that cannot exist on the phase line R.

Many interesting features of single neuron dy-

namics can be illustrated or explained using two-dimensional systems. Even neuronal
bursting, which occurs in multi-dimensional systems, can be understood via bifurcation
2D  two-dimensional
models   and  lsystems.
imit  using
cycles:  
In this chapter we introduce methods of phase plane analysis of two-dimensional sys-
analysis
tems. Mostofconcepts will be illustrated the INa,p +IK -model in Fig. 4.1:
This model is equivalent in many respects to the well-known and widely used
ICa +IK -model proposedleak Iby Morris
z }| L { z
and Lecar
instantaneous
}| { z }| { Remember:  
INa,p(1981) toIKdescribe voltage oscillations INa,p  inward  (depolarizes  V)  
in the Cbarnacle
V̇ = I giant
° gL (Vmuscle
°EL ) °fiber.
gNa m1 (V ) (V °ENa ) ° gK n (V °EK ) ,  (4.1)    IK  outward  (hyperpolarizes  V)  
ṅ = (n1 (V ) ° n)/ø (V ) , (4.2)

4.1 Planar Vector Fields

having leak current IL , persistent Na+ current INa,p with instantaneous activation ki-
netic and a relatively slower persistent K+ current IK with either high (Fig. 4.1a) or
low (Fig.Planar  
Two-dimensional vector  
4.1b) threshold fields  
dynamical
(the achoices
nd  nullclines  
twosystems, also incalled
result planar systems,
fundamentally are often written in
different dynamics).
formof the INa,p +IK -model is a two-dimensional vector (V, n) 2 R on the phase
the state
The 2

plane R2 . New types of equilibria, orbits, and bifurcations can exist on the phase plane
In  a  general  
that cannot exist on 2 D  sphase
the ystem  
line R. Many f (x, y) ,features ofnulclines  
ẋ =interesting single neuron are  dy-
the  curves  in  the  (x,y)  plane  defined  
namics can be illustrated or explained using two-dimensional systems.
ẏ = g(x, y) , by  fEven
(x,y)=0,  
neuronal  and  g(x,y)=0    
bursting, which occurs in multi-dimensional systems, can be understood via bifurcation
analysis of two-dimensional systems.
where the functions
wo-Dimensional
This modelSystems
f and g describe the evolution of the two-dimensional state variable
is equivalent in many respects to the well-known and widely97used
I(x(t), y(t)). For any point (x0 , y0 ) on the phase plane the vector (ff(x,y)=0  
Ca +IK -model proposed by Morris and Lecar (1981) to describe voltage oscillations
(x0 , y0 ), g(x !0  x, y-­‐nullcline  
0 ))
0.7
in the barnacle giant muscle fiber.  
0.6
93
(b) g(x,y)=0  !  y-­‐nullcline  
spike
4.1 (c)
Planar Vector Fields downstroke
e

absolute
n-nullclin

0.5 refractory
K+ activation variable, n

Two-dimensional dynamical systems, also called planar peak -­‐ The  

systems, are often written
of spike
in intersecEons  of  the  nullclines  define  
the form 0.4
the  equilibrium  points.  
ẋ = f (x, y) , -­‐ The  nullclines  define  regions  in  the  state  
0.3
ẏ = g(x, y) , variable  (x,y)  where  the  flux  is  purely  
V-n

(d) (a)
verEcal  or  horizontal.  
ullc

0.2 spike
li

where the functions f and g describe the evolutionupstroke

of the two-dimensional state variable
ne

relative
any point (x0 , y0 ) on the phase(regenerative)
(x(t), y(t)). Forrefractory plane the vector (f (x0 , y0 ), g(x0 , y0 ))
0.1

93
resting
0
-80 -70 -60 -50 -40 -30 -20 -10 0 10 20
membrane voltage, V (mV)
 Two-Dimensional Systems 101

