5.5 Appendix
5.5.1 Appendix I: The Proof of Proposition 5.4
This section gives the proof of Proposition 5.4. To begin with, the following claims are obvious and will be useful.
Claim 1 If z = 0, u6= u∗, we have
z0 = `(w)q(u) has the same sign with [w· (u − u∗)]. (5.61) Claim 2 If z = 0, u = u∗, we have z0= 0 and
z00= `(w)q0(u)v has the same sign with − v. (5.62) Claim 3 If v = 0, g(w)− p(u) 6= 0, u 6= 0, we have
dv0= h(u)[g(w)− p(u)] has the same sign with [g(w) − p(u)]. (5.63) Claim 4 If v = 0, g(w)− p(u) = 0, u 6= 0, we have v0= 0
dv00= h(u)g0(w)z has the same sign with z. (5.64) Claim 5 If v6= 0, g(w) − p(u) = 0, we have
[g(w)− p(u)]0 = g0(w)z− p0(u)v > 0,if z≥ 0, v ≥ 0, (z, v) 6= (0, 0). (5.65) [g(w)− p(u)]0 = g0(w)z− p0(u)v < 0,if z≤ 0, v ≤ 0, (z, v) 6= (0, 0). (5.66)
The exit set W−must be a subset of ∂W . To find W−, we need to analyze the dynamics of (5.31) on each portion of ∂W .
In the portion ∂R\ P . Set ∂R = R0∪ R1∪ R2∪ R3 with
R0={(u, v, w, z) : u = 0, g(w) − p(u) ≤ 0, v ≤ 0}, R1={(u, v, w, z) : u = u∗, g(w)− p(u) ≤ 0, v ≤ 0}, R2={(u, v, w, z) : u < u∗, g(w)− p(u) = 0, v ≤ 0}, R3={(u, v, w, z) : u < u∗, g(w)− p(u) < 0, v = 0}.
Now we investigate the behavior of solutions on each Ri. On region R0, we consider the following two subsets.
(1) Assume u = 0, v < 0.
Since v < 0, we know that u+< 0 and this implies this set belongs to W−. (2) Assume u = 0, v = 0.
In this case, (0, 0, w, z) will stay at W for any (w, z)∈ R2. Thus, (0, 0, w, z)∈ J2. On region R1, we consider the following four subsets.
(1) Assume u = u∗, w = w∗, v < 0.
If z≤ 0, then
[g(w)− p(u)]0 = g0(w)z− p0(u)v≤ −p0(u)v < 0.
Hence, the any trajectory of solutions will enter the region R. On the other hand, if z > 0, then w+> w∗ and this implies that any trajectory of solutions will enter the region P .
(2) Assume u = u∗, w < w∗, v < 0.
In this case, it’s easy to see that any trajectory of solutions will enter region R.
(3) Assume u = u∗, w = w∗, v = 0.
If z = 0, it’s obviously that (u∗, 0, w∗, 0) /∈ W−. If z > 0, then w+> w∗, v0 = 0,
[g(w)− p(u)]0= [g0(w)z− p0(u)v] = g0(w)z > 0,
dv00= cv0+ h0(u)v[g(w)− p(u)] + h(u)[g(w) − p(u)]0 = h(u)[g(w)− p(u)]0> 0.
Thus, v+> 0, u0 > 0 and u+> u∗. Then, any trajectory of solutions will enter the region S.
If z < 0, similar to case of z > 0, we can obtain w+ < w∗ and u+ < u∗. Hence, any trajectory of solutions will enter the region R.
(4) Assume u = u∗, w < w∗, v = 0.
In this case, we have
dv0= cv + h(u∗)[g(w)− p(u∗)] < cv + h(u∗)[g(w∗)− p(u∗)] < 0
and this implies v+ < 0. Since u0 = v < 0, we then obtain u+ < u∗ and g(w)− p(u∗) <
g(w∗)− p(u∗) = 0. Hence, any trajectory of solutions will enter the region R.
On region R2, we consider the following two subsets.
