1
Consider the system
๐ฅโฒ=2๐ฆโ3๐ง,๐ฆโฒ=3๐ฅโ๐ฆโ3๐ง,๐งโฒ=โ2๐ฅ+2๐ฆโ๐ง.
(a)Write the matrix of this system, and find and factor its characteristic
polynomial. Then find an eigenvector for each eigenvalue.
Solution
The matrix of the system is
๐ด=โโ
โ
โโ02โ33โ1โ3โ22โ1โโ
โ
โโ .
Expanding the determinant gives
det(๐๐ผโ๐ด)=๐3+2๐2โ5๐โ6.
Trying integer roots, ๐ =2 works: 8 +8 โ10 โ6 =0. Dividing by ๐ โ2
leaves a quadratic that factors as well:
det(๐๐ผโ๐ด)=(๐โ2)(๐2+4๐+3)=(๐โ2)(๐+1)(๐+3).
The eigenvalues are 2, โ1, and โ3. Solving (๐ด โ2๐ผ)๐ฃ =0, the first
two rows say โ2๐ฅ +2๐ฆ โ3๐ง =0 and 3๐ฅ โ3๐ฆ โ3๐ง =0. Together these force ๐ง =0
and ๐ฅ =๐ฆ. Solving (๐ด +๐ผ)๐ฃ =0, the last two rows give ๐ง =๐ฅ and ๐ฆ =๐ฅ.
Solving (๐ด +3๐ผ)๐ฃ =0, the last row gives ๐ฆ =๐ฅ โ๐ง, and then the first row
becomes 5๐ฅ โ5๐ง =0, so ๐ฅ =๐ง and ๐ฆ =0. This gives the eigenvectors
๐ข1=โโ
โ
โโ110โโ
โ
โโ ,๐ข2=โโ
โ
โโ111โโ
โ
โโ ,๐ข3=โโ
โ
โโ101โโ
โ
โโ ,
with eigenvalues 2, โ1, and โ3, respectively.
(b)Find the solution with ๐ฅ(0) =1, ๐ฆ(0) =0, ๐ง(0) =0.
Solution
Write the initial state in terms of the eigenvectors. Solving
๐1๐ข1 +๐2๐ข2 +๐3๐ข3 =(1,0,0) gives ๐1 =1, ๐2 = โ1, ๐3 =1:
โโ
โ
โโ100โโ
โ
โโ =๐ข1โ๐ข2+๐ข3.
Each eigenvector component evolves by its own exponential, so
๐(๐ก)=๐2๐ก๐ข1โ๐โ๐ก๐ข2+๐โ3๐ก๐ข3,
that is,
๐ฅ(๐ก)=๐2๐กโ๐โ๐ก+๐โ3๐ก,๐ฆ(๐ก)=๐2๐กโ๐โ๐ก,๐ง(๐ก)=๐โ3๐กโ๐โ๐ก.
As a check, ๐ฆโฒ =2๐2๐ก +๐โ๐ก, and
3๐ฅ โ๐ฆ โ3๐ง =3๐2๐ก โ3๐โ๐ก +3๐โ3๐ก โ๐2๐ก +๐โ๐ก โ3๐โ3๐ก +3๐โ๐ก is
the same.
(c)Describe what happens to solutions as ๐ก โโ. Which initial states
give solutions that approach the origin? What happens to every other
solution, and how can you tell from the initial state which way it
heads?
Solution
The three eigenvectors form a basis, so every initial state is
๐0 =๐1๐ข1 +๐2๐ข2 +๐3๐ข3. Solving for the coefficients in terms of
๐0 =(๐ฅ0,๐ฆ0,๐ง0) gives
๐1=๐ฅ0โ๐ง0,๐2=๐ฆ0+๐ง0โ๐ฅ0,๐3=๐ฅ0โ๐ฆ0,
and the solution is
๐(๐ก)=๐1๐2๐ก๐ข1+๐2๐โ๐ก๐ข2+๐3๐โ3๐ก๐ข3.
As ๐ก โโ, the last two terms decay to zero. So everything
depends on ๐1.
If ๐1 =0, the solution approaches the origin. These are exactly the
initial states with ๐ฅ0 =๐ง0: the plane spanned by the two contracting
eigenvectors ๐ข2 and ๐ข3.
If ๐1 โ 0, the decaying terms die out and the solution grows like
๐1๐2๐ก๐ข1. It moves off to infinity along the line through ๐ข1:
in the direction of ๐ข1 when ๐ฅ0 >๐ง0, and in the direction of โ๐ข1
when ๐ฅ0 <๐ง0. The plane ๐ฅ =๐ง separates these two outcomes. The
solution in the previous part has ๐1 =1, and its ๐2๐ก terms carry
it out along ๐ข1.
