Chapter 3: Euclidean Space 41 SECTION D Linear Independence By the end of this section you will be able to understand what is meant by linear combination and linear independence show that given vectors are linearly independent or dependent prove properties about linear independence D1 Standard Unit Vectors in n What does the term standard unit vectors mean? From the last section we know that unit vectors are of length 1 and the standard unit vectors in n are column vectors with 1 in the kth position of the vector ek and zeros everywhere else. y e2 = z 0 1 e1 = e3 e1 1 0 x x e2 kth position y 0 ek 1 0 Fig 35 Why are these standard unit vectors important? Well we can write any vector of n in terms of the standard unit vectors as we showed in Exercise 3a question 14. We proved the following Proposition: x1 Proposition (3.20). Let u be any vector in n given by u xk . Then we have x n u x1e1 x2e2 xk ek xnen Proof. See solution to question 14 of Exercise 3a. xk ek xnen is called a This representation u x1e1 x2e2 and en . We can write this linear combination of the standard unit vectors e1 , e2 , u x1e1 x2e2 x3e3 xnen in matrix form as u e1 The matrix e1 e2 e2 e3 e3 en x1 en x n has its first column entries e1 , its second column is e2 , , and the nth column is en . Moreover we can say this representation u x1e1 x2e2 xnen of a vector is unique. Chapter 3: Euclidean Space 42 Proposition (3.21). Let u be any vector in n then the linear combination u x1e1 x2e2 xk ek xnen is unique. What does this proposition mean? Means that there is only one way of writing any vector u of n in terms of the standard unit vectors e1 , e2 , and en . Proof. Let the vector u be written as another linear combination: u y1e1 y2e2 yk ek ynen What do we need to show? Required to show that the scalars are equal: y1 x1 , y2 x2 , y3 x3 , and yn xn . Equate the two linear combinations because both are equal to u: x1e1 x2e2 xnen y1e1 y2e2 yn e n x1e1 x2e2 x1 y1 e1 xnen x2 y2 e2 Writing out e1 , e2 , y1e1 y2e2 xn yn en yn e n O O Factorising , en and O as column vectors in the last line we have 1 0 0 0 0 0 1 0 0 Remember ek x1 y1 x2 y2 xn yn 1 0 0 1 0 0 We can write this in matrix form as 0 x1 y1 0 x1 y1 0 1 0 I is the x y x y 0 1 0 0 0 2 2 2 2 identity implies I 0 matrix 1 xn yn 0 0 xn yn 0 and xn yn 0 which gives our Hence we have x1 y1 0, x2 y2 0, x3 y3 0, and yn xn . result: y1 x1 , y2 x2 , y3 x3 , Therefore the given linear combination, u x1e1 x2e2 xk ek xnen , is unique. ■ D2 Linear Independence Example 12 Determine the linear combination of the standard unit vectors in n which gives the zero and xn in the following: vector. This means find the values of the scalars x1 , x2 , x3 , x1e1 x2e2 xk ek xnen O Solution and en we have Substituting e1 , e2 , 1 0 0 1 x1 x2 0 0 0 0 0 0 xn 1 0 Chapter 3: Euclidean Space 43 We can write this in matrix form as 1 0 x1 0 [This is Ix O ] 1 xn 0 0 In compact form we have Ix O where I is the identity matrix. Ix O gives x O . From the zero vector x O we have x1 0, x2 0, x3 0, and xn 0 . We say that the standard unit vectors e1 , e2 , and en are linearly independent which means that any one of the vectors ek cannot be made by a linear combination of the others. These standard unit vectors are not the only vectors in Definition (3.22). Generally we say vectors v1 , v 2 , v3 , n which are linearly independent. and v n in n are linearly independent the only real scalars k1 , k2 , k3 , and kn which satisfy: k1v1 k2 v 2 k3 v3 kn v n O are k1 k2 k3 kn 0 What