TABLE OF CONTENTS Preface to Fifth Edition PART I Teaching Tips 1.

Preparing to Teach an Applied Combinatorics Course for the First Time. . . 2. General Teaching Suggestions . . . . . . . . . . . . . . . . . . 3. Goals of this Text . . . . . . . . . . . . . . . . . . . . . . . 4. Student Background . . . . . . . . . . . . . . . . . . . . . . 5. Note Taking . . . . . . . . . . . . . . . . . . . . . . . . . 6. Homework Assignments and Reviewing for Tests . . . . . . . . . . . PART II Section-by-Section Commentary on Text Prelude . Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Chapter 6 6 Chapter 7 8 Chapter 8 10 Chapter 9 12 Chapter 10 13 Chapter 11 Appendices . . . . . . . 24 . . . . 16 18 19 . 21 . . . . . . . 22 . . . . . . . 23 . . . . . . . . . . . . . . . . . . . . . . . 1 2 2 3 4 4 5

PART III Sample Syllabi 1. Sample One-Semester Graph Theory/Counting Course . . . . . . . . . 26 2. Sample One-Semester Course emphasizing Graph Theory . . . . . . . . . 26 3. Sample One-Semester Course emphasizing Counting . . . . . . . . . 27 PART IV Solutions to Exercises Prelude . Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 . . 28 . . 37 . . 48 . . 55 . . 61 Appendices . . . . . . Chapter 6 Chapter 7 Chapter 8 Chapter 9 Chapter 10 Chapter 11 . 90 . . . . 69 74 78 82 . . . . . . . 86 . . . . . . . 88 . . . . . . . . . . . . . . . . . . . . . . . .

PREFACE TO FOURTH EDITION The fifth edition of this text has a number of changes from the fourth edition. A new chapter on computer science connections to enumeration has been added (much of its contents appeared in the similar chapter in the second edition that was dropped in subsequent editions). There is a new ‘Prelude’ at the beginning of the text on the game of Mastermind, which serves as a friendly warm-up to combinatorial logic. There is a new section in the Network Algorithms chapter on the Transportation Problem. A new appendix on computational complexity was added. Finally, two pages about chromatic polynomials were added to the section on graph coloring. Dozens of additional exercises were incorporated. The numbers were changed in dozens of other easy counting exercises to reduce the use by students of solutions passed down from previous years. However, only minor changes have been made to the rest of the text. A webpage with corrections to the fifth edition will be maintained at the author’s website, which can be found by typing ‘Alan Tucker’ into a search engine such as Google. Instructors using this text are encouraged to e-mail the author at tucker@ams. stonybrook.edu, with corrections or comments. Thank you.

1

Teaching Tips

Part I Teaching Tips
1. Preparing to teach an applied combinatorics course for the first time. The good news first. There is no challenging new material an instructor needs to learn to teach an undergraduate applied combinatorics course. Most chapters in this book are built around a few simple formulas or principles. This is a problem-solving text, not a theory text. Advance preparation consists primarily of working out for oneself problems in the book. The principal attribute required of an instructor for this course is a disciplined, logical mind. Now the bad news. There is much that occurs in the classroom in this course for which an instructor cannot prepare. The critical tasks are: a) determining the reason underlying a student's wrong answer or misconception about a counting principle; and b) being able to spot alternate forms of an answer--expressions which look very different from the standard answer but still are correct. Some exercises have 4 or 5 equally correct analyses yielding different looking expressions. The combination of "debugging!' mistakes and unexpected forms of an answer arise in combination when a student uses an original analysis but makes mistakes along the way. In this course, the job of the instructor is not so much to present a right analysis of problems as to help students learn how to devise a right analysis for themselves. This means, answering student questions (especially about homework problems) should be a major part of the class time. Trying to answer all these questions promptly in class is clearly fraught with danger. The best way, time permitting, to answer a "why can't you do it this way" question is for the instructor to work slowly through the problem, starting from first principles and using a small concrete example, and follow the student's suggested line of attack until a fault is found or a valid, correct answer is obtained. The instructor should encourage other students to participate in the analysis. Sometimes the instructor may get stuck and should quickly say that he/she needs time to think about the question in order to give a clearer answer. Don't try to hide the fact that the material can be tricky for the instructor, too. One of the best ways to prepare for unforeseen questions about homework is to try to solve exercises in advance many different ways. An even better tactic is to encourage students to come to the instructor's office to ask some of their questions about homework before it is discussed in class. The same questions inevitably come up again in class, and now the instructor is all set with pretested, clear answers.

2. General teaching suggestions
This text is designed to be a resource for an enjoyable teaching experience for the instructor in combinatorial problem-solving—which in turn usually guarantees an enjoyable learning experience for the students. Happily, most mathematicians enjoy solving combinatorial problems. An applied combinatorics course using this text is ideally suited to a style of interactive mathematics teaching that gives students a more 2

recurrence relations. many exercises. Several of these topics have special pedagogical value. consciously and unconsciously.Teaching Tips active and. in recurrence relation models. No matter how important the mathematical methods. hopefully. They should foster student interest in the subject with their enthusiasm for teaching the material and with their encouraging answers to student's questions (even silly questions). logical analysis of "word" problems. typically a problem-solving point of view. and the inclusion-exclusion formula. The most important technique for a combinatorial problem-solver is simply disciplined. a problem-solving approach is desirable (similar to the standard problem-solving approach in calculus). First. and in probability and statistics. or learning more about them. in generating function models. in operations research. The way for a student to learn this logical thinking is by working many. To some. all applied mathematics courses should motivate topics with applications and develop concepts intuitively at first. students often avoid using them. Polya's formula is a practical (and fairly painless) introduction to group theory. the goal of developing such logical reasoning in a problem-solving framework should be foremost in the instructor's mind at all times. A secondary goal of this text is to introduce students to the basic concepts and tools of enumeration and graph theory: counting methods such as generating functions. Goals of this text This text introduces students to combinatorial problem-solving. and basic types of graphs. Students will use this reasoning often. but first of all. more rewarding role in the classroom. such as planar graphs and trees. and the uses of graphs in computer science and operations research. 3 . Second. A logical mind will serve a person well in any field. The inclusion-exclusion formula is a grand exercise in applied set theory. introductory courses in any mathematical subject. especially courses serving many non-math majors. in computer science. such a goal makes this text little more than an IQ enrichment book. they are likely to give the instructor constructive feedback that will make the course even better. An instructor's enthusiasm for an elegant theorem may influence some students to study the mathematics for its own sake. Such reasoning is the foundation for building simple mathematical models of problems– models implicit in counting expressions built of sums and products of binomial coefficients. Whatever the interpretation. 3. or in graph models. If students believe that the instructor is interested in their ideas and questions. recurrence relations develop recursive reasoning (so important in computer science). an instructor's outlook in class should reflect the needs and point of view of the majority of students. if they disliked the way they were taught. The following principles underlie this style of teaching. instructors should seek to develop good rapport with their students. In a first undergraduate course in an applied subject. For example. should try to leave students with a favorable attitude toward the subject and the learning experience. Third.

Conversely. as in most other math courses.. then the 2 -1 answer will be obvious and the student may write down only the numerical answer with no explanation. how many different non-empty subsets (of any size) n are there of the integers 1. Related difficulties may arise for math majors used to theorem-proving courses.g. A combinatorics instructor often hears the complaint from students that the textbook is unfair because the homework exercises are so different from the worked-out examples. etc. “choose which 4 of the 6 apples”) with an arrow from the phrase to that term in the expression. there are some students for whom combinatorial problem-solving comes very easily. If a student understands the binary n sequence model quickly. They may have trouble developing their own simple models for analyzing word problems." But several days later when students read over their notes. Note-Taking A correct solution to most combinatorics problems is often "obvious" once seen. card problems and logical puzzles. The answer is 2 -1. and the instructor should pause at times to allow students the chance to write out explanations. with real-world problems.and there are 2 n-digit binary sequences and n n hence 2 subsets. They are not accustomed to having to figure out how a problem should be solved.they expect that the homework problems should be solvable by mimicking calculations in the examples. applying a given integration technique to various functions. the n expression 2 -1 may now mean nothing.. However." The instructor at first must remind students to be careful to take good notes. Student Background Many students find the transition from calculus courses to a combinatorics course very frustrating. n. As a consequence. . They have become conditioned to technical "plug in" problems. Consider the problem. . students who coast along will see their grades deteriorate when other students catch and surpass them by dint of hard work. 2. 5. it is helpful to annotate the answer by writing a brief descriptive phrase for each term (e. nor may the phrase "model as binary sequence.Teaching Tips Finally this text seeks to achieve these learning objectives in as appealing a manner as possible. which can be explained: any subset (including the empty set) can be represented as an n-digit binary sequence with the n i-th digit = 1 if and only if i is in the subset. When the answer to a problem is an expression involving products and/or sums of binomial coefficients. 4. . students often copy down little or no explanation of the reasoning behind an answer.. 4 . or possibly the phase "model as binary sequence. 2 -1 of which are non-empty. This writer's experiences in teaching applied combinatorics consistently re-affirm the importance of giving problems interesting settings.

the problems are the message. class discussion of the assignment. it is by finding errors in an initial guess that one gains the insight that leads to the correct analysis. wrong ideas must be corrected and refined. It is helpful to hand out a copy of last year's test (or a realistic sample test). there were a homework assignment. 5 . Along the way. then counting the homework grade on the second assignment would be reasonable. If for each section. Mathematicians do not divine right answers. (Getting students to propose answers that are likely to be wrong for constructive analysis is a tricky business that requires good student-teacher rapport. and finally a second assignment on the material. students are eager to solve lots and lots of practice problems to prepare for tests. They have ideas and sort through them in an evolutionary way until all the steps needed for an answer are appropriately assembled. if necessary. Students discover that they now can solve problems that tricked them earlier.) Reviewing for tests is an important learning period in the course. Typically. The old adage "learn from one's mistakes" is the essence of modeling. Students learn only a modest amount from seeing right answers. The week before a test. The most important aspect of discussing students' homework solutions is the wrong answers. This writer often creates a de facto second assignment by going over half of the exercises in an assignment in one class and then letting students have extra time to go back and rework. Exams are meant primarily to measure a student's proficiency at solving problems. Homework Assignments and Reviewing for Tests In this course.Teaching Tips 6. Most of the course should be centered around the homework and how a student learns from homework mistakes. the other exercises and hand in the whole assignment in the next class. The instructor should help students learn to find right answers from initial wrong answers.

which avoids the litany of definitions and theorems found in most introductions to graph theory. Section 1. A relatively informal approach is used. The goal of this chapter is to develop: (i) familiarity with graphs. and (iii) some 'common sense' about exploring and analyzing simple graph properties.3 presents an important simple theorem about graphs. chosen from 6 colors (repetition is allowed). say. Section 1. Although graphs are assuming an ever-growing role in computer science and have always been important in operations research. Section 1. Given a set of guesses and their scores. Guesses at a solution are scored with two markers that tell (i) how many pegs in the guess are correct. 6 . logical examination of possibilities.4 examines an important graph property. Chapter 1: Elements of Graph Theory This chapter introduces students to graphs. This chapter's problems aim to illustrate simple uses of graphs and to build a general familiarity among students with graph models. (ii) facility with constructing graph models.2 uses the question of when are two graphs isomorphic to explore the intrinsic structure of graphs.1 illustrates the diverse uses of graph models. and all its sections should be covered. some students will never have seen graphs before. The examples and exercises are generally easier than the problems in later chapters. Even small graphs are such complicated structures that brute force enumeration of possible subgraphs to. Mathematical insights are required to simplify and efficiently sort through the possibilities. The problem-solving analysis required in this chapter is efficient. It involves determining a sequence of 4 colored pegs. The whole concept of a non-numerical mathematical model will be new to them. and (ii) how many pegs are part of the solution but are in the wrong position in the guessed solution. planarity. This is the most important chapter about graphs. Section 1. a crucial part of the combinatorial reasoning in later counting chapters. This author uses challenging Mastermind exercises as supplementary problems that are posted on the course website once a week (Thursdays at 10 pm) with extra credit points on the next exam awarded to the first 10 students to submit solutions.Commentary on Text PART II: COMMENTARY ON TEXT Prelude: The Game of Mastermind The game of Mastermind was introduced in the 1970’s. verify a particular graph property is normally infeasible. The combinatorial logic involved is solving such problems is the core of the reasoning skills used throughout this text. but ad hoc. students are asked to reason out what the correct solution must be.

Commentary on Text Section 1.1: Graph Models The section quickly introduces the concept of a graph (vertices, edges, adjacency relation, and little else). Note that, as defined in this book, graphs cannot have loops or duplicate edges. Example 4 introduces an exploratory but disciplined analysis of an edge covering problem. The unexpected relation between edges covers in Example 4 and independent sets in Example 5 should be emphasized. While Example 6 introduces a directed graph problem, it should be noted that directed graphs get very modest treatment in this text. Many other good examples of graph models may be found in the exercises. Exercises: Exercises 1-6, 21, 26, 27 present other graph model problems. The other exercises mimic and extend the models introduced in the section's examples. Section 1.2: Isomorphism This section examines the structure of graphs, via the problem solving vehicle of asking whether two graphs are really different or are just "rearrangements" of each other. To answer such questions, it is necessary to begin introducing some graph terminology, such as, degree of a vertex and complete graph. The informal use of some graph terms and the arguments using these terms in Chapters 1 and 2 may be too lacking in rigor for some instructors. However the machinery needed to make all definitions and proofs rigorous would greatly slow down the pace of the course and sidetrack the purposes of the course. Chapters 3 and 4 have more rigorous presentations. The introduction of the complement of a graph is an important secondary theme in this section. Exercises: Exercise 14 anticipates the parity Corollary in the following section. Section 1.3: A Simple Counting Formula The theorem and corollary in this section are the only general results known to apply to all graphs (by 'general', we mean without specially defined terms). The corollary is used to prove the nonexistence of certain types of graphs, as demonstrated in Example 2. Example 4 is a very appealing application of the parity Corollary although it requires a little time to present. The author finds that it is well received by students– the time to present it is worth the effort. Exercises: Exercise 8 is a true real-world scheduling application of the parity Corollary. In the early 1970’s, the NFL owners once designed scheduling parameters for pro football games that violated the Corollary.

7

Commentary on Text Section 1.4: Planar Graphs Planar graphs are of practical importance because they arise so often in operations research problems. Planar graphs are also of historical importance in graph theory because much of the development of the subject centered around the question, are all planar graphs 4-colorable. This section surveys some ways of determining whether a graph is planar. At first, an ad hoc geometric approach is used. Next Kuratowski's forbidden configuration is tried. Then some theory is developed: Euler's formula for planar graphs and a powerful corollary. From this corollary, many graphs can be shown to be non-planar with a simple algebraic test. This section is a case study in the power of a little good theory over ad hoc arguments. Exercises: The exercises develop several extensions of the theory and other basic concepts of planar graphs, e.g., properties of dual graphs in Exercise 10 and Platonic graphs in Exercise 24. Variations of Exercise 7 are a favorite for this author on exams. Exercises 25 and 26 integrate ideas in section 1.3 and section 1.4.

Supplement: Supplementary Exercises This section presents a collection of additional graph theory problems. The problems examine a variety of graph terms, such as strong connectivity, cut-sets, and automorphisms. Exercise 5 uses the Pigeonhole Principle. Exercise 20 is a cute logical puzzle that can be analyzed with the help of graphs. Exercise 30 is a well-known Ramsey theory question, whose solution (given in Part 4 of this Manual) is famous for its elusive simplicity. Exercise 32 is a 'trick' problem whose answer is a set of isolated vertices.

Chapter 2: Covering Circuits and Coloring
This chapter presents three topics in graph theory, Euler cycles, Hamiltonian circuits, and graph coloring. These subjects are all over 100 years old, originally arising as games or semi-recreational mathematics, but have extensive uses today in operation research. Logical arguments are more carefully constructed and stated in this chapter, although only four theorems, the Euler Cycle Theorem, the Hamilton Tournament Theorem, the Polygonal Triangulation Theorem and the Five Color Theorem, are actually proved. The three topics presented here are inherently appealing because of their recreational flavor. Further, what is fun is also seen to have practical applications. With this motivation, students are asked to perform the careful logical analysis of the possibilities required in ad hoc testing for Hamiltonian circuits and coloring. Section 2.4 on coloring theory can be skipped or covered briefly in most courses.

8

Commentary on Text Section 2.1: Euler Cycles Euler cycles are a pictorial topic with a simple, complete theory. This section presents an application to street sweeping and uses it to develop an alternate proof of the Euler cycle theorem. Exercises: There are several fairly easy variations of the Euler cycle theorem in the exercises: Exercises 8, 12, 13, 17a and 18. Exercise 19 is a famous application of Euler cycles. Section 2.2: Hamilton Circuits The simple theory of Euler cycles is now replaced by the more normal graphtheoretic situation of complicated ad hoc arguments, here needed to determine whether a graph has a Hamilton circuit. This section's focus is on thorough analyses to prove a particular graph cannot have a Hamilton circuit. The section present four theorems, only one of which is proved, about Hamilton paths and circuits. The section ends with an example (new to this edition) applying Hamilton paths to the construction of Gray codes. Exercises: The graph in Exercise 4o (it also appears in Exer. 1o in section.2.3) is called Petersen's graph; it has many "'bad" properties. Exercises 9 and 10 present some "tricks" that can be used to prove quickly that certain graphs cannot have a Hamiltonian circuit. Exercise 15 about a knight's tour is a well-known result but finding the tour is tricky. The following heuristic works: starting from any vertex, for the next move always choose to go to the square from which the number of possible subsequent moves is minimal. (Note: Ball and Coxeter's Mathematical Recreations and Essays, U. of Toronto Press, p.l75-184, has a fascinating discussion of the history of the knight's tour problem, including contributions of Euler and DeMoivre). Section 2.3: Graph Coloring This section presents logical ad hoc arguments for determining the chromatic number of a graph and illustrates the practical uses of coloring with examples in operations research and computer science. Although the garbage collection model takes a whole page to explain, students have always found it very interesting. Exercises: The trick to showing the graph in Exercise 1q cannot be 3-colored is first showing that a 3-coloring of the left half (vertices a,b,c,d,e,f,g,h) forces a and h to have different colors (as if they were joined by an edge), and similarly for i and p in the right half. Exercise 7 is a good review of geography. Exercises 10-14 are simple color modeling problems. Section 2.4: Coloring Theorems This section starts with an interesting problem in geometry. Then presents some results in coloring theory, which was initially motivated by a century of research on the Four-Color

9

It is possibly the most powerful use of a graph model known to this author—powerful in the sense of organizing the information in a combinatorially rich situation so efficiently that the problem can be solved by inspection in a few minutes once the graph representation of the puzzle is constructed. Section 3. try a 1-3 interchange at a and a 1-4 interchange at a. Sometimes the author presents this model at the end of the last class to end the course on a high note. depth-first search and breadth-first search. a baby sister of the famous Four Color Theorem. Namely. now try a 2-3 interchange at e and a 2-4 interchange at b to make e 3 and b 4 and permit x to get color 2: the error is that the 2-3 interchange at e may break the 1-3 path from a to c so that the 2-4 interchange at b may change d to 2. Exercises: Virtually all the exercises in this section are proofs. The Five Color Theorem. the instructor should buy his or her own set of cubes and solve them in class.4 uses trees to analyze the behavior of several sorting algorithms. It finishes with new material about chromatic polynomials.3 illustrates two different uses of trees in attacking complex operations research problems.Commentary on Text problem. in Figure 2. Section 3. There are enough components to the model that a good 20 minutes should be allocated to present the model and its analysis. Practice arranging the cubes into the correct pile a few times before class. if both fail (there is a 1-3 path from a to c and a 1-4 path from a to d). This chapter is essential material for computer scientists (although much of it may also be covered in computer science courses). 10 . the Branch and Bound method and heuristic approximation. number some corner of each cube 1 through 4. The sample operations research problem chosen here is the famous Traveling Salesperson Problem. The trees are used to build search algorithms and to analyze these procedures. The class usually bursts out in applause when the puzzle is solved.15 suppose e is color 2 and G-x is 4-colored.2 introduces spanning trees and the two most common types of searches. Appendix: Graph Model for Instant Insanity This clever model speaks for itself. Section 3. Chapter 3: Trees and Searching This chapter begins with some terminology and basic theorems about trees. Rather than using the sample cubes in the text. To keep the cubes distinct. The original false 1876 proof of the Four Color Problem by Kempe used the same argument that works successfully in the Five Color Theorem proof. The rest of the chapter applies trees to various searching problems. is proved.

) There are four theorems giving formulas involving various parameters of trees. Tree traversals are also discussed.2: Enumeration with Trees This section presents depth-first and breadth-first search methods with recreational examples and introduces the concept of a spanning tree. Although most computer science students will have studied this material before. Exercises: Exercise 5 presents an important production scheduling problem. Rather than sort the whole list at once. Theorem 4 is used in Section 3.4: Tree Analysis of Sorting Algorithms Binary trees are a convenient model for representing most sorting algorithms.3: The Traveling Salesperson Problem This section first presents the Branch and Bound technique. Section 3.Commentary on Text Section 3. Such complicated operations research problems are often attacked with heuristic procedures that yield near optimal solutions.1: Properties of Trees This section begins with a collection of tree-related definitions. The beginning of the chapter has been re-worked with a new theorem about the equivalence of different definitions of a tree. called heaps. this brief revisit with its tree approach should be valuable to them. while basically straightforward. Exercise 32 presents a one-pass algorithm for finding an Euler cycle in a directed graph. Exercises: Exercises 3-18 are theory problems and exercises 19-27 involve applications. The problem used to illustrate the technique is the famous Traveling Salesperson Problem. a tree-based procedure to search for optimal solutions in combinatorial operations research problems. a heap sort picks out one element at a time from a heap. The proof bounding the accuracy of this heuristic. Theorem 5 presents an interesting way to associate a unique numerical sequence with a labeled tree.4 to show that any algorithm for sorting n items needs at least log (n!) [ = O(n log2n)] comparisons in the worst case. This section concludes with a sorting procedure using partially ordered trees. (These definitions are collected together in the Glossary at the end of text. Exercises: Exercise 30 presents a binary sequence characterization of binary trees. 11 . Section 3. They are also a necessary tool for analyzing the efficiency of sorting algorithms. is a bit difficult for students to follow and may be skipped. Section 3. The second part of this section presents a tree-based heuristic for the Traveling Salesperson Problem.

