig. 7CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 73 a “Yes” vote, Let us construct a switching circuit which will pass current when and only when a majority votes “Ye Let A stand for “member 1 approves”, B for “member 2 approves”, and C for “member 3 approves Then a necessary and sufficient condition for a majority vote is (A&B)v (ARO) (BEC) A corresponding circuit is shown in Fig, 4.9. However, the given statement form is logically equivalent to (A&(BVC)v (B&C), having the simpler circuit of Fig. 4-10, 2 ? 3 > —}—_ pr» —_ vn _1L. ¢ B e LANs Fig. 4-9 Fig. 4-10 Example 46. A light in a room is to be controlled independently by three wall switches (not to be confused with switches of a circuit), located at the three entrances of the room. This means that flicking any one of the wall switches changes the state of the light (on to of, and of to on). Let us design a circuit which allows current to fiow to the light under the required conditions. ‘Let A stand for “wall switch 1 is up”, B for “wall switch 2 is up”, and C for “wall switch 3 fs up”. In the truth table of Fig. 4-11, we wish to construct a statement form f(A, B,C) for the required switch- ing circuit. 1A, B,O) a (2) « @) (CG) m © @) Fig. G11 Fig. 6-12 ae ‘The requirement on f(A, B,C) is that its truth value should change whenever the truth value of one of A, B,C changes. We arbitrarily assign the value T to f(A, B,C) when A, B, C are all T (the frst row); thus the light will be on when all wall switches are up. Then we proceed down the truth table, changing the truth value of /(A, B,C) whenever the truth value of precisely one of A, B, C changes. We have indi- cated such a procedure in Fig, 4-12 by writing to the left of each row a number showing at what step the truth value for that row has been determined. Another way of deseribing the assignment of truth values is to note that T is assigned when an odd number of statement letters have the value T. We find the resulting statement form by the method developed in the proof of Theorem 1.8; this amounts to forming the disjunction of the truth assignments in the rows to which a T is attached: (AGBEO V (TAB ABEO V (TALBE NO) v (AL BEC) logically equivalent to [A&(BEO)v (AB& ACY] v [TAK(ABEO) v (BEC) 413 below. This having the cireuit shown inSWITCHING CIRCUITS AND LoGIc CIRCUITS (CHAP. 4 4 B Cc > oe 2 2 > > > Fig. 4-13 43 BRIDGE CIRCUITS Sometimes a series-parallel circuit can be replaced by an equivalent circuit which is not a series-parallel circuit. Example 47. A series-parallel circuit corresponding to the condition [A & (Bv B)] v [C&(1Bv Ev D)] is given in Fig. 414, This in equivalent to the circuit shown in Fig. 4-15. Clearly, the only paths through this circuit are A&B, AGE&D, A&E& 1B, CRE&B,C&D, C& 1B. Hence, a condition for flow through this cireuit is (A&B) v (A&E &D) v (A&E & 1B) v (C&E&B) v (C&D) v (C& 1B), which is logically equivalent to [A &(Bv B)) v [C&(1Bv Ev D)). oF _ Cy 3 e Ww Fig. Fig. 4-15, The circuit of Fig. 4-15 is an example of a circuit which is not a series-parallel circuit, Such circuits are called bridge circuits. In Example 4.7, the bridge circuit had fewer switches (6) than the corresponding series-parallel circuit (7). Another example of a bridge circuit is given in Fig. 4-16. A corresponding statement form is (A & [Dv (C&E)]) v [B& (Ev (C&D))], whose series-parallel circuit is shown in Fig. 4-17. a2 “+ — c sf P Fig. 416 D a c & E 2 }~ e DCHAP. 4} SWITCHING CIRCUITS AND LOGIC CIRCUITS ic) Notice that any bridge circuit determines a truth function. A statement form for this truth function is obtained by finding all possible paths through the circuit. For example, the bridge circuit displayed in Fig. 4-18 corresponds to the statement form (A&B&CED) v (A&BENC) v (1BECED) Notice that the path 1B > C>71C is impossible, since it contains a formula and its negation. B & Ww 44 LOGIC CIRCUITS ‘The processing of information is one of the most important roles of the modern digital computer. For this purpose, special devices are available. An and-gate operates on two or more inputs A:,...,A, and produces their conjunction A:&Ar&... & Aq. An and-gate is denoted (6). AL Ar Ss SW rete te An More precisely, each input A; has the form of a physical quantity (say, voltage level), of which we choose to distinguish two states, denoted 0 and 1. The state 1 occurs if A: is true, and the state 0 if A; is false. The output of the and-gate is likewise in two possible states, 0 and 1: it is 1 if and only if A: &Az&...&A, is true, and it is 0 if and only if A, & Aa& ... & A, is false. Often the state of an input or output is taken to be 1 if it is transmitting current and 0 if not. Arithmetically, the output of an and-gate is the product of the inputs. Another common element of a logic circuit is an or-gate (V). If the inputs are Ay, ..., An (m=2), then the output is Aiv Azv «+: v An AL a SO bet te La An Fig. 4-20 ‘Thus the output is 1 if and only if the output of at least one A, is 1. Arithmetically, the output is the maximum of the inputs.6 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 ‘An inverter 2) is a device which has one input A and produces as its output 1A. Thus the output is 1 if the input is 0 and the output is 0 if the input is 1. A logic circuit is defined as a circuit constructed from various inputs by means of and-gates, or-gates, inverters, and possibly also other devices for performing truth- functional operations. The actual electronic (or mechanical) devices used to construct and-gates, or-gates, and inverters vary with the state of technology. For this reason, it is most convenient to ignore (as far as possible) questions of hardware (diodes, transistors, vacuum tubes, etc.). This also holds for our treatment of switching circuits. Readers interested in the physical realization of switching and logic circuits can consult [53] and [18]. Example 48. To construct a logic circuit producing the output A,+> Ay, notice that A, <> As is logically equivalent to (Ay & Ag) v (1A; & 1Ay) (Fig. 4-22) as well as to (4,&Az)v (Aiv Ay) (Pig. 4-23). Clearly the second logie circuit is simpler. a Ai&Ay NS areas \1A; & TAy 4 pare As / Fig. (22 A ABA yA) Aide © A Fig. 23 Example 49. Construct a logic eireuit producing (A, & Ay) v TAs (42849) ® Fig. 4-24CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 7 Notice that, instead of a logic circuit, one could construct a series-parallel switching cireuit through which current flows if and only if (1) is true (see Fig. 4-25). Fig. 4-25 Example 4.9 indicates that the same effect can be obtained by logic circuits as by series- parallel switching circuits. Indeed, connection in series corresponds to an and-gate, while connection in parallel corresponds to an or-gate. 45 THE BINARY NUMBER SYSTEM We are accustomed to using the decimal number system. Thus 84,062 stands for the number 2+ 6-10 + 0-10? + 4-10°+3-10'. In general, any positive integer can be rep- resented in one and only one way in the form Oy + +10 + oer 10® + +++ = ay 10 where 0a =9 for 00. This number is denoted aeds-1--axtido in standard decimal notation. However, for any integer r > 1, every positive integer n can be represented uniquely in the form 5 Qo + ayer + ager? + ee — ane where 00. This can be proved by induction on x. In particular, every positive integer can be represented in binary notation: Gy + G12 + 3122 a +++ + Ome 2™ where 0=a,=1 for 0=i=m and an=1. Example 4.10. ‘The number 23 (in decimal notation) has the binary representation 10111, ie. + 2%+2+1. The decimal number 101 has the binary representation 1100101, ie. 2°+ 25+ 22+ 1. A procedure for finding the binary representation of a number m is to find the highest power 2" which is =n, subtract 2" from n, then find the highest power 2! which is =n—2", ete. Further examples: Decimal Notation Binary Notation Decimal Notation Binary Notation 1 1 n 1011 2 10 16 10000 3 aw 35 100011 4 100 52 110100 5 101 ut 110101 6 10 7 m1 8 10002 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 46 MULTIPLE OUTPUT LOGIC CIRCUITS Occurrence of the same logic circuit as part of other logic circuits suggests the use of logic circuits with more than one output. Example 4.11. ? 2aQ=O-—> Fig. 4.26 Example 412. ‘Two numbers in binary notation are added in the same way as numbers in decimal notation. Binary notation: 100101 Decimal notation: 37 oii 23 311100 00 If we just consider the addition of one digit numbers, 0 and 1, we have the following values for the sum digit s and the carry digit c. rjafofa o}1 fifo 1fofifo olololo Thus s corresponds to the exclusive-or (which we shall denote A +B), while c corresponds to the conjunetion. If we wished to construct a separate logic circuit for s we would obtain A> O-@—O-— Ba s=A4+B Fig. 4-27 Similarly we can construct a .ogic circuit for ¢:CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 9 However, we can combine these two circuits into a single multiple output circuit: A Fig. 4.29 ‘The circuit in Fig. 4-29 is called a half-adder. If we wish to add two single-digit numbers A and Bt, while taking into account a carry-over C from a previous addition, we obtain the table Thus ¢ corresponds to the statement form (C& (A+B) v (A+B) & 10) which is logically equivalent to (A+B) +C. The carry-over ¢ corresponds to the statement om (A&B) v (C&(A+B)) ‘We can use the circuit constructed for A +B in Fig. 4-27 to obtain the following diagram corresponding to the above statement. A © @©-—@-—® 2 = (A BBV (C&(A+B)) S=(AtB4C Fig. 4-30 The circuit of Fig. 4-80 is called a full adder. tActually, A is the proposition that the first number is 1, and B is the proposition that the second num- ber is 1.80 SWITCHING CIRCUITS AND LOGIC CIRCUITS [CHAP 4 We can construct a circuit for adding two three-digit binary numbers A:4,Ay and —l_|—: I++ B:B:B, Let ——»|FA represent a full-adder, and let HA i—e —-| jee represent a half-adder. Then the sum is represented in Fig. 4-81 by cs:8150. 4,————| % HA 2,———>| fo ———————__—_______ +, A Fra 8,————_—_——| _ — ~s Ay FA —_——_——_+« By Fig. 6-31 4.7 MINIMIZATION The cost of constructing and running a switching circuit or logic circuit depends upon the state of technology and therefore varies with time. However, at a given time, some circuits will be less expensive than other equivalent circuits. Example 413. ‘The circuit (Fig. 4-32) corresponding to 1A v (B&A) is more expensive than the equivalent circuit (Pig. 4-93) corresponding to 1A vB, since the latter contains fewer occurrences of statement letters and fewer connectives. In general, decreasing the number of connectives (i.e. gates and inverters in a logic cireuit) lowers the cost, other things being equal, and decreasing the number of occurrences of statement letters also lowers the cost, other things being equal. These two criteria usually are not the only measures of cost; the special hardware used for constructing circuits imposes other criteria, —o a SS ~ ? Fig. 4-32 Fig. 4-33 The minimization problem consists of determining methods for finding a simplest (i.e. cheapest) circuit equivalent to a given circuit (or finding all simplest circuits equivalent toa given circuit). Since all that matters about the given circuit is the truth function that it determines, the minimization problem amounts to. finding one or all simplest circuits defining a given truth function. In Example 4.18, the circuit of Fig. 4-83 is clearly the simplest circuit corresponding to the truth function represented by 1A v B. Of course, for any given truth function, one can find a circuit representing the truth function and then check the cost of the finite number of all simpler or equally simple equivalent circuits. This method will yield all simplest equivalent circuits, but, for three or more variables, the application of this method often will be so involved and long that it becomes practically unfeasible. Therefore what we are seeking is a fast, convenient and practical way of finding one or all simplest circuits for a given truth function.CHAP. 4) SWITCHING CIRCUITS AND LoGic CIRCUITS 81 Let us assume that we are given a truth function. There are several forms of the minimization problem. (1) Find the “simplest” disjunctive normal form (dnf) representing the given truth function. (11) Find the “simplest” series-parallel switching circuit (or logic circuit) representing the given truth function. (III) Find the “simplest” switching circuit (either series-parallel or bridge) representing the given truth function. Example 414. Consider the truth function given by the following table. A dnf for this function is (A&B&C)v (A& 1B&C)v(A&BE IC), which is logically equivalent to (A&B) v (A&C). It is easy to check that this is the simplest dnf for g, thus solving Problem (I). The corresponding series-parallel circuit is given in Fig, 4-34. However, a logically equivalent statement form is A &(BvC), which has the simpler series-parallel cireuit shown in Fig. 4-35. It is obvious that this circuit cannot be replaced by a simpler one. Thus, a minimal dnf solving Problem (1) need not be a solution of Problem (II), that is, » solution of Problem (II) need not be a dnf. 2 2 — | |. | ne ™, 14———-, ©—C- ) —_—- a Fig. 4-36 — =——@ 1c —__+82 SWITCHING CIRCUITS AND LOGIC CIRCUITS {CHAP. 4 Notice that we have not pictured inverters. This common convention stems from the fact that the presence or absence of a negation sign often results from an arbitrary decision as to which of two contradictory assertions is to be labeled by a letter, say A, rather than by 1A. Since the number of negation signs often depends upon arbitrary decisions, it is advisable to consider negations of letters as initial inputs, on a par with letters, and not to count the number of inverters in computing the cost of the circuit. (2) In solving minimization problem (II) for the case of logic circuits, we must consider arbitrary statement forms (not just dnf’s). In this case, inverters are counted in computing the cost, since negations may be applied not only to letters but also to arbitrary statement forms. However, if we are only interested in switching circuits, consideration is restricted to statement forms in which negation is applied only to statement letters. Notational Convention. In writing statement forms, it is often convenient to omit the conjunction sign &, and to write A instead of 1A. Example 4.15. ABC y ABC instead of (A& 1B&C) v(1A&B& 10) ABy ABCD v AC instead of (3A& 1B) v (1A&B& ICED) v (1A &O) (AvByAvBv 6) instend of (1AVB)& (Av 1BY 10) In Example 4.15 and in the sequel, we adopt the convention of omitting the parentheses around the disjuncts of a disjunctive normal form. Thus, we have written ABC v ABC instead of (ABC) v (ABC), and AB v ABCD v AC instead of (AB) v (ABCD) v (AC). This alternative notation saves time and space, and is customary in work on circuits. 48 DON’T CARE CONDITIONS In many problems involving design of circuits there are certain conditions which are impossible or for which no requirement is made concerning the operation of the circuit. Such conditions are called don’t care conditions. Example 4.16. ‘The awitching circuit of Fig. 4-37 has a corresponding statement form A(BvC)v ABC. In the spe- cial case where A is “z is an even integer”, B is “z is a perfect square”, and C is “x is an integer divisible by 4”, the three conditions ABC, ABC, ABC are impossible. Hence there is no danger if we build a circuit which happens to allow current to flow if some of these impossible conditions occur. In particular, a circuit corresponding to the statement form A(Bv C) v ABC v ABC v ABC will accomplish the same task as the original circuit, But this statement form turns out to be logically equivalent to A v B, which has the much simpler circuit of Fig. 4-38. (To derive the logical equivalence, notice that ABC v ABC is logically equivalent to AB, while A(BVC)v ABC in logically equi A(Bv Cv BC) and therefore to A. We are left with Av AB, which is logically equivalent to Av B.) ¢ Loy yn — Fig. 37 Fig, 4-38CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 88 Example 4.16 shows that addition of don’t care conditions sometimes allows simplifica- tion of circuits. Later (Section 4.17), we shall learn a technique enabling us to choose those don’t care conditions which lead to maximal simplification of the circuit. Example 4.17. ‘The decimal digits 0 to 9 can be represented in binary notation as follows: Decintae neaaeel ee) 1 2 3a 4 5 6 7 8 9 Binary Note goat | oot0 | oot ae | o1o1 | o110 | o111 | 1000 | 1001 Consider the sentences: A The first (right-most) binary digit is 1 B The second binary digit is 1 C The third binary digit is 1 D_ The fourth binary digit is 1 Then ABCD corresponds to 0 ABCD corresponds to 1 ABCD corresponds to 2 ABCD corresponds to 9 In terms of inputs A, B, C, D, let us construct a switching circuit which passes current if and only if the number represented is 6, 7, or 8, i.e. under the condition ABCD v ABCD v ABCD. If the inputs A, B,C, D are such that they always represent a number between 0 and 9, then we can ignore the six possibilities 1010, 1011, ..., 1111 (i.e. the binary rep- resentations of 10 through 15). Hence the don’t care conditions are ABCD, ABCD, ABCD, ABCD, ABCD, ABCD. In particular, we can use ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD ‘Thus we are using four of the six don’t care conditions.) This statement form is logically equivalent to BC v AD. (This is left as an exercise. It can be done laboriously by a truth table, or much more easily using well-known logical equivalences from Chapter 1.) The cireuit for BC v AD is given in Fig. 4-39. B ce B ce BD a D a B ¢ D Fig. 439 Fig. 440 If we had not made use of the don’t care conditions, our original statement form ABCD v ABCD v ABCD could have been reduced to BCD v ABCD, with the costlier circuit shown in Fig. 4-40.84 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 5 4.9, MINIMAL DISJUNCTIVE NORMAL FORMS Given any dnf #, say ABC v ABD v ABCD v ABCD. Let le = the total number literals (i.e. letters or negations of letters) in, and de = the total number of disjuncts of In the example above, le=14 and de =4. For dnf’s @ and ¥, we say that is simpler than ¥ if and only if le The Quine-MeCluskey Method for Finding All Prime Implicants of a Non-Tautologous Full Dnf o: Let be ¥v -*+ voy (1) List yy -- ee (2) If two fundamental conjunctions ¢ and x in the list are the same except that $ con- tains a certain letter unnegated while x contains the same letter negated, add to the list the fundamental conjunction obtained by eliminating from ¢ the letter in which ¢ differs from x. Place check marks next to ¢ and y.CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 87 (3) Repeat the process indicated in (2) until it can no longer be applied. Fundamental conjunctions which already have been checked can be used again in applications of (2). The unchecked fundamental conjunctions in the resulting list are the prime implicants of ®. (This assertion will be justified after consideration of a few examples.) Example 423. Let # be ABC v ABC v ABC v ABC v ABC. Start with ABC ABC ABC Abc ABC Application of (2) yields ABC ABC Apc ¥ Abc ¥ Abe vy Notice that ABC and ABC yield BC, ABC and ABC yield AC: ABC and ABC yield AB. Now (2) is no longer applicable, Hence the prime implicants are ABC, BC, AC, AB. Example 424. Let © be ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCBy ABCD (2) List {asco ABCD Lasoo { abcD | ABCD | ABod ABeD 4B6b Notice that the disjuncts are listed in groups: first, those with no negations, then those with one negation, ete. Since process (2) is applicable only to a pair of fundamental conjunctions which differ by one in the number of negations, in seeking to apply (2) we need only compare fundamental con- junctions with those in the next group. (2) Application of process (2) yields ABCD {a ieee y ao BoD ¥ Ac ABCD y mee ABCD y ad ABCD y BéD y BCD v ABD88 SWITCHING CIRCUITS AND LOGIC CIRCUITS [CHAP. 4 Further application of (2) produces ABCD ¥ ee y AD fae y acon ABCD ¥ acp ¥ Abed ¥ Be ABCD y¥ ABD y¥ ABcD y¥ BeD {ae v BCD BCD y aBb Process (2) is no longer applicable. Hence the prime implicants are ABC, BCD, BCD, ABD, AD. Justification of the Quine-McCluskey Method: All fundamental conjunctions in the list logically imply ©. (It is only necessary to observe that application of process (2) to funda- mental conjunctions which logically imply @ yields a fundamental conjunction which also logically implies ¢.) Every prime implicant y of @ will appear unchecked. (For, by Theorem 44, all completions of y will be disjuncts in . Hence y will eventually appear in the list after suitable applications of process (2) to these completions eliminate all the letters not in y. itself will never be checked, since if it were it would not be a prime implicant of 4.) On the other hand, no fundamental disjunction ¢ which is not a prime implicant will remain unchecked. (For, » must include a fundamental conjunction $ which is a prime implicant of & By Theorem 4.4, all completions of ¢ originally appear in the list. By suitable applications of process (2) to these completions, we obtain a fundamental disjunction x differing from ¢ in precisely one letter. Then application of process (2) to y and x imposes a check on ¢.) Limitation of the Quine-McCluskey Method: One must start with a full dnf. If we are given a dnf which is not full, we must expand it into a full dnf. This tedious and long process can be avoided by another procedure which we shall study later. 4.12 PRIME IMPLICANT TABLES Once we have obtained all prime implicants of a given statement form ©, we must find out which disjunctions of prime implicants are minimal dnf’s. Theorem 45. Let @ be a non-tautologous full dnf, and let ¥ be a dnf. If ¥ is a minimal dnf for , then each disjunct of @ includes a disjunct of ¥. Proof. Assume not. Let ¢ be a disjunct of @ which does not include any disjunct of ¥. Hence each disjunct of ¥ differs in at least one literal from ¢. But then the assignment of truth values making ¢ T makes v F, and therefore makes @ F. But ¢ is a disjunct of ¢; so @ must be T, which is a contradiction. > Our overall strategy can now be made clear. We choose a disjunction ¥ of prime impli- cants so that every disjunct of the full dnf ¢ includes a disjunct of ¥. (Clearly, ¥ is logically equivalent to @. Since ¥ is a disjunction of prime implicants of ©, ¥ logically implies &. On the other hand, since each disjunct of ® includes a disjunct of ¥, © logically implies ¥.) Among all such ¥’s we find the minimal ones. We shall indicate techniques for narrowing this choice to a relatively small number of ¥'s.CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 89 Construct a matrix, one row for each prime implicant of # and one column for each disjunct of ©. Place a cross (Xx) at the intersection of a row corresponding to a prime implicant ¢ and a column corresponding to a disjunct y of ® such that y includes ¢. This matrix is called the prime implicant table for . Example 4.25. Let @ be ABC v ABC v ABC v ABC. Using the Quine-McCluskey method, we obtain ABC y AB y ac hess y 4B ABC y Be ‘Thus the prime implicants are AC, AB, BC, and the prime implicant table is shown in Fig. 4-41. ABC ABC ABC ABC ac x x AB x x BC x x Fig. 4-41 We shall now describe various operations performed on prime implicant tables in order to obtain minimal dnf’s. Core Operation. Assume there is a disjunct ¢ of @ such that the column under ¥ con- tains a single cross. Let ¢ be the prime implicant corresponding to this cross. belongs to what we shall call the core of 6. By Theorem 4.5, ¢ must be a disjunct of every minimal anf for @ We eliminate the row corresponding to ¢ as well as all columns containing a cross in the row corresponding to ¢. (Since ¢ must be a disjunct of every minimal dnf, the condition of Theorem 4.5 is met for the disjunets heading any such column.) Example 425 (continued). In Fig, 4-41, the columns under ABC and ABC have a single cross each. Hence AC and BC belong to the core, But all the columns contain a cross in the rows corresponding to AC and BC. Hence AC v BC is the unique minimal dnf for +. Example 426. Let © be . ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD Find the prime implicants: ABCD ABD vy BD ABCD ¥ cD ABCD ¥ BCD ¥ ABop ¥ ABC ABCD ¥ BCD ¥ 4Bcb y Asc ABCD ¥ ABD ¥ Aged y Acb ince the prime implicants are BD, ACD, ABC, ABC, ACD.90 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 The prime implicant table is ABCD ABCD ABCD _Abcp__Apcp AscD__ABcp ABCD ap | x x x x Aco x ® dpe 1 ® ABC x © aco ® We draw circles around crosses which are the only ones in a given column. In this case there are such crosses in the columns under ABCD, ABCD, ABCD, ABCD. Hence ABC, ACD, ABC and ACD are in the core. We draw a square around each cross in any row in which there is an encircled cross. In Example 4.26 we then have a square or circle in every column. Thus every disjunct of # includes a prime implicant in the core. Hence ABC v ACD v ABC AGD is the unique minimal dat for ®, ‘The results of Examples 4.25 and 4.26 are exceptional. Sometimes a single application of the core operation is not sufficient. Wider coverage is afforded by adding the following two operations. Dominant Column Operation. If a column f has a cross in every row in which a column a has a cross, then we can eliminate column g. (To satisfy the condition of Theorem 4.5, we have to use a prime implicant included in the fundamental conjunction heading column . Then by assumption the same prime implicant is included in the fundamental conjunc- tion heading column 2.) Dominated Row Operation. If the row corresponding to a prime implicant y, has a cross in every column in which the row corresponding to a prime implicant y, has a cross, and if the number of literals of y, is smaller than that of y,, then we eliminate the row corresponding to ;,. (For, if a minimal dnf had y, as a disjunct, replacing y, by y, would lower the cost, contradicting the assumption that y, is minimal.) Example 4.27. Let # be ABCD v ABCD v ABCD v ABCB v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD First we obtain the prime implicants: ABCD ¥ {aco y {a abco ¥ az y BC aBcD ¥ Bop y ie pcp y abc y ad Apcb y ABD ¥ ABCD BcB y ABCD ¥ ach y 4Bép ¥ 4zc y ABCD y a ased v ABB y ABCD y | ABD y ABC y ie vy ed {acoCHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 91 Thus the prime implicants are BCD, ACD, AC, BC, AB, AB. The prime implicant table is ABCD ABCD ABCD ABCD ABCD ABCD ABCD ABCD ABCD ABCD ABCD Bed x x ACB x x ac x x x x BC x x x x as x 1 ® a AB x x The columns under ABCD and ABCD have unique crosses, which we circle, Thus AB and AB belong to the core. Put a square around each cross in the rows belonging to AB and AB. Hence all the columns containing a circle or square can be eliminated. ‘The new table ABCD ABCD ABCD BCD x ACB x Ac x x BC x x ‘The columns under ABCD and ABCD have crosses in the same row. Hence by a dominant column opera- tion we may eliminate either column, say, the one under ABCD. This leaves us with the new table: ABCD a) BCD x @ ACB x @ 4c x @) BC x Fig, 4-42 ‘None of our operations are applicable to this table. However, notice that in order that each column include some disjunet of the required minimal dnf, we must have ((1) or (2) and ((3) or (4))f, (Q)v 2) & ((3)v (4). This is logieally equivalent to (2) &(3)) v (1) &(4)) v (2) & (3)) v (2) & (4) ‘Thus there are four different ways of choosing the rest of the prime implicants, and we obt four anf’s: AB AB BCD v AC ABy AB v BCD v BC ABy ABv ACD v AC ABy ABy ACB v BC jes for minimal dnf’s, Since they all have the same cost, all four are mini- ‘These are the only possibi mal di We shali call the method we have used for handling the table of Fig. 4-42 the Boolean method. Example 4.28. Let # be ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v AR¢D v ABCD v ABCD v ABCD a disjunct in the required minimal dnf, by (2) we mean that ‘By (1), we mean that row (1) appears row (2) appe disjunct, ete.92 SWITCHING CIRCUITS AND LOGIC CIRCUITS We obtain the prime implicants: (CHAP. 4 ABCD ¥ {4 vy BD ABCD ¥ BCD ¥ ac {isco wey Bed y ey abép ¥ nep vy ABcD y pa ABép v ABD vy ABCD y ACB y {Sten y Abe y ABCD y 4cb BCD Wass Thus the prime implicants are ABC, ACD, BCD, ABD, BD, AC. The prime implicant table is ABcbD ABCD ABCD ABCD ABCD ABCD ABCD ABCD AbcD ABCD ABC x AcB BOD ABD BD Ac © x & © © x x @ Fa Circle the crosses which are unique in their columns. Hence BD and AC are in the core. Place squares around all crosses in the rows of BD and AC. We then eliminate the rows of BD and AC, and all columns containing squares or circles, ‘The new table is ABcb__Abcb ABCD a asc} x @ Acbl x x @) BCD x @ 48d x x Fig. 4-48 Now we can apply the Boolean method used in Example 4.27. We obtain [(1) v (2)] & [(2) v (4)] & [(3) v (4)], which is equivalent to (0) & (2) & (2) v [C0) & (2) & (A) v FCA) & CA] v [C1) & (8) & (4)] v (2) & B)] v (2) & (A) v [(2) & (9) & (4)] But, taking into account the fact that, when (2) is used, (1) is not required, and that, when (4) is used, (3) is not required, we obtain (&B)v (&M)v (H&M) ‘Thus there are three possibilities for minimal dnf's: BDv ACw ACD y BCD BDv ACv ACD v ABD BDv AG v ABC v ABD Sinee the costs are the same, all three are (the only) minimal daf's for #.CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 98 Instead of the Boolean method, we may use the so-called branching method for handling the table of Fig, 4-43. We take a column with a minimal number of crosses. In our example, there are two crosses in each column; so we may choose any column, say the one under ABCD. To ensure that ABCD includes ‘a prime implicant of the sought-for dnf’s, we may use either ABC or ACD. Hence we obtain two tables (in general, if there are n crosses in the column, we would obtain n tables) as follows: in the left-hand table (Fig. 4-44) we assume that ABC is taken as a disjunct and we eliminate the row containing ABC as well as every column containing 2 erots in that row. For the righthand table (Pig. 4-8), we do the stme with ACD. ABcD__ ABCD In the left-hand table, we can again apply the Boolean method or branching, but in this simple case it is obvious that only the choice of ABD will yield minimal cost. In the right-hand table, we can choose either BCD or ABD, Thus we have the following possibilities: From the left-hand table, BD v AC v ABC v ABD. From the right-hand table, BD v AC v ACD v BCD and BD v AC v ACB v ABD. identical with the result of the Boolean method. 4.18 MINIMIZING WITH DON’T CARE CONDITIONS Let us assume that we must find a minimal dnf for a statement form 4, assuming that the additional fundamental conjunctions y,,...,y, are don’t care conditions. Then we can adapt the method used in the preceding section in the following manner. Find all prime implicants of @vy,v --- vy, However, in constructing the prime implicant table, use columns only for the disjuncts of #, not for y,,...,¥, (since we are concerned only that each disjunct of # include some prime implicant of the required minimal dnf'si. Example 429. Let & be oo - ABCD v ABCD v ABCD v ABCD v ABCD Let the don't care conditions be: ABCD, ABCD, ABCD, ABCD, ABCD. By the standard procedure (left as an exercise for the reader), we find that the prime implicants of # v ABCD / ABCD v ABCD v ABCD v ABCD are: BD, AC, ABC, ACD, BCD, ABD. ‘The prime implicant table is asco abd _Ascp Ascb__ Abc BD x x Ac x x ABC x AcB x x BCD x ABD x Now AC has crosses wherever BCD has, and AC has fewer literals than BCD. Hence by a dominant row operation we eliminate the row of BCD, obtaining94 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 asco ABCD Ascp ABCB_ ABcD ABc x x AcD x x Abb x Now the column of ABCD has a unique cross. Hence we place AC into what we call the secondary core. (AC must be a disjunct of every minimal dnf.) Then by the core operation we drop the row of AC, ‘together with the columns under ABCD and ABCD: Asch _ ABcD Now we ean apply either the Boolean method or the branching method. However, it is clear that the first and third rows yield the only minimal dnf. Hence the unique minimal dnf is AC v BD v ACD. 4.14 THE CONSENSUS METHOD FOR FINDING PRIME IMPLICANTS Given two fundamental conjunctions ¢, and v,. If there is precisely one letter p which occurs negated in one of x, and ¢, and unnegated in the other, then the fundamental con- junction obtained from c,c, by deleting p and j and omitting repetitions of any other literals is called the consensus of c, and y». Example 4.30. (i) The consensus of ABC and ABD is ACD. (ii) The consensus of AB and ABCD is BCD. (iii) There is no consensus of ABC and ABD. (iv) The consensus of A and AB is B. (*) The consensus of A and AB is B. Theorem 4.6. The consensus ¢ of y, and y, logically implies y, v ¥_. Proof. Consider any truth assignment making ¢ true. Let p be the letter occurring negated in y, and unnegated in y,. If pis T, then y, is T. If pis F, then y, is T. In either case, Y,V ¥y 18 T.) Corollary 4.7. If ¢ is the consensus of y, and y, then ¥, vy, is logically equivalent to WV vev de Consider the following two operations transforming a dnf into a logically equivalent dnf. (i) Eliminate any disjunct which includes another. (ii) Add as a disjunct the consensus of two disjuncts, if that consensus neither is identical with nor includes some disjunct of the given dnf. Given a dnf ©. (If we are given an arbitrary statement form, first transform it into a logically equivalent dnf.) The consensus method consists of applying operations (i) and (ii) until these operations are no longer applicable. The result turns out to be the disjunction of alll prime implicants of 4.CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 95 Example 431. Let be ABC v ABCD v AB v ABC v ABCD. By (i), ABC v ABv ABC v ABCD (ABCD includes AB). By (ii), ABC v AB v ABC v ABCD v AC (Consensus of ABC and AB). By (i), AB v ABCv ABCD v AC (ABC includes AC). By (ii), AB v ABC v ABCD v AG v BED (Consensus of AB and ABCD). By (i), ABv ABCy AG v BCD (ABCD includes BCD). By (ii), ABy ABC v AC v BCD v BG (Consensus of ABC and AC). By (i), ABv AC v BCD v BC (ABC includes BC). By (ii), AB v AC v BCD v BC v CD (Consensus of BCD and BC). By (), ABV AC v BG v ED (BCD includes CD). ‘Thus, the prime implicants are AB, AC, BC, CD. Example 432, Let @ be AB v ABCD v ABC y BD. By (ii), AB v ABCD v ABC v BD v ACD (Consensus of AB and ABCD). By (i), ABv ABCv BDv ACD (ABCD includes ACD). By (i), AB v ABC v BD v ACD v BC (Consensus of AB and ABC). By (i), ABy BD v ACD v BC (ABC includes BC). By (ii, ABv BD v ACD v BC v AD (Consensus of AB and BD). By (ii), ABV BD v ACD v BC v ADv ABC (Consensus of BD and ACB). By (ii), ABv BD v ACD v BC v AD v ABC v AC (Consensus of AB and ABC). By (i, ABv BDv BCv ADv AC (ABC includes AC; ACD includes AC). By (ii), AB v BD v BC v ADv AC v CD (Consensus of BD and BC). Hence the prime implicants are AB, BD, BC, AD, AC, CD. Example 433. Let # be ABCD v ABCD v ABCD v ABC v ACD. By (i), ABCD v ABCD v ABC v ACD (ABCD includes ACD). By (ii), ABCD v ABCD v ABC v ACD v ACD (Consensus of ABCD ard ABC). By (i), ABCD v ABC v ACD v AGD (ABCD includes ACD). By (ii), ABCD v ABC v ACD v ACD BCD (Consensus of ABC and ACD). Cb, BCD. ‘Hence the prime implicants are ABCD, ABC, ACD, Justification of the Consensus Method. (1) The process must come toan end. Since there are only a finite number of dnf’s using the letters of the given statement form 4, we must show that there can be no cycles in the application of (i) and (ii). Once we drop a fundamental conjunction ¢ by (i), then ¢ can never reappear by virtue of (ii). For, in all future steps, there will always be a fundamental conjunction which is included in ¢. Hence if there were a cycle, it would consist solely of applications of (ii). But (ii) increases the number of disjuncts, (2) Every prime implicant of ® occurs as a disjunct in the dnf ¥ remaining at the end of the process. Assume the contrary. Hence there must be a fundamental conjunction which has the maximum number of literals among all fundamental disjunctions + such that: (a) + includes ¢; (b) + includes no disjunct of ¥; (c) the letters of + oceur in @ Notice that ¢ is such a fundamental conjunction 7. Clearly, by (a), @ logically implies ¢.96 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 Also, @ cannot contain all the letters of # (Otherwise, by (b), 6 would logically imply the negation of each disjunct of ¥, and therefore would logically imply 1®, But only con- tradictions logically imply both © and 11@, and no fundamental conjunction is a contradic- tion.) Let A bea letter of not in 9. By the maximality of @, A@ and Ad must lack one of the properties (a)-(c). The only one they can lack is (b). Hence there are disjuncts y, and ¥, 0f ¥ such that Ad includes y, and A@ includes v,. By property (b) of 6, A must be a literal of y, and A must be a literal of y,. Since A# includes y, and A@ includes y., y, and y, do not have any other literals which are negations of each other. ‘Then the consensus » of y, and % is included in @, and therefore, by (b), includes no disjunct of ¥. Hence an application of (ii) can be made to y, and y,, contradicting the assumption that the process has ended. (8) Every disjunct ¢ of the dnf ¥ remaining at the end of the process must be a prime implicant of ©. Otherwise ¢ would include some prime implicant y. By (2), y would be a disjunct of the final dnf, and operation (i) would still be applicable. 415 FINDING MINIMAL DNF’s BY THE CONSENSUS METHOD If we have obtained all the prime implicants of a statement form @ by the consensus method, the problem remains to find the minimal dnf’s. Of course, we could construct the full dnf for @ and then apply the methods already described. However, constructing the full dnf for @ sometimes would involve a long and tedious process, and it would be conven- ient to have ways of producing minimal dnf’s without going through that process. One such method is to eliminate superfluous literals and disjuncts from the disjunction of the prime implicants, obtaining irredundant dnf’s. Then one can compare the irredundant dnf’s and pick out the minimal ones.t Example 434. We have already found (Example 4.33) that the dnf ABCD v ABCD v ABCD v ABC v ACD has as its prime implicants ABCD, ABC, ACD, ACD, BCD. Now we shall eliminate superfluous disjuncts from ABCD v ABC v ACD v ACD v BCD To determine whether a given disjunct ¢ is superfluous in a dnf @v we check whether ¥ is logically equivalent to #v¥. This holds if and only if ¢ logically implies ¥. But the latter holds if and only if the result is a tautology whenever, for each literal p in ¢, we replace » in ¥ by T and the negation of p by F. Hence we construct the following table. In the row corresponding to a fundamental conjunction ¢ we caleulate what each of the other disjuncts must be when ¢ is T. Then we check to see whether the disjunction of the results in that row is a tautology. In the table above, this holds only in the last row. Hence BCD is the only superfluous disjunct. All the other disjuncts must occur in every minimal dnf, Thus we are reduced to ABCD v ABC v ACD v ACD ‘The method we shall outline is due to M, J. Ghazala (26).CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 97 To eliminate superfluous literals, remember that a literal p is superfiuous in py v ¥ if and only if y logi- cally implies pv ¥. A quick check shows that none of the literals is superfluous. Hence we have a unique irredundant dnf, which must be the only minimal dnf. Example 4.35. Consider the statement form : Bcv BCv BDv CDv AD Since the consensus method yields no additional disjuncts, this is already the disjunction of all the prime implicants of #. For finding superfiuous disjuncts, we construct a table as in the preceding example, ¢ Bc FF D AD o: BC} F D F AD BD F é c A superfluous «cop| & F B A superfluous #4p| Bc Be Bc Since the disjunctions of the terms in the first, second and fifth rows respectively) are not tautologies, BC, BC and AD are not superfiuous and occur in all minimal dnf's for #. Let a mean that 2 occtrs as 2 disjunct in a given dnf for #. Hence from the third row, if % then es, for, if 5 does not oceur in the given dnf for ©, then both ¢» and 4 must occur. (Otherwise, when oy is T, then @ would not necessarily be T, contradicting the fact that y logically implies +.) Thus oy v (oye) is true. Similarly, from the fourth row, #4 v (ees) is true. Hence we must have (ay v apesMey V 2403) which is equivalent to os ¥ eseres05 Hence the two irredundant daf's are BCv BCvCDv AD and BCv BC v BDv AD. Since these are of equal cost, they are both minimal dnf’s, Notice that in this example we could have guessed this immediately from the third and fourth rows. Example 436. Let the prime implicants of a statement form be DE, CDE, ACD, ACE, ABD, ABE, BCD, BCE. (Observe that the consensus method is no longer applicable to the disjunction ¢ of these fundamental con- Junetions, and therefore the latter are all the prime implicants of ¢. ts 6 er o BI F coB| F PAB 4co| Fé E F superfluous AcE 5B F F superfiuous ab] FF cé P ase} DF F superfluous acp| wy & 4B A AB EB —superfiuous ace| B Fr dD A AD A dD superfluous oy are true, Prom ixth row, e6V oes Rows 3, 446,78 show that ACD, ACE, ABE, BCD, BCE are superfluous, Thus ey the third row, 3 V e204 is true. From the fourth row, «,V oe, is true. From the is true. From the seventh row,98 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 is true. This has been obtained by finding those subsets of the entries in the seventh row, the disjunction of which is a tautology and such that no proper subset of this subset has the same property. This process ‘ean be earried out by constructing a table for the entries in the seventh row similar to the one constructed above. From the eighth row, e4 Y e107 V o10ges V ei0ar6 V o17405 V is true, Hence we have ereseleaY enraleg V 2109) (0¢¥ e105) exeaeeY eters exeelenY ey0r ciara ersereY oveeesY e40) (ory eye ome This is equivalent to wieees V ovezeees (When multiplying out, we see that these are two of the disjuncts in the expansion, and, since all the other disjuncta include one of these, all other disjuncts may be dropped.) ‘Thus the irredundant dnf’s are Dev cpEv ACDv ABD and DEv CDE v ACE v ABD Notice that none of the literals is superfiuous, Since the costs of these dnf’s are equal, both are minimal dnf’s, 4.16 KARNAUGH MAPS There is a pictorial method for obtaining minimal dnf’s which is convenient for problems involving at most six statement letters.t Let us start with the case of two statement letters. In this case the Karnaugh map is based upon Fig. 4-46. A Each square represents the fundamental conjunction whose conjuncts are the literals standing at the head of . the row and column determining the square. To rep- resent a full dnf we place a check in each square cor- responding to a disjunct of ®. Fig. 446 ‘Example 4.37, AB v AB is represented in the Karnaugh map of Fig. 4-47 and AB v ABv AB in the Karnaugh map of Fig. 4-48, aa 4 = L Fig. (47 By adjacent squares we mean squares which have a side in common. Clearly, a single square represents a fundamental conjunction consisting of two literals, while two adjacent squares differ in one statement letter and therefore represent a single literal. Thus in Fig. 4-47 we have AB v AB, which is logically equivalent to B. In Fig. 449 we notice two pairs of checked adjacent Fig. 449 +See Karnaugh [42]. Another pictorial method, somewhat less graphic than Karnaugh’s, has been given by Veitch (93).CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 99 squares. We place loops around the checks in each pair. Then the corresponding dnf is logically equivalent to B v A. (B corresponds to the horizontal loop, and A to the vertical loop.) Clearly, B v A is minimal. Now let us turn to the case of three statement letters (Fig. 4-50). Each square rep- resents the conjunction of the fundamental conjunctions heading the column and row intersecting in that square. AB AB AB 4B Example 438. ABC v ABC v ABC is represented in Fig. 4-51. AB AB AB OB Example 439, ABC v ABC v ABC v ABC v ABC is represented in Fig. 4-52. AB AB OAB AB T 7 clyly y | ély ly jj | Fig. 4-52 In Fig. 4-50, by adjacent squares we mean squares which differ in precisely one literal. Thus two squares which have a side in common are adjacent. (Observe that we have used the labeling AB, AB, AB, AB so that as we move from one square to an adjoining one, only one literal changes.) In addition, in the first row the left-most square ABC is adjacent to the right-most square ABC; and in the second row, ABC is adjacent to ABC. This amounts to an identification of the left-most vertical line with the right-most vertical line. Pictorially we can imagine the left-most vertical line glued to the right-most vertical line so as to form a cylinder (Fig. 4-53). On the cylinder, adjacent squares are adjacent in the usual geometric sense.100 SWITCHING CIRCUITS AND LOGIC CIRCUITS [CHAP. 4 In Fig. 4-50, a single square represents a fundamental conjunction of three literals, while two adjacent squares differ in one literal and therefore represent a fundamental conjunction of two literals. Example 4.40. 4-54 shows ABC v ABC which is logically equivalent to AB. 4B AB 4B AB TTT Example 4.41. ABC v ABC is represented in 4-55 and is logically equivalent to BC. AB AB AB AB Fig. 4.55 Furthermore, four squares forming a square array or arranged in one row represent a single literal. Example 442. AB AB AB AB c viv e viv J Fig. 456 Fig. 4-56 exhibits ABC v ABC v ABC v ABC, which is logically equivalent to B. Example 443. AB AB OAB OAs ely]y ° Lied | | Fig. 437 In Fig. 4-57 we see ABC v ABC v ABC v ABC, which is logically equivalent to A.CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 101 Example 4.44, AB AB AB AB eé;vilyiviy Fig. 458 In Fig. 4-58 we have ABC v ABC v ABC v ABC, which is logically equivalent to C. Example 445. AB AB AB AB cly y ely y Fig. 4-69 Fig. 4-59 representa ABC v ABC v ABC v ABC, which is logically equivalent to B. Notice that if we picture Fig. 4-59 on a cylinder, the four checks form a square array. ‘The technique for minimization is straightforward. We draw loops around single checks, or pairs of adjacent checks, or groups of four checks (forming a square array or arranged along a row), in such a way that every check belongs to at least one loop. We try to make maximal use of groups of four checks or two checks so as to minimize the number of disjuncts and literals} Example 4.46, AB AB AB AB ¢ lf > elY| [- Fig. 4-60 ‘The Karnaugh map of Fig. 4-60 represents ABC v ABC v ABC v ABC. The unique minimal dng is ABy ACw ABC. AB corresponds to the vertieal loop, and AC to the horizontal loop. Example 447. 4B AB AB AB Fig. 4-61 Fig. 4-61 represents ABC v ABC v ABC v ABC v ABC, There is a unique minimal dnf: Bv AC. B corresponds to the four checks ABC, ABC, ABC, ABC, while AC corresponds to the horizontal loop covering ABC and ABC.102 SWITCHING CIRCUITS AND LOGIC CIRCUITS [cHaP. 4 Example 4.48, ‘The unique minimal dnf is Av C. Example 449. 4g [> Fig. 468 In this ease, ABC ean be combined with either ABC or ABC. Hence we have two minimal dnf's: ACy Adv BE Let us consider now Karnaugh maps for four statement letters (Fig. 4-64). 4B Ab Ab AB cD cB cD eD Fig. 464 Again, adjacent squares are those which differ in exactly one literal. In particular, ABCD and ABCD are adjacent, as are ABCD and ABCD. This amounts to identifying the left- most and right-most vertical lines, and identifying the lowest and highest horizontal lines. Pictorially we can imagine that we have glued together the left-most and right-most vertical lines, and the lowest and highest horizontal lines, to form a doughnut-shaped surface (called a torus). On this doughnut, adjacent squares are adjacent in the usual geometric sense. ‘A single square represents a fundamental conjunction of four literals. A pair of adjacent squares represents a fundamental conjunction of three literals. Four squares, in a square array or along a single row or along a single column, represent a fundamental conjunction of two literals. Finally, eight squares arranged in two adjacent columns or in two adjacent rows represent a single literal.CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS Example 450. ‘The Karnaugh map in Fig. 4.65 represents AB. Example 451. Fig. 4-66 represents AD. Example 4.52. Fig. 4-67 is the Karnaugh map for BD. Example 458. Fig. 4-68 represents 8. cD cD cB ep cD oD oD cD cb cD cD cD cD ep 4B ab AB AB y y y y Fig. 4.65 AB AB AB 4B y t viy Pig. 4-66 AB ab AB AB ia eel een) y y 467 AB ab AB AB vyiy viv 108104 SWITCHING CIRCUITS AND LOGIC CIRCUITS [cHAP. 4 Example 454. AB AB AB AB cD [ ¥Jy¥ivi¢v cD cb elyi|yi|viv Fig. 4-69 Fig. 4-69 represents D. Minimization techniques for four statement letters are similar to those for three, Example 455. 