(1) Assume 0 < u < u∗, g(w)− p(u) = 0, v < 0.
will enter the region P . If z = 0, we have
z0 = cz + l(w)q(u) > 0 + l(w)q(u∗) = 0
and this implies z+> 0 and w+> w∗. Then any trajectory of solutions will enter the region P . If z < 0, it is easy to check that
[g(w)− p(u)]0= g0(w)z− p0(u)v < g0(w)z < 0.
Hence [g(w)− p(u)]+< 0 and any trajectory of solutions will enter the region R.
(2) Assume 0 < u < u∗, g(w)− p(u) = 0, v = 0.
If z≥ 0, by the same arguments as (1), any trajectory of solutions will enter the region P . If z < 0, we have
[g(w)− p(u)]0= g0(w)z < 0⇒ [g(w) − p(u)]+< 0, dv0= cv + g(w)− p(u) = 0,
dv00= cv0+ h0(u)v[g(w)− p(u)] + h(u)[g0(w)z− p0(u)v]
= 0 + 0 + h(u)g0(w)z + 0 < 0,
and this implies v+< 0. Hence, any trajectory of solutions will enter the region R.
On region R3, we have 0 < u < u∗, g(w)− p(u) < 0, v = 0. Thus,
dv0 = cv + h(u)[g(w)− p(u)] = 0 + h(u)[g(w) − p(u)] < 0.
Hence, any trajectory of solutions will enter the region R. The proof is complete.
In the portion ∂S\ Q, we set ∂S = S1∪ S2∪ S3 with
S1={(u, v, w, z)|u = u∗, g(w)− p(u) ≥ 0, v ≥ 0}, S2={(u, v, w, z)|u > u∗, g(w)− p(u) = 0, v ≥ 0}, S3={(u, v, w, z)|u > u∗, g(w)− p(u) > 0, v = 0}.
Now we investigate the behavior of solutions on each Si. On region S1, we consider the following four subsets.
(1) Assume u = u∗, w = w∗, v > 0.
If z≥ 0, then
[g(w)− p(u)]0 = g0(w)z− p0(u)v≥ −p0(u)v > 0.
Hence, the any trajectory of solutions will enter the region S.
On the other hand, if z < 0, then w+< w∗ and this implies that any trajectory of solutions will enter the region Q.
(2) Assume u = u∗, w > w∗, v > 0.
In this case, it’s easy to see that any trajectory of solutions will enter region S.
(3) Assume u = u∗, w = w∗, v = 0.
If z = 0, it’s obviously that (u∗, 0, w∗, 0) /∈ W−. If z > 0, then w+> w∗, v0 = 0,
[g(w)− p(u)]0= [g0(w)z− p0(u)v] = g0(w)z > 0,
dv00= cv0+ h0(u)v[g(w)− p(u)] + h(u)[g(w) − p(u)]0 = h(u)[g(w)− p(u)]0> 0.
Thus, v+> 0, u0 > 0 and u+> u∗. Then, any trajectory of solutions will enter the region S.
If z < 0, similar to case of z > 0, we can obtain w+ < w∗ and u+ < u∗. Hence, any trajectory of solutions will enter the region R.
(4) Assume u = u∗, w > w∗, v = 0.
In this case, we have
dv0= cv + h(u∗)[g(w)− p(u∗)] > cv + h(u∗)[g(w∗)− p(u∗)] = 0
and this implies v+ > 0. Since u0 = v < 0, we then obtain u+ > u∗ and g(w)− p(u∗) >
g(w∗)− p(u∗) = 0. Hence, any trajectory of solutions will enter the region S.
On region S2, we consider the following two subsets.
(1) Assume u > u∗, g(w)− p(u) = 0, v > 0.
If z < 0, then u > u∗and g(w)−p(u) = 0 imply that w < w∗. Hence, any trajectory of solutions will enter the region Q.
If z = 0, we have
z0 = cz + l(w)q(u) < 0 + l(w)q(u∗) = 0
and this implies z+< 0 and w+< w∗. Then any trajectory of solutions will enter the region Q.
If z > 0, it is easy to check that
[g(w)− p(u)]0= g0(w)z− p0(u)v > g0(w)z > 0.
(2) Assume u > u∗, g(w)− p(u) = 0, v = 0.
