Asymptotic properties of the spectrum of neutral delay differential equations

Asymptotic properties of the spectrum of neutral delay differential

equations

Y.N. Kyrychko^ K.B. Blyuss

Department of Mathematics, University of Sussex,
Brighton, BN1 9QH, United Kingdom

P. Hovel and E. Scholl

Institut firr Theoretische Physik, Technische Universitat Berlin,
10623 Berlin, Germany

January 31, 2012
Abstract

Spectral properties and transition to instability in neutral delay differential equations are
investigated in the limit of large delay. An approximation of the upper boundary of stability
is found and compared to an analytically derived exact stability boundary. The approximate
and exact stability borders agree quite well for the large time delay, and the inclusion of a
time-delayed velocity feedback improves this agreement for small delays. Theoretical results
are complemented by a numerically computed spectrum of the corresponding characteristic
equations.

1 Introduction

An important area of research in physics and engineering is control theory, and a recent monograph
by Scholl and Schuster jjj gives a good overview of the developments in this field. From a dynamical
systems perspective, one could consider stabilisation of unstable fixed points or unstable periodic
orbits [2, 3, 4J. In fact, even when these unstable periodic orbits are embedded in a chaotic attractor,
they can still be stabilised by weak external forces, as it has been first proposed in a seminal paper
by Ott, Grebogi, and Yorke [5]. Since then, several other methods of controlling unstable motion
have been proposed. One of them is a time-delayed feedback control proposed by Pygaras in [6],
which can be easily implemented in a wide range of experiments and is non-invasive, i.e., it vanishes
as soon as unstable motion becomes stable [IlElElQiiiniQaQSlCg This method

utilizes a difference between a signal at the current time and the same signal at some time ago. The
scheme can be improved by introducing multiple time delays into the control loop |19j . Further
considerations of multiple delay control, also referred to as extended time-delay autosynchronisation
can be found in [U ED 1221 E3 EU •

From theoretical point of view, introduction of a time delay into the system leads to an infinite-
dimensional phase space and transcendental characteristic equations [251126] . This adds a significant

'Corresponding author. Email: y.kyrychko@sussex.ac.uk

1

difficulty to the stability and bifurcation analyses of such systems. Some analytical results on time-
delayed feedback control can be found, for instance, in [27J ESI GHl [3Ql GU G2]- In the case of linear
time-delayed systems with non-delayed highest derivative, one can use the Lambert function to
find the solutions of the corresponding characteristic equation [6'6\ 134] . However, for the neutral
equations, i.e. equations which have time delays in the highest derivative term, this approach
fails, and other approaches should be used. Furthermore, since neutral delay differential equations
(NDDEs) often possess discontinuities in their solution, the numerical treatment and bifurcation
analyses of such equations are much more involved than those of regular delay differential equations
(DDEs). For instance, the existing packages for bifurcation analysis of DDEs, such as DDE-
BIFTOOL [35] and PDDE-CONT [36] currently are unable to perform continuation for neutral
systems.

This paper is devoted to the analytical and numerical analysis of a time-delayed system of
neutral type. Such models arise in a variety of contexts, such as biological and population dynamics
models, see, for example, [37 1 138 1 [39], Balanov et al. [40] , for instance, derived a neutral DDE as a
model for torsional waves on a driven drill-string. Another example studied by Blakely & Corron
|41j is a model of a chaotic transmission line oscillator, in which an NDDE was used to correctly
reproduce experimental observations of fast chaotic dynamics. The system to be studied in this
paper was first introduced by Kyrychko et al. |42j in the context of hybrid testing, where it proved
to be a good physical model for the description of the effects of actuator delays. It is noteworthy
that in hybrid testing, quite often one encounters significant time delays due to actuator response
time. Furthermore, actuator delay strongly depends on the stiffness of the system, hence it can
vary considerably even in different runs of the same experiment. For this reason, the actual values
attained by the actuator time delay can be quite high |441 1431 146] .

