Section 9.2: Matrices Definition: A matrix is a rectangular array of numbers: a11 a12 · · · a1n a21 a22 · · · a2n A = .. .. .. . .. . . . . am1 am2 · · · amn In general, aij denotes the (i, j) entry of A. That is, the entry in the ith row and jth column. The matrix A can be denoted using the shorthand A = (aij ). Definition: Two m × n matrices A = (aij ) and B = (bij ) are said to be equal if aij = bij for each ordered pair (i, j). Definition: If A = (aij ) is an m × n matrix and c is a scalar, then cA is the m × n matrix whose (i, j) entry is caij . This is known as scalar multiplication. Definition: If A = (aij ) and B = (bij ) are both m × n matrices, then the sum A + B is the m × n matrix whose (i, j) entry is aij + bij for each ordered pair (i, j). Example: Consider the matrices 1 2 A= 0 3 0 1 B= 2 4 (a) Find A − B + 2C. (b) Find a matrix D such that A + C = 2A − B + D 1 1 2 C= . 1 −1 Definition: If A = (aij ) is an m × n matrix and B = (bij ) is an n × r matrix, then the product AB = C = (cij ) is the m × r matrix with entries cij = n X aik bkj = ai1 b1j + ai2 b2j + · · · + ain bnj . k=1 This is known as matrix multiplication. Example: Consider the matrices 1 2 A= 3 4 and 2 2 B= . 1 3 Compute the matrix products AB and BA. Note: In general, matrix multiplication is NOT commutative! That is, AB 6= BA. Example: Consider the matrices 3 −2 A = 2 4 1 −3 −2 1 3 and B = . 4 1 6 Compute the matrix products AB and BA. 2 Notation: If A is an n × n matrix, it can be multiplied by itself a number of times. If k is a positive integer, then Ak = AA · · · A} . | {z k times Example: Find an expression for An if 1 1 A= . 1 1 Definition: The transpose of an m×n matrix A is the n×m matrix B with entries bij = aji . That is, transposition interchanges the rows and columns of a matrix. The transpose of A is denoted by AT or A0 . Example: Consider the matrices 1 2 A= 3 4 and Verify that (AB)T = B T AT . 3 2 1 B= . 5 3 Just as the number 1 acts as an identity for multiplication of real numbers, there is a special matrix I that acts as an identity for matrix multiplication. That is, IA = AI = A for any n × n matrix. Definition: The n × n identity matrix is the matrix In = (δij ), where 1, i = j δij = 0, i 6= j That is, In is the matrix with 1’s on the main diagonal and 0’s elsewhere. For example, 1 0 0 1 0 I1 = [1], I2 = , I3 = 0 1 0 0 1 0 0 1 If the size of In is clear from the context, then we simply write I. Definition: An n × n matrix A is said to be nonsingular if there exists a matrix B such that AB = BA = I. The matrix B is called the inverse of A and denoted by A−1 . Example: Verify that the matrices 1 2 3 A = 0 1 4 0 0 1 and are inverses of each other. 4 1 −2 5 B = 0 1 −4 0 0 1 Definition: An n × n matrix is called singular if it does not have an inverse. 1 0 Example: Verify that the matrix A = is singular. 0 0 For every n × n matrix, we can defined a scalar whose value will tell us whether the matrix is singular or nonsingular. Definition: The determinant of a 2 × 2 matrix a11 a12 A= a21 a22 is a scalar defined by det A = a11 a22 − a12 a21 . Theorem: If A is an n × n matrix, then A is nonsingular if and only if det A 6= 0. Moreover, if A is a 2 × 2 matrix a11 a12 A= , a21 a22 then its inverse is −1 A 1 a22 −a12 = . det A −a21 a11 5 Example: Determine whether each matrix is nonsingular and find its inverse, if it exists. 3 5 (a) A = 2 4 2 4 (b) B = 3 6 2 −1 (c) C = 1 3 1 2 (d) D = 0 3 −2 6 (e) E = 3 −9 6 The inverse of a square matrix A can be calculated by augmenting A with the n × n identity matrix and performing Gaussian elimination. 2 5 Example: Show that A = is nonsingular and find its inverse. 