Aceptado Manuscrito

El desarrollo de la estrategia de precios para optimizar los beneficios totales en una batería del vehículo eléctrico cerrado cadena
de suministro bucle

Xiaoyu Gu, Petros Ieromonachou, Li Zhou, Ming-Lang Tseng

PII: S0959-6526 (18) 32558-7

DOI: 10.1016 / j.jclepro.2018.08.209

referencia: JCLP 13993

Para aparecer en: Journal of Cleaner Production

Fecha 23 de septiembre de 2017 Revisado:: Recibido

Fecha 16 de agosto de 2018 Fecha de aceptación: 20

Agosto 2018

Por favor citar este artículo como: Gu X, Ieromonachou P, L Zhou, Tseng ML, desarrollo de estrategia de precios para optimizar los beneficios totales
en una batería del vehículo eléctrico cerrado bucle de la cadena de suministro, Diario de Producción Limpia ( 2018), doi: 10.1016 /
j.jclepro.2018.08.209.

Este es un archivo PDF de un manuscrito inédito que ha sido aceptado para su publicación. Como un servicio a nuestros clientes estamos
proporcionando esta primera versión del manuscrito. El manuscrito será sometido a la corrección de estilo, composición y revisión de la prueba
resultante antes de que se publique en su forma final. Tenga en cuenta que durante los errores en el proceso de producción se pueden descubrir lo que
podría afectar el contenido, y todos los avisos legales que se aplican a la revista pertenecen.
 ACEPTADO MANUSCRITO

El desarrollo de la estrategia de precios para optimizar los beneficios totales en una batería del
vehículo eléctrico cerrado cadena de suministro bucle

Xiaoyu Gu a, *, Petros Ieromonachou b, Zhou Li b, Ming-Lang Tseng c

una Escuela de Economía y Administración de la Universidad de Nanjing de Ciencia y Tecnología, Nanjing

b Departamento de Sistemas de Gestión y Estrategia, Universidad de Greenwich, Londres

c Universidad Lunghwa de Ciencia y Tecnología, Taoyuan

*
Autor correspondiente.

direcciones de correo electrónico: x.gu@njust.edu.cn (Xiaoyu Gu), ip07@gre.ac.uk (Petros Ieromonachou), zl14@gre.ac.uk

T
(Li Zhou), tsengminglang@gmail.com (Ming-Lang Tseng)

IP
abstracto

CR
Este documento estudia una de tres periodos vehículo eléctrico de reciclaje de la batería y la cadena de suministro reutilización de bucle cerrado que
consta de un fabricante de la batería y un reconstructor. A diferencia de otros productos y las investigaciones existentes, baterías de vehículos eléctricos
utilizados pueden ser reutilizados para otros fines instante antes de reciclar, como el almacenamiento de energía. Con el fin de optimizar los beneficios
totales en toda la cadena de suministro en diferentes períodos de baterías de uso, este trabajo se desarrolla la estrategia óptima de precios entre el

US
fabricante y reconstructor, analiza las relaciones entre el rendimiento de retorno, la tasa de clasificación, tasa de reciclaje con el fin de optimizar el
beneficio total en diferentes período. El resultado sugiere que, comparando con la nueva fabricación de baterías, el reciclaje de la batería y la reutilización
contribuiría a reducir el consumo de materia prima, por tanto, reducir el impacto medioambiental, pero no pueden obtener beneficios financieros. Se
observa también que aunque la cadena de suministro de bucle cerrado es no lineal complicado, algunas relaciones entre los parámetros pueden ser
AN
tratados como lineal o cuadrática. Los resultados de esta investigación ayudarán a los médicos a comprender mejor toda la cadena de suministro de ciclo
cerrado con el fin de mejorar su colaboración.
M

Keywords: Closed-loop supply chain, Electric vehicle battery, Recycle, Reuse, Profit

1. Introduction
Currently, Electric Vehicles (EVs) considered as one of the future development directions for the automotive industry. According
ED

to International Energy Agency (2016), from 2005 to 2010 the number of EV sales worldwide, which includes both battery EV
and Plug-in Hybrid EV (PHEV), has increased from 1,670 to 12,480. By 2015, EV sales reached 1,256,900 which is almost 752
times than 10 years ago.
T

One of the most important parts on EV is the battery. Here are two main reasons: Firstly, approximately 50% of the cost of an
EP

EV is attributed to the battery (Lih et al., 2012). And secondly, unlike gasoline vehicles (GVs) that have a short refuelling time
(5 minutes), the EV charging time is long. A typical EV model (Nissan LEAF 40kWh) takes 8 hours to charge from empty with
a 6kW home charging point or 40 minutes super charge from empty to 80% capacity of electricity (Nissan,
C
AC

2018). Nevertheless, an EV cannot use the original battery until its end of life. Normally, due to performance and safety
concerns, the EV battery has to be removed when its capacity falls to 70 ∼ 80% (McIntire-Strasburg, 2015). Moreover, with the
increasing popularity of EVs, more and more batteries will need to be replaced. Discarding these batteries would constitute bad
environmental practice, with more far reaching long term effects. Used batteries must be recycled or reused rather than
discarded (Yu et al., 2013).

In many countries, similar to normal batteries, it is not allowed to put used automotive batteries through to landfill or incineration.
Instead, various EV battery collection and recycle schemes have been set up. For instance, in North America, Tesla, working
with Kinsbursky Brothers, recycles about 60% of its battery packs; in Europe, Tesla started working with Umicore on recycling
(Kelty, 2011); Nissan and Volkswagen require their EV customers to return used batteries to licensed points or local authority
battery collection schemes (Nissan, 2015; Volkswagen, 2016).

In addition, some organisations have already noticed the reuse of EV batteries when the EV industry just started. In the early
2010, the US National Renewable Energy Laboratory has undertaken a
 ACCEPTED MANUSCRIPT

project on EV battery reuse (Newbauer and Pesaran, 2010). The report of the project showed that secondary use of EV
batteries is necessary and the recycled EV batteries can be reused in the following ways: (a) grid-based stationary use, such as
energy time shifting, renewable capacity firming; (b) off-grid stationary use, for instance, as backup power and remote
installations (see Heymans et al. (2014) as well); and (c) mobile, for example, as commercial idle management or public
transportation. These applications for second use of EV batteries would significantly increase the total lifetime value, both
economically and environmentally. It is also gratifying that, currently, more and more EV manufacturers are considering the
secondary use of EV batteries: BMW and Nissan are expected to reuse returned batteries as home energy storage (Ayre, 2016;
Dalton, 2016). Chevrolet has set up an energy storage station using old EV batteries at the General Motors facility in Michigan
(Voelcker, 2016). In a summary, as can be found, recycling of EV battery has been widely accepted and operated by the EV
companies. The companies are also aware of the potential value of secondary use of EV batteries. However, the effect of
collaboration between reusing and recycling of EV batteries is lack of employment.

T
IP
Accordingly, this work develops an EV battery closed-loop supply chain (CLSC) model and investigates the pricing strategy to

CR
optimise the total profits of the supply chain. In detail, this research attempts to answer the following research questions: (1)What
are the relationships between relevant parameters and profit for EV battery manufacturer and remanufacturer? (2) How to
balance the accuracy and complexity of the result. In other words, how to simplify the relationship to the level that general

US
practitioners can understand when making a decision.

The rest of the paper is organised as follows. The next section reviews some relevant papers. Section 3 describes the model and
AN
derives the optimal quantity, the optimal purchase price and the maximised profit for manufacturer and remanufacturer,
respectively. Section 4 analyses the relationship between each parameter and the optimal profit in both period 2 and 3. Section 5
conducts some numerical experiments to express the findings graphically. Section 6 concludes and discusses the limitation of
M

the research.

2. Literature review
There is not much literature relating to EV battery recycle and reuse. Richa et al. (2014) forecast the value and quantity of EV
battery waste and then stated as a suggestion that, in order to increase the economic efficiency, the EV end-of-life battery
ED

management system must include an increase in reuse avenues before recycle or disposal. And Lih et al. (2012) discussed the
technology challenges, cost issues and business model for the EV battery secondary use applications. The results show that,
second use of EV battery is a perfect win-win deal which will probably create long-term and stable profits. The research also
T

estimates that, the profit rate could reach around 35% in the 15 service years of a 10kWh Li-ion battery pack. Neubauer and
EP

Pesaran (2011) estimated the impact of EV battery second use on the initial cost of PHEV/EV batteries to automotive consumers
and explored the potential applications for grid-based energy storage. Although second use of battery is not expected to
significantly affect today’s PHEV/EV prices, it has the potential to become a common component in the future EV battery life
C

cycles and to transform markets in need of cost-effective energy storage. Some research also studied the reuse of EV batteries
with focusing in the aspect of technology. For example, Tong et al. (2017) proposed a solar energy time shifting and demand side
AC

management system for secondary use of EV batteries with objectives to maximize economic benefits, minimize grid energy
consumption, or a balance of both. In terms of energy storage, Patten et al. (2011) suggested a wind energy storage system to
increase energy capacity factor, improve utilization, and make more efficient use of EV batteries prior to recycling.

As can be seen, there is few research examining how reused and recycled EV batteries affect the operational performance and
profit of a CLSC jointly. In other words, from the first use on the EV to reuse for other purposes and then entering the recycle or
remanufacture process, the EV battery CLSC is considered as a multi-period CLSC, which is also supported by Yu et al. (2013).
As a matter of fact, there are large volumes of papers studying the CLSC from the multi-period perspective. For instance,
Majumder and Groenevelt (2001) studied a two-period competition model between an original equipment manufacturer (OEM)
and a local remanufacturer by fixing the total cost for dealing with
 ACCEPTED MANUSCRIPT

the returned items. The model developed by Mitra and Webster (2008) analysed the regulation of remanufacturing activities.
Moreover, Ferrer and Swaminathan (2010) analysed the (re)manufacturer monopoly environment from a multi-period planning
horizon, and develop a strategy in order to optimise the profit for the firm. However, these studies did not take into
consideration about the process of secondary use of product. This means that, existing research cannot reflect the features of
EV used battery CLSC from return, reuse to recycle.

Therefore, the literature review suggests that, there is little research studying the EV battery combined with recycling and second
use processes. On another aspect, existing CLSC models are not able to reflect the practices of used EV battery reuse and
recycling and the characteristics of such CLSC; that is, unlike normal goods, EV batteries cannot be reused for their original
purpose when it degrades down to two thirds of full capacity, which significantly complicates CLSC operations. Moreover,
majority of the mathematical models and their results from the relevant literature appear to be too complicated for general

T
practitioners to understand, e.g., Cai et al. (2014) and Bulmus et al. (2014), which significantly limits the application and
implication of these research outcomes. Hence, this study aims to fill the CLSC research gap in EV battery reusing and recycling

IP
and to help managers better understand the CLSC. The objective of this paper is to design a model to describe a three- period
EV battery CLSC, then explores the interrelationship between EV battery manufacturer and remanufacturer, explains the

CR
reasons why recycling is still underdeveloped, and how profit can be increased by using right pricing strategy.

US
Similar with Bulmus et al. (2014), we involve EV battery manufacturer, who produces new EV batteries. Then we enrol more
participants in the second period: the EV battery manufacturer (same as period 1), and the remanufacturer, who collects used
AN
batteries and sorts them into high-quality and low-quality returns (Cai et al., 2014). Then, reusable batteries will be selected from
high-quality returns, i.e. second market for reusing battery. To reflect the specific feature of used battery, the author proposes a
third period where reused batteries are collected for recycling. In this study, we aim to optimise the total profit in the entire
M

Supply Chain (SC) by taking into consideration the return rate, sorting rate, processing cost and recycling rate.