It is important to note that NDDEs are different from DDEs in that they may possess a
continuous as well as a point spectrum and their stability properties are far from being completely
understood. Here, we investigate two different kinds of time-delayed feedback, both of which arise
naturally in experimental settings. The first of these includes time delay in the feedback force,
while the second introduces a velocity feedback. To understand the stability properties of the
system, we will analyse asymptotic behaviour of the eigenvalue spectrum and identify regions of
(in)stability in terms of system's parameters. The validity of these results will be compared to the
numerical solution of the corresponding characteristic equation. For the particular system under
investigation it is possible to find the stability spectrum analytically, and therefore it serves as a
perfect test model for which it is possible to compare exact and approximate stability boundaries.
It will be shown that although the approximation may deviate quite significantly from the exact
boundary for small time delays, it gives a good agreement for larger time delays. Therefore, this
approximation can be used for systems described by neutral DDEs with large time delays, where
it is impossible to find the stability boundary analytically.

2 Stability analysis

2.1 Delayed force

Consider the following NDDE [12]:

z(t) + 2Qz{t) + z(t) + pz(t - r) = 0, (1)

where dot means differentiation with respect to time t, and r is the time delay. In the context
of hybrid testing experiments on a pendulum-mass-spring-damper system, £ stands for a rescaled

2

damping rate, andp is the mass ratio. Introducing v(t) = z(t) and u(t) = v(t)+pv(t), this equation
can be rewritten as a system of differential equations with a shift:

z{t) = u(t) — pv(t — r),

u(t) = -2C[u(t)-pv(t-r)]-z(t), (2)
v(t) = u(t) — pv(t — r).

With the initial data (z(0), «(0)) = (zo,^o) € M x M and v(s) = 4>(s) G C[— r, 0], this system can
be first solved on < t < r interval, then on r < t < 2r and so on, provided the following sewing
condition is satisfied: 0(0) = uq — p<j){— r). This condition ensures that there are no discontinuities
in the solutions at t = kr, k E Z + . For arbitrary initial conditions the sewing condition does not
hold, and leads to jumps in the derivative of the solution [43J.

The equation ([I]) has a single steady state z* = 0. The stability of this steady state is determined
by the real part of the complex roots A S C of the corresponding characteristic equation

A 2 + 2CA + 1 + pA V Ar = 0. (3)

As it was already mentioned in the introduction, the existing bifurcation packages for the analysis
of delay equations, such as DDE-BIFTOOL |35| and PDDE-CONT [36], currently do not provide
capabilities of calculating eigenvalues for NDDEs. One of the reasons for this lies in the so-called
behavioral discontinuity, a feature unique to NDDEs as compared to standard DDEs. This refers to
the fact that even when all characteristic roots are stable for r = 0, for r being small and positive
infinitely many of these roots may have unbounded real parts. In other words, a small variation of
the time delay leads to an infinitely large root variation [UJ SB] •

Several methods based on linear multi-step approach and pseudospectral differentiation have
been recently put forward which provide an efficient tool for computing the characteristic spectrum
of NDDEs [HI [501 EJJ • We have used this method to compute the spectrum of Eq. ^ , which
is shown in Fig. [lj It can be observed that for small time delays (Fig. [if a)) the steady state is
stable, as all the eigenvalues are in the left half-plane. As time delay increases, a pair of complex
conjugate eigenvalues crosses the imaginary axis, as demonstrated in Fig. [TJb) ) and (c), leading
to an instability. As time delay increases further still, the unstable eigenvalues return to the left
half-plane, thus restoring the stability.