1 3 Example: Find the inverse of 1 −1 −1 A = 2 −1 1 . −1 1 −1 7 Consider a system of m linear equations in n variables a11 x1 + a12 x2 + · · · + a1n xn = b1 a21 x1 + a22 x2 + · · · + a2n xn = b2 .. . am1 x1 + am2 x2 + · · · + amn xn = bm . This system can be expressed in the a11 a12 · · · a21 a22 · · · A = .. .. .. . . . am1 am2 · · · matrix form AX = B, where a1n x1 b1 a2n x2 b2 , X = , B = .. .. . .. . . . amn xn bm Note: If we have a system of n linear equations in n variables, then it can be written in matrix form AX = B, where A is an n × n matrix. If A is nonsingular, then the system has a unique solution X = A−1 B. Example: Consider the linear system 2x + y = 2 4x + 6y = 8. (a) Write this system in matrix form. (b) Solve this system using the inverse matrix. 8 Consider the linear system defined by AX = ~0. Clearly X = ~0 is a solution of this system. It is called a trivial solution. If A is nonsingular, then the trivial solution is the only solution. Indeed, X = A−1~0 = ~0. However, if A is singular, then there exists a nontrivial solution X 6= ~0. Theorem: Suppose that A is an n × n matrix, and X and ~0 are n × 1 matrices. Then the equation AX = ~0 has a nontrivial solution if and only if A is singular. Example: Determine whether 1 1 A= 2 1 is nonsingular. If so, compute its inverse. In either case, solve AX = ~0. Example: Determine whether 1 3 B= 1 3 is nonsingular. If so, compute its inverse. In either case, solve BX = ~0. 9 The Leslie Matrix Suppose that we would like to model the population size of a species with discrete breeding seasons and age-dependent reproductive fitness (e.g. perennial plants, cicadas). Let N (t) denote the population size at generation t = 0, 1, 2, . . .. Then the population size at generation t + 1 is a function of the population size at the previous generation, N (t + 1) = f (N (t)). An equation of this form is called a difference equation or recursion. Recall that the difference equation modeling discrete exponential growth is N (t + 1) = RN (t), t = 0, 1, 2, . . . where N0 ≥ 0 is the initial population size and R > 0 is a growth parameter (average number of offspring per individual per breeding season). The solution of this difference equation is N (t) = N0 · Rt If R > 1, then the population experiences exponential growth over time, and if 0 < R < 1, the population decays exponentially over time. 300 5 4 Population Size Population Size 250 200 150 100 2 1 50 0 0 3 2 4 6 8 0 0 10 Time 2 4 6 8 10 Time The discrete exponential growth model (as well as the Beverton-Holt, discrete logistic, and Ricker logistic models) assume that all individuals have the same reproductive fitness. However, in many species, reproduction is age dependent. Thus, we will consider an agestructured, discrete-time model introduced by Patrick Leslie in 1965. Since only females produce offspring, we will model only the female population. The number of individuals is determined at the end of each breeding season. 10 Suppose that the population is divided into m+1 distinct age classes and let Nx (t) denote the number of females of age x = 0, 1, 2, . . . , m at time t = 0, 1, 2, . . .. Then the vector N0 (t) N1 (t) N2 (t) N (t) = .. . Nm (t) describes the number of females in each age class at time t = 0, 1, 2, . . .. Let Pj ∈ [0, 1] denote the fraction of females of age j at time t that survive to time t + 1. Then N1 (t + 1) = P0 N0 (t) N2 (t + 1) = P1 N1 (t) .. . Nm (t + 1) = Pm−1 Nm−1 (t) Let Fj ≥ 0 denote the average number of female offspring per female of age j which survive to the end of the next breeding season. Then N0 (t + 1) = F0 N0 (t) + F1 N1 (t) + F2 N2 (t) + · · · + Fm Nm (t). These equations can be written in matrix form as N (t + 1) = LN (t), where F0 F1 F2 P0 0 0 0 P1 0 L=0 0 P 2 .. .. .. . . . 0 0 0 · · · Fm−1 Fm ··· 0 0 ··· 0 0 ··· 0 0 .. ... . · · · Pm−1 0 The matrix L is called the Leslie matrix. The Leslie matrix model is illustrated below. 