3. Model description
We consider a three-period model to describe an EV battery manufacturing/ remanufacturing system as shown in Fig. 1.
ED

Initially, the demand for EV battery raw material is based on the quantity of required EV batteries. Furthermore, the demand for
EV batteries depends on the EV market size.
T

Fig. 1: A three-period model in manufacturing/remanufacturing system

EP

• In period 1, all EV batteries are made from raw materials. Battery manufacturing quantity is based on EV demand.

• In period 2, batteries are made from raw materials, and high- and low-quality returns. First, a proportion of θ of used EV
C

batteries is returned. We categorise returns into two classes (Cai et al., 2014; Gaines and Singh, 1995): a proportion ( α ) of
high-quality returns and a proportion (1 α − ) of
AC

low-quality returns. Then, high-quality returns are sorted again: a proportion of β will be reused, meanwhile (1 β
− ) of them will be recycled directly. Because of depletion in the battery recycling
process, we set λ land h λ as the remanufacturing rate for low- and high-quality returns. This
indicates that λ l( or λ h) of low- (or high-) quality returns can be recycled to materials.

• In period 3, batteries are made from raw materials, high- and low-quality returns and reused batteries. Those reused batteries
reach their end of life and will be recycled as well. The recycling rate for reused batteries is u
λ . The other returns will be recycled as indicated in period 2.

The notations are listed in Table 1.

Table 1: Notations Input

parameters
 ACCEPTED MANUSCRIPT

θ Battery return yield α High-quality used battery sorting

rate

β Reusable battery sorting rate c ntr Cost for new EV battery (including
material cost and manufacturing
cost)

c Remanufacturing cost for high- quality c Remanufacturing cost for

h l
returns low-quality returns

cu Remanufacturing cost for reused battery λh High quality returns recycling rate ( 0
< <λ h) 1

T
λl Low quality returns recycling rate ( 0 λu Reused EV battery recycling rate ( 0
< <λ l) 1 < <λ u) 1

IP
M EV market size

CR
EV

Decision variables

p ; i = 1, 2,3 EV price in period i q ; i = 1, 2,3 EV demand in period i

US
EVi EVi

p i i; = 1, 2,3 Battery price in period i q ini ;= 1, 2,3 Battery quantity made from raw
material in period i
AN
q ili ;= 2,3 Quantity of batteries re- manufactured q ihi ;= 2,3 Quantity of batteries re-
from low-quality returns in period i manufactured from high- quality
returns in period i
M

q
3u
Quantity of batteries re- sil i; = 2,3 Price of purchasing low- quality
manufactured from reused returns in period i
batteries in period 3

sihi ;= 2,3 Price of purchasing high-quality returns s

3u
Price of purchasing reused
in period i batteries in period 3
ED

Objective variables

Π ;i i = 1, 2,3 Total profit in period i

T
EP

Intermediate variables

v Customer’s willingness to pay for the EV H Coefficient between battery material

quantity and EV sold quantity:
C

q Hq
i
= H EVi
,( > 0)
AC

δm Coefficient between EV sale price and the k For simplification, suppose

value of battery on EV:
k = δm / ( HM )
pi = δ p
m EVi
, (0 <<
δm 1)
EV

Π in; i = 1, 2,3 Profit from new battery Π il; i = 2,3 Profit from low-quality returns
manufacturer in period i re-manufacturer in period i

Π ih; i = 2,3 Profit from high-quality returns Π u3 Profit for reused battery returns
remanufacturer in period i remanufacturer in period 3

3.1. Nash equilibrium in period 1

 ACCEPTED MANUSCRIPT

The EV market size is defined as M . Similar with Ferguson and Toktay (2006) and Debo et al.
EV

(2005), both EV price p and customer’s willingness v are uniformly distributed between 0 and 1
EV 1

(i.e., v ∈ [0,1] and p ∈ [0,1] ) . By adopting same utility-based approach as Bulmus et al. (2014),
EV 1

customers utility of buying EV is ( vp

−
EV 1
) . Therefore, the quantity of EV which is sold in this
period becomes

qM EV 1
=
EV
(1 − p
EV 1
) (1)

The demand for battery material is based on the demand quantity of EVs, that is q Hq
1
= H
EV 1
( > 0) .

T
And the EV battery price accounts for m δ of the total EV price ( p = δ p
m EV 1
) . In period 1, all EV

IP
1

batteries are made from the raw materials, that is q = .qHence,

1
through substituting 1 q and p
1 n 1

into Eq. 1, we have

CR
q HM
=
1n EV
(1 − pδ/
1 m ) (2)

US
Let k = δm / ( HM EV
) , and through formula transformation, the battery price in period 1 is

p =−
δm kq (3)
AN
1 1n

Battery manufacturer’s profit is the sale price minus both the new EV battery cost (including both raw material cost and
manufacturing cost), then multiply by the quantity of sold. Through substituting Eq.
M

3, the profit can be expressed as

Π Π= 1 1n
= ( p c−q 1
ntr ) 1n
= (δ m
− kq
1n
− c ntrq ) 1n
(4)

3.2. Nash equilibrium in period 2

Similar to period 1, the entire demand for EV depends on market size and EV price in period 2:
ED

q
EV 2
= M
EV
(1 − p
EV 2
) (5)
T

With q Hq
=
2
and p = δ p
m EV 2
, the quantity of EV batteries required in this period is
EP

EV 2 2

q HM
= 2 EV
(1 − p
EV 2
/δm ) (6)
C

Let k = δm / ( HM EV
) , then we can derive the EV battery cost function by inversing Eq. 6:
AC

p =−
δm kq 2 (7)
2

In this period, θ of batteries will be returned. These returned batteries will be sorted into three classes: reusable returns,
high-quality returns and low-quality returns. As shown in Fig. 1, those reusable returns will be reused to other places, for
example, as energy storage. Both high- and low-quality returns will be recycled into battery materials directly. Therefore, in
period 2, the battery materials come from three sources: raw natural materials, and material recycled from both high-quality
returns and low-quality returns. The amount of raw natural materials required amounts to the material requirement for making a
battery minus the quantity of materials recycled from the returned batteries:

q = −q 2−q q (8)
2 n 2 l 2h

We can derive the inverse the demand function Eq. 8 by substituting Eq. 7 as follows:
 ACCEPTED MANUSCRIPT

p2 =−
δm kq
( + q q+ ) (9)
2 n 2 l 2h

The total return at period 2 is the return rate θ multiplied by the quantity of battery material in the previous period, i.e.
H θq EV 1
. As mentioned, ( −
1α ) of them are classified as low-quality returns.
For the other returns, β of them are high-quality returns which will be recycled, while ( 1 β− ) of
them will be sorted as reusable returns. The demands for high-quality returned batteries and low- quality returned
batteries are respectively:

•=qHq θα ( 1 − β ) (1 − s 2 h )
•• = 2 h EV 1

(10)
•• q H q θ (1 − α ) (1 − s )

T
2 l EV 1 2l

IP
The quantity of materials made from different categories of returns are the quantity of returns multiplied by the
returned batteries recycling rate, λ hand l λ . The profit for new battery

CR
manufacturer and the low-quality battery remanufacturer is defined as battery sale revenue minus recycling cost and

returned battery purchase cost. The profit functions for the new battery manufacturer (

Π 2n ), low-quality and high-quality battery

remanufacturer in period 2, i.e. Π l2and Π h2 , are
US
AN
•Π = ( p c −q pntrc s) q p c
2n 2 2 n

••
•• Π 2l
= (λ l −− ) (11)
M

2 l 2 l 2l

•• Π 2h
= (λ h 2
−−
h
s q)
2 h 2 h

In summary, the total profit in period 2 could be

Π Π=Π Π 2 2 n
+ 2 l
+ 2h
(12)
ED

The existence and uniqueness of Nash equilibrium in period 2 can be proved (see Appendix A) The optimal total profit is
∂ Π Π Π ∂0 ∂
achieved by using first-order condition, that is
2
=
2
=
2
= .
∂q ∂s ∂s
T

2 n 2 h 2l

* * * * *
Because of the length limit of paper, the optimal values q, s, s, q, q are shown in detail
EP

2 n 2l 2h 2 h 2l

from Eq. A.43 to Eq. A.47 in Appendix A as well.

3.3. Nash equilibrium in period 3

C

Similarly, the EV quantity in period 3 is

AC

q
EV 3
= M
EV
(1 − p
EV 3
) (13)

With q Aq
= 3 EV 3
and p
3
= δ p
m EV 3
, the total demand for EV batteries in this period is

q HM
=3 EV
(1 − pδ
3
/ m ) (14)

Let k = δm / ( HM EV
) , then we can achieve the price function by deriving from Eq. 14:

p3 =−
δm kq 3 (15)

In this period, the battery material consists of raw natural materials, high and low-quality returns and end-of-life reused battery
returns. The demand for batteries made from raw natural materials is total market demand minus all EV batteries made from
returns:
 ACCEPTED MANUSCRIPT

q3n = −q −
q−q
3 3 l
q (16)
3 h 3u

And the price

p =−
δm kq
( 3 n
+ q 3ql + +q ) (17)
3 3 h 3u

The return quantity in period 3 is new batteries manufactured in period 2 multiplied by the return rate,
i.e. H θq EV 2
. In this period, all returns in the three categories (low-quality, high-quality and reused

returns) will be recycled with the quantity:

•=q 3Hl q θ (1 − α ) (1 − s )
•• =
EV 2 3l

T
••q=H q θα (1 − β ) (1 − s ) (18)

IP
3 h EV 2 3h

•• q Hq θ αβ
EV
(1 − s
3u
)

CR
3 u 2

El beneficio de todo para el nuevo fabricante del producto es una nueva demanda de baterías EV multiplicada por la ganancia cada nuevo EV de la batería
que se puede ganar en la fabricación. Los beneficios para las baterías hechas de retornos reciclados o reutilizados son los ingresos menos todos los costos.

US
Suponiendo la ecuación. B.48, los beneficios son:

• O = ( pcqqp
- )
3 n
3
NTR 3n
AN
•• O = l ll - +( c )
sqqp 3l
•• 3 l 3 l 3 l 3l
(19)
•• O 3h
= l h
- +( c cuadrados ) 3h
M

3 h 3 h 3 h

•• O 3u
= u
QP - +( c u cuadrados )
3 u 3 3 u 3u

El beneficio entero es un beneficio para la suma del fabricante / remanufactureros:

P O=O O O+
3 3 n 3l
+ 3 h
+ 3u
(20)
ED

La existencia y unicidad del equilibrio de Nash en este período se pueden encontrar en el Apéndice B. Mediante el uso de la condición de primer
orden para adquirir el beneficio óptimo para cada agente, los valores óptimos *
q,3 n
*
s,
3l

*3l
s,
* *
s, q, q,
* *
T

q en el período 3 se expresan a partir de la Ec. B.53 a la Ec. B.59 en detalle en el Apéndice B.

3h 3u 3h 3u
EP

4. Discusión
En esta sección, vamos a analizar las relaciones entre los parámetros (es decir, yo , una , b , l, l l, h l, u
C

c,NTR
c,l c,h c)u y la ganancia total, tanto en período de 2 y 3. Dos definiciones y lemas son primero
AC

presentada como preparación:

Definición 1. Según Fraden (2004) y Cooper (1970), linealidad se define como una relación de desviación máxima entre la curva de
práctica y equipada línea recta con la salida de escala completa, es decir

n = max YY
( N / A) * 100% .