In the case when the mass ratio p in Eq. exceeds unity, the steady state is unstable for any
positive time delay r. It is worth noting that for \p\ < 1, the steady state may undergo stability
changes/switches as the time delay is varied. To understand the dynamics of the system in the
neighborhood of these stability changes, one can use the framework of pseudocontinuous spectrum
used by Yanchuk et al. \31\ [52] for the analysis of scaling behavior of eigenvalues for large time
delays, who followed an earlier work of Lepri et al. [53J on scaling of the spectra. Following this
approach, one can express the asymptotic approximation of the eigenvalues for large r as

x = ^ + i (n + -cp)+o(\), (4)

T \ T ' ( "~ 2

where 7, f2, and <f> are real-valued quantities, which are associated with the real and imaginary part
of the eigenvalue A, respectively. Substituting this representation into the characteristic equation
Q, gives to the leading order in C(l/r):

1 - ft 2 + 2i(tt - P n 2 e-^e-^e- inT = 0. (5)

3

E

10
5

-5

4

• •

1 1
(a)-

% ^ *

, i 1 1 , i ,

, 1 ,

E

1

1 1 1

(b)-

, 1 f 1 , 1 ,

, 1 ,

■2 -1.5 -1 -0.5 0.5 1 -2 -1.5 -1 -0.5 0.5 1

Figure 1: (Color online) Spectrum of the characteristic equation ^ for different time delays: (a)
r = 2.5, (b) r = 3.32, (c) r = 5, and (d) r = 7. Parameter values are: £ = 0.1, p = 0.3. The solid
lines show the asymptotic pseudocontinuous spectrum given by Eq. ([7]).

±1, ±2, ±3,.... in equation ([5j), we can simplify this equation

0.

By choosing Q = = 2im/T, n
to

i - n 2 + 2i(n - P n 2 e -^e~^ = o. (6)

From (|4j) it follows that Re(A) ~ r f(Q)/r and Im(A) ~ £1 up to the leading order, and therefore the
eigenvalues A accumulate in the complex plane along curves (7(0), O), with the real axis scaling as
rRe(A). Solving equation Q gives an expression for the real part 7 of the eigenvalue as a function
of the Hopf frequency

7 (0)

- In — ^
2 p 2

4<C 2 - 2 1

1 + — 1

n 2 o 4

(7)

A steady state can lose its stability via a Hopf bifurcation, at which point the tip of curve 7(0)
will cross the imaginary axis. If this happens, there will be an interval of frequencies Oi < $7 < O2,
for which 7(0) > and 7(^1) = 7(^2) = 0. This instability can be prevented, provided the interval
of unstable frequencies Oi < f2 < O2 lies inside the interval [f2 n °, O n ° +1 ] for some no [31] ■ Here,
fii 5 2 are two positive roots of the equation 7(0) = 0, which can be found from Eq. (FtT) as

O 2
"1,2

1

1 - 2C 2 ± v / (l-2C 2 ) 2 -l+P 2

1 -p 2 L

For further analytical progress, we expand this expression for small values of £, which gives

(8)

4

Figure 2: (Color online) Comparison of the approximate upper bound of stability according to
Eq. (10) (dotted line) with an exact stability boundary for £ = 0.1 in the (r,p) plane. The

grayscale (color code) encodes the value of the largest real part of the complex eigenvalues A.

An = Oi - n 2

P

C 2 / i

+

p

(9)

Since the actual values of the frequencies are 0( n ) = 27rn/r for any integer n, the distance between
any two successive frequencies is 2tt/t, and hence the necessary condition for stability Af2 < 2tt/t
can be written as

1 1 CV 1 1 \

Tt^p ~ VT+p ~ J \VT+p + V^p) < 27r/r ' (10)

For large enough time delay r, p asymptotically approaches a lower bound of stability which
corresponds to Ail = 0. It can be obtained from Eq. (Tsl) by using

(^1-^2)^1 + ^2)

n 2 2

1 + P 2

which yields

P

2Ca/1 - C 2 ~ 2C-

(11)