Exercise:  Plot  the  nullclines  in  the  (V,n)  plane  (n  as  a  funcEon  of  V)  of  the  INa,p+IK  model,  and  draw  
two  trajectories  in  the  plane  of  variables  starEng  with  (V0,n0)=(-­‐60  mV,0.2)  and  (V0,n0)=  (-­‐30  mV,
0.2).  Draw  also  the  V-­‐Eme  plot  for  both  trajectories.  
Model  parameters:  I=0,  C=1,  EL=-­‐78  mV,  gL=8,  gNa=20,  gK=10,  ENa=  60  mV,  EK=-­‐90  mV.  
m∞  (V)  has  V1/2=  -­‐20,  k=15.  n∞  (V)  has  V1/2=  -­‐45,  k=5  and  τ(V)=1.  
stable unstable
Plot  the  soluEons  of  the  fast  INa,p+IK  model  for  the  same  iniEal  condiEons  as  before  but  with  
Figure 4.9: Limit cycles (periodic orbits).
injected  current  I=40.  What  do  you  think  that  has  changed?.  
4.1.3 Limit cycles
A trajectory that forms a closed loop is called a periodic trajectory or a periodic orbit
Limit  (the
cycles  
latter is usually reserved for mappings, which we do not consider here). Sometimes
periodic trajectories are isolated, as in Fig. 4.9, sometimes they are part of a continuum,
as in Fig. 4.13, left. An isolated periodic trajectory is called a limit cycle. The existence
Limit  cycles   are  cycles
of limit closed  
is a(major
periodic)  
feature aofnd   isolated  (no  
two-dimensional other  
systems thatccannot
losed  exist
orbits  
in R1i.n  the  vicinity)  
If the initial point is on a limit cycle, then the solution (x(t), y(t)) stays on the cycle
soluEons  oforever,
f  the  and
system   variables  
the system exhibits periodic behavior; i.e.,
x(t) = x(t + T ) and y(t) = y(t + T ) (for all t)
Two-Dimensional Systems 101
for some T > 0. The minimal T for which this equality holds is called the period
of the limit cycle. A limit cycle is said to be asymptotically stable if any trajectory
with the initial point sufficiently near the cycle approaches the cycle as t ! 1. Such
asymptotically stable limit cycles are often called limit cycle attractors, since they
“attract” all nearby trajectories. The stable limit cycle in Fig. 4.9 is an attractor.
The limit cycle in Fig. 4.10 is also an attractor; It corresponds to the periodic (tonic)
spiking of the INa,p +IK -model (4.1, 4.2). The unstable limit cycle in Fig. 4.9 is often
called a repeller, since it repels all nearby trajectories. Notice that there is always at
least one equilibrium inside any limit cycle on a plane.
In Fig. 4.11 we depict limit cycles of three types of neurons recorded in vitro.
Since we do not know the state of the internal variables, unstable
stable such as the magnitude of the
activation and inactivation of Na+ and K+ currents, we plot the cycles on the (V, V 0 )-
plane, where V 0 is theFigure 4.9: Limitofcycles
time derivative (periodic
V . The orbits).
cycles look jerky because of the poor
 Necessary  condiEons  for  limit  cycles:      1.-­‐  At  least  2  variables.  
                                                                                                                                         2.-­‐  Non-­‐linear  system.  

IntuiEvely,  there  will  be  a  limit  cycle  if:    

1.-­‐  We  have  an  unstable  equilibrium  point  (flux  outwards).  
2.-­‐  We  have  a  closed  region  around  the  unstable  point  where  the  flux  points  inwards.  