Prim's algorithm. This is followed by the correct approach. It is a good example of operations research algorithms and serves to prepare students for the labeling procedure used in the augmenting flow algorithm in section 4. the proof is literally constructed by the augmenting flow algorithm.3 which introduces the basic theory of network flows. Section 4. The section begins by introducing the terminology of network flows and basic properties of flows and cuts. Search trees underlie the shortest path. The theory in this section 12 .3.5 (new) uses trees in a more central fashion to solve the Transportation Problem. an intuitive but faulty approach to flow building is illustrated. Matching problems are an application of flows in which negative labeling arises naturally.4 applies flows to the combinatorial theory of matching. Finally. the results in both case are uncharacteristically (for graph theory) simple. If time is tight. several basic variations of network flows are modeled by standard single source-single sink flows. Exercises: Exercises 12 and 13 present another basic variation on flows.Commentary on Text Chapter 4: Network Algorithms The theory of network flows is one of the great accomplishments in graph theory. Section 4. The presentation here is mathematically rigorous. Section 4. one can skip the proof of Prim's algorithm. Section 4. Section 4.4: Algorithmic Matching The theory of network flows is applied in this section to solve matching problems and to prove the two fundamental theorems of matching theory. maximal flow and matching algorithms. the Augmenting Flow Algorithm of Ford and Fulkerson.3: Network Flows This long section develops the basic theory of network flows. An important pedagogical aspect of network flows is the use of an algorithm to prove a theorem (the Max Flow-Min Cut Theorem). Minimal spanning tree algorithms are somewhat like Euler cycles. exercises 18-40 are theory problems. Corollary 2a should be carefully explained. both in terms of mathematical elegance and practical uses.2: Minimal Spanning Trees The two standard algorithms for finding a minimal spanning tree are presented and the validity of one. This interplay of algorithms and theory is typical of much recent theory in computer science and operations research.1: Shortest Paths Dijstra's shortest path algorithm is presented. for it is the key to understanding the flow algorithm. This chapter first "warms up" for the augmenting flow algorithm with sections on shortest path and minimal spanning tree algorithms. Next. Exercises 1-17 are computational problems. is proved. The main part of this chapter is section 4. capacitated vertices. Section 4.

much of the learning will occur after assignments are done and homework is discussed in class. On the other hand. and only a few basic formulas: just lots of examples to help prepare students to do their own problem-solving. Section 4. then the instructor will be better prepared to help others with their questions during the class. exercises 15-24 are theory problems.Commentary on Text should be skipped for students with limited mathematical maturity. (Historical Note: the author’s father. they may become so uncertain of themselves that they no longer can solve easier problems. As students worry about more complicated constraints found later in the chapter. A W Tucker.the 'mathematical' (early) elimination of teams in a league championship. However. the problem solved here. Students easily misinterpret problems. It introduces general problem-solving in combinatorial enumeration.3 should then be treated as just one more example of network flows. Most of the exercises is examples of transportation problems.) Exercises. It is helpful to establish some personal rapport with students if uncertainty is setting in. can the instructor be in a position to help. Only after the student slowly explains the way he or she analyzed the problem. Example 4 is perhaps the most involved example in the text and illustrates the complex models that can be made with network flows. was first attracted to the latter subject by the transportation problem with its rich graph-theoretic structure. the first two pages of section 4.5: The Transportation Problem (New) This section presents the classical operations research analysis of the Transportation Problem with an emphasis on the underlying bipartite graph and the central role of trees in finding an initial solution and finding improved solutions. Chapter 5: General Counting Methods for Arrangements and Selections This is the most important chapter in the text. Exercises: Exercises 1-14 are computational problems. and even extensive discussion of an exercise in class may not clear up the confusion. the columns associated with edges of a tree form a basis for the solution space. 13 . Rather. There is no theory. Tell them this material is tricky for professors too (it is!) and encourage them to come to office hours with their questions. who was trained as a topologist but became one of the leaders in developing mathematical programming.is very interesting. It is virtually impossible to teach students the right way to do problems in advance. Some students will have difficulty solving problems whose analysis does not mimic an example in the text. If some students come to see the instructor with questions on the homework before the homework is discussed in class. in operations research texts trees are studied from a linear algebraic viewpoint—in the constraint equations.

g.27. If you throw just one dart and it hits the bull's eye. 34.2.WW. you were lucky but have learned little about controlling the flight of a dart. 28.WB <. A good analogy is learning to throw darts. Even though combinatorics is full of cute short solutions.13. Example 49 presents the game of Swap. e.1 through 5. The progression of increasingly more involved problems from Section 5. . manageable subproblems. blues move right in a similar fashion. It is hard for students to anticipate such mistakes. in this writer's opinion.Commentary on Text It is extremely important to stress the value of learning from one's mistakes.e. BBWB_ <. 14 . the game is played: BBB ->_WWW.. Section 5. WWW_BBB. e. how many distributions are there of 20 identical administrators into .2.1 and 5. 59. humorous settings are used in Section 5.WWW. Professor Mindthumper and names from Tolkien's "Lord of the Rings". before it is assigned. But if you start with a poor throw and successively improve your aim until you hit the bull's eye.1: Two Basic Counting Principles The Addition and Multiplication Principles. Section 5. 61.1 in this edition. 47. To lighten the possible tedium of combinatorial word problems. 31. should be devoted to discussing homework. with n white pegs and n blue pegs separated by a space. W_ <.. Instead.4. BB_B <. Incomplete and incorrect decompositions are the cause of most counting errors. Exercises: Exercises 30-31. discussing the faults in wrong solutions so that students will be able to avoid them in the future. Note that the solutions to selected exercises appear in a Supplement at the end of the chapter. 28. 33. 59. The set of solved exercises is: Section 5.WB -> WB > WB -> _.6 in the fourth edition has been moved to section 10. 65. then you have really learned to throw darts. 34. 77. 52.W (3 successive moves).3. 29.5 is the standard development found in most combinatorics texts. For n=3.5 on binomial identities may be skipped. 55. Later problems in this chapter will involve repeated. Note that the old section 5. whites move left either one step (to an open space) or two steps (jumping over a blue peg to a following open space). This game should be explained in class. for breaking problems into disjoint or sequential subproblems.g. BB -> W_WBW. i. . they must learn from their mistakes.25.1. Section 5. _B <. the counting itself is easy once one knows what to count. the best way first to approach any counting problem is by using these two principles to produce a thorough case-by-case decomposition into small.45. Part d) of Example 4 illustrates a common error that should be emphasized..WB <. This is why a large amount of class time in the course. 64. 33.WB <. Section 5. intermixed use of these two principles. WWWB -> _BB. B -> _WBWBW. are the fundamental building blocks of all combinatorial enumeration. Section 5. BBWB -> W_W. 30. 44 and 45 are primarily modeling or interpretation problems.WBB. 51. The instructor may want to continue the levity by changing the setting of subsequent examples in the book.

counting poker hands with exactly one pair. 15 .Commentary on Text Section 5. Assigning some review exercises from previous sections along with new problems is worth considering here. it has many possible right and wrong analyses. certain consecutive pairs are forbidden. Example 8 presents a measure of voting power. and the model of integer solutions of an equation is introduced. which has important uses in game theory and politics. Exercises: The first 17 exercises are similar to the examples. To assist students. Example 6 is a counting problem that arises in quality control probabilities. Exercises: The exercises in this section contain many different types of constrained arrangement and selection problems. At each stage. but the formula for selections with repetition is usually new.3: Arrangement and Selections with Repetition Most students have seen the formula for arrangements with repetition. The remaining exercises involve more complex constraints and. This mistake of implicitly ordering the same set several different ways (picking a first part and then a second part of a set different ways) will be made over and over again in homework solutions. are quite difficult word problems. The section extends and compounds the examples in the previous section.2: Simple Arrangements and Selections Simple arrangements and selections are the small. For several sections. The worked examples are of little help in attacking many of the later problems. The equation model is central to the development of generating functions in the next chapter. Section 5. students have been faced with increasingly complicated problems. This section introduces basic types of unconstrained and constrained arrangement and selection problems. is a pet problem of this author. a table at the end of the section summarizes different counting formulas for ordered and unordered sets. identical and distinct objects are intermixed. manageable subproblems into which complex counting problems are commonly decomposed. they had to move on without time to master fully the current problems. Exercise 55 is the famous Birthday Paradox. Exercise 65a. in general. There are more complex constraints. Some students will get overwhelmed with the amount of ingenuity required and the variety of these exercises. Section 5.4: Distributions This section presents equivalent distribution models for arrangements and selections with repetition. Exercises 60 and 61 introduce two important combinatorial problems in sampling theory. The most common error in enumerating unordered sets is addressed in the Set Composition Principle and is demonstrated in Example 5d. This section introduces problems involving simple constraints with repetition. This section is the most difficult part of the course for most students.

Example 47-49 are pet problems of the author (variants of which always appear on his tests).Commentary on Text Exercises: Exercises 31-34 provide straightforward. Another polynomial expansion problem is restated as a combinatorics problem. as shown in Example 4 and 5. However. Section 6.5 shows how summation problems can be solved with generating functions.1 explains how ordinary generating functions can model problems of selection with restricted repetition. e.5: Binomial Coefficients This section begins with an explanation of how binomial coefficients get their name. The polynomial model is the sole message of this section. the combinatorial reasoning used to verify them has much in common with the reasoning used in previous sections. the Binomial Theorem can easily be explained at this point from scratch. it would be better to concentrate on a solid mastery of ordinary generating functions. Section 6.5.4 to solve a special class of recurrence relations. Although binomial identities do not appear to be problem-solving.4 introduces exponential generating functions.2 is concerned with techniques for evaluating coefficients in given generating functions. Section 6. identity (8) provides a convenient way to evaluate combinatorial sums.4. This section is needed for section 7. well-structured model. Section 6. if section 5. Section 6.5 on generating function methods for solving recurrence relations. and combinatorial problems are restated as polynomial multiplication problems.5 was skipped. yet valuable. this chapter attacks problems with a very specific. as well as Ferrers graphs of partitions.g. Then the symmetry relation and Pascal's recurrence relation are given. Example 51 presents an important alternate way to approach non-consecutivity problems (the method in Example 9 is more flexible. The modeling performed by these functions is a bit more difficult to understand. i. practice in translating between different arrangement and selection models. Chapter 6: Generating Functions Generating functions are probably the most important enumeration model in combinatorics. Many courses will cover just these two sections and possibly also section 6.k)'s arise in the binomial expansion. students are shown how to build generating function models for selection problems with restricted types of repetition.3 looks at partitions and their generating functions. Moreover. that of polynomial multiplication. 16 .e. In contrast to the unstructured problem solving of Chapter 5. Then the direction of the modeling is reversed.. This summation technique is used in section 7. The rest of the section concerns binomial identities. how the C(n.1: Generating Function Models In this section. easily accommodating the added constraint of at least two symbols between each vowel). Many courses will stop at this stop. The section starts with a review of the combinatorial explanation of the Binomial Theorem given in section 5. Section 5. These first two sections are the core material of this chapter.. Section 6. In many undergraduate courses.

The last page on combinatorial identities can easily be skipped. Exercise: The first 19 exercises are fairly straightforward. primarily as an example of infinite generating functions. and 44 cover the basic properties of probability generating functions. Exercises 38-42. Section 6. Any one of these three concepts may prove to be a major stumbling block for students. Exercise 44 is a basic result of queuing theory. and conversely. Exercises: Exercise 13 is an interesting use of Ferrers graphs. Exercises: Exercises 11 and 15 present variations of the expansion formulas given in this section.Commentary on Text The three key concepts in this section are: a) polynomial multiplication generates a set of formal products. Exercises 43 and 44 present examples of the "chain rule". and c) the number of such sums of exponents is just an integer-solution-of-equation problem. any integer-solution-ofequation problem can be viewed as a sum of exponents problem.2: Calculating Coefficients This section presents the necessary algebraic manipulation for evaluating particular coefficients of common types of generating functions. Section 6. As these techniques are demonstrated. The topic is a nice extension of the ordinary generating function model introduced in the beginning of this chapter.for generating functions. Letting rote algebraic manipulations do the work of the combinatorial analyses in Chapter 5 is a good example of the algebraic spirit of modern mathematics. The fact that each formal product now contributes a value of n!/el! er! to the coefficient makes exponential models much harder for students to understand. 17 . their combinatorial interpretations are also discussed. as in expansion (3). to prove that integer multiplication is commutative (this same problem is given as an induction proof exercise in Appendix A. Section 6. Exercises 20 and 21 are important examples of faulty generating functions. from polynomial expansion to counting problem and from counting problem to polynomial expansion. The best cure is for the instructor and a student to go slowly over the modeling process with a sample problem in both directions. b) these formal products are characterized as a sum of exponents.4: Exponential Generating Functions This section begins with a careful discussion of a sample exponential generating function model. The section finishes with the Ferrers graph diagram for representing partitions and partition identities. Even the very last exercises on multivariate generating functions are not hard.3: Partitions A brief discussion of partitions is given.2).

However. Then the "summation operator" 1/(l-x) is used to sum such summands. again a little private help should quickly straighten things out. recurrence models may confound some students at first.1: Recurrence Relation Models This section is the heart of the chapter. Like generating functions. While recurrence relations are a fundamental tool in computer science for analyzing recursive algorithms.. b) medium-sized problems are readily solved numerically by iteration. and so they will not see the great practical importance in these relations that the instructor does. If enough parameters are used. The section concludes with a discussion of the connection between recurrence relations and difference equations. and may be the only section some courses cover in this chapter. Exercises: Exercise 7 is the original Fibonacci problem about multiplying rabbits. Solving the relations is of limited importance because: a) an applied combinatorics course normally emphasizes the modeling side of problem-solving. Section 6.5. This chapter's primary objective is teaching students how to build recurrence relation models. A warning about compound interest relations: Many students have yet to worry about savings accounts. Section 7. The material in this section is used for solving recurrence relations in sections 7. one simply determines parameters in "cookbook" families of solutions. the problem after which the Fibonacci relation is named. virtually any counting problem can be modeled with a recurrence relation. assigning it is a must (unless covered in class). such relations have much wider applicability in all applied combinatorics than is generally realized. A recurrence relation incorporates in a formal mathematical equation the combinatorial reasoning for decomposing a counting problem into similar subproblems.4 and 7. Chapter 7: Recurrence Relations. with the exception of generating function methods.5: A Summation Method This section shows how to build a generating function with coefficient an equal to a given combinatorial expression.2.Commentary on Text Exercises: Exercises 21 and 22 continue the probability generating function exercises in section 6. it raises an ancillary issue that can partially sidetrack students who should be concentrating on finding recurrence relation models. once a recurrence relation is available. Exercises 28 through 37 involve primarily multivariate relations and 18 . Its secondary objective is to present a brief survey of methods for closed form solutions to recurrence relations. and c) the solution techniques. are of limited general pedagogical value. This last material can be skipped. etc. A variety of examples of recurrence relations are given. mortgages.

2: Divide-and-Conquer Relations This section presents "cookbook" solutions to a class of recurrence relations that arise frequently in the analysis of recursive algorithms. when in doubt.5) or generating functions (section 6. While important. the idea behind Strassen's famous fast matrix multiplication (that multiplies two n x n matrices with less than n3 multiplications). Section 7.3: Solutions of Linear Recurrence Relations Linear recurrence relations are the simplest form of recurrence relations. Example 4 is fairly complicated and the instructor may want to skip it (to avoid scaring students). either using binomial identities (section 5. Section 7. If k ≠ 1.5: Solutions with Generating Functions This section shows how recurrence relations can be converted into equations for associated generating functions. Section 7. the instructor should spend a few minutes in class sketching the form of solution to linear recurrence relations. Many problems presented in Chapters 5 and 6 can be solved more easily now. this material is difficult for most undergraduates who are not upper-division math majors. at best. students saw briefly in calculus and quickly forgot).5). Chapter 8: Inclusion – Exclusion Formula In this chapter. that the number of objects without property A is the total number minus the number with property A. Examples 2 and 5 use partial fraction decompositions (a technique that. The key pedagogical points in this section are the similarity of the forms of solution for linear recurrence relations and linear differential equations. the counting principle of complementation. It could require upwards of a week of the course to present this section properly to other students. consider using the Heavyside method (discussed in most calculus texts). Example 3 demonstrates.Commentary on Text simultaneous relations. Section 7. it helps to add additional variables or supplementary equations. Students should be told that. the relation is just a summation problem. This conversion is the discrete counterpart to transform techniques for solving differential equations. If k = 1. Even if this section is not assigned. to determine coefficients in partial fraction expansions. in a simple form. is generalized to situations involving many properties. it is necessary to have previously discussed summation. and that the general solution to a linear recurrence relation is really a family of solutions. The "theory" of this chapter consists of working out in 19 . one for each possible set of initial conditions. an will be essentially the same type of function as f(n). Exercises 46 and 47 are quite tricky. In the latter case.4: Solutions of Inhomogeneous Recurrence Relations This section presents solution techniques for inhomogeneous first-order linear recurrence relations: an = kan-1 1 +f(n).

27. A second problem is trusting in a settheoretic formula. Exercise 31 looks at sequences of distinct subsets. Section 8. expressed in terms of the sizes of various set intersections. Exercises 20. Exercise 15 is a nice variation on the standard inclusion-exclusion type of problem (here the 'answer' is given and N must be determined).3: Restricted Positions and Rook Polynomials This section on rook polynomials presents a clever model for restricted position problems using chessboards and generating function models. To solve a given problem with this formula.g.1: Counting with Venn Diagrams This section slowly generalizes the principle of complementation from one property to two properties and then to three properties. and 28 ask for unions instead of intersections. and b) count the number of objects with various subsets of the properties– these latter counting problems are usually simple Chapter 5-type counting problems. Exercises: Exercise 14 has tricky wording. Section 8. This material is not difficult.Commentary on Text general set-theoretic terms a formula. Section 8. A few students may be bothered by the generality of the summation notation in the formula. The counting subproblems in Exercises 12. combinatorial sets whose elements are themselves combinatorial sets. The last section develops a nice mini-theory for solving arrangement problems with restricted positions. namely.. similar problems had to be carefully thought out and broken into appropriate subcases. Exercises 31-34 involve a new type of combinatorics problem. 27.2: Inclusion-Exclusion Formula Now the students should be ready for the general inclusion-exclusion formula. Theorem 2. cannot be solved by inclusion-exclusion but rather require ad hoc Boolean algebra analysis.1. can be skipped for most audiences. like Example 6. 24. Each step of the theory is easy to 20 . not n properties). e. Some students already have trouble in the 2-property case with defining the two properties that are not to hold for the elements to be counted. Some students will already have forgotten the necessary Chapter 5-type counting methods needed for solving some of the subproblems arising in the use of the inclusionexclusion formula. Exercises: Exercises 33-36. it is really a selection-with-repetition problem with at least one of each type. a student must: a) define a group of properties. Now students are conditioned to such an approach rather than "plugging into" a formula. for the number of objects that have none of a group of properties. Exercise 30 is a tricky "recursive" inclusionexclusion problem (see its solution in Part 4 of this Manual). Exercise 25a looks like a derangement problem but is not (there are n -1 properties. at the end of the section. 28. but time constraints may force the last section to be skipped. In Chapter 5. and 29 require combinatorial insights developed in Chapter 5. The proof of the general Inclusion-Exclusion formula can be sketched as a generalization of the 3-set proof in section 8. This is a matter of logical negation in set theory and propositional logic.

The approach here now works well for the author. a proof of the lemma should be given. Section 9. Burnside's Lemma is the only part of this section that has been difficult for students at the author's institution. Examples 2.1: Equivalence and Symmetry Groups This section introduces the sample problem of two colorings of the corners of a square. A heuristic argument for this algebraic lemma is presented. of tinker toys) of a square to show how motions transform one coloring pattern of the square's comers into another pattern. However. or covered lightly. The key idea is that with "multiplicities" counted. and 4 about enumerating symmetries can be left for out-of-class reading.the latter permutations generate the equivalence classes we want to count. If most students have previously had a modern algebra. a problem in which generating functions play an important role. 3. It is important here to emphasize the difference between a motion's permutation of the corners of the square and a motion's induced permutation of the colorings of the corners-. Polya taught this topic to Stanford undergraduates in 1969 (a course for which this author had the privilege to be the grader). illustrated with the color class of C10. One of the pedagogical goals of this chapter is to present groups in a practical and natural setting (as opposed to the more formal setting in a modern algebra course). It is most helpful to have a physical model (say. every equivalence class will have size 8 for the square coloring problem. The concepts of an equivalence relation and a group are presented and linked together. the instructor 21 . Exercises: Exercise 4 is a valuable warm-up for cycle decompositions in sections 9. This chapter has the only extensive theoretical development in the enumeration half of this book.d. because of its abstract foundations in group theory. 8b. Polya's formula is usually not discussed in sophomore/junior-level applied combinatorics courses. after about 20-30 minutes of theoretical discussion. Chapter 9: Polya's Enumeration Formula This chapter discusses a counting problem in applied group theory. and 10 do not decompose into disjoint subboards (but 8 is not hard to do by inspection). The emphasis is on understanding the theory by using it to solve problems. In any case. This concrete approach follows very closely how G. Exercise 12 on the non-commutativity of symmetries should be assigned or covered in class.3 and 9.4.e. The case of non-disjoint subboards (starting after Example 1) may be skipped.2: Burnside's Lemma The section on Bumside's Lemma is possibly the most difficult section in the book. Exercises: Exercises 2c. the presentation "by example" used here is much easier to follow than the standard treatment.Commentary on Text follow but the final result is far from obvious. Exercises 16-25 involve basic group theory. The basic definitions and theoretical development are presented by examples without formal proofs. Section 9.

Commentary on Text should turn to applying the lemma's formula. It is desirable to get maximum student participation in filling in the various entries in the table. Chapter 10 Computer Sciences Approaches to Enumeration This chapter looks at enumeration from three computer science-based points of view.3: The Cycle Index The section is a self-explanatory development of the role of the cycle structure representation of a motion in counting the number of colorings left fixed by a motion. Section 9. Then it illustrates how to use these algorithms to enumerate outcomes in earlier counting problems. Exercise 13 continues the theoretical development of the theory exercises in section 9. what 22 .1: Generating Permutations and Combinations This section gives algorithms for listing lexicographically permutations and rcombinations of an n-set. Section 10. Section 9. a physical model of the cube is desirable. If Example 4 is discussed. which range from very concrete to very abstract. what is the next outcome in this subclass. Linking combinatorial analysis with programming has an important dividend: it is often helpful to think of building a formula for an enumeration problem as a dynamic process. Exercises: Exercise 8 presents a constraint most easily handled with the inclusionexclusion formula.. Writing out these algorithms precisely in some programming language is important for computer science students.2 introduces formal languages and shows how a combinatorial collection can be viewed as the sentences of a properly defined grammar. as if one were a computer that was printing out page after page of all possible outcomes. Section 10. what outcome to print next (e.1 presents sequel algorithms to enumerate all permutations or r. Section 10. This material can be integrated into the first two sections of Chapter 5.combinations of a given set of elements. The examples in this section are easy for students to read out of class.g.3 to obtain the final Polya's formula. Exercises: Exercise 11 is a cute (and tricky) combinatorics problem.3 introduces finite-state machines and looks at the problem of designing a finite-state machine to recognize sequences that are members of a given combinatorial collection.1.4: Polya's Formula This section extends the development in section 9. Section 10. but one should be covered in class. or is this subclass of outcomes completed and if so. The fundamental question for the computer. Students should not worry about remembering the explanation of the lemma.

The definitions and proofs are all based on recursive constructions and thus have substantial pedagogical value. This section would normally be skipped. The ultimate goal is a winning strategy for the game of Nim.3: Finite-State Machines This brief section takes a more theoretical approach to enumeration. Exercises: Exercises 9 and 10 about finding a given permutation's or combination's position in a lexicographic list are good problems for advanced students. generally requires the same type of analysis as needed to obtain a formula to count all outcomes. and Grundy functions. These building blocks are typically the same building blocks arising in a counting formula. contain some interesting theory. Mie recursive constructions need to be constantly illustrated with specific examples. The existential proof of a kernel-based winning strategy and the 23 . Grammars inherently involve a construction of the deserved combinatorial collection out of building blocks. Exercises. Chapter 11 Games with Graphs This chapter develops the theory of progressively finite games and applies this theory to find winning strategies in certain 2-person games. The exercises all involve building grammars for specified combinatorial collections. by building finitestate machines to recognize sequences in a given combinatorial collection. while recreational in spirit. Section 10.Commentary on Text subclass should be listed next). However. grammars can help develop counting formulas. Just enumeration algorithms can motivate the type of thinking needed to produce a formula to count all outcomes.2: Formal Languages and Grammars This section introduces the formal structure of languages and grammars and then designs grammars whose sentences are a given combinatorial collection. The two sections on progressively finite games.1: Progressively Finite Games This section introduces the basic concepts of progressively finite games: kernels. Further. levels. most mathematics majors never will take such a course and finite-state machines (and automata theory) are one of the most mathematically rich areas of theoretical computer science. Finite-state machines are much more complicated to design than grammars. finite-finite machines are a basic topic that is carefully treated in theory of computation courses in computer science. especially for computer scientists. Section 10. The sentences can also be viewed as the leaves in the tree whose internal vertices are production rules in the associated grammar. Section 11.