Figure 2 shows the plot of the approximate stability boundary (10) as a function of time delay

r for a given small value of the damping £. The grayscale (color code) in this figure indicates the
value of the largest real part of the eigenvalues in the spectrum of the characteristic equation Q
for each value of p and r. As it follows from Fig. [2j the analytically derived formula (10) for the
maxima on the stability boundary deviates from the exact stability peaks (which correspond to

5

E

10
5

E

■10
10

5

-5

■10

1 1 \ 1 1 1 1

1 i 1
(a)-

, i T i , i ,

, 1 ,

V

1 i 1

(c)-

r

, 1 ,

1 i

1 1 1 1

1 i 1

(b)

1,1,

, 1 ,

: 1

1

(d)

r

, 1 ,

2 -1.5 -1 -0.5 0.5 1 -2 -1.5 -1 -0.5 0.5 1

Figure 3: (Color online) Spectrum of the characteristic equation (13) or different time delays: (a

t = 2, (b) r = 2.725, (c) r = 3.5, and (d) r = 5. Parameter values are: & = 0.25, (2 = 0.24,

p = 0.3. The solid lines show the asymptotic pseudocontinuous spectrum given by Eq. (15)

codimension-two Hopf bifurcation) for small delays, but it provides a good approximation for large
time delay r. Appendix A contains an exact analytic expression for the stability boundary in terms
of system's parameters.

2.2 Delayed viscous damping

In order to analyse the influence of velocity feedback on the stability of NDDE, we modify Eq. ([!])

as

z(t) + 2&z(i) + z{t) + pz(t - r) + 2( 2 z(t - r) = 0. (12)

This equation was introduced in Ref. [42] . where it was shown that depending on the difference
between two damping parameters £i and (2, the stability domain may shrink and even split into
separate stability regions in the parameter plane (the so-called death islands). The characteristic
equation now modifies to

A 2 + 2CiA + 1 + p\ 2 e- XT + 2C 2 Ae" Ar = 0. (13)

Figure [3] shows the numerical approximation of the roots of this equation in the neighborhood of
the origin. From this figure it follows that similar to the situation without velocity feedback, the
system undergoes successive stability switches as the time delay is varied.

6

-0.1
-0.2
-0.3
-0.4
-0.5

Figure 4: (Color online) Comparison of the approximate upper bound of stability according to

Eq. ( 17) (dotted line) with an exact stability boundary for (j = 0.25 and (,2 = 0.24 in the (r,p) plane.
The grayscale (color code) encodes the value of the largest real part of the complex eigenvalues A.

Assuming in Eq. ( 13 ) the same asymptotic behavior Q of the eigenvalues for large time delay
(i.e., the real part of the eigenvalue scales as 1/r), gives to the leading order

0.

(14)

with the constraint VL = = 2im/T, n = ±1, ±2, ±3,.... One can solve this equation for the
real part of the eigenvalue 7 at the Hopf bifurcation as a function of frequency O, as

7 (fi)

1 (l-ny + ^fo 2

2 11 p^ + 4C 2 2 ^ 2 '

(15)

Transition to instability occurs when 7(0) = 0, which gives the expression for instability fre-
quencies

O 2

1 + 2(1 - 2(f ±

l + 2C 2 -2C 2 ) 2 -l+p

1 -p 2

In a manner similar to the analysis of the delayed force feedback, one can make further analytical
progress by assuming that both damping coefficients are small: \(i\ <C 1, IC2I *C 1. The necessary
stability condition Af2 = f^i — f2 2 < 2-k/t gives the following asymptotic approximation for the
maxima of the stability boundary:

1

1

Ci 2 - C 2 2

a/i — p V 1 + P

1

+

1

a/1 +P y/1 ~P

< 2vr/r.