2
Consider the system
๐ฅโฒ=2๐ฆ+2๐ง,๐ฆโฒ=๐ฅโ๐ฆโ2๐ง,๐งโฒ=โ2๐ฅโ๐ฆ.
Let ๐ด be the matrix of this system. In Chapter 9, a quadratic factor
with no real roots led us to an invariant plane. Here is the procedure.
First complete the square to
write the factor as (๐ โ๐)2 +๐2 with ๐ >0. Then find a
nonzero vector ๐ฃ solving
((๐ดโ๐๐ผ)2+๐2๐ผ)๐ฃ=0,
and set
๐ค=โ1๐(๐ดโ๐๐ผ)๐ฃ.
Then ๐ด๐ฃ =๐๐ฃ โ๐๐ค and ๐ด๐ค =๐๐ฃ +๐๐ค. So ๐ฃ and ๐ค span a plane
that the flow never leaves, and in the basis ๐ฃ,๐ค the matrix of ๐ด on
this plane is
(๐๐โ๐๐),with flow๐๐๐ก(cosโก๐๐กsinโก๐๐กโsinโก๐๐กcosโก๐๐ก).
On this plane, states rotate at angular speed ๐ while scaling by
๐๐๐ก.
(a)Write the matrix of this system, and find and factor its characteristic
polynomial. What motion does each factor describe? For the linear
factor, give the exponential rate along its line. For the quadratic
factor, give the scaling rate and the time for one full turn in its
plane.
Solution
The matrix of the system is
๐ด=โโ
โ
โโ0221โ1โ2โ2โ10โโ
โ
โโ .
Expanding the determinant gives
det(๐๐ผโ๐ด)=๐3+๐2โ2.
Trying integer roots, ๐ =1 works. Dividing by ๐ โ1 and completing the
square,
det(๐๐ผโ๐ด)=(๐โ1)(๐2+2๐+2)=(๐โ1)((๐+1)2+1).
The linear factor ๐ โ1 gives an invariant line along which states grow
like ๐๐ก. The quadratic factor has no real roots; it has ๐ = โ1 and
๐ =1. It gives an invariant plane on which states rotate at angular
speed 1 while shrinking like ๐โ๐ก. One full turn takes time 2๐,
and each turn shrinks a state by the factor ๐โ2๐ โ0.002.
(b)Find an eigenvector ๐ข for the linear factor. Then use the procedure above
to find vectors ๐ฃ and ๐ค spanning the invariant plane.
Solution
Solving (๐ด โ๐ผ)๐ข =0, the first row says โ๐ฅ +2๐ฆ +2๐ง =0, so ๐ฅ =2(๐ฆ +๐ง).
The last row says โ2๐ฅ โ๐ฆ โ๐ง =0, which becomes โ5(๐ฆ +๐ง) =0. So ๐ง = โ๐ฆ
and ๐ฅ =0, and we can take
๐ข=โโ
โ
โโ01โ1โโ
โ
โโ .
For the plane, ๐ = โ1 and ๐ =1, so we need (๐ด +๐ผ)2 +๐ผ. Multiplying
out,
๐ด+๐ผ=โโ
โ
โโ12210โ2โ2โ11โโ
โ
โโ ,(๐ด+๐ผ)2+๐ผ=โโ
โ
โโ000550โ5โ50โโ
โ
โโ .
Its nonzero solutions are the nonzero vectors with ๐ฅ +๐ฆ =0, so the
invariant plane is ๐ฅ +๐ฆ =0. Choose ๐ฃ =(0,0,1). Then
๐ค=โ(๐ด+๐ผ)๐ฃ=โโโ
โ
โโ2โ21โโ
โ
โโ =โโ
โ
โโโ22โ1โโ
โ
โโ .
As a check, ๐ด๐ฃ =(2, โ2,0) = โ๐ฃ โ๐ค and ๐ด๐ค =(2, โ2,2) =๐ฃ โ๐ค, matching
๐ด๐ฃ =๐๐ฃ โ๐๐ค and ๐ด๐ค =๐๐ฃ +๐๐ค with ๐ = โ1, ๐ =1. The
eigenvector ๐ข does not lie in the plane, since its coordinates satisfy
๐ฅ +๐ฆ =1.
(c)Find the solution with ๐ฅ(0) =2, ๐ฆ(0) = โ1, ๐ง(0) =0.