Exercises: Exercise 6 is a concrete example of part b (the omitted part) of the proof of the Theorem. like the problem of counting imprecise sets. is raised as a reminder to students that questions must be well posed before one can employ the combinatorial reasoning taught in this text. Example I illustrates the 'real-world' issue of how much information is needed to solve a counting problem. and with it the operation of digital sum. which should be review material but unfortunately is not for many students. This problem. an important topic in combinatorial theory. A. Appendix 4 presents the Pigeonhole Principle. since they will be used many times in chapters 5 and 8. Notation gets quite involved and the proof of the Theorem is beyond the grasp of most undergraduates. Exercise 5a consciously has contradictory data. some of the requested subsets cannot be generated (parts d and e).2: Mathemstical Induction Anyone who takes a course in discrete mathematics should be exposed to induction arguments.2: Nim-type Games This section introduces the complicating generalization of direct sums of games. Appendices Appendices 1 and 3 have background material on set theory and probability. Appendix 2 presents induction. Exercises: Note that in Exercise 2. Appendix 5 (new) gives a brief overview of computational complexity and NP-completeness. Exercises: Exercises 14 and 16 contain two important theoretical extensions (needless to say. If an instructor wants to put more emphasis on induction.Commentary on Text inductive construction of a unique kernel are friendly examples of important proof techniques. and one or two induction exercises from this section can be assigned 24 . they get a feeling for the idea behind the proof. The set complementation laws (deMorgan's Laws) are stressed. It is intended that these appendices will not normally be covered in class. they at least should be developing a familiarity with induction arguments for later courses in computer science and other areas of discrete mathematics. While students may be asked to do few induction proofs themselves in this course.1: Set Theory This section presents basic set-theoretic terms. extra examples (chosen from the exercises) can be done in class. Section 11. the proofs are quite difficult). A. although exercises from them might be assigned. However by learning how to play winning Nim. notation. and operations.

Exercises: Exercise 18 is a famous Ramsey-theory question with a slick proof. the 5-color theorem (section 2. in giving this definition. 3rd ed. we immediately raise the problem of distinguishing outcomes. and several theorems in chapter 10. 25 . A. which was requested by several reviewers. It is important for students to be appeal how incredibly difficulty it is to solve large instances of such problems. New York. yet still elementary.Commentary on Text every week for the first month. objects are always distinguishable (even if they have the same shape and color). This problem becomes fairly subtle when identical objects are involved. Note that identical objects arise implicitly in some probability problems. Exercises: Exercise 15 proves the commutativity of integer multiplication. A.4). e. and minimal colorings.3). The instances studied of those problems were snall-sized and chosen to be tractable to ad hoc combinatorial analysis.4 examine basic properties of probability generating functions. However. a bound on the height of m-ary trees (section 3.2)..2 and 6. Induction arguments are used in this text to prove Euler's formula for planar graphs (section 1. edge coverings. 1998). provides the necessary background to appreciate what is means to say that a problem is NP-complete.g. (North Holland. However it is a traditional combinatorics topic whose advanced theory has blossomed in recent years. Probabilities are used in Chapter 5 to give interesting settings to counting problems. This appendix.4: Pigeonhole Principle This brief section on the Pigeonhole Principle is independent of material in the rest of the text. a well-known fact that few can prove (this fact also has a simple combinatorial proof. whereas in combinatorics one allows identical objects.4). existence of a Hamiltonian circuit. namely. see R.5 Computational Complexity and NP-Completeness Many of graph problems studied in chapters 1 and 2 are NP-complete. such as counting all sequences with k (identical) heads and n-k (identical) tails when a coin is flipped n (distinguishable) times. Introduction to Combinatorics. or elementary events.3: A Little Probability The objective of this section is to present the definition of probability used in this text. Exercises 24 and 25 are classic examples of faulty induction arguments. Some instructors may find this section appealing because Pigeonhole exercises require a type of creative abstract reasoning generally missing in the rest of the book.1). discussion of the Pigeonhole Principle. Some advanced exercises in section 6. For a fuller. Games with Graphs. the validity of Prim's minimal spanning tree algorithm (section 4. see Exercise 13 in section 6. Brualdi. from compound events. A. A point that can cause some confusion is that in probability problems. the fraction of 'favorable' outcomes. The idea can be explained in half a minute in any subsequent course that may need it.

2.4.Sections 1.Sections 8.Sections 1.3 Week IV.Sections 3. Week I.1. One-semester 14-week Course with 40% graph theory and 60% counting. 5.1 Week III.Review and Test 2 (see note above with Test 1) Week XII.3. 1.Sections 8. 6. 1.2 Week XIV. Week VII.Review and Test 1 Note: extensive time for review before a test can be profitably used by students to work lots of problems from old tests or review sheets.2 Week V. 2. 1. 2.3 and Review 2.1.2.1.4.4 Week IX.2 Week VIII.3.1.2 Week II Sections 1.3 Week II. 7. One-semester 15-week Course emphasizing Graph Theory or One-quarter Graph Theory Course (using first ten weeks) Week I.Sections 2. 3.4 Week VI. 8.2 26 .4 Week XI.Sections 6.Sections 7.Sections 5.3 Week XIII. 7. 3.2.Sample Syllabi PART III: THREE SAMPLE COURSE SYLLABI 1.3.1.Sections 5. 2.1 Week X. 1.1.Sections 2.Sections 3. 6. Week III.Sections 5. 5.Sections 1.5.2.1.

3.VIII follow Weeks VII .Sections 3.1.Review & Test Week XI.1 and in Week VIII cover 8. 3.1.XV follow approximately Weeks I .Sections 7.Sections 6.Sections 3.2 Week XIV.3 Weeks IX.1.3 Week XV.Sections 9.2.1. 9. 8. 27 .Sections 2.IV in Course 1 above. 5.3.3.2 3.3 Week X. with time at end for review.1 Weeks XI . 7. 5. 4. 8. 3.Test and Section l. except in Week VII cover 7.3.4 and Review Week XI.Sections 5.4 Week V.Review & Test Note: extensive time for review before a test can be profitably used by students to work lots of problems from old tests or review sheets. 3.2. 7.2.2 Week XII.Sections 4.Sample Syllabi Week IV.1.1. 2. Week VI.3 (up to Max Flow/Min Cut Algorithm) Week IX. 6.4 Week XIII. 4.Sections 4.4 Week VIII.1.XIV in Course 1 (syllabus above).Section 9.5 and 8.Sections 8.Sections 5.5 Week VII. One-Semester Undergraduate Course emphasizing Counting or One-Quarter Enumerative Combinatorics Course (using first 10 weeks) Weeks I .2.4 Week X. 9.

W Bu R W. Bu Bu Bu Bk or Bk Bu Bu Bu or W Y W W 9. Bk Bk R Bu.(John.D). W W R Bu. 19.e must be matched with just b and e. Chapter One Solutions F 82 • 85 A • 90 77 91 82 90 •C E • Section 1. several routes. Bu Bk Bk Bu or Bk Bu Bu Bk. Bk Bu Bk R. 1040. 9. Many possible guesses: possibilities are 7.1: 1a) B• 82 77 85 D • b) 2 (C. 5a) many possibilities.c.Solutions Part IV Solutions to Problems Prelude Solutions 1. 4a) W b) 2 days. W O G O Bu. 6.Mary) or (John. c) a. Bu Bk R Bu. 10. b) min = 4. 3. DCBA. 4. 7. G Bk G Y or G W G Bu. 13a) Three black and one white. Mary. (C. B-b. CBAD. Two of one color and one of a second color. 12. A B 2a) D J b) ADCB. O R Y Bu P. several possibilities 6a) 5 vertices. D-c} (and other possibilities). {A-a. b) 14. Y Bu Bu W. Bu O W Bk or Bu Bu Y O. b) 9 vertices. 2. W Bu R Bu. C M R T S 3a) a 5-circuit. c) yes.E). 5. Bu Bu Bk R.Wendy). 15. R Bu G Y. 28 . G Y R Bk. c) yes. b) A and C each can only fill job b. Bu O W Bk or Bu Bu Y O 11. C-d. 8. BADC. Y W G R or W R Bu R. 9a) not possible (odd number of vertices. 17. DBAC.

1. {b.f) and (d. Many possibilities. (5. c) 3 (the edges incident to any vertex). {b. b) 7 splits.c.j}. (a. {c. {c. O • • O • O • 19. 17a) and b) C. 14. shortest 11. {b. 15a) 4 other pairs. (e.→C). 3.h}.g) and (c.c}.D..a.f}. k. minimal block surveillance. many possibilities.e).e}. (d. minimal corner surveillance.d).b. b) {b. b) 8. 20. .k}. {e. E. b) {b.B. (3. {c. {c.4).Solutions • 10a) • A • B • C • D • Z b) add edges from c.e) and (c. {h.f).4).4).E}. i+3. M G 12a) A C b) 1. {c.j}. c. 16a) (a.. {d.4). 16 and edges from i to i+1.e} b) (i) {a.j}. between z and n or o.g}.e}.4). 22.{b.k}. (ii) {A. (8. 29 . 13a) vertex = variety of chipmunk.j}.f}.→C) and (A. • • • • O O O • • c) O • O • O • O • Block surveillance vertices = • Corner surveillance vertices = O • O • O 21a) 5: squares (2. then make edges (A.d. if A splits into B and C.i}. T J E B D F H I K L N O Q P R S U V X W Y Z path has 11 edges. {c.h.e.d}.f}. and t to top vertex.b).d). i+2.3.. (e.6. 23a) (i){b. 26a) Directed graph with vertices named 0. 2.k}. d. (ii) {C. (4. • O a) • • • • b) • • • O 20.

2. 13-18. e-2. h) yes. a-3. c-3. subgraphs of vertices of degree 3 do not match g) no. d-3. building on the isomorphism in Example 2. c) yes. b-6.2: 1. g-1.b. k) yes. only vertices 2. c-4. e-4. d-7. 9. 25-30 isomorphic. c-7. c-4. degree-2 vertices are adjacent only in left graph. 8. 30 .c. b-2.e. a-2. a-1. l) yes. b-4. b-8. vertices form a triangle only in right graph. i-9. 5-degree and degree-3 e) No. h-10. • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • plus complements of first 4 graphs • • • • • • • • • 3a) and b) Many possibilities. Section 1. 27a) b) not possible. i) no. right graph has one more edge than left graph. j-3. degree-2 vertex in a triangle only in right graph. d) no. 5a) yes. c) no. f-8. k-5. f-6. 4. 8.a)) while all circuits have even length in right graph. no. other two distinct. h) no. a-1. h-4. f-9. a-7. c-5. 9-16. f-6. b) yes. c-5. e-4. g-11. d-1. only left graph has triangles. 19-24. j) no. b-3. 16. 31-36. g-2. c-5. g-6. b-5. 37-42 mutually isomorphic. f-3. a-5. b-3. d-4. 7-12. 6a) no. degree-3 vertices are adjacent only in left graph.Solutions S B A A B A A B A A B A b) Winning positions are 4. (a. a-1. f-4. e) no.5 of degree 2. e-6. degrees do not match. e-6. a-6. d-2. the left graph has circuits of length 5 (e. d-5. 12. 7. b) yes. complement of left graph is 2 4-circuits. f-8. d-1. b-1.. g) no. e-6. d-7. Graphs 1-8.g.d. 17-24 mutually isomorphic. Graphs 1-6.. b-3. f) yes. a-1. h-5. i-2. b → c but 5 ← 2. c-2. i-9. c-3. e-2. No. e-1. f) yes. only right graph has triangles. etc. d) no. f-2.

1. three in each. edge = joining a and b if b is adjacent to neighbor of a. 9. no. c) no..M) b) (A. d) vertex = vertex in given graph. i. yes. edge = game.U)-(I.S)-(M. b) yes. no. 14a) b) no. 13a) no. deg-4 vertex adjacent to all but deg-1 vertex.W)-(C. n is odd and so parity of degrees is preserved in complement— Answer: n . b) No.3: 1a) 12. a vertex with out-degree 2. 12.X)-(E.Solutions 10a) no.V)-(A. the subgraph consists of 13 vertices. 6. since sum counts each edge once. c) Section 1.M). has an edge going to a vertex. there is a path of length two joining a and b— note that this "graph" can have loops (is a multigraph. 12. sum of in-degrees (or out-degrees) = number of edges. with out-degree 2. b) vertex = person. 2a) vertex = person. only vertex 5 has all edges directed inward.R)-(C. then e = 2 vp edges (where 1 1 2 v is an integer since p is odd).Z)-(B.Y)-(G. 5. 10a) (A. 8. 6. Within a given conference. see Section 2.X)-(C. to a vertex’s degree— no.Y)-(D. edge = joining siblings.1.Z)-(C. not 1. c) no model. yes.U)-(B. 3.e.V)-(M.W)-(H. 12. each of degree 11— contradicting Corollary. 31 .1) but loops contribute 2. No way to get to such a vertex. 7. 4. b) 9. edge = do not know each other. Make graph with vertex = team. if v vertices. c) 8 or 10 or 20 or 40. but no such vertex in left graph. and then deg-3 vertex cannot be adjacent • • • • • • to deg-1 or deg-2 vertices.T)-(B. deg-5 vertex adjacent to all others. Solve for n in terms of m in the formula m = n(n 1)/2. n(n – 1)/4. 13.X)-(F.

delete a and (b. n = 8. and t are all ≤ 2. delete (d.{c.i}. f) possible.f }. delete (b. g) not possible (parity violation). By same reasoning as in Parallel Climbers Puzzle.c).h}.{d. and so adding edge (x.e}. v = 6. 7a) possible. j) delete (f. r = 10. r = 8.e. c) 1. 6. (d.g}. of common edges in the two circuits.1) = 6. d) not possible.g): {a.j. 5a) n ≤ 4.c).c.g}. d) 2.e. h) no.3. delete (a.e). delete (b.19b.d.f. d) no.f. b) no.e. e = 12.{b. i) no.g}. 12a) 1.f. {a.g):{a. g) no.g}.e. (b.e).c).h}. 9a) degree of vertices in K5 is 4 ⇒ degree of vertices in L(K5) is 2 x (4 .y) changes nothing.3: {c. K5 = {a.d}. e = 20.g): {a. 8.b). (f. v = 7. (e. 4. 15. b) 1.f}.b. many possible K3.f. (i. bipartition in K3. e = 10.h}.c. h) possible. b) 10d) K4.s. 32 . r = 10.d). {e.t equal 1.g.c.c.b). f) no. k) no. b) r or s ≤ 2.s. (c. Kr. delete edges (c.4: 1a) d e • • b b) 3a) yes.d.f. v = 7.e}.6. c) yes e) no.h.b.g): {a.d}. b) no.c.Solutions 14a) yes.d.f): {a. l) no. i) possible.f}. a • • c • • f A • • • B C • D • F • G E Section 1. L(K5) has v = 10 and e = (1/2)Σdeg = 3v = 30 ⇒ e >3v . c) not possible.{c. j) not possible.e): {a.f. 11a) new circuit length = sum of two circuit lengths – 2x(no. x and y must be in same component of G. b) circuit length = sum of number of boundary edges of each enclosed region minus 2x(number of edges interior to circuit). b) possible.{b. interchange positions of c and d and of e and f in Figure1.{b. or ii) r.{b. e = 12.e). delete (a.h}. r = 8.e}. (e.s.h}. e) possible.d.t is planar if either i) two of r.j). delete (a. (b.f):{a.{d.

3). If two vertices of degree 0. edge = committee overlap (person).1. 1 3.v + 2 = v + 2 ≤ 14. 24d) (d1.3). 14a) if not triangular. v . (3.K4. (5.6 = 3n .6) = 6v . G and G together can have at most 2x27 = 54 edges.3 and K5 are critical nonplanar. deg ≥ 5 and so 5v ≤ sum of degrees = 2e ≤ 2(3v . 26.v + c + 1 16. If false.5).v + 2 ⇒ e = 3v . 22b) s = no.6). 4.1 values ⇒ some value must be repeated. n = 12. use same argument as in 5a). 18a) If false.3 (≤ 3v . 25. v ≤ 2 x (no.6) .1 = e .6) = 6v . (3. of pairs of mutually intersecting circles) = 12. that is. The number of edges in b) Using a). b) 3r = sum of boundaries of regions = 2e ⇒ 3 e = r = e . c) replace inequalities by equalities in b). of hexagons. 15a) r = e .v + 1 = p+l+1. Answer = r . e = 2 Σdeg = 2v.v + 2 = 2v .octahedron. — 4s + 6(v . G is union of circuits. e ≥ 11.cube.6) = 6v .6s ≤ -12 or s ≥ 6.3). have n vertices and degrees range from 1 to n .6 = 27.dodecahedron. — G and G is 55 (=1/2 sum of degrees = 1/2(11x10) ). but by Corollary e ≤ 3(11) . r = e . 19.4). that is. 8. deg ≥ 6 and so 6v ≤ sum of degrees = 2e ≤ 2(3v .12. (4. and so if both are planar. done. an interior chord 2 can be drawn. Use same argument as in 5a) but at i-th vertex on path pick an edge that does not go to 33 . if one vertex of degree 0 ignore. Corollary becomes e ≤ 3v -3c . 6. 5v ≤ 6v-12 or 12 ≤ v— impossible. v = p + 2l. e = 2 Σdeg = 2p + 3l.6 and r = 3 2 e= 3 2 (3n .icosahedron. n numbers assuming n . vertex = committee. 1 1 Supplement: 1. 5.12 ⇒ 4s . 7.Solutions 13a) K3.s = no. of squares.s) = sum of degrees = 2e ≤ 2(3v . Graph is K7 with e = 1 2 Σdeg = 2(6v) = 21.12— impossible.d2) are (3.

v of non-adjacent vertices in G. x2. (x. only if: pick a. 19a) yes. Vertices in different components of G are directly adjacent in G . Same reasoning as in Exer. b) yes. Then x = x2.. and continue process . y = x1. odd number of vertices of odd degree c) no. 20. see Exer.2. then successively add (and consistently direct) side path from a vertex in the current strongly connected subgraph to another vertex in this subgraph. if y beat every competitor that x then y would have a greater score than x— not possible and so for some w. then x's removal disconnects G. xk = v. 11. if: a. 7 in Section 1.6. 17. between first and second visit to z a circuit is formed.e. there exists a pair u. 10. . b in different components of G-x. 15. .on or not on the circuit—would not disconnect the graph. → → 34 .. remove x0 and remaining graph must have a vertex x1 of rank 0. the removal of any edge-. vertices in same component are joined in G by a path of length 2 via any vertex in other component of G. 12. and z = x3. b) no. 22. With such a circuit. 9a) yes. consider a vertex y which beat x. If G is not a complete graph. V2 is other vertices. (y. . (i. x)). e. trace out any sequence of edges and eventually a vertex z will be repeated. Megabucks shakes 3 hands. y). 16. one can find a shortest path between u and v: u = x1. them some vertex x0 has rank 0 (see Exercise 9a) and so all other vertices have edges directed to x0 and hence have rank ≥ 0. b in different components of G-x.. 14a) possible.. w) and (w. 16. if no directed circuits. i-(d-1)-th vertices. e = 1 2 Σdeg = 13 ⇒ e > 3v . i-2. Since G is connected. → 21. x3. — — 18. sketch of proof: find a circuit in G and direct edges consistently to yield initial strongly connected subgraph. if: suppose not strongly connected with no path from a to b− let V1 consist of a and all vertices that can be reached by a directed path from a.Solutions i-1. only if: obvious.g. . b) not possible.

With more edges. c) Two sets of 3 concentric circles. i. Pick any vertex. c form blue triangle. if: obvious. Assume by symmetry that a majority (at least 3) of edges incident at x are red.Exer. 35a) • ab ac • • bc 35 . if there is a red edge between any two of a. edges in two complete subgraphs are maximized if one subgraph is Kn-1 and other is isolated vertex (K1). Following hint. and 32. n and n-1 are not each divisible by 2. set m.g). b. b) circuit starts on one side of cut-set S and each time it crosses S it must cross back to finish on side where it started. graph must be connected. 33.n/2. move xn from graph and let xn-1 be vertex with 0 out-degree in remaining graph. b) Position circles at vertices in Fig.e. each circle in one set intersects all circles in other set. continue indexing in this fashion. otherwise a. Most edges in a bipartite n-vertex graph occurs if n/2 vertices 1 1 2 1 1 in each part and all possible joining edges. 28. let 3 red edges from x go to vertices a. call it x.a). 24a) Four mutually intersecting circles. 30. c. 31a) 5-circuit or 3-edge path. n = 4k or n = 4k + 1. 7 isolated vertices. number of edges in G = 1 of edges in Kn) = 2(2n(n-1)) = 4 n(n-1). so either n or n-1 must be divisible by 4. 25a) (h.even number of edges of S.which has e = (2 n)2 = n2.Kn-1 has e = 2 Σdeg = 2(n-1)(n-2) .. 59). (h. b.Kn/2. only if: let xn be vertex with 0 out-degree (if no such vertex. 1 2(number 1 1 b) By hint.3. d) if odd number of edges of cycle are partitioned into circuits (part d) ). 1. then some circuit must have odd number of edges. if G has n vertices. there is a directed circuit. • x • • • • y b) c) similar argument to part a). c then red triangle (with x) results. Edge cover n. indep. 27. ad • • de • bd • cd 34.Solutions 23a) repeatedly remove side circuits until trail from x to y has no repeated vertices. b.

since e = 1 2 Σdeg.Solutions b) each edge of Kn is incident to n-2 other edges at each endvertex. 36 . yielding n2 edges. all deg = 2 ⇒ G is any circuit.(C1∩C2) has even degree and then repeatedly trace a path (without repeating any edge) until a vertex x is revisited and remove the circuit formed between the first and second visit to x. c) can be no vertices of degree 0 or 1 and so all degrees ≥ 2. 37a) b <--> c. a and d fixed. 2(n-2) incidences in all. e <--> f. 38a) and b)) are each bipartite graphs with different numbers of vertices on each side. b) none. 36. 40. and each edge has two endvertices and the corresponding vertex in L(G) is in two such complete subgraphs. edges incident to a common vertex form a complete subgraph in L(G). sketch of proof: to maximize number of edges without a triangle build a complete bipartite graph. c) a <--> c. then to have e = v (so that G and L(G) to have same number of vertices). b and d fixed. 2n-vertex complete bipartite graph has most edges with n vertices on each side. Sketch of proof: show that each vertex in (C1∪C2) . 39.

an Euler trail corresponds to a sequence of races in which each racer is in two consecutive races. 2a) n odd. make degree even at all other vertices and degree also even at new vertex (or else graph would have only one odd-degree vertex. by connectedness of graph the cycles overlap and can be combined into one large cycle. c) many possibilities 37 . c) r and s even 3. b) many possibilities with b and f as end vertices.Solutions Chapter Two Solutions Section 2. 2 times (to link two pairs of the 6 5a) No. add k edges joining pairs of odd-degree vertices so that the resulting graph has all even degrees. e. 12a) trace a directed path (without repeated any edge) and must terminate at vertex you started. at the single vertex fulfills all the requirements of an Euler cycle. 11. a 10-edge path. joining these edges at odd-degree vertices together (without changing the parity of the degree of other vertices) requires a set of paths. violating Corollary in Section 1. the 0-edge path starting and ending b) many possibilities. 8. 7. build Euler cycle. forming a set of cycles. K2.1: 1a) many possibilities. No. build a graph with a vertex for each racer and an edge for each race. odd-degree vertices.3). the 8 vertices corresponding to squares on edge of board one away from corner have degree 3.. b) at each vertex. 4. then remove k added edges from Euler cycle leaving k trails covering all edges of original graph. once bridge crossed there is no way back to starting vertex. for. pair off incoming and outgoing edges. an isolated vertex added to a connected graph with even degrees now has an Euler cycle but is not connected. build graph with vertex for each chessboard square and an edge joining two squares that are linked by a knight's move. this graph has the desired Euler trail because only A and F have odd degree. a set of deadheading edges must have one edge at each odd-degree vertex. 9. 6.g. 10. b) yes. yes. then add side tours by same process as in undirected case.