Fig. 2: Ejemplo de linealidad

Fig. 2 es el diagrama esquemático para la linealidad. A través de regresión lineal, la línea ajustada óptima puede ser resuelto. Para simplificar,
se dibuja un línea ajustada sencilla simplemente conecta dos puntos en la curva que un punto de cruce entre
x = y0la curva original y un punto de cruce entre x = y1 el original
curva. Es fácil demostrar que la linealidad óptima de la curva es siempre menor o igual a la linealidad de la línea ajustada sencilla.
Si la linealidad n = , la0 función es completamente lineal. En esta investigación, se
afirman que la curva es aproximadamente lineal si la linealidad de la línea ajustada simple es 10% ± .
Definición 2. Según los algoritmos (Sedgewick, 1988), red de búsqueda es un ergódico mundial
 ACEPTADO MANUSCRITO

método de búsqueda utilizado para encontrar el valor objetivo.

En esta investigación, red de búsqueda se utiliza para encontrar el máximo y el mínimo de linealidad por el recorrido a través de la búsqueda del valor de
todos los parámetros posibles en base a la resolución.
Lema 1. Función con formato

NN +x N xfx+
2

() =
0 1 2
(21)
DD +x
0 1

Si ( NN
0
x
+ 1
) N 2x
2
fx enfoques
sostiene, a continuación, ()
NN +xfx
() ≈+ 0 1
(22)
DD x

T
0 1

IP
Lema 2. dado q EV
< M EV
y0 <<,
d m0 1 < Hkq 1
< sostiene.
EV

CR
= dm ( HM ) en Hkq, obtenemos Hkq .
EV
Prueba: Sustituyendo k / EV
= m
A medida que el quanitity de EV
M dEV
EV EV

es menor que el tamaño del mercado ( q < M EV

) y el precio de la batería es menor que el precio de todo EV

US
EV

qMδ
(0 < <
d m), we1have 0 < 1. Hence, 0 Hkq < 1.
EV
< m
<
EV

EV
AN
Based on proposed definations and lemmas, the monotonicity and linearity are discussed and analysed. Due to the length
limitations of research paper, all proofs between paramters in period 2 and 3 are shown in appendix C and D respectively. In the
appendix, we discuss the relationship between parameters (return yields, recycling rates and remanufacturing costs) and total
M

optimal profit in both period 2 and 3, especially the range of first-order derivative and range of linearity. Although the profit
functions are too complex, they can be simplified for approximation. The character of first-order derivative represents the
monotonicity. The range of linearity indicates the possibility that a relationship is supposed to be linear. With all proofs in
appendix C and D, to summarise, all relationships can be encapsulated in Table 2. We use “L” to express the linear relationship;
and “Q” to show the quadratic relation; then “ ↑ / ↓ "Representan la correlación positiva / negativa; mientras que "N / A" muestra
que la relación es inaplicable.
T ED

Tabla 2: Resumen de teoremas

EP

yo ab l l l
l h u
c NTR cl ch cu
C

P
*
↓ & L↑y ↑ ↑ N/A Q ↓ y Q ↓ y QN / A
2

L L
AC

↓ & LL Q ↓ yQ↓yQ ↓ yQ
*
P 3
L ↑ ↑ ↑

Por lo tanto, se llega a seis observaciones, como se muestra en las proposiciones siguientes:

Propuesta 1. La relación entre yo , una , b y el beneficio óptimo total en el período de 2 / período de 3 (es decir,

P /2 II) puede
3
ser tratado como lineal, basado en 10% ± condición límite de linealidad.

Propuesta 2. La relación entre la totalidad del coste ( NTR c, c,l c,h c)u y la ganancia óptima P, 2 P es
3

cuadrática.

Propuesta 3. b tiene una correlación positiva con II. 2

Propuesta 4. Todos los índices de reciclaje l se correlacionan positivamente con tanto P y2 3 II.
 ACEPTADO MANUSCRITO

Propuesta 5. Las relaciones entre yo , c,l c,h c uy P, 2 W 3son negativos.

Propuesta 6. Otras relaciones dependen de diferentes situaciones y valores iniciales. Sobre la base de las

proposiciones, se consideran las siguientes discusiones Insight gestión:

• Cuanto mayor sea el rendimiento EV retorno batería yo , cuanto menor es la ganancia óptima total en ambos periodo 2 y 3. Esto indica que los
rendimientos no son capaces de generar más beneficios para el SC. Este hallazgo es similar con Tierney (1996), quien argumentó que el reciclaje
puede no ser una actividad valiosa desde el punto de vista económico. Sin embargo, se aprecia que el aumento del reciclado es capaz de reducir el
consumo de nuevos productos y materiales (Thomas, 2003). Mientras tanto, este hallazgo explica por qué el reciclaje de baterías EV no se han
adoptado ampliamente.

• La relación entre el rendimiento de retorno de la batería reutilizable b and optimal profit in periods 2 and 3 is different. It has a

T
positive linear relationship with the optimal profit in period 2. This means the more reusable EV batteries the higher profit in period

IP
2. But in period 3, which is more complex than period 2, the trend of relationship depends on the initial value of parameters.
Therefore, in order to increase the overall profit, EV battery should not be used till the end of its life cycle. Instead, it should go to

CR
2nd stage of reuse when it reaches 60-70% of full capacity.

• Recycling rates for different quality of returns λ l, λ hand u λ have a positive relationship with

US
optimal profit. This is because the higher quality of returns, the less effort (hence lower cost) needed in the recycling process,
resulting in higher recycling efficiency. As an example, Gaines (2014) looks ahead at how to improve the recycling efficiency
technically.
AN
• According to King et al. (2006), remanufacturing could be the best solution to deal with the returns. And furthermore, in EV
battery CLSC, the higher costs of recycling operations (i.e. c l , c hand u c)
M

the lower optimal profit in both period 2 and 3. Producing new EV battery is more profitable than recycling used batteries.
Therefore, how to reduce the recycling cost in the entire CLSC is a considerable problem. There are some solutions, for
example, to develop better technologies (Hutchinson, 2008) or export recycling operations to the countries with lower
processing costs (Geyer and Blass, 2010).

• Although relationships between parameters and entire profit are complicated, our research finds that they actually can be
ED

simplified as linear or quadric. For instance, relationships between θ , α , β and

Π 2, Π can
3
be considered a linear relationship and relationships betwee ntr c, c l , c h, c uand Π 2,
Π can be treated as quadric. This simplification would be considerably helpful to managers when analysing the SC and
T

making decisions.
EP

5. Numerical experiments
C

The previous section analysed and discussed the relationships among parameters. This section will have some numerical
experiments as the implementation of the model. Section 5.1 proposes a numerical example as a case to explain how the
AC

model can be used in industry. And section 5.2 describes the use of methodology in practice and verify the relationships as
shown in Table 2.

5.1. Numerical example

According to International Energy Agency (2017), EV market size is predicted to 18,000,000 in 2020,
i.e. M =EV 18,000,000 . And EV battery price accounts for around 30% of electric car price i.e.

δ m= 0.3 . According to Fred Lambert (2017) and Mark (2014), taking Tesla Model as an example, the whole value for each
EV battery is £11700. According to Binks (2016) and Will Date (2015), it averagely costs £860 to process a used battery. We
assume the low-quality used battery recycling processing cost is £950, high-quality is £850 and reusable battery is £800. To
normalize the cost into the same scale, without losing generality, set
cntr= 0.2 as benchmarking, other costs against the

benchmark cl = 0.2*950 /11700 0.016

= , c h= 0.2*850 /11700 0.015
= ,
 ACCEPTED MANUSCRIPT

cu= 0.2*800 /11700 0.014

= . We can come up with a numerical example to demonstrate the model.

All input parameters of numerical example are summarised in Table 3:

Table 3: Numerical example parameters

M= 18000000 H= 4 δ m= 0.4 θ = 0.3 α = 0.8

EV

β = 0.7 cntr= 0.2 c = 0.015 c = 0.016 cu= 0.014

h l

λ l= 0.8 λ h= 0.9 λ u= 0.85

In period 1, with equations from Eq. 1 to Eq. 4, the optimal quantity for battery raw materials (for period 1, this is also the

T
q q= = ×n
7
optimal total quantity) is 1 1
1.8 10 . The optimal sale price is

IP
p1 =17550
6
and the optimal profit in this stage is Π Π=1.8 110
1n
=× . Moreover, the optimal EV
6
sale quantity is q=× 4.5 10 .

CR
EV 1

*
While in period 2 we substitute q as the initial input parameter for this period. By applying Eq.
EV 1

A.43 to Eq. A.47, the optimal values are as follows in Table 4. Table 4: Optimal

q=
*
values in period 2

1.70 10
×
7
q=
*
4.73 10
×
5
q=
* US
5.48 10
×
5
q 2=1.81 10 ×
* 7
AN
2 n 2l 2 h

* * 6 * 6
p=
2
17526 Π 21.36
= 10 × q=
EV 2
4.52 10
×
M

*
And in period 3, we substitute q as initial EV quantity in this period. By applying Eq. B.49 to Eq.
EV 2

B.59, the optimal values in this period are as shown in Table 5. Table 5:

Optimal values in period 3

q =n q =l
* 7 * 5 * 5 * 6
3
1.58 10
× 3
4.75 10
× q=
3h
5.50 10
× q=
3u
1.30 10
×
ED

20475 * 6
q3=1.82 10 × p=
* 7
Π 3799660
*
q=
*3
= 4.54 10
×
EV 3

In a summary, in this numerical example, the optimal manufacturing yield and the optimal price in each of periods can be
T

derived. Also, in this example, as can be seen, from period 1 to period 3 quantity of batteries made by the raw material is
EP

decreasing and the total profit is decreasing as well. This indicates that, with the increasing returned EV batteries, battery
production is less dependent on natural resources, but in the meantime, the total profit is reduced because of the cost of
remanufacturing.
C
AC

5.2. Analysis
Based on the initial numerical input in Table 3, this subsection shows the relationship between parameters and total
profits by with figures.

5.2.1. Period 2

• Relationship between θ , α , β and total profit in period 2

Fig. 3: θ , α , β vs total profit in period 2 The function between battery return yield θ , high-quality used battery sorting rate α , reusable

battery sorting rate β and the optimal total profit

*
Π ,2 as shown in appendix (see Eq. C.60, Eq. C.67, Eq.

C.74), are expressed as the original curves in Fig. 3. And the fitted line with ‘*’ are drawn through linear regression. In
addition, three functions below express the profit functions and fitted lines.
 ACCEPTED MANUSCRIPT
13
1.95883 10×
(θ)
* 9
Π 2
= − 137616 θ θ+θ − 5.09116 10×
+ 3846.15 (23)
6
≈ − 1.46 10 + ×1.80 10
6
×

13
1.06 10
×
(α) α
* 5
α − 2.95 10
× 7.46
≈ 10 + ×7.64 10
9 5
Π 2
= − 72527.4 × (24)
3604.68 −α

14
2.1422 10
×
2 ( )
* 5
β + ×8.22 10
10
1.29 10 7.71 10
5
Π β = − × ≈ × (25)
16668.7 −β

• Relationships between l λ , λ hand total profit in period 2

T
The figure and equations below show functions λ land h λ with optimal profit in period 2 (As shown

IP
in Eq. C.80 and Eq. C.88 in the appendix), as well as the fitted line by linear regression.