4B AB AB AB ay 7 | CB y al ep ar, Fig. 4-70 The unique minimal dnf is AD v BD. Observe that the four squares in the column under AB are not joined by a loop, since the corresponding fundamental conjunction AB would be superfluous. Example 456. AB AB AB AB cD ¥ cD y, cb éb Fig. 71 The check in ABCD cannot be combined with any other. Hence ABCD must be in any minimal dnt, ‘The check in ABCD can be combined only with the check in ABCD. Hence BCD mast be a disjunct of any minimal dnf. Similarly, the check in ABCD can be combined only with the check in ABCD. Hence ‘ACD is a disjunct of any minimal dnf. Now the checks in ABCD and ABCD already have been covered. ‘Thus the unique minimal dnf is ABCD v BCD v ACD Examples 4.55-4.56 illustrate the method to be used. For each checked square, deter- mine whether there is a unique largest combination of checked squares containing it. If 80, put a loop around that combination. To avoid superfluous disjuncts, first handle eachCHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 105 checked square whose unique largest combination consists only of itself; then, among the remaining uncovered checks, handle those whose unique largest combination consists of two checks; among the still uncovered checks, handle those whose unique largest combination consists of four checks, etc. For any remaining checked square, determine all the possible largest combinations containing them, and, among the corresponding dnf’s,* find the minimal ones. Example 457. ° (ALTA o Key oi eb Fig. 412 Considering ABCD, we see that ABC must be a disjunct (covering ABCD and ABCD). Looking at ABCD, we note that ABC must be a disjunct (covering ABCD and ABCD). None of the other three checks belongs to a unigue largest combination. The only uncovered check is ABCD, which can be combined either with ABCD or with ABCD. Hence we obtain two minimal dnt’ ABCyv ABCV ACB and ABC ABC BCD In the case of five statement letters, we can use a three-dimensional Karnaugh map (Fig. 4-73). AB Ab__AB_ AB ©D, cD, cD, cD, cof cB, of fpf oe Y Fig. 4-73 Some choices among the remaining combinations may render superfluous some of the disjuncts already obtained (ef. Problem 4.22).106 SWITCHING CIRCUITS AND LOGIC CIRCUITS [cHap.4 The usual definition of adjacent squares implies that corresponding squares in the two planes (e.g. ABCDE and ABCDE) are adjacent. Combinations of sixteen squares are pos- sible and yield a fundamental conjunction consisting of a single literal. Combinations of eight squares yield fundamental conjunetions of two literals, ete. For six statement letters one could use four planes, but in that case, and even more so for larger numbers of statement letters, the geometrie picture often is too complex to permit easy construction of minimal dnf’s. 4.17 KARNAUGH MAPS WITH DON'T CARE CONDITIONS If we are given a full dnf , together with various don't care conditions, we construct a Karnaugh map by placing checks in the squares corresponding to the disjuncts of ® and crosses in the squares corresponding to the don’t care conditions. In constructing minimal dnf’s, we are free to use any of the crosses which allow us to form larger combinations of squares. Example 4.58. Let # be . . a . ABCD v ABCD v ABCD v ABCD v ABCD and assume that ABCD, ABCD, ABCD, ABCD, ABCD are don’t care conditions, The-Karnaugh map is shown in Fig. 4-74. AB AB AB Ai oo} v |(v\ fal ob y WJ Fig. 4-74 First we handle the checked squares which belong to unique largest combinations (possibly including srosses). Thus ABCD belongs to @ unique largest combination: {ABCD, ABCD}. Hence ABC must be a sjunct of all minimal dnf's. Likewise, ABCD belongs to a unique largest combination (the second column), and therefore AB must be a disjunct of all minimal dnt's. The other checks do not belong to unique largest combinations. ‘The only check still not included in a loop is ABCD. For the latter, there are two possible combinations of four squares. Hence we may use either AD or BD. Thus there are two minimal dnf's: ABCv ABv AD and ABCv AB v BD. Example 459. Let @ be ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE vy ABCDE v ABCDE Lat the don’t eare conditions be ABcDE, ABCDE, ABCDE, ABCDE, ABCDE, ABCDECHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 107 AB_AB_ AB_ AB Fig. 4-75 ‘We seek the checks belonging to unique largest combinations. First, ABCDE belongs to such a com- ination (the four corners of both planes). Hence BD is a disjunct of all minimal daf’s. We get the same result from the checks for ABCDE, ABCDE, ABCDE, ABCDB. The checks _ middle aguares of both planes belong to a unique Baquare combination: (ABCDE, ABCDE, ABCDE, ABCDE, ABCDE, ABCDB, ABCDE, ABCDB), yielding the disjunct BD. ‘This belongs to two 2-square combinations. Hence we must have either BDv BD v ABCE BDv BD v AcDE 4.18 MINIMAL DNF’s OR CNF's Given a statement form ®, we can obtain the minimal dnf’s for # and we also can obtain the minimal dnf’s for 14. But the minimal dnf’s for 1@ yield minimal enf’s (conjunctive normal forms) for &. Example 4.60, Recall that a enf is a conjunction of one or more disjunctions of one or more literals. ‘The enf (Av Bv 0) & (Av B)& (BV) has as its negation atc ahoR ABév abv Bc ‘Thus by comparing the costs of the minimal dnf's and minimal enf's for +, we can obtain those state- ‘ment forms which are minimal among all dnt’s or ent’s for +. Example 461. Let © be oo _ ABCD v ABcD v ABCD v ABCD v ABCD v ABCD v ABCD If we examine the Karnaugh map for # (Fig. 4-76, below), we find that there are three minimal dnf's: ABD v ACB v AcD v BCD ABD ACD v ABD v ABC ABD y ACD v ABC v BCD108 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 AB AB AB AB cD v ‘| oD L vty | olyiy ep v Fig. 416 On the other hand, 14 has the Karnaugh map (Fig. 4-77) obtained by putting checks in the empty squares and erasing the checks already present in Fig. 4-76. AB AB AB AB o}l}yviy cl v v &D viv ely iv iy Fig. 4-17 From Fig. 4-77, we see that the minimal dnf’s for 1 are AD v ABC v ABD v ABC AD v BCD v ABD v ABC AD v BCD v AGD v ABC ADv BCD y ACD v BCD Thus the minimal enf's for @ are (Av D) & dv BVO) & (AvBv D) & (AVBVO) (Av DB) & Bv Cv D) & (AVBVD) & (AvBVO) (Av D) & (By Cv D) & (AvCvD) & (AVBVO) (Av B) & By Cv D) & (Av Cv D) & (By Cv) Since these are cheaper than the minimal dnf's for #, these are minimal among all dnf’s or enf’s for 4. ‘The procedure in the above example for finding the minimal statement forms among all dnf’s or cnf’s for @ does not provide a general method for finding minimal statement forms for @ (i.e. minimal series-parallel switching circuits, or minimal logic circuits). For example, (A&B) v (C&(DvE)) is minimal, but it is neither a dnf nor a cnf. Final Remarks: (1) We have indicated methods for finding minimal dnf’s (or minimal dnf’s or enf's). This constitutes a solution of Problem (I) on page 81, although possiblyCHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 109 not the best solution (cf. Remark (2)). No reasonably good general solutions for Problems (Il) or (ITI), page 81, are known. (2) The methods we have given for finding minimal dnf’s require us to find all prime implicants. (This is not true of Karnaugh maps, but these are useful only for statement forms involving at most five or six statement letters.) How- ever, there are certain cases in which the number of prime implicants is so large that our methods are not practical.t There is need then for a method for finding minimal dnf’s which does not use the set of all prime implicants, but no general method of this kind is available. Solved Problems SWITCHING CIRCUITS. SIMPLIFICATION 4.1. Replace the series-parallel cireuit of Fig. 4-78 by a simpler bridge circuit. Solution: Consider the bridge circuit shown in Fig. 4-79. The paths through this cireuit are A&B&C, A&D&E, AGDEC,AGBEC, ALB&E. Hence a condition for passage of current is (A&BEC)V (A&D&E)V (A&DEC)v ALBEC)v A&BEE) which is logically equivalent to [A&B &C] v [(Bv 6) & (A &D) v (A&B). But this is a con- dition for passage of current through the given circuit, ne ~ ~~] Fig. 4-79 tFridshal [24] states that, for nine statement letters, the full dnf whose fundamental conjunctions are those with 1, 3, 4, 5, 6, or 8 negated literals has 1698 prime implicants. The full dnf whose fundamental 5,6, or 7 negated literals has 765 prime implicants, and its negation has ants. conjunctions are those with 0, the same number of prime impl110 42, SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 A committee consists of the chairman, president, secretary, and treasurer. A motion passes if and only if it receives a majority vote or the vote of the chairman plus one other member. Each member presses a button to indicate approval of a motion. Design a switching cireuit controlled by the buttons which passes current if and only if a motion is approved. Solution: Let C,P,S,T stand for “The chairman approves”, “The president approv condition for ‘approval is ete. Then the [c&Pvsv nN] v Pesan ‘which has the corresponding series-parallel circuit shown in Fig. 4-80. P e +— § r P s r Fig. 480 LOGIC CIRCUITS. BINARY NUMBER SYSTEM 43. Let a non-negative integer less than 10 be given by its binary representation a3a20,00. (For example, if the integer is 3, then as =a: and a:=a.=1; while if the integer is 9, then a:=ao=1 and a=a,=0.) If A; is the statement that a is 1, construct a logic circuit corresponding to the condition that the given integer is prime. Solution: ‘The prime integers are Decimal Notation | 2 3 5 7 Binary Notation | 0010 | 0011 | o101 | o111 A corresponding statement form is Ag & [(1Ay& A, & Ag) V (14g & AY HAG) Vv (Ag TALE AY) Y (Ag Ay EAQ)] which is logically equivalent to As & ([Ao& (Ary pl! v [1A & A, & 143) A corresponding logie circuit isCHAP, 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS iu 44, 45. Justify the following algorithm for translating a number x from decimal into binary notation. Use two columns. Place z at the top of the left column (in our example, z= 43). Divide x by 2, putting the re- mainder 7 in the right hand column and the quotient qo in the left hand column below the given number, Repeat this process with qo, etc. Stop when we get a quotient 0, The resulting binary number is to be found by reading the right hand column from the bottom up (in our example, 101011). Solution: # = Wyte rm = Oord = +r ny = Oorl Gena = Manat ree Tey = Ord Gear = 21 they me = 0orl 1 = 2-041 ‘Then © = Wlqtr) +r, = dat 2r tr = de ryt int ry = Bigg + Ary + Or, +79 = 1695+ Bry + Ary + Bry + To = Se ee ee ‘Thus the binary expansion of = is Irn, aro Construct a logic circuit for adding 1 to a four-digit binary number aya2a,d0. Solutior Let A, stand for “a, is 1”. Let b,bgbgbby be the result of adding 1, and let B, stand for “bis "Then Vg and the carry Co= (Ay & VO) v (1A & Cg) = Ay + Co and the carry Cy = Ay& Cy = A, & Agi (Az & 1C,)v (1Ag& Cy ApEC, = A,B ALAS (Ay&e 1G) v (1Ay& Cy) = Ag+ Cy and the carry Cy = Ay& Cy = Ar& Az & A, & Ay = By A+; andthe carry Cz If we use@ to designate the circuit of Fig. 4-27, we obtain Ao @———> \__________-s, Ay —?] 4 ps é Ay +)——-8, Byn2 46. 47. SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 Describe a method for reducing the operation of subtraction of binary numbers to the use of addition only. Solution: ‘Assume given two numbers x and y in binary notation, Let us assume that they are at most sdigit numbers. Example: 2 = 11010 y= 1101 (n=5) Change all digits of y to their opposites; in our example, 10010. Add this new number z to 2. 11010 +_10010 101100 ‘Add 1 to the result: 101101. Omit the leading 1: 101, This is z—y. ‘What we did in obtaining from y was to form (2*—1)-y=J11...1-y. Adding this to z wales yielded 2 + [(2*—1)—y] = (2~y) + (2-1), Addition of 1 then gave (=~ y) +2", of the leading 1 finally reduced to z—y. nd omission Another example: 10100. Then = 001011. = 101111 and y= 101111 + 001011 111010 1 T1101. Answer: 11011. The purpose of reducing subtraction to other operati its implementation by logic circuits. 5 (addition, adding 1, etc.) is to facilitate ‘What does the process described above yield when y is greater than z? Assume that a number between 0 and 9 is given as a four-digit binary number. Employing the notation of Example 4.17, make use of don’t care conditions in order to construct a simple switching cireuit (or logic cireuit) for the condition that the given number is a prime. Solution: ‘The condition for being a prime is ABCD v ABCD v ABCD v ABCD ‘The don’t care conditions correspond to the numbers 10 through 18: ABCD, ABCD, ABCD, ABCD, ABCD, ABCD Of these six conditions, we select three and obtain the dnf: ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD (A method for choosing the proper don’t care conditions is presented in Section 4.17.) This state- ment form is logically equivalent to BC v AB (exercise for the reader) and hence to B(Cv A), which has the switching circuit ———_ XX“ __CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 113 MINIMAL DISJUNCTIVE NORMAL FORMS. PRIME IMPLICANTS 48. 49. 4.10. 41. ‘To test whether a fundamental conjunction y is superfluous in y v ®, show that it suffices to replace all occurrences in ® of unnegated letters of y by T and occurrences of negated letters of y by F, and then observe whether the result is a tautology. Example: To see whether AC is superfluous in AB v AC v BC, we obtain TBv TF v BT, ie, Bv B, which is a tautology. (Note that any T in a fundamental con- junction may be dropped, and any fundamental conjunction containing an F as a conjunet also may be omitted.) Solution: We must see whether y logically implies , i.e. whenever y is T, then © also must be T. When y is T, all the unnegated letters of y are T and all the negated letters are F. When this truth assignment is made in , we observe whether the result is always T, ie. whether the result is a tautology, If y is a fundamental conjunction and « is a literal, to test whether a is superfluous in ay v &, show that it suffices to replace all occurrences in av # of unnegated letters of y by T and occurrences of negated letters by F, and then to observe whether the result is a tautology. Example: Is B superfluous in ABv AB ABC? We obtain TB v By FBC, ie. Bv B, which is a tautology. Soltis We must see whether y logically implies av ©; i.e, whenever y is T, then av ¢ also must be T. But, if y is T, the unnegated letters of y are T and the negated ones are F. Then the result of the indicated substitution must always be T, i.e. must be a tautology. Prove that if one of the disjuncts of a dnf ® is not a prime implicant of #, then a literal of that disjunct is superfluous. Hence an irredundant dnf © must be a dis- Junction of prime implicants of #. Solution: Let ¢ be y-v ¥, where y is not a prime implicant of 6 This means that there is a fundamental conjunction @ which is proper part of ¥ and such that ¢ logically implies ¥v¥. Let @ be any literal of y which is not a literal of ¢, and let be obtained from y, by deleting a. Thus ¢ is included inf. Hence # logically implies ¢, and therefore ¢ logically implies yv ¥. From this we may conclude that # logically implies av ¥, Thus a is superfluous in apV¥, ie. in Find an irredundant dnf logically equivalent to ABC v ACD v ABD v ABC v BCD Solutio (1) ACD is superfiuous (since Bv B is a tautology). ‘This leaves ABC ABD v ABC v BCD (2) B is superfiuous (since BC v Bv BC is a tautology). This leaves ABC v ADv ABC v BCD (3) BCD is superfuous (since Av A is a tautology). This leaves ABC v AD v ABC which is irredundant,4 4.12, 4.13, SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 Find all prime implicants of ((Av B)~C)v ABC by the Quine-McCluskey method. Solution: First we must expand the given statement form into a full dnf: Av B)v Cy ABC ABv cv Abe Apc v ABC y ABC y ABC y ABC v ABC ‘Then ABC ¥ BC y c Apc y aa v A Waee v ABy Bey acy Abc ¥ Boy ABC y acy lis 7 ‘Thus there are two prime implicants: C, A. Find all minimal dnf’s logically equivalent to ABCDE v ABCDEy ABCDEy ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE using the Quine-MeCluskey method and prime implicant tables. Solution: First we find the prime implicants. ABCDE ¥ {4208 {4ncoe v ACDE | ABcpE y ABCE [AscDe y BcDE + ABCBE y¥ AcbE | apcbE aoe [4BepB y Ase lascbs y ABCDE y¥ (iscae y ‘The prime implicants are ABCDE, ABDE, ACDE, ABCE, ACDE, BCDE, ACDE, ABDE. The prime implicant table is ABCDE ABCDE ABCDE AbCDE ABCDE ABCDE ABCDE ABCDE ABCDE ABCDECHAP. 4) Aus. SWITCHING CIRCUITS AND LOGIC CIRCUITS 115 Hence the core consists of ABCDE, ABDE, ACDE, ACDE, ABBE, Thus we obtain the new table: ABCDE ABCDE Jt is clear from this table that a minimal dnf is obtained only if we choose ABCE (since all other ways of covering both columns would require two disjuncts, each having four literals). Hence there is @ unique minimal dnf: ABCDE v ABDE v ACDE v ACDE v ABDE v ABCE Find the minimal dnf’s for the dnf ABCDE v ABCDE v ABCDE vy ABCDE v ABCDE v ABCDE v ABCDE v ABCDE with the don’t care conditions ABCDE, ABCDE, ABCDE, ABCDE. Solution: First find the prime implicants: ‘Therefore the prime ADE. We obtain the ABCD BCDE AbcE Ascp BcDE ABCD AcbE ACDE ADE { ABCDE ABCDE ABCDE ABCDE ABCDE ABCDE AbCDE AbCDE Agcbé ABcoE AbCDE AbCDE v RASLR RRR RR ABDE ¥ ADE ACDE ¥ ABCD ACDE ¥ BCDE ABDE ¥ ABCE Ascp laeae AacD AcbE Abe implicants are ABCD, BCDE, ABCE, ABCD, BCDE, ABCD, AcDE, ACDE, e prime implicant table: ABCDE ABCDE ABCDE ABCDE ABCDE AécBE ABCDE ABCDE x x Ea © Hence ABCD and ADE belong to the core. We obtain the following new table, (Notice that row ABCE has been dropped, since it is empty.)116 SWITCHING CIRCUITS AND LOGIC CIRCUITS ABCDE ABCDE (CHAP. 4 ABCoE ABCDE ‘We may now use the Boolean minimal dnf is obtait would require more than two disjuncts) or branching. method. ‘orb z However, in this case, it is cbvious that the sd using ABCD and ABCD (since any other way of covering all the eclumi Hence there is a unique minimal dnf: ABCD v ADE v ABCD v ABCD 4.15. Find all solutions of Problem 4.7, using the Quine-McCluskey method and a prime implicant table. Solution: ‘The given dnf is ABCD v ABCD v ABCD v ABCD, The don't care conditions are: ABCD, ABCD, ABCD, ABCD, ABCD, ABCD. First we find the prime implicants. ABCD ABCD ABCD ABCD \ Agcp ABCD ABCD ABCD ABcD ABed Hence the prime implicants implicant table Be v ABC ABD ACD BCD ABD ACD ABC BeD ABC BoD Asp Acb BCD Ase RVR RRR { are AB, AC, BD, CD, ABCD ABCD @ ‘Thus BC is in the core, We then obtain the table AB AC x x y AB ¥ Ac v BD v cD y BC v v v v v v v v v BC; and we obtain the following prime ABCD aBcd ABCDCHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 17 4.16. ‘Thus we may choose either AB or AC, and there are two minimal dnf’s: BCv AB (which factors into B(Cv A)), and BCv AC. Notice that in Problem 4.7 we obtained only BCv AB. How the other answer BC v AC cannot be factored, and therefore the first answer gives us the simplest statement form (although both give equally simple dnt’s). Using the Quine-McCluskey method and prime implicant tables, find all minimal dnf’s for ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG with the don't care conditions ABCDEG, ABCDEG, ABCD£G, ABCDEG, ABCDEG, ABCDEG, ABCDEG, ABCDEG Solutio ‘The prime implicants turn out to be BODG, CDEG, CDBG, BDEG, ADEG, BCDE, ACEG, ADEG, CDEG, ACD (Verification is left as an exercise for the reader.) We then draw up the prime implicant Table I. Table T ABCDEG ABCDEG ABcDEG ABcpec _aBcpkc _ABcDEG BCDG x x eDEG x x cDBe x x BbEG x x pee x x x BODE ACKe ADEG epee aco x x x x ABCDEG ABCDEG ABCDEG ABCDEG aBODEC ABCDEG +“ BCDG CDEG CDEG BDEG ADEG BCDE x x ACEG x x ADEG x CDEC x x AcD x x118 SWITCHING CIRCUITS AND LOGIC CIRCUITS [CHaP. 4 No applications of the core operation are possible, However, we can eliminate the following col- umns by dominant column operations: ABCDEG (since it dominates ABCDEG), ABCDEG (since it dominates ABCDEG), ABCDEG (since it dominates ABCDEG), ABCDEG (since it dominates ABCDEG), ABCDEG (since it dominates ABCDEG), ABCDEG (since it dominates ABCDEG). ‘Thus, we obtain Table II. BoDG DEG cDEG BDEG ADEG BODE AcEG ABEG epee AcD In Table ADEG (both dominated by ACD). We can also drop the first row, since it is Table 11 ABCDEG ABCDEG _ ABC: II we can apply the dominated row operation to eliminate the rows of CDEG and mpty. Thus we obtain Table ITI. Table I AbCDEG ABcDEG ABCDEG ABCDEG ABCDEG _ABCDEG CDEG BDEG BeDE AcEG ADEG CpEe AcD x ® @ In Table III, the first and third columns have unique entries. Hence ACD belongs to the secondary core. We then can drop the last row and the first, third and fourth columns, obtaining Table IV.CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 119 417. Table IV AscoéG aBCDEG ABcDEG Clearly, pplication of the Boolean method to Table IV yields elght different minimal dnf's (by choosing either the first or second row, either the third or sixth row, and either the fourth or ceed ACD v cbBGY ACEEy BeDE ACD v CDBG ACEG v CDBG ACD v CDEG v ADEG v BCDE ACD v CDEGY ADEG v éDBE AGD v BDBGy ACEC y BéDE ACD v BDEGY ACEO v DEE ACD v BDEG v ADEG v BCDE ACD v BDEG v ADEG v CDEG Observe that ACD must be present because it is in the secondary core. Using the consensus method, find all minimal dnf’s for the dnf of Problem 4.11. Solution: (1) ABC v ACD v ABD vy ABC v BCD (2) ABC v ACD v ABDv ABC v BCD v ABD (Consensus of ABC and BCD) (8) ABC v ACD v ABDv ABC v BCD v ABD v ACD (Consensus of ABD and BCD) (@) ABC y ACD v ABD ABC v BCD v ABD v ACD v AD (Consensus of ACD and ACD) (©) ABCv ABC v BED AD (since ACD, ABD, ABD, ACD all include AD) Operations () and (i), page 94, are no longer applicable. Hence ABC, ABC, BCD, AD are the prime implicants, ‘Let us draw the following table: superfluous From the third row, #V exes, and we obtain eyx(oyv oe,Jou Which is equivalent to eyez. Hence the only irredundant dnf is ABC v ABE v AD ‘Therefore this is the only minimal dnf.120 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 4.18, Apply the consensus method to find all minimal dnf's for ABC v BD v ACD v ABC Solution: () ABC y BD v ACD v ABE (®) ABCv BD v ACD v ABC v ABD (Consensus of ABC and ACD) (8) ABC v BD v ACD v ABC v ABD v ABC (Consensus of BB and ACD) (© ABC y BD v ACD v ABC v ABD ABC v ACB (Consensus of BD and ABC) (6) ABC v BD v ACDv ABC v ABD ABC v ACD v AB (Consensus of BD and ABD) (6) BB v ACD v ABC v ACB v AB (ABC, ABD, ABC all include AB) () BD AéD v ABC ACB v ABv BCD (Consensus of ACD and ABO) Operations (i) and (il), page 94, are no longer applicable, Hence the prime implicants are BD, 40D, ABC, ACD, AB, BOD. ‘Now we construct the following table: “1 BD ACD ABC 2 superfluous superfluous superfivous be ey 8 > ry CT F superfluous Hence we have the result exlenv erelon oure(oa ¥ eveadesleeV 6903) whieh is equivalent to ereg9er6 V er0z03%5 ¥ ‘Therefore there are three irredundant dnt’ BD v ABCy ABy BCD BB v ABC v ABy ACD BD v ACDv ABy BCD Since they are of equal cost, all three are minimal dnf's, KARNAUGH MAPS 4.19. Using a Karnaugh map, find all minimal dnf’s for ABCD v ABCD v ABCD v ABCD v ABCD v ABCD ABCD v ABCD Solution: Draw the Karnaugh map:CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 121 420. 