each graph has many Hamilton paths- . each vertex must have even degree.. A directed graph has an Euler trail if and only if at all but two vertices indegree = outdegree and at those two. b) If every vertex has odd degree in G. c) applies to Euler trails. and associated vertex for E in line graph has degree d+d' . 4. many possibilities. then when just before using last edge E before being forced to stop at starting point. Section 2.g. 14.2 other edges..a) 000 • 00 b) concatenating the first digits of each node on an Euler cycle produces desired sequence. indegree and outdegree differ by one. a) 2. straightforward. 16a) if edge E is incident to two vertices with degrees d. 19. 38 . remove added edge yielding desired Euler trail. b) if at every stage graph of remaining edges is connected.g. many possibilities. e.2 (as defined in part a) ) will be even.2 (even degree). b) sketch of proof: this rule can been shown to guarantee that Fleury's algorithm is being 01 • 001 011 010 101 11 • 111 100 10 • 110 used (Exercise 17). e. c) many possibilities. 17a) many possibilities. d' (both even degree). b) 3.Solutions 13. then d+d' . 15. Euler cycle contains a trail from any vertex to any other vertex and every trail contains a path (earlier exercise). no such Euler cycle containing all vertices. a-g-c-b-f-e-i-k-h-d-j-a. 18a) by using each edge twice. E is only edge remaining and once taken there are no remaining edges.2: 1. then E is incident to (d-1) + (d'-1) = d+d' . proof: add on extra edge so that indegree = outdegree at two unbalanced vertices and resulting graph has Euler cycle. cycle 00-00-01-11-11-10-01 (-00) produces 00011101.

no circuit: three cases at o. 2) deleting e-f.b) => subcircuit (b-c-e-d-b). at f. must use 1-6.e-c forces subcircuit a-c-d-b-a. j-f. forcing i-h-c. forcing subcircuit k-g-c-h-i-d-a-b -k. just 1) use n-o-g and delete o-i. f) path: b-c-d-e-n-i-f-h-g-a-m-j-l-k. and subciruit a-e-j-k-g-f-m-a. forces d-e-f and f-l-a. deleting at c. require subpath c-a-b-d or c-b-a-d (or same path is reversed order). delete e-k and k-l. by same reasoning. e) path: a-b-g-i-f-e-d-h-c.by symmetry use k-m-i. no circuit: rule 1 at d. delete m-g. no circuit: rule 1 at a and f. d need 6 edges collectively connected to b. must use 1-7 and 1-8 – 3 edges must be used at 1.e. at c by rule 1 must use c-a. no circuit: by rule 1 use a-m-b. by symmetry at b choose b-a and delete b-c (rule 3). no circuit: rule 1 at b. at j. no circuit: rule 1 at 4. d) path: 7-1-2-3-4-5-6. by symmetry use j-k (and j-d) delete j-i. no circuit: same reasoning as in Example 3. no circuit: at m. forcing j-e-a. f forces subcircuit a-d-g-f-a. 1) deleting e-b. c-n-d. delete at a.i. delete d-c and f-a. e. 3 edges at j. combining these 3 types of subpaths yields a circuit excluding j.e can each take one more edge but c. no circuit: rule 1 at d. j) path: l-j-a-b-c-f-e-d-g-h-i-k-m. e-f (or by symmetry 39 . no circuit: at e. by rule 2. now b.i and to g. g-k-j. c) path: f-a-d-g-b-e-h-c. g. l) path: e-a-b-c-k-j-l-o-i-h-g-f-n-m-d. must use exactly one of k-j and j-i. c.Solutions a) path: b-a-c-e-d. at k.i. subcircuit a-b-c-i-mk-j-d-e-f-l-g-a. d yields 3 edges at a and e. o-l. forcing b-f.e. b. g) path: a-b-c-d-e-f-l-k-j-i-h-g-m. use m-l-j-i-k. 3) deleting e-b.or else subcircuit 6-3-4-5-6. no circuit: to visit a.. c-d. forcing i-h-c-g-k. delete g-h. b-c. other two cases left to reader. f. h) path: d-i-e-j-h-c-g-k-f-a-b.h. same argument applies to f. at c. b) path: a-d-g-h-e-b-c-f-i. m) path: f-a-c-e-b-d-g-h-i. i) path: j-c-a-b-d-i-f-h-g-e. and now a has degree 3.h. delete c-b. e-g forces subcircuit a-f-h-g-d-i-a. delete e-f. (a. no circuit: three cases at j: 1) use e-j-f and delete j-h. by sym. k) path: a-m-b-f-e-j-k-g-h-c-n-d-i. at g by symmetry use g-f.

Solutions e-c, e-g) and using e-c, e-g forces path d-b-a-f-h, delete a-c, a-i (rule 3), forces c-d-i-h (and more) using 3 edges at d, and finally 4) deleting e-b, e-g (or by symmetry e-c, e-f) and using e-c, e-f forces a-b-d-g, delete d-c, d-i (rule 3), forces a-i, a-c using 3 edges at a, n) g-a-b-d-c-f-e-j-k-q-p-o-n-l-m-h-i, no circuit: use h-i-j (rule 1), two choices at h (vertical edge or h-g)- vertical edge, by symmetry choose h-b and delete h-g, h-m (rule 3), by rule 1 at g and m use a-g-l-m-o, delete l-n, at n use o-n-q, delete o-p, at p use q-p-j, delete q-k and j-k (now k has degree 1), similar reasoning when h-g is used at h; o) path: a-b-c-d-e-j-g-i-f-h; no circuit: either 2 ro 4 edges used between outer pentagon (a, b, c, d, e) and inner star (f, g, h, i, j): case 1- 2 edges used: at least one edge of outer pentagon not used (rule 2), by symmetry a-e not used ⇒ use b-a-f, d-e-j; delete b-g, c-h, d-i (case 1 assumption) ⇒ use remaining edges at b, c, d, g, g, h, i ⇒ 3 edges forced to be used at f and j; case 2- 4 edges used: by symmetry assume a-f not used ⇒ by rule 1 use e-a-b and h-f-i ⇒ 2 edges used at e and b- delete e-d and delete b-c ⇒ at c and d must use i-d-c-h ⇒ subcircuit i-d-c-h-f-i, p) path: d-a-h-k-i-e-b-f-c-g-j; no circuit: 3 cases at f: case 1) use b-f-i ⇒ delete f-c, f-i ⇒ use a-c-g, g-j-k ⇒ delete g-h ⇒ use a-h-k forming subcircuit a-c-g-j-k-h-a, case 2) use b-f-c ⇒ delete f-i, f-j ⇒ use e-i-k, k-j-g ⇒ delete k-d, k-h ⇒ use e-d-a, g-h-a ⇒ subcircuit a-h-g-j-k-i-e-d-a, case 3) use b-f-j ⇒ delete f-c, f-i ⇒ use a-c-g, e-i-k; now two subcases at b: 3a) use b-a ⇒ delete b-c, a-d, a-h ⇒ use k-d-e ⇒ subcircuit d-e-i-k-d, 3b) use b-e ⇒ delete b-a and e-d ⇒ use a-d-k ⇒ delete a-h ⇒ use g-h-k ⇒ subcircuit a-c-g-h-k-d-a; 5. path: m-b-c-d-e-q-a-l-k-j-o-p-n-f-g-h-i, no circuit: by symmetry at p, any delete any edge, choose p-m, at m and c rule 1 forces subcircuit m-b-c-d-m; 6. at b must use b-g or else subcircuit a-b-c-h-g-f-a is forced, also at b use b-a (symmetric

40

Solutions to b-c) and delete b-c and use c-d, delete a-e and use d-e-j, delete f-g (rule 2) and use f-j, now have subcircuit a-b-g-h-c-d-e-j-f-a; 7a) rule 1 at f, h, j forces subcircuit e-f-g-h-i-j-e,

b) by symmetry at p, use m-p-n and delete p-o forcing k-o-i, at m and n, if both use edges going up (to g) then subcircuit p-m-g-n-p, if both m, n use edge down then subcircuit p-m-k-o-i-n-p, so by symmetry use m-g and n-i, delete i-h, i-c, i-j forcing g-h-b-c-d-j-k, plus k-o-i-n-p-m-g, yields a subcircuit; 8a) graph has just three 4-circuits which cannot be split evenly between inside and outside, b) graph has three 6-circuits and two 3-circuits, 6-circuits dominate and cannot be split, c) graph has nine 4-circuits which cannot be split evenly; 9a) Hamilton circuit alternates between red and blue vertices and so must have equal numbers of each, b) follows from part a), c) (i), (ii) have odd number of vertices,

(iii) has 9 vertices in one part and 7 vertices in other part of bipartition; 10a) e - e' is number of edges that are available for a Hamilton circuit, must be enough to visit all vertices or else no Hamilton circuit can exist, b) if two vertices x, y of I are adjacent, then edge (x,y) contributes to count for both deg(x)-2 and deg(y)-2, c)(i) I = {a, c, e, g, i, n), e - e' = 24 - 12 = 12 < 13 = v, (ii) I = {a, e, f, g, k}, e - e' = 18 - 8 = 10 < 11 = v, (iii) I = {b, d, f, g, i, k, p}, e - e' = 27 - 12 = 15 < 16 = v;
1110 1111 1101 1010 0100 1011 1001 edges between vertices at 0001 corresponding positions on the two cubes

0110 0101

1100 0111

11a)
0000

1000 0011 0010

b) many possibilities: 1-0000, 2-0001, 3-0011, 4-0010, 5-0110, 6-0111, 7-0101, 8-0100, 9-1100, 10-1101, 11-1111, 12-1110, 13-1010, 14-1011, 15-1001, 16-1000; 41

Solutions

12. Follows immediately from interpretation of vertices of hypercube as binary n-tuples and x adjacent to y if they differ in one just bit. 13. See graph in Odd Solutions of text;

14. Exer 7 in Section 1.2 has all graphs (see Exer.12 in Section 1.2), all have Ham. circuits. 15. many solutions but very tedious, the following heuristic works: at each stage, look at all possible squares (not yet visited) a knight's move from current square and move to a square with the minimum number of possible squares for the next move; 16. associated graph has vertices = desks and edges = moves and is bipartite (see Exer. 9); solution forms a set of circuits which must use same number of 'red' and 'blue' vertices (see Exer . 9)— impossible since 25 vertices total; 17a) rook: starting at upper left corner, go down first column, square by square, at bottom of first column move right to bottom of second column, move up second column, square by square, continue going down and up successive columns finishing at top square of right column from which one moves back to top square of left column, king: similar to rook,

except in columns 2, 3, . . . , n-1, avoid top square, when top square of right column reached, move left along top squares of each column to return to top left square, b) rook: not possible (variation on reasoning in Exer. 14), king:(1,1)-(2,1)-(2,2)-(3,1)...(n,2), now go left and right covering rows 3 through n avoiding left square in each row except row n, finishes at (n,1) and now move up first column to starting square; 18. The graph is bipartite with 15 vertices, 8 on one side and 7 of the other, the middle inside cube and any corner cube are on different sides, but any Hamilton path would have to start and end on the side with 8 vertices; 19a) n!, b) form the Hamilton circuits by

placing vertices in a circle: first circuit formed by joining consecutive vertices, second circuit formed by joining vertices two positions apart (one intervening vertex), third circuit formed by joining vertices 3 positions apart, and so on (no subcircuits because n prime); 42

a.c. f) 3. b. then any or all edges incident to x can be inserted between e and e' in the circuit in L(G). e. Use the order of the vertices on a Hamilton path (guaranteed by Theorem 4) to provide the ranking. d. easily verified for n = 3..cannot be two last vertices. many possibilities. e' have common endvertex x. has K3. then these other vertices must all be adjacent to y and so deg(y) = n-1. if consecutive edges e. answer: 8 days. it can have at most one vertex y of degree < n/2 but if all other vertices have degree n-1. h) 5. e) 2.b. since G has ⎛2⎞ +2 ⎝ ⎠ edges. b) 4. further x must be adjacent to two consecutive vertices in H' and can be inserted between them to form a Hamilton circuit in G.Solutions c) form complete graph with professors as vertices. if x. 22. now x and y can only be on a Hamilton path if they are the last vertex on the path. assume true for n-1 and consider graph G with n vertices. y nonadjacent. Section 2.d. c) 2. f different colors so that c requires a fourth color. using part b) to form 8 Hamilton circuits out of the edges of K17.3: 1a) 3. e.g. a..clearly no Hamilton circuit. direct all edges inward at both x and y. g) 4. isomorphic to graph in 1b). c) many possibilities. 43 . so the desired x n-1 exists and it can be checked that G-x has at least ⎛ 2 ⎞ +2 edges and so by induction ⎝ ⎠ G-x has a Hamilton circuit H'. b) initially form circuit of edges in L(G) corresponding to edges on Hamilton circuit in G. 20a) result immediate. odd circuit. 21. an attempt to 3-color outer 6 vertices gives b.g. b) consider Kn-1 plus an edge from one vertex in Kn-1 to an nth vertex. g form a K4. 23.f form a K5. observe that G is either Kn (which obviously has a Hamilton n circuit) or must have a vertex x with n/2 ≤ deg(x) ≤ n-2— if not. e. d) 4. 24a) proof by induction on n.

vertices = ships. g}. b) 3. colors = open areas. (a. colors = piers. no. c) 4.f). vertices = animals. 5a) {a.f).e.d}. f. 4 colors (with 3 colors. outer pentagon (odd-length circuit) cannot be 2-colored.Solutions i) 4. graph has 5-wheel.c}. p act like a K4. 44 . f. {c.e). {j. 13. h. edges=overlap in time.f}. (b. d. b. G must be same color).c).e). 12a) vertices=experiments.h}. (a.e}. Nevada (or Kentucky or West Virginia) and its neighboring states have duals that form odd-length wheels. of Kn =n (n odd).b. (b).b. k) 2. edges = overlapping visit. b) b. 7. d fails. j) 2.k}. {b.c).d). (q).a).{a. vertices = classes. 9. A. (e. 4a) {a. Day 4: (d. 14. an attempt to 3-color the sequence of vertices a. edges = incompatible animals.g}. n) 4. (b. 2a) 3. e. {f. b) give regions inside even number of circles one color and regions inside odd number of circles the other color.{g.i form a K4. colors=observatories.d).c}.c. A B C D F E G b) . {g. =n–1(n even). b) {a. b) yes.e.h. p) 4. {c. 6a) Day 1: (a. h}. (p). 11.b.g. colors = times. i. 16. B must have different level numbers. an attempt to 3-color vertices clockwise around the circle starting at a forces j to have same color as a m) 4. (d. (b. 3. Day 3: (c. adjacent vertices A. o) 3.e). graph has 5-wheel. no. q) 4.f).b). now in the interior of the nth circle complement the two colors (replace interior disk by 'negative' of current interior). D. 10.f). c) Edge chrom. e.f}. (c. Day 5: (e. a. colors = days of the week 15a) yes. e.{d. (g). {d. day 2: (b. a 3-coloring forces a and h to have different colors. (a. g.i}. 3.d). f. similarly for i and p so that a. c. (c. vertices = banquets. e. since one beat the other and so the winner must be at a higher level.f). g}. edges = student in both classes. edges = overlapping rooms. l) 4. h}. c) {a. 8a) assume map with n-1 circles has regions 2-colored.

has at least two vertices of degree 2. although possibly one such vertex in each subgraph is incident to edge e (cannot be both endvertices of e or subgraph would not be triangulated). 2. then if G-x could be k-1 colored. proceed as in Theorem 5 using induction and the fact that there is a vertex x of degree 5. each component of the graph must be a path or an evenlength circuit. any Km. b) largest possible size of an independent set in a graph consists of a given vertex and all vertices not its neighbors. 8a) the set of vertices of a given color form an independent set and so each color class has size ≤ q. each of the two subgraphs. b) if deg(x) ≤ k-2.g. Combining these two facts. then (no. but since 6 colors are available.4 = 12.. 7a) 3-color one triangle and then extend by successively coloring a vertex that lies on a triangle with 2 previously colored vertices. length of shortest circuit ≥ 4. thus q ≤ (n-d) and result follows from part a). the max degree id 2.n for m.Solutions Section 2. then k colors are needed to color one of G's components and removing a vertex from another component will not reduce χ(G). of vertices). follows immediately. If the edge chromatic number of its components. 6. 5. also G could be k-1 colored by giving x one of the k-1 colors not used by one of x's k-2 (or less) neighbors. The chromatic no. by induction. 3.4: 1. of G is the maximum of the chromatic nos. False. n ≥3. is 2. 9a) if not connected. and so combined original graph has at least two vertices of degree 2. c) K3. x is immediately colored with a color different from those used by its 5 neighbors.n. 45 10.3 or any Kn. c) if x disconnects so that G-x has components G' and G' then one of G'∪x or G"∪x requires k colors and removing a vertex from the other component will not reduce χ(G). then the graph is bipartite. then e ≤ 2v . if bipartite. of colors)x(size of a color size) ≥ (no. b) process in part a) yields unique coloring. Use induction argument with a splitting edge e as in the proof of Theorem 1. If the vertex chromatic number is 2. 4. e. .

(k2. if interchange fails then do 2-4 interchange as in proof. if: label with numbers that are the length of the longest path starting at the vertex. but the Four Color Theorem says that any planar graph can be 4-colored (i. from Exercise 9b from Section 1.(k-1)2+ (k-1)(k-2)2]. 12. and by Exercise 8a. new inequality follows immediately from b) and fact that a2 + b2 ≥ 2ab. v) k(k-1)2(k-2)+k(k-1)(k-2)3. k-1 and direct edges from largest to smaller numbers: 16. by induction we can color G-x and G -x. with a total of n colors for both graphs (possibly less). as required. 13. if x's 4 neighbors all have different colors do same 1-3 interchange between colors of two opposite neighbors of x. 15. 14a) (i) k(k-1)3. (iv) k[(k-1)4.6k + 8) = 0 when k = 4 and so Pk(G) = 0 for k = 4. i. graph is bipartite. for any given vertex x. — — — — — — any circuit enclosing regions with even boundaries has even length. (iii) k(k-1)3(k-2). there would have to be additional triangles). c) square both sides of inequality.4. size of largest independent set in G. length longest paths since → edge (x. χ(G)q ≥ n.e. 17.. 46 . assume for n-1 and consider an n-vertex graph G. the two vertices chosen to be nonadjacent (if this nonadjacency were not possible. 2-colorable. χ(G) + χ(G ) = 2. has a positive number of 4-colorings). adjacent vertices must have diff. then the resulting graph has no odd circuits and can be 2-colored. b) χ(G ) ≥ size of largest complete subgraph in G = q.. y) implies that x's longest path will be greater than y's longest path length. and so all circuits have even length and then by Theorem 2 of Section 1. (ii) k(k-1)(k-2)(k-3).Solutions 11a) for n = 1. 1. now use third color for the two removed vertices. x has a total of n-1 edges in G and G and so is not adjacent to one of the n colors classes in one of G or G and can be added to that color class. remove one vertex on each odd-circuit. . . only if: let colors be numbers 0. mimic proof of Theorem 5. although in the other graph x may require an additional color— for a total of n+1.e. .3.

2 .G.G. b.4 . 5a) yes. b) having an Euler circuit has no relation to having a factor.4 . 6a) (a.1 .3 .B .4 .G. then in some vertex in Gk+1 is adjacent to vertices with k different colors and so Gk+1 requires k+1 colors. where A: G .2 . g). G .1 .W . A.G.G with either G .B .Solutions 18. D: G .E or D. b) none. j. d.R .1 .2 .2 .3 .W .R . F or B.B .4 . (i.B .Instant Insanity: 1. W 2-loop and G .4 .W . E: G .1 . D or A.G. if k colors are needed to color Gk.B .B .B .1 .W .R .4 . h. G . h.3 . F. R 1-loop. a Hamilton circuit is a factor.3 . (e.2 .B .1 . C or B. f.1 .G.R.R . W 4-loop. l).4 .R .1 .4 . F: G .W . g).B . Chapter 2 Supplement-.R .2 .3 .G.W . E or C. e.2 .W .3 .2 . C: G . k). B: G . 3.R .3 .4 . f). R 3-loop.1 . The new graph Gk+1 has a vertex adjacent to every possible combination of one vertex in each of k copies of Gk. 2.G or G .G.3 .B .2 . b.G. d.W . (c.W . 47 .3 .4 .2 .B . c) (a.1 .3 .R .

1).1/n} = 1/m. besides for those in this tree.y) forms 13. Start at any vertex and trace a trail (no repeated edges).1)/m}/n = 14.1)/(m . 5a) no circuits means unique path from root to each vertex (two paths would yield a circuit). 6.1) = (ml .v + 2 ⇒ e = v . 10. l + i = n = mi + 1 ⇒ l . b) since G is connected. no odd circuits ⇒ bipartite (Theorem 2 of Sect.1)/(m . i/n = (by Corollary part c) {(n . 1 m {1 .1) and n = l + i = l + (l . a circuit. but this subgraph has one fewer edges than vertices (by Theorem 1) and if G contained other edges.3) ⇒ 2-colorable. height h ⇒ each path to a leaf passes through at most h internal vertices with a choice of m children to go to at each internal vertex. it would have as many vertices as edges. 3) and so at least n/2 vertices are in one color class. e. and so 1 = r = e .i ⇒ i = (l . circuit would result) and graph is finite.1) ≤ (mh+1 . so the graph is connected and circuit-free. there is already a path between x and y.1 = mi . A tree is a 2-colorable (Exer. l ≤ mh ⇒ (by Corollary part b) 11. now start trail-building again at x to get a second vertex of degree 1.1)/(m . 48 . 9. 8. b) 2. r =1.1.1)/(m . no circuits ⇒ no K3. n = (ml . smallest 2.1). 1. 7. 21 4. Since adding any edge (x. trail must end at a vertex x of degree 1. Largest n-1. now use part a).g.1: 1a) c) 3. removal of an edge on circuit does not disconnect G ⇒ G has no circuits.Solutions Chapter Three Solutions Section 3.1)/(m . 15. since no vertex ever visited twice (if so. c) if G has a circuit. a subset of edges can be chosen to form a tree containing all vertices of G.1 in planar fashion). given l. parts of Corollary all follow immediately from the formulas n = l + i and n = mi + 1.3 or K5 configuration (or just draw tree as in Figure 3. for part b). for a total of at most mh paths to a leaf.

1.some other yg must have a leaf at level k-1 in T. ei = vi . if only subtree of yh has leaf at level k (leaves on other subtrees at smaller levels). 49 . n = (ml-1)/(m-1) = (57. 3. take 9 on right side and apply weighings in Example 4. . 17. ans: $5859. 23a) 24. or else yh.30.1)/4 = 19531. now level k in new tree rooted at yh. If x is a center of T with neighbors y1. . e = v . ym and two of the subtrees rooted at the yi's contain leaves at the maximum level k. leaves = letters. then 3rd item. 18.1 and summing e = Σei = Σ(vi -1) = n . . if balanced take other 9 with 3 vs. not x. each leaf = n-digit binary sequence ⇒ 2n letters. 20. 25a) ⎣log2 n+1 2 ⎦ = ⎣log2(n+1)⎦ . then 2nd item. y2. do same for left side. c) 12.Solutions 16. 21. Unbalancing a tree: makes sum of level numbers larger and so smallest sum occurs when binary tree is balanced in which case each level number is ⎣log2l⎦ or ⎣log2l⎦ + 1.t: for each tree. etc. 19a) internal vertices are +'s 3 + b) ⎡log2100⎤ = 7.1. d) 8 losers tournaments. sequential search. . b) 16. is + + 9 + 1 2 5 the unique center of T. if right side rises up. then yh could also be a center.t.check 1st item. and sum of level numbers is at least l ⎣log2l⎦. then x is the unique center (replacing x by any of the yi's would yield a tree with height k+1. the tournament winner retains his/her new can and so number of matches is n . if left side rises. 24. 22. b) A B C D E 26a) weigh 9 against 9.

e.1)..2: 1a) any 8-vertex path. for example with one on each side. e. any 6-vertex path. 5...g..Solutions b) Inductively follow procedure in part a).g. e. 27a) 1 2 4 T 1 2 1L 3H 2L 4L all 4H 2H fair 3L 1H b) first weighing is either one coin on either side or two coins on either side. a-b-c-d-e-f-g-h. m even => n odd. i = mi +1. x19. j f c) many possibilities.g.. x23 isolated vertices. connected (lots of spanning trees). 28.g. a b c d h e i g k a e. d) many possibilities. 2a) a star (a vertex with 7 edges to other vertices) b) many possibilities. places of d and c can be interchanged in last tree. a a a b c b d c e c) d e b c d e . e. b) many possibilities. By Theorem 3.g. 1 2 8 5 4 3 1 8 3 5 7 6 29a) 1 4 6 2 5 3 b) 6 7 c) 2 4 Section 3. 4 components. x17. in either case some outcome has more than 3 possibilities that must be distinguished in the one additional weighing (impossible. 4. b) all trees on 4 vertices (see solution to Exercise 1a) in Section 3. initially dividing coins into 3 piles of 3n-1 and 12 3T putting a pile on each side of scales. a-b-d-c-h-k-i-g-e-f-j. 50 . 3a) all trees on 5 vertices (see solution to Exercise 1b) in Section 3. c) many possibilities.1). h f g e c d b d) a 5-edge path centered at the root vertex with an additional pendant edge at the root. there are 5 possibilities if the scales balance.

