See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259328615

Hybrid Control Strategy for Thermal Cycling of PCR Based on a Microcomputer

Article  in  Instrumentation Science & Technology · February 2007

DOI: 10.1080/10739140601000723

CITATIONS READS

3 71

2 authors:

Xianbo Qiu Jingqi Yuan

Beijing University of Chemical Technology Shanghai Jiao Tong University
65 PUBLICATIONS   1,153 CITATIONS    166 PUBLICATIONS   1,798 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Full scope simulation of coal-fired power plants View project

coal fired power plant, modelling, monitoring, operation optimization View project

All content following this page was uploaded by Jingqi Yuan on 19 August 2015.

The user has requested enhancement of the downloaded file.

This article was downloaded by: [Shanghai Jiaotong University]
On: 14 January 2009
Access details: Access Details: [subscription number 731868277]
Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Instrumentation Science & Technology

Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713597258

Hybrid Control Strategy for Thermal Cycling of PCR Based on a Microcomputer

Qiu Xianbo a; Yuan Jingqi ab
a
Department of Automation, Shanghai Jiao Tong University, Shanghai, P. R. China b State Key Laboratory of
Bioreactor Engineering, East China University of Science and Technology, Shanghai, P. R. China

Online Publication Date: 01 February 2007

To cite this Article Xianbo, Qiu and Jingqi, Yuan(2007)'Hybrid Control Strategy for Thermal Cycling of PCR Based on a
Microcomputer',Instrumentation Science & Technology,35:1,41 — 52
To link to this Article: DOI: 10.1080/10739140601000723
URL: http://dx.doi.org/10.1080/10739140601000723

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
 Instrumentation Science and Technology, 35: 41–52, 2007
Copyright # Taylor & Francis Group, LLC
ISSN 1073-9149 print/1525-6030 online
DOI: 10.1080/10739140601000723

Hybrid Control Strategy for Thermal

Cycling of PCR Based on a Microcomputer
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

Qiu Xianbo
Department of Automation, Shanghai Jiao Tong University,
Shanghai, P. R. China

Yuan Jingqi
Department of Automation, Shanghai Jiao Tong University,
Shanghai, P. R. China and State Key Laboratory of Bioreactor
Engineering, East China University of Science and Technology,
Shanghai, P. R. China

Abstract: In the polymerase chain reaction (PCR), the reference temperature profiles
of thermal cycling must be tracked rapidly and precisely so as to amplify DNA in
samples, strictly following the exponential function of 2n (n: thermal cycles) as
quickly as possible. For this purpose, a model based hybrid controller to follow up
setpoint changes is developed which is easily implemented with a microcomputer
ADuC834 for its simple configuration and low cost of computation. The hybrid
control strategy combines feedforward, feedback, and Bang-Bang actions to achieve
better setpoint response and steady state behavior. Based on the steady state model
of the thermocycler, the feedforward action serves as the model predictive control
for each setpoint, while the feedback controller is used as the local control action to
accelerate ramp rates and reduce the tracking errors. The model of the thermocycler
is identified with step response data at different operation regions. Experimental
results demonstrate that, besides the improvement of the tracking accuracy, both the
rise time and the decay time of each thermal cycle have been reduced significantly
with the hybrid control strategy, compared with the conventional PID. Therefore,
improved efficiency and accuracy of the polymerase chain reaction is achieved.

Keywords: PCR, Thermal cycling, Hybrid control strategy, Feedforward action,

ADuC834

Address correspondence to Prof. Yuan Jingqi, Department of Automation, Shanghai

Jiao Tong University, 200240, Shanghai, P. R. China. E-mail: [email protected]

