NOTES ON LINEAR ALGEBRA

 

CONTENTS:

[9] BASIS VECTORS

 

 

[9] BASIS VECTORS

Consider the vector

(2)

(5)

 

This means two units in the x direction, five units in the y direction.

 

 

 

 

 

 

 

 

 

 

 

 


Graphically, we see we can write it as a vector in the x direction, and a vector in the y direction. Let

                                                (1)                    (0)

                                    Ex  =    (0)        Ey  =    (1)

 

be the unit vectors in the x direction and the y direction. We will show that they are a basis for the plane. What this means is that we can write any vector as some copies of Ex and some copies of Ey.

 

For example,

                        (2)      (2)       (0)         (1)         (0)

(5)  =  (0)  +   (5)   =  2 (0)  +  5 (1)   =  2 Ex  + 2 Ey

 

How did we get this? We’re trying to write the vector (2,5) as some number of copies of (1,0) and some number of copies of (0,1).

 

So we’re trying to solve

 

            (2)      (*)      (0)

(5)  =  (0)  +  (**)

 

So, what does * and what does ** equal?

 

Let’s look at the x component, the ‘top’. Then 2 = * + 0, so * = 2.

Let’s look at the y component, the ‘bottom’. Then 5 = 0 + **, so ** = 5.

 

Let’s do another example.

 

                        (7)      (7)       (0)         (1)         (0)

(3)  =  (0)  +  (3)   =  7 (0)  +  3 (1)   =  7 Ex  + 3 Ey

 

 

Again, let’s go through the computation as to how we found it:

 

            (7)      (*)      (0)

(3) =   (0)  +  (**)

 

Let’s look at the x component, the ‘top’. Then 7 = * + 0, so * = 7.

Let’s look at the y component, the ‘bottom’. Then 3 = 0 + **, so ** = 3.

 

Now, this leads us to conjecture:

 

            ANY vector in the plane can be written as some number of copies of Ex and some number of copies of Ey.

 

NOTE: just because we’ve checked this for several vectors, doesn’t mean we’ve proven the theorem. For example, I might conjecture every 2x2 matrix is symmetric. Why? Well, look at some matrices:

 

            (2 3)     (5 0)     (5 5)     (7 1)     (2 0)     (2 1)     (12 92)

            (3 2)     (0 5)     (5 5)     (1 7)     (0 5)     (1 2)     (92 12)

 

But this is absurd! Consider

 

            (1 0)                 (1 2)                 (5 9)

            (2 1)                 (3 4)                 (3 3)

 

So we must be careful not to be misled by checking certain special cases. It is a very good idea to test a theorem or conjecture by looking at certain specific cases. This helps lead you to what should be true, but you must prove it in the end.

 

So, in our case, we must show that, given any vector (x,y) in the plane, we can find numbers a and b (where a and b will depend on x and y) such that

 

            (x)                       (1)             (0)

(y)        =          a (0)     +   b (1)           =          a Ex  +  b Ey

 

 

Now, for the two vectors Ex and Ey, it is easy to find an a and a b. Just take a = x and b = y.

 

 

 

 

 

 

 

 

 

 

 

 


Let’s consider a slightly more exotic example.

 

                                    (5)                                            (12)

            V1       =          (0)                    V2       =          (10)

 

Are V1 and V2 a basis? Before showing it is, before showing that we can write any vector as copies of V1 plus copies of V2, let’s do a specific example first. Consider the vector (1700,-500)

 

So we want to solve

 

            (1700)                                        (5)                     (12)

            (-500)              =                      a (0)       +        b (10)

 

We are looking for ‘a’ and ‘b’. We have two equations:

 

(Eq1.1)            1700    =          5a + 12b

 

Unfortunately, this isn’t too easy to just look at and see what ‘a’ and ‘b’ are. Let’s look at the second equation:

 

(Eq1.2)            -500     =          0a + 10b

 

This we can easily solve. We get 10b = -500, or b = -50. Now that we know b, we can substitute this back into (Eq1.1):

 

                 1700 = 5a + 12(-50)

            è 1700 = 5a - 600

            è 2300 = 5a

            è       a = 460

 

So, we get

 

            (1700)                                        (5)                     (12)

            (-500)              =                      a (0)       +        b (10)

 

or

 

            (1700)                                        (5)                       (12)

            (-500)              =                460  (0)        +      -50 (10)

 

So

 

            (1700)

            (-500)              =                460 V1   -    50 V2

 

 

Why does this work? Why are V1 and V2 a basis? Notice that while V2 has a piece in the x direction and a piece in the y direction, V1 only has a piece in the x direction. So if we have a vector (x,y), we must find an a and a b such that (x,y) = aV1 + bV2.

 

Right now, we don’t have to actually find an a and a b, but just show that we could. We show that we determine b first, and then can find a. Since V2 has a y component, we multiply it by whatever is needed to equal the y component of (x,y). We now have b V2. This has the same y component as (x,y), but may not have the correct x component.

 

But this is no problem, as we can still add a number of copies of V1, which is only in the x direction. So we can add whatever we need to correct the x component.