4.21. AB AB AB AB cD v cB ob ép vy Handling the checks whose unique largest combination consists of two squares, we obtain the four loops indicated in the diagram. Since all checks are covered, the unique minimal dnf is ABC y ACB v ABC v ACD Notice that, although each of the four checks in the middle belongs to a unique largest combination of four squares, that combination is not required, since all the checks in it already have been covered. Use a Karnaugh map to find all minimal dnf’s for the dnf of Problem 4.18: ABC v BD v ACD v ABC Solution: To use a Karnaugh map we need not expand the given daf into a full dnf. It suffices to place check in every square containing one of the disjuncts. (For example, ABC generates the two checks in ABCD and ABCD; BD generates four checks, etc.) AB AB Ab AB —— cD ld cD v |ly vNM eoiWiv iv There are no isolated checks and no checks with a unique largest 2-square combination, However, there are checks belonging to unique largest 4-square combinations. The first column gives AB, and the other d-square combination yields BD, The remaining three checks produce three minimal dnf's: AB vy BD v ACD v ABC (ABCD combines with ABCD, and ABCD with ABCD) AB vy BD v BCD v ABC (ABCD combines with ABCD, and ABCD with ABCD) ABv BB v BCD v ACB (ABCD combines with ABCD, and ABCD with ABCD) This verifies the solution of Problem 4.18. By means of a Karnaugh map, find all minimal dnf’s for ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE vy ABCDE y ABCDE122 4.22, SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 Solution: ‘The Karnaugh map AB AB AB AB ev fv fv cb, y cf” v/v G ep, viv v/¥ v/v e V v/v There is one isolated check, yielding the disjunct ABCDE, There is one check (ABCDE) belonging to a unique 2-square combination, which yields the disjunct ACDE. There are two checks (ABCDE and ABCDE) belonging toa unique S-square combination, which yields BC. ‘Another check (ABCDE) belongs to a unique &-square combination, yielding BD. Since all the checks are covered, we have obtained a unique minimal dnf: ABCDE v ACDEv BC v BD Use a Karnaugh map to find all minimal dnf's for ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE Solution: ee AB AB 4B AB cD, v/y¥ cb v/v cp y Hi ‘There are four checks belonging to unique largest 2-square combinations (ABCDE with ABCDE, ABCDE with ABCDE, ABCDE with ABCDE, and ABCDE with ABCDE), yielding the dis- juncts ABCE, ABCD, ABCD, and ABCD. ‘There is a unique largest 4-aquare combination containing ABCDE, and this yields the disjunet BDE. Now there is one check still uncovered: ABCDE. ‘There are apparently two equally simple ways of covering this check: (1) ABCDE with ABCDE, yielding the disjunct ACDE. (2) ABCDE with ABCDE, yielding the disjunct ABCE.CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 123 However, notice that in ease (2) the square combination (ABCDE, ABCDE, ABCD, ABCDE} hhas been covered by four different 2-square combinations, and the disjunct BDE becomes super- fluous. Thus there is actually only one minimal daf: ABCE v ABCD v ABCD v ABCD v ABCE 4.23. Using a Karnaugh map, find all minimal dnf’s for ABCD v ABCD. ABCD assuming ABCD, ABCD, ABCD, ABCD, ABCD, ABCD, ABCD, ABCD are don’t care conditions. Solution: ‘We use checks for the disjuncts of the given dnf, and crosses for the don’t care conditions. AB AB AB AB cD viv |x cb l|xiyv x ® Dib _L The check in ABCD is in » unique largest 4-square combination, yielding the disjunct AB. The only uncovered check is in ABCD. This belongs to two 4-square combinations, Hence there are two minimal daf's: ABV AG and AB BC. 4.24. Use a Karnaugh map to find all minimal dnf’s for ABCDE v ABCDE v ABCDE v ABCDE y ABCDE v ABCDE with the don’t care conditions ABCDE, ABCDE, ABCDE, ABCDE, ABCDE, ABCDE. cD/| ¥ cD, x IT} €p, ¥ Solution: op124 SWITCHING CIRCUITS AND LOGIC CIRCUITS [cHAP. 4 ‘There is an isolated check at ABCDE. Hence this must be a disjunct. ‘The checks at ABCDE and ABCDE belong to unique largest 2-square combinations, producing the disjuncts ABCE and ABCE. The cheek in ABCDE belongs to a unique largest 4-square combination, yielding the dis- junet BDE. The only uncovered check is ABCDE. ‘This belongs to two Z-square combinations, Yielding either ABCE or ABCD. Hence there are two minimal dnt ABODE v ABCE v ABCE v BDE v ABCE ABCDE v ABCE v ABCE v BDE vy ABcD 4.25, Find all minimal dnf’s or cnf's for ABCD v ABCD ABCD v ABCD v ABCD vy ABCD v ABCD Solution: Draw the following Karnaugh map. AB AB AB AB ro) y ob y cD y v eD vivly ea The cheeks in ABCD, ABCD and ABCD belong to unique largest 2-square combinations, yield- ‘ing the disjuncts ABC, ABC and ABC. The remaining check in ABCD belongs to three 2-square combin Hence there are three minimal dnd’s: ABE v ABC v ABC v BED ABC y ABC v ABC v ABD ABC y ABC v ABC ACD To find the minimal enf's, we draw the Karnaugh map for the negation: AB AB AB AB co ly|y y cD ¥iv v ly oe ely ABED belongs to a unique largest 4-square combination (the first column), yielding the diajunct AB. ABCD belongs to a unique largest 4-aquare combination, producing the disjunct AC, and ABCD belongs to unique largest 4-square combination, yielding BC. The only uncovered check is ABCD, which is isolated. Hence the unique minimal dnf for the negation is AB vy ACv BC v ABCD ‘Therefore the unique minimal enf is (Av By v OMB v OAV BV CD) ‘This is simpler than any of the minimal dnf's. Hence it is the minimal dnf or enf.CHAP. 4] SWITCHING CIRCUITS AND LoGIC CIRCUITS 125 426. Find all minimal dnf’s or enf’s for ACDE v ABCDE v ABCDE ., ABCD y ABCDE v ABCDE with the don’t care conditions ACDE, ABCE, BCDE, ABDE, ABCDE, ABCE, ABCDE. Solution: First we draw the Karnaugh map, AB_AB_AB_ AB The check in ABCDE is in a unique largest 2-square combination, yielding the disjunct ABCD. The check in ABCD# is in w unique largest 4-square combination, yielding the disjunct BCE. ‘The other checks are covered by two 8-square combinations, yielding BD and AC. Hence there is eae ane ABCD v BCE y BB v AC w ‘The Karnaugh map for the negation is: AB_AB_AB_ AB cD, v y op, xf xv caf yf x v“ff* efy] Vv ]V¥ ‘The check in ABCDE belongs to a unique largest 2-square combination, producing the dis- junet BCDE. The check in ABCDE belongs to a unique largest 4-square combination, yielding ABD. ‘The remaining checks are covered by two unique largest 8-square combinations, generating AC and AC. Hence there is a unique minimal dnf for the negation: BCDE v ABD AC v AC and therefore a unique minimal enf for the given daf: (Bv Cv Dv By Av By DAV CYA VC) @ This cnf and the minimal dnf (1) are of equal cost, Hence there are two minimal dnf's or enf’s: (1) and (2).126 SWITCHING CIRCUITS AND LOGIC CIRCUITS [oHap. 4 Supplementary Problems SWITCHING CIRCUITS. SIMPLIFICATION. BRIDGE CIRCUITS 4.21, Write down a statement form representing a condition for flow of current through each of the following series-parallel circuits. @ > aw — ie — 2 — Nw or 5 LJ o B 2 D ~ NS we aw D 428. Write down a statement form representing a condition for flow of current through each of the following bridge circuits. (a) 2 2 —_»_ L. ¢ e % 7? wt »tl, 429, Draw # switching cireuit having the following corresponding statement forms. (0) (BVA) & (ABV) v (C&A) @) A& BAC D) v (IAB BVO)CHAP. 4) SWITCHING CIRCUITS AND LOGIC CIRCUITS 127 430. Replace the following series-parallel swit ig cireuits by simpler bridge circuits. (@ > — SP — |. > 2 ‘Use at most five switches. (6) a B 7 > | > > |. I Use at most ten switches. © é EB a 18 B — c re? D B E c 1 Use at most seven switches. 431. Is there an equivalent bridge cireuit simpler than the series-parallel cireuit of Fig. 413 (Example 4.6) 432. A light is to be controlled from two wall switches such that flicking one of the wall switches changes the state of the light (“on” to “off”, or “off” to “on”), Construct a switching circuit that will allow current to flow to the light under the given condition. (Hint: Compare Example 4.6.) 43% A municipal board consists of the Mayor, President of the City Council, Comptroller, and three Borough Presidents, The Mayor has two votes and all the others one vate, A motion obtaining 8 majority passes except that any motion opposed by all three Borough Presidents fails. Write a switching circuit which will indicate passage of a motion,128 SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP. 4 LoGic CIRCUITS 4.4, Construct logie circuits corresponding to the following statement forms, () (A&B) v (B&(CV 14) (@®) (4>B)v IC 435, Write down statement forms corresponding to the following logic circuits. (@) a ® BINARY NUMBER SYSTEM 436. 43. 438, 439, 440, 4a. a2, ‘Write the binary notation for the following numbers given in decimal notation: 35, 74, 155, 820. ‘Write the decimal notation for the following numbers given in binary notation: 10001101, 10110, 111011, Write the ternary notation (base 3) for the numbers given in Problem 4.36. Write the decimal notation for the following numbers in ternary notation: 12011, 222110, 10110, Solve Problem 4.36 for base 6 and base 8, instead of base 2. Do the following additions in the binary system (and check by going over to the decimal system). @ mor ) ~—— 11000 +1011 + 101110 Do the following multiplications in the binary system (and check by going over to the decimal system). (11101 () 11000 x 1011 x 1010CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS 129 443, (o) Let a non-negative integer less than 10 be given in binary notation: oyax0jty, Letting A, stand for “a; is 1”, construct a logic circuit producing the statement that the given integer is a per- fect square, (2) Same as (0), except that the resulting proposition states that the given integer is even. (c) Same as (0), except that the resulting proposition states that the given integer is a perfect eube, 444, (a) Using half-adders and full adders, draw a logic circuit which carries out the addition of two four-digit binary numbers. (6) Same as (a), except that three two-digit numbers are to be added. 445. Translate the following decimal integers into binary notation using the method of Problem 4.4: 21, 59, 124, 446. Translate the decimal numbers of Problem 4.48 into ternary notation, using a method analogous to the one given for binary notation. 447. Perform the following binary subtractions directly and also by the method indicated in Problem 4.6. (@) 1100110 () 1110001 (© 101 = q1011 — 1011100 — 11010 MINIMAL DISJUNCTIVE NORMAL FORMS. PRIME IMPLICANTS 448. Show that ABcb v ABCD v ABED v ABCD v ABCD v ABcD v ABCD fs logically equivalent to BCv AD (ef. Example 4.17). Under the same assumptions as in Problem 4.7, use don't eare conditions to find a simple switehing clrcuit for the following properties: (6) the given number is odd; (b) the given number is composite (i.e. has a divisor different from 1 and itself). 450, Which of the following dnf’s are simpler than the dnf ABC v AB v BCD? (a) ABv ABC; () Av By CDv AC; (e) ABCD v ABC; (a) ABCD v AC v AB. 451. Which of the following are prime implicants of the dnf ABC v AB v BCDY (0) A, (6) AG, (e) ABD, (@) BCD. (Note: ‘There are prime implicants which do not oceur in this list) 452. Show that @ is logically equivalent to yv @ if and only if y logically implies #, 458. Determine whether the fundamental conjunction AB is superfluous in (a) ABC v AB v BC; (b) ACv BC v AB, 454. Determine whether the fundamental conjunction BC is superfluous in (a) ABCD v ABCD v ABCD v BC; (6) BCv ABD BD v ACD. ave 455. Show that.y v ¢ is logically equivalent to ayv¢ if and only if y logically impli 456, Determine whether the first occurrence of the literal C is superfluous in (@) ABC v AB Y CA; (6) ACv BC AC. 457. Determine whether the first occurrence of the literal B is superfluous in (@) ABC v BC v AC; () ABCY BCY AC.180 458, 459, 41. 483, 464, 465, 496, 4st. 40. an, an. SWITCHING CIRCUITS AND LOGIC CIRCUITS [CHAP. 4 Find irredundant dnf's logically equivalent to (a) ABCD v ABD v CD v ABD v ABC; (b) ABD v AC v ABD v BCD v BD. Prove that a full dnf containing n letters is not a tautology if and only if it has fewer than 2" disjuncts. Find full dnf's logically equivalent to (a) ABD v ABC; (8) Av BC v AC. Carry out the proof of Theorem 4.4, using Lemmas 4.2-4.3. Show that if ya logically implies # and yz logically implies #, then y logically implies +. Find all prime implicants of the following statement forms, using the Quine-McCluskey method. () ABOO&AE () ABC ABC v ABC v ABC () ABCD v ABCD v ABCD v ABCD v ABCD (® ABCDE v ABCDE v ABCDE v ABCDE v ABCD v ABCDE Do the irredundant dnf’s of Problem 4.58 contain all their prime implicants? Prove that a statement form is logically equivalent to the disjunction of all its prime implicants. Find all the prime implicants of the df in (a) Problem 4.11, (6) Problem 4.51. Construct the prime implicant tables for the dnf’s in Problem 4.636, ¢,¢. Find the minimal dnf's for the daf’s in Problem 4.636, 0,d, using the Quine-McCluskey method and prime implicant tables. Find the minimal dnt's for the following dnt's, using the Quine-MeCluskey method and prime im- plicant tables, (@) ABCDE v ABCDE v ABCDE v ABODE v ABCDE v ABCDE v ABODE v ABODE v ABCDE v ABCBE (®) ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD (©) ABCDEG v ABCDEG v ABCDEG v ABCDEGv ABCDEG v ABCDEG v aBcDBG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABCDEG v ABcDEG v ABCDEG v ABCDEG v ABCDEG v ABéDEG (® AB¢bB v ABCDE v AbcDE v ABGDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDEv ABCDE ‘Are the irredundant dnf’s of Problem 4.58a, minimal daf's? Give a full argument showing that, if # is a full dnf and, in the prime implicant table for , every column contains an entry from a row corresponding to a fundamental conjunction in the core, then the disjunction of the members of the core is the unique minimal dnf for & (ef. Examples 4254.26). Verity the assertion in Example 4.29 that members of the secondary core must be a disjunct of every minimal dnf.CHAP. 4] SWITCHING CIRCUITS AND LOGIC CIRCUITS iat 473, am, 45. 476. 4m. 478, 479, Find minimal dnf’s for the following dnt’s with don’t care conditions. Use the Quine-McCluskey method and prime implicant tables, (a) ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE, with don't care conditions ABCDE, ABcDR, ABCDE, ABCDR, ABCDE, ABCDE. (0) ABCDE v ABCDE v ABCD v ABCDE v ABCDE v ABCD: ABCDE, ABCDE, ABCDE, ABCDE, ABCDE, ABCD, aBCDB. (@) ABCDEG v ABCDEC v ABCDEG v ABCDEG v ABCDEG v ABCDRE v ABCDEE v ABCDEG v ABCDEG, with don’t care conditions ABCDEG, ABCDEG, ABCDEG, ABCDEC, ABCDEG, ABCDEG, ABCDEG, ABCDEG, ABCDEG. with don’t care conditions Show that if one fundamental conjunction y, includes another yz, then yz is logically equivalent to von By the consensus method, find all prime implicants of the following dnf’s: () ABC v ABCD v ABC y BCD; (b) ABCD v AB v BCD v ABCD v ABCD. By the consensus method, find all prime implicants of the dnf's in: (@) Problem 4.11, (b) Problem 4.18, (c) Problem 4.63b, (d) Problem 4.694, c. Check the solution to Problem 4.18 by expanding the original dnf into a full dnf and using the Quine-McCluskey method. Apply the consensus method to find all minimal dnf's for: (0) AB y BCE v ABCB v ABCDE; (6) ABC y BCB v ACD v ABB; («) AG v BC v BD BD. Check your results by using the Quine-McCluskey method. Apply the consensus method to find all minimal dnf’s for: (a) dnf in Problem 4.13; (8) dnf’s in Problem 4.694, b,¢,d; (c) dnf's in Problem 4.75a, b. KARNAUGH MAPS 4.80, 41. 492. Using Karnaugh maps, find the minimal dnf's for: (a) ABC v ABC v ABC v ABC v ABC (&) ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD (© ABCD v ABCD v ABCD v ABCD v ARcD v ABCD v ABCD (@) ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v ABCD v Ago v ABCD v ABéD v ABCD (© ABcDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABODE v ABCDE v ABCDE v ABCDE v ABCDE v ABCBE v ABCBE v ABCDE () ABy ACDE v ABD v CDE v ABD Solve Problem 18 by means of a Karnaugh map. By use of Karnaugh maps, solve Problems 4.694, 6, ,d and 4.78a, 5.182 4.83. 484. 485. 4.86. 4st. SWITCHING CIRCUITS AND LOGIC CIRCUITS (CHAP.4 Using a Karnaugh map, find all minimal dnf's for ABCDE v ABCDE v ABCDE v ABCDE vy ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE sjuncts.) (Be sure that you are not using superfiuous Draw a Karnaugh map for six statement letters, and try to use it to solve Problem 4.69¢ and Problem 4.16. Using Karnaugh maps, find all minimal dnf’s for the given statement forms, with the indicated don’t care conditions, (a) ABC v ABD v ACD v ABCD, with the don’t care conditions ABCD, ABCD, ABCD. (®) ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE v ABCDE, with the don’t care conditions ABCD, ABCDE. (@) ABC v ABCD v AC, with the don’t care conditions ABCD, ABCD, ABCD, ABCD. Find all minimal daf's or cnf's for the statement forms in (a) Problems 4.13, 4.18, 4.19, 4.21, 4.22. (®) Problems 4.636, ¢,d, 4.69, b, ¢,d, 4.78a, b,¢, 4.800-f. Find all minimal dnt’s or enf’s for the following statement forms, with the indicated don’t care conditions. (a) ABCD v ABCD v ABCD, with the don’t care conditions ABCD, ABCD, ABCD, ABCD, ABCD, ABCD, ABCD, ABCD, as in Problem 4.23. (8) Same as Problem 4.24. (© Same as Problem 4.16, (@) Same as Problem 4.14. (¢) Same as Problem 4.730, 8,¢.Chapter 5 Topics in the Theory of Boolean Algebras 5.1 LATTICES A lattice is an ordered pair (L,<) consisting of a non-empty set L together with a partial order < on L satisfying: (L4) For any 2 and y in L, the set {z,y) has both a least upper bound (lub) and a greatest lower bound (eib). ‘We have seen in Chapter 8 (Theorem 3.9) that a Boolean algebra B determines a lattice (B, =), with zvy and zay as the lub and glb respectively. Therefore it should cause no confusion if, for any lattice (L,=) and for any x and y in L, we use zvy and x,y to denote the lub and glb of {2,y), respectively. Example 5. The set (a, 6, ¢, d, ¢, f} is not a lattice with respect to the partial order pictured in Fig. 6-1. For, {a, 6} has no lub. ee oa NN Fig. 5-1 Theorem 5.1. For any elements z, y, 7 of a lattice (L, =): (0) zaz=2 and 2vz=z (Idempotence) (0) way=yaa and zvy=yv2 (Commutativity) (e) @A(yaz) = (zay)az and zv(yve) = (2vy)ve (Associativity) (@) ea(evy) = 2 and tv (cay) = 2 (Absorption) () 2 (eazeynz & evzz1>z,>---.t All the terms of this sequence are distinct (by Theorem 3.8), contradicting the fact that there are only a finite number of elements in the algebra. > Given an element z of a Boolean algebra B, we define ¥(z) to be the set of all atoms b of Bsuch that b Theorem 5.6. Every finite Boolean algebra B has 2* elements, where the positive integer nis the number of elements in the set A of atoms of B. Proof. By Theorem 6.4, B = (B,a,v,’,0,1) is atomic, and, by Lemma 5.5, ¥ is a one-one function from B into the set P(a) of all subsets of A. Now let C be any subset of A. Since is finite, so is A and therefore also C. Thus C = {b,..., bs}. Let 2 = dives by Then ¥(z)= (by,-.-0x)=C. (For, bf biv-+-vbe=a for all i, Thus Cc¥(z). On the other hand, if bE¥(2), then b=2=biv--+vby. Hence b= baw = babies vb) = (abs v “++ v (babi) Now if b were different from all the b/s, then each b~bi=0 and we would have b=0v--+v0=0 which is impossible. Thus d=, for some i, ie. ¥(z)CC.) We have proved that ¥ is a one-one correspondence between B and the whole set P(A). Since A has n elements, (A) has 2* elements and therefore B also must have 2* elements.» Something more can be said about the function ¥. Theorem 5.7. If B is an atomic Boolean algebra, then the function ¥ is an isomorphism from @ into the Boolean algebra (A). If @ is a finite Boolean algebra, then ¥ is an isomorphism from B onto P(A). Proof. Remember that A is the set of atoms of B, and ¥(z) = (b: bEA&by means y <2; 2 #y means T(z = ‘We use the usual conventions: #>y mesns y= ; 2 ty means Ve The second part of Theorem 5.7 shows that any finite Boolean algebra is isomorphic with a Boolean algebra of all subsets of a set A. This turns out not to be true for arbitrary infinite Boolean algebras, although, as we shall see later, any Boolean algebra is isomorphic with a field of sets (ie. to a subalgebra of the Boolean algebra of all subsets of a set). 33 SYMMETRIC DIFFERENCE. BOOLEAN RINGS In a Boolean algebra @ we define the operation + of symmetric difference as follows: ety = (zay')v (ey) In the Boolean algebra P(A) of all subsets of a non-empty set A, r+y=zAy (cf. Section 2.6). In the Boolean algebra of statement bundles (cf. Example 3.5), [Al+{B] = {A+B}, where the second + is the exclusive-or connective. Theorem 5.9. The operation + has the following properties. @) ety =yt+e QO) s+0=2 () ste =1 @ z+(y+z) = (ety) +e () BAY+2) = (ery) + (er2) ()e+z=0 (9) et+y=atery=e (i) lt2 =a @ e+yseryaate Q t=zertz=0 Proof. @ ety = (tay)v(ery) ute = (Yar) v (yar) = (ynz)v (yar) = (Bay) v (any) (0) B40 = (2A0)v(e'A0) = (eal)vO = zal = 2 () B42 = (tay) v (eax) = (aa)val = ave =1 (@) z+ (y+z) + ((yaz’) v (y'n2)) (aly nz) v (y n2))") v (za [(yn#) v Wrz) [ea (y'v 2) a (uve) v [(e ayaz’)v (2 ay az)] = [aay ae) v (yaz))] v [la ayae) v @ au’ n2)] = (Bay azyy (@Ayaz)v (er ryaz)y (2 AY’ nz)188 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 6 On the other hand, (r+y) +2 = 2+(2+¥). But, to calculate z +(r+y), we use the equation just found for z+ (y+z) after substituting z for x, = for y, and y for z. ‘We obtain et (ety) = (ena ayy (enzan)v (ef azay) v (eazy) (eay’ nz’) v (eayaz) v (a aynz’)y (a! ay’ az) = 2+(y+2) (2) zAtyte) = BA((Yrz)y (AB) = (@aUar’)y (BAY rz). (aay) + @az) = [way v @az)] v [Way a (@r2)] = [env a(e'vey)v [le vu) a (eazy) = [aceate’vey] v [er vu) a2)A2)] = [yawae]y [zeny ae] = (aya) (eay'az) () w+a = (tar)v (eng) = 0v0 = 0. (9) Assume z+y = x+z, Then z+(c+y) = 2+(+z2). Hence (eta)ty = (cta)+z, Ofy=O0te, yore (i) Add z to both sides of (c) and use (f). () Add x to both sides of 2 +y = 2 and use (f). (/) Taking y=0 in (i) yields s=2-+>2+2=0, If we exchange y and z in (@) and set y=0, weobtain 2+2=0>2=2.) By a ring, we mean a structure % = (R, +, X, 0), where F is a set, + and x are binary operations on R, and 0 is an element of R, such that Q) @ ty) tz = 2+ (yt2); Q@ ety=yts (3) 2+0= 35 (4) for any 2, there is a unique element (—z) such that 2 + (—2) = 0; (6) @xy) xz = ex (yXx2); (6) 2X (y+2) = (ex) + (2X2); (1) (yt2) xo = (YX2)+ (2x2). A ring & is said to be commutative if and only if, in addition, (8) exXy=yXxz, A ring & is said to have a unit element if and only if there is an element 1 in R such that (9) 2x1 =1xe=2, (Clearly, there cannot be another unit element u, for we would then have u=1%u=1) In Theorem 5.9 we have already verified that a Boolean algebra determines the com- mutative ring (B, +, a, 0) with unit element 1. ‘This enables us to apply the highly developed algebraic theory of rings to the study of Boolean algebras. But we also can give a more Precise characterization of Boolean algebras in terms of rings, in the following way.CHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS. 189 A ring ® = (R,+,%,0) is said to be a Boolean ring if and only if it satisfies the identity wes for all z. (Here we employ the usual abbreviation: w=2xe) Theorem 6.10. Let ® = (R, +, X,0) be a Boolean ring. Then 0 (@) 242 a= () sty=0 024 (@ xy=yxz (Thus the ring must be commutative.) Proof. First, we observe that for an arbitrary ring the cancellation law z+y= +2 > y=z holds. For, if c+y=2+2, then (-2) + (+) = (-2) + (2+2) (2) +2) +y = ((-a) +2) +2 O+y = 042 vse From the cancellation law it follows that waater2z=0 (a) For, if s=z+z, then 2+0=2-+2, and the cancellation law yields =0. a) ota = (etz)X (eta) = ateattsttat = atatete. By (), st+2 = 0. >) Since x+2=0, 2=(-a) by the uniqueness assumption for (—z) (ef. Axiom (4) for rings). ic) If ©+y=0, then, again using 2+2=0, we conclude that z=y by the uniqueness assumption of Axiom (4). D (ety) = (ety) x (ety) = + (eXy) + (ye) + yt = (ety) + (eXy) + (Xz). By (1), 0 = (e@xy)+(yX2). Hence by (c), 2Xy = yXz.p Theorem 5.11. Let & = (R, +, X, 0) be a Boolean ring with unit element 10. If we define walte, saysaxy, evy = et+yt(exy) then @ = (R,A,v,’,0,1) is a Boolean algebra. Proof. We must verify Axioms (1)-(9) for Boolean algebras. havysyve y vy = utyt(exy) = ytat(yX2) = yve 2. say = yx. This is just (d) of Theorem 5.10. 3 =+(yv2) = (way)v (zaz). Ba(yve) = 2X (yte+(yXz)) = (2X4) + (wX2) + (eX (YX 2) (way) v(eaz) = (eX) + (@X2) + ((X») x (@X 2) = (ex y) + (Xz) + (eX yX2) = (xX) + (eX2)+ (eX yX2)140 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 (4) #v (uaz) = (eva) a (eve). ey (yaz) = 2+ (yxe) + (exyX2) (evy) (eve) = (levy) az] v (levy) az] by (3) = [eaa)y Waal] y [lerz)v(yrz)] by (8) = [ev(uxa)] v [lex2)v (yx2)] = [tex + @x2Xxy)] v exe) + wx2) + (exyxz4) = [zt (ex) + (exy)] v [lex 2) + (yX2) + (ex X2)] = ev (exe) + (yx2) + @XxyX2)] = tt (eX2) + YXz) + (eX YXz) + (BX [eX 2) + (Xz) + (ZX YX) B+ (eX 2) + (yX2) + (2XYX2) + [x (Xz) + (@X (YX 2) + (@X(eX¥Xxz))] by GB) BH (@X2) + (yX2) + (@XYX2) + (eX 2) + (eXYX 2) + (2XYX 2) 2+ (yX2yt (@XyX2) + [lexe) + (2X 2)] + [eXUx2) + (@xYX2)] W 2+ (yx2) + (@xyX2) () v0 = 9. zv0= 2+04(2x0) = 2+0405% (Note that we have used the fact that, in any ring, xx0=0. To see this fact, observe that 2x0 = zx (0+0) = (zx0)+ (2x0); and then by (1) in the proof of Theorem 5.10, rx0=0) (6) tal =a. This is just zx 1= 2, which follows from the definition of a unit element. (1) eve = 1, aval = etal t (exe) = 2+(1+2)+ (ex (1 +2) 1+ (e+2) + ((@X1) + (exa)) = 1+0+(2+2) =14040=1 @) az =0. EAg = 2x (1t2) = (ex1)+(exa) = 242 =0 (9) 0%1. This is an assumption of the theorem. > ‘Thus we see that a Boolean ring with nonzero unit element determines a Boolean algebra, and vice versa any Boolean algebra determines a Boolean ring with nonzero unit element (essentially Theorem 5.9). 54 ALTERNATIVE AXIOMATIZATIONS There are numerous axiom systems for Boolean algebras.t The following system is ¢ variation of one due to Byrne (101). A Byrne algebra is a structure (B, »,’,0) where B is a set, » is a binary operation on B "ig a singulary operation on B, and 0 is an element of B, satisfying the postulates: +The one we have used (Axioms (1)-(9)) is due to Huntington [121]. For systems proposed up to 1982 ef. Huntington (122). For later work, ef. Sikorski [148], p. 1, footnote 1.CHAP. 5} TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (Bl) ray = yar (B2) (B83) (Ba) (B5) Let us introduce a few definitions. Definitions. 