does this mean? The only solution to the linear combination k1v1 k2 v 2 kn v n O is when all the scalars k1 , k2 , k3 , and kn are equal to zero. In other words you cannot make any one of the vectors v j say, by a linear combination of the others. We can write the linear combination k1v1 k2 v 2 k3 v3 v 1 v2 The first column of the matrix v1 v3 v2 v3 kn v n O in matrix form as k1 0 k 0 vn 2 kn 0 v n is given by the entries in v1 , the second column is given by the entries in v 2 and the nth column is given by the entries in v n . Recall from chapters 1 and 2 that a linear system Ax O has the unique solution x 0 matrix A is invertible or det A 0 [Not zero]. How is this related to linear independence? v n then these vectors v1 , v 2 , … and v n are Well if the matrix A v1 v 2 linearly independent Ax O has the unique solution x O . What does this mean? v n is invertible or Means we only need to check that the matrix A v1 v 2 det A 0 [not zero] for the vectors v1 , v 2 , … and v n to be linearly independent. Example 13 1 2 Show that u and v are linearly independent in 2 and plot them. 1 3 Solution Using the above definition (3.22) and writing out the linear combination we have k u c v O [Using k and c as scalars] Substituting the given vectors u and v into this k u c v O gives Chapter 3: Euclidean Space 44 k 2c 0 1 2 0 k c or k 3c 0 1 3 0 1 2 k 0 Let A u v and x . We need to solve Ax O where O . 1 3 c 0 1 2 Since det A det 1 3 1 2 5 is non-zero so the only solution is 1 3 x O [Because Ax O and det A 0 then x O ] which means the values of the scalars are k 0 and c 0 . Hence the linear combination k u c v O yields k 0 and c 0 therefore the given vectors u and v are linearly independent because all the scalars, k and c, are equal to zero. Plotting the given vectors we have: 2 v 3 1 u 1 Fig 36 This means that vector u is not a scalar multiple of vector v and vice-versa. In 2 the vectors u and v are linearly independent means that they are not scalar multiples of each other. Plotting linear independent vectors: v Linearly independent vectors u and v u Fig 37 D3 Linear Dependence What does linear dependence mean? Definition (3.23). Conversely we have the vectors v1 , v 2 , v3 , linearly dependent the scalars k1 , k2 , k3 , k1v1 k2 v 2 k3 v3 and v n in n are and kn are not all zero and satisfy kn v n O From chapters 1 and 2 we know that the linear system Ax O has an infinite number of solutions if det A 0 . This means that if k1 A v1 v2 vn , x k n and v n are linearly dependent. Why? and det A 0 then these vectors v1 , v 2 , v3 , Because det A 0 means the linear system Ax O has an infinite number of solutions so there must be non-zero solutions, that is x O . Hence there are non-zero k’s which satisfy Chapter 3: Euclidean Space 45 k1v1 k2 v 2 k3 v3 kn v n O Example 14 3 1 2 Show that u and v and plot them. are linearly dependent in 1 1/ 3 Solution Using the above definition (3.23) and writing the vectors u and v as a linear combination we have k u c v O [k and c are scalars] Substituting the given vectors 3 1 0 ku cv k c 1 1/ 3 0 You can check the determinant of A u v is equal to zero. However this does not give us the relationship between the two vectors u and v. We can write out the argumented O and carry out row operations to determine the relationship between u and matrix A v. The augmented matrix for this is given by R1 3 1 0 Because u v O R 2 1 1/ 3 0 where R 1 and R 2 represent rows. Carrying out the row operation 3R 2 R1 gives k c 3 1 0 3R 2 R1 0 0 0 