Several possible solutions.0-(0. several depth-first search paths. but a depth-first search would use such an edge.Solutions depth-first spanning tree for other component has path x2-x4-x6-x3-x9-x12-x8-x10-x5-x15x20-x14-x16-x18-x22-x24-x26-x13.4) 3 (0. 13. some reached vertex would have an edge to an unreached vertex.4) (6. removal of C could not disconnect graph (spanning tree's edges connected graph).3)-(2.0) 0 0 1 2 3 4 5 6 7 (7. from x14 to x7 to x21. E 15.2)-(4. 10a) breadth-first spanning tree finds shortest paths to all vertices. graph has no circuit.3). 17. but a spanning forest of k trees on n vertices has n-k edges ⇒ one tree. 6. since it has no circuits. 11. 1 (6.3) 2 (2. (0.0)-(0. b) immediate.5). 7.4) (7. and from x22 to x11.0)-(0. right-hand-wall rule is same as depth-first search which takes leftmost branch at every intersection (corner).5)-(7.4)-(4. minimizing height b) any tree with 2 or more edges.0)-(2. otherwise a second tree could be formed. 4 (0.3) 12.0)-(4.4) —now 2 qts in 10-qt pitcher. one way continues search in 51 . 18. b) (0. b) 4. 16a)(0. subgraph generated by edges will be a set of spanning trees. plus edges from x20 to x25.0) S 14. 9a) if not all vertices reached.0)- (5. if C has no edge of a spanning tree T.2)-(5.

B.4)-(2. 19. Not possible.d.i.b. 25.).).f.d.0)-(1.e.C. restart 7-min hourglass. Begin crossings as in part a). b) a.Solutions Example 2: (2.M. then ab cross and a returns.D. Start with ABCDabcd on near shore.2)-(7.b. turn 7-min hourglass which will run for one more min.K.h.F. ad across yields ABCD/abcd.I.f.g. c) H.f. B.i. at 8 min.A.i. then take D across and bring G back. C and associated husbands be a.j. a return.k. start both 4-min and 7-min hourglasses.E.k and d. after 7-min. next C and D cross (10 min. A and B cross (2 min. abc across.h. and take G.K.C.a.4).G.c.g.d.e. restart 4-min hourglass. similarly.m.l.N.l.f.c. 26a) a. finally ac cross. 23 Let C1 be the one cannibal that can row.M and A.B. 20.N.m.g. then when M1M2C1C2/M3C3 on near shore.).) and B returns (2 min. First Cc cross and C returns.G.e.b.) and A returns (1 min.l. c.0)-(0.L.h and g.I. 21. G (goat). after 4 min (when smaller hourglass empties). k.J.H. take G across and return.j. a returns.a.L.4)-(5. then take T across. 52 .j. Now continue as in part a) with indices 1 and 3 interchanged.c.k. 24. Bb return. c return and abc across completes crossing.i.b. T (tin can). BCD cross yields Aa/BCDbcd. b. perform round trip of M1C1 across and M3C3 return to obtain M2M3C2C3/M1C1.j.l. then BC cross and Bb return. Let wives be A. D.F.e. then ab cross and a returns.E. and b (boatsman) on near shore.J. finally A and B cross again (2 min. Start with D (dog). 22.2)-(5. AB across. then AB cross and c returns. return.c.h.

5.. 6. i. would use the two cheapest edges at each vertex. a feasible tour would have to use at least 2 second-choice entries producing a new "ideal" cost of 17 1/2.Solutions sin + _ -1 + + x + c a d e c e a _ b 27.. i. cost is 20: 1-4-2-6-3-5-1. 8b) (i) 1-6-3-2-4-5-1. this ideal tour costs 1/2(sum of the 2 cheapest edges at each vertex) = 16 1/2 for Figure 3. equivalent to traveling salesperson problem except subcircuits allowed. cost is 11: 1-3-2-4-1. 2-1. b) 1-4-5-3-2-6-1.1). in each row. however this set of edges uses 3 edges in first column and in third column.3: 1.16). An "ideal" tour. if more than one circuit results. min tour is 1-4-3-2-5-6-1 with cost of 18. Section 3. cost is 14: T1-T5-T4-T3T2-T1. cost is 16: 1-2. (ii) 6-5-2-3-4-1. a -1 + x b 28. 2 2 2 2 1 2 53 .4-3 (use subset of 0 entries in Fig. 5b) of Section 3. then some edge e is on one circuit and not on other circuit. 3. 2.23 [divide by 2 since each edge counted twice]. c) 2-3-5-1-4-2. 4. call it S. 2 1 2. but removal of e leaves a circuit— impossible since n-1-edge spanning subgraph of an n-vertex graph must be a tree (see solution of Exer.e. 9. 3-4. 7a) 1-4-3-2. at least 18.e.

At each stage one compares first (smallest) remaining item on each list and selects the smaller to be the next item on the merged list. 4a) Assume a balanced heap (a heap that is a balanced tree) on the first k items has been built.1) log2n! = n log2n. 9b) O(nlog2n). If the initial heap is balanced (as described in Exer.4: 1. so average number of comparisons = average leaf level = (by Exer.e.. 54 . there will be nlog2n comparisons. 3. A binary comparison tree has n! outcomes or leaves (as noted at the beginning of this section). 2. b) 1 part a) 6 6 9 6 6 5 0 9 9 6 9 . 13 in Section 3. constructing initial heap is similar.. 4a). either item can be considered the smaller item. i. 6.Solutions Section 3. attach (k+1)-st item (new leaf) as the child of an internal vertex with one child or else as the child of a former leaf at the largest level... Outcomes are ordered lists. In case of a tie on a comparison. for n iterations of removing the root and readjusting the heap. now repeatedly compare new item with its parent and interchange the two items if parent is smaller. 5 0 4 9 7 6 2 3 1 8 5 5. then largest level number is log2n and the number of comparisons to adjust the heap each time the root is removed will equal the largest level number. log2n.

c) 15: e-j-i-h-g. c) 6: L-c-e-g-j-m-W.W). 10a) If multiple shortest paths. and adding (a.g. Define a tree by letting the first label of a vertex be its parent. (f. d) Modify part b) by deleting (a. (g. E. b) 6: L-c-d-f-g-k-W. (d.j).x) from a vertex y (y is closer to a than x) plus the distance d(y) of y from a equals m+1— this is exactly the test that the algorithm performs.W) by (m.d). length 14: c-d-h-k-j-m. a directed circuit from a to b to c to a with two edges of length 1 and the third edge of length –3.h). c) 13: L-a-c-d-f-g-k-W.b). (k. c) L.k). b) 60: replace (k.i). 5a) 5: L-b-h-j-m-W. shortest paths using the first k vertices 9 Many possible examples as middle vertices on these paths). Induct on k (i.1: 1.. 7. 3a) 31: L-c-d-f-g-k-W. 5. b) 32: L-b-h-j-m-W. 4 paths. cost is 63. then a vertex x will be distance m+1 from a if and only if the length k(y.l). (j.f).i).. 8. b) 21: d-c-h-m-r. 4a) R=57: L-c-d-f-g-k-W. order vertices so that desired vertex is first label in all cases where there is a choice. b) R=63: L-c-d-f-i-k-W.c). 3ab) 39: path N-b-c-d-e-g-j-m-R plus edges (d. 11.m). (k.j (other possibilities).e.e). 2.k). follows immediately. Same answer as in 1a) 4. choice of which path depends on which vertex is used for first label (when more than one choice possible). 31: L-c-d-f-g-k-W. If algorithm has found shortest paths to all vertices ≤m units from a. (g. 6. Section 4. d) L-c-d-f-g-k-W. (e. 55 . (g.Solutions Chapter Four Solutions Section 4.2: 1a) 59: path L-a-c-d-h-f-g-i-k-W plus edges (b. 2a) 19: a-f-g-l-q-v-w-x-y.

max flow = 150. b) now possible. 15.d}. 8. 56 .b.d each into 11a) 3 (3 edges leaving L).c}.b. c) d • 3 7 i • • • x 6 e• 3 j • • o • t 6 • y 6. P={a. P = {W}. P={a.b. vertex.b. a• 2 2 — — f • 5 7 • k 3 p • 4 5 8 • u 8 b • 7 8 g l q 11 • 14 • 10 • 12 4 m • 10 r • 4 14 8 n 5 7 • 7 13 10 10 s 4 9 8 16 1 •v 3 c• 4 12 h • • w 4 5a) max flow = 13.g}has capacity flow).c}. 11a) If the set of shortest edges do not form a circuit and they are not all in T'. c.b. 26 in Section 3.e. See solution of Exer. max flow = 50. 12. 9. Section 4. P= {a. Yes. 7. 5. then add an omitted one and remove a longer edge in the resulting circuit (like in proof of Prim's algorithm) to obtain a shorter minimal spanning tree— impossible. 7. b) max flow = 24.d.3: 1. 2a) max flow = 19. P = {f. quintuple flow in part a).c. 4a) max flow = 19. only if: verified in part b). P={a.Solutions 6. Set capacities of edges out of a and into z equal to 100 (equivalent to unlimited 8a) not possible. P={a.z}. build successive flow paths by choosing leftmost unsaturated edge leaving each vertex.c}. P={a}. 3. Let initial edge be prescribed edge and then continue with Prim's algorithm.c.b. Send 3 messengers along each path used in single-messenger problem.2.d}. split b. modification: add longest edge that does not form circuit. 10. more flow must go into h than can leave h. max flow = 13. cut with P={a. b) Same reasoning as a). d) Impossible. c) if: part b) is property of minimal spanning tree used to prove validity of Prim's algorithm. Build paths by choosing leftmost unused edge leaving each 10. Prim's algorithm gives minimal spanning tree and it will proceed in a unique fashion if all edges have distinct costs. 50. c) max flow = 14. b) 15.

z}. route 400 on a0-d2-z3-z4. 400 on a0-a1-d3-z4. 19a) Algorithm tries to reduce flow in incoming edge before 21. If each flow path crosses a cut once. b) Maximum a-z flows need not violate condition (b) but maximum z-a flows could. 27. b) yes. instead of increase flow along a flow path. max flow = 2100. 13. 28. 30a) edge sets where in-degree = out-degree at each intermediate vertex. go. e}. at most. Number of edge checks ≤ 2x(number of edges). result is max flow-min cut theorem for this network. for modeling the vertex capacity constraint. every edge is checked once at each endvertex. 12). one xi for incoming edges and one xo for outgoing edges with an edge of capacity 5 joining the two vertices. parent being its first label. then the capacity of the cut = sum of values of such flow paths = value of flow. b) Take flow in a). fo. max flow = 36. 26. 16. other edges no flow. P = {fi. max flow = 40.Solutions two vertices. and add a 2-unit flow path a-d-f-z. 22a) No flow in only edges into a or only edges out of z. min cut theorem for model in Example 6. P={a. 23a) Put flow of 1 in the two edges between c and e.bi. 400 on a0-a1-b2-z4. co. adding flow to outgoing edge. 300 on a0-c2-z4. 200 on a0-b1-c2-d3-z4. — 29a) This is max flow- b) Make edge capacities 1 and build vertex constraints of 1 (see Exer. 15. 3. di. max flow = 2200 (1100 leaving a on Monday and on Wednesday).bo. 12. A set of edges disconnecting a from z obviously cuts all flow paths from a to z. max flow = 12. a min cut must be a cutset since it disconnects flow paths between P and P and no subset of min cut does this. 31. then use same algorithm to reduce. except a and z). — 14. define tree with a vertex's 17. 32a) Have source with edge of capacity 1 going into 57 .ci. 400 on a0-b1-z3-z4. One needs to build an initial feasible flow (requires at least one edge entering and one edge leaving each vertex. see Exer.

c-A+. Only possible to be co-champions. remove these edges.F & J. but each can only win once. C-F (1). by Example 3 there is a pairing for the first night. Give edges from a and into z the appropriate supplies and demands. 8. Lo . Lo-J.'s can only hire at most 6 (one from each univ. C-J. z-a+ . d-D+. Lions and Tigers play each other 3 times. D-Bi-. f-E+. yes • L 3. D . D-Bi. C . Bi-A+. C-a. Middle edges still ∞. 33a) In step 2a of augmenting flow algorithm. Not possible. J-C+. Bo -F & G.3 2. A-G (1). Section 4. D-La (3). b) E-a+. first labels are (except for a. and make edge capacities 0 in bipartite graph. otherwise use same algorithm. E-e. all second labels are 1): Bo-a+. Lions and Tigers must play each other 3 times but each team can only win once. jointly with the other three teams. there is a pairing 58 . Only possible to be co-champions.4: 1a) several possibilities. new matching: A-c.T 2. e-E+.2 3. • T 7a) Can be co-champions with either Lions or Tigers.Bi & La. jointly with the other three teams. Bo-F (1). 10. Lo-F (2). now again by Example 3. C-F-. A . b) using pairings A-G. 9.2 • z 6.Solutions each vertex on one side of bipartite. z-La+. a-A+. 3. C-J (3).) B • 9.4 2 2 • 0 3 • L.T B. G-Bo+. Lo-J-. Lo-La. Bo-G (2). 2a) several possibilities. A-e-. Lo-J (2). A-G-.L • B. 5. no. new matching A-G. each vertex of degree n. see b).Bi & G. schools with demand of 7 Ph. Bo-F.3 a • 0 0 7. B-b.J & La. 11ab) make complete bipartite graph. have a similar sink for other set of vertices.0 5. La-Lo+.D. incoming flow cannot be reduced below its lower bound value. D-f -. b) Not possible. D-f. 4. D-Bi (2). F-Bo+. one solution: A-Bi (3 dates).

4a) x11 = 10. x21 =20. 15. x22 = 30. continue in this fashion. b) x12 = 30. x23 = 20. each with capacity m/3 and unit capacity edges to each university’s graduates. x12 = 10. x21 = 20. x21 = 20. x22 = 10. where s(S) is the set of successors of vertices in S (vertices with an edge coming in from a vertex in S). for each left-side vertex x. x13 = 20.j in the matrix has a value between k and k+1. which corresponds to a set of edges in original graph with one edge out of and one edge into each vertex. x31 = 30. 3a) x11 = 30. x31 = 10. z) is in the cut.= 20. x22 = 20. x12 = 30. x21 = 20. x22 = 10. b) x12 = 30. the edge from a to X-vertex i has a lower bound of (integer) k and an upper bound of k+1 if the sum of the entries in the i-th row is between k and k+1. b) x11 = 20. 14. b) x11 = 20. tj). 59 . x22 = 30. b) many possibilities (easily solved by inspection). tj) is in P. the edge from Y-vertex j to z is similarly defined. Make a bipartite-type matching network with X-vertices representing rows and Y-vertices representing columns. 5a) x13 =30. Start with a standard set-of-distinct-representatives matching network. x23 =10. x31. x23 = 10. |S| ≤ |s(S)|. x31. Necessary and sufficient condition: for any set S of vertices.Solutions for the second night. this condition guarantees a complete matching in hinted bipartite graph. rplace the source a with 3 sources (one for each university). x32 =20.j) has a lower bound of (integer) k and upper bound of k+1 is entry i. x32 = 20. → 18. 2a) x12 = 30. form a set of x's (right-side) neighbors. 17. x22 = 20. 13. x12 = 20. x33 = 10. since then the edge ((ti. Section 4. 12a) none.= 30. x21 = 10. the capacity of the cut is increased if vertex (ti. and the edge (i.5 1a) x12 = 30.

x23 =30. b) x13 = 40. b) x13 =30. 60 . x42 =30. x33 =20. x32 = 10. 7a) x11 =30. x42 =30.Solutions x33 =10. x21 =20. x13 = 30. x13 =10. x33 =30. x31 = 20. b) x12 =20. x31 = 20. x32 =10. 6a)x12 = 10. x33 = 10. x21 = 30. x34 = 20. x33 =20. x21 = 30. x31 =10. x23 =10. x34 = 20. x21 =30.

10 . b) 7 x 8 . (3 x 3 x 2 + 3 x 3!)/63 . 45.3 x 42. 37. 14. b) 1 x 48 + 12 x 47 = 612. See Supplement at end of Chapter 5. 2 x 85/100 x 99. 43. See Supplement at end of Chapter 5. (n + 1)4 .26) + 4 x (64. b) 1/36. 35. 16a) 362. 39. c) 36. b) 6. 64 . 9a) 4 x 47 = 188. b) 20 x 19 x 518. c) 9 x 8 x 7 .7 x 6 x 5. 11. 8a) 14. 6. 52 x 48/52 x 51.1. 27 . count ways of picking middle value and then arranging them (3x2 possibilities when middle value same as largest or smallest).2) x (n .1296) . 36. 26 x 25 x 24 x 23 x 22. 2 {28 x (64 . 45 .22) (Queen on edge of board) + 20 x (64 . 2 40. 265. 4 x 6 . 32. 26. 1 61 . each friend is or is not in the subset.4 x 3 x 2 2 2 or 2 x 3 x 4 x 3 + 3 x 2 x 4.1).(4 x 4 +24x 6 +36x 9)/2.Solutions Chapter Five Solutions Section 5.28)}.5 x 4 x 3. 15 x 10 x (14 + 9). b) 9 x 7 x 5. 4 x 10 . c) 192. 4. 15. 2 3 c) 9 x 8 x 7 x 1)/9 x 9 x 8 x 8.1 . 262 x 212. position 9. 8 23a) 9 x 8 x 7. 10 x 9 x 8n-2.4) x 6 + (m .1: 1a) 20.2) x 9)/2. 2 x 3 x 3 x 2 x 4 . 1030. n x m (4 x 4 + 2(n + m . 2a) 41.6. (112 + 102)/212. 52 x 262. 9 x 7 x 5 x 3. (12 + 22 + 32 + 42 + 52 + 62 + 52 + 42 + 32 + 22 + 12)/362 = 146/1296. 56. 44. 16x 8 x 7/2. 15.1.2)/2. 7 x 10 x 6 x 4. 22. 210– 1. 27. 25. 10. (3 + 3 + 1)/80. 33. 12. 5 21. b) (26 + 262 + 263) x (2 x 10 + 3 20. 6 x 5 x 4 . 42. b) 21. 31. d) 122. 10 x 9 x (2 – 2)/2. b) 196. 19. 41. 1 146 17a) 10 x 9 x 8 x 8. 5. x 102 + 4 x 103).first position 8 and then 29.pick the possible smallest value (3 choices).7. See Supplement at end of Chapter 5. n x (n . 5+ 52+ 53+ 54+ 55. b) 9 x 8 x 1/9 x 8 x 8. n xm x(n + m . 34. 24. 6 x 5 x 4 . c) 3 x {9 x 8 x (7 + 5) + 7 x 6 x (9 + 5) + 5 x 4 x (9 + 7)}. d) 2(1 . 3 x 29. 38. 12. 4 x 3 x 82/104 . 6 x 5. 18a) 2 x 263 x 103. 7a) 9 + 7 + 5. 3 30. 13.24) (Queen one away from edge of board) + 12 x (64 . 2 x 25/50 x 49. 3. 28a) 720.

3). 29. c) C(26. d) 13 x C(4. 18. C(13. 40.2) x C(13.2) x C(4. b) 26!/2 .2) x 4!/6! = 1/2 28. C(10.8) x 308/C(360. b) 1-vote 1/6. 25a) 2 x 5!/6! = 1/3.5). b) 33/83. 62 . 12a) C(14. C(13. b) C(9. f) 45 x C(13. 13!4/52!/4!13.3) x C(4.60).8) x 2n-8. 19. + 8)2. P(26.2) x C(5. c) C(5. 2 34. c) C(10.3). d) 1-vote 396/7!.8). P(15.5)/C(52.4)}/104. 2 9.5) x 10!.10).9)/109.5) – C(7. C(26. 26.5). 4 x 3! x C(7.2) x C(6. 42a) 4 x C(13.6). 39.10!/4 41a) {(3 x 4 + 2 x 5) x 8!}/10!. d) {C(4. C(12. 8. 1 .2) x 24 .2)x{C(6.3).6)} + C(4. 4.1) x C(5.7)/2.2)/C(50.2) + 21 x 6!.a) {C(4.2).5).1)3/C(52. 8!. 32.2) x C(4. 21. 2 x C(5.5).2) + C(20.3) .9)]/29.5).3) + C(5. b) 13 x 48/C(52.5) + C(6.2)/C(13. 16a) 48/C(52.2)2 x 44/C(52. 2-vote 1056/7!. 3 c) 1-vote 468/7!.2) + 35. P(20.4) + C(6. 2 x 5!2/10!.3). b) {30 × C(20.2). c) C(10.5).2). 3. P(8.2). 27a) C(n.1) x C(6. C(21.7) + C(9.4) /104.5). 2 x 4!2. 5 x C(4.3)x C(11.5). c) C(13.3) + C(9.1) + C(5.5). 2-vote 1/3. 24.5). . 2 x 4!/6!.5). 2 x 3 x 6.4). C(5.7).5)12/C(360.4). e) 10 x 45/C(52. [C(9.2) x C(5.3) x P(6. 11a) C(10.a) 30 × C(20.8).3).3) + 5C(4. 31a) 1-vote 1/6. C(3.2) x C(8. P(7. 10.P(10. 14a) C(10.2) x 3!. C(5. 5.5). b) 5!x 5!.3).4) x P(6. C(n. 33a) 26!/2.Solutions minus case of no one invited. C(12.5) x 21!. 30.5). 12 b) C(6.3)}/C(60. b) 2 x 8. 2-vote 1/2. 46. C(5.4) x C(5.2) x C(6. b) C(14. 36. 20a) 8!3. 2. b) {1 x 9 x 8! + 8 x 8 x 8!}/10!.3) + m x C(n-m.3) + C(5. 15.5). 52! 6. 23.8) + C(9.3) x C(5.3) x P(21.2) × (24 – 2) /104. 22.2)} + C(6. 7.3) × 9 + C(4.3) x C(6.2: 1. 47. C(30. C(18. Section 5.3) x C(5.2) × 92 + C(4. (1 + 2 . 17a) C(4.2) . 11!/4! .3x5} + {C(4.4) + C(9. b) C(n-m. .8). 2-vote 864/7!.C(5.4) x P(21.2) x 44. 38.2)/C(52. b) C(10. 13.6). 37. 3! x 6! x 8! x 5!.3) x 12 x C(4.