41
 42 Q. Xianbo and Y. Jingqi

INTRODUCTION

As an important tool for the replication of DNA in biomedical research, PCR can
accurately identify and synthetically amplify specific molecules of DNA with tiny
amounts of samples.[1] To get polymerase chain reaction effectively, the tempera-
ture of the vessels with DNA samples must be changed quickly, repeatedly, and
accurately between the three stages, namely low temperature for annealing,
medium temperature for extension, and high temperature for melting. Tempera-
ture control in conventional thermocyclers or micromachined DNA polymerase
chain reaction chips is mostly performed with a PID controller for its simple
algorithm, robustness, and stability.[2 – 4] Because of the small thermal capacity
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

of the vessels, the small amount of reagents and the small thermal contact resist-
ance between chips and heaters, the micromachined chips can achieve much
faster temperature changes than conventional thermocyclers.[5,6] There are
several ways of heating and cooling in the conventional thermocyclers with
fixed chambers. One of them is to control the chamber temperature by alternating
the injection of hot air and ambient air as the medium for temperature transfer,
which achieves very rapid temperature exchange rates because of the small
thermal capacity of air.[7] Another is to control the vessels’ temperature by
heating or cooling the heating block. Its advantages include being simple in
structure, diverse in function, easy and cheap in use, but the large thermal
capacity of the heating block prevents the vessels’ temperature from changing
fast; as a result, the average ramp rates are low.[8 – 10] That means significant
cycling time, in this case, is spent non-productively in transition between
melting, annealing, and extension. On the other hand, tuning of a PID controller
is usually a compromise between the ramp rate and the overshoot; therefore, the
actual ramp rate will not reach the physical limit of the control system.
In this paper, the proposed hybrid control strategy is characterized with
applying Bang-Bang control and feedforward/feedback control to reduce
the time taken by a standard PCR. Besides the Bang-Bang control, the feedfor-
ward/feedback hybrid control strategy also helps to reduce the amplification
time with small overshoot.
Before PCR experiments, the thermal cycling settings must be decided pre-
viously, which makes the reference profile of the reaction temperatures become a
priori. This enables the design of an optimal open-loop power for the setpoint
temperature with the inverse dynamic method, which serves as the feedforward
action to improve the dynamic performance of temperature tracking.
Meanwhile, the feedback control action is used to reject the disturbances and
reduce the steady state errors. The model of the thermocycler is identified
with step response data at different operating regions. In the DNA amplification
system, the hybrid control strategy is realized with an embedded system with the
microcomputer ADuC834. Experimental results indicate that both the ramp rates
and the tracking accuracy of the thermal cycling of PCR are improved signifi-
cantly, compared to the best tuned PID, so that the hybrid control strategy
yields higher efficiency and accuracy of PCR, indeed.
 Hybrid Control Strategy for Thermal Cycling 43

EXPERIMENTAL

The heating block of the thermocycler, which consists of TE modules, a

sample block, and a heat exchanger, is the carrier for gene amplification. Ther-
moelectric (TE) modules have several outstanding advantages, such as small
size, simple installation, and reliable operation. TE modules may be used for
heating and cooling, depending on the direction of the applied current flow.

Peltier-Effect Heat Transfer

Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

As current passes through the TE module in the direction shown in Fig. 1, QL

is the absorbed heat on the cold side, and QH the released heat on the hot side.

Figure 1. Heat transfer due to Peltier effect.

QL equals the Peltier-effect heat aITL subtracting the Joulean-effect heat

1/2I2R and the transfer heat K(TH –TL):
1
QL ¼ aITL
I 2 R
KðTH
TL Þ ð1Þ
2

where a is the Seeback coefficient,[11] R the electrical resistance, I the

operating current, K the thermal conductivity, TH and TL the temperatures
on hot and cold side, respectively.
The applied voltage V is:
V ¼ aðTH
TL Þ þ IR ð2Þ
The input power P is:
P ¼ VI ¼ aIðTH
TL Þ þ I 2 R ð3Þ
and the thermal power given out of the hot side QH is:
1
QH ¼ P þ QL ¼ aITH þ I 2 R
KðTH
TL Þ ð4Þ
2