From the top row we have 3k c 0 which implies c 3k . Let k 1 then c 3k 31 3 Substituting our values k 1 and c 3 into k u c v O gives u 3v O or u 3v We have found non-zero scalars, k 1 and c 3 , which satisfy k u c v O therefore the given vectors u and v are linearly dependent and u 3v . Plotting the given vectors u and v we have: R1 3 u 1 1 v 1/ 3 Fig 38 Note that u 3v which means that the vector u is a scalar multiple 3 of the vector v. If vectors u and v in 2 are linearly dependent then we have k u c v O where k 0 or c 0 That is at least one of scalars is not zero. Suppose k 0 then Chapter 3: Euclidean Space 46 k u c v c u v Dividing by k k This means that the vector u is a scalar multiple of the other vector v which suggests that u is in the same (or opposite) direction as vector v. Plotting these we have: u Fig 39 v Linearly dependent vectors u and v Of course in the above Example 14 we could have let k 2, 3, , 666 etc. Any non-zero number will do! Generally it makes the arithmetic easier if we use k 1 . Example 15 2 3 0 Determine whether the following vectors u 1 , v 1 and w 0 in 3 are 0 1 0 linearly dependent or independent. Solution Consider the following linear combination k1u k2 v k3w O where k’s are scalars. Substituting the given vectors u, v and w into this, k1u k2 v k3w O , yields Let A u Expanding along this row 2 0 3 0 k1u k2 v k3w k1 1 k2 1 k3 0 0 0 1 0 0 3 0 2 k1 and v w 1 1 0 x k2 . We need to solve Ax O : 0 1 0 k 3 3 0 2 3 2 det A det 1 1 0 1 det 0 2 2 1 0 0 1 0 Because det A 2 is non-zero therefore Ax O only has the solution x O . Solution x O means k1 k2 k3 0 , that is all the scalars are zero: k1u k2 v k3w O implies k1 k2 k3 0 Hence the given vectors u, v and w are linearly independent. Example 16 2 1 3 0 Determine whether the following vectors u 1 , v 1 , w 0 and x 2 in 0 1 0 3 3 are linearly dependent or independent. Solution Chapter 3: Euclidean Space 47 Consider the linear combination k1u k2 v k3w k4 x O What do we need to find? Need to determine whether all the scalars k’s - k1 , k2 , k3 and k4 are zero or not. Substituting the given vectors into this linear combination: 2 1 0 3 0 k1u k2 v k3w k4 x k1 1 k2 1 k3 0 k4 2 0 0 1 0 3 0 The augmented matrix of this is given by R1 3 0 2 1 0 Using u v w x R2 1 1 0 2 0 O R 3 0 1 0 3 0 Carrying out the row operation R 2 R 3 gives: k1 k2 k3 k4 3 0 R2 R3 1 0 R3 0 1 From the bottom row we have k2 3k4 R1 0 0 5 0 0 3 0 0 which gives k2 2 1 3k4 . Let k4 1 : 3k4 3 1 3 k2 From the middle row we have k1 5k4 0 implies that k1 5k4 5 1 5 . The top row gives 3k1 2k3 k4 0 . Substituting k1 5 and k4 1 into this: 3 5 2k3 1 0 implies k3 8 Our scalars are k1 5 , k2 3 , k3 8 and k4 1 . Substituting these into the above linear combination k1u k2 v k3w k4 x O gives the relationship between the vectors: 5u 3v 8w x O or x 5u 3v 8w Since we have non-zero scalars (k’s) so the given vectors are linearly dependent. We can write the vector: x 5u 3v 8w 1 x 2 3 3 u 1 0 0 v 1 1 2 w 0 0 Fig 40 In general if vectors u , v and w in 3 are linearly dependent then we have Chapter 3: Euclidean Space 48 k1u k2 v k3w O gives k1 0 or k2 0 or k3 0 That is at least one of scalars is not zero. Suppose k1 0 then k1u k2 v k3w k k u 2 v 3 w Dividing by k1 k1 k1 This means that the vector u in 3 is a scalar multiply of the other two vectors v and w. Hence we can make vector u out of vectors v and w. In the above Example 16 we have x 5u 3v 8w which means we can make the vector x out of the vectors u, v and w. D4 Properties of