Solution
Solving ๐1๐ข +๐2๐ฃ +๐3๐ค =(2, โ1,0) gives ๐1 =1, ๐2 =0, ๐3 = โ1:
โโ
โ
โโ2โ10โโ
โ
โโ =๐ขโ๐ค.
The line component evolves as ๐๐ก๐ข. The plane component has
coordinates (0, โ1) in the basis ๐ฃ,๐ค, and the planar flow carries
these to
๐โ๐ก(cosโก๐กsinโก๐กโsinโก๐กcosโก๐ก)(0โ1)=๐โ๐ก(โsinโก๐กโcosโก๐ก).
So
๐(๐ก)=๐๐ก๐ข+๐โ๐ก(โsinโก๐ก๐ฃโcosโก๐ก๐ค),
that is,
๐ฅ(๐ก)=2๐โ๐กcosโก๐ก,๐ฆ(๐ก)=๐๐กโ2๐โ๐กcosโก๐ก,๐ง(๐ก)=โ๐๐ก+๐โ๐ก(cosโก๐กโsinโก๐ก).
As a check, ๐ฅโฒ = โ2๐โ๐ก(cosโก๐ก +sinโก๐ก), and
2๐ฆ +2๐ง =2๐๐ก โ4๐โ๐กcosโก๐ก โ2๐๐ก +2๐โ๐กcosโก๐ก โ2๐โ๐กsinโก๐ก is the same.
(d)Describe what happens as ๐ก โโ to solutions starting on the
invariant line, solutions starting in the invariant plane, and solutions
starting anywhere else.
Solution
A state on the line, ๐๐ข, moves straight out along the line as
๐๐๐ก๐ข, growing exponentially.
A state in the plane spirals into the origin. It turns once every 2๐
units of time and shrinks by the factor ๐โ2๐ with each turn.
Every other state has a component on the line and a component in the
plane. The line component grows like ๐๐ก, while the plane component
spirals in and dies out. So the spiraling fades, the distance from the
solution to the line goes to zero, and the solution heads off to
infinity along the line. The solution from the previous part shows all
of this: the terms ๐โ๐กcosโก๐ก and ๐โ๐กsinโก๐ก are the dying
spiral, and the ๐๐ก terms carry the state out along ๐ข.
Which way it heads depends on the sign of the line component ๐1. The
plane is ๐ฅ +๐ฆ =0, and ๐ข has ๐ฅ +๐ฆ =1. So if ๐0 =๐1๐ข +๐ with ๐ in
the plane, then ๐ฅ0 +๐ฆ0 =๐1. Therefore:
if ๐ฅ0 +๐ฆ0 >0, the solution heads out in the direction of ๐ข;
if ๐ฅ0 +๐ฆ0 <0, it heads out in the direction of โ๐ข;
if ๐ฅ0 +๐ฆ0 =0, the state lies in the plane and spirals into the
origin.
The invariant plane separates the two outcomes.
3
In each part, write the matrix as a sum of two pieces that commute, and
use this to compute its exponential.
(a)Compute ๐๐ก๐ต, where
๐ต=โโ
โ
โโโ2100โ2100โ2โโ
โ
โโ .
Solution
Split off the diagonal:
๐ต=โ2๐ผ+๐,๐=โโ
โ
โโ010001000โโ
โ
โโ .
A multiple of the identity commutes with every matrix:
( โ2๐ผ)๐ = โ2๐ =๐( โ2๐ผ). Multiplying out the powers of ๐,
๐2=โโ
โ
โโ001000000โโ
โ
โโ ,๐3=0.
Every higher power is zero as well, so the exponential series stops
after three terms:
๐๐ก๐=๐ผ+๐ก๐+๐ก22๐2=โโ
โ
โ
โโ1๐ก๐ก2/201๐ก001โโ
โ
โ
โโ .
Since the two pieces commute, ๐๐ก๐ต =๐โ2๐ก๐ผ๐๐ก๐ =๐โ2๐ก๐๐ก๐:
๐๐ก๐ต=๐โ2๐กโโ
โ
โ
โโ1๐ก๐ก2/201๐ก001โโ
โ
โ
โโ .
๐ถ=โโ
โ
โ
โ
โ
โ
โโ0210โ2001000200โ20โโ
โ
โ
โ
โ
โ
โโ =(๐
๐ผ0๐
),๐
=(02โ20),
where ๐ผ is the 2 ร2 identity and 0 is the 2 ร2 zero
matrix.
Solution
Split ๐ถ into its block-diagonal part and the rest:
๐ถ=๐ท+๐,๐ท=(๐
00๐
),๐=(0๐ผ00).