Poincaré-­‐Bendixson  theorem    
 102 Two-Dimensional Systems

The  INa,p+IK  model  with  repeEEve  firing  (a  stable  limit  cycle).  Note  that  along  the  axis  limits  the  
flux  points  inwards  (if  value  of  n  is  extended  above  0.7).    
0.7

0.6

ne
n-nullcli
K+ activation variable, n 0.5

0.4

0.3

V-n
ullcl
0.2

ine
0.1

0
-80 -70 -60 -50 -40 -30 -20 -10 0 10 20
membrane voltage, V (mV)
0
membrane voltage, V (mV)

-20

-40

-60

-80
0 5 10 15 20 25 30
time (ms)

Figure 4.10: Stable limit cycle in the INa,p +IK -model (4.1, 4.2) with low-threshold K+
current and I = 40.
 Stability  of  equilibrium  points  in  a  system  of  2  variables    
u = ueq + Δu

u! = f (u , v; µ ) v = veq + Δv ∂f eq ∂f eq
Δu! = Δu + Δv
v! = g (u , v; µ ) Linearize  around   ∂u ∂v
equilibrium  point   ∂g eq ∂g eq
Δv! = Δu + Δv
∂u ∂v

" %
$ ∂f eq ∂f eq '
" Δu! % $ ∂u ∂v ' " Δu %
Δx! = J ⋅ Δx $ '=$
# Δv! & $ ∂g eq ∂g eq ' ⋅ $ Δv '
$ '' # &
# ∂u ∂v &

Jacobian  or  stability  matrix.    

 Find  solu;ons  of  a  linear  system:  a  problem  of  eigenvalues.  

Find  direcEons  of  exponenEal  approximaEon/divergence  from  the  equilibrium  point  

(transform  to  variables  called  eigenvectors  that  diagonalize  the  Jacobian)    

" Δy! % " λ 0 % " Δy1 %

Δy! = λΔy $ 1
'=$ 1
$ Δy!2 ' $ 0
'⋅$ '
λ2 '& $# Δy2 '&
# & #
λ1t
Δy1 (t) = Δy1 (0)e
λ2t
Δy2 (t) = Δy2 (0)e

Solve  the  characteris"c  equa"on  

to  find  the  eigenvalues  λ  
det(J − λI ) = 0
With  2  variables:  

τ + τ 2 − 4Δ
λ1 =
2
τ ≡ Tr(J)
τ − τ 2 − 4Δ
λ2 = Δ ≡ det(J)
2

Possible  cases  

1.-­‐Real  eigenvalues  
λ+ > 0
Saddle  
2 λ− < 0
τ − 4Δ > 0
λ1 < λ2 < 0 Stable  node  
λ1 ≠ λ2
λ1 > λ2 > 0 Unstable  node  
 2.-­‐Complex  eigenvalues:  

Re(λ1 ) = 0 Center  
τ 2 − 4Δ < 0
* Re(λ1 ) < 0 Stable  spiral  or  
λ1 = λ2 stable  focus  

Re(λ1 ) > 0 Unstable  spiral  or  

unstable  focus  

Use  Euler  formula     eiIm( λ )t = cos(Im(λ )t) + isin(Im(λ )t)

Special  case:  
τ 2 − 4Δ = 0 λ1 = λ2
Star  node  (stable  or  unstable)  
Bifurca;ons  in  2D.  
We  have  the  same  types  of  bifurcaEons  of  equilibria  that  appear  in  1D,  but  with  2D  we  
can  also  have  bifurca;ons  of  limit  cycles  (oscillaEons  can  be  turned  on  and  off).  

Hopf  bifurca;on:  A  stable  spiral  becomes  unstable.  

This  happens  when  two  complex  conjugate  eigenvalues  change  the  sign  of  their  real  part.  

Im  λ   Im  λ  

Re  λ   Re  λ  

det( J ) > 0
in  2  variables,  this  implies  that   for  some  values  of  the  parameters  
Tr (J ) = 0
 Summary  of  equilibria/bifurcaEons  for  2D  systems  as  a  funcEon  
of  the  trace  aSystems
Two-Dimensional nd  determinant  of  the  Jacobian   109

τ
(real positive eigenvalues)
=0

saddle-node bifurcation
eigenvalues
unstable node τ2 − 4∆

unstable focus
(complex eigenvalues,
positive real part)