Linear Dependence Proposition (3.24). Let v1 , v 2 , v3 , and v n be vectors in n . If one (or more) of these and v n are linearly vectors, v j say, is the zero vector then the vectors v1 , v 2 , v3 , dependent. Proof. Consider the linear combination k1 v1 k2 v 2 k3 v 3 kjv j kn v n O (*) In (*) take k j 0 [Non-zero number] and all the other scalars equal zero, that is k1 k2 k3 k j 1 k j 1 kn 0 Since v j O we have k j v j O which means that all scalars in (*) are not zero ( k j 0 ). By definition (3.23) we have that the vector v1 , v 2 , and v n are linearly dependent. ■ What does Proposition (3.24) mean? and v n , one of these is the zero vector then they If amongst the vectors v1 , v 2 , v3 , are linearly dependent. and v m be distinct vectors in n . If n m , that is Proposition (3.25). Let v1 , v 2 , v3 , value of n in the n – space is less than the number, m , of vectors, then the vectors v1 , v 2 , v3 , and v m are linearly dependent. Proof. and v m : Consider the linear combination of the given vectors v1 , v 2 , v3 , k1v1 k2 v 2 kn v n kn1v n1 km v m O (*) The number of equations is n because each vector belongs to the n – space n but the number of unknowns k1 , k2 , k3 , , kn , kn1 , , km is m. Writing this out we have v11 v21 v12 v k1 k2 22 v1n v2 n v n 11 vn1 v v n 1 2 kn n 2 kn 1 v v nn n 1n vm1 0 vm 2 0 km vmn 0 n equations m unknowns By the following Proposition (1.24) case (II) of chapter 1: In a linear system, if the number of equations is less than the number unknowns then the system has an infinite number of solutions. (3.23) If non-zero k’s satisfy k1v1 k2 v 2 kn v n O then vectors v’s are dependent Chapter 3: Euclidean Space 49 In our linear system the number of equations n is less than the number of unknowns m. Therefore we have an infinite number of k’s which satisfy (*) and this means all the k’s are not zero. By Definition (3.23) the given vectors are dependent. Example 17 Determine whether the following vectors 2 2 2 4 u 3 , v 19 , w 6 and x 2 7 5 3 5 3 in are linearly dependent or independent. Solution Since we have 4 vectors u, v, w and x in 3 and 4 3 therefore by the above Proposition (3.25) we conclude that the given vectors u, v, w and x are linearly dependent. Normally we write the vectors as a collection in a set. A set is denoted by { } and is a collection of objects. The objects are called elements or members of the set. We can write the set of vectors v1 , v 2 , v3 , and v n as S v1 , v 2 , v3 , , vn We use the symbol S for a set. Proposition (3.26). Let S v1 , v 2 , v3 , , v n be n vectors in the n - space be the n by n matrix whose columns are given by the vectors v1 , v 2 , v3 , n . Let A and v n : A v1 v2 vn If the reduced row echelon form (rref) of A contains a zero row then the set of vectors S v1 , v 2 , v3 , , v n are linearly dependent. Proof. See Exercise 3(d). SUMMARY and v n be vectors in n and k1 , k2 , k3 , and kn be scalars. Let v1 , v 2 , v3 , Consider the following linear combination k1v1 k2 v 2 k3 v3 kn v n O (*) kn 0 (all scalars are If the only solution to this (*) is k1 k2 k3 and v n are linearly independent. zero) then the vectors v1 , v 2 , v3 , and kn are not all zero satisfy (*) then the vectors If the scalars k1 , k2 , k3 , v1 , v 2 , v3 , and v n are linearly dependent. and v n is zero then they are linearly If one or more of the vectors v1 , v 2 , v3 , dependent. and v m in n where n m then these vectors If there are m vectors v1 , v 2 , v3 , v1 , v 2 , v3 , and v m are linearly dependent.