{C(30.2) x C(28.2) x C(4.2) + 1}/C(11.3) x C(8.1.Solutions b) {4 x 133 x C(13. See Supplement at end of Chapter 5. 50. 64a) 211.3).2) x C(21.2) x C(45. 60a) 1 . C(n.4) + C(10. c) as in b). b) {(2 x 23 + 3 x 22) + (22 + 2 x 2) + (2 x 2 + 1) + 2 + 1}/28 71.3) x C(4. x C(5.2) + C(6. 72. {4 x 4 x 3 + 2 x 4 x 2 x 1}/C(11.m + n). 66.1) + 4 x C(4.5) + C(13. See Supplement at end of Chapter 5.3). 54.6). b) {C(26.4) x C(28. 67a) C(45.k) x 10n-k.4).2) + 2 x [C(6.2) × 2 × C(7.k . x 5!/8!.2) + 222 x 23. 59.2) + C(6.k). 57.2) x C(8. C(8. 61a) C(10.2) × C(5. 77.2) x C(13.1) x C(5.3) + C(45. 43a) [C(4.6) + C(5.1) x 10/P(50. 56. 68.2) x C(24.5) x C(11.1).2) x C(k-10. C(100.2) x 23 + 23 x C(22. 63 .{C(40. 45. C(12. 2 x 2 (n-k-1) x (n-3)!. 63.C(40. b) 3 x C(30.1)2 x C(26. break into cases on the values mod 4): C(22. 2n-1. 49.2) x (20 . 44.5)].1)4x C(36.6) + 5 x C(10.8) x C(56.8).2) x C(26.k)/C(20.10) x (k . 52.1)! x 10/P(50. 62.k).1) x C(2.5) + 4 x C(10. C(6. c) P(40.2) 76.3). 7 × 6 × 4.3) + 5 x {C(11.4) x C(11. 51a) see Supplement at end of Chapter 5. b) C(6. P(C(n. b) 211.9)}/C(50.2) x C(11.5).3). end of Chapter 5.3) x C(21.4)} + C(3. 69a) (2 x 23 + 3 x 22)/28.2) x C(4. C(11.2).10) + 10 x C(40. 75a) C(10.3) + C(4. 65. 5!/2! x 7!/2!2!.C(8. C(64. b) k = 100. C(32. See Supplement of Selected Solutions at 74.3)].6). 55.7) + C(5. See Supplement of Selected Solutions at end of Chapter 5.3) .2) x 22 + C(23.2)2x C(26.3)+ C(23. 53. 48a) 1/4.5).1)}/C(52. m = 12. P(C(5.3). d) C(40.1)] + 6 x C(4. C(12.4).8) + 5 x C(21.2) × C(3.4).1)3 (each different value mod 3). C(5.2) + C(3.2)2 x 132}/C(52. 73. 70.18)/C(k. See Supplement at end of Chapter 5. 8! x [C(21. 46. C(k . 4 x 3 x 8!/2!2!.3).5).3) + C(13. b) C(10.3) x C(4. C(8.5) × C(4.5). See Supplement at end of Chapter 5.3) x 3! 47.5)/C(50.10).k .1)3]/C(52. 58.3).2). {C(4.2)}/4!.20).3).2) x C(2.2) x C(4.3) (three integers with same value mod 3) + C(30.4) .1) + C(6. b) 1 .8 x 5.5) + C(12.

Σk C(3. 11. 3a) 37. 8a) 9! x [1/5!2!2! + 2/3!4!2! + 2/4!3!2! + 1/3!3!3!].8)]/C(8 + 25 .r+s). 6!/2!3 + 3! x (6!/3!2!1!). end of Chapter 5. C(r + n . C(5. C(10. 9) . 8).12!/3!. C(8 + 3 .1. 3 x 6!/3!3! + 5. 5)×C(3+3 –1. 12. C(50.3 x C(3 + 3 . 2. 3). 17.. etc.1. see Supplement of Selected Solutions at 26a) 49. 3).1. 8).1. 21. 7. 2).1. 8). 20a) 13!/5!4!3!1!. 25.2) × C(8 + 2 – 1). 8) + C(6 + 12 . 11!/4!4!2!1!.k) x C((10 . b) 13!/9!3!1!.3)/2!3}.2 x 3 x 5!/4!1! + 2 x 2 x 5!/3!2! + 2 x C(3. 9a) C(8 + 3 . 10.2)/3!2! + C(6. 4. and k3 3's. C(8 + 25 . left side looks at all cases of k1 1's. 15.1. 16.k)).3: 1. 4) +C(2 + 12 .1. 8).2) x C(5. 10!/2!2! – 9!/2!2!.2. Counts all (1. consider the cases of: (i) 4 of one letter and 1 of another letter.1. 4). 22a) 3 x 10!/4!3!2!1! b) 2 x 10!/4!3!2! – 9!/4!2!2!.1.3)-sequences of length10 two ways: finger's rings. 24.1.6) x [C(6. 2 x 6!/2!2!1!1!. b) C(2 + 3 . (iv) 2 of two letters and 1 of another. 5!/1!2!2!.1. pick on even number of odd numbers [C(8 + 12 . and (iii) 2 of one letter and 1 of three others.1. C(5.2) x 9! x {6/4! + P(6. 6) x C(2 + 13 . b) Consider sequence of first finger's rings. while right side counts all (unrestricted) 64 . 18. a slash. 13.1. 6. b) 35[3 × (6!/4!1!1!) + 3! × (6!/3!2!1!) + (6!/2!2!2!)]. 15) .a) C(5+3 –1. then second 27. (ii) 3 of one kind and 2 of another letter. 6) + C(8 + 13 . a slash. C(4 + 6 . (iii) 3 of one letter and 1 of two others.1.1.1. 23a) 10!/4!3!2!. 8).k) + 3 . 4).1. k2 2's. (10 .1. 9). subtract cases where one party has a majority (8 or more) C(9 + 3 .Solutions Section 5.C(4 + 3 .2) x 8!/2!2!1!4 + 6 x 8!/3!1!5].4) x C(5. 19. 4) x C(4 + 13 . 3 x 10!/2!4!. 14.2) x 5!/3!1!1! + C(3.1. 2) x C(6 + 13 .1) x 9! x { 3/7! + 3!/6!2! + 3!/5!3! + 3/5!2!2! + 3/4!4!/ 3!/4!3!2! + 1/3!3} + C(5. r).1. b) C(15 + 3 . 2) + C(4 + 12 .2) x 2 x 5!/2!2!1! + 3 x 5!/2!1!3.1. C(9 + 3 . b) (7!/3!2!2)/37.

(21-4)) x 21! x 5!. 13. (5-2)) x 5! b) C(9.2) x C(13. 6) x C(3 + 3 . (10!/2!5)/5! 31. see Supplement of Selected Solutions at end of Chapter 5. second sum when last c is at end of sequence: Σ C(3.7) x C(6 + 3 . 2a) 416. 65 . 17. C((21-4) + 6 . c) 16!/4!4.1. 5 . 15).(2-k))! 34. 3 x C(21.2) x 16!/6!6!2!2!. C(15 + 3 .Solutions 10-digit sequences of 1's.1. Supplement of Selected Solutions at end of Chapter 5.13).1. 5 .1. 13). 30. c) C(13 + 5 . Since the set of books a person receives in not ordered.k)21!/7!k!(2-k)!(8-k)!(6 . 3). See Supplement of Selected Solutions at end of Chapter 5.7) x C(6 + 3 . b) [13!/5!5!2!1! x 39!/8!8!11!12!]/52!/13!4. this strategy violates the Set Composition Principle in Section 5. 16. Section 5.(3-k))! + Σ C(2. See 32.4) × 8!/2!.1. 9a) C(9.3)2/C(52. 6.13).3 x 2 x 6!/2!]. 2) x C(6 + 4 . 2's.4) × 8!/2! × 5!/2!2!. C(8 + 4 .1. 14. 11.k)21!/7!k!(3-k)!(8-k)!(6 . 8. 3a) C(13.8) x C(2 + 4 . 36). (14-5))/C(20. 12a) C(28 + 5 .1). b) C(7 + 3 . 29. 7). d) 4! x (13!/4!3!3)4/52!/13!4.1). 6).1.5) x C(13. count outcomes with one pair of vowels consecutive and subtract off outcomes with 3 vowels in a row 3 x [7!/2! . 33. 5.1. c) 4 x C(48.1. This strategy gives each person a distinct first book and then possibly additional books. b) [52!/8!37!4]/3! x 4!. 7. b) C(28 . 32). 5! × C(6 + 6 – 1. 28). 3's.1.1.1. must have b or d after each c (except possibly last c). 4a) C(7 + 3 . See Supplement of Selected Solutions at end of Chapter 5.9)/C(52. 10.1.4: 1a) C(36 + 4 -1.1.14).2.1.14) x 27. 1 – [10! – 10 × 9 × (10!/2!)]/1010. 28. c) C(32 + 4 -1. 6) x C(7 + 3 . add 'slack' variable: C(100 . b) 1. C((14-5) + 7 .1. sum is over number of b's following c's (first sum when last c followed by b or d. 15a) [52!/13!4]/4!. C((5-2) + 4 .1. 6). b) C(4.

5k + 1) .1. xi ≥ 2. b) distributions of 3 5 identical objects into 3 boxes with at most 5 objects in first box.1. 20) . (20-k)). x1 ≤ 5.1. c) C(40 + 5 . 7) . (15-3k)) Chapter 5. etc. 23.1. 7) x C((20-k) + 4 . 66 .5 x C(11 + 5 . 32a) arrangements of length n of n letters with repetition. 35.15). 26. 20.1.1. 24.1. 31a) distributions of 8 distinct items into 3 boxes. 30). Σi=1 xi = 5.5k + 1) + Σk=0 (10 . 40.1. b) C(n.2 x C(4 + 2 . x2 ≤ 4. 37a) C(7 + 4 . 31)]. 9).20) x C(20. 6 39.1. b) C(7 + 5 .4 x C(3 + 4 . Σ C(15-3k) + 4 .1.1. 9).1.1. 20) - 5 x C(9 + 5 . 21.5k + 1) . 3 x C((30-16) + 3 .1. k).3) x C(9 + 4 .1. (30-16)). 2 x C(7 + 4 – 1.2) x [{Σk=9 C((40-3k) + 2 . b) 5 x [C(32 + 4 . C((7-3) + 5 . b) C(20 + 5 . 7 30a) Σk=0 (35 . 42a) C(20 + 5 . k).1. c) C(13 + 4 . Σk=0 C(7 + 3 .1.1. 9). 7 2 b) Σk=0 (35 . c) Distribute teddies and fill out each child with lollipops: 43. (7-3)) x 7!/13!/4!2!2! . (40-3k))} . b) C(10 + 5 – 1. C(55 + 5 . 7) + 1. 36a) C(30 + 5 . Σi=1 xi = 6.Solutions 18a) P(n.1.1. 55). Σi=0 C(k + 2 . See Supplement of Selected Solutions at end of 28. 11) + C(5. 32) + C(31 + 4 . 7).1). b) P(4. See Supplement of Selected Solutions at end of Chapter 5. 4)]. C(7 + 3 .1. 13) . 29. 10). 13). 31 33a) distributions of 6 identical objects into 31 boxes.1.1. Σi=1 xi = 18.1. b) distributions 9 distinct items into 3 boxes with two items in the first box.15). 38.2 . [10!/2!5]/510. C((7-3) + 5 . b) selection of 6 18 objects from 6 types with at least 2 of each type.7) x C(13 + 3 .1.4)..1. (12-2k)). 4 x C(4 + 4 . (7-3)) x 7!/4!2!1! 22a) C(15 + 5 – 1. x3 ≤ 2. C(30. 27.1).1.5 x C(19 + 5 .2) x C(2 + 5 . 19a) 43 x C(9 + 4 . 2). Σi=1 xi = 30. 25. C(30 + 3 .1. etc.1. 40) . b) arrangements of 15 letters using 3 letters of each of 5 types. 5 34a) selection of 30 objects from 5 types.1. 30). 7 41. k) x C((12-2k) + 2 . 13 d) C(5.

4!x7!/2!x C(5 + 3–1. C(7 + 4 – 1.5).25!/6!6!1!] + 25!/10!5!3.4). C(n.k) x C(13 . b) 2 x 5! x {C(3 + 5 . 49.3) + 6 x C(n + 1. 2 x C(1 + 3 – 1. 12) + 14]. 3! x 25!/16!8!1! + 3 x 25!/14!7!4! x 24 + 3 x [25!/12!6!7! x (27 . 4) + C( 5 + 5 .2) + (4!/2!) x 22]. b) Σk=0 C(20.5 x 2n.1. C(13. k . 3-tuples where a + b + c = 15 with no letter greater than 7. n = 4. Section 5. b) 4 x n + 21 x C(n + 1.2m + 1).1.4).5: 8. Σk=0 C(13. (n-2m)) {simplifies to C(n + 1. r . ∑ k =0 5 C(k + m -1.6) x C(5 + 3 – 1.13).5) x C(3 + 4 – 1. b) 5 x 6 x 8 . 67 . 5) + 47. 59.1. 4! x C(4 + 5 .1.2) + 18 x C(n + 1.1.5).Solutions 6 c) Σi=0 C((20-3k)+ 3-1.3) x P(21. C(3 + 4 – 1. 3). 10) + 2 x 3 x C(11 + 4 . 4). 50a) 9!/2!2! . See Supplement of Selected Solutions at end of 54.1) x C(n + 1. Σ 15!/a!b!c! summing over all a. 46. 51a) 5! x C(9.2. + 5 x C(12 + 3 .1.1. 57.k)xC((r-k) + (n-m) . 10 10 56a) Σk=0 C((20-2k) + (m-2) . 3)}.7) x 9!/6!. 58. c. 44.3).2) + 6 x C(n + 1. Chapter 5. 4) + C(3 + 5 .13-k) x C(26+k. (20-2k). of Selected Solutions at end of Chapter 5. 3) + C(4 + 5 .11)4 x [C(4. 5) + 2 x C( 4 + 5 .8) + 5 x C(21. See Supplement 63. 11b) C(n + 1. C(40. 16!/4!4 x 36!/9!4+ 13 53. See Supplement of Selected Solutions at end of 62.1. C(4.k) x C(20 . 43a) 5 x 6 x 8.1. 64.1. 5).k.11) x P(13.n + m).2) x 16!/5!5!3!3! x 36!/8!8!10!10! + 4 x 3 x !16!/5!4!3!3! x 36!/8!9!9!10! + 4 x 3 x 16!/6!4!3!3! x 36!/7!9!10!10! + 4 x 16!/7!3!3 x 36!/6!10!3 55.9!/4!.1.3). See Supplement of Selected Solutions at end of Chapter 5. 13!/6!7! x [C(10 + 5 -1. b) 1.2) .k) x C(26.1.k) x (m-2)20-2k . 64.2) .k) x C(39. 1) x 5!/2!. c) (n . b) 21! x 5! x C(22. Chapter 5. 12a) 72 x C(n + 2. 8![C(21. 11) 61. 52. 3) x 5!/3!.2) x P(21. (20-3k)).7)] + P(5. P(5. 65. 60. b. (r-k)).m) x C(r + m .1) x 13! x 39!/52!.1. 13a) 2n + n x 2n-1. C((n-2m) + (2m+2) . 45.1. 48.2 x C(n + 1. 5)} c) 2 x 5! x {2 x C(5 + 5 .

1 2 f) (2n+11-1)/(n + 1). b) n x (n . 68 .C(2n. c) 3n. 21. g) (n + 1) x 2n.Solutions 14a) 0. d) 3 x n x 4n-1.1) x 2n-2. C(2n + 2. e) 0. n + 1) . 20a) k = n/2 (or nearest integer). n). 0. 23. b) k = n.

d) 15 products. 18b) x10(1 + . xxx2. c) (x + x3+ x5+ . )2(x + x3+ x5+ . . c) (1 + x + x2 + . 1x3x. . . . x3x. . . . of x6. (1 + x5+ x10+ . . 1x3x1. )2. . x2xx.)n. + xs2))( + x + x2 + x3)q-2. x2x2. . b) (x3 + x4 + x5+ x6)4. . . ). + x8)2. )7. )4. . b) (x + x2 + x3+ x4)3. . )2(1 + x + x2 + . (1 + x + x2 + . x311x. + x7+ x8). . coefficient of x27. )(x + x2 + x3 + . . i=1 16. 6. . xx311. 2a) (1 + x + x2 + x3 +x4 )5. . + x13)4. )4. 8a) (1 + x + x2 + . . (xr1 + xr1+2 + . 11. . 14a) (x + x3 + x5)3(x2 + x4 + x6)3. coefficient of x18. . + x9)6. 10. x1x3. )x (1 + x + x2 + . )2(1 + x)2. c) (x2 + x4 + x6 + x8)(x3+ x5+ x7)(x2 + x3 + . . . . . 3a) (1+x + x2 +x3+ x4+ x5)(1 +x + x2 + x3+ x4)2. 1xx3. 19. . . (1 + x)u(1 + x + x2)v(1 + x + x2 + x3)w. . . 1x2x2. b) (x + x2 + .coefficient of x27. )4. x2x21. (1 + x + x2 + . d) (x + x3+ x5+ . . . + xs1)(xr2 + xr2+2 + . (b) 5 products—1x4. )n-2. . . . . . 9. x3x11. )4. x31x1. 1x31x. 1x3x1. d) (1 + x + x2 + . . )5. 1x411. 15. e) (x + x2+ x3+ . .xi]. 13. . x31x1. (1 + x + x2 + .x411. corresponds to first picking 10 balls (two in each box). . )8. + x5)(1 + x + x2 + . + x7)5. xx31. . x311x. . . . 6 b) ∏ [(x + x2 + x3 + x4 + x5 + x6) . xx3. . coef. 11x4. x31x. . coefficient of x27. (c) 7 products—x4111. )5. (1 + x2 + x4 + . + x5)5. (1 + x + x2 + . . . 12. . . )4(1 + x + x2 + . 1x31x. x3x1. (x-3 + x-2 + x-1+ 1 + x+ x2+ x3)4. c) (1 + x + x2 + . . x21x2. xx2x.Solutions Chapter Six Solutions Section 6. . 11x2x2. . 69 . 1x41. )(x+ x3). 7. (x + x2 + x3+ x4+ x5+ x6)n. . . )(x + x3+ x5+ . .1: 1a) 7 products—xxxx. 17. 5. (x3 + x4 + x5 + . 1x4. (1 + x + . )4. d) (1 + x + . . . b) (x +x2 + x3+ x4+ x5)(x +x2 + x3+ x4)x (x + x2 + . 18a) coef. 4a) (1 + x + x2 + x3 + x4+ x5)5. of x20 in (x2 + x3 + . .

(10-6)). 12). C((r . . d) {x3z4 + x3z5 + x4z3 + x4z5 + x5z3 + x5z4) x (y3 + y4 + y5)}n. 10. 3.3)x3)50. define f1. 1 + x + x2 + . 27. 7) .3). . 16.x180)/(1 .3 x C(11 + 4 . 4).Solutions coefficient of x15. C(21 + 3 .C(5 + 3 . c) (x3y3 + x4y4 + x5y5)3 x (x3 + x4 + x5)n-2(y3 + y4 + y5)n-2(z3 + z4 + z5)n. .8) + C(m. b) C(10 + 4 . 12a) (1 . 10) – C(6. . r cannot appear in generating function.x4).3)xC(3 + 6 –1. 10).4 x C(3 + 4 . 20.1. 25. f2. 5) – C(9. 11).35) + 7 -1. 0. 3). .2: 1. c) ∏ {1 + xi + xi2 + . . 7. 7) + 6 x 1. C(7 + 6 . + C(7.1.n). . e3 = f1 + f2 + f3. 2).1.1. (xy + xz + yz)8. C(9 + 4 . 12).2) x C(3 + 7 – 1.)(1 + x5 + x10 + . c) C(12 + 8 . 26a) {(x3 + x4 + x5)(y3 + y4 + y5)(z3 + z4 + z5)}n.1. C(6 + 7 . 15) – 9 x C(10 + 9 – 1. e4 = f1 + f2 + f3 + f4. b) 1. . e) b12 x C((12. d) 412 x C(12 + 5 . + xm)n. C((10-6) + 3 . . 17a) (x2 + x3+ x4+ . C(p.C(6 + 3 -1.1. 70 . )3.1. 5). . generally xn-5.x20). f3 f4 so that e1 = f1. (r . b) (1 + 5x + C(5. 10!/2!3!5!.C(2 + 6 . b) {(x3y3 + x4y3 + x5y3 + x4y4 + x5y4 + x5y5)(z3 + z4 + z5)}n. 13) – 7 x C(8 + 7 – 1. 10) . C(13 + 7 –1. 15a) 0.x28)/(1 . 24) – 6 x C(17 + 6 –1.6) + C(m.1. . 9).1. f1 + f2. )(1 + x10 + x20 + . 10) + C(9. (1 + x + x2 + . C(24 + 6 –1. 13. d) C(11 + 2 . 24a) coef of x20 in (1 + 5x)50. b) (1 + x12 + . 22. . 8.4).1. + xii}. 4. . 14) – 4 x C(7 + 4 – 1. + xm2)n. 12). 8) 6. 21) . C(14 + 4 – 1. C(15 + 9 – 1. ).3). C(5 + 9 – 1. 11) + 3 x C(12 + 2 . 11a) C(10 + 10 . .2) x C(10 + 6 –1. b) x20(1 . 14. 17) + C(6.2)x2 + C(5. C(m.m). cannot have a variable number of factors. 28a) (x1 + x2 + .1. .2) x 9.9) .35)). 6) .1. 5.1. 2. C(9 + n -1.1. c) 0. i=1 n 29. 23. e2 = 21. now e1 + e2 + e3 + e4 = 4f1 + 3f2 + 2f3 + 1f4. 3).1. 6) e) C(4 + 4 . n! Section 6. )5(1 + y + y2 + y3)5.1.

30. i=1 13 25.C(4. (5) is not defined for x = -1. 2+2. (1 . 6) + 4 x 6 x C(3 + 10 . 22.1. 1+1+1+1+1+1. m + r).(t/2)s+1 m ( 1 . 28.1. 3) + 4 x C(2 + 10 .1). n1).3 x C(6 + 5 . 6) + 3 x C(2 + 5 . )2. 3) .1. C(14 + 10 . . 10). 2a) 1 1 1 3 5 2 1 . ∑ (-1) i x C(50. 3)]. if r odd. . Section 6. . 7) .7 x C(3 + 7 .C(n1 + 6.x6)-1 = 1 + (x + x2 + x3+ x4+ x5+ x6) 33a) C(m + n.C(2 + 3 .1. 4+2.15) + C(n.1. r/2).x6 .1. C(8 + 7 .1. .9) 27. 9) + C(4.2) x C(3 + 4 . 23. 10. C(10 + 5 . 14) .4 x C(9 + 4 . 12) .1. b) 0. 3+1+1+1. )(1 + x + x2 + .(t/2) ) . )2.x5 . . pn.Solutions b) (1 + x + x2)(1 + x + x2 + . . 4+1+1.1. 18a) C(r . 37. 5+1. 8 . 17) .1. .[C(10 + 4 .x2 . . .1.1. 1 .1. 19. 5)}/C(25.4. C(6 + 3 . b) (1 + x)(1 + x )(1 + x ) .1. 21a) C(15 + n . C(12 + 5 .1. 9) + C(10.10 x C(9 + 10 .2) x C(3 + 6 . 3+2+1. + C(6.1. r even. 4)] x [C(15 + 4 .C(7 + 3 . 2) + C(6.x3 . 60.1. 2+2+1+1. 10) . PX(t) = 1 (2 )m 40b) pn + p2 + p2n2 or pqn + p2n2. 1 36.1.1.1.1.6 x C(9 + 6 .1. b) C(n + 1.3i) + 50 .i) x C((40 .8).8. 7). 6) . 34a) C(n2 + 6. 7) 20. b) C(n. (40 . 26. 8) . C(15 + 10 . 1+1+1+1. 10) - 6 x C(7 + 10 .x4 . 38c) (2 t)5/(1 . . .1.1. {C(17 + 9 .x4 1 . + C(5. 29. 10) . C(10 + 3 .1.2) x C(2 + 5 . b) C(r/2 + 8 . 15) + C(10 + n . 2+1+1+1+1. 3). 15) .1. r/2).1.1. 3)]/2.1.x . C(n.1. n2) . 3+1. 15) . [C(15 + 6 . m/p.6. 15) . 10) . c) 3n.2) x C(6 + 10 . 11) + C(9.1.C(3 + 3 . c) (1 + x2 + x4 + . b) 11 partitions.9 x C(11 + 9 . (10-2k)). + (x + x2 + x3+ x4+ x5+ x6)2 + . 2).5 x C(7 + 5 . 2).1.2) x C(3 + 10 .1. m + 1). Σ C((10-2k) + 2 . d) (pt)m/(1 . 35b) 2.4 x C(4 + 4 . 39b) n/2. .3i)).1.1. 2). 2+2+2. 2+1+1.1.10).2) x C(5 + 9 . 1-x 71 .1. 24. 3).1.3: 1. a) 5 partitions. n even.qt)m-1.4 x C(10 + 10 .t)5.2). 3+3.