System Model Identification

Based on the energy balance, a high order nonlinear dynamic model of the TE
module system has been developed by Huang.[11] This complicated model has
been simplified to a linear second-order model using the small-signal
 44 Q. Xianbo and Y. Jingqi

linearization method. Here, in order to simplify model development, the

system model is identified based on the step response tests. Because the
thermal cycling of PCR is usually carried out above the ambient temperature,
a series of step response tests at different operating regions are conducted
above the ambient temperature. Figure 2 presents the experimental results at
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

Figure 2. Step responses of the heating block at the two extreme operating points.

the two extreme operating regions of thermal cycling. The data show that
the system model can well be approximated by a first order model with the
transfer function of:
kp
GðsÞ ¼ ð5Þ
Ts þ 1
where, s is the Laplace operator, kp and T model parameters. The two model
parameters are identified with the least squares regression. Both the gain kp
and the time constant T depend on, not only the properties of the heat
block, especially the TE module, but also the operating regions. Table 1
shows the identified model parameters at different operating temperatures.

Table 1. Model parameters at different operation regions

Operation temperature (8C) kp (8C/V) T (s)

39.98 4.52 47.3

49.02 4.87 48.4
58.75 5.13 49.7
69.01 5.40 50.8
79.80 5.64 52.2

The nonlinear dynamics of the heating block is obvious because the

model parameters in Table 1 change along with the operating regions. Both
the gain kp and the time constant T increase with the increasing operating
temperature. This results from the fact that the input power of the TE
module varies with the square of the applied voltage.
 Hybrid Control Strategy for Thermal Cycling 45

Controller Design

The thermal cycling of PCR requires tracking the periodically changed

setpoint temperature profile precisely and rapidly. Figure 3 shows the
typical temperature trajectory of PCR, in which, each setpoint temperature
corresponds to a different stage.
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

Figure 3. Typical temperature trajectory of PCR.

Because the time constant of the system model is large, a traditional PID
controller is not able to track the rapidly changing temperature profile accu-
rately. The best tuned PID controller functions, such as the dashed line
shown in Fig. 8. The rise time is as high as 20s and the overshoot or undershoot
is not small enough either, which results in low PCR efficiency.
In order to obtain better performance of thermal cycling, a model based
hybrid control strategy, which consists of Bang-Bang, feedforward, and
feedback control actions, is proposed. The Bang-Bang, as well as the feedfor-
ward control action is responsible for fast tracking, while the feedback control
action is for rejecting disturbances and eliminating the tracking errors.
Such hybrid control strategy is depicted in Fig. 4, where, G(s), C(s), and
F(s) are transfer functions of the plant, feedback controller, and feedforward
controller, respectively. And, r is the reference profile, e the servo error, u
the plant input, y the plant output, uBB the Bang-Bang action, uFF the feedfor-
ward action, and uFB the feedback action.

Figure 4. Configuration of the hybrid control strategy.

 46 Q. Xianbo and Y. Jingqi

The hybrid control strategy functions follow. If the absolute value of

servo error e is greater than a preset value 1, then uBB is switched on, so
that a maximum ramp rate is obtained; otherwise, uBB is switched off, while
uFF and uFB are switched on. The control action u is, therefore:


uBB; jej . 1
u¼ ð6Þ
uFF þ uFB; jej  1

The switching threshold 1 is chosen as a compromise between ramp rates and

overshoots. In practice, 1 may be tuned to get high ramp rates with very small
or, even, no overshoot. Moreover, it was found that the changes of 1 in a small
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

region have almost no influence on the control performance.

As shown in Fig. 4, the feedforward action plays, not only a role of model
predictive control, but also a role of consistent disturbance rejection to the
fluctuation of ambient temperature.

Bang-Bang Controller

According to Eq. (4), the maximum operating current of the actuator should
be applied for achieving the maximum heating capacity for sample block
heating. For sample block cooling, the pumped heat from the sample
block, together with the input power of the TE, is dumped to the hot side
of the TE and then dispersed into the ambient through a heat exchanger.
If the current applied to the TE is very large, the dispersion of the accumu-
lated heat on the hot side will be not so quick that the cooling capacity of
the cold side is wakened due to heat conduction from the hot side,
K(TH 2 TL), see Eq. (1). On the other hand, the cooling capacity of the
TE system is not a monotonic function of the applied current. Figure 5
shows the relationship between the applied cooling voltage and the transi-
tiontime from the high temperature (T1) to the low temperature (T2).