CR
Fig. 4: λ l, λ hvs total profit in period 2
10
8.81 10 λl + ×1.06 10
7
×
( λ)
* 6
1.46 10
× 1.57≈10 λl + ×1.24 10
7 5
Π (26)

US
2 l
= − ×
λ lλ( l
− +2 )667.68

9
8.79 10 λh + ×8.78 10
7

* × 6
λh
AN
(λ h ) 1.46 10
× 1.88≈10 + ×1.20 10
7 5
Π 2
= − × (27)
λ hλ ( h
− +2 )556.59

• Relationship between l c , c and total profit in period 2

M

Through Eq. C.102 and Eq. C.108, we can describe the quadratic relationships using the figures and expressions below:

Fig. 5: c l , c vs total profit in period 2

h

( c)
* 62 5
= − 0.27 10 0.46 10 c l + ×1.37 10
6
Π × cl − × (28)
ED

2 l

( c)
* 62 6
= − 0.32 10 0.54 10 c h + ×1.37 10
6
Π 2 h
× ch − × (29)
T

( Π 2 (θ) =)
*
With the concept of linearity, in period 2, the linearities are summarised as η 0.67% ,
EP

( Π (α) =) ( Π ( β) =) ( Π ( λ)l= ) (Π ( λ h=) )

* * * *
η 2
0.71% , η 2
0.87% , η 2
2.95% , η 2
1.37% . This

means, although the relationships are complex, they can be treated as linear. And policy makers will be easier to analyse the
C

profit. In this specific case, in order to increase the revenue, the rate of high- quality used battery and reusable battery should
be increased and the waste in remanufacturing processes should be decreased. For example, encourage EV users to use
AC

batteries carefully and test them in time so more surplus value could be remained at reusable level; while the remanufacturer
should improve the techniques in recycling and remanufacturing. What is more, the cost for the remanufacturing has quadratic
relationship with the total profit
*
Π .2 And the higher cost, the less total

optimal profit.

5.2.2. Period 3

All relationships and description functions for period 3 are given below.

• Relationships between θ , α , β and total profit in period 3 The functions between Battery yield θ , high-quality used battery
sorting rate α , reusable battery sorting rate β and optimal profit in period 3 ( * 3
Π ) are shown in Eq. D.114, Eq. D.120, Eq. D.127.

Fig. 6: α , β , θ vs total profit in period 3

 ACCEPTED MANUSCRIPT

Fig. 6 shows the functions and the fitted lines. The mathematical expressions are shown below.

− 13 6 75 26 4
• 3.13 10
× θ − 1.54 10
× θ + ×1.89 10 θ •
• 31 3 37 2 41 •
θ
41
• + ×9.68 10 θ + ×1.43 10 θ + ×5.02 10 − 2.72 10
× •
(θ)θ = − (30)
*3
Π
19 3 26 2 35
θ
31
2.46 10 3.47 10 8.70 10 1.51 10
×
4
× θ − × θ − × −
6
≈ − 3.36 10 θ + ×1.81 10
6
×

− 14 6 75 27 4
• − 1.66 10
× α − 1.47 10
× α − 3.24 10
× α •
• 33 3 38 2 42 •
α
41
• + ×2.02 10 α − 3.15 10
× α − 5.53 10
× − ×
8.59 10 •
(α)α =+ (31)
*3
Π
20 3 27 2 37
α
33
4.42 10 5.87 10 + ×1.79 10 1.12 10
×
4
× α − × α −

T
5
0.61 10 α + ×7.51 10
5
≈ ×

IP
− 14 33 3
β •
27
2.40 10 + ×2.47 10 + ×6.33 10 8.89 10
×
6 7 5 4
• × β β β −
• 44 •

CR
β
39 42
• + ×3.13 10 7.39 10 + ×2.37 10
2
β − × •
( β)
*3
Π = (32)
20 3 38
(β β )
28 34
+ ×5.14 10 3.12 10 + ×2.16 10 + ×2.88 10
4 2
β − × β

5
≈ − 0.21 10 β + ×8.14 10
5
×

• Relationships between l λ , λ h, λ uand total profit in period 3

The figure below shows the relationships between high-quality, low-quality, reused EV recycling rate ( l
US
AN
λ , λ h, λ u) and optimal profit in period 3 ( * 3 Π ). The functions are described in Eq. D.134, Eq.

D.142, Eq. D.150. The linear regressed lines are also shown in the figure. Fig. 7:
M

λ l, λ h, λ uvs total profit in period 3

86 75 13 4 14 3
• 2.32 10
× λ l
− 1.73 10
× λ l
− 4.66 10
× λ l
+ ×5.28 10 λl •
• 19 2 20 •
λl
20

• − 2.18 10
× λ + ×1.70 10 + ×6.50 10 •
( λ)l =•
*3 l
Π
64 73 12 2
λ
6
− 8.04 λ
5
+ ×3 10 λ − 1.69 10
× λ + ×1.44 10 λl • (33)
ED

l l l l
• 12 14 •
• − 2.88 10
× λ l + ×9.62 10 •
5
1.66 10 λ l + ×6.64 10
5
≈ ×
T

85 76 13 4
• 2.30 10
× λ − 1.71 10
× λ − 3.87 10
× λh •
EP

h h
• 14 3 19 2 20 •
λh
20

• + ×4.38 10 λ − 1.51 10
× λ + ×1.18 10 + ×3.53 10 •
(λ h ) = •
*3 h h
Π
64 73 12 2
λ
6
− 8.03 λ
5
+ ×2.50 10 λ − 1.41 10
× λ + ×1 10 λh • (34)
C

h h h h
• 12 14 •
• − 2× 10 λ h + ×5.57 10 •
AC

5
2.11 10 λ h + ×6.13 10
5
≈ ×

9
8.78 10 λu + ×3.64 10
7

* × 7
Π 3 (λ u ) = − 1.48 10
×
λ uλ ( u
− +2 ≈
)239.11 (35)
5
4.37 10 λu + ×4.32 10
5
×

• Relationships between l c , c , c and total profit in period 3

h u

Fig. 8: c , c , c vs total profit in period 3

h l u

( c)
* 52 5
= − 2.74 10 4.69 10 c l + ×8.07 10
5
Π 3 l
× cl − × (36)
 ACCEPTED MANUSCRIPT

( c)
* 52 5
= − 3.27 10 5.41 10 c h + ×8.08 10
5
Π 3 h
× ch − × (37)

Π
*
3 ( c) u
= − 0.76 10
×
62
cu − 1.28 10
×
6
c u + ×0.82 10
6
(38)

( Π (θ) =) ( Π (α) = )
* *
Similar with period 2, in period 3, the linearities are η 3
0.58% , η 3
1.26% ,

( Π 3 ( β) =) ( Π 3 ( λ)l= ) ( Π 3 ( λ h=) ) ( Π 3 ( λ u=) )

* * * *
η 0.83% , η 1.61% , η 1.87% , η 0.72% .

Therefore, these relationships can be treated as linear as well. In this period, the more reusable batteries, the less total profit. It
means that, although reusable batteries are benefit to the environmental sustainability, it may reduce the total profit in period 3.
The remanufacturing costs for different quality returns are quadric with the total profit in period 3.

T
To conclude, in this numerical analysis, through observing formulas from Eq. 23 to Eq. 38, all equations revalidate the

IP
propositions in section 4. Moreover, Fig. 3 to Fig. 8 show the relationships between independent variables and dependent
variables, Π and
2
3 Π . Although profit functions

CR
themselves are complex and non-linear, as can be seen from Fig. 3 to Fig. Fig. 6, in this given case, all sorting rates ( θ , α , β ) and all

recycling rate ( l λ , λ h, λ u,) can still be regressed with a straight line. All
the cost has quadratic relation with total profit.

6. Conclusion
US
In this paper, we proposed a three period EV battery closed-loop supply chain to describe the return, reuse and recycle
AN
remanufacturing process. Differing from other product, EV battery should be disassembled from the vehicle when its capacity
falls to 70% ∼ 80%. But it can be reused for other purposes. In period 1, all batteries are made from raw materials. In period 2,
some used batteries are returned and they are sorted to high-quality and low-quality used batteries while in this period, some
M

batteries are selected for reusing. And in period 3, after reusing, used battery has to be recycled. The Nash equilibrium between
(re)manufacturers in period 2 and period 3 are and the optimal (re)manufacturing quantities and acquisition prices are derived.
We then analysed the relationships between used battery return yield θ , high-quality sorting rate of used battery α , reusable
sorting rate β ,

recycling rate ( l λ , λ hand u λ ), ( re-)manufacturing costs ( l c , c hand u c ) and total optimal profit in
ED

both period 2 and period 3 ( Π and

2
3 Π ). As complexity of the CLSC model, these relationships are

difficult to describe. Through the normalization process, costs and prices are distributed into [0,1], by using the grid search
T

method, portfolio of all parameter values are traversed.

EP

In comparing with the existing models, such as Neubauer and Pesaran (2011) and Tong et al. (2017), it can be found that EV
battery CLSC is still a relatively new topic. Firstly, current research on EV batteries is more focused on technology. Secondly,
the existing CLSC models, such as Bulmus et al. (2014) and Cai et al. (2014), are not able to describe the features of EV
C

battery, e.g., the combined features of reusing and recycling. We envision that, the model developed in this paper fills the
research gap in the EV battery CLSC and EV industry.
AC

Through discussion, the findings can be summarised as: (1) The sorting rate has linear relationship with optimal profit. This
conclusion simplifies the difficulties of enterprise managers to analyse the EV battery supply chain; (2) In the EV battery CLSC,
the more return batteries, the less profit. This finding also explains why recycling is not widely accepted by manufacturers even
though more returns will reduce the consumption of natural resources. Therefore, government should try to take some incentives
to increase the economic benefits of recycling; (3) The more reusable batteries, the more profit in period 2. So, reusable returns
are encouraged. Hence, EV service providers should encourage customers to use the battery carefully; (4) The higher recycling
rate and the lower recycling processing cost, the higher profit. This requires increasing efficiency and technological innovation in
recycling operations; (5) The relationship between high-quality return yield and profit is uncertain but it is still encouraged in order
to increase the sorting rate for the reusable batteries. Therefore, to conclude, this research develops a unique EV battery CLSC
model which reflects the special characteristics of EV battery. All these findings have answered the research questions and will
help
 ACCEPTED MANUSCRIPT

EV battery manufacturers and remanufacturers make better decision in cooperation. And to sustain recycling business,
governments support is vitally important to keep the business going.

7. Limitation and future research

This research has proposed a CLSC model to illustrate the whole life cycle of EV batteries. It takes more consideration on
economics while the environmental influence of EV batteries is less discussed. Therefore, future research will try to quantify the
environmental impact in the CLSC. Moreover, how to determine the balance between the economy and the environment in EV
battery CLSC is a future research direction as well.

8. Reference

T
Ayre, J., 2016. Nissan Reuses EV Batteries For Home Energy Storage xStorage. URL:

IP
https://cleantechnica.com/2016/05/15/nissan-recycles-ev-batterieshome-energy-storage/.

CR
Binks, S., 2016. Proceso de reciclado de baterías. URL:

US
https://www.removemycar.co.uk/battery-recycling.html.

Bulmus, SC, Zhu, SX, Teunter, R., 2014. La competencia por los núcleos en la refabricación. European Journal of Operational
AN
Research 233, 105-113.

Cai, X., Lai, M., Li, X., Li, Y., Wu, X., 2014. adquisición óptima y política de producción en un sistema de fabricación /
M

refabricación híbrido con adquisición núcleo en diferentes niveles de calidad. European Journal of Operational Research 233,

374-382.