0 Andronov-Hopf bifurcation ∆

saddle-node bifurcation
saddle
(real eigenvalues, different signs) stable focus
(complex eigenvalues,
negative real part)

stable node τ 2 − 4∆
(real negative eigenvalues) =0
0 210 5: Excitability

Figure 4.15: Classification of equilibria according to the trace (ø ) and the determinant
(∆) of the Jacobian matrix L. The shaded region corresponds to stable equilibria.

approach the saddle equilibrium along the eigenvector corresponding to the neg-
ative (stable) eigenvalue and then diverge from it along the eigenvector corre-
BifurcaEon   diagram  
sponding to the of  positive
the  H(unstable)
odgkin-­‐eigenvalue.
Huxley  Focus
model   as  a  funcEon  of  a  
(Fig. 4.18): The eigenvalues are complex-conjugate. Foci are stable when the
constant  aeigenvalues
pplied  current   by  areal
have negative n  parts, and unstable when the eigenvalues have pos-
external  electrode  
itive real parts. The imaginary part of the eigenvalues determines the frequency
of rotation of trajectories around the focus equilibrium.

When the system undergoes a saddle-node bifurcation, one of the eigenvalues becomes
m
zero and a mixed type of equilibrium occurs — saddle-node equilibrium, illustrated
in Fig. 4.14b. There could be other types of mixed equilibria, such as saddle-focus,
 0.6

e clin
n-null
0.5

K+ activation variable, n
Simulate  the  INa,p+IK  model  with  the  same  parameters  except  for  EL=-­‐80  mV,  V1/2=-­‐25  for  n∞(V),  
0.4

and  τ(V)=0.152,  with  an  applied  current    of  I=3  pA.  Use  as  iniEal  condiEons    (V0,n0)=(-­‐70  mV,  0)  and   0.3

V-n
(V0,n0)=(-­‐50  mV,0).  Plot  both  (V,t)  trajectories  and  the  system  nullclines  and  explain  what  you  see.  Now  

u
llcl
0.2

ine
trix
use  as  iniEal  condiEon  (V0,n0)=(-­‐10  mV,  0).  What  happened?  

ara
0.1

sep

trix
ra
pa
0

se
rest
attraction
domain

Bistability  between  rest  and  spiking  states   -80 -70 -60 -50 -40 -30
membrane voltage, V (mV)
-20 -10 0 10 20

-10

membrane voltage, V (mV)

A  stable  limit  cycle  (an  akractor)  can  coexist  with  a  stable  equilibrium  (another  akractor).  Now,  
both  akractors  are  separated  not  by  a  single  unstable  equilibrium  as  in  1D,  but  by  a  trajectory  
tangent  to  the  stable  eigenvector  of  a  saddle  point.  This  is  called  the  stable  manifold  or   rest spiking

Two-Dimensional Systems 115 -70

separatrix,  and  separates  domains  of  akracEon.   0 5

time, t (ms)
10 15

Figure 4.23: Bistability of rest and spiking states in the INa,p +IK -model (4.1, 4.2) with
0.6
high-threshold fast (ø (V ) = 0.152) K+ current and I = 3. A brief strong pulse of
e clin

current (arrow) brings the state vector of the system into the attraction domain of the
n-null

0.5 stable limit cycle.

K+ activation variable, n

0.4 B C A to spiking

membrane voltage
0.3 x)
tri A
ra
pa
V-n

e C
(s B
u

ld separatrix
llcl

0.2
v2

if o
ine

to rest

an
to spiking to saddle
trix

m
le
ara

ab
d
unstable manifol
0.1 unstab

st
l em
sep

trix

a n if o v1
ld
ra
pa

0 to rest
se

rest
attraction
domain time

-80 -70 -60 -50 -40 -30 -20 -10 0 10 20

membrane voltage, V (mV)
Figure 4.24: Stable and unstable manifolds to a saddle. The eigenvectors v1 and v2
-10
membrane voltage, V (mV)

correspond to positive and negative eigenvalues, respectively.

rest spiking
-70
0 5 10 15
time, t (ms)