Figure 5. Voltage applied for optimal cooling.

 Hybrid Control Strategy for Thermal Cycling 47

Therefore, to achieve the maximum average cooling rate, an optimal current,

rather than the maximum operating current of the actuator, should be
applied. That means an asymmetric Bang-Bang action is needed. In our
practice, maximal power input of 15 V is applied for heating, while only
212 V is applied for cooling.

Feedforward Controller

The tracking error function to be minimized:[12]

Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

e 1
GðsÞFðsÞ
¼ ¼ ð1 þ GðsÞCðsÞÞ
1 ð1
GðsÞFðsÞÞ ð7Þ
r 1 þ GðsÞCðsÞ

For accurate tracking, the Laplace transform of the ideal feedforward control
action should be the inverse of the plant transfer function.
T 1
FðsÞ ¼ G
1 ðsÞ ¼ sþ ð8Þ
kp kp
However, such ideal feedforward control action is found to become easily
saturated, due to the large gain T/kp (e.g. 52.2/5.64 at 808C).

Figure 6. Applied voltages and the corresponding temperatures in the steady-state.

Figure 6 presents the relationship between the applied voltage and the cor-
responding steady-state temperature of the heating block when the heating
block is heated with different voltages from the ambient temperature. To
simplify the design of feedforward action, a linear fitting model is
developed. The fitted model equation is:

T ¼ 24:26 þ 4:79u þ eðuÞ
2:36 ð9Þ

 48 Q. Xianbo and Y. Jingqi

where, T is the steady state temperature, u the applied voltage, e(u) the linear
fitting error function, which is caused by the nonlinearity of the system model.
e(u) ¼ 2.36 at u ¼ 0. The constant 24.26 is the initial temperature of the heat
block at u ¼ 0, i.e., the ambient temperature.
From Eq. (9), the applied voltage for each setpoint temperature is deduced
with the inverse dynamic method, see Eq. (10) and Eq. (11). Therefore, the
applied steady-state voltage of the TE module for the desired steady-state
temperature of the heat block can be estimated with Eq. (11), which serves
as the feedforward action to speed up dynamic response of the thermal
cycling.
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

Tsetpoint
Tambient 2:36
eðuÞ

uFF ¼ þ l; l¼ ð10Þ
p p
uFF ¼ KF ðTsetpoint
Tambient Þ þ l; KF ¼ 1=p ð11Þ

In Eqs. (10) and (11), Tsetpoint and Tambient are the setpoint temperature and the
ambient temperature, respectively. (p ¼ 4.79) is the ratio of the temperature
change DT over the applied voltage change Du. p has the same physical
meanings as the model gain kp. KF is the gain of the feedforward controller,
which is tuned offline with Eq. (11). When the TE module undergoes any
deterioration, the system dynamics will change accordingly. Therefore, it is
necessary to identify the model parameters periodically to retune the gain of
the feedforward controller.
In feedforward action, l is the only parameter to be tuned. It not only
serves as compensatory power to optimize the calculated part of the feed-
forward action, but also compensates the moderate nonlinearity of the
model gain. Because the fitting error function e(u) is unknown for each
setpoint, l is tuned online at different setpoints in the thermal cycling
experiments.

Feedback Controller

The feedback controller is used to accelerate ramp rates, reject disturb-

ances, and reduce the tracking errors. A two-degree of freedom (2-DOF)
PID controller is adopted, of which the derivative action is only applied
to the output y, instead of the error e, so as to avoid derivative
kick due to abrupt changes of the setpoint in the thermal cycling, see
Eq. (13).
Because, the integral action of simple PID control at steady-state is
approximately equivalent to the feedforward action, see Eq, (11), integral sep-
aration is applied in the feedforward/PID hybrid control strategy to prevent
 Hybrid Control Strategy for Thermal Cycling 49

integral saturation, see Eq. (12).