1 for 0” avy for (aay a 2=y () 2 22 (f) tay eazeynz & 2vz=yv2)142 (@) (o) () @) @ Co) (9) (h) @ a (k) oO (m) (n) (0) @) (@) r) ) @ TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 (w) (@<2&y=2) + tvy=z (2) @=2&2=y) > ee any (y) ta(z'vy) = tay (a) Ba(yv2) = (@ay)v (eaz) (2) tv (yz) = (@vy) a (zv2) (és) Axioms (1)-(9) for Boolean algebras hold. Proof. aaa = x, Hence by (B4), zaz’ = 0. This follows immediately from (B4) and the definition of <. This follows immediately from (B3) and the definition of =. Assume e 35 IDEALS i An ideal of a Boolean algebra B = (B,»,v,’,0,1) is a non-empty subset J of B such at: () @EF& yEI) > zvyes (i) (@EJ & yEB) > aayes It is clear that (ii) is equivalent to (i) (@EJ & y=2) > yer For, assume (ii) and let s €J & ya, Then since y=z, we have y=zay. Hence by (ii), y EJ. Conversely, assume (ii’) and let eJ & yEB. Since zany =z, it follows by (i) that zayeJ. Notice that 0 belongs to every ideal, since 0 <2 for all z. Example 55. (0) is.an ideal. Example 5.6. B is itself an ideal. Every ideal different from B is called a proper ideal. In particular, {0} is a proper ideal. Note: An ideal J is proper if and only if 1@J. For, if 1¢J, then JCB. Conversely, if 1 EJ and if y © B, then y=1. Hence by (i), y€B. Thus J=B.144 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 Example 5. If A is a non-empty set and if P(A) is the Boolean algebra of all subsets of A, then the set J of all finite subsets of A is an ideal. J is proper if and only if A is infinite. (More generally, if m is any infinite cardinal number, the set of all subsets of cardinality less than m is an ideal.) Example 58. Given u€B, the set J, of all v=u is an ideal. For, if 1, =u and v,=1, then %Vv,*w; and if vuand y=», then yu. The ideal J, is called the principal ideal generated by u. Theorem 5.13. Given any subset C of a Boolean algebra , the intersection W of all ideals J containing C (i.e. such that CCJ) is itself an ideal containing C. W is said to be generated by C, and is denoted Gen (C). Proof. There is at least one ideal containing C, namely B itself. Assume z and y are in W, and z is in B. If J is any ideal containing C, then c€J & yEJ. Hence zvy J. Likewise, zaz€J. Thus zvy and zz are in W, and therefore W is an ideal.» Theorem 5.14. Given any subset C of a Boolean algebra, the ideal Gen (C) consists of all elements of the form (nam)y rv Qaam), EI where a1,...,2€C and w,...,s are arbitrary elements of B. Proof. Let D be the set of elements of the given form. The join of any two elements of Dis clearly again of the same form and therefore also in D. In addition, for any y € B, the meet yr ((urnm)y =v Wena) = (ya(yiaa))v ov (ya(uente)) = (yam) ami) v oe v (yam) Az) is again in D. Thus D is an ideal. Since z= 1,2, every member z of C is in D. Every member (y: 021) v «++ v (Yea) of D belongs to any ideal J containing C, for, since 2 € C, it follows that yz, €J and therefore that (#:m:) v «+ v (eau) J. Hence D is the intersection of all ideals containing C.> Theorem 5.15. Given any subset C of a Boolean algebra ®, the ideal Gen(C) consists of the set E of all y such that ye mveve where 71, ..., 2. are arbitrary members of C. Proof. Wf y Corollary 5.16. If C is any subset of a Boolean algebra @, then the ideal Gen (C) generated by C is a proper ideal if and only if vie 1 nv for all zy, ...,2% in C. Proof. Note first that 1 CHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 145 Theorem 5.17. If J is an ideal of a Boolean algebra B, and if y€B, then the ideal Gen (JU(y)), generated by J plus y, consists of all elements of the form ayve where z €B and z €J, Proof. Let H be the set of all elements (zy) v 2, where z€ Bandz€J. First,y€H, since y= (1Ay)v0. Also, if v€J, then v=(Ony)vvEH. Thus JU{y)CH. In addition, (21a) vas) v ((e2ay)v aa) = ((e1v 22) a4) v (tives) CH Also, if x € J, then wa(ery)ve) = ((waz)ay)v (was) EH Hence H is an ideal. Finally, if J is an ideal containing JU{y), and if 2 €J, then (¢ny) v= €/; thus 12H.» Corollary 5.18. If J is an ideal of a Boolean algebra @, and if y@B~J, then the ideal Gen (J U (y}) generated by J plus y is a proper ideal if and only if zvy #1 for all z in J, ie. if and only if, for all z in J, y #2. Proof. By Corollary 5.16, Gen(JU{y}) is proper if and only if aiv-+-vaxvy #1 for all x, ...,2% in J. (Note that if tv---vmvy%l, then aiv---va01.) But since Jig an ideal, z=2v--+v2 is also in J. Hence the indicated condition is equivalent to saying that zvy 1 for all z in J, The additional remark follows from the fact that zvy=1 is equivalent toy <2. Definitiog. An ideal M of a Boolean algebra B is said to be maximal if and only if M is a proper ideal and there is no proper ideal J of @ such that MCJ. Theorem 5.19. Given a proper ideal M of a Boolean algebra 3. Then M is maximal if and only if, for any y in B, either y €M or y’ EM. Proof. (a) Assume M is maximal. Assume y ¢ M. We must show that y’€M. Let I = Gen (MU{y)). Then J is an ideal of B such that MCI. Hence by the maximality of M, = B. There- fore by Corollary 5.18, y’ =z for some in M. Since M is an ideal, y’ EM. (0) Assume that y€M or y EM for all y in B. Assume MCJ, where J is an ideal. We must prove that J=B. Let y€J~M. Since y€M, y'EM. Hence y’€J. Then l=yvy EJ. Therefore J=B.> Definition. An ideal J in a Boolean algebra is prime if and only if, for any 2 and y inB, (t€J&yEI > ery Ed. Theorem 5.20. A proper ideal J in a Boolean algebra B is maximal if and only if it is prime. Proof. (2) Assume J is maximal. Assume z¢J& y@J. Then by Theorem 5.19, 2’€J and y' EJ. Hence 2’vy’ EJ, Since J is proper, cay =(2’vy/y €J. Thus J is prime. (0) Assume J prime. Given y€B. By Theorem 5.19 it suffices to show that yeJ or vEJ. Assume y¢J and y'€J. Since J is prime, O=yay’€J which is @ contradiction. »146 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 ‘Theorem 5.21. The maximal principal ideals are the principal ideals J., where u is an atom. Proof. (a) Assume u is an atom. To prove Jy maximal, we shall use Theorem 5.19. Assume yeB and y@Ju. Then y+w'. Hence uy’. Since u is an atom, u yee () cary & yes > cayz. @) ary & asd) > (ery & tAGEsyrd & zvaHsyvd). Proof. (e) s+2=06. () ety = ute () ste= 24042 = rt(ysy tz = (ety) +(yt2). (@) Assume z=,y and a,b. Since 2’ +y/ = 2+y, it follows that x’ =,y'. Next, notice that for any z, (az) (yz), since (2az)+(yaz) = (z+y) az. Hence ERassyaa = anyesbay = yard Lastly, ava = (tna) mi (y aby = yvb > Definition. Ie) = @w:2=s9) [2] is called the equivalence class of x modulo J. Clearly, z € [z].CHAP. 5) TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 1a Theorem 523. (a) [2] = [2] oz =z. (©) [2] * [2] > [z]ne] = 9. Proof. (a) First, assume (z] =z]. Since #€ [2], we obtain x €(2], = =,2, Then for any y, 2 =,2. Second, assume Elz] > 2esy > 2 ary > ye [lz] Hence [z]C[z]. Similarly, [2]¢[z]. ‘Therefore [2] (0) VE [z]N fe] + (@mry & emyy) + ee > [e]= [21> Given an ideal J of the Boolean algebra B. Let B/J be the set of equivalence classes modulo J. We define operations 4,, v,, ’ on B/J as follows: (la, fy) = fay] le] vs ty) = fev yl (els = [] These definitions are meaningful by virtue of Theorem 5.22(d). For example, if X and Y are equivalence classes modulo J, we select any €X and any y€Y. If m€X and EY, then [2,4] ={2iayi] by virtue of Theorem 5.22(d) and Theorem 6.23(a). Thus [24] does not depend upon the particular x chosen in X or the particular y chosen in Y. This shows that our definition of a, makes sense. Let 0, stand for (0), and let 1, stand for [1]. Clearly, 0,=J and 1, = (y: y’ J). Let B/J stand for (B/J, Ay vy,'s, 0,,1,). ‘Theorem 524. If J is a proper ideal of a Boolean algebra , then B/J is a Boolean algebra (called the quotient algebra of B by J). Proof. We must check the axioms for Boolean algebras. 1 [ely ty) = levy] = [yva] = [w)v Ee) Axioms (2)-(8) are proved in a similar manner. (This is left as an exercise for the reader.) 2 0,%1), since 0,1. (This follows from the equation 0+1=1 and the assumption that J is a proper ideal.) » Example 5.11. Let J be the ideal {0}. Then z=, y if and only if + =y. Thus the elements of B/J are the single cans (z}, where 2€B, The function f(x) = (z) is an isomorphism between @ and B/{0). Theorem 5.25. B/J = (0,, 11} if and only if J is a maximal ideal. Proof. If J is maximal, then, for any 2, either 2€J or 2€J. Hence [z]=0, or . izi=. Conversely, assume that for every x in B, [2]=0, or [2]=1). Hence 2€J or x GJ. Therefore by Theorem 5.19, J is maximal. > Remark. If c€B, then [z]=2+J, where 2+J stands for (r+u: uéJ). For, on the one hand, y € [2] +z+yEJ>a+y=u for somewin J+ y=z+u for some uJ. On the other hand, if y= 2+ for some u in J, then y+2 = uJ, and therefore y € [2].148 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 5.7 THE BOOLEAN REPRESENTATION THEOREM The theory of Boolean algebras is intended to be a generalization of the algebra of sets. We already have proved (Theorem 5.7) that every finite Boolean algebra is isomorphic to the Boolean algebra of all subsets of some set A (namely, A is the set of atoms). This is not true for all Boolean algebras. For, in Problem 5.16 we cite an example of an atomless Boolean algebra 3; and since the Boolean algebra P(A) of all subsets of any set A is atomic, B cannot be isomorphic to P(A). However, we shall show in what follows that every Boolean algebra is isomorphic to some field of sets. To this end, we shall need the following general mathematical principle. Zorn's Lemma: Given a set Z of sets such that, for every C-chain C in Z,* the union ave 4 is also in Z. Then there is an C-maximal set M in Z, ie. MEZ, and, if A is any set in Z, then M¢ A. A proof of Zorn’s Lemma (based upon the use of a more transparent assumption, the so-called axiom of choice) is given in Appendix C. Theorem 5.26 (Maximal Ideal Theorem). If J is a proper ideal in a Boolean algebra B, then there is a maximal ideal M in B such that JCM (i.e. every proper ideal can be extended to a maximal ideal). Proof. Assume J is a proper ideal. Let Z be the class of all proper ideals K in B such that JCK, Now assume that C is an C-chain in Z Then U_ J is a proper ideal (ef. Problem 5.34) containing J. Hence by Zorn’s Lemma there is a maximal set M in Z. But M2J and M is a maximal ideal in @ (for, if M* is any proper ideal such that MC MS, then M*€Z and hence M = M*).> Example 5.12, Consider the Boolean algebra P(A) of all subsets of an infinite set A. Let J be the ideal of all finite aubsets of A. By Theorem 5.26, J can be extended to a maximal ideal M. M eannot be a principal ideal. (For, a principal maximal ideal consists of all subseta of A not containing some fixed element 6 of A (ef. Example 5.9), while every singleton {6}, being finite, belongs to J and therefore to M. Hence if M were principal, we would have A = {b}U(A ~{5})EM, and then M would not be a proper ideal.) No way is known for describing such an ideal M in a constructive way, i.e, no property is known for which the sets satisfying this property form a maximal ideal M containing J as a subset. Corollary 5.27. Every Boolean algebra has at least one maximal ideal. Proof. The set (0} is an ideal. Hence by Theorem 5.26, {0} can be extended to a maximal ideal.» Corollary 528. If C is any subset of a Boolean algebra B such that for any ws, ...,Um in C, wiv +++ vmx, then there is a maximal ideal M containing C. Proof. By Corollary 5.16, the ideal Gen(C) generated by C is a proper ideal. Then by Theorem 5.26 there is a maximal ideal M containing Gen(C), and therefore also containing C.> ‘By an C-chain C in Z we mean a subset of Z such that, if AEC and BEC and AB, then either ACB or BCA. More generally, if R is a binary relation on a set W, then an R-chain in W is a subset of W on which R is transitive, connected, and antisymmetric (ie. 2Ry&yRz ~ ==).CHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 4g Corollary 5.29. If z is a nonzero element of a Boolean algebra @, there is a maximal ideal M not containing the element 2. Proof. x’ x1. Hence by Corollary 5.28 there is a maximal ideal M containing 2’. Hence z€M. (Otherwise, 1 = 2v2’EM)> Theorem 530. (Stone's Representation Theorem). Every Boolean algebra B is isomorphic toa field of sets. Proof. For each z in B, let =(z) be the set of all maximal ideals of @ such that z¢M. Clearly, =(1) is the set 2 of all maximal ideals of 8, while =(0) is the empty set. For any 20, 3(2) is non-empty, by Corollary 5.29. Also, =(2’) = =(2), since every maximal ideal “M contains precisely one of the elements « and x’. Finally, =(zvy) = =(z)U=(y). (First, if z@M, then zvy@M, since savy. Likewise, if vEM, then zvy@M. Hence (y) Cz(avy). Conversely, if z€M&yEM, then zvyEM. Hence if zvy eM, then z€M or y€M. Therefore =(zvy) C =(2)UZ(y).) Thus = is an isomorphism of B into the Boolean algebra ?(21) of all subsets of the set Vf of all maximal ideals. The range of 3 is a field of sets isomorphic to B. > Corollary 5.31. For any sentence A of the theory of Boolean algebras, A holds for all Boolean algebras if and only if A holds for all fields of sets. ‘By a sentence of the theory of Boolean algebras we mean either an equality +=«, where + and o are Boolean expressions, or an expression obtained from such equalities by applying the logical connectives and the quantifiers “for all x” and “there exists an 2”.) 58 INFINITE MEETS AND JOINS Given a Boolean algebra B = (B, a,v,’,0,1) and a subset A of B. If the least upper bound (lub) of A exists, it is denoted V and called the join of A. Thus in order for y ta to be the join of A, it is necessary and sufficient that: (a) # There are various extensions of the Duality Principle. The transformation from A to the dual A* exchanges = and >. (For, r=y o zay=2, and the dual of zay=z is 1+y=2, which is equivalent to z=.) In addition, the taking of duals interchanges the general notions of meet and join. Example 5.16. The second part of De Morgan’s Laws (Theorem 5.32) ) Az = (y,@) is the dual of the first part: @ Vr = (a «) Hence our proof of (a) automatically is also a prooftvf (b). Example 5.17. From = 0, it follows by duslity that A # = 1. Similarly, = = 0 follows from 5.10. INFINITE DISTRIBUTIVITY Theorem 535. (a) 2. Vi u= Vy (eau). Oey Awa A, eva (These equations are intended to mean that, if the left side is meaningful, then the right side is also meaningful and the two sides are equal.) Proof. Since (6) is the dual of (a), it suffices to prove (a). As a preliminary, note that anb=c + b anc=b (52) c= Assume now that y wexists. If veA, then v< yu and therefore TAU SrA Vu152 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 Thus x» Vu is an upper bound of the set Z of all zu, where u€ A. Now assume w is an upper bound of the set Z. Then for all uA, 2au “ea nea Remark, If the right-hand side of (a) or (b) of Theorem 5.85 is meaningful, the left-hand side need not be. For example, if 7=0, then y (cau)=0, but Vu need not exist if B is not complete. wea nee By means of the ordinary distributive laws, one obtains identities of the form (en van ++ vty) 0 (Gav tev + vee) Am (Gm md Zink) = Yen, HAMA AEnIQ) For example, (envau) a (env aeevae) = (onze) v (en ater) v (Gua zn) v (Bie zn) v (12 A was) v (12.0 22s) ‘One even can extend these identities to the following infinite case. (uaz) Guaza) a9 Cy) = Ya, in ean no 0 uy) (68) where the join on the right is taken over all possible terms zu, 0 cay 0° A ey, with i:€Ay, ..., je@As and where we assume that all the joins on the left-hand side exist. This identity is proved by induction on k, using Theorem 5.35(a) (plus generalized associ- ativity; ef. Problem 5.37), For any sets S and W, let S” stand for the set of all functions from W into S. Assume given a function assigning to each we W and s€S an element z,, of a given Boolean algebra B, Consider we (Ya?) = Mee (ade 2) 4) where the join on the right extends over all functions f € S¥. Definition. If m and n are cardinal numbers, the Boolean algebra B is said to be (m, n)-distributive if and only if, whenever W hes cardinal number m and S has cardinal number n and te,s is any assignment of elements of B such that the left-hand side of (6.4) makes sense and each term A v,yo) on the right-hand side makes sense, then the right- hand side makes sense and the equation (5.4) holds. @ is said to be completely distributive if and only if B is (m,n)-distributive for all cardinals m and n. @ is said to be m-distributive if and only if @ is (m, m)-distributive. ‘We have seen above (equation (5.8)) that if m is finite every Boolean algebra is (m, n)- distributive, no matter what n is. Obviously (m, n)-distributivity also holds when n= 1. However, if m is infinite and n=2, then a Boolean algebra need not be (m, n)-distributive, even when n is finite,CHAP. 5} TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 153 It is easy to see that, if is (m, n)-distributive, then (c) p=m + B is (p, n)-distributive; (0) p B is (m, p)-distributive. To verify (a), it suffices to choose a set W* such that WAW*=Q and WUW* has cardi- nality m, and then to extend the given assignment by letting ,.=1 for all weW*. To Prove (b), we need only choose a set S* so that SNS*=Q and SUS* has cardinality n, and then extend the given assignment by letting zv,.=0 for all s€S*. From (2) and (8) it follows that ® is completely distributive if and only if B is m-dis- tributive for all m. It can be shown that & is m-distributive if and only if B is (m, 2)-distributive (cf. [140] and [149}). Example 518. The field P(X) of all subsets of a non-empty set K is completely distributive. This follows from the tact that (yt) = the (04 seam) eo always holds for arbitrary sets 7, Remark. By the Duality Principle, it is not necessary to give separate consideration to the dual of (5.4): ete (dst) 7 de Ov eee) 5.11 m-COMPLETENESS Let m be an infinite cardinal number. A Boolean algebra @ is said to be m-complete if and only if every subset of B having cardinal number =m possesses a least upper bound (lub) and a greatest lower bound (gib). By De Morgan's Laws (Theorem 5.32), for m-completeness it suffices to know only that every subset of B having cardinal number =m possesses a lub (or that every subset of B having cardinal number =m possesses a glb). If Xs is the cardinal number of the set of integers, it is customary to use the term o-algebra instead of 8e-complete Boolean algebra. Thus @ is a o-algebra if and only if every denumerable subset of B has a lub. We shall use the term m-complete field of sets for a field of sets F such that any subset of F of cardinality y. Hence min(z,max(y,2)) = min(z,2) and min (x, max (y,z)) = min(z,y). Therefore min (x, max (y, 2)) > max (min (z, y), min (z, 2)). Now assume min (z,max(y,2)) > max (min (z,y), min(z,2)), and we shall derive a contradiction. It follows from our assumption that = > max (min (z.y), min(z.:). Therefore z > min(z,y) and 2 > min(z,2). Hence > y and z > z Consequently max (min (z,y), min(e,2)) = max(y,2) and min (z, max (y,2)) = max(y,2), contradicting our assumption. The other dis- tributive law (L6) must hold by virtue of Theorem 5.2. For (A, =) to be complemented, there would have to be a least element 0 and = greatest ele ment 1. Also, for any # in A, max(z,z’)=1 and min(z,z) =0. But either » = max(z, 2’) or #=min(2,e’), Hence #=1 or z=0. Therefore A would have to contain at most two elements. (®) By Part (a), such » lattice is given by (I, =), where 7 is the set of all integers (positive, zero, and negative) and = is the usual order relation on integers. Show that lattices can be characterized by the six laws: @ avy =yve | commutative laws (0) tAy = yar (2) ev (uve) = (evu)ve aa associative laws (@ ayant) = (wavaz (©) tv(eay) = 2 () ta(evy) =e in the following sense: if (L, a, v) is a structure such that » and v are binary operation on the set L satisfying the laws (a)-(f), and if we define =y by zay=z, then (L, =) is a lattice with Iub({z,y))=2vy and glb({2,9))=20y. } absorption laws Solution: ‘We already know, by Theorem 6.1, that any lattice satisfies (a)-(f). Conversely, assume that (L,a,v) is a structure satisfying (a)-(/) and define ry by zay=. Notice that zy zvy=y. (For, assume zy, ie. zay=z. Then by (¢) and 0b), y=yVVAs) =¥VE= zvy. Conversely, if y=avy, then, by (), ==2Alevy)=zAy, be zy.) Now, () #42, be eazea, (Por, by (f) and (¢), #= 2A (ev (ena) = 2A) (i) (@4y&ys2) + 242, (Por, we are given 2Ay = 2 and yar =y, Then 202 = G@awnz= zalyas) = zAy = 2, ie 22)156 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 (ii) @ (eave =aa(yv2) Solution: Assume zx, Hence s (zay)ve = 2a(yv2) (b) Prove that distributivity implies modularity. Solution: (a) This follows immediately from Problem 5.6. 'b) This is a direct consequence of Problem 5.5(6(v)). 58 Determine which of the lattices appearing in Problems 5.1(c,d) and Example 5.2 are modular. ‘Solution ‘The lattice of Problem 5.1(c) is This is not modular, since b=d and bv(end)*(bve)nd, (Namely, bv(end)=bva=b and (bve)Ad=end=d) The lattice of Problem 5.1(d) is ‘This is not modular since it contains the lattice of Problem 5.6(c) as a sublatticet, and we have just seen that the latter is not modular, The lattice of Example 5.2 is ‘This lattice is modular. To see this, we must verify Sa + eatyv2) <(eayve determined by a subset of L closed under » and ~, ‘tBy a sublattice of a lattice (L,+) we mean a lat158 59. 5.10. SL. TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 It is obvious that if z=, the result #A(yVv 2) <(zay)v is an immediate consequence of the absorption laws, Therefore we may assume that z <<, It is obvious that this implies that = 0 or e=1, But if 2=0 the inequality reduces to zay“2ny, and if 2=1 the inequality reduees to yvs#yva, Show that a lattice is modular if and only if it satisfies the identity @atevm) vu = (evu)a(evy) Solution: Assume L is modular, In the condition for modularity of Problem 5.7(a), substitute = for y, y for z and zvy for x, The antecedent then becomes the true statement y “=v y, and we obtain (atevMvy = evn aleve) Conversely, assume that the indicated identity is true. Exchanging y and z, we obtain the Identity Wrevays = wa) atevay @ Now, to prove modularity, we assume 21 >a, >¢0. Let vo x~2y m= e~ Ay w= My~ ey... Then WAY, whenever #% j. ° Exhibit a Boolean algebra which is not isomorphic to any Boolean algebra of the form P(A). Solutio Consider the Boolean algebra & of statement bundles, based upon the propositional calculus (cf. Example 35). B is denumerable. For, since there are denumerably many statement forms, there ean only be countable number of statement bundles, However, distinct statement letters determine distinct statement bundles, since distinct statement letters are not logically equivalent, Hence there are denumerably many statement bundles. Assume @ is isomorphic to some P(A). ‘Then A must be infinite, for otherwise P(A) would be finite and so would . Thus P(A) is denumerable (since it is isomorphic with @) and A is infinite. By Cantor's Theorem (Problem 2.22), A must have smaller cardinality than P(A). A is equinumerous with a subset of (A), and there- fore A must be denumerable, But then A would be equinumerous with P(A) (since both are denum- erable), contradicting Cantor’s Theorem. (What we have shown, by means of Cantor's Theorem, is that no denumerable Boolean algebra can be isomorphic with any P(A), and we also exhibited a particular denumerable Boolean algebra.)CHAP. 5! TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 161 5.16. Show that the Boolean algebra B of statement bundles (cf. Example 3.5) is atomless. Sotation: Assume [A] is an atom. Then [A] #0g, and A cannot be a contradiction. Let A, be any Satement letter not occurring in A. Then A does not logically imply Ay for we can assign Ay the value F and assign the statement letters in A suitable values oo that A is T. In addition, A, & A is not a contradiction, since we can make A, T and, at the same time, make AT. Also, A, & A Jogically implies A. Thus 0g <(A,&A] oy = of oO rtwew © Etwaw =0 © (Aw) + aw) = raw =o ‘When £A.' =0, all solutions are of the form ((s-+w) v2) Aw for arbitrary 2. But [le+1) Aw] v [enw] = (wre) v (enw) = waleve) ‘The equation Aw’ =0 always has solutions + and the solutions are all elements yaw for arbitrary y. Thus syne w= wAleve) = wAleVYAu)) = wAev¥vw) = walevy’) for arbitrary x, y and w ) UAE = UAW & [CUA AAU)! Y Uru) A(urz] [Urdatu'y wi] y [(urw)aww'y eh) SO UAtAw)y (uawar) =0 OF uAlenwiv wre) © une =0 A solution exists if and only if 0 = (w +2)’. Hence a solution always exists, The solutions are w= 2A (wz) for all z,1,, 5.21. Find all solutions of the system u = (zeaw)vo @ 2 = (vaw)vu @) wv seve © Solution: (@) is equivalent to (Ale Vw Av) v Walerwve) = 0 (@) (6) is equivalent () is equivalent to ee ® (©) is equivalent to e [wv nazan]y [wlav'ac'va)]164 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 Hence we must solve the system (a’), (8), (c’). By Problem 5.19(t), this system is equivalent to the single equation ur [e vw) Av) v (WA [enw vo) v (EA[ Vw) An) Vv [wv a rzaw] v [w'as'ale'va)] = 0 which is equivalent to (HA ME V0) Ao] v [wl Av) v (ul a {lew 0} v [en (o'v wi] v [wv vag] v [w'Avrz'}) = 0 Hence o is [(e' vw) A0'] v [uw av’), and oy is [enw va] v en (v'v wi] v [Gov v) Az] v [wav az] By easy calculation, op=vvevw' and o,=0. Hence a solution exists if and only if vvrvw' =v. But the latter equation is equivalent to v’A(evw') By Problem 5.20(6), substituting 0” for u, 2v w' for z, and 0 for w, we obtain the solution v= #A (ev w= za(e' aw), ie. v=2'vevw'. The solution for u is w= (egy) Aci = (vvev wry) A. Thus the solutions are u=v=2'vzvw' for arbitrary z,2,w. AXIOMATIZATIONS 5.22, In our axiom system for Boolean algebra (cf. Section 3.2), prove the independence of each of Axioms (1)-(4) and (7)-(9). Show that each of Axioms (5)-(6) is not independent. (A member A of a system JV’ of axioms is said to be independent if and only if A is not provable from the set J/~ (A) of the other axioms.) Selution: Axiom (I): =v. Complement: 0/=1, 1" vv. Define a model ({0,1),A,v,',0,1) as follows: =0, Join: vio 1 o}o o ataoa In the above table, the value of vy is to be found at the intersection of the row to the right of z and the column under y, Thus Ov0=0, Ov1=0, 1v0=1, Ivi=a Meet: ajo 4 ofo o rto o Thus 0A0=0, OAL 100 Jalsa ‘Then Axioms (2)-(9) hold, but Axiom (1) is false. (1) fails because 1v0=1%0=0V1. To see that Axiom (2), 2Ay=yAz, holds, observe that all meets are 0. Verification of the other axioms is left as an exercise for the reader. For the independence of Axiom (2), use 0'=1, 1/=0, and the operations vio. ajo a olo 4 ofo o adaroa ata1oaCHAP. 5) TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 165 For the independence of Axiom (3), 2A (yv2)=(eAx)v(eAs), use 0'=1,1'=0, and vio1 ajo 2 ofo a oj1 0 atron ato 4 Note that 0A (Ov 1) =0A1=0, but (On0)v(0A1)=1v0 For the independence of Axiom (4), xv (ys) =(zvw)A(eva), use 0'=1, 1'=0, and vjo 1 ajo 4 ofo 1 ofo o rt1i oo 1jo 1 For the independence of Axiom (7), zv2’=0, let the domain of the model be P(A), where A is any non-empty set, take A to be M, and v to be U. Let 0 be @ and let 1 be A. However, let for all z. For the independence of Axiom (8), use the same model as for Axiom (7), except that =1 for all z. For the independence of Axiom (9), use the model (Q}, with a, v, ‘asn, U,and— (9N@= 99 =H=9); 0=1=9. To show that Axiom (5), #v.0= 2, is provable from the rest, note first that 2v1=1 for all z. For, 1 = eve = ev(e'al) = eve) a(evl) = 1A(evl) = av Hence £V0 = ev(zaz) = (eal) (ear) = ZA(lv2) = eal =e ‘To show that Axiom (6), 7A1=<2, follows from the rest, “dualize” the proof just given for the axiom 2v0=2, Thus, firs, <\0=0 for all z. For, 0 = Ak = ZA(e'v0) = (enz)v (EA0) = OV(zA0) = AD Hence EAL = zaleve’) = (evO)A (eve) = 2v(0Az) = zv0 =a Detailed verification that the examples in the independence proofs satisfy the remaining axioms ig left to the reader. IDEALS 5.23. If C is a subset of a Boolean algebra @ and if an ideal J contains C and is contained in every ideal containing C, show that J is the ideal Gen (C) generated by C. Sotuti We have to show that J is equal to the intersection H of all ideals containing C. Since J is contained in every ideal containing C, it follows that JGH. On the other hand, since J is itself an ideal containing C, HCJ. Therefore, J =H. 5.24, If J is an ideal of a Boolean algebra B and y & B, prove that Gen (JU {y}) is a proper ideal if and only if y' € J. Solutior ‘Assume y/€J, Hence 1 = yvy' € Gen(JU{y)), and therefore Gen (JU(y}) is not a proper ideal. Conversely, assume that Gen (JU{y}) is not a proper ideal. Then 1€Gen(Juty}), and by Theorem 5.17 there exist 2€ B and 7€J such that 1=(zay)vz. Hence v= VAL = VACEADY2) = WAEAW)V (AR) = Az But since z€J, y'Az€J and therefore y'€J.166 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS [cHaP.5 5.25, By a congruence relation R on a Boolean algebra B we mean a binary relation R on B 5.26, 527. satisfying the following properties. (a) 2Re (reflexivity) (0) «Ry > yRx (symmetry) (c) (2Ry & yRz) + xRz (transitivity) (@) Ry > (Ry & (e02)RY 2). Define Je = (x: 2R0}. Prove: (i) Jx is an ideal. (i) Je is a proper ideal if and only if 1(0R1). (iii) 2 =s,y if and only if zRy. Solution: (i) By (@), 0€Jg. Assume #€Jp and yEJq. Hence 2RO and yRO. By (2) 2'R1, and, again by @, @'Av)Ry'. But from (2), yRO implies y’B1. Hence by (c), (2’ a 'VR1, and, again by (d), (@'AvYRO, ie. zvy Edy. Now assume that z€Jp and s€B. Then 20 and, by (d), (en2)R0, ie. eA E (i) This follows imme contain 1. (ii) Since + is definable in terms of the meet and complement, it follows by (d) that 2Ry + (+ WRY+v), ie. zRy>(e+yR0. But (2+ y)RO is equivalent to x+y €Jg, which in turn is equivalent by definition to = Jn ¥. tely from the fact that an ideal is proper if and only if it does not A subset F of a Boolean algebra B is said to be a filter if and only if: (i) F is non- empty; (ii) (QF &YyEF) + 2ayEF; (ii) re &YEB>zvy EF. By an wtra- filter we mean a proper filter which is contained in no other proper filter. Prove: (a) F isa filter if and only if F’ = {2': 2€F) is an ideal. (0) F is an ultrafilter if and only if F” is a maximal ideal. (€) Assumption (iii) in the definition of filter may be replaced by (it) zEP &2=y > yer Solution: (@) Assume F is a filter. Given 2€F', yEF', 2€B. Then #/€F and y'€F, Hence 2’ ay €F and 2'v2'@F, Therefore zvy = (2'ay'' EF" and az = (eve EF’. Thos F’ is an ideal, The converse is left as an exercise for the reader. (®) This is an immediate consequence of (a). (©) The equivalence between (iii) and (ii’) follows from the equivalence between zy and evyey Call a Boolean algebra @ simple if and only if {0} is the only proper ideal. Prove that @ is simple if and only if B= (0,1). Solution: Clearly, if B = (0,1), then (0) is the only proper ideal. Conversely, assume @ is simple. Let be any element of B different from 1. Then the principal ideal J, is a proper ideal, since 1+ 2. Since @ is simple, J, = {0). But 2€J, and therefore == 0, ie. B= (0,1). In a Boolean algebra B, a set C of ideals of @ is said to be an C-chain of ideals if and only if, for any J; and Jz in C, either JiCJ2 or J2CJ;. (This amounts to saying that the relation ¢ totally orders C.) Prove that the union of an C-chain of ideals is again an ideal, and, if each ideal in C is proper, so is the union.CHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 167 Solution: Let H= UJ, where C is an C-chain of ideals. Given 2 and y in H, and sin B. Then ree EJ, and y@J, for some J; and Jz in C. Since C is an C-chain, either JyCJp or JpCJy, say, TyCJq, Hence zE J, and yE Jy Since Jy is an ideal, zvy € Jy and zAzEJ,, But since J,cH, we obtain: 2vy@H and zAs€H. Hence H is an ideal. If each ideal in C is proper, then 1€J for each J in C. Hence 1H and therefore H is also proper. If R = (R,+, x, 0) is a commutative ring, then a non-empty subset J of R is called a ring-theoretic ideal if and only if @) Es & ye) > 2-yeL; (ii) @EJ & 2ER) > 2xzEJ. Prove that if J is a subset of a Boolean algebra B = (B,n,v,',0,1), and R (B, +, 4,0) is the corresponding Boolean ring, then J is an ideal of B if and only if J is a ring-theoretic ideal of R. Solution: Notice that, for Boolean rings, (—y) = y, and therefore we may replace z—y in condition (i) by x+y. Now assume that J is an ideal in B. We already know that (c€J& yEJ) + 2+yEJ, which is condition (j), while condition (ji) reads (e€J & 2€B) > zn2€ J, which is part of the definition of an ideal. Conversely, assume that J is a ring-theoretic ideal, By (ii), (¢€J & 2€B) + zA2€J. Now it remains to show that (EJ & yEJ) > zvyEJ. So assume EJ &yEJ. By Gi, zAvES, and, since zvy = 2+y + (zAy), we may conclude by (i) that zvy EJ. Show that, if J is an ideal of a Boolean algebra @, then 2 =,y if and only if there exists some element z in J such that zvz=yvz. Solution: Assume #5, y, ie z+y@J. Let z=2+y. Then eves av(ety) = 2vy = yv (tty) = ave Conversely, assume 2vz=yv2 for somezinJ. Then BAW = (evaay valay = zayl se Hence tay €J. Also, way = ate) = eave) = Ae se Hence x'ay€J. Therefore x+y = (zay)v (2 ay) EJ. QUOTIENT ALGEBRAS 33L. Let B be the Boolean algebra P(A), where A is some infinite non-empty set, and let J be the ideal of finite subsets of A. Prove that the quotient algebra @/J is atomless. Solation: Given an element (X] of @/J such that [X] x [0]. Hence X is infinite, Then there is an infinite set Y such that YCX and X~Y is infinite, (To see this, enumerate a subset of X, (ay,0,,...}, and lot ¥ = (4,a,05,-..}.) Since ¥ is infinite, [¥] # 0g/z. Also, since ¥ CX, [¥] = [X]. However, since X~¥ is infinite, X+YEJ, ie, X*,Y. e [X] «[¥]. Thus avs <[¥] <[X], and therefore [X] cannot be an atom.168 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS [cHAP. 5 582. Given Boolean algebras of = (A, rcp Voss’, 0.4104) and B=(B,rg,vg,'2, 0g, 1g). A function f from A into B is called a homomorphism from of into @ if and only if @) fer) = (F@ys; Gi) fer y) = fle) re fly). If such a function has its range equal to all of B, then it is called a homomorphism from of onto , and @ is called a homomorphic image of of. A homomorphism f is an isomorphism of of into @ if and only if f is one-one, and f is called an isomorphism of of onto @ if and only if f is an isomorphism of of into B and the range of f is B. We say that of and ® are isomorphic if and only if there is an isomorphism of of onto B. If f is any function from A into B, by the kernel K, we mean the set {2: 2€A & f(z)=0g}. Prove: (@) If f is a homomorphism from of into 8, then Q) feva) = He) ve Kyi 2) Meta) = fz) +a fy; (8) 4) = % and fly) = (0) If f is a homomorphism from of into @, the range f(A) determines a subalgebra of B. (0) If f is a homomorphism from of into B, and C determines a subalgebra of @, then fC) = (2: cEA & f(z)EC} determines a subalgebra of of. (@) The identity mapping [, is an isomorphism from eA onto of. (Hence the relation “isomorphic” is reflexive.) (e) If f is an isomorphism from ef onto , then the inverse function f-1 from B onto A is an isomorphism from @ onto of. (Hence the relation “isomorphic” is symmetric.) () If / is a homomorphism from of into B, and g is a homomorphism from @ into an algebra C = (C,A¢,v¢,’C, Oc, 1c), then the composition g of is a homomorphism from of into C. In particular, if f and g are isomorphisms onto, then 20 is gf. (Hence the relation “isomorphie” is transitive.) (9) A homomorphism f from of into B is an isomorphism of of into @ if and only if the kernel Kj= (0,,). (h) If J is an ideal of of, then the function f(z)=2z+J for all 2 in A is a homomor- phism from of onto c4/J (called the natural homomorphism from of onto cf/J). () If Bis a homomorphic image of of, then there is an ideal J of of such that B and eA/J are isomorphic. Solution: @ Ha veg v) = Maoh Agy y'ot Vor) FAA VAR = (la’et) ng flor) = (MEN8 Ag fUY'2¥9 = Mz) ve Mv) ‘The proof for + is similar, since + is definable in terms of A, v,'. Now Wn) = Merce 2A) = fle) rg fleet) = fit) rg flz'® = Og Hence fe) = 10.7) = /0,4)'2 = 09 = 1g.CHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 169 5.33. (®) From now on we shall omit subscripts cf and ®, wherever this is not likely to cause confusion, sume now that f is 8 homomorphism from of into B. Assume %, vE/(A). Then w= f(z) MH) for some z, y in A. Hence unv=fl2)afty) = fe ry) € f(A). Similarly, = fle)E f(A). Hence f(A) determines « subalgebra of B. w= fle) (©) Assume z,y €f-4C). Then fle ry) = fe) f(y) EC, since C is closed under A. Similarly, fa’) = fla) EC, since C is closed under complementation. Thus zy € f-¥C) and 2’ f-"(C). (@) This is obvious. (¢) Given wv €B. Then J0-Mu nv) = 400 = {E-HO) AST-HO) = FF-H(u) 0 FH) Since f is one-one, f-Yu nv) = f-Mu)Af-Hv). Similarly, ff-1W)) = w= Uf-Uu) = WG-(u))). Since f is one-one, f-\w’) = -*(u)y. “ (@2Alery) = olflerw) = o(f=) > fu) = off2) oF) = (9° Me) 0 (oflv) Similarly, (eNe) = oe) = ofley) = (ey = (oeNcey”’ = f(0). Since f is one-one, 2=0, Con- (®) Assume f is one-one, and let €K, Then f(2) = (2) + fle) = 0. versely, assume K, = (0}, and assume (2) = f(y). Then f(z +y) = f(z) +fv) Thos z+y EK; but, since K,=(0}, 2+ =0, which is equivalent to z= w Keay) = ay) tT = [ens] = fla [y) = fe) as) Similarly, Key = at +d = [2] = [al = fay (9 Assume f is a homomorphism from ef onto . Let J=K, J is an ideal. (For, if zy €J and EA, then f(zv y) = f(z) v fy) =0v0 1d fle nz) = fe) a fla) = OAs) =0) For any « in A, we define F([z}) = f(z). This definition is independent of the choice of the partieu- lar representative 2 in [z], since tyes > flety=0 > fit fy) =0 > fe) = 1H) Now Fife] A{ue)) = Fileaw)) = Kenu) = fle)» foo) = Piel) a File) Similarly, F(z} = File) = fle) = fo = File)’ To see that F is one-one, we check that the kernel of F is {0}. Assume Fi({z])=0. Then M2) =0. Hence s€K;, and therefore [2] =0, That the range of F is B follows from the fact that the range of f (a) Let } be a nonzero element of a Boolean algebra B. Let By denote (2: 2 waveyy & (uavtva,0 0) Since (=) = [z], s.a2 5,0. Since J Assume now that (s] is an upper bound of the [: only if wav’ sy 0. (For, [ul =[o} < [unc] (UA) + (LAD) By 0 UAT +4) By 0 wre ideal, VY (a nz’) 0, te. aA \/ x =, 0, Hence ly «| < (2). That the natural ‘ y if mapping ¢(z) = 2+J is a e-homomorphism of of onto of/J is an easy consequence of the fact m= [on] = ye For any subset CC P(K), the intersection of all o-subfields of P(X) containing C is itself a o-subfield containing C. Solution: ‘The intersection H clearly is closed under denumerable joins and complements. The e-subfield Hin called the o-rubfield generated by C. Given a subset C of a w-algebra cf. The intersection D of all o-ideals containing C is itself a c-ideal containing C (called the e-ideal generated by C). The elements of D are all those z= V « for elements « in C (1a’ fore some @'va,,%1; say, av ayy) #1. Then 1 > av aya, = @' vaya v0 = av ayay¥ ox = \(0'voysv ay) Hence 1> a’ Vv ayj3)V Gaya) for some j= f(2); etc. We obtain a sequence Gyr dyyay,--- SUch that 1 > ev ayy ¥ Gye ¥ *** ¥ Gayay for each Therefore among 6,1», Gaya» --+ neither @ nor a complement of any ayy) occurs. Therefore there ‘exists a selection Y containing a and all ay) Thus ¥€r(a), but Y €r(ayq;), contradicting (1). Remark: This result of Loomis fails to hold for non-denumerable cardinalities (125).CHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 179 Supplementary Problems LATTICES 554. jagrams are lattices? Among the i) are complemented, (iv) are ‘Which of the partially ordered sets given by the following lattices, which (i) have a zero element, (ii) have a unit element, modular, (v) are distributive? @ . ) @ (@) Which of the following structures (L,#) are partially ordered sets, totally ordered sets, lattices, distributive lattices, lattices with a zero element, lattices with a unit element, complemented lat- tices? For those which are lattices, describe the operations . and v. (a) Lis the set of all finite subsets of a set A and < is the inclusion relation C. +b) Same as (a), except that A itself is also a member of L. (c) Lis the set of complex numbers and a+bite+di ase, ‘4) Lis the set of all complex numbers and a+ bi = ¢+di <> (a Lis the set of all subsets of a set A, and = is c. © Lis the set of positive integers and zy if and only if y is an integral multiple of z (i.e. = divides y). 3 Assume (A, ) is a given partially ordered set. Let C be a fixed set. Let L be the set of all fanetions fom C into A. For any f and g in L, let {=o if and only if f(s) = (2) for * Assume (A, =) is a given partially ordered set. Let L=A, and xe () EAyaz=avyve > = Prove that the following inequalities hold in any lattice. (@) (2Ay)v Ad) = (ev rtyv yy (©) AW Vv WAs)v ena) = (ev) AWE) AGE) Prove that each of the following conditions is equivalent to distributivity of a (@) (@rny42&eeyva) > 2ee ©) @arsynz & avesyve) + 24 (Hint: Use (a) (©) (@as=yar & eve=yve) + e=y (Hint: In one direction, use Problem 5.11(c).) @ evW ava) rceve) = Av) v WAR V (eA2) Complete the proof of Problem §.11(). Give an example of a distributive Inttice (L,=) Incking both zero and unit elements such that = is not a total order on L. If (L, =) is a lattice, prove that = totally orders L if and only if all subsets of L are sublattices (i.e. are closed under » and v). Prove that any distributive (modular) lattice can be extended to a distributive (modvlar) lattice with zero and unit elements simply by adjoining such elements if they are not already present. Prove that a lattice is modular if and only if it satisfies the law Ay &ave=yveb easy) > (eas (Hint: In one direction, use Problem 6.10.) A lattice (L, =) is said to be complete if and only if every subset of L has a lub and » glb. Prove that in order to verify completeness it suffices to show that every subset has a lub or that every subset has a glb. Given lattices (Ly, ;) and (Ly, =,) and a funetion f from L, into Ly. Definitions. J is an erder-homomorphiem if and only if ©“, > (2) p(y). J is a meet-homomorphiem if and only if f(z 41») = f(z) A2flu). fis a join-homomorphism if and only if f(z ¥1¥) = f(=) v2 fly). J is @ lattice-homomorphiem if and only if f is both a meet-homomorphism and a join-homomorphism. Prove: (a) Every meet-homomorphism is an order-homomorphism. (8) Every join-homomorphism is an order-homomorphism. (©) Every lattice homomorphism is a meet-homomorphism. (@ Every lattice homomorphism is a join-homomorphism. (€) ‘The converses of (a)-(d) do not hold, (Counterexamples may be found using lattices of at most four elements.) ( For one-one functions f from L, onto Lz, the notions of order-, meet-, join, and lattice. homomorphism are equivalent. (9) Any order-isomorphism from a Boolean algebra of onto a Boolean algebra ® is 8 1 isomorphism, not only is it a latticeshomomorphism, but it also preserves complemerts Bootear182 5.3. TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS (CHAP. 5 A lattice (L,=) is said to be relatively peeudo-complemented if and only if, for any z,y in.L, the set (2: yaz=2} has a lub, (Such a lub is denoted y > x.) By a preudo-Boclean algebra, we mean a relatively pseudo-complemented lattice possessing a zero element 0. Prove ({142)): (a) In a relatively pseudo-complemented lattice: () 22 y>x ifandonly if yar=e. i) The distributive laws hold, (iil) For any z, 2 > z is a unit element 1. (iv) y > 2 =1 if and only if y =z, (@) x=y itandonly if 2 >y=y d2=1, i) ydisa Wil) 1>rs2 il) ya >2) =e (ix) fey, then y>2z= @) a=y>e xi) ya >a) Gi) GY>aarae Gil) @ Py AE>2) = 2>yA2) Gir) >a dz) = Wve) de Ox) 2332) = ean d2 = 9394) av) (> ayo) = 2dr ei (2 py) = (yD 2) D(z 2) vy rlxny). Dz and 2>222>y. wae 1b) In a pseudo-Boolean algebra, we define: —x===>0. Then: and -1=0 avy seedy fe 23¥ 4-H ) (CW) = enw) = > (2) (i) Ode (xii) A subset F is a filter if and only if 1€F and (EF & 2 >yEF) = yer (©) In 4 Boolean algebra, show that y >= = y'vz. (d) (For those readers acquainted with elementary point-set topology.) Show that the lattice of ‘alt open sets of a topological spa pseudo-Boolean algebra, where A => B is the interior of AJB. By taking the special case of the real line, show that the assertions 2v (2) =1 and ~-x= 2 do not hold for all pseudo-Boolean algebras. (e) By & Brouwerian lattice we mean a lattice with a unit element such that, for any z,y in the lattice, the glb of the set {z: y= vz} exists (and is denoted y+ z). Show that the notion of Brouwerian lattice is dual to the notion of pseudo-Boolean algebra in the sense that a lattice (B,=, is a pseudo-Boolean algebra if and only if the lattice (L, =) is a Brouwerian lattice. (For a study of Brouwerian lattices, ef. [133].) (A In the Boolean algebra of statement bundles (cf. Example 3.5), for any statement forms A and B, what is the interpretation of [A] > {B)CHAP. 5) TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 183 ATOMS 5.76, 87. 5.8, 5.19. 5.80, sal. 5.820, Prove that two finite Boolean algebras are isomorphic if and only if they have the same number of atoms, Show that there is no Boolean algebra containing 28 elements. In the Boolean algebra of all divisors of , wh what are the atoma? Note that z=y<>z di n is a square-free integer > 1 (ef. Problem 3.8), 5 ¥- (a) How many subalgebras are there of the Boolean algebra of all subsets of a four-element set? (®) Given a Boolean algebra B with 2k elements, show that the number of subalgebras of @ is equal to the number of partitions of a set with & elements (where » partition is a division of the set into one or more disjoint non-empty sets). If of and @ are Boolean algebras with the same finite number 2* of elements, how many isomor- phisms are there from of onto B? Consider the Boolean algebra given by the field of sets consisting of all finite unions of left-open intervals of real numbers (cf. Problem 2.68). What are the atoms of this algebra? Does the set of atoms of = Boolean algebra always have a supremum? Show that every atomic, uniquely complemented lattice with zero and unit elements is isomorphic tp a fleld of sets and is therefore a Boolean algebra. (Hint: Use the proof of Theorem 5:7.) SYMMETRIC DIFFERENCE. BOOLEAN RINGS 5a. 55. 5.36. 531. 588. sat. In an arbitrary Boolean algebra: (a) Does the distributive relation 2 +(y 2) (@) Is @#y use) + etusyto valid (@ + YA (z+ 2) hold? In the Boolean algebra of all divisors of », where » is a square-free integer > 1 (cf. Problem 3.3), find an arithmetic formula for the symmetric difference =+y. In any Boolean algebra, prove: (@) ev(ety)=2vy, (b) the “dual” of z+y is (2+) ?_ Is there any difference In a-ring, 2—y is defined to be z+(-y). In a Boolean ring, what is (— between z—y in the ring-theoretic sense and x ~y as defined in Problem Prove that the uniqueness of (—2) in Axiom (4) for rings need not be assumed (ie. it can be proved from the other axioms). (a) Give an example of a Boolean ring without a unit clement. (4) Show that every Boolean ring without # unit element can be extended (by addition of new elements and extension of the ring operations to the enlarged set) to a Boolean ring with unit element, Prove that the original ring is a maximal ideal of the extension. In the axioms for Boolean rings, show that z+y=y+z is not independent. Let & =(R,+,,0) be a commutative ring with unit element 1. We say that an element 2 in R in idempotent if and only if z=. Let R* be the set of idempotent elements of R. For sny = and y in R*, define s@y=2+y—2zy. Prove that (R*,@, X,0) is a Boolean ring with unit element 1. Express the Boolean operations »,v,’ in terms of the original ring operations +, *.184 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS [CHAP. 5 5.92, For any Boolean expression r(u), prove that ru) = by + rau) where b; and by are fixed elements of the Boolean algebra. 598, Solve the following equations. (@ uaw=0 @) warew (© utw =r (for u and w in terms of r) (@ @awv (enw) =r (for wand w in terms of p, 6, ) 594. Solve the simultaneous equations 5 eab=0 for 2 in terms of b and c, What further conclusion follows if bc? 585. Solve the following system of equations. =) 5.96,_,If r(u) is a Boolean expression and 6 is an element of # Boolean algebra B, and if +(0)ar(1) = 5 = 1(0)vr(1), prove: (@) r(u) = has a solution. Find all solutions, (®) If x{w) = 6 has a unique solution, then, for any ¢, r(u)=e has a unique solution (namely, +(e). (c) If @ is finite and there are k atoms which are # r(0)'+r(1), then r(u) = has 2* solutions. AXIOMATIZATIONS 597, In our axiom system for Boolean algebras, determine whether Axiom (5) can be proved from Axioms (1)-(4), (7)(9). (See Problem 6.22.) 5.98. Determine whether or not each of the axioms (B1)-(B5) for Byrne algebras (cf. Section 5.4) is independent. 599, Show that the following variation of the axioms for Byrne algebras (cf. Section 5.4) also may serve ‘as an axiom system for Boolean algebras. Consider structures (B, A,’) satisfying (B1), (B2), (BB), and () zay = sae & eayae (Hint: Prove that 22’ =waw’ for all s and w, and introduce 0 by definition as being equal to this common value of all z3'.) 5.100, (a) Prove that the following is a system of axioms for Boolean algebras. Wl) fay =yae (D2) zAWA®) = (eawat (D3) (Avy AG’ ayy = (b) Investigate the independence of Axioms (D1)-(D3). 5.01, Prove the independence of Axioms (a)-(f) for lattices in Problem 5.4. 5.102, [121] Let L be a complemented lattice. Show that L is » Boolean algebra if and only if, for any and y in L and for any complement z of y, zAy=0 ¢> 2 =z,CHAP. 5; TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 185 5.103. {101'. Given a structure @=(B,A,'. Prove that @ determines a Boolean algebra if and only if the following two laws are satisfied. (a) r=rAy & ray a i) (en yaza(yazng S10. (1227, Prove that a structure following three laws are satisfied. =(B,v,") determines 2 Boolean algebra if and only if the (a) rvy = yve () (evy)vz = zv(yve) () vv vive = IDEALS, 5.105, If A is an infinite set, prove that the ideal of all finite subsets of A is not a principal ideal field of sets P(A). 5.106. Prove that a nonempty subset J of a Boolean algebra is an ideal if and only if the condition avyEl @ (ces &yE) is satisfied, 3.107. What is the ideal generated by the empty subset of a Boolean algebra? 5.08, Given a Boolean algebra B. For any ideals J and J of @, let Jvl = fevy: 26] & yen (@) Prove that Jv/ is an ideal. (8) Show that the set of ideals of @ forms a distributive lattice under the operations of 9 and v. Are there zero and unit elements? Is the lattice complemented? 