Cooper, WD, 1970. técnicas de instrumentación y medición electrónicos. Prentice Hall. Dalton, A., 2016. BMW i3 reutilizar
ED

las baterías de almacenamiento de energía en el hogar. URL:

https://www.engadget.com/2016/06/21/bmw-will-repurpose-i3-batteriesfor-home-energy-storage/.
T
EP

Debo, LG, Toktay, LB, Van Wassenhove, LN, 2005. Segmentación del mercado y la tecnología de productos Selección de productos

remanufacturable. Gestión de la Ciencia 51, 1193-1205. Ferguson, ME, Toktay, LB, 2006. El efecto de la competencia en las
C

estrategias de recuperación. Producción y gestión de operaciones 15, 351-368.

AC

Ferrer, G., Swaminathan, JM, 2010. La gestión de productos remanufacturados nuevos y diferenciados. European Journal of

Operational Research 203, 370-379.

Fraden, J., 2004. Handbook of sensores modernos: la física, diseños y aplicaciones. Springer Ciencia y Business Media.

Fred Lambert, 2017. Tesla ahora asegura la reducción de costes de la batería 35% en Gigafactory 1 'haciendo alusión a un costo avance por

debajo de $ 125 / kWh - Electrek. URL:

 ACEPTADO MANUSCRITO

https://electrek.co/2017/02/18/tesla-battery-cost-gigafactory-model3/.

Gaines, L., 2014. El futuro de la automoción reciclaje de baterías de iones de litio: Trazando un curso sostenible. Materiales y

Tecnologías sostenibles 1, 2-7.

Gaines, L., Singh, M., 1995. Energía y los impactos ambientales de la producción de baterías de vehículos eléctricos y el reciclaje. Informe

Técnico. Documento Técnico SAE.

Geyer, R., Blass, VD, 2010. La economía de la reutilización y el reciclaje del teléfono celular. International Journal of Advanced

T
Manufacturing Technology 47, 515-525.

IP
Heymans, C., Walker, SB, Young, SB, Fowler, M., 2014. Análisis económico de segundo uso baterías de vehículos eléctricos para el

CR
almacenamiento de energía residencial y de nivelación de carga. Política energética 71, 22-30. Hutchinson, A., 2008. es el reciclaje vale la

pena? pm investiga su impacto económico y ambiental. Mecánica Popular.

US
Agencia Internacional de la Energía, 2016. Global EV de Outlook 2016 más de un millón de coches eléctricos. Agencia
AN
Internacional de Energía, 2017. Global EV de Outlook 2017: Dos millones y contando. Publicaciones AIE, 1-71doi: 10.1787 /

9789264278882-en.
M

Kelty, K., 2011. Programa de reciclaje de baterías de ciclo cerrado de Tesla. URL:

https://www.tesla.com/blog/teslas-closed-loop-battery-recyclingprogram.
ED

Rey, AM, Burgess, Carolina del Sur, Ijomah, W., McMahon, CA, 2006. La reducción de los residuos: la reparación, reacondicionamiento,

refabricación o reciclar? Desarrollo Sostenible 14, 257-267.

T

Lih, WC, Yen, JH, Shieh, FH, Liao, YM, 2012. En segundo lugar el uso de jubilados paquetes de baterías de iones de litio de
EP

vehículos eléctricos: desafíos tecnológicos, análisis de costos y modelo de negocio óptimo, en: Computer, los consumidores y control
C

(IS3C ), 2012 Simposio Internacional de IEEE. pp. 381-384. Majumder, P., Groenevelt, H., 2001. La competencia en la refabricación.

Producción y Dirección de Operaciones 10, 125-141.

AC

Marcos, M., 2014. El backgground fot Tesla Model S Batería. URL:

https://xueqiu.com/8940100817/30082286.

McIntire-Estrasburgo, J., 2015. La batería del vehículo eléctrico puede y debe ser reciclado. URL:

https://cleantechnica.com/2015/07/23/electric-vehicle-battery-canrecycled/.

Mitra, S., Webster, S., 2008. La competencia en la refabricación y los efectos de las subvenciones del gobierno. Revista

Internacional de Producción Economía 111, 287-298.

 ACEPTADO MANUSCRITO

Neubauer, J., Pesaran, A., 2011. La capacidad de las estrategias de la batería segundo uso para impactar plug-in precios de los vehículos eléctricos y sirven

aplicaciones de almacenamiento de energía de servicios públicos. Journal of Power Sources 196, 10351-

10358.

NEWBAUER, J., Pesaran, A., 2010. PHEV / EV Li-Ion batería Proyecto de segundo uso, NREL (National Renewable Energy

Laboratory) (cartel). Informe Técnico. National Renewable Energy Laboratory (NREL), Golden, Colorado.

Nissan, 2015. reciclaje de su antiguo batería de coche. URL:

T
https://www.nissan.co.uk/recycle-your-old-car-battery.html.

IP
Nissan, 2018. 2018 Nissan Leaf. URL:

CR
https://www.nissanusa.com/vehicles/electric-cars/leaf/rangecharging.html.

US
Patten, J., Christensen, N., Nola, G., Srivastava, S., 2011. Electric vehicle batterywind storage system, in: Vehicle Power

and Propulsion Conference (VPPC), 2011 IEEE, IEEE. pp. 1–3. Richa, K., Babbitt, C.W., Gaustad, G., Wang, X., 2014. A
AN
future perspective on lithium-ion battery waste flows from electric vehicles. Resources, Conservation and Recycling 83,

63–76. Sedgewick, R., 1988. Algorithms. Pearson Education India.

M

Thomas, V.M., 2003. Product self-management: evolution in recycling and reuse. Environmental science & technology 37,

5297–5302.
ED

Tierney, J., 1996. Recycling is garbage. New York Times Magazine 30, 24–29. Tong, S., Fung, T., Klein, M.P., Weisbach,

D.A., Park, J.W., 2017. Demonstration of reusing electric vehicle battery for solar energy storage and demand side
T

management. Journal of Energy Storage 11, 200–210.

EP

Voelcker, J., 2016. Reusing electric-car batteries: great idea, lots of practical challenges. URL:
C

http://www.greencarreports.com/news/1103363{\_}reusing-electric-carbatteries-great-idea-lots-of-practical-challenges.
AC

Volkswagen, 2016. Battery Recycling. URL: http://www.volkswagen.co.uk/aboutus/environment/battery-recycling.

Will Date, 2015. Battery recycling costs expected to shoot up. URL:

http://www.letsrecycle.com/news/latest-news/battery-collectioncosts-expected-to-shoot-up/.

Yu, H.J., Zhang, T.Z., Yuan, J., Li, C.D., Li, J.M., 2013. Trial study on ev battery recycling standardization development, in:

Advanced Materials Research, Trans Tech Publ. pp. 2170–2173.

 ACCEPTED MANUSCRIPT

Supplementary Material

Appendix A. Optimal values calculation and proof for period 2

By substituting Eq. 9, and supposing A Hh q=
2
θα (1 − β ) , A H
l
=q 2
θ ( 1 − α ) and also
EV 1 EV 1

A A=A
2 2 h
+ 2l
, las funciones de beneficios para el nuevo fabricante de la batería ( P n),
2
baja calidad y alto

reconstructor batería de la calidad en el periodo 2, es decir, P l2 y P h,2 son

P 2n
= ( pcqq
-
2
NTR ) 2 n

= 2 n
(- k A
( q 2
+- 2 n
Un2 hh
s
2
- Un2 ls2 l
) - +c NTR
dm )

T
P
••••••••2 l
= (l l
- - = personal
ordenador Un )

IP
2 l 2 l 2l

(A.1)
••••• ( 1 - s 2 l ) ( l ll d(
cuadrados - kA
( q2 + - Un2 2hh
s - Un2 ls2 ) ) - -c s2l )
2 l m 2 n l l

CR
W 2h
= ( h
- - = personal
ordenador
2
cuadrados
h 2 h
) 2 h

la 2 h (1 - s2 h ) ( mh s( - kA
( q2 + - Un2 hh
s - Un2 ls2 ) ) - -c s2h )

US
m 2 n 2 l h

En resumen, el beneficio total en el periodo 2 podría ser

OO =O O 2 n + +
AN
2 2 l 2h

• q ( S M-S- k A
( s - Un2 ls2 + Un +q - ) c NTR ) •
2 hh
2 2n
•
2 n m l 2
•
(A.2)
s2h )•
M

= +• la 2 h ( 1 1- s2 h ) ( M s( - -k A
( s 2 hh
2
- Un2 ls2 + Un q - -2 n ) )
+ c
•
h m l 2
• h

• •
+ la 2 l ( - s2 l )( ( - -k A
( s - Un2 ls2 + Un q - -2 n ) )
+ c s2l )
• l m 2 hh
2 l 2 l •

La matriz de Hesse para II (Ec.

2
A.2) es
ED

• 0 kA 2 h
• kA 2 l
• •
H (P ) = kA - 2 Un kA
( l - 1) - Un kA 2 ( l l + l ) (A.3)
2
• 2 h 2 h 2 hh 2 h
• l h

• kA - Un kA ( l lh + - ) ( 2 l l - 1 ) ••
2 Un kA
T

• 2 l h 2 2 l l 2 l l
EP

tenemos - 2 Unh kA
( l 2 2 hh
->1y)0 - 2 Unl kA
( l
2 2l l
->.1 Por
)0 lo tanto, H ( P 2 ) es positiva
semidefinida y la existencia y unicidad de equilibrio Nash pueden ser probadas. El beneficio total óptimo se logra mediante el
II II II 0∂
C

∂ ∂
uso condición de primer orden, es decir
2
=
2
=
2
= :
∂q ∂s ∂s
AC

2 n 2 h 2l

••••
- -Un( ks 2 h
* * *
)(
*
•
h )
∂P
•
2 h
- s2AA
l
q+
2 l 2
+- 2 n
k 1 - s2Un 2h
l h
2
= • • = 0
- -k (1
* *
•
l )
∂q • s Un l - -c NTR kq 2 n + d m
•
2 n
2l
• 2 l

••••••
•
(m s( - - Un( ks )) ) ••
* * * *
∂P
- la 2 l l m 2 h
- s AA2 ql + 2
+ - -2 n c s2l
2 • 2 h 2 l
(A.4)
= 0
l

=
∂ s2 h • •
- 1) •
* * *
• + -k
•
(1 h )
s AA 2 2 h 2 lh
l + Un kq
2 n
2 l
+ -(1 l )
s 2Un kAl ( 2 2l
l l •

) •• =
•••• •
(m s( - - Un( ks 2 h ))
* * * *
P - Un2 h h m 2 h
- s2AA
l
q+
2 l 2
+-- c s2 h
•
∂ = ∂2
2 n h
0
s2l • •
- 1) •
* * *

•
• + -k
•
(1 l )
s AA 2 2 h 2 l
l l
+ Un 2kq
n 2 h
+ (1 - s2Un) kAh ( h 2 2h
l h •

Las expresiones para valores óptimos son

 ACEPTADO MANUSCRITO

k (A l - +cch+ + 1) 2 ( kA 2 l + d m ) ) •
2
•
2 l l
( kA 2 h
( NTR
• •
+ la 2 l l ( kA ((l cch + - 2 c NTR + 2) + -cc
-h 2)
• l 2h h l l •
• - 2( c kA +- 2 c NTR + d lm + 1)) •
2l l
• •
• - +cch- + l
2) 2 ( kA 2h h
- -1) 2 (2 ch - c NTR + d m
) 2+ mhm
s )•
•
- 2( Un c( +l - +1) Un2 l c( - -1)) 2 ( Un2 kA (l - 1) l •
2h h
•
2 h h h l h
•
•
+ kA (l - 1) l + 2)) •
* • 2l l l •
q = (A.5)
2
2 k( kA (l - +1) kA 2 l ( l - +1) 4)
2n 2