8
< 0; jeðkÞj . M
>
 
uI ðkÞ ¼ Kc T X eðiÞ þ eði
1Þ
k
1 ð12Þ
>
: T ; jeðkÞj  M
i i¼d 2 1 þ eðiÞ2 =1
Td
uPID ðkÞ ¼ uP þ uI þ uD ¼ Kc eðkÞ þ uI ðkÞ
Kc ðyðkÞ
yðk
1ÞÞ ð13Þ
T
In Eqs. (12) and (13), T, Kc, Ti, Td are the sampling period, proportional,
integral and derivative gains, respectively, y(.) the actual output, e(.) the
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

error between the setpoint and the actual output, d ¼ Minfk, je(k)j  M g,
je(k)j  1, k ¼ 1, 2, 3 . . . , 0 , M , 1 and M is the switching threshold for
the use of integral separation, which is tuned by trial and error, 1 is the
switching threshold defined in Eq. (6). Trapezoidal integration is used to
avoid the effect of sudden changes of setpoint or output. In addition, a
nonlinear adjustment (1/(1 þ e(i)2/1)) is used to suppress overshoot,[13]
which makes the integral action smaller for larger errors.

RESULTS

The hybrid control strategy is carried out with an embedded system with the
microcomputer ADuC834. Temperatures are measured with high precision
thermistors (+0.058C). The control signal of D/A output is amplified to
provide a linear drive current for the TE modules. Both the feedforward con-
troller and the 2-DOF PID controller are tuned at different setpoints, and then
their gains are saved in the Flash Data Memory of ADuC834. Figure 7
presents the schematic of the embedded system with the microcomputer and
other main components.

Figure 7. Schematic of the embedded system.

Experiments demonstrate that the hybrid control strategy performs much

better than the best tuned PID controller, see Fig. 8. The rise time and the
overshoot are decreased significantly. Besides, both the heating rate and the
 50 Q. Xianbo and Y. Jingqi

steady state precision are improved also. Especially, the undershoot of cooling
is reduced significantly by the feedforward/PID hybrid control.
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

Figure 8. Experiments of thermal cycle tracking.

In Fig. 8, the improved ramp rates are very close to the system physical
limits of ramp rates that are achieved with Bang-Bang control, which is ben-
eficial to decrease the time taken to complete a standard PCR protocol. In
contrast, the output temperature of the conventional PID can hardly reach
steady state when the temperature holds are relatively short (e.g., 948C).
In order to evaluate the efficiency of the hybrid control strategy, two real-
time PCR experiments with different amount of initial HBV DNA samples
(including: 2  107, 2  106, 2  105, 2  104, and 2  103 copies/mL)
are designed and carried out, respectively, with PID control and hybrid
control. Figure 9 presents the analytical results of the two experiments
under the same thermal cycling settings. Thus, Ct (Y-axis) is the number of
cycles at which the fluorescence values reach the threshold value and the
value of X-axis the initial DNA copies.

Figure 9. Experimental results for gene amplification.

 Hybrid Control Strategy for Thermal Cycling 51

In Fig. 9, the comparison of the slopes of the linear regression curves

shows that the efficiency of PCR has been improved by the hybrid control.
The slope of the linear regression curve of PID control is 23.10, which is
far from the ideal value of 23.32. That means, with the conventional PID
control, the DNA samples are not amplified strictly following the exponential
function of 2n (n: thermal cycles), i.e., low efficiency of DNA amplification.
On the other hand, with the hybrid control strategy, the slope of the linear
regression curve is 23.26, which is very close to 23.32.
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