5.109. A non-trivial finitely-additive measure on a Boolean algebra @ is a function « from B into the set of non-negative real numbers such that @) levy) = we) + ay) if 2ny (2) wis not a constant function, (a) If » is « non-trivial finitely-additive measure on a Boolean algebra , prove: (0) =0 Gi) aleve vay) = wey) too tale) AE any Gil) ey + Hl2)4 aly) (ie) az vos vm) = ale) + oo + alae) (0) Jy = (2: sl2)=0) is a proper ideal of B. (vi) If 4 is 2-valued, ie. if the range of » consists of two numbers (one of which, by (i), must be 0), then J, is « maximal ideal, (wil) If » is bounded, i.e. the range of is bounded above, and if we define xz) = a(z)/a(1), then y is a non-trivial finitely-additive measure such that 0 (x)=1 and 91) =1. 0 for imi. (®) IFM is a maximal ideal of B, and if we define o if vem 1 if sem for any « in B, show that » is a non-trivial 2-valued finitely-additive measure on B. (©) If B is the Boolean algebra P(K), where K is a finite set, and if we define »(A) = the number of elements in A, prove that » is a non-trivial finitely-additive measure on 3. (a) Prove that every Boolean algebra admits a non-trivial 2-valued finitely-additive measure. wa)186 5.10, TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS [CHAP. 6 Given a subset D of a Boolean algebra @, recall (ef. Problem 3.5) that the intersection of all sub- algebras containing D is a subalgebra Gp, called the subalgebra generated by D. (@) Prove that the clements of Gp are all elements of the form ¢,V-++ vc, where each = dyacead,, and either dj,€D or dj €D, i.e. the elements of Gp are finite joins of finite meets of elements of D and complements of elements of D. (6) Show that the subalgebra Gp of P(K) generated by the set D of all singletons consists of all finite and cofinite sets. Prove that Gp is complete if and only if K is finite. (6) We say that a set D is a set of generators of the Boolean algebra @ if and only if the sub- algebra Gp generated by D is the whole set B. We say that @ is finitely generated if and only if there is a finite set of generators of 8. Prove that every finitely generated Boolean algebra, is finite, (@ We say that a set D of generators of a Boolean algebra @ is a free set of generators of @ if and only if, for every function h from D into a Boolean algebra C, there is an extension g of h which is a Boolean homomorphism of @ into C. The Boolean algebra is said to be free if and only if there is a free set of generators of &. (For any non-negative integer k, if B is a Boolean algebra having a free set of k gen- erators, prove that @ has 2) elements. (ii) Is every subalgebra of a free Boolean algebra also free? (lif) Show that the cardinal number of any infinite free Boolean algebra @ is equal to the cardinal number of any free set of generators of 3. (ivy Show that if D, and Dz are free sets of generators of the same Boolean algebra , then D, and D, have the same cardinal number. (v) If his & function from a free set D of generators of a Boolean algebra @ into » Boclean algebra C, show that there is a unique homomorphism g from into C such that g is an extension of h. (vi) If Dy is « free set of generators of , and Dz is a free set of generators of 3, and D; have the same cardinal number, then , and B, are isomorphic. (wil) Let D be a set of generators of a Boolean algebra B. Show that D is a free set of generators of @ if and only if, for any uj,...,% in B, if wED or w/ED for each %, then uy Asis An, #0. (citi) For any cardinal number m, prove that there is » Boolean algebra having a free set of generators of cardinality m. (Hint: Generalize the Boolean algebra of statement bundl (Example 3.5) by using a set of statement letters of cardinality m instead of a denumer- able set of statement letters.) (ix) When is P(K) a free Boolean algebra? nd D, QUOTIENT ALGEBRAS Baul. sa2, 5413, Let J be a proper ideal of a Boolean algebra @ and let K be a proper ideal of the quotient algebra B2=G/J. Let JO = (2: 2€B &e+JEK), Prove that JO is a proper ideal in @ and that the function y from B/JO onto B*/K defined by HerID) = (2+J+K is an isomorphism. In a Boolean algebra B, B is an ideal. What is 8/B? Why isn’t it a Boolean algebra? (For those readers acquainted with elementary point-set topology.) Let X' be a topological space. (@) Prove that the set F of subsets of ‘Q° having nowhere dense boundary is a field of sets. (®) Prove that the set N of nowhere dense sets of is an ideal of F. (©) Show that the quotient algebra F/N is isomorphic to the algebra of regular open sets of (ct. Problem 5.47). (Hint: If AGF, show that there is a unique regular set A, such thi A+A, is nowhere dense.) BiCHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 187 BOOLEAN REPRESENTATION THEOREM 5.114. Prove that every ideal of a Boolean algebra @ is principal if and only if @ is finite. 5.115. For each ¢ in a set S assume that there is an associated Boolean algebra @,. By the Cartesian product [] @, we mean the set of all functions f defined on S such that, for each s€S, f(s) €B,. We define Boolean operations on the Cartesian product in componentwise fashion, eg. if f and g are in the Cartesian product, let f A g be the function defined on S such that (fn 9)(#) = f(t) Ag, o(8) for each s€S. (a) Prove that the Cartesian product of Boolean algebr (0) Prove that a finite Boolean algebra of cardinality 2* is isomorphic to a Cartesian product of 1 copies of the Boolean algebra (0,1). (©) If the cardinal number of a set A is m, prove that the Boolean algebra (A) is isomorphic to 4 Cartesian product of m copies of {0,1). (@ Show that every Boolean algebra is isomorphic to a subalgebra of a Cartesian product of copies of (0,1). Boolean algebra, 5.416. (a) Prove that if and y are distinct elements of a distributive lattice, then there is a prime ideal containing one of = and y but not the other. (The notion of prime ideal, although originally defined for Boolean algebras, also makes sense for Inttices. Hint for the proof: Since zy, we may assume y +2. Let Z be the set of all ideals containing © but not y, and apply Zorn’s Lemma.) (8) Prove that any distributive lattice is lattice-isomorphic to a lattice of sete. (Hint: To each clement 2 of the lattice associate the set of proper prime ideals not containing z.) (©) Prove the converse of (a), i.e. a lattice is distributive if, for any two distinct elements of the lattice, there is a prime ideal containing one of the elements but not the other. S417. Although every Boolean algebra is, by Theorem 6.30, isomorphic to a field of sets, show that there are Boolean algebras B for which there is no isomorphism to a field of sets preserving infinite Joins and meets. (Hint: Consider the regular open algebra of the real line (Problem 6.47).) 5.118, [137]. Prove that a distributive lattice is a Boolean algebra if and only if every proper prime ideal is maximal. INFINITE MEETS AND JOINS 5119. Give an example of a Boolean subalgebra ( of a Boolean algebra @ such that some subset Hof C has a lub in C but not in B. 5.120, In any lattice, prove (assuming that all the indicated lub’s and gib’s exist): (0) If ey for all weW, then A te A vy and V ze OVA One = AV one: seswew wewses O A tev A te A, ov eew” ” wew wew @ Verte) = OV ter V ver wew wew wew id for all lattices. (¢) None of the inequalities in (a)-(d) can be changed into equalities v 5.2. An ideal J of a Boolean algebra is said to be complete if and only if J is closed under arbitrary joins of its elements. (a) Prove that every complete ideal is a principal ideal. (8) If J is a complete ideal of a complete Boolean algebra B, prove that the quotient algebra BIJ is complete,188 TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS [cHaP. 5 5.422, (a) Prove that the set of sublattices of a lattice is» complete lattice with respect to the inclusion Az. ) @ () Let relation C. What are the operations of join and meet? Prove that the set of subalgebras of a Boolean algebra is a complete lattice with respect to ‘the inclusion relation C. Describe the join and meet operations. Assume that g is a closure operation on a set L partislly ordered by +, i.e. g is a function from L into L such that: @) 24¥ > 92) = 0) @ 2 = ote) @) gle) ‘An element = in L is said to be g-closed if and only if g(2)= 2. Prove: () © is g-closed if and only if = = g(y) for some y. (i) If 1 fs a maximum element of L, then 1 is g-closed. ote) (ii) A glb of a set of g-closed elements is also g-closed. In particular, if @ and y are g-closed, no is = Ay (if it exists). (iv) If, =) is a complete lattice, then the set C of g-closed elements is a complete lattice with respect to the original ordering = on L, and, for any subset ¥ of L, ACv = Aby and ver = #(vE¥) ver yer sey ver = be a partial order on L. For any YCL, let ¥* = the set of upper bounds of Y and Y* = the set of lower bounds of ¥. We say that Y is a cut if and only if ¥ Prove: @ yerw @ ycr® (ii) XCY > (Wea YeCx) + (xe (iv) Xb= Xbb & Xe= Xoo () X= Xb | XH = Xe (vi) Every set X° is a cut, (vii) ‘The function g(X) = X™ is a closure operation on (P(L),C). The g-closed elements are the cuts. (will) The net C of all cuts is a complete Inttice with respect to C. Meets are intersections and vy ¥=( u_¥)". The function /, such that f(s) = (2) for any 2 in L, is an vex vex order-isomorphism of (L,+) into the complete lattice (C,C); preserves all meets and joins already existing in'(L,#), ie. if z= VY a, in L, then fle) = V fie.) in C, and similarly for meets, aes «ea (ix) Let £=(L,4) determine a Boolean algebra, {.c. it is a distributive, complemented lat- tice, Define, for any XCL, X* = (y: yaz=0}. Then @) XX = (0) = 0¢ @) (Kuxe = 0) @) (Kux = L = Ie (4) Xe = Xe © (ey = (aye (@) X** is an ideal of L. (1) ‘The funetion F such that F() jeal J of -£ is 2 Inttice-homomorphism from the distril ttice of all ideals of onto the lattice of cuts. (® Hence C forme a complete Boolean algebra, and the lattice-isomorphism f of (vill) is ‘2 Boolean isomorphism. ‘Thus every Boolean algebra is embeddable in a complete Boolean algebra in such a way that unions and meets are preserved [131]. Yo & Xbcym)CHAP. 5] TOPICS IN THE THEORY OF BOOLEAN ALGEBRAS 189 51m, In Example 5.19, prove that the collection of all singletons {x}, where 2€ K, has no lub in F (in addition to not having its union in F). m-COMPLETENESS. o-ALGEBRAS 5.125. 5.128. 512. 5.128, 5.129, Let F be the collection of all subsets A of a given set K such that the cardinality of A or the cardinality of A is = a given infinite cardinal number m. Prove that F is an m-complete field of sets, What happens if we change = to m and m

4 v 38 & 2 1 1 ‘The rank of a statement form A will be (@) 0, if A is 2 statement letter; (b) the rank of the principal connective of A, otherwise. ‘We shall describe our procedure for eliminating parentheses by induction on the number of occurrences of connectives in the statement form. (The description will appear com- Plicated, but the simplicity of the procedure will become apparent after a few examples.) Clearly, if A has no connectives, it is a statement letter and has no parentheses. Assume that we have described the procedure for all statement forms having fewer than k occurrences of connectives, and assume that A has k occurrences of connectives. Case (i): A is a denial (1B). If B itself is of the form (716), then we apply our procedure to (16) and omit the outer parentheses (if any) of the resulting expression, obtaining some expression D. The final result is taken to be (1D). If B is not of the form (116), and the application of our procedure to B yields E, then the final result is (71). Case Ais (BC), where ais +»,-+, v or & We apply the procedure for eliminating parentheses to B and €, obtaining B* and C*, At this stage we have (B*aC*). We drop the outermost pair of parentheses (if any) from B* if the rank of « is greater than or equal to the rank of B. We drop the outermost pair of parentheses (if any) from C* if the rank of a is greater than the rank of C. (II) After completion of (I), we omit the outermost pair of parentheses (if any). Examples. AL (101A), Applying (1), Case (i), twice, we obtain (171A). Then by (II) we have 1A. In general, no parentheses are needed to separate successive negations. AZ ((AvB)~0). ‘The principal connective is +, of rank 4. Since (Av B) has rank 3, we obtain, by (1), Case (ii), (AvB~0), and finally by (11), AvB~C. 190APPENDIX Al ELIMINATION OF PARENTHESES 191 AS. (AV (B>C)). The principal connective is v, of rank 3, Since (BC) has rank 4, we do not drop the parentheses from (B~C). Thus in this example (I) allows no elimination of parentheses, and by (II) we obtain Av (EO). AS. (UB 1A) © B) (1B) &C)), = is the principal connective. Application of (I) to ((Bv A) ¢>B) yields (Bv A +> B), and application of 'T to 1B. &C) yields (1B &C), Since the rank of <> is greater than that of +, we leave the outermost pair of parentheses of (Bv A <> B). However, since the rank of & is less than that of -, we omit the ‘outermost pair of parentheses from (18 &C). The final result is (BV A © B) +> 1B &C. AS WALB)VO. The right-most v is the principal connective. Since (Av B) has rank equal to that of v, we may drop she parentineses, obtaining (Av Bv C), and finally by (II), Av Bv C. Ak IAv(BVO). The left-most v is the principal connective. Since (B v C) has rank equal to that of v, we cannot drop she parentheses. Thus (I) yields no elimination of parentheses, and by (II) we obtain Av (Bv 0). The results given in Examples A.5 and A.6¢ illustrate the principle of association to the left. Thus Av BvC stands for (Av B)vC). Likewise A+B C stands for ((A > B) *C), and the same holds for the other binary connectives. Association to the left is due to our agreemen’ that in (BC), we omit outer parentheses from B if rank (B) B&B «+ DvB, « is the connective of highest rank, Hence the left side of « should be (A > B&B) and the right side should be Dv B. Thus we have (4+ B&B) + (DvB). Within (A> B&B), ~ is stronger, and we obtain (A> (B&B) o (DvB)). In ordinary arithmetic, without realizing it we already have been taught an analogous ranking of arithmetic operations. Addition is strongest, then comes multiplication, and finally exponentiation. For example, 4 + 2-5 stands for 4+ (2-5), and not for (4+2)-5, while 6 - 2 stands for 5 (2%), not for (5°2)°. Sometimes, especially in long statement forms, for the sake of clarity we can keep some parentheses which could be omitted according to our conventions. For example, in A.T above we might write Av(BvC)v (B& 171C) instead of Av(BvC)vB& 11.192 Al. ELIMINATION OF PARENTHESES [APPENDIX A Solved Problems Describe an algorithm (i.e. effective procedure) for determining whether a given expression is a statement form, and for determining the principal connective when the expression is a statement form. Solution: ‘The deseription is given by induction on the number of occurrences of connectives. If there are no connectives, then the expression is a statement form if and only if it is a statement letter. ‘Assume now that an expression A has k connectives, where k>0, and that our algorithm already has been defined for expressions with fewer than k occurrences of connectives. If A does not begin with a left parenthesis and end with a right parenthesis, then A is not a statement form. If A does have the appropriate initial left and terminal right parentheses, omit them, obtaining an expression B, Case 1. Bhas the form 16. If € is a statement form, then A is a statement form and its principal connective is 7, If € is not a statement form, then neither is A. Case 2B ia not of the form 16, For each of the binary connectives [1 in 8, if B is €D) and if © and D are statement forms, then A is also a statement form with principal connective [)- On the other hand, if the indieated condition does not hold for any of the binary connectives () in 3, then A is not a statement form, Eliminate as many parentheses as possible from: (a) (Av B) > (10) v (1B) &) (b) (A& (1(1B))) @ (Bo (CvB))) (0) (Bo (Cv B)) o (A&(1(18)) Solution: (a) The second v is the principal connective, Since v has higher rank than &, we can omit the outer parentheses from ((1B) & C). Since > has higher rank than v, we cannot drop the outer parentheses from ((A v B)->(710)). But within ((A vB)+(1C)) we can drop the outer paren- theses from (A v B), since + has higher rank than v: (AvB (10) v (1B)&C Finally, we can drop the parentheses around the denials: (AvB> 10) v 1BEC (®) © is the prineipal connective. We can omit the outer parentheses of (A & (71(71B))), since hhas higher rank than &: A&(1(1B)) & (BO (CvB) However, we cannot omit the outer parentheses of (B +> (Cv B)), since the latter has the same rank as ¢> and appears on the right hand side of the biconditional. Within (B «> (Cv B)) we an drop the parentheses of (Cv B), and, on the other side, we can drop the parentheses around the denials: A& IBS (BeCvA) (c) This is the same as (b) except that the two sides of the biconditional have been reversed. Since (B © (Cv B)) now appears on the left hand side of the biconditional, we can drop the outer parentheses: Bo (CvB) & A&(1(78) As before, we then obtain BOCVBOALNIB In practice, we would never eliminate all the parentheses in part (¢), since their complete elimination will not facilitate the interpretation of the original statement form.APPENDIX Aj ELIMINATION OF PARENTHESES 198 A3. Describe an algorithm for determining whether a given expression A has been obtained from a statement form as a result of applying our conventions for elimi- nating parentheses (and, if it has, find the statement form). Solution: ‘We shall describe, by induction on the number of symbols of A, a procedure which either will nd the statement form abbreviated by A or will eventually tell us that there is no such statement form. Clearly, if A has one symboi, then A abbreviates a statement form (A itself) if and only if ‘Ais a statement letter. Now assume that A has k symbols (where &>1) and that our algorithm has been defined for all expressions with fewer than k symbols. Case 1. A either does not begin with a left parenthesis or does not end with a right paren- thesis. (Hence if A does abbreviate a statement form F, then the outer parentheses of F were omitted.) Case 1a. A is of the form 17... 1(B). If B abbreviates a statement form o (and hence B does not have the outer parentheses of @), then A abbreviates (1(7(...(716)...))). Otherwise, A is not an abbreviation of » statement form. Case 1b. A is of the form 11...18, where B is a statement letter. Then A abbreviates CGC)... Case le. A is not of the form 17... 7(8), and not of the form 17... 1B (where 8 is a statement letter). For each occurrence of a binary connective # in A, represent A as C#D. If (i) © abbreviates a statement form 4, (i) D abbreviates a statement form J, (iii) © contains the outer parentheses of M (if any) if the rank of # that of #, (iv) D contains the outer parentheses of J (if any) if the rank of J is greater than the rank of #, then A abbreviates (M #4), If no occurrence of a binary connective in A satisfies (i)-(iv), then A does not abbreviate statement form. greater than or equal to Case 2. A is of the form (8). Case 2a. Omit the initial left and terminal right parentheses, and then apply Case 1 to 8. If B abbreviates a statement form G, then A abbreviates C also. Case 2b. If Case 2a doos not show A to be an abbreviation of a statement form, apply the procedure of Case le. If we obtain a statement form D, then A abbreviates D. Examples. (a) Be>CvB. Case te applies, First, we try ¢> as principal connective. The left side is B, and the right side Cv B, which abbreviates (Cv B). Since (Cv B) has rank less than that of ©, the outer parentheses around CvB have been legitimately omitted, Thus we obtain @'o (ceva). (©) A&1BE&A. Case tc applies. First, we consider the left-most & The left side is A and the right side is 184A, Hence we must consider 18&A. Again, Case 1c applies and we consider & ‘The left side is 1B (which is an abbreviation of (71B)) and the right side in A. Thus 1B& A abbreviates (1B) & A), but, since ((18) & A) has rank equal to that of &, clause (iv) has been violated. Next, we try the second & The left side is A & 1B, which is easily seen to abbreviate (A &(71B)). Hence the original statement form is ((A & (71B)) & A). (©) VIAVB. Case 1c applies. The left side of v is 171A, which by Case 1b abbreviates (ACA), Hence we obtain ((1(14)) vB). (@) Av Av). Case 1c applies and we look at the first v. The right side is 1(Av ©). To the latter, Case 1b applies and we see immediately that AVC is a statement form. Hence the original statement form is (Av (Av C))).194 AM Aa ELIMINATION OF PARENTHESES {APPENDIX A Supplementary Problems Find the ranks of the following statement forms. @) Ay 0) (BE ©) (BVO); @ (A>(1B)&A); (@) (A> ((1B)&AY). possible from the following. Eliminate as many parentheses (@) (1A) v (B&O) > (4 (1B) 0) (A> (IBY) v (114 &EY) (2) (A> (14) +B) (A+B) @ (14) (4>8) Determine whether each of the following expressions is an abbreviation of a statement form, and, if so, construct the statement form. (@) Av IB+(BeOOes (@) A+B+T1AVB () WAvB)v A&B (@ AvBv (C+D) Show that if A is a statement form, then there is at most one connective [] such that A is of the form (BT}C), where B and C are statement forms. Show that the algorithm of Problem A.3 is correct, {.e. when B is an abbreviation of a statement form A, then the algorithm applied to B yields A as its only answer, and when ® is not an abbrevia- tion of a statement form, then the algorithm says so.Appendix: B Parenthesis-free Notation ‘We may avoid the use of parentheses if we redefine the notion of statement form as follows: (i) every statement letter is a statement form; (i) if A and B are statement forms, so are 1A, & AB, v AB, + AB, <> AB, Examples. ‘The statement form ((1A)&(Bv A) would be rewritten as &1AVBA. The statement form (A> ((1B) © (Av) would be rewritten as +A IBV AC. This way of writing statement forms is sometimes called Polish notation (in honor of its inventor J. Lukasiewicz). Examples. ~vABv 1C& BC is Polish notation for the statement form (Av B)-+((10) v ((1)&C))). Sim- ilarly, &A11B<>Bv CB is Polish notation for ((A &(1(1B))) + (B + (Cv B))). Solved Problems B.l. Write the following statement forms in Polish notation. (@) (VA) v (B&O) > (A (1B) (0) (A> (1(1B) v (1A & CO) (©) (A> (1A) > B) > (A> B)) Solution: (@) +v TA&BCOAIB () v+ATIBIaAC () +++ A1AB+AB B2. Find the statement forms whose transcriptions into Polish notation are (0) +VATB&OBCA (0) v1VAB&AB (b) + ABV11AB (a) vv AB>CD Solution: (a) (Av (1B) + (BOO) &A)) (8) (A+B) = (AAA) vB), (0) (AVE) v (A&B) @ (Av B)v (C+D) 195196 PARENTHESIS-FREE NOTATION [APPENDIX B B.3. Describe an algorithm for determining whether a given expression is a statement form in Polish notation and for constructing the corresponding statement form in its original notation. Solution: Assign the integer —1 to statement letters, 0 to 1, and +1 to the binary connectives &,v,-+, <>, ‘Then we have Theorem: An expression A in a statement form in Polish notation if and only if (1) the sum #(A) of the integers assigned to all the occurrences of symbols in A is (II) the sum #(8) of the integers assigned to all the occurrences of symbols in every proper initial segment 8" of A is non-negative. Example. vv AB1B&1BC Proof of the Theorem. First, we shall prove that any statement form in Polish notation satisfies ‘conditions (I) and (II), This is shown by induction on the number k of occurrences of connectives in A. If there are no connectives, A is a statement letter and conditions (1)-(II) are obvious. (In this case there are no proper initial segments of A.) Now assume that the result has been established for all statement forms having fewer than k occurrences of connectives (1). By our new definition of statement form, A is of one of the forms 18, &BC, v.BC, > BC, <>6C, where B and C are statement forms (having fewer than k occurrences of connectives). Use of the induc- tive hypothesis now yields (1)-(II). (For instance, if A is &BC, then #(a) = 1+ #(8) + #(C) 1+(-D+CD= jelding (I). If D is I segment of A, then either D is & and ‘#(0) is +10; or Dis &D, (where D, is 1+ #0) = 140=1>0; or Dis &Band #0) =1+ #(8) & BC, (where C, is a proper initial segment of C) and #(D) = 1+ #(B) + #(€,) = 14 (-1) + #(C,) = 0+ #(C,) = #(C,) = 0.) Conversely, let us assume now that an expression A satisfies (I)-(II). We prove that A is a statement form in Polish notation by induction on the number k of symbols in A. k= 1: then by (1), A is a statement letter, Induction step: assume that k>1 and that the result holds for all expressions having fewer than k symbols. Case 1: A is