2 h h l

c kA (l ch (l l - +1) ( l c NTR ( - l )•
2
• (- - +1) - 1) (

T
2 h l h h l l h
• •
+ c NTR + l + 1) (+ l c NTR ( l - +1) - l 3)) •

IP
• h h h h

• + c NTR ( l - +( -c l
+ 1) 2 l 1) d + 2) •
* • l l m •
s = (A.6)

CR
kA 2 l ( l
2
- +1) - +1) 4)
2l 2
2 ( kA 2 h
(l h l

c kA (l cl (l l - -1) (( l l + c NTR + 3) •
2
•
2
(- l h l
- +1) h
- 1) ( l l
c NTR - 1) h
•

US
•
+ ( c NTR + 1) (l + c NTR + 1) l
2
• h l
) •
• - 2 A 2k (l - -1)4 c h + -2 c NTR 2d +4 •
* • h m •
s = (A.7)
AN
kA 2 l ( l
2
- +1) - +1) 4)
2
(l
2h
2 ( kA 2 h h l

* *
h ( )
q = la 2 1 - s (A.8)
M

2 h 2h

* *
l (
q = la 2 1 - s ) (A.9)
2 l l 2

Apéndice B. óptima valores de cálculo y la prueba en el período 3

Para simplificar, vamos AHl q= yo (1 - una) , AHh q= Iowa (1 - b ) y AHu q= i ab y
ED

3 EV 2
3 EV 2
3 EV 2

AAAA
=+3 3l 3 h 3u
, entonces la ecuación. 18 se puede reescribir como:
T

• =q la 3 l (1 - s3l )
3l
•• =
EP

••q= 3h
la 3 h (1 - s3 h ) (B.1)

•• q 3u
la 3 u (1 - s3 u )
C

Los beneficios para las baterías hechas de retornos reciclados o reutilizados son los ingresos menos todos los costos. Suponiendo la ecuación. B.1, los
AC

beneficios son:
 ACEPTADO MANUSCRITO

P 3 n
= ( PCQ
-
3 NTR ) 3n

= (d l m
- kq( 3 n
+ - AA
3
s 3 l 3l
- Un3 hs
3 h
- Un3 us
3u
) - cq ) NTR 3n

P 3 l
= l l QP
3 l
- •(+c Un ) 3l
3 l 3l

1- s
cuadrados ( ) (d - kq( + - AA s - Un3 hs - Un3 us ))•
•
l 3 l m 3 n 3 3 l 3u
• 3l 3l 3 h

=•
••
• - +( c ) ( 3l 1 - s
s Un )
• l 3l 3l •

P 3h
= l lh QP - +( c- Un )
3 h 3 h 3 h 3h

T
• 1 s
cuadrados ( ) (d - kq( + - AA s - Un3 hs - la 3 u s ))•

IP
•
h 3 h m 3 n 3 3 l 3h 3u
• 3 h 3l

=
• ••
• - +( c ) ( 3h 1 - s
s Un ) •

CR
• h 3 h 3h

P 3u
= l lu QP - •(+c u Un )
3u 3 3u 3u

) (d ))•

US
1- s
cuadrados ( - kq( + - AA s - Un3 hs - Un3 us
•
u 3 u m 3 n 3 3 l 3u
• 3u 3l 3 h

=•
••
• - +( c u ) ( 3u 1 - s
s Un ) (B.2)
•
••••••••••••••••••••••••• 3u 3u •
AN
El beneficio entero es un beneficio para la suma del fabricante / remanufactureros:

P O=O O O+ + + 3u
M

3 3 n 3 l 3 h

• Un qk
( s(
3 n 3 hh
3
+ Un3 l3s l
+ Un3 uu
3
s - - Un q 3 3n
• ) - +c NTR
dm )
• •
• + Un
3 h
s( 3 h
- 1 ) (c + s3 h )+ Un
3 l
s( 3l
- 1 ) ( c + s3l ) •
•
h l
•
• + Un s( - 1 )1 (1c u + s3u ) • (B.3)
=- 3 u 3 u

• •
la 3 hhl ( s 3 h -1 ) ( k A s3 hh + Un3 l3s + -Un-3 uu Un q 3 ) + d m )
ED

s
• 3 l 3 • 3 n

• - la 3 l l ( s3 l - ) ( k A s3 hh
3
+ Un3 l3s + -Un-3 uu
3
s Un q 3 ) + d m ) •
•
l l 3 n
•
• - Unu s( - )l (k A + Un3 l3s + -Un-3 uu Un q + d ) •
T

• 3 3u u ( 3 hh
3
s l 3
s 3 ) m • 3 n
EP

In period 3, much like period 2, the Hessian matrix for Π ,3 i.e., Eq. B.3, is

H (Π 3) =
C

• 0 kA (1 + λ h ) kA (1 + λ l ) kA 3 u ( 1 + λ u ) •
AC

3 h 3 l

• •
•
kA 3 h
(1 + λ h ) − 2 A kA
h
( 3 3 hh
λ − 1) − kA Ah 3 3 l
( λ hλ + − l
) kA Ah 3 3u ( λ hλ + u ) • (B.4)
• kA (1 + λ l ) − kA Ah ( λ hλ + − ) 2 A kA( λ − 1) − kA Al ( λ lλ + ) •••
3 l 3 3 l l l 3 3 l l 3 3u u
•••
kA 3 u
(1 + λ u ) − kA Ah 3 3 u
( λ hλ + − u
) kA Al 3 3 u
( λ lλ + − u
) 2 A kA
u
(3
λ
3u u
1−) •

We have −2 A kA
h
(λ
3 3hh
−1
>)0, − 2 A kA
l
(λ 3 3l l
−1
> )0
and −2 A kA
3u
(λ 3u u
−1
>)0. Therefore

H ( Π 3 ) is positive-semidefinite and we have existence and uniqueness of Nash equilibrium in this

 ACCEPTED MANUSCRIPT

period. By using first-order condition to acquire the optimal profit for each agent:

• ksA
•••••••••• + s A s3 l A+ A q A3 h hλ •
* * * * *
( 3 h 3 u
−− 3 3 n ) +( −k s 1)
•
3 h 3l 3u 3h
•= 0
•
+ ks 3 l
*
− 1) A 3 l λ + −k s
(
*
1) A 3 u uλ −−
c ntr kq 3 n
*
+ δm •
• l 3u •
•
• A c( ( k s( A + s A s3 l A+ A− q− ) + δ ) + kq A
* * * * * *
+ −s ) A3l λ 3l
•=
3l l 3 h 3 u 3 3 n m 3 n
• l 3l 3 h 3l 3u
0
• ••
AA − 1) A3l λl − A3A λ
* * * *
k s( − 1) k s( k s( − 1) + −( s 1) A3 l
2
• − λ −
• 3 h
3 h 3 lh
3l 3u
l 3 uu 3l •
(B.5)
•
• Ac( ( k s( A + s A s3 l A+A− q− ) + δ ) + kq A
* * * * * *
••• + −s ) A 3 h λh 3h
•= 0
3 h 3 h 3u 3 3 n m 3 n
• h 3 h 3 h 3l 3u

• ••
AA AA λ
* * * *
k s( − 1) k s( − 1) A3hλ k s( − 1) + −( s 1 ) A3 h
2
• − λ

T
− −
• 3l
3 h 3 l l 3 h
h 3u
3 h 3 uu 3h •
••••••

IP
•
• Ac( A 3 u uλ ( k s( A + s A s3 l A+A− q− ) + δ ) + kq A
* * * * *
)
*
+ −s u 3n 3u
•= 0
3u u 3 h 3 3 m 3 n
• 3u 3 h 3l 3u

••

CR
•
AA − 1 ) A 3 u uλ
* * * *
k s( − 1) k s( − 1) A3A3 − k s( + −( s 1 ) A3 u
2
• − λ − λ
• 3 h
3 h 3 uh 3l
l ul 3u 3u •

Through solving the above equations, the optimal values in period 3 are

US
2
• k (A kAh ( 3 3l
(λ h
( − c+l c+ + ntr
1) ( ( λ hλ l
c hc + − l
2 c ntr + −2)+ − c hc l
2) •
• •
+ λ lλ( ( − c+hc+ + −ntr− + 1 ) c hc 2 )) (λu ( − c+hc+
2
• l l
kA 3u ntr
1) •
•
AN
•
+ λ uλ( ( h c h − 2 c ntr + cu + 2) + c hc − − u+ 2) ( ( λ hλ c ntr − +c u 1)
• h •
•
− c+hc− + A3ch ( + −λ h
•
2)) 2 λ ( − c+ 2 c ntr + −( λ 1)δ ) ) 2(− 1)
•
u h h h m h
•
M

))•
• 2
+ A3cl ( − 1+ ) A cu ( − 1+ ) A 3 u uλ ( kA ( − c+l c+ + ntr
1 )2( kA 3 u + δ m
•
l 3 u 3 l
•
• + A 3 l λ l ( kA 3 u
( λ( l c ntr − +c u− +1)− + c l c u
2) 2 ( kA 3 l λ l − 1) •
• •
•
− 2 (cl − 2 c ntr + δ m
+ 1 )2+ λ δl m ) + A 3 u uλ ( kA 3l
(λ l
(cl − 2 c ntr + cu + 2 ) •
• 2 •
+ c l c− − u− 2) 2 ( kA − 2 c ntr + cu + δ + 1 )) 2+ kA 3 h ( λ − 1) λ h
• m
•
ED

3 u h

•
A kA + kA 3 l ( λ l •
− 2 ( (λ − 1) λ − 1)λ + kA 3 u ( λ u − 1 )λu + 2)
•
3 3 h h h l
•
•
c ntr + 4 δ2 m( •
• − 4 •
T

*
q3n = 2
k kA − 1)
+ − 1)
+ kA 3 u ( λ u − 1)
+ 4)
2 2
(λ kA (λ
EP

3 h h 3 l l

(B.6)

2
• kA 3ch (λ − 1)
+ ch (λ − 1)( λ − 1)
+( λ c ntr ( − λ+ h+) + c ntr λh 1) •
C

(− l h h l l
• 2 •
+ λ h ( c ntr ( λ − 1+)− +λ 3) kA cu ntr λ l + kA λ ( − c+l c+ 1) •
AC

• h h 3 3u u ntr

• •
+ kA λ ( − c ntr ( λ + 1 ) + cu (λ l −1
+ )2+ − −c l λ 3) kA cu u l λ − kA 3cu l
•
3 uu l l 3
•
2
• + kA 3cu u + 2 kA 3 l
λ l
− 2 kA 3 l
λ l
− 2 A 3k ( λ l − 1+ ) kA 3u l
λ + 2 c ntr λ l − 4 c l •
•
•
••
* • + 2 c ntr + 2λδ
lm
− 2δ m
+ 4 •
s = (B.7)
2
− 1)
+ − 1)
+ kA 3 u ( λ u − 1)
+ 4)
2 2
2( kA kA (λ
3l

3 h
(λ h 3 l l
 ACCEPTED MANUSCRIPT
2
• kA 3cl ( − h
(λ l
− 1)
+ cl (λ h
•
− 1)( λ l
− 1)
− (( λ l c ntr − 1) λ h + c ntr + 3)
• • 2
+ ( c ntr + 1 ) λ h + ( c ntr + 1 ) λ l ) + kA 3cu ntr h λ + kA 3 u uλ ( − c+hc+ ntr
2