(17)

7

The expression (17) can be further simplified for large time delay in a manner similar to ( 11 ), which
2vW

C|)(i-C 1 2 + C|)-2 v ^f

t 2

gives p

It is noteworthy that the inequality (17) provides a good approximation for the stability bound-
ary even when actual values of damping coefficients (j and £2 are large, as long as the difference
(£ 2 — £f ) is small by the absolute value. Figure [4] shows an excellent agreement between the
asymptotic approximation (17) and the exact stability boundary, especially for sufficiently large
time delay. In Appendix A it is shown how the exact stability boundary varies depending on
parameters, and in particular on the relation between the two damping coefficients.

3 Conclusions

Time delays are an intrinsic feature of many physical, biological and engineering systems, and
in recent years the analysis of such systems has led to many interesting and important findings.
There are systems where the time delay is present intrinsically due to processing times, mechanical
inertia etc., and there are those where the time delay is introduced externally in order to stabilise
unstable periodic orbits and steady states. Therefore, a better understanding of delay differential
equations will provide a clear picture of the system's stability and controllability. In this paper we
have concentrated on the analysis of two neutral delay differential equations. We have shown that
depending on the time delay r, the systems exhibit stability switches, where stability is lost /regained
depending on the time delay. In the case of delayed velocity feedback, the interplay between the
time delay and the two damping coefficients gives different stability regimes in the parameter plane,
and for some parameter values the stability area collapses into separate islands. We have derived
an asymptotic approximation of the stability peaks for large time delays, based upon universal
scaling arguments, and have compared this approximation with the exact stability boundary. The
results agree quite well even when the time delay is not too large, and give excellent agreement for
large delays. The results presented in this paper include numerical simulations of the characteristic
spectrum and constitute the first attempt to approximate stability peaks for neutral DDEs. As it
has already been mentioned in the introduction, neutral DDEs arise naturally in a wide range of
physical problems, which makes the approach developed in this paper a useful tool for the stability
analysis of such systems.

Acknowledgements

Y.K. was supported by an EPSRC Postdoctoral Fellowship (Grant EP/E045073/1). The authors
would like to thank Dimitri Breda for providing assistance with numerical computation of the roots
of characteristic equations. This work was partially supported by Deutsche Forschungsgemeinschaft
in the framework of Sfb 555.

A Exact stability boundary

To find an exact analytical expression for the stability boundary, one has to substitute A = icu into
the characteristic equation ([3]). After separating real and imaginary parts, this gives [53]

1 — uj 2 — puj 2 cos(wr) = 0, , ,

2(u + puJ 2 sin(ojT) = 0. ^ '

Squaring and adding these equations gives

(l-p 2 )u; 4 + (4C 2 -2)w 2 + l = 0, (19)

8

Figure 5: Exact stability boundary of the characteristic equation Q in (r,p) parameter plane for
£ = 0.1. The steady state is stable in the shaded area.

This equation can be solved as

W?2

l-p 2

l-2( 2 ± VU-2C 2 ) -1+P 2

(20)

In fact, Eq. (19) provides an expression for stability boundary value of p as parametrized by

the Hopf frequency ui:

= -^y/u* + 2u>* (2C 2 -1) + 1.

The corresponding value of the time delay at the stability boundary is derived from Eq. (|18|)

1

2 C^

Arctan — r ± 7rA;

1

(21)

(22)

where k = 0, 1, 2, ... and Arctan denotes the principal value of arctan. Figure [5] illustrates the
dependence of critical mass ratio p on the time delay r which ensures the stability of the steady
state. It is noteworthy that if \p\ > 1, the steady state is unstable for any positive time delay r;
on the other hand, if \p\ < 1 and £ > l/y/2, then the steady state is asymptotically stable for any

9

Figure 6: Exact stability boundary of the characteristic equation (13) in (r,p) parameter plane.
The steady state is stable in the shaded area. Parameter values are: (a) £i = 0.25 and £2 = 0.24,
(b) Ci = 0.23 and ( 2 = 0.25.

positive time delay r. For ( < l/y/2, there is a lower bound on the value of p m i n = 2£y / l — ( 2 , so
that an asymptotic stability is guaranteed for all r > provided p < p m i n [42J.