of the form 78, It is then easy to show that the truth of (I)-(II) for A implies the truth of (1)-(II) for 8, and hence, by inductive hypothesis, B is a statement form. Therefore 18 is a statement form. Case 2: A is of the form BC, where B is a statement letter. ince 8 is a proper initial segment of A. Case 8: A is of the form [1C, where (1 is one of the binary connectives &, v,~-, «>. There must be a shortest proper initial segment 8 of © such that #(8)=—1. For, as we move from left to right in A, the sum of the symbols begins at +1 (the integer for []) and ends with ~1 (= #(A)), and moving from one symbol to the next either leaves the sum unchanged or changes it by +1 or —1, ‘Hence we must finally arrive at the first proper initial segment of A whose sum is 0. This proper initial segment is of the form (1B, where 8 is the shortest proper initial segment of C such that #(8)=—1. Then 8 satisfies (I). As for (11), consider any proper initial segment D of B, Then ina proper initial segment of A if and only if A is of the form BC, where C contains at least one symbol.APPENDIX B] PARENTHESIS-FREE NOTATION 197 Ba Bs. since (is a shorter proper initial segment of A than (8, #(C1D) is > 0, and #(D) = 0. Hence (11) holds for B, and, by inductive hypothes a statement form in Polish notation. Let © be BE. Thus Ais OBE. Since —1= #(A) 1+ (-1) + #(€) = #(6), E satisfies (1) ‘As for (II), let F be any proper initial segment of £. Then UBF is a proper initial segment of A. By (1) for A, 0-< #(C) BF) = 1+ #(B) + #(F) = 14 (—1)+ #(F) = #(F). Thus (I) holds for E, and, by inductive hypothesis, £ is a statement form in Polish notation, and therefore so is (1 BE. Notice that the second part of the proof of the theorem gives a method for constructing the corresponding statement form in our original notation (since it locates the statement forms out of which our given statement form is constructed). Example. &A71B@>BVCB ‘The first proper initial segment whose sum is 0 is 4A 17B, Thus we have (& A178) ¢ (BV CB). In &A7B, the first proper initial segment whose sum is 0 is &A. Hence we obtain A& 118. In > BY CB, the first proper initial segment whose sum is 0 is 2B, and we obtain Be (vCB). Thus, 20 far, (4& 178) + (B<>(vCB). Finally, vCB corresponds to CB, and the statement form in our original notation is (A & 11 B) « (B (Cv B)). Supplementary Problems Write the following statement forms in Polish notation, (@) B& WAV) (@) (A+B) + (BA) & (AB) (©) [(A+B)>(C>D)) + [E+{(D +A) >(C>A))) rmine whether each of the following expressions is a statement form in Polish notation, and, if it is, find the corresponding statement form in our original notation. (@) + 1vABeOCa AB (@) 1&A+B>A (©) vv>+AB AA BBA (d) +A>BC+& ABCAppendix: C The Axiom of Choice Implies Zorn’s Lemma By the axiom of choice we mean the assertion that, for any set x, there is a function f (called choice function for z), defined on P(z)~ (QD), such that, for any non-empty subset uof 2, f(u) eu. We shall say that a collection A of sets is well-ordered by inclusion if and only if A is an C-chain and every non-empty subset C of A has a least element 0 (ie. if u€C, then Cu). Given a collection A well-ordered by inclusion, and given any set y in A, the segment determined by y (denoted Seg (A, y)) is defined to be the set of all z in A such that zCy. Notice that, if Sis a section of A (i.e. if S is a subset of A such that (y€S & zCy) + z€S)) and if S*A, then S=Seg(A,u), where u is the least element of A~S. Theorem. The axiom of choice implies Zorn’s Lemma. Proof. Assume that a set Z of sets has the property that, for every C-chain C in Z, the union U_ A is also in Z. Let F be a choice function for Z. Thus if 9» DCZ, then FD)ED. Let us assume that Z has no C-maximal element. We shall derive a contradiction from this assumption. For any y in Z, the set ¥ of all elements z of Z such that yCz is non-empty (since there are no C-maximal elements). Let f(y) =F(¥). Thus for any y in Z, f(y) is an element of Z such that yC f(y). By a ladder we mean any subset L of Z such that L is well-ordered by inclusion, and, for any EL, f(_.U _u) =. (By hypothesis, we know that Uw EZ, since esas west Seg (L, z) is an C-chain.) Let L be the set of all ladders. (1) Given two different ladders L; and Lz, we shall show that one of them is a segment of the other. Let K be the set of all uGZi0L2 such that Seg (L1,u) = Seg (Lz,u). Clearly, K is a section of both Ly and L;. Hence if K=L, or K=Ls, then one of L; or Ly is a segment of the other. Thus we must show that KC, and KC Lz do not simultaneously hold. To this end, assume KC Z, and KC Iz. Let u; be the least element of L1~K, and let tw be the least element of Za~K. Then Seg(L,u;)=K and Seg (L2,u)=K. By definition of ladder, (ve) = 1 cotoom™) and m= a) ~ 1(.ee*) 198APPENDIX C] THE AXIOM OF CHOICE IMPLIES ZORN’S LEMMA 199 Hence w= uz and Seg (Z1,%) = K = Seg(Ls,u.). Thus u:€K, which contradicts the fact that u €L.~K. (2) The union of all ladders H =U), Zn is again a ladder. For, by (1) it is clear that H isan C-chain. H is well-ordered by inclusion. (In fact, if @*WCH and ué W, then wEL, for some L, €L and the least element of WNL; must be the least element of W.) Finally, for any z€H, EL, for some L,€L, and, by (1), Seg (H,z) = Seg (L:,2). Henee = = A 00) ot) = 1 Unt): 6 Seats 2) Since H is an C-chain, the union v= uu € Z, by hypothesis. Hence HU (f(v)} is a ladder, and therefore f(v) €H. “e It follows that f(v) Cv, contradicting vC f(v).>e. Proof of the Schroder-Bernstein Theorem Lemma. Let (Z,<) be a complete lattice and let ¢ be a function from L into L such that ¢ is order-preserving, ie. 2 4(z)=¢(y). Then ¢ has a fixed point b in L, ie. 9(0) =. Proof. Let W = (a: z€L&z: From b= 4(0) & 6(0) [Bs] CflZs] + ¥~ fds) CY ~ 12s) ~» of¥ ~ f[Zs]] C o[¥ ~ £[Z1]] > X~ol¥ ~ MZ} CX ~ oY ~ £125] #2) 23) Hence ¢ is an order-preserving function from the complete lattice (P(X), C) into itself. Hence by the lemma above, ¢ has a fixed point Z*, i.e. = o(2*) = X~ ol ~ F(Z") Therefore ol¥~f[2*]] = X~2* It is easy now to verify, using Fig. D-2 as a guide, that the following function h, _ [fe itzez* Me) = {ie if 2eX~Z* is a one-one correspondence between X and Y.> 200Bibliography The Bibliography is divided into two parts: the first on switching circuits, logic circuits and minimization, the second on Boolean algebras and related areas. L L 2 10. nL 12, 18. 14, 15. SWITCHING CIRCUITS, LOGIC CIRCUITS AND MINIMIZATION Adam, A.: Truth Functions and the Problem of their Realization by Two-terminal Graphs, Akadémiai Kiadé, Budapest, 1968. Adefio, S. A., and C. F. Nolan: Principles and Applications of Boolean Algebras for Electronic Engineers, Wiffe Books, London, 1966. Akers, S. B,, Jr: A Truth Table Method for the Synthesis of Combinational Logie Circuits, IRETEC,* EC-10, 604-615, 1961. Ashenhurst, R. L.: The Theory of Abstract Two-terminal Switching Networks, Harvard Comp. Lab. BL-5, VII, 1-15, 1954, Ashenhurst, R. L.: The Decomposition of Switching Functions, Proc. Int. Symp. ‘Theory of Switching, 1957, Annals, Harvard Comp. Lab., 1959. Beatson, T. J.: Minimization of Components in Electronic Switching Circuits, AIEE, I, Comm. and Electr., 77, 288-291, 1958, Booth, T. M. The Vertex-Frame Method for Obtaining Minimal Propositional-Letter Formulas, IRETEC, Vol. EC-11, 144-154, 1962. Calabi, L.: A Solution of the Minimization Problem for Boolean Formulas, Parke Math. Lab. Rept. 7-8471, 1960. Calabi, L.: Relations between Switching Networks and Boolean Formulas, Parke Math. Lab, Rept, 8-2471, 1961. Calabi, L., and J. A. Riley: Inessentiality in Minimal Networks and Formulas, IRETEC, EC-11, 5, 711-713, 1962. Calabi, L., and J. A. Riley: The Algebra of Boolean Formulas: Some Criteria for Minimality, Proc. Third Ann. Symp. Switching Circuit Theory & Logical Design, S-141, 33-47, 1962, Calabi, L., and E. W. Samson: On The Theory of Boolean Formulas: Minimal Including Sums, I, J. Soc. Industr. Appl. Math., 11, 212-284, 1963. Caldwell, S.: Switching Circuits and Logical Design, Wiley, New York, 1958. Chang, D. M. ¥., and T. H. Mott, Jr.:_ Computing Irredundant Normal Forms from Abbreviated Presence Functions, IEEETEC, EC-14, 885-842, 1965. Choudhury, A. K., and M.S. Basu: A Mechanized Chart for Simplification of Switch- ing Functions, IRETEC, EC-11, 713-714, 1962. TIRE stands for Institute of Radio Engineers, recently changed to Institute of Electrical and Electronics Engineers (IEEE). TEC stands for Transactions on Electronic Computers. 201202 16. 17. 18. 19. 20. 21. 22. 23, 24. 25. 26. 27. 380. 31. 82. 33. 35. 87. BIBLIOGRAPEY Chu, J. T.: A Generalization of a Theorem of Quine for Simplifying Truth Functions, IRETEC, EC-10, 165-168, 1961. Cobham, A. R., R. Fridshal and J. North: Switching Circuit Theory & Logical Design, Proc, 2nd Ann. Symp. Switching Theory & Logical Design, Detroit, AIEE, New York, 1961. Curtis, H. A.: A New Approach to the Design of Switching Circuits, Van Nostrand, Princeton, 1962. Das, S. R., and A. K. Choudhury: Maxterm Type Expressions of Switching Functions and their Prime Implication, IEEETEC, EC-14, 920-923, 1965. Dietmeyer, D. L., and J. R. Duley: Generating Prime Implicants via Ternary Encoding and Decimal Arithmetic, Comm. Assoc. Comp. Mach., 11, 520-528, 1968. Dunham, B., and R. Fridshal: The Problem of Simplifying Logical Expressions, J. Symb. Logie, 24, 17-19, 1959. Flegg, H. G.: Boolean Algebra and its Applications, Wiley, New York, 1964. Logic of Computer Arithmetic, Prentice-Hall, Englewood Cliffs, N. Flores, 1963, Fridshal, R.: The Quine Algorithm, Summaries of Talks, Summer Inst. Symb. Logic, Cornell, 211-212, 1957. Ford, L. R., and D. R. Fulkerson: Flows in Networks, Princeton Univ., Princeton, 1962. Ghazala, M. J.: Irredundant Disjunctive and Conjunctive Normal Forms of a Boolean Funetion, JBM J. Res. and Dev., 1, 171-176, 1957. Gimpel, J. F.: A Reduction Technique for Prime Implicant Tables, Proc. Fifth Ann. Symp. Switching Circuit Theory & Logical Design, S-164, IEEE, 174-182, 1964. Gimpel, J. F.: A Method of Producing a Boolean Function Having an Arbitrarily Prescribed Prime Implicant Table, IEEETEC, EC-14, 3, 485-488, 1965. Gimpel, J. F.: A Reduction Technique for Prime Implicant Tables, IEEETEC, EC-14, 4, 585-541, 1965, Hall, F, B.: Boolean Prime Implicants by the Binary Sieve Method, AIEE Trans. I, Comm. and Electr., 80, 709-713, 1962. Hammer, P., and S, Rudeanu; Boolean Methods in Operations Research, Springer Verlag, New York, 1968. Harris, B.: An Algorithm for Determining Minimal Representations of a Logic Func- tion, IRETEC, EC-6, 2, 103-108, 1957. Harrison, M. A.: Introduction to Switching and Automata Theory, McGraw-Hill, New York, 1965. Hill, F. J., and G. R. Peterson: Introduction to Switching Theory and Logical Design, Wiley, New York, 1968. Hockney, R.: An Intersection Algorithm Giving All Irredundant Forms from a Prime Implicant List, IEEETEC, EC-11, 2, 289-290, 1962. Hoernes, G., and M. Heilweil: Introduction to Boolean Algebra and Logic Design, McGraw-Hill, New York, 1964. Hohn, F. E.: Applied Boolean Algebra, Macmillan, New York, 1960.6 8 t 6 & 56. 87. 58. BIBLIOGRAPHY 203 Hohn, F. E., and L. R. Schissler: Boolean Matrices in the Design of Combinational Switching Circuits, Bell System Tech. J., 84, 177-202, 1955. Holst, P. A: Bibliography on Switching Circuits & Logical Algebra, IRETEC, EC-10, 638-661, 1961. House, R. W., and T. Rado: A Generalization of Nelson’s Algorithm for Obtaining Prime Implicants, J. Symb. Logic, 80, 8-12, 1965. Hu, S.T.: Mathematical Theory of Switching Circuits & Automata, Univ. of Califor- nia Press, Berkeley, 1968. Karnaugh, M.: The Map Method for Synthesis of Combinational Logic Circuits, Trans. AIEE, I, 72, 593-599, 1953, Karp, R. M.: Functional Decomposition & Switching Circuit Design, J. Soc. Ind. Appl. Math., 291-335, 1963. Kautz, W. H.: A Survey & Assessment of Progress in Switching Theory & Logical Design in the Soviet Union, IEEETEC, EC-15, 164-204, 1966. Keister, W. A., A. Ritchie and S. Washburn: The Design of Switching Circuits, Van Nostrand, Princeton, 1951. Lawler, E. L.: An Approach to Multi-Level Boolean Minimization, J. Assoc. Comp. Mach., 11, 283-295, 1964. Lawler, E. L., and G. A. Salton: The Use of Parenthesis-Free Notation for the Auto- matic Design of Switching Circuits, IRETEC, EC-9, 342-852, 1960. Luceio, F.: A Method for the Selection of Prime Implicants, IEEETEC, EC-15, 2, 205-212, 1966. Lupanov, 0. B.: On the Synthesis of Contact Networks, Doklady Akad. Nauk, USSR, 119, 1, 23-26, 1958 (Russian). Lupanov, 0. B.: On Asymptotic Estimates of Complexities of Formulas which Realize Functions of the Algebra of Logic, Ibid. 128, 8, 464-467, 1959 (Russian). Lupanoy, 0. B.: On Realization of Functions of the Algebra of Logic Using Formulas of Limited Depth in the Basis of & v, and~, Ibid., 186, 5, 1041-1042, 1961 (Russian). Maley, G. A.: Simplifying Switching Circuits with Boolean Algebra, Electrotechnology, 67, 101-106, 1961. Marcus, M.: Switching Circuits for Engineers, Prentice-Hall, Englewood Cli 1962, McCluskey, E. J.: Minimization of Boolean Functions, Bell System Tech. J., 35, 1417-1444, 1956, McCluskey, E. J.: Minimal Sums for Boolean Functions Having Many Unspecified Fundamental Products, Proc. Second Ann. Symp. Switching Circuit Theory & Logical Design, AIEE, 10-17, 1961. McCluskey, E. J.: Introduction to the Theory of Switching Circuits, McGraw-Hill, New York, 1965. McCluskey, B. J., and T. C. Bartee: A Survey of Switching Cir Hill, New York, 1962. Meo, A. R.: On the Determination of the ps Maximal Implicants of a Switching Func- tion, IEEETEC, EC-14, 6, 830-840, 1965. NJ, it Theory, McGraw-59. 60. 61. 62. 65. 66. 67. 68. 69. 70. m1. 72. 73. 14. 7. ‘76. 77. 78. 79. 80. 81. BIBLIOGRAPHY Mileto, F., and G. Putzolu: Average Values of Quantities Appearing in Boolean Func- tion Minimization, IEEETEC, EC-13, 2, 87-92, 1964. Mott, T. H., and C. C. Carroll: Numerical Procedures for Boolean Function Minimiza- tion, IEEETEC, EC-18, 4, 470, 1964. Mott, T. H.: Determination of the Irredundant Normal Forms of a Truth Function by Iterated Consensus of the Prime Implicants, IRETEC, EC-9, 2, 245-252, 1960. Moortgat, M. J.: Simplification of Boolean Functions, Gen, Electric Co. Tech. Informa- tion Series R64CD18, Falls Church, Va., 1964. Muller, D. E.: Application of Boolean Algebra to Switching Circuit Design and to Error Detection, IRETEC, EC-8, 8, 6-12, 1954. Muller, D. E.: Complexity in Electronic Switching Circuits, IRETEC, EC-5, 1, 15-19, 1956. Nelson, R. J.: Weak Simplest Normal Truth Functions, J. Symbolic Logie, 20, 8, 282-284, 1955. Nelson, R. J.: Simplest Normal Truth Functions, Ibid., 20, 2, 105-108, 1955. Ninomiya, I- A Study of the Structures of Boolean Functions and Applications to of Switching Circuits, Mem. Faculty Eng. Nagoya Univ., 18, 2, 149-363, Peterson, W. W.: Error-correcting Codes, Wiley, New York, 1961. Petrick, S. R.: A Direct Determination of the Irredundant Forms of a Boolean Fune- tion from the Set of Prime Implicants, Air Force Cambridge Res. Center Tech. Report 56-110, April 1956. Phister, M.: Logical Design of Digital Computers, Wiley, New York, 1958. Prather, R. E.: Introduction to Switching Theory: A Mathematical Approach, Allyn and Bacon, Boston, 1967. Pyne, I. B, and EB. J. McCluskey: An Essay on Prime Implicant Tables, J. Soc. Industr. Appl. Math., 9, 604-681, 1961. Pine, I. B, and E. J. McCluskey: The Reduction of Redundancy in Solving Prime Implicant Tables, IRETEC, EC-11, 4, 478-482, 1962. Quine, W. V.: The Problem of Simplifying Truth Functions, Amer. Math. Monthly, 59, 521-581, 1952. Quine, W. V.: A Way to Simplify Truth Functions, Ibid., 62, 627-681, 1955. Quine, W. V.: On Cores and Prime Implicants of Truth Functions, Ibid., 66, 755-760, 1959. Quine, W. V.: Two Theorems about Truth Functions, Bol. Soc. Math. Mexicana, 10, 64-70, 1953. Richards, R. K.: Arithmetic Operations in Digital Computers, Van Nostrand, Princeton, 1955. Roth, J. P.: Algebraic Topological Methods for the Synthesis of Switching Systems, I, Trans. Amer. Math. Soc., 88, 301-326, 1958. Roth, J. P.: Algebraic Topological Methods in Synthesis, Proc. Int. Symp. on Theory of Switching, Harvard Univ., 1959. Rudeanu, S.: Axiomatization of Certain Problems of Minimization, Studia Logica, 20, 37-61, 1967.82. R 87. 89. 90. 91. 92. 97. BIBLIOGRAPHY 205 Samson, E. W., and B. E. Mills: Circuit Minimization: Algebra and Algorithm for New Boolean Canonical Expressions, Air Force Cambridge Res. Center Tech. Report, 54-21, 1954, Samson, E. W., and R. Mueller: Circuit Minimization: Minimal and Irredundant Boolean Sums by Alternative Set Method, Air Force Cambridge Res. Center Tech. Report, 55-109, 1955. Scheinman, A. H.: A Method for Simplifying Boolean Funetions, Bell System Tech. J., 41, 4, 1887-1346, 1962, Schubert, E. Ju: The Relative Merits of Minimization Techniques for Switching Cireuits, IEEETEC, EC-12, 321-822, 1968, Seshu, S., and M. B. Reed: Linear Graphs and Electrical Networks, Addison-Wesley, Reading, Mass., 1961. Semon, W.: Synthesis of Series-parallel Network Switching Functions, Bell System Tech. J., 87, 4, 877-898, 1958, Semon, W.: Matrix Methods in the Theory of Switching, Proc. Int. Symp. Switching Theory, Harvard, 1959, Shannon, C, E.: A Symbolic Analysis of Relay and Switching Circuits, Trans. Amer. Inst. Electr. Eng., 57, 713-728, 1938, Shannon, C. E.: The Synthesis of Two-terminal Switching Circuits, Bell System Teck. J., 28, 59-98, 1949. Torng, H. C.: Introduction to the Logical Design of Switching Systems, Addison- Wesley, Reading, Mass., 1964. Vasil’ev, Yu. L.: Minimal Contact Networks for Boolean Functions of Four Variables, Doklady Akad. Nauk, USSR, 127, 2, 2420245, 1959 (Russian), Veitch, E. W.: A Chart Method for Simplifying Truth Functions, Proc. Assoc. Comp. Mach., 127-133, 1952, Washburn, S. H.: An Application of Boolean Algebra to the Design of Electronic Switching Circuits, AIEE, I, Comm. & Electr, 72, 880-388, 1953. Yablonskii, S. V.: The Algorithmic Difficulty of Synthesizing Minimal Switching Circuits, Problems of Cybernetics, 2, 401-457, 1961 (Translated from Russian). Zacharov, B.: Digital Systems Logic and Circuits, American Elsevier, New York, 1968. Zuraviev, Yu. I: On Various Concepts of Minimality for Disjunctive Normal Forms, Siberian Math. J., 1, 609-610, 1960 (Russian), II. BOOLEAN ALGEBRA AND RELATED TOPICS (Sikorski [1958] is the best and most authoritative text. Halmos [1963] is very stimulating.) 98. 100. 101. Birkhoff, G.: Lattice Theory, American Math Society Colloquium Publications XXV, New York, 1940, Second Edition, 1948, 1961. Birkhoff, G., and G. D. Birkhoff: Distributive Postulates for Systems like Boolean Algebras, Trans. Amer. Math, Soc., 60, 3-11, 1946, Birkhoff, G., and O. Frink: Representation of Lattices by Sets, Trans. Amer. Math. Soe. 64, 299-816, 1948, Byrne, L.: Two Brief Formulations of Boolean Algebra, Bull. Amer. Math. Soc., 52, 269-272, 1946,206 102, 108. 104, 105. 106. 107. 108. 109. 110. 11. 112. 13. 114. 15. 116. 117. 18. 119. 120. 121. 122. 123, 124, 125, 126. BIBLIOGRAPHY Byrne, L.: Short Formulations of Boolean Algebra Using Ring Operations, Canadian J. Math., 8, 81-83, 1951. Chang, C. C.: On the Representation of a-complete Boolean Algebras, Trans. Amer. Math. Soc., 85, 208-218, 1957. Chang, C. C.: Algebraic Analysis of Many-valued Logics, Trans. Amer. Math. Soc., 88, 467-490, 1958. Cohn, P. M.: Universal Algebra, Harper & Row, New York, 1965. Curry, H. B.: Foundations of Mathematical Logic, McGraw-Hill, New York, 1963. Dilworth, R. P.: Lattices with Unique Complements, Trans. Amer. Math. Soc., 57, 128-154, 1954, Dilworth, R. P.: Structure and Decomposition Theory of Lattices, Proc. Symp. Pure Math., Lattice Theory, 3-16, 1961. Dunford, N., and M, H. Stone: On the Representation Theorem for Boolean Algebras, Rev. Ci. Lima, 43, 748-749, 1941. Dwinger, P.: Introduction to Boolean Algebra, Wirzburg, 1961. Enomoto, S.: Boolean Algebras and Fields of Sets, Osaka Math. J., 5, 99-115, 1953. Frink, 0.: Representations of Boolean Algebras, Bull. Amer. Math. Soc., 47, 755-756, 1941. Gardner, M.: Logic Machines and Diajbams, McGraw-Hill, New York, 1958. Goodstein, R. L.: Boolean Algebra, Pergamon, Oxford, 1963. Halmos, P. R.: Algebraic Logic, Chelsea, New York, 1962. Halmos, P. R.: Lectures on Boolean Algebras, Van Nostrand, 1968. Henkin, L.: La Structure Algebrique des Theories Mathematiques, Gauthier-Villars, Paris, 1956. Henkin, L., and A. Tarski: Cylindrie Algebras, Proc. Symp. Pure Math. II, Lattice Theory, 83-133, 1961. Hermes, H.: Einfihrung in die Verbandstheorie, Springer Verlag, Berlin, 1955. Horn, A., and A. Tarski: Measures in Boolean Algebras, Trans. Amer. Math. Soc., 64, 467-497, 1948, ‘Huntington, E. V.: Sets of Independent Postulates for the Algebra of Logic, Trans. Amer. Math. Soc., 5, 288-309 (1904), Huntington, E. V.: New Sets of Independent Postulates for the Algebra of Logic, Trans, Amer. Math. Soc., 35, 274-804, 557-558, 971, 1983. Jonsson, B.: A Boolean Algebra without Proper Automorphisms, Proc. Amer. Math. Soc., 2, 766-770, 1951. Jonsson, B., and A. Tarski: Boolean Algebras with Operators, Amer. J. of Math., (I) 78, 891-989, 1951; (II) 74, 127-162, 1952. Karp, C. R.: A Note on the Representation of a-Complete Boolean Algebras, Proc. Amer. Math. Soc., 14, 705-707, 1963. Kolmogorov, A. N.: Foundations of the Theory of Probability, Chelsea, New York, 1950.iz. 128. 129. 130, 331. 182. 139. 140. 141. 142. 143, 144. 145, 146. 147. 148. 149, 150. BIBLIOGRAPHY 207 Kripke, S.: An Extension of a Theorem of Gaifman-Hales-Solovay, Fund. Math., 61, 29-32, 1967. Kuratowski, K., and A. Mostowski: Set Theory, North Holland, Amsterdam, 1968. LeVeque, W. G.: Topics in Number Theory, Vol. I, Addison-Wesley, Reading, Mass., 1956. Loomis, L. H.: On the Representation of o-complete Boolean Algebras, Bull. Amer. Math. Soc., 63, 757-760, 1947. MacNeille, H.: Partially Ordered Sets, Trans. Amer. Math, Soc., 42, 416-460, 1937. McKinsey, J. C. C., and A. Tarski: The Algebra of Topology, Ann. of Math., 45, 141-191, 1944. McKinsey, J. C. C., and A. Tarski: On Closed Elements in Closure Algebras, Ann. of Math., 47, 122-162, 1946. McKinsey, J. C. C., and A. Tarski: Some Theorems about the Sentential Calculi of Lewis and Heyting, J. Symb. Logic, 18, 1-15, 1948. Mendelson, E.: Introduction to Mathematical Logic, Van Nostrand, Princeton 1964. Mostowski, A.: AbzihIbare Boolesche Kérper und ihre Anwendungen auf die allge- meine Metamathematik, Fund. Math., 29, 34-53, 1937. Nachbin, L.: Une propriété caractéristique des algebres booleiennes, Portugal Math., 6, 115-118, 1947. Neumann, J., von, and M. H. Stone: The Determination of Representative Elements in the Residual Classes of a Boolean Algebra, Fund. Math., 25, 353-376, 1935. Newman, M. H. A: A Characterization of Boolean Lattices and Rings, J. London Math. Soc., 16, 256-272, 1941, Pierce, R. S.: Distributivity in Boolean Algebras, Pacific J. Math., 7, 983-992, 1957. Pierce, R. S.: Introduction to the Theory of Abstract Algebras, Holt, New York, 1968. Rasiowa, H., and R. Sikorski: The Mathematics of Metamathematics, Polish Academy of Sciences, Monografie Matematyczne, Warsaw, 1963. Rieger, L.: Algebraic Methods of Mathematical Logic, Academic Press, New York, 1967. Rosenbloom, P.: The Elements of Mathematical Logic, Dover, New York, 1950. Rudeanu, S.; Boolean Equations and their Applications to the Study of Bridge Circuits, I, Bull. Math. Soc. Math. Phys. R.P.R., 3, 447-478, 1959. Scott, D.: The Independence of Certain Distributive Laws in Boolean Algebras, Trans, Amer. Math. Soc., 84, 258-261, 1957. Sikorski, R.: On the Representation of Boolean Algebras as Fields of Sets, Fund. Math., 35, 247-258, 1948. Sikorski, R.: Boolean Algebras, Springer Verlag, Berlin, 1958; 2nd Edition, 1964. Smith, E. C., and A. Tarski: Higher Degrees of Distributivity and Completeness in Boolean Algebras, Trans. Amer. Math. Soc., 84, 280-257, 1957. Stabler, E. R.: Boolean Representation Theory, Amer. Math. Monthly, 51, 129-132, 1944,208 11. 152, 153. 154. 155. 156. 187, 158. BIBLIOGRAPHY Stone, M. H.: The Theory of Representations for Boolean Algebras, Trans. Amer. Math. Soc., 40, 87-111, 1936. Stone, M. H.: Applications of the Theory of Boolean Rings to General Topology, Trans, Amer. Math. Soc., 41, 321-364, 1937. Stone, M. 807-816, 1938. Szasz, G.: Introduction to Lattice Theory, Academic Press, New York, 1963. The Representation of Boolean Algebras, Bull. Amer. Math. Soc., 44, ‘Tarski, A.: Zur Grundlegung der Boole’schen Algebra I, Fund. Math. 24, 177-198, 1936. Tarski, A: Der Aussagenkalkiil und die Topologie, Fund. Math. 81, 103-184, 1938. Tarski, A: Cardinal Algebras, Oxford Univ. Press, New York, 1949. Tarski, A.: A Lattice-Theoretical Fixed-Point Theorem and its Applications, Pacific J. Math., 5, 285-810, 1955.