Multiplying block by block,
๐ท๐=(๐
00๐
)(0๐ผ00)=(0๐
00),๐๐ท=(0๐ผ00)(๐
00๐
)=(0๐
00).
These agree because the same block ๐
appears in both diagonal
positions, and ๐
๐ผ =๐ผ๐
.
The matrix ๐ท is block diagonal, so we exponentiate each block. The
block ๐
is rotation with ๐ =0 and ๐ =2, so
๐๐ก๐
=(cosโก2๐กsinโก2๐กโsinโก2๐กcosโก2๐ก),๐๐ก๐ท=(๐๐ก๐
00๐๐ก๐
).
For ๐, multiplying block by block gives ๐2 =0, so the series stops
after two terms:
๐๐ก๐=๐ผ+๐ก๐=(๐ผ๐ก๐ผ0๐ผ).
Since ๐ท and ๐ commute, ๐๐ก๐ถ =๐๐ก๐ท๐๐ก๐:
๐๐ก๐ถ=(๐๐ก๐
00๐๐ก๐
)(๐ผ๐ก๐ผ0๐ผ)=(๐๐ก๐
๐ก๐๐ก๐
0๐๐ก๐
).
Written out in full,
๐๐ก๐ถ=โโ
โ
โ
โ
โ
โ
โโcosโก2๐กsinโก2๐ก๐กcosโก2๐ก๐กsinโก2๐กโsinโก2๐กcosโก2๐กโ๐กsinโก2๐ก๐กcosโก2๐ก00cosโก2๐กsinโก2๐ก00โsinโก2๐กcosโก2๐กโโ
โ
โ
โ
โ
โ
โโ .
4
A 6 ร6 matrix ๐ด has characteristic polynomial
(๐+3)(๐โ1)(๐2+4)(๐2+2๐+5).
We do not need the matrix itself to understand the motion of ๐โฒ =๐ด๐.
(a)Make a table with one row for each factor. In each row, record whether
the factor gives an invariant line or an invariant plane, whether the
motion there grows, decays, or neither, and the period of any
oscillation.
Solution
Each linear factor ๐ โ๐ gives an invariant line along which states
scale by ๐๐๐ก. Each quadratic factor with no real roots, written
as (๐ โ๐)2 +๐2, gives an invariant plane on which states rotate at
angular speed ๐ while scaling by ๐๐๐ก; one full turn takes time
2๐/๐. Here ๐2 +4 =๐2 +22 has ๐ =0, ๐ =2, and completing
the square gives ๐2 +2๐ +5 =(๐ +1)2 +4, with ๐ = โ1, ๐ =2.
| Factor | Line or plane? | Grows, decays, or neither? | Period of oscillation, if any |
| ๐ +3 | Line | Decays, like ๐โ3๐ก | None |
| ๐ โ1 | Line | Grows, like ๐๐ก | None |
| ๐2 +4 | Plane | Neither | ๐ |
| ๐2 +2๐ +5 | Plane | Decays, like ๐โ๐ก | ๐ |
The dimensions add up to 1 +1 +2 +2 =6, so these two lines and two planes
account for the whole state space.
(b)Describe what happens as ๐ก โโ to the solution starting from each
of these states:
a state on the invariant line from ๐ +3;
a state in the invariant plane from ๐2 +4;
a state with components only in the two invariant planes;
a state with a nonzero component on the invariant line from ๐ โ1.
Solution
No factor repeats, so the two lines and two planes together give a basis
of the state space. Every initial state splits into a component in each
piece, and each component evolves on its own.
The solution decays to the origin like ๐โ3๐ก, without
oscillating.
The solution circles forever, returning to its starting state every
๐ units of time.
The component in the plane from ๐2 +2๐ +5 spirals in to the origin,
while the component in the plane from ๐2 +4 keeps circling. So the
solution settles into circling with period ๐. (If the component
in the plane from ๐2 +4 is zero, the solution simply spirals into
the origin.)
The component on the line from ๐ โ1 grows like ๐๐ก. The components
in the two planes stay bounded or die out, and any component on the
line from ๐ +3 dies out. So the solution grows without bound,
heading off in the direction of the line from ๐ โ1.
5
Every cell carries instructions for building proteins in its genes. To
make a protein, the cell copies a gene into a molecule called mRNA, then
builds the protein from that copy. Some proteins are repressors: a
repressor binds to the DNA at the start of another gene and blocks that
gene from being copied. The more repressor there is, the less of the
other protein gets made. Proteins don't last forever, either. The cell
breaks them down, and they are diluted as the cell grows and divides.