 

Now, let’s prove that V1 and V2 are a basis. So, given a vector (x,y) we need to find a and b such that (x,y) = a V1 + b V2. Now, a and b will be different for different values of x and y. Really,

 

                        a = a(x,y)

                        b = b(x,y)

 

So let’s find them!

 

            (x)                       (5)                 (12)

            (y)        =          a (0)     +      b (10)

 

So we must solve

 

            (Eq1.3)            x = 5a + 12b

            (Eq1.4)            y = 0a + 10b

 

We can solve (Eq1.4) easily, getting b = y/10. Substituting this into (Eq1.3) yields

 

             x = 5a + 12b

             x = 5a + 12(y/10)

             x = 5a + 1.2y

            5a = x - 1.2y

             a = (x - 1.2y) / 5

 

So we have succeeded in finding a and b, given x and y!

 

            a = (x - 1.2y) / 5

            b = y / 10

 

Note that a and b are different for different values of a and b. For the example we did before, namely (x,y) = (1700,-500) what should a and b be?

 

Well, the formulas above give

 

            a = ( 1700 - 1.2*(-500) ) / 5  =  2300 / 5  =  460

            b = -500 / 5  =  -50

 

And this agrees with what we calculated before!

 

 

 

From what we’ve just seen above, we might be led to expect that any two vectors in the plane are a basis. A quick example shows that there are certain pairs of vectors that cannot be a basis. Consider, for example,

           

                                    (2)                                            (4)

            U1       =          (1)                    U2       =          (2)

 

Now, U1 and U2 are along the same line, as U2 is just twice U1. Any multiple of U1 will be in the same direction as U1; any multiple of U2 will be in the same direction as U2. Hence the sum of a multiple of U1 and a multiple of U2 will still be in the direction of U1.

 

Why does this mean U1 and U2 cannot be a basis? Just take any vector (x,y) that’s not in the direction of U1. Then as multiples of U1 and U2 are still in the direction of U1, we cannot get (x,y).

 

 

 

                                                            y-axis                                          direction of U1 and U2

 

 

 

 

 

 


                                                                                   

                                                                                                            x-axis

 

 

 

 

 

 

 

 

 

Again, any number of copies of U1 will still be in the direction of U1. If U2 is in the same direction as U1, then copies of U1 plus copies of U2 will still be in the direction of U1.

 

So, this quick sketch shows why certain pairs cannot be a basis.

 

We have the following:

 

            THEOREM: Let W1 and W2 be any two vectors that are not in the same direction (ie, that do not lie on the same line). Then W1 and W2 are a basis.

 

Let’s assume one of the vectors is in the direction of the x axis, and draw a picture.

 

 

 

 

 


                                                                                                                        W2

 

 

 

 

 

 


                                                                                                      W1

 

 

 

 

 

 

I don’t really want to go into a theoretical, rigorous proof, so I’ll just do it in the case when W1 is in the direction of the x axis.

 

Let’s just do a sketch. (Sorry for the pun). The first vector W1 equals, say,  (W1­x, 0).

 

W2 has a non-zero component in the y direction. We’re trying to get the vector (x,y). Let’s say W2 = (W2x , W2­Y). We need to solve

 

            (x,y)  =  a W1 + b W2

            (x,y)  =  a W1 + b (W2x , W2­Y)

 

            If we take b = y / W2­Y,  (we can divide by W2­Y as it is not zero) then we get

 

            (x,y) =  a W1 + y / W2­Y (W2x , W2­Y)

            (x,y) =  a W1 + (y W2x / W2­Y, y W2­Y / W2­Y)

            (x,y) =  a (W1­x, 0)  + (y W2x / W2­Y, y)

            (x,y) =  (a W1­x, 0)  + (y W2x / W2­Y, y)

 

So the y component on the Left Hand Side equals the y component on the Right Hand Side. We now solve for a:

 

            (a W1­x, 0)  =  (x,y) - (y W2x / W2­Y, y)

            (a W1­x, 0)  =  (x   -   y W2x / W2­Y, 0)

 

Then we can solve for a:

 

            a =  (x   -   y W2x / W2­Y)   /   W1­x.

 

A similar argument would work for any two vectors W1, W2 that are not in the same direction.

 

 

The last thing we’re going to do is how to find a and b for W1, W2. I’ll give a method (using Gaussian Elimination) that will always work, although I won’t prove why.

 

Let W1 = (A, B) and W2 = (C,D).

 

We want, given the vector (x,y), to find a and b such that

 

            (x)                      (A)                    (C)

(y)        =          a (B)      +        b (D)

 

So we have

 

            (x)                    (aA)                 (bC)

(y)        =          (aB)     +         (bD)

 

 

            (x)                    (aA + bC)

            (y)        =          (aB + bD)

 

 

            (x)                    (Aa + Cb)

            (y)        =          (Ba + Db)       

 

 

            (x)                    (A   C) (a)

(y)        =          (B   D) (b)

 

 

Now we use Gaussian Elimination to solve! So, we say: what must we multiple the first row (A  C)  by so that, when we add it to  the second row (B  D)  we get  (0 something).

 

So A*m + B = 0, so we multiply the first row by -B/A. Etc...