In the case of time-delayed viscous damping, the characteristic equation ( 13 ) at the points of
stability changes can be written as

1 — lo 2 — pco 2 cos lot + 2&lo sin lot = 0,
2(i + pcu sin lot + 2^2 cos lot = 0.

(23)

Squaring and adding these two equations gives the following parametrization of p by the Hopf
frequency:

"1,2

1

l-p 2

1 - 2d 2 + 2C 2 2 ± V (1 " 2 Ci 2 + 2C 2 2 ) 2 " 1 +P 2

(24)

Similar to the previous case, one can derive parametric expressions for p(uS) and t(lo) from Eq. (23):

p(co)

LO'

l) 2 + 4( Cl 2 -Cl

The corresponding value of the time delay at the stability boundary can be found as

t(lo)

1

LO

2im — arccos

P (l - LO 2 ) - 4ClC 2

P 2 lo 2 + 4C|

n = 1,2,3.

(25)

(26)

Figure [6] demonstrates how stability boundary is affected by the relation between £1 and £2- In
particular, we note that when (j = £2 the stability boundary touches the r-axis (p = 0), and for
C2 > Ci) the stability area consists of non-overlaping death islands, inside which the oscillations are
damped, and the steady state is stable.

10

References

[1] E. Sch611 and H. G. Schuster (eels.), Handbook of Chaos Control, Wiley- VCH, Weinheim, 2007.

[2] L. M. Pecora and T. L. Carroll, Driving systems with chaotic signals, Phys. Rev. A 44 (1991),
pp. 2374-2383.

[3] E. R. Hunt, Stabilizing high-period orbits in a chaotic system: the diode resonator, Phys. Rev.
Lett. 67 (1991), pp.1953-1955.

[4] W. L. Ditto, S. N. Rauseo, and M. L. Spano, Experimental control of chaos, Phys. Rev. Lett.
65 (1990), pp. 3211-3214.

[5] E. Ott, C. Grebogi, and J. A. Yorke, Controlling chaos, Phys. Rev. Lett. 64 (1990), pp. 1196-
1199.

[6] K. Pyragas, Continuous control of chaos by self-controlling feedback, Phys. Lett. A 170 (1992),
pp. 421-428.

[7] K. Pyragas and A. Tamasevicius, Experimental control of chaos by delayed self- controlling
feedback, Phys. Lett. A 180 (1993), pp. 99-102.

[8] S. Bielawski, D. Derozier, and P. Glorieux, Controlling unstable periodic orbits by a delayed
continuous feedback, Phys. Rev. E 49 (1994), pp. R971-R974.

[9] T. Pierre, G. Bonhomme, and A. Atipo, Controlling the chaotic regime of nonlinear ionization
waves using the time-delay autosynchronization method, Phys. Rev. Lett. 76 (1996), pp. 2290-
2293.

[10] K. Hall, D. J. Christini, M. Tremblay, J. J. Collins, L. Glass, and J. Billette, Dynamic control
of cardiac alternans, Phys. Rev. Lett. 78 (1997), pp. 4518-4521.

[11] D. W. Sukow, M. E. Bleich, D. J. Gauthier, and J. E. S. Socolar, Controlling chaos in a fast
diode resonator using extended time-delay autosynchronization: Experimental observations and
theoretical analysis, Chaos 7 (1997), pp. 560-576.

[12] O. Liithje, S. Wolff, and G. Pfister, Control of chaotic Taylor- Couette flow with time-delayed
feedback, Phys. Rev. Lett. 86 (2001), 1745-1748.

[13] P. Parmananda, R. Madrigal, M. Rivera, L. Nyikos, I. Z. Kiss, and V. Gaspar, Stabilization of
unstable steady states and periodic orbits in an electrochemical system using delayed- feedback
control, Phys. Rev. E 59 (1999), pp. 5266-5271.