In 2000, Michael Elowitz and Stanislas Leibler used these ingredients to
build a biological clock from scratch in the bacterium E. coli. They
chose three repressor genes that occur in no natural clock and arranged
them in a ring:
the protein LacI represses the gene for TetR;
the protein TetR represses the gene for CI;
the protein CI represses the gene for LacI.
They called the circuit the repressilator. To watch it, they added a
gene for a green fluorescent protein that TetR also represses, so a cell
glows when its TetR level is low. Under a microscope, individual cells
blinked on and off, with a period of around 150 minutes. That is about
three times as long as it takes the cells to divide.
Why would this ring oscillate? Suppose a cell has a little extra LacI.
Extra LacI shuts down TetR production, so TetR falls. With less TetR, the
gene for CI is released, so CI rises. Extra CI shuts down LacI
production, so LacI falls. The original excess of LacI has come back
around the ring as a shortage of LacI. Each step takes time, because
proteins are built and removed gradually, so the levels keep
overshooting. Whether these swings die away or grow depends on how
strongly each repressor acts. That is what we can decide with linear
algebra.
Here is a simplified model, which tracks only the proteins. In a steady
state, all three proteins sit at the same level, with production exactly
balancing removal. Let ๐ฅ, ๐ฆ, and ๐ง be the deviations of LacI, TetR,
and CI from that steady level. We measure time in units of a protein's
lifetime, so each protein is removed at rate equal to its own deviation.
Each production rate is a decreasing function of the repressor upstream
of it. For small deviations, we replace that curve by its tangent line,
whose slope we write as โ๐พ. This gives
๐ฅโฒ=โ๐ฅโ๐พ๐ง,๐ฆโฒ=โ๐ฆโ๐พ๐ฅ,๐งโฒ=โ๐งโ๐พ๐ฆ.
For example, the term โ๐พ๐ฅ in the equation for ๐ฆโฒ says that extra
LacI (๐ฅ >0) lowers TetR production. The constant ๐พ >0 measures how
sharply production responds to its repressor.
(a)Write the matrix of this system, and find and factor its characteristic
polynomial. Don't multiply out (๐ +1)3. Instead, write ๐ =๐ +1 and use
the identity
๐ 3+๐พ3=(๐ +๐พ)(๐ 2โ๐พ๐ +๐พ2),
which you can check by multiplying out the right side. Then complete the
square in the quadratic factor.
Solution
The matrix of the system is
๐ด=โโ
โ
โโโ10โ๐พโ๐พโ100โ๐พโ1โโ
โ
โโ .
Expanding along the first row,
det(๐๐ผโ๐ด)=detโโ
โ
โโ๐+10๐พ๐พ๐+100๐พ๐+1โโ
โ
โโ =(๐+1)3+๐พโ
๐พ2=(๐+1)3+๐พ3.
With ๐ =๐ +1, the identity gives
๐ 3+๐พ3=(๐ +๐พ)(๐ 2โ๐พ๐ +๐พ2).
Completing the square in the quadratic factor,
๐ 2โ๐พ๐ +๐พ2=(๐ โ๐พ2)2+๐พ2โ๐พ24=(๐ โ๐พ2)2+34๐พ2.
Substituting ๐ =๐ +1,
det(๐๐ผโ๐ด)=(๐+1+๐พ)((๐+1โ๐พ2)2+34๐พ2).
Since ๐พ >0, the quadratic factor has no real roots.
(b)The quadratic factor describes oscillation in an invariant plane. Find
its frequency and its scaling rate in terms of ๐พ.
Solution
Comparing (๐ +1 โ๐พ2)2 +34๐พ2 with
(๐ โ๐)2 +๐2 gives
๐=๐พ2โ1,๐=โ32๐พ.
In this plane, the protein levels oscillate at angular frequency
โ32๐พ while scaling by ๐(๐พ/2โ1)๐ก.
(c)If the oscillations in the invariant plane die out, the protein levels
settle back to their steady state, and the clock stops ticking. For the
circuit to work as a clock, the oscillations must grow, so that they
become large enough to notice and measure. For which values of ๐พ does
this happen?
Solution
The oscillations grow when the scaling rate is positive:
๐=๐พ2โ1>0โบ๐พ>2.
For ๐พ <2 they die out, and at ๐พ =2 they neither grow nor decay. The
linear factor gives an invariant line along which deviations decay at
rate 1 +๐พ, for every value of ๐พ, so it never gets in the way. The
circuit can work as a clock when ๐พ >2: production must respond to the
repressor steeply enough.