10a) 3r . b) 2610 . . = (1 .x )(1 . 6. Section 6. 9a) 2210 + 4 x P(10.4 x 2r + 6 + (-1)r) . 1 13a) ex . odd-parts diagram.x 7b) (1 + x)(1 + x 1 .2r + 1)/2. 8. . b) 1/4(3r + 2 + (-1)r).x8 1 .4 x 2310 + 2210. 2 4r.2 x e(n-1)x/n + e(n-2)x/n.x6 1 . b) 4(3r . + x )(1 + x + x ) .4) x 226.x3 )(1 . b) (x3 + x6)/{(1-x2)(1-x4)(1-x6)}.2) x 228 + 4 x P(10. .x4 ) .3 x C(r. (1 . (1 + x)5e21x.x)(1 . odd-parts diagram. etc. x2 x3 x4 5.4: 1. .Solutions 3. . b) 1 .x5)(1 .x4 1 . 48 .2 x 2 + 2.3r . c)same as b).x3) . .x2) . delete the first row and column of the self-conjugate diagram and use the number of dots in the reduced self-conjugate diagram to define the second row of the distinct. .x3)(1 . 72 .2) x C(r .x2 )(1 . (x + 1!2! + 1!3! + 1!4! + 2!3! + 2!4! + 3!4! )e . .2). (4r . b) Σ (-1)k x C(n.k) x rn-k.x .38 + 8 x 37. r r 7a) 1/2(3r + 1).x10)(1 .x)(1 . (1 + x + 2! + 3! + 4! )13.x25) 1 6a) 1 1 2)(1 .4 x 2510 + 6 x 2410 .1) x 229 + 6 x P(10.2. . (1 .x5). 3.x)(1 .2) x 2r-2 + 3 x C(r. 1 x16 6 9 12 24 7 14 4a) 2)(1 . 19a) multiply the partition generating function = (after canceling) 1 . c) 3 . (1 .x7) .x)(1 .x 5.x3)(1 . (2! + 3! + 4! )6. (1 + x + x2 + x3)(1 + x2 + x4 + x6)(1 + x3 + x6 + x9) .3) x 227 + P(10. b) (1 . .x 2 2)(1 + x3) .1)n. . (1 . . 1 11. coefficient of x13/13!. (1 + x + 2! + 3! + 4! + 5! )5. x2 x3 x4 x5 x2 x3 x4 2. 21a) let the number of dots forming the first row and first column (= 2k -1 if first row and column have length k) in a self-conjugate Ferrers diagram be the length of the first row in the distinct. x3 x4 x5 x5 x6 x7 3x 12. (1 .x2(x + x + x + . . 4a) ex. . 1 (given just before Example 1) times 1 .xn) .

x)3. e) 4!x4/(1 . e) 4! x C(n + 1. m2 = m* = (n2 . b) 13(n + 1). d) m1 = m* = μ. 14. (2n + 1)n(n + 1) .ar-1.x) 8. r3 = P(r.1). c) 3x(1 + x)/(1 . 2 4a) r2 = P(r. 21.x)2 + 7/(1 . 1-x 5a) (4x2 . 9x3 8x2 2x + 4 3 + (1 .n)/4. C(n. b) (1 . x(3-x)/(1 . 2 1 Section 6.2) + 7(n + 1).x). b) loge(1-x).2(n . eμ(x-1). c) 3.x).1)n-m.m) x (ex .2).5).1 .2)r]/nr.3x + 1)/(1 . 15a) ex .x)5.x)3. 6. ar .2) + P(r.x) (1 . b) 13/(1 .x)2 .1)r + (n . b) 2a) C(n + 1. 73 .x)3.x)2.2) + P(r.5: 1a) x/(1 .Solutions b) [nr . 22c) m1 = n/2. h(x) xh(x) -1 or x . (1 .1).x)-1.3) + P(r.1)/x. d) 3x/(1 . d) C(n + 1.

k = an-2. b) an = an-1 + an-3 + an-5 .an-4 . 21. an. an = 2an-1 + 1. an = an-1 + an-2. 7. an = bn-1 + cn-1.k-1 + an-k. 6a) an = an-1 + an-2. an = 2an-1 . 35. 5.an-9 c) an = 3an-1 .1: 1.k-1 + an-1. cn = n-digit binary sequences with no 74 .bn-1 . 31. an = 2an-1 + an-2 . an = an-1 + 3n-1. an = 2an-1 + 4n-1 or an = 4an-1 + 2n-1. 32. 19a) an = an-1 + an-3 + an-5. 25. 29. ak = k+1.2.Solutions Chapter Seven Solutions Section 7.m.k-1. . b) a12 = 1.an-3. an = 3an-1 .k-1 + an-4.3)-digit sequences with no consec. an. 33.k.3000.cn-1. 30. bn = such sequences starting with a 1. an = such sequences starting with a 0.m.bn-1 .m-1. 3. 22.k. an = an-1 + an-2.k-1 + an-6.k = an-1.k-1. an = 2an-1 + 2n-1.k = an. b) an = an-2 + 2an-3 + 4an-4. c) an = an-1 + 2an-2 + 2an-3 + 2an-4 + .k-1 + an-2. an. an = an-1 + 2n . c) an = bn-1 + cn-1. (2nd and 3rd terms count sequences that start with 001 followed by (n . bn = cn = 4n-1 . an = 1.k-1 + an-3. 34a) an = 3an-1 + bn-1. an = 2an-1 + 2an-5 + an-10 + an-25.k = an-1. bn = 2bn-1 + 2an-1. b) an = 2an-1 + 2an-2.06(an-1 +50).k-1 + an-4. 16a) an = 1.an-6. 12.an-3 + 2n-3. 20a) an = 2an-1 + an-2. an = (2n . a1 = 4. n > k. b) 43. 17. 9. 2a) an = an-1 + an-2 + an-4. an = an-1 + an-2. 23.k = an. .1)an-1. 24. 11.m. 26. 13a) an = an-1 + n. c) an = an-1 + an-3 + an-5 . 27. an = 2an-1 + an-2. an. 28. an = 4an-1. bn = 3bn-1 + an-1. 15.m. b) an = 2an-1 + 2bn-1.k-1. an.k-1 + an-1. b) a5 = 10. 18. an = 2an – 1 + 2an – 2 + 2an – 3.cn-1.k-1 + an-3. 0's. an = an-1 + an-2.08an-1 . 10. b) 21 years.k-1 + an-2. 4a) an = an-3 + an-5 + an-10. bn = cn = 2n-1 .

2. n ≥ 2. an = an-3. Δpn = Δfn/fn. bn = an-1 + bn-1. an = 2an/2 + 1. 42. c) 3n2 . + a2an-2. e) An4 – 5n/7. an = 1 9 3a) an = an/10 + 1. 36. k) + f(n . b) 2n .Solutions consecutive 0's.k. 1 4.1). an = 50n⎣log2n + 2⎦. nickel or dime on successive days with a penny on the first day.3: 1. 9a) Pick largest from first half and largest from c) an = n – 1. 2. an = 2an/2 + n . b) an = 2an/2 +1. c) an = 2n – 3.0 = an-1. pn = ways to hand out a penny. n even. f) An2-3n. an. Section 7. 6. an = an-3 +⎣ 4 ⎦. dn = pn-10 + nn-10. an = 2an/2 + 50n. 39.0 + an-1. . an=an-1+2.1. an = an-2a2+ an-4a4+ an-6a6+ . + a3an-1. n not a multiple of 5.1. 0. b) An1/2 . an = 3 x 2n. an = 2an/2 + 100⎣log2n⎦. an = n . an = n log2n . second half and compare. 38. an = bn-1 + cn-3 .1. an = an-1+1. an = 2an-1.2: 1a) An . an = 3an-1+2an-2+ .k.0 (an.1.k = an. . 46a) an = a0an-1+ a1an-2+ . 44a) f(n. an = 100[3n – ⎣ln2(n-2) ⎦]. 41a) 3. an. c) the first k integers in the sequence will form a set of consecutive integers. the (k+1)-st integer can be the next larger or the next smaller number to extend this consecutive set. c) A + 4n . 40.5.k.k. 75 3a) an = 5 4n + 2 3 5 (-1) n . b) an = 2n-1. an = an-1a1+ an-2a2+ an-3a3+ .n + 1. . 43a) an = 2an-1. n+1 47.k. b) an = 10an/10 + 1. a0= a1= 1. pn = nn-1 + dn-1.k.1.1.⎣log2n⎦. + 2a1 +1.k-1.1) = such sequences starting with a 1 (0). where an. nn = pn-5 + dn-5. 2. . 8b) an = 2an/2 + 3. d) An – 2. 6n – 6.1. 45.0 + an. + an-1a0 . . cn = cn-1 + cn-2. n a multiple of 5 but not 10 or 25. n a multiple of 10 or 25.k) = f(n .1 = an-1. k . . 37.3n + 1. b) bn = an-1 (by part a).k. 2. nn and dn are defined similarly. an = an-1a3+ an-2a4+ an-3a5+ .08)n x 1000. an = log10n. a1 = 1. 48.k. n-9. (1. . Section 7. + a1an-1.2n. an = an-1.an. . 5.

2 12. an = (1/6)(3+2 x 31/2)[(1 + 31/2)n +(1 . Section 7. 3a) g(x) . 5 1 b) an = -3410 x 1.3) + C(n + 1. an = an-1+ 2. an = -3(-1)n + 2n + 6. 6. (1 .3n . an = 2an-1 + 2n-1. an = C(n.2 .x .x . 7. an = 0.2.2).1 .1) . . c) 3 x 2n . d) 15 x 2n .3) + n + 1.1 = xg(x)2. 3 x 2n .1) .3) + 3. an = 2an-1 .Solutions b) an = 1.8n . 18.20 + 40 x 2n. an = 2+ 2 4 (1 + 1 1 4. c) g(x) . 19a) an = 1 . c3 = 12. an = C(n + 1. 4.1 . d) an = 2n + n x 2n. b) an = 2C(n + 1.2x2g(x) + 1 .4: 1a) an = 3C(n.1 =xg(x) + 1 .x .5: 1a) g(x) .3) . c) an = 2. c) an = 2 x C(n + 1.2. 2) + n. an = 2an-1 + an-2.n .2n .2 n + 1. 1 b) g(x) . Section 7. 1 11. c1 = 9. 5. 2.an-2 + 10 x 2k.x) 2a) an = 2n + 1. 0. an = 960n .05n + 200n + 4200.1 . 11. b) 3 2n + 3(-1) n.1 = 2xg(x) + 1 .x . 9a) -3n + 1.3C(n + 1. an = 2C(n + 1.2) + 1.2).1 . an = 2an-1 + 2an-2. n-3 5.1 = xg(x) + 2x2 1 3 .2) . an = 2n + n x 2n-1. d) an = 2 n2 . n even.2x . 17. d) g(x) . pn = 3 x 2n .1 = 3x(g(x) . c) g(x) . b) an = n! . c) an = 6C(n + 2. pn . 7. 2x 2x2 b) an = 2 x 2n . 8a) an = -3200 x 1. c2 = -18. c4 = -19. b) g(x) .31/2)n].2n2 . an = 2n.n + 1. 3) + 1. c1 = 7. an = an-1 + 2C(n.pn-1 = 2(pn-1 . 3. c2 = 10.2 4 (1 - 1 1 ( 5 2 + 2 5 ) .x)2. 1 2) n + 2) n . an = 1250(1. 4a) Fn(x) = xFn(x) 76 .04)n -1250. an = an-1 + C(n.2(n2 + 3n) .x = 1 .12. A1 = 2.1. an = 3n .05n + 210n + 4410. 6.k) + 1}.2x(g(x) . n odd.pn-2). an = an-1 + Σk=1 {k x (n . A2 = 1 1 1 ( 5 2 -2 5).2) + 1.4) + C(n. A x 2n + B x 2n + 2 n + 3 25 4 13.x2 = (g(x) .

Fn(x) = Fn-1(x)/(1 .k . k .Solutions (-2x)n 2n . an = a1an-1 + a2an-2 + . an = 3an-1 .k-1 or 13.1.k = n x an-1.1) .k = an-1. (1 + x) dx = Fn-1(x).(r . a2 = 1. an. an-1a2.k-1 + qan-1. an = bn = cn = 3n-1. 77 . 3 Fn(x) 16a) sn. an = 15(4n . similar to recurrence relations in Example 5 except 2 with 4n-1 replacing 3n-1 and a1 = 0 instead of a1 = 1 (still b1 = c1 = 1). an = n .k = C(k + n . an.k). a1 = 1. n ≥ 2.2xFn-1(x). n . b) g(x) = ee -1. an-1a1. Fn(x) = xFn(x) + Fn-1(x).an-3. Fn(x) = (q + px)n. . bn = an-1 + bn-1 + cn-1.1).r = sn-1. Fn(x) = 1 7.k-1 + an-1. 11.1 C(2n . Fn(x) = . n even. x 15a) an = Σ C(n . a0 = -1.2. (1 . .1. 10. an = 15(4n .r-1. 1 + 3x 8. an = n C(2n .4) . n ≥ 3. n odd.k = pan-1.3xFn(x).k + k x an-1. 14. .k. . 12a) an.3x + x3 1 9. n . g(x) = . a1 = a2 = 0 .k = an.k-1.r-1 . an = a2an-1 + a3an-2 + .2).1)an-k.4. cn = an-1 + bn-1 + cn-1.x). an = an-1 + bn-1 + cn-1. 1 .1)sn. an. b) Fn(x) = 2Fn-1(x) . an.x)n (1 + 3x)n .

1: 1. {C(52. (200 + 500 – 100)/1000.10) + 24. 3 x 9n . 40. 28. . 1 .10!. 22. 20.6) – C(40.2) x 1. 3. 1 . 14.2) x 4! .9 x 8 x C(3.52 x 264.4) x C(7.10) – [C(27. 69 . 6.C(6.C(13.2: 1.3 x 9m + 3 x 8m – 7m.6). b) 26! . 30a) 0. 6!/2! + 7!/3! – 5!.7) x 4!.[C(11. 9. 3 x 5! – 3 x 4! + 3!.4) x 7!/2! + C(11.150. 320 . 45.3) x 8!/2!2] . 30 .5!3. a contradiction. 17. 5.100 .70 .22!.2) x 49 3. 2.6) x 5! + C(11.C(6. 6!/2!3 .140 . 29.6) x C(4.10) + C(25.C(6. 27.3 x 80 + 3 x 30 . C(52.C(13. 33. N(Y) = N(Y-K) + N(Y∩ K) = 50 + 20. 8. 19.200 . Section 8. 31.yields 70% who like chess.4) x 2! . b) 40%. [2 x C(11. 13. C(37.15 . 109 .40 + 28 + 20 + 8 . 7. 10. b) 30.1/n!. [2 x 7!/4! + 5!/2!] + 4!.4. b) 200 .23! + 22!. b) 80 .10)] + C(15. 3n . 11a) 200 . 21.56 .2)).60. 12a) 2 x 24! . 3 x C(6.10 .6 + 5 + 3 + 2 .1) x 59 + C(6. 23.23 x 2!.{2 x C(6.2 x 9! + 8!. 15!/3!5 . 10! .3) x 39 + C(6. 15a) 200 .200 .6).3 x 2n-1. 4 x (93 . 78 .2 x 30 + 15.3 x 8n + 7n.1.15.(9 + 1)/29.Solutions Chapter Eight Solutions Section 8.6) – 3 x C(48. 2. 2 x 5 x 265 .20 = 25% like chess and tennis but not bridge and similarly 25% like chess and bridge but not tennis and 20% like all three.3) x 4!] + C(11. 2 x 8!/4!2! + 10!/4!4! - 25.3 x 220 + 3 x 1. 25.5) x 6!/2! + C(11.4) .3 x 2 x 4! + 3 x 22 x 3! . 280 .10) + C(12. 5. 13 x C(48.6) + 3 x C(44.3) x 3! + 6!/2!2}+ C(6. 2.1)6}/C(52.3 x 5! x 10!/2!5 + 3 x 5!3 . 34a) 39. C(22.10) + 1.3 x 5!/2!2 + 3 x 4!/2! .24! . 4. 36. n = 100.3!. 600 .100 + 30.5). 32. 26. b) 12. 18.4) x 29 . 20. 35. 5! . 16a) yes. 10m .5) .

k) x C(5(n-k).4) .C(6 + 4 .1.1) x C(37 + 4 .n.2) x P(5.4) x P(5.5) .9) + C(4.2) x C(6 + 4 . 28) . b) {C(4. 16.C(4.C(5.1) x C(39.13).5) + C(6.1.1.{(C(6. 10!/2!5 x {10!/2!5 – 20.2) x n + 2 x n2} . 17) + C(4.13).3).2) x n3 . 17. 21.5) x 5!. 6) + C(11 + 4 . 5a) {C(52. 13.C(17 + 4 . C(5.C(12 + 4 .5)} x n. 22.3) x 4!/2!2 + C(5.2) x C(6 + 4 .1.1.C(5. 630 . 68) . 12.C(5.1) x n4 + C(5.4) . ≈ 26!2/e. 6.1) x n4 + C(5.C(6. 84 .1.2) x 82 .C(5.13) . C(11 + 4 . 26! .3 x C(17 + 6 . 14.{(C(7.1/e. 18.1. 17) . b) C(68 + 4 .1) x 5 x 8!/2!4 + C(5.1.(21 + 15 + 9 + 6) . r).3) .2) x n + 2 79 .C(5.2) x n3 .2) x 9!/3!3 – C(5. 22) .2) x C(26. 12) .3) . 11) + C(6 + 4 .1. 9.4) . C(24 + 6 .3. 1 .C(4.13) . ≈ 10!2/e.3) x n2 + {C(5. 7. c){C(52. 6).2) x 4!/3!1! + 10. 11.C(4. 23a) n5 .1) x C(48.2) x n3 . 28) .{(C(5.4) .13) + C(4.13)}/C(52. 10a) C(28 + 4 . 6) + 1.1.1) x C(9 + 3 .C(5.2) x C(44.5)}.{3 x 23! + 24!} + {2 x 20! + 2 x 21!} . c) C(28 + 4 .11) .3) x C(13.4) + C(10. b) n5 . c) n5 .3) x C(13.3) x 8 + 3 x 82) + {C(6.17) + 3 x C(10 + 6 .6)} x 8. 19.1.2) x C(26. 6).1.C(22 + 4 .1) x n2 + n3} + {(C(5.C(6.2) x 6!/2!3 .1.2) x 8.1.13) .1) x 8!/2!4 + C(5.C(4.13) + C(4.3) x 6!/3!2 + 5 x 5! – 5!. Easier to solve without inclusion-exclusion (4 digits appear once or 1 digit appears 3 times or 1 digit appears 4 times): P(10.1.3) x P(5.k) x (n .3) x C(40.4) .1) x C(39.k) x (2n .C(4.C(3 + 6 .3) x P(5. assume n even: Σ (-1)k x C(n/2. 15!/3! 5 – 5 x 5 x 12!/3!4 + C(5.1.3) x n2 + 3 x n3} + {(C(7. 8.13)}/C(52.4) .13) + C(4.18!. 37) + C(4.(315 + 210 + 126 + 90) + (105 + 63 + 45 + 42 + 30 + 18) . 10!/2!5 .1) x C(17 + 4 .k)!/2!n-k.13) .C(4.1) x n4 + C(7.2) x P(5.2) x 6!/2!3 .C(4.1.13).14) x n + 14 x n2} .13)}/C(52.C(7.3!.{(C(7.10) .24) .1) x 83 + C(6.C(5. 15. Σ (-1)k x C(n.13) .3) .13) + C(36.C(5. Σ (-1)k x C(n.1.k)!.3) x 4!/2!2 + C(5. 9!/3!3 .C(5.Solutions 4.3 x 7!/3!2 + 3 x 5!/3! .1.C(4. 27.

6) + C(2.2 x 6!/1!1!1!3! + 5!/1!1!3!} .2).7!/1!3!3!} 29.2) x 53. 2 x C(5.7 x 5! + 18 x 4! . 3. 6. D5 = 44.k)r }/n!. 31.8) + C(6.8). D7 = 1854. D4 = 9. D8 = 14.4) 33.2) x P(C(4.4) .k)!.1 . D3 = 2. 80 9. m6 .3) x 8!/2!2 + 3 x C(5.{(C(8. b) 10/20. Σk=3(-1) k+1 x C(k-1.7)} x n 24.496. D10 = 1. 4. 34.k) x n!/j!}. d) 1 + 8x + 14x2 + 4x3.7) + 1) x 5. 28.1 . D9 = 133.1) x C(4.6) . C(P(6.3) x C(n.4) x 5 + 4 x 52} + (C(8.2) x C(26.13 x 2! + 2 x 1!.3).7) . 26.k) x (2n . 25a) Σ (-1)k x C(n .2) x P(C(5.1)! .4) – 6 x P(C(5. 7.3!}.k) x (21-k)n. D2 = 1.7 x 4! + 16 x 3! .2) x 9!/2!3 – n n 47.6 x m5 + C(6. 6! x (6! . C(52. b) 1 + 7x + 17x2 + 17x3 + 6x4. 27. {Σ (-1)k x C(n. then Σ (-1)k x C(6. C(5.833.3) x m3 + C(6.4 x 6!.7 x P(C(6. 5. 5 x 5 board with darkened squares on main diagonal.5) . P(C(7.6) . Σk=3(-1) k+1 x C(k.2 x 1!)/296.C(8. 36c) D1 = 0. Σk=0 k x {Σj=k(-1) j-k x C(j.46 x 4! + 32 x 3! .5) x m + C(6.25) x 5 + 25 x 52} . 35.k)r.9 x 6! + 30 x 5! .2) = 21.3). e) 1 + 12x + 46x2 + 56x3.C(4.8) – 6 x C(P(5. 32. c) 1 + 7x + 12x2 + 4x3.k) x (n . 7! . D6 = 265.k) x (n .k)!. P(C(6.21 x 3! + 11 x 2! .k)r. + 3 x {7!/1!1!1!1!3! .C(6. Dn.3).k) x (n .3) .3).C(7.4) x 7!/2! .k)!}.{(C(8.1.6) .4) x 52 + 4 x 53} + {(C(8. + C(6. 9!/3!3 .7) + C(7. 5! – 8 x 4! + 20 x 3! – 16 x 2! + 4 x 1!.{6! .3) x C(13. 5! .3 x 5! + 3 x 4! . 3.k) x (n .Σ C(n.1.C(8.2) x C(n. 5!n x {(2n .3).6) x m.961.2). 5! – 8 x 4! + 22 x 3! – 24 x 2! + 9 x 1!.6).3 x {8!/1!1!3!3! .4).7). given that C(2 + 6 . Section 8. 39.Solutions x n2} + {C(7. 38. Σ (-1)k x C(n. . 2a) 1 + 8x + 21x2 + 20x3 + 6x4. 8a) 6/20.1) x 54+ C(8.3). 55 . n-1 n-1 37.C(6.334.2) x C(P(4. 30.k) x 2k x (2n .2) x m4 .4) x m2 .2).3: 1.8 x 2!.