[14] J. M. Krodkiewski and J. S. Faragher, Stabilization of motion of helicopter rotor blades using
delayed feedback - modelling, computer simulation and experimental verification, J. Sound Vib.
234 (2000), pp. 591-610.

[15] M. G. Rosenblum and A. Pikovsky, Controlling synchronization in an ensemble of globally
coupled oscillators, Phys. Rev. Lett. 92 (2004), pp. 114102.

[16] O. V. Popovych, C. Hauptmann, and P. A. Tass, Effective desynchronization by nonlinear
delayed feedback, Phys. Rev. Lett. 94 (2005), pp. 164102.

11

[17] S. Schikora, P. Hovel, H. J. Wiinsche, E. Scholl, and F. Henneberger, All-optical noninvasive
control of unstable steady states in a semiconductor laser, Phys. Rev. Lett. 97 (2006), pp.
213902.

[18] T. Erneux and T. Kalmar-Nagy, Nonlinear stability of a delayed feedback controlled container
crane, J. Vibr. Contr. 13 (2007), pp. 603-616.

[19] J. E. S. Socolar, D. W. Sukow, and D. J. Gauthier, Stabilizing unstable periodic orbits in fast
dynamical systems, Phys. Rev. E 50 (1994), pp. 3245-3248.

[20] K. Pyragas, Control of chaos via an unstable delayed feedback controller, Phys. Rev. Lett. 86
(2001), pp. 2265-2268.

[21] N. Baba, A. Amann, E. Scholl, and W. Just, Giant improvement of time-delayed feedback
control by spatio-temporal filtering, Phys. Rev. Lett. 89 (2002), pp. 074101.

[22] O. Beck, A. Amann, E. Scholl, J. E. S. Socolar, and W. Just, Comparison of time-delayed
feedback schemes for spatiotemporal control of chaos in a reaction- diffusion system with global
coupling, Phys. Rev. E 66 (2002), pp. 016213.

[23] J. Unkelbach, A. Amann, W. Just, and E. Scholl, Time-delay auto synchronization of the
spatiotemporal dynamics in resonant tunneling diodes, Phys. Rev. E 68 (2003), pp. 026204.

[24] T. Dahms, P. Hovel, and E. Scholl, Control of unstable steady states by extended time-delayed
feedback, Phys. Rev. E 76 (2007), pp. 056201.

[25] J. K. Hale, Theory of functional differential qquations, Springer, New York, 1977.

[26] J. K. Hale and S.M. Verduyn Lunel, Introduction to functional differential equations, Springer,
New York, 1993.

[27] K. Pyragas, Analytical properties and optimization of time-delayed feedback control,
Phys. Rev. E 66 (2002), pp. 26207.

[28] W. Just, S. Popovich, A. Amann, N. Baba, and E. Scholl, Improvement of time- delayed feedback
control by periodic modulation: analytical theory of Floquet mode control scheme, Phys. Rev. E
67 (2003), pp. 026222.

[29] P. Hovel and J. E. S. Socolar, Stability domains for time-delay feedback control with latency,
Phys. Rev. E 68 (2003), pp. 036206.

[30] P. Hovel and E. Scholl, Control of unstable steady states by time-delayed feedback methods,
Phys. Rev. E 72 (2005), pp. 046203.

[31] S. Yanchuk, M. Wolfrum, P. Hovel, and E. Scholl, Control of unstable steady states by long
delay feedback, Phys. Rev. E 74 (2006), pp. 026201.

[32] B. Fiedler, V. Flunkert, M. Georgi, P. Hovel, and E. Scholl, Refuting the odd-number limitation
of time- delayed feedback control, Phys. Rev. Lett. 98 (2007), pp. 114101.

[33] R. M. Corless, G. H. Gonnet, D. E. G. Hare, D. J. Jeffrey, and D. E. Knuth, On the Lambert
W function Adv. Comput. Math. 5 (1996), pp. 329-359.