•
1) •
• + kA 3 u uλ ( − c ntr ( λ h + 1 ) + c u ( λ h − 1+ )2 ch +− λ h− 3 ) kA 3cu u h λ •
• •
2
• − kA 3cu hkA +c 3 uu
+ 2 kA λ
3 hh
− 2 kA λ
3 hh
− 2 A k
3
( λ h
− +
1 ) kA λ
3u h
•
• •
* • + 2 c ntr λ h − 4 c h + 2 c ntr + −2 λ δh m 2δm + 4 •
s = (B.8)
2
− 1)
+ − 1)
+ kA 3 u ( λ u − 1)
+ 4)
2 2
(λ kA (λ
3h
2( kA 3 h h 3 l l

• λ u ( k( A 3 h
( λ( h c hc − + ntr− + +1)+ c hc ntr
1) ( ( A 3 l λ l
c l c− + ntr− 1) cl •
• 2 •
+ c ntr + −1) 2 A3u − 2 A) 23 + ( c ntr + δ ) ) (+ k A (λ ( c ntr − +c u 1)
•

T
• m 3 h h

( c hc + − ) + −c h+
c (λl •
23
• − λ 2 cu + 3 ) A ( c ntr − +c u 1)

IP
h ntr u l
• 2 •
•
− λ l
( c l c+ − ntr
2 cu + 3 ) + −c +
c l u
) 2 A) 23 + kA 3 u uλ •

CR
• ••
* •
+ 2 ( c ntr − 2 cu − +
δm 2)
s = (B.9)
2
− 1)
+ − 1)
+ kA 3 u ( λ u − 1)
+ 4)
2 2
2( kA (λ kA (λ
3u

3 h h 3 l l

US
q A=
* *
3 l 3 1l ( −s
3l
) (B.10)

* *
h ( )
q = A31 −s (B.11)
AN
3 h 3h

* *
u ( )
q = A31 −s (B.12)
3 u 3u
M

Appendix C. Proof for relationships between parameters and total optimal profits in
Period 2
Relationship between θ and 2 Π
The total optimal profit in period 2 can be rewritten as
ED

2
N 10N + θ + N θ
Π 2 (θ) =
11 12
(C.1)
4k ( − 4+ D 11 θ )
T

N 10 7
EP

0 ≤ ≤ 0
−
1.23 10
× ≤ N 11
≤ 24
•••• (C.2)
0 N
≤D ≤ 0
C

••• 12
−7
−1 ≤ ≤ − 2.86 10
×
AC

11

Expressions of N 10, N 11, N 12, and 11 D are shown from Eq. C.4 to Eq. C.7 in Appendix. Based on
the range above, we have

N 12≈0, N 10≈0and Lemma 1, N α

(i) Because of Π 2 (θ) =
11

4k ( − 4+ D 11 α )

(ii) By using grid search, the range of first order of Π on

2
θ is

(θ) (θ)
−8
− 6 / 4( k ) ≤ Π ′2 ≤ − 8.53 / 4( k ) × 10 , also Π ′2 < .0Therefore, the relationship is
decremented.

(iii) For simplification, we assume the fitted line crosses ( a1b1, and
) ( a b, , where
2 2 )
10 a=,
b1 = Π 2
( a) 1
, a 2=1 , b2 = Π 2 ( a 2 ) . And the fitted straight line is
 ACCEPTED MANUSCRIPT

b2 b− b2 b−b •
Π 2
(α) =
1
α•+−• 2
1
a2 •
a 2a− 1 • a 2a− 1 • ( C.3)

Through using gird search and based on the assumption above, the linearity range is
− 5.57% ≤ η 2 θ ≤ .0So, θ and Π can
2
be treated as linear relationship.

2 ( )
Therefore, the relationship between Πθ and θ is negative linear.

2
N 10
= − 4( c ntr − δ m ) (C.4)

2
• 4( Hkq EV 1
) ( − αβ
+ −α αβ( •
)c h + −α β
( 1) c( h(c ntr
λ h + 1)

T
• •
+ −( λ h 1)δ m − + 2) ( α c ntr ( β β+ λ −( λ δ δ1) h ( − h m + m + 1) ) •
•

IP
• 2 •
N 11 = − (−β 1) c ntr hλ δ +β−λ mβ ( ( ) ) (− −α 1) (c l − −α 1) c ntr (C.5)
2
1) h −
• •

CR
2
• · λ lδm −− ( α λ 1) l ( ( c ntr − 1 ) c ntr − ( c ntr + 1 ) δ m ) + −( α 1) c l •
• •
• (
• · c
ntr ( λ l + 1 ) (+ − λ l 1)δ m − − 2 )+ + c ntr δ m 1) ••

US
2 2
N 12
( α α β1 ) (
=− − 1) ( Hkq EV 1 ) ( − c l ( λ h − 1+) ch ( λ l − 1+) ( c ntr − 1 ) ( λ hλ − l )) (C.6)

2
D 11Hkq
= ( ) (αβ α+β− ( 1 ) (λ − 2 ) λ hα+ −( 1)λ − 2 (α − 1 ) λ l − 1) (C.7)
AN
EV 1 h l

Relationship between α and 2 Π

M

2
N 20N + α + N 22 α
Π 2 (α) =
21
(C.8)
4k D
( D20 + 21
α )

− 4≤ NN
20
≤ 20
ED

− 24
≤ ND 21
≤ 21 0
− 33
D22 ≤ −
≤ ••••• 2.27 10
× (C.9)

− 5≤
T

20
≤− 4

••••− 1≤ 1
EP

21
≤

We find The first derivative − 5.25 / 4( k ) (≤ ) Π ′2 α ≤ 6.25 / 4( k ) and the linearity is

C

− 5.57% ≤ η 2α ≤ 5.57% . Therefore, the relationship between α and Π can

2
be treated as linear.
AC

4( θ( Hkq )c l ) (c l c + 1) (+ − λ l δ m − 2) •
2
•
EV 1
− θ ( Hkq EV 1 ntr
(λ l
1)
• •
N 20
= +• c ntr (δ θ (
m
( Hkq EV 1
) (λ l − 1 )λ l
+ −2 )( θ Hkq EV 1 ) ( λ l + 1) ) • (C.10)
• •
• + c ntr (θ( Hkq ) λ l − 1)
+( ) (− ( δ mθ Hkq − 1)
+ δ m ) ) ••
2
θ Hkq )(λ
• EV 1 EV 1 EV 1 l
 ACCEPTED MANUSCRIPT

2
θ ( Hkq
••••••••••••
EV 1
) ( −4 −β−β ( 1) c( h( θ Hkq EV 1
) (+ θ Hkq EV 1
)(λ l
− 2) λ l + 4) ••••••••••••

+cl (− − ) λ hβθ− ) (+ θ Hkq ) − 4)

2
(β 1 ) θ( Hkq EV 1
) (λ h − 2 ( Hkq EV 1 EV 1

+ −2( β 1) c( h(Hkq
θ EV 1
) (c l λ h
− 1)( λ l
− 1)
+( λ h
c ntr (θ( Hkq EV 1
)

−θ ( Hkq EV 1
)λ l + 2) ( +θ Hkq EV 1
)(λ l −1
+ )2 ) δ m
+ θ ( Hkq EV 1
) ( c ntr − 1 )
2
·( λ l − 1 )λ + l
2 ( c ntr − −
δ m+ 2 ) 2 (( λ lβ θ − 1) ( Hkq EV 1
) ( c ntr − 1 ) λ h

+ 2 ( c ntr ( − c+ntr+ + δ m
1 ) δ m ) ) 2+ ( c c l ntr
(( β θ− 1) ( Hkq EV 1
) 1( λ h − )
N 21 = (C.11)
·( λ hλ − + l ) ( 2 λ l
+ 1 )) (+ − β θ 1) ( ( − Hkq EV 1
) ) 1( λ h
− ) ( λ hλ − l )
•

T
• 2
+2 (λ l − −4) ) c ntr hλ + ) c ntr hλ
2 2

• − 1 )δ m
βθ ( Hkq EV 1
2 βθ ( Hkq EV 1 •

IP
• 2 2 •
+ θ ( Hkq ) c ntr hλ β λc ntrδ h m + 4 β λc ntr
δh m
2 2

• EV 1
− 2θ ( Hkq EV 1 ) c ntr hλ − 4
•
•

CR
• 2
+ 4 c ntr hλm δ β λc ntr h + β λc ntr h + − 4 c ntr hλ
2 2
− 4 c ntr hλm δ − 4 4 4 c ntr hλ
• •
2
+ λ lβ ((θ 1) ( ( − Hkq ) ) ( c ntr − 1)
− 4 ) 4+ β c ntr •
2
• − c ntr mδ
EV 1
• •
• 2 ••
) λ hθ + Hkq βδ

US
+ βλ δh m − )
2

• − βθ ( Hkq EV 1
( EV 1
λ
)4 h
− 4 λδ
hm
4 m

2
= −( β θ − 1)
+( − 1)
+ ( c ntr λ λ− ))
2 2
N 22
1) ( Hkq EV 1
) ( − cl (λ h
ch λ l
− 1)( h l
(C.12)
AN
D 20
θ (
=− ( Hkq EV 1
) ) −( θ Hkq EV 1
) (λ l − 2 )λl − 4 (C.13)

D = θ ( Hkq ) ( β β+ −( 1) (λ − 2 )λh + −( λ 2)λl ) (C.14)

M

21 EV 1 h l

Relationship between β and 2 Π

N 30N k+ D D31 β
Π 2 ( β) = (C.15)
4 ( + β )
ED

30 31

− 4≤ N 30
≤ 20
−7
T

− 24
≤ N 31
≤ − 1.64 10
×
(C.16)
EP

− ≤5 •••• D ≤−4
••• 30
−7
3.29 10
× ≤ D 31
≤1
C

9.22 10
× /4 ) (≤ ) Π ′2 β ≤ 6.25 / 4( k ) and the linearity is
− 8
We find The first derivative ( k
AC

− 10
7.44 10
× ≤ η 2β ≤ 5.57% . Therefore, the relationship between β and Π can
2
be treated as positive
 ACCEPTED MANUSCRIPT
2
•
22 2 22 2 2
• − αθ ) + 2α θ ) c ntr hλ αθ+ ( Hkq )
2 2 2
( Hkq EV
c ntr hλ ( Hkq EV EV 1
•
1 1
•
2
) c ntr hλ αθ+ c hHkq ) (θ αθ
− ) •
2 2 2 2 2
• ⋅c
ntr h
λ − 2 αθ ( Hkq EV 1
( Hkq EV 1
) (( EV 1
• •
− (−α θ 1) ( Hkq 1 ) θ(
2
• EV 1
)(λ l
− 2) λ l
+ −4)− ( α Hkq EV 1 ) cl •
• •
⋅ ( αθ ( Hkq )(λ − 2) λ hαθ+ ( Hkq ) + 4) 2+ ( αθ Hkq )ch
• EV 1 h EV 1 EV 1
•
• ⋅ (( α θ− 1) ( Hkq ) (c l λ − 1)( λ − 1)
+ λ ( c ntr (( α θ− 1) ( Hkq ) •
EV 1 h l h EV 1
• •
• − (−α θ 1) ( Hkq EV 1
) λ l − +2)−( α θ 1) ( Hkq EV 1
)(λ l
− 1)
− 2) δm •
• •
+ −( α θ 1) ( Hkq ) ( c ntr − 1)( λ − 1) 2λ − c ntr + 2δ − (α
+ 4) 2+ 1 )
• EV 1 l l m
•
N 30 =
• ⋅θ ( Hkq ) (λ lαθ ( Hkq ) ( c ntr − 1)
2
λ + 2( c ntr ( − c+ntr+ δm 1) + δ m ) ) •