If z≤ 0, by the same arguments as (1), any trajectory of solutions will enter the region Q.
If z > 0, we have
[g(w)− p(u)]0= g0(w)z > 0⇒ [g(w) − p(u)]+> 0, dv0= cv + g(w)− p(u) = 0,
dv00= cv0+ h0(u)v[g(w)− p(u)] + h(u)[g0(w)z− p0(u)v]
= 0 + 0 + h(u)g0(w)z + 0 > 0,
and this implies v+> 0. Hence, any trajectory of solutions will enter the region S.
On region S3, we have u > u∗, g(w)− p(u) > 0, v = 0. Thus,
dv0 = cv + h(u)[g(w)− p(u)] = 0 + h(u)[g(w) − p(u)] > 0.
Hence, any trajectory of solutions will enter the region S.
In the portion ∂P\ R, we set ∂P = P0∪ P1∪ P12∪ P2∪ P3∪ P13∪ P123 with P0={(u, v, w, z)|u = 0, w ≥ w∗, z≥ 0},
P1={(u, v, w, z)|u = u∗, w > w∗, z > 0}, P12={(u, v, w, z)|u = u∗, w = w∗, z > 0},
P2={(u, v, w, z)|u ∈ (0, u∗), w = w∗, z > 0}, P3={(u, v, w, z)|u ∈ (0, u∗), w≥ w∗, z = 0} P13={(u, v, w, z)|u = u∗, w > w∗, z = 0} P123={(u, v, w, z)|u = u∗, w = w∗, z = 0}.
Now we investigate the behavior of solutions on each Pi. On region P0, we consider the following three subsets.
(1) Assume v < 0.
Since v < 0, we know that u+< 0 and this implies this set belongs to W−. (2) Assume v = 0.
In this case, (0, 0, w, z) will stay at W for any (w, z)∈ R2. Thus, (0, 0, w, z)∈ J2.
(3) Assume v > 0.
Since v > 0, we know that u+> 0. (5.61) implies z+ > 0 and w+> w∗. Thus, the trajectory of solutions will enter the region P .
On region P1, we have g(w)− p(u) > 0 and consider the following three subsets.
(1) Assume v < 0.
Since v < 0, we know that u+ < u∗ and this implies the trajectory of solutions will enter the region P .
(2) Assume v = 0.
(5.63) implies v+> 0 and u+> u∗. Thus, the trajectory of solutions will enter the region S.
(3) Assume v > 0.
Since v > 0, we know that u+> u∗. Thus, the trajectory of solutions will enter the region S.
On region P12, we have g(w)− p(u) = 0, w+> w∗ and consider the following three subsets.
(1) Assume v < 0.
Since v < 0, we know that u+ < u∗ and this implies the trajectory of solutions will enter the region P .
(2) Assume v = 0.
(5.64) implies v+> 0 and u+ > u∗. (5.65) implies [g(w)− p(u)]+> 0. Thus, the trajectory of solutions will enter the region S.
(3) Assume v > 0.
Since v > 0, we know that u+> u∗. (5.65) implies [g(w)− p(u)]+ > 0. Thus, the trajectory of solutions will enter the region S.
On region P2, we have w+> w∗. Hence, the trajectory of solutions will enter the region P .
On region P3, (5.61) implies z+ > 0 and w+ > w∗. Hence, the trajectory of solutions will enter the region P .
On region P13, we have g(w)− p(u) > 0 and consider the following three subsets.
(1) Assume v < 0.
Since v < 0, we know that u+< u∗. (5.62) implies z+> 0. the trajectory of solutions will enter the region P .
(2) Assume v = 0.
(3) Assume v > 0.
Since v > 0, we know that u+> u∗. Thus, the trajectory of solutions will enter the region S.