12

A. Amann, E. Scholl, and W. Just, Some basic remarks on eigenmode expansions of time-delay
dynamics Physica A 373 (2007), pp. 191-202.

K. Engelborghs, T. Luzyanina, and G. Samaey, Technical Report No. TW-330, Department
of Computer Science, K.U.Leuven, Belgium (2001).

R. Szalai, PDDE-CONT: A continuation and bifurcation software for delay- differential equa-
tions. \http://seis. bris.ac.uk/~rs l909/pdde, 2005.

G. Bocharov and K. P. Hadeler, Structured population models, conservation laws, and delay
equations, J. Diff. Eqns. 168 (2000), pp. 212-237.

K. P. Hadeler and G. Bocharov, Where to put delays in population models, in particular in the
neutral case, Can. Appl. Math. Quart. 11 (2003), pp. 159-173.

K. P. Hadeler, Neutral delay equations from and for population genetics, Electron. J. Qual.
Theory Differ. Equ. 11 (2008) 1-18.

A. G. Balanov, N. B. Janson, P. V. E. McClintock, R. W. Tucker, and C. H. T. Wang,
Bifurcation analysis of a neutral delay differential equation modelling the torsional motion of
a driven drill-string, Chaos, Solitons & Fractals 15 (2003), pp. 381-394.

J. N. Blakely and N. J. Corron, Experimental observation of delay-induced radio frequency
chaos in a transmission line oscillator, Chaos 14 (2004), pp. 1035-1041.

Y. N. Kyrychko, K. B. Blyuss, A. Gonzalez-Buelga, S. J. Hogan, and D. J. Wagg, Real-time
dynamic sub structuring in a coupled oscillatorpendulum system, Proc. R. Soc. A 462 (2006),
pp. 1271-1294.

V. Kolmanovskii and A. Myshkis, Applied theory of functional differential equations, Dor-
drecht, Kluwer, 1992.

M. Nakashima and N. Masaoka, Real-time on-line test for MDOF systems, Earthquake Engng.
Struct. Dyn. 28 (1999), 393-420.

A. P. Darby, M. S. Williams, and A. Blakeborough, Stability and delay compensation for
real-time substructure testing, J. Eng. Mech. 128 (2002), pp. 1276-1284.

J. Zhao, C. French, C. Shield, and T. Posbergh, Considerations for the development of real-time
dynamic testing using servo-hydraulic actuation, Earthquake Engng. Struct. Dyn. 32 (2003),
pp. 1773-1794.

N. Olgac and R. Sipahi, The cluster treatment of characteristic roots and the neutral type
time-delayed systems, ASME J. Dyn. Syst. 127 (2005), pp. 88-97.

D. A. W. Barton, B. Krauskopf, and R. E. Wilson, Homoclinic bifurcations in a neutral delay
model of a transmission line oscillator, Nonlinearity 20 (2007), pp. 809-829.

K. Engelborghs and D. Roose, On stability of LMS methods and characteristic roots of delay
differential equations, SIAM J. Num. Anal. 40 (2002), pp. 629-650.

D. Breda, Solution operator approximations for characteristic roots of delay differential equa-
tions, Appl. Num. Math. 56 (2006), pp. 305-317.

13

[51] D. Breda, S. Maset, and R. Vermiglioa, Pseudo spectral approximation of eigenvalues of deriva-
tive operators with non-local boundary conditions, Appl. Num. Math. 56 (2006), pp. 318-331.

[52] M. Wolfrum and S. Yanchuk, Eckhaus instability in systems with large delay, Phys. Rev. Lett.
96 (2006), pp. 220201.

[53] S. Lepri, G. Giacomelli, A. Politi, and F. T. Arecchi, High- dimensional chaos in delayed
dynamical systems, Physica D 70 (1993), pp. 235-249.

[54] Y. Kuang, Delay differential equations with applications in population dynamics, San Diego,
Academic Press, 1993.

14