CONCLUSION

To improve the performance of the thermal cycling of PCR, a model based

hybrid control strategy for fast temperature tracking, with very low over-
shoots and offsets, is developed. With the step response data of the heating
block, a reduced first-order model is established at different operating
points. It is demonstrated that the hybrid controller, which is composed of
Bang-Bang, feedforward, and feedback control actions, yields more rapid
and more accurate temperature tracking, when compared with the best
tuned conventional PID. Especially, the feedforward action increases the
bandwidth of the system, which speeds up the command-following
response. Experimental results indicate that the hybrid control strategy not
only reduces the time to complete a standard PCR protocol, but also
improves the efficiency and accuracy of gene amplification. Because of the
simplicity of the control laws, the hybrid control strategy is easy to
implement on the microcomputer.

ACKNOWLEDGMENTS

We thank Mr. Li Qi and Mr. Zou Long for doing the experiments. Financial
support of Natural Science Foundation of China (Grand No. 60574038)
and the Open Project Program of the State Key Laboratory of Bioreactor
Engineering, East China University of Science and Technology, is
acknowledged.

REFERENCES

1. Saiki, R.K.; Scharf, S.J.; Faloona, F.; Mullis, K.B.; Horn, G.T.; Erlich, H.A.;
Arnheim, N. Enzymatic amplification of b-Globin genomic sequences and restric-
tion site analysis for diagnosis of sickle cell anemia. Science 1985, 230,
1350– 1354.
2. Pramod, K.U. An inexpensive PC based thermocycler for diagnosis based on poly-
merase chain reaction. Proc. RC IEEE-EMBS & 14th BMESI. 1995, 490– 491.
 52 Q. Xianbo and Y. Jingqi

3. Yoon, D.S.; Lee, Y.S.; Cho, H.J.; Sung, S.W.; Kwang, W.O.; Cha, J.H.; Lim, G.B.
Precise temperature control and rapid thermal cycling in a micromachined DNA
polymerase chain reaction chip. J. Micromech. Microeng. 2002, 12, 813– 823.
4. Zou, Q.B.; Miao, Y.B.; Sridhar, U.L.; Chong, C.S.; Chai, T.C.; Tie, Y.; Leng, C.H.;
Lim, T.M.; Heng, C.K. Micro-assembled multi-chamber thermal cylcer for low-
cost reaction chip thermal multiplexing. Sens. Actuat. A 2002, 102, 114– 121.
5. Debjani, P.; Venkataraman, V. A portable battery-operated chip thermocycler
based on induction heating. Sens. Actuat. A 2002, 102, 151– 156.
6. Eric, T.L.; Simpson, P.C.; Richard, A.M. Monolithic integrated microfluidic DNA
amplification and capillary electrophoresis analysis system. Sens. Actuat. B 2000,
63, 138–146.
7. Idaho Technology Inc. The LightCyclerTM –the smartest innovation for more
Downloaded By: [Shanghai Jiaotong University] At: 07:21 14 January 2009

efficient PCR. Biochemica 1998, 2, 4 –7.

8. Hansen, J.; Nussbaum, M. Application of Bismuth-Telluride thermoelectrics in
driving DNA amplification and sequencing reactions. 15th Intl. Conf. Thermoe-
lect. 1996, 1, 256– 258.
9. Chen, Q.G.; Wu, Z.P.; Lin, Y. Intelligent PID controller for PCR process. IEEE
Intl. Conf. Syst. Man. Cybernet. 1998, 1, 828– 831.
10. Technical profile: Mx4000w Multiplex Quantitative PCR System. Stratagene
Corporation, 2004.
11. Huang, B.J.; Duang, C.L. System dynamic model and temperature control of a
thermoelectric cooler. Intl. J. Refrig. 2000, 23, 197– 207.
12. Meystel, A.; Uzzaman, S. The planning of tracking control via search. Proc. Intl.
Symp. Intell. Cont.; Chicago, USA, 1993; Vol. 8, 554–559.
13. PID Control Toolkit for G Reference Manual; National Instruments Corporation,
2000.

Received June 1, 2006

Accepted October 21, 2006
Manuscript 1531

View publication stats