In the portion ∂Q\ S, we set ∂Q = Q0∪ Q1∪ Q12∪ Q2∪ Q3+∪ Q3−∪ Q13+∪ Q13−∪ P123with Q0={(u, v, w, z)|u ≤ u∗, w = 0, z = 0},
Q1={(u, v, w, z)|u = u∗, w < w∗, z < 0}, Q12={(u, v, w, z)|u = u∗, w = w∗, z < 0}, Q2={(u, v, w, z)|u > u∗, w = w∗, z < 0}, Q3+={(u, v, w, z)|u > u∗, 0 < w < w∗, z = 0} Q3−={(u, v, w, z)|u > u∗, w < 0, z = 0} Q13+={(u, v, w, z)|u = u∗, 0 < w < w∗, z = 0}.
Q13−={(u, v, w, z)|u = u∗, w < 0, z = 0}.
Now we investigate the behavior of solutions on each Pi.
On region Q0, The trajectory of solutions are invariant in V and included in J10. On region Q1, we have g(w)− p(u) < 0 and consider the following three subsets.
(1) Assume v > 0.
Since v > 0, we know that u+ > u∗ and this implies the trajectory of solutions will enter the region Q.
(2) Assume v = 0.
(5.63) implies v+< 0 and u+< u∗. Thus, the trajectory of solutions will enter the region R.
(3) Assume v < 0.
Since v < 0, we know that u+< u∗. Thus, the trajectory of solutions will enter the region R.
On region Q12, we have g(w)− p(u) = 0, w+< w∗and consider the following three subsets.
(1) Assume v > 0.
Since v > 0, we know that u+ > u∗ and this implies the trajectory of solutions will enter the region Q.
(2) Assume v = 0.
(5.64) implies v+< 0 and u+ < u∗. (5.65) implies [g(w)− p(u)]+< 0. Thus, the trajectory of solutions will enter the region R.
(3) Assume v < 0.
Since v < 0, we know that u+< u∗. (5.65) implies [g(w)− p(u)]+ < 0. Thus, the trajectory of solutions will enter the region R.
On region Q2, we have w+< w∗. Hence, the trajectory of solutions will enter the region Q.
On region Q3+, (5.61) implies z+< 0 and w+< w∗. Hence, the trajectory of solutions will enter the region Q.
On region Q3−, (5.61) implies z+ > 0 and it does not enter Q immediately. Since u > u∗, we know that it does not enter P or R immediately. We consider the following four subsets.
(1) Assume v < 0.
Since v > 0, it does not enter S immediately. Thus, the trajectory of solutions does not exit W immediately and it is included in J12.
(2) Assume v≥ 0 and g(w) − p(u) < 0.
Since g(w)− p(u) < 0, it does not enter S immediately. Thus, the trajectory of solutions does not exit W immediately and it is included in J12.
(3) Assume v = 0 and g(w)− p(u) > 0.
(5.63) implies v+> 0. Thus, the trajectory of solutions will enter the region S.
(3) Assume v = 0 and g(w)− p(u) = 0.
We have [g(w)− p(u)]0= 0 and [g(w)− p(u)]00= g0(w)z0. By (5.61), [g(w)− p(u)]+> 0. On the other hand, dv0= dv00= 0 but dv000= h(u)g0(w)z0> 0. That implies v+> 0 and the trajectory of solutions will enter the region S.
On region Q13+, we have g(w)− p(u) < 0 and consider the following three subsets.
(1) Assume v > 0.
Since v > 0, we know that u+> u∗. (5.62) implies z+< 0. the trajectory of solutions will enter the region Q.
(2) Assume v = 0.
(5.63) implies v+< 0 and u+< u∗. Thus, the trajectory of solutions will enter the region R.
(3) Assume v < 0.
Since v < 0, we know that u+< u∗. Thus, the trajectory of solutions will enter the region R.
On region Q13−, we have g(w)− p(u) < 0 and consider the following three subsets.
(1) Assume v > 0.
Since g(w)− p(u) < 0, we know it does not enter S immediately. Since v > 0, we know that u+> u∗and it does not enter P or R immediately. (5.62) implies z+> 0 and it does not enter Q immediately. Thus, the trajectory of solutions does not exit W immediately and it is included in J11.
(2) Assume v = 0.
(5.63) implies v+< 0 and u+< u∗. Thus, the trajectory of solutions will enter the region R.
(3) Assume v < 0.
Since v < 0, we know that u+< u∗. Thus, the trajectory of solutions will enter the region R.