11a) 4 x 5 board with darkened squares in 4 positions just to right 5 c) Σj=k(-1) j-k x C(j. k . Σ C(n.k) + C(2n .k. 13.1.Solutions + 10.k) x C(n .k)2 x k! x xk. b) (x + 1)4.1)}xk.1.k .j)!. 81 . 2 x 2 array of darkened squares and "L" (a column of 3 squares beside a single square) both have 1 + 4x + 2x2.j) x (n . n 14c) Σk=0 {C(2n . of main diagonal.

d) yes. b) 4 symmetries. c) (ad)(bc). (d) not transitive. and axis 30˚ counterclockwise of vertical. e) yes. (b) yes. 5a) all Ci left fixed. (e) not transitive. e. any rotation with any flip. c) 1 symmetry (just the identity). f are flips around vertical axis. axis 30˚ clockwise of vertical. ⎡1 0⎤ ⎥ . column second symmetry: a b c d e f a a b c d e f b b c a e f d c c a b f d e d d f e a c b e e d f b a c f f e d c b a c) let a = ( 1 2 3 4 ). g) (1346)(257). d = ( 1 2 3 4 ). 1234 2143 3412 4321 b) straightforward. c) yes. ⎝ C1C3C2 C5 C4 C6 C9C8 C7 C11C10C15C14 C13C12C16 ⎠ • o 2 • • • • • 3 o o • 4 • • o 5 o o • 6 o o o 7 • o o 8 o 6a) • 1 b) (i) (1)(234)(567)(8). d) π2. a b c d a a b c d b b a d c c c d a b d d c b a 12. 120˚. b) yes. 0. 10a) π1. ⎛ C1C2 C3 C4 C5 C6 C7C8 C9 C10C11C12 C13C14C15C16 ⎞ . (c) not transitive. 1. c = ( 1 2 3 4 ). b. 0. π2 ⋅ π5 = π5 ⋅ π2. b = ( 1 2 3 4 ). in general. e) (ab)(cd). b) π7. 82 . f) (ad)(bc). b) ⎜ ⎝ C1C3C4 C5 C2 C7 C8 C9 C6 C11C10C13C14 C15C12C16 ⎠ ⎛ C1C2 C3 C4 C5 C6 C7C8 C9 C10C11C12 C13C14C15C16 ⎞ d) ⎜ . d) (a)(b)(c)(d). and 240˚.1: 1 (a) not symmetric. c are rotations of 0˚.Solutions Chapter Nine Solutions Section 9. 2a) no. (ii) (1)(2)(34)(57)(6)(8). c) π2. Example 3). ⎢ ⎣0 1 ⎦ 3a) 6 symmetries (as in 4a) (abcd). row is first symmetry. 11a) a. respectively d. b) (ac)(bd).

hexagon of R-W-B-R-W-W and R-B-W-R-W-W. b) 92. 2(3 x 2n-1) . d) 1 10 1 12 1 ( m + m 5 + m 6 + m 7 ) . f) same as b). 55. π2 = (12)(3). π3 = (13)(2). f) b) (m 16 + 2 m 4 + m 8 + 4 m 9 ) .1. b) 2195. e) ( m + 2 m 2 + 2 m 4 + m 6 + 6m 7 ) .. b) 2C(n.2). and 2(5n + 3 x 5(n-1)/2) n even.Solutions 13. 5. d) yes. 5a) 6(m4 + 2m2 + 3m3) . ψ(π7) = ψ(π8) = 12. G'. c) {π1π6}. c) cyclic color sequence on 4a) 315. 51.π7}. c) 2(m5 + m3) . other subgroups are {π1πi}.π3. 1 b) ψ(π1) = C(12 + 3 . 4(264 + 2 x 216 + 232) .π6} or {π1. c) in addition to subgroups in b) and G. 13b) {π1. 6a) 28. 22a) 2n . only left structure (right structure has 6 isomers). 1 1 4.π7.π3. 1 2 1 7a) 6(m6 + 2m2 + 3m4) .3: 1. 20a) no. 3 . 2(3 x 2n-1 + 3 x 2(n-1)/2) . b) 136. 8 4 6a) 8(x4 + 2x4 + 3x2 + 2x2x2) . c) 8(38 + 2 x 32 + 34 + 4 x 35) = 954. 6. 12a) 4(455 + 2 x 49 + 7) = 140. 2(5n + 5n/2) n even. 3(315 + 2 x 35) . 1 11. π4 = (1)(23). b) 21. b) 70.2: 1a) 24. n even. 8. 1 2. 33. c) 5934. π5 = (123). for i = 3. ψ(π5) = ψ(π6) = 1. d) 16107. 10. b) yes. 8. c) 2(m6 + m4) . G". π6 = (132). b) e) 1 1 7 (m + 2m2 + 2m3 + 4m4 + 3m5). n odd. Section 9. 1 7. 280. b) 8(m12 + 2m3 + 3m6 + 2m7) . 21b) {π1. b) 4(165 + 2 x 25 + 5) = 55. ψ(π2) = ψ(π3) = ψ(π4) = 7. 5.π8}. 1 1 answer: 6(91 + 7 + 7 + 7 + 1 + 1) = 19. and other ψ(πi)'s = 0.π5. 4 12 8 1 1 83 . 12 1 1 1 1 12 1 1 3 6 7 (m + 2m + 3m + 2m ) . 2a) 208. 2 Section 9. ψ(π3) = 6. c) no. 3a) 130. ψ(π1) = 18. 12) = 91. 9a) in cycle form: π1 = (1)(2)(3). d) (m8+ 3m4). 1 1 answer 8(18 + 6 + 12 + 12) = 6.

7. c) (b + w)(b 2+ w2)2}.1) x m) . (ii) 9. {(b + w)11 + 10(b11 + w11)}. 4. + 2(b2 + w2)8}. (iii) 258. Section 9. b) 4 {(b + w)8+ 2(b4 + w4)2+ (b2 + w2)4}. 15.1) x m + p x m(p+1)/2) .4: 1. 11a) 2(28 + 24) = 136. 10. 1 p 2p(m + (p . f) same as b). b) 1 8 {(b + w)12 + 1 {(b + w)6 + (b + w)2(b 2+ w2)2}. b) 15. 12. 1 5. 1 2 1 b) 4(28 + 24 + 0 + 24 ) = 72. b) (i) 1.(a) 1 6 {(b + w)4 + 2(b3 + w3)(b + w) + 3(b2 + w2)(b + w)2}. c) f) 8 {(b + w)16 + 2(b4 + w4)4 + (b2 + w2)8 + 4(b2 + w2)7(b + w)2}. (iii) 147. (ii) 6. d) 2 1 10+ (b2 + w2)5}+ (b + w)2(b 2+ w2)4 + (b + w)4(b 2+ w2)3}. e) 1 {(b + w)12 + 2(b6+ {(b + w) 4 12 6)2 + 2(b3+ w3)4 + (b2 + w2)6 + 6(b2 + w2)5(b + w)2}. 1 1 13. b) 1 1 1 9. 3. e) {(b + w)7 + 2(b6+ w6)(b + w) 4 12 + 2(b3 + w3)2(b + w) + 4(b2 + w2)3(b + w) + 3(b2 + w2)2(b + w)3}. 1 2a) 6 {(b + w)6 + 2(b6 + w6) 1 + 2(b3 + w3)2 + (b2 + w2)3}. 15a) p(mp + (p . b5 + b4w + 2b3w2 + 2b2w3 + bw4 + w5. b) 9 {(b + w)9 + 6(b9 + w9) + 2(b3 + w3)3}. 84 1 . c) d) 1 10 1 11 {(b + w)10 + 4(b10 + w10) + 4(b5 + w5)2 + (b2 + w2)5}. b4 + w4 + r4 + b3w + b3r + bw3 + w3r + br3 4) 4 {(b + w)16 + 2(b4 + w4)4 1 + wr3 + 2b2w2 + 2b2r2 + 2w2r2 + 2b2wr + 2bw2r + 2bwr2 . w 2(b4 + w4)3 + 3(b2 + w2)6 + 2(b2 + w2)5(b + w)2. 10a) 24(x4 + 6x4 + 3x2 + 8x1x3 + 6x2x2) . ψ(πi) = number of cycles of length 1.(a) 1 6 1 1 {(b + w)6 + 2(b3 + w3)2 + 3(b2 + w2)2(b + w)2}. 6a) and c) 4 {(b + w)5 + 2(b + w)(b4 + w4) + (b + w)(b2 + w2)2}. d) 1 1 {(b + w)8+ 3(b2 + w2)4}. 1 {(b + w)5 + 2 b) 8 {(b + w)12 + 2(b4 + w4)3 + 3(b2 + w2)6 + 2(b2 + w2)5(b + w)2}.. 4(74 + 3 x 72) = 637.Solutions 8a) (i) 2.

1 = 1 + b + 3b2 + 3b3+ 5b4 + 3b5 + 3b6 + b7 + b8.x2)(1 . b) 8 {(r + w + rw)4 + 2(r4 + w4 + r4w8) + 3(r2 + w2 + r2w2)2 + 2(r + w + rw)2(r2 + w2+ r2w2)}. 11. + 6(b2 + w2)(b + w)2}. 13. 1 b) 24 {(b + w)6 + 6(b4 + w4)(b + w)2+ 3(b2 + w2)2(b + w)2 + 6(b2 + w2)3 + 8(b3 + w3)2}. ak. . .n equals the number of partitions of n into k or less parts.x)(1 . c) b4w4 + b4r4 + w4r4 + b4w3r + b4w2r2 + b4wr3 + b3w4r + b2w4r2 + bw4r3 + b3wr4 + b2w2r4 + bw3r4 + b3w3r2 + b3w2r3 + b2w3r3. 16. 24 {(b + w)4 + 6(b4 + b4) + 8(b3 + w3)(b + w)+ 3(b2 + w2)2 13a) if not a cyclic rotation of all corners. b) 216. 24 {(b + w)12 + 6(b4+w4)3 + 3(b2 + w2)6 + 6(b + w)2(b2 + w2)5 + 8(b3 + w3)4}. 1 cycle would have to divide p — impossible. 10a) b2w2r + b2wr2 + bw2r2. 14a) 1 8 {(1 + b + b2)4 + 2(1 + b4 + b8) + 3(1 + b2 + b4)2 + 2(1 + b + b2)2(1 + b2+ b4)}. 11. 1 1 9a) 12 {(b + w)4 + 8(b3 + w3)(b + w) + 3(b2 + w2)2}.k)/p.xk) number of partitions of n into parts of size k or less: gk(x) = 85 . which equals the 1 (1 .Solutions 8.x3) . (1 . 15a) 36. 20. the length of the b) C(p. b) no positive terms.

1456. a) s ⇒ tt ⇒ 01tt ⇒ 01t10t ⇒ 01010t ⇒ 010101.. s2 → Ts3. b) s ⇒ tt ⇒ 01tt ⇒ 010t ⇒ 01010t ⇒ 010101. 3. and p → cj.1: 1. b) s → Rppp. ti. cards in suit I}. adj. b) s ⇒2s ⇒22s ⇒ 220t ⇒ 2201t ⇒ 22010t ⇒220102t ⇒ 220102. dhj. b) s → si . 2. 7a) s → pppp. 1256. and p → cj. 1235. s → t2BC.. 12. 2345. t3 → Tt4. u2 → BB. 1234. t → 1t”. t → 0t. s → t2u2. s → t2v2. etc. t → 1. s → Ts2. a) s ⇒ tt ⇒ 0t ⇒ 001t ⇒ 0010. s’ → 1s. 5. t → 1. 86 . wxyz. many possibilities. s → ABv2.2 1a) s ⇒1s ⇒101s ⇒1011s ⇒101101s ⇒101101. e. 1346. s → tv. s3 → THH. 5a) 123456789(10). ahj. where i ranges over the 4 card values. t → 0. c) same as b) but cj ranges over all colors. 1236. s → ttttt. b) abcd. e) s ⇒ 0s1 ⇒ 00s11 ⇒ 000s111 ⇒ 00001111. d) s ⇒t3p8 ⇒ c3sc3hc3cp8 ⇒ c3sc3hc3cc8hc8d. s2 → Ht3. 1234. c) s ⇒ tt ⇒ 0t ⇒ 001t ⇒ 00110t ⇒ 001100. s → pppR. t” → 0t. s → Ht2. (10)987654321. s2 → Tt3.g. 1245. 3456. s → 1t. 2. aehc. v → 1. s → Au2C. 1345. v → 1v. t → 0. adh. 1342. t’ → 1t. where cj ranges over the five non-Red colors. 6. aech. i = the four suits. Section 10. t → 0. 3. 2356. aceh. c) s ⇒2s2 ⇒20t3 ⇒202t4 ⇒2020t5 ⇒ 20202. 10. t2 → Tt3.1356. b) s ⇒ tt ⇒ 10tt ⇒ 100t ⇒ 1001. s → 0t. 1243. where cj ranges over the six colors. s → 1t. s3 → Ht4. 1246. t → 0t’. s → 0s’. t→ 1t. s → ppRp. 8a) s → fi. si → {all subsets of 5 9. 4. 4. v2 → CC. fI → ciscihciccihdi. s” → 0s. t4 → TH. s → pRpp. 2456. ache.. s → u2v2. t → 0t. 11. → AA.Solutions Chapter Ten Solutions Section 10. 2346. etc. where di ranges over the 48 cards whose value is not i. t → 0. 1324. s → 1s”.

N). 0 yields (s/0) and 1 yields (s/1). R yields (r/N) and non-R yields (t4/N). 0 yields (w/0) and 1 yields (w/1). R or non-R yields (r. from s3.N). -2. next state is ti+j and i + j is printed 7. -1.R yields (r. 1.N) and non-R yields (end. from t2. from w. States of machine are ti. from s4. c) (to/–)→ (d1/-)→ (d1/-)→ (d1/-)→ (d2/-)→ (d2/-)→ (d2/-)→ (s5’/6)→ (s4/5)→ (t1/4).0).N).Y)l from t.Y). 3a) (to/–)→ (t1/6)→ (t1/-)→ (t1/-)→ (t1/3)→ (t1/-)→ (t1/-)→ (t1/1)→ (t1/-)→ (t1/-). 3. 5. from r. 11a) starting from s. reading R yields (t2/N) and any non-R yields (s2/N). 2. where k = i + j (mod 6): 9. b) Starting from s.N) and non-R yields (t3. 4. from t4. R yields (r. 0. 0 yields (w/1) and 1 yields (s/0). 87 . when j (=-1 or +1) is read in state ti. b) (to/–)→ (d1/-)→ (d1/-)→ (d1/-)→ (s4/5)→ (t1/4)→ (t1/-)→ (t1/1)→ (t1/-)→ (t1/-). R yields (t4/N) and non-R yields (s4/N). 2. and 2 yields (s. from t3. 3. from s2. 1 yields (r. from s. reading 0 yields (t. any input yields (t. and from r.Y) and non-R yields (r. from u. any input yields (r. next state is tk and k is printed. reading 0 yields (s/1) and 1 yields (u. States of machine are ti. 5.Y).Y). when j is read in state ti. b) (to/–)→ (s1/1)→ (t1/1)→ (s2/2)→ (t2/2)→ (s3/3)→ (s3/3)→ (s3/3).Solutions Section 10. R yields (end.N). 1.3: 1a) (to/–)→ (s1/1)→ (s1/1)→ (t1/1)→ (t1/1)→ (s2/2)→ (s2/2)→ (t2/2). R yields (t3/N) and non-R yields (s3/N). i = 0.N). From t0. i = -3.

(1. b) f. g. over 40}. 4. (7. that is.0).3). 6a) (0.1: 1a) {a. (0. by symmetry assume g(a) = 0. suppose x and y are adjacent because there is an edge from x to y. 15.1). 12. longest path is at least length k (maybe there is another longer path starting at x). 2.8). (1. 16. second player wins at 4.2). 7. first player to 2. (5. 4.Solutions Chapter Eleven Solutions Section 10. e) if a not in K. (4. (6.d}.8). but now two kernel vertices are adjacent. if only if W(S) = S.10).3).2). (7. a. b) (0.1). 5. 21. (5. (6. c) x = 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 g(x) = 1 1 2 0 0 1 1 2 2 0 1 0 0 2 1 1 0 0 2 1 1 0 2 2 x = 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 over 40 g(x) = 1 0 0 0 2 1 1 0 0 2 1 1 0 2 2 1 1 0 9. follows from parts a) and b) of Exercise 12 since g(x) = k means there is a path of length k starting at x while l(x) is length of longest path starting at x. (7. b. 13. then g(e) = 1.0). (3.5). then g(e) = 0.9). (1. 18. A goes to 2 . (2. (2. then b in K — impossible since a in K. g(b) = g(d) = 1 (or interchange values).5).7).2). 27.if a is K (kernel). 6. second player wins at 9. then g(d) = 0. then g(c) = 1. (3. (2.c} or {b. for example. first player wins by moving to kernel.1). graph in Exercise 1b).6). move to multiples of 5k + 1 (initial position is win for second player). (6. 36. c) no kernel. a. 24). then g in K. f. first player to 11.4). second players wins at 16. (8. B must go to 4 or else A will win. (5. 32. then d not in K. 11. 17. 25. 8a) g(a) = g(c) = 0. 11. second player to 6. h. 26.6). 31. similar sort of argument (also involving c.0). 9. {3. 12. 10. d. then h not in K. 3. then x must have a 88 . b) no Grundy function. then first player to 7 or 8. consider directed 5-circuit b. S is a kernel if and only if all vertices not in S have an edge to a vertex in S while no vertex in S has an edge to a vertex in S. kernel = {0.7). first player to 1 or 5.

5). remove 2 from 1st pile. call it xi. (5. 8). f = o_o. c) 2. f. 3rd. call it x2. (0. must have an infinite number of vertices reachable from it. 11. remove 2 from pile 3. s(f) = {g}. remove 1 from pile 3. from 3rd pile. must have an infinite number of vertices reachable from it. (1. 6a) remove 3 from pile 2 or 3. g(e) = 2. (0. then trivially c' + cj = c' + dj . 17. 5a) 0. f. remove 2 from pile 1. g(b) = g(c) = g(d) = 3. s(e) = {g. c) 4. if there were an infinite number of vertices. g(a) = g(g) = 1. add nickel to 3rd pile. c) remove 1 from pile 2 or 3. if cj = dj . b) 2. 13a) remove 3 balls along one of the 3 lines formed by balls on one of the 3 side of the arrangement. (0. 4). 5). g(f) = g(h) = 0. and so on. without end. b) remove 2 from pile 2 or 3. s(c) = s(d) = {e. (2. 7a) 3. 6). 7. if c' + cj = c' + dj . 2). 3rd or 4th pile. 19. 89 . d) 1.2: 1a) 7.Solutions larger level number than y and a different Grundy value from its successor y. h}. that is. then one of the finite number of starting vertices. (3. b) 2. (4. s(h} = ∅. remove 2 from 3rd pile. c = o_oo. call it x3. they must be equal. and one of the finite number of successors of x2. remove 4 from 2nd. remove 3 from 4th pile. remove 1 from pile 4. s(g) = h. b) 0. f}. (2. remove 1 from pile 3. d) 0. Section 10. e. g = o. d. d) d. b) (0. then cj and dj must have 1's in the same positions in their binary representations. d) 1. s(b) = {e. 0). 4). remove 3 from 3rd pile or 2 from 4th pile. h = _ (win). c) 2. remove 2 from 4a) 0. d) 0. b) 0. c) 1. 6). s(a) = {b. c) 0. 9). e = oo. g}. 2a) 0. Immediate that the proposed strategy works. (3. or 4th pile. 1). c. b = ooo. g}. (4. must have an infinite number of vertices reachable from it. 3). remove 1 3a) 3. Let a = oooo. 7). (2. b) 2. and one of the finite number of successors of x1. d = oo_o.

_ ABC ABC here all expression involve ∩. 18. 3.C. 24. — 2a) A ∩ B ∩ C. 7.N(R ∩ M) = 4 . or 2 memory. 13.2 = 2. 12. and clearly N(R ∩ M ) = 2. 4. 25. 6.Solutions Appendices Solutions Section A. 23. c) 1. 29.8 = 4. — ————— ————— ————— __ ABC ABC _ ABC _ ABC __ 9. 27. b) yes. e) none. 5. N(R ∩ M ) = N(R ) . 7a) A B . 8a) A ∪ B. rechargeable and 2 no-memory. 5a) impossible.3.1: 1a) 12. non-rechargeable and 2 no-memory. then N(R ∩ M) = N(M) . see Figure A1. b) (A ∪ B) ∩ (A ∩ B).2 = 2.8 . ———————— d) all 1 ≤ k ≤ 29 except 12. 10.A) = A. 22. 2 memory. — — — — — — — — — 4. b) A B c) ( A ≈ B ) ∩ ( A ↔ B ) = ( A ↔ B ). f) A ∪ (A ∩ B ∩ C). 8. rechargeable. 9. d) none. 11. 28. 15. we are given that N(R ∩ M) = 2 as well as that N(M) = N(R) = N(M ) = N(R ) = 4. c) 20 -15 = 5. 20. c) A ∩ B . 27. 17. 19. — — __ ___ ABC ABC 90 . so that AB C = A ∩ B ∩ C . 14. 26. b) A ≈ B ≈ C . 21. 3.(B .N(R ∩ M) = 4 . 16. 27 b) 2. non-rechargeable. d) A . 20 .

24. 3 x 132 x 39. but for n = 1.393.(50 + 40 . not n ≥ 2. prob(E0) =1/8. 13. 15a) all sequences with k tails. 1/2. 6a) 2/(5 x 4). define A2 similarly.4)/50. 4. (2 x 2 x 2)/4! = 1/3. 17a) E = S ∪ H ∪ C. 523 . 3a) 1/6. 1 . (as if eyes were open). 14. b) 1/2. prob(E2) = 3/8. b) (4 x 3)/(5 x 4) = 3/5. prob(E3) = 1/8. let A1 = outcomes with 1 or 2 on first die.2: 23. prob(E1) = 3/8. b) 2 x 1/3 x 2/3 = 4/9. (2 + 6)/(5 x 4). a multiple of i). = 3/10. and one head followed by a 91 . 1 . ———————— c) E = (S ∩ H ∩ C ) ∪ (S ∩ H ∩ C) ∪ ( S ∩ H ∩ C). 2/3. one can prove that there are a finite number of binary sequence of length ≤ n. for example. b) E = (S ∪ H ∪ C) ∩ ( S ↔ H ↔ C ).393 . c) and d) unshaded part of A in b). — — — Section A.4 = 20. 9.20)/100 12. 25.133.55/65.Solutions A B 10a) obvious b) C . 7a) 1/3 x 1/3 = 1/9. then must count N(A1 ∪ A2) = 12 + 12 . 16. step (2) assumes proposition has been proved for n = 2— not true. 523 . the initial step only assumes that n ≥ 1. one can only prove by induction a property that is a function of n. all sequences of 5 selections numerator is N(D3 ∪ D4): (16 + 12 . Section A. 11. if Di = integers from 1 to 50 divisible by i (or equivalently. 2. k ≥ 0. b) 18/36 = 1/2. 5a) 1/6. 12a) a) and c) both equal B. c) 3/36 = 1/12. an-2 is undefined. 10. 8. (3 x 6 + 1)/216.3: 1.

Solutions head. n + 1. implying ai-aj = 0 mod n-1. if any difference is 0 mod n-1. say. b) all positive integers. each dummy must go to all 6 real printers. i + 5. Section A. since C(5. Printer i is connected to computers i.4: 1. 4. i + 2. 92 . i + 4. 12. i + 3. k ≥ 0. 24 connections. consider the n-1 differences. we are done.3) = 10 < 12. 6.if dummy printer A connected to just 5 real printers. ai-a1 and aj-a1 must be the same mod n-1. 16. c) all ordered pairs of positive integers. i + 1. a set of calls to dummy A and those 5 real printers could not be handled. 6. d) all sequences of k black balls. modulo n-1. (8 + 10 + 3 x 11) + 1 = 52. 15. otherwise two of the differences. of the first number subtracted from each of the other n-1 integers. followed by a red ball.

Sign up to vote on this title
UsefulNot useful