11 Example: Assume that a population is divided into three age classes and that 80% of the females of age zero and 10% of the females of age one survive until the end of the next breeding season. Moreover, assume that females of age one have an average of 1.6 female offspring and females of age 2 have an average of 3.9 female offspring. If the initial population consists of 1000 females of age zero, 100 females of age one, and 20 females of age two, find the Leslie matrix and age distribution at time t = 3. 12 Example: Assume that a population is divided into four age classes and that 65% of the females of age zero, 40% of the females of age one, and 30% of the females of age two survive until the end of the next breeding season. Moreover, assume that females of age one have an average of 2.8 female offspring, females of age 2 have an average of 7.2 female offspring, and females of age 3 have an average of 3.7 female offspring. If the initial population consists of 1500 females of age zero, 500 females of age one, 250 females of age two, and 50 females of age three, find the Leslie matrix and age distribution at time t = 3. 13 Example: Consider the Leslie matrix 0 5 0 L = 0.8 0 0 . 0 0.3 0 Determine the number of age classes in the population, the fraction of one-year-olds that survive until the end of the next breeding season, and the average number of female offspring of a two-year-old female. Example: Consider the Leslie matrix 2 3 2 0.4 0 0 L= 0 0.6 0 0 0 0.8 1 0 . 0 0 Determine the number of age classes in the population, the fraction of two-year-olds that survive until the end of the next breeding season, and the average number of female offspring of a one-year-old female. 14 Stable Age Distribution Example: Suppose that a population is divided into two age classes with Leslie matrix 1.5 2 L= , 0.08 0 and suppose that N0 (0) = 100 and N1 (0) = 100. (a) Find the age distributions for t = 0, 1, 2, . . . , 7. The following table summarizes the age distributions for t = 0, 1, 2, . . . , 7. Time, t 0 1 2 3 4 5 6 7 N0 (t) 100 350 541 868 1388 2221 3553 5685 N1 (t) 100 8 28 43 69 111 178 284 (b) Compute the successive ratios q0 (t) = N0 (t) N0 (t − 1) and q1 (t) = N1 (t) N1 (t − 1) for t = 1, 2, . . . , 7, and determine lim q0 (t) and lim q1 (t). t→∞ t→∞ The following table summarizes the successive ratios for t = 0, 1, 2, . . . , 7. Time, t 1 2 3 4 5 6 7 q0 (t) 3.5 1.55 1.6 1.5991 1.6001 1.5997 1.6001 q1 (t) 0.08 3.5 1.536 1.605 1.609 1.604 1.596 Both ratios seem to be approaching 1.6. Therefore, we suspect that lim q0 (t) = lim q1 (t) = 1.6. t→∞ t→∞ 15 (c) Compute the fraction of females which are age 0, p(t) = N0 (t) N0 (t) + N1 (t) for t = 0, 1, 2, . . . , 10, and determine lim p(t). t→∞ The following table summarizes the fraction of females which are age 0 for t = 0, 1, 2, . . . , 7. Time, t p(t) 0 0.5 1 0.9777 2 0.9508 3 0.9528 4 0.9526 5 0.9524 6 0.9523 7 0.9524 This fraction seems to be approaching 0.9524. Therefore, we suspect that lim p(t) = 0.9524. t→∞ In the long run, approximately 95.24% of the population is age 0. Although the population is increasing in size, the fraction of females in age class 0 seems to converge. This constant fraction is called the stable age distribution. If we start the population in the stable age distribution, the fraction of females in age class 0 will remain the same, about 95.24%, and the population will increase by a constant factor of 1.6 each generation. A stable age distribution for this population is 2000 N (0) = . 100 It follows that 1.5 2 N (1) = LN (0) = 0.08 0 2000 3200 = . 100 160 The fraction of females in age class 0 remains the same 3200/3360 = 2000/2100 = 0.9524. Similarly, 2000 3200 1.6 = 100 160 yields the same result. Therefore, if N denotes a stable age distribution, then LN = λN, where λ = 1.6 and N = [N0 , N1 ]T , with N0 /(N0 + N1 ) ≈ 95.24%. 16