T
EV EV 1
•
1 h
•
• + −2( α θ 1) ( Hkq ) (c l c ( αθ ( Hkq )(λ − 1)( λ hλ − + l λ l + 1)) •

IP
EV 1 ntr EV 1 h
) 2(
• •
αθ
+− ( ( Hkq )) ( λ − 1)( λ hλ − + l ) 2( λ − 1) δ − +4) 4 αθ ( Hkq cEV ntr1 )
• EV 1 h l m •

CR
•
⋅λ δ
2
− 4 αθ ( Hkq ) c ntr hλmδ + 4 αθ ( Hkq
2
) c ntr hλ αθ− 4 ( Hkq ) c ntr hλ •
hm
•
EV 1 EV 1 EV 1
•
• − (−α θ 1) ( Hkq EV 1
) (λ lαθ
2
( Hkq EV 1
) ( c ntr − 1)
+ 4
2
c ntr mδ ) 4− ( θ Hkq EV 1
) c ntr •
• •

US
22 2 2 2
+ 8 c ntr mδ αθ ) λ αθ+ ) λ − 4 αθ ( Hkq ) λ hδm •
2 2 2
• − 4 c ntr − ( Hkq EV 1 h
( Hkq EV 1 h EV 1
• •
linear. •
+ 4 θ ( Hkq )+ 4θ ( Hkq )δ − 4δ m
2 •
• EV 1 EV 1 m •
AN
(C.17)

αθ ( Hkq ) ( c((h α θ− 1) ( Hkq α θ 1) ( Hkq λ l − 4) •

2
•
EV 1 EV 1
) (+ − EV 1
)(λ l
− 2)
M

••••• •••••••••••••
2
+ −( α θ 1) ( Hkq + 2 ( ( c h − −α θ 1) ( Hkq
− 1) ) (c l λ h − 1)
2

EV 1
) (c l λ h EV 1

⋅( λ l
− 1)
+ λ hα((θ − 1) ( Hkq EV 1
) ( c ntr − 1) λ l
+ c ntrHkq
(( EV 1
) ( θ αθ
− ) + 2)

+ −( α θ 1) ( Hkq EV 1
) + 2 δ) m −−
(α 1 ) θ( Hkq EV 1 ) ( c ntr − 1 ) ( λ l − 1) λ l
N 31 = •+ (C.18)
•• − 2( c ntr −−
δ m− 2)) 2 1 ( α − ) θ( Hkq EV 1 ) c( l c ntr
− 1) (λ h
− 1 ) ( λ hλ − l )
ED

2
α θ− 1) ( Hkq λ hλl λ λ+α θh ( ((h 1) ( Hkq ) ( c ntr − 1)
2

•• − 2( EV 1
) ( c ntr − 1) − EV 1

2 2
4 c ntr mδ +
) 4( c ntr ( − c+ntr+ + δ m
1) δ m
) ) 1+ −(( α ) θ Hkq EV
) ( c ntr − 1) λ l
•• + 1
T

4( c ntr − −
δ m• 1))
EP

2
D 30
= θ ( Hkq EV 1 ) ( − αλ h
+ 2 αλ αh + −( 1) (λ l
− 2 ) λ θ) + − ( ( Hkq
l EV 1 )) − 4 (C.19)
C

2
D 31
= αθ ( Hkq EV 1
) (λ h − 1) (C.20)
AC

Relationship between l λ and 2 Π

2
N 40N + λ + N 42 λ l
Π 2
( λ)l =
41 l

2
(C.21)
4 ( D40 +
kD 41
λ l
+ D 42 λ l )
 ACCEPTED MANUSCRIPT

− 4≤ N 40
≤ 20
−8
− 16
≤ N 41
≤ − 5.69 10
×
−8
6.11 10
× ≤ N 42
≤ 4
(C.22)
− 5≤ D 40
≤−4
−8
8.69 10
× ≤ D 41
≤ 2
− ••••••••••• −8
1 ≤ D 42
×
≤ − 9.08 10

8.68 10
× /4 ) (≤ )Π ′2 λ l ≤ 2.33 / 4( k )
− 8
We find The first derivative ( k and the linearity is

0% ≤ η 2λl ≤ 24.72% . Therefore, the relationship between l λ and Π can be treated as positive.

T
2

α β( ) c( h ( α θ− 1 ) ( 4 )−( α θ 1 ) (
2
)−+ ) cl
2
•••••••• − 1 ) θ( Hkq Hkq Hkq •••••••••••

IP
EV 1 EV 1 EV 1

⋅ (α( β θ − 1) ( Hkq EV 1
) (λ h
− 2) λ α+β−θ (
h
1) ( Hkq EV 1
) − 4)

CR
+ 2α( β θ − 1) ( Hkq EV 1
) c((h α θ− 1) ( Hkq EV 1
)c
(l λ h − 1)

+ c ntr ( λ αθ
(− h
( Hkq EV 1
) (+ θ Hkq EV 1
) + 2 )2+ ) + −( α θ 1) ( Hkq EV 1 )λh

US
N 40
=+ 2( λ h
− 1) δ m
4) 2 1 ( α θ− ) (
−+ Hkq EV 1 ) c c( l n tr
(2 − α β( − 1)θ (C.23)
•⋅
••+( Hkq EV 1
) (λ h
− 1) λ) α
( β+ −θh
1) ( Hkq EV 1
) (λ h
− 1) 2λ h
− δ m − 4)
AN
α β( θ − 1) ( Hkq )λ
(( hα θ
2
1) ( Hkq − 1)
− 4 c ntr mδ )
2
− ) ( c ntr
• EV 1 EV 1 •
•
− 4α β( •
− 1 ) θ( Hkq ) λ( h ( c ntr − 1 ) c ntr − ( c ntr + 1 ) δ m )
•
EV 1
•
M

• ••
+ 4 ( c ntr − δ ) ( − c+ntr θ ( Hkq ) ( αβ − 1+) δ )− 4θ ( Hkq ) ( αβ − 1 )
• m EV 1 m EV 1

2
2( α θ− 1) ( Hkq EV 1
) (α( β − 1) θ( ( − Hkq EV 1
) ) c h + c l ( α( β − 1) θ •
• •
•
⋅( Hkq EV 1
) ( c ntr − 1) λ h
+ −c ntr (2 α β( − 1) θ( Hkq EV 1
)) •
N 41 = • + α β( 1) θ( Hkq ) + −2 δ) m α β( 1 )(θ Hkq ) c hc( ( λ h − 1) • (C.24)
ED

− EV 1
− EV 1 l
• •
2
• − (c
ntr
− 1)( λ h
+ −1)) α β( − 1 )(θ Hkq EV 1
) ( c ntr − 1 ) λ h •
• •
+ 2( c ntr ( − c+ntr+ + ))
δ m• 1) δm •
T

•
EP

( α θ 1) ( Hkq ) (α( β θ − 1) ( Hkq ) ( c h c−

2
N 42
=− EV 1 EV 1
+ntr− 1) 4 c ntr mδ ) (C.25)

D 40
= α β( − 1 ) θ( Hkq EV 1
) (λ h
− 2 ) λ θ+ h
( Hkq EV 1 ) ( αβ − 1
− )4 (C.26)
C

D = − 2 ( α − 1 ) θ( Hkq ) (C.27)
AC

41 EV 1

D 42
= −( α 1 )θ
( Hkq EV 1 ) (C.28)

Relationship between h λ and 2 Π

2
N 50N + λ + N 52 λ h
Π 2
(λ h ) =
51 h
(C.29)
2
4 ( D50 +
kD 51
λ h
+ D 52 λ h )
 ACCEPTED MANUSCRIPT

− 4≤ N 50
≤ 20
−8
− 16
≤ N 51
≤ − 7.66 10
×
−8
6.92 10
× ≤ N 52
≤ 4
(C.30)
− 5≤ D 50
≤−4
−8
9.10 10
× ≤ D 51
≤ 2
− ••••••••••• −8
1 ≤ D 52
×
≤ − 8.87 10

9.04 10
× /4 ′2 ( λ h ) ≤ 2.33 / 4( k ) and the linearity is
− 8
We find The first derivative ( k ) ≤ Π

− 24.25% ≤ η 2λ ≤ 24.25% . Therefore, the relationship between h λ and Π can be treated as

T
h 2

positive.

IP
α β( 1) θ( Hkq α θ− 1) ( Hkq • ) (+ − α θ 1) ( Hkq )
2
• − ch
) ((
EV 1 EV 1 EV 1
•

CR
•
• ⋅ ( λ l − 2) λ l − + 4) 2 α β( − 1 ) θ( Hkq EV 1 ) c( h ( α − 1 )(θ Hkq EV 1 ) •
• •
• · ( λ l − 1 ) ( c l − ( c ntr − 1 ) λ l ) + α θ 1) ( Hkq EV 1 ) c l
2
( 2 c ntr − −δ m+ − 2 ) ( •

US
• 2 •
⋅ ( α( β − 1) θ( Hkq EV 1 ) − + 4)−( α θ 1) ( Hkq EV 1 ) λ( l(α β − 1) θ( Hkq EV 1 ) •
• (C.31)
N =
•
50
• 2
·( c ntr − 1−) 4 c ntr mδ ) 2(+ − α θ 1) ( Hkq EV 1 ) (c l c ntr ( λ(2l − αβ ( − 1)
• •
AN
• ⋅θ ( Hkq ) ) + 2) ( + −α β 1) θ( Hkq ) λ
2( + − λ 1) δ − 4) •
EV 1 EV 1 l l m
• •
• − 4 ( α − 1 ) θ( Hkq EV 1 ) λ( l ( c ntr − 1 ) c ntr − ( c ntr + 1 ) δ m ) +( −4 c ntr δm ) •
M

•⋅−+
• ( c ntr θ ( Hkq EV 1 ) ( αβ − 1) + − ( θ Hkq EV 1 ) ( αβ − 1)
δ m )4 ••

• 2 α( β θ − 1) ( Hkq EV 1
) (c((h α θ− 1) ( Hkq EV 1
) (λ( l − c+l c− ntr
1) •
•
+ −c l c ) 2+ c ntr + −( α θ 1) ( Hkq ) + 2 δ) m( + − α θ 1)
N • = •⋅ ntr EV 1
(C.32)
51
( ( Hkq ) ) (c l α θ 1) ( Hkq ) (c l c + 1)
− −( α
2
− +− − 1)( λ 1)
ED

EV 1 EV 1 ntr l
• 2 ••••
• ⋅θ ( Hkq EV 1
) ( c ntr − 1) λ l
+ 2( c ntr ( − c+ntr+ + δ m
1) δ m )) •

α β( θ − 1) ( Hkq ) (( α θ− ) ( c l c−
2
N = 1) ( Hkq +ntr− 1) 4 c ntr mδ ) (C.33)
T

52 EV 1 EV 1
EP

2
D 50
= θ ( Hkq EV 1
) ( αβ − 1+ )−( α 1 ) θ( Hkq EV 1
)λ l
− 2 (α − 1 ) θ( Hkq EV 1 )λl − 4 (C.34)

D 51
= − 2α β( − 1 ) θ( Hkq EV 1 ) (C.35)
C

D = α β( − 1 ) θ( Hkq ) (C.35)
AC

52 EV 1

Relationship between ntr c and 2 Π

The equation below shows the relationship between ntr c and Π .2 Expressions from 60 N to 60 D are
presented from Eq. C.39 to Eq. C.42 in appendix as well.

N 60N +c + N 62c
2
ntr
)=
61
Π 2 ( c ntr (C.37)
4 kD
ntr
60

By using gird searching method again, the ranges for each character are
 ACCEPTED MANUSCRIPT

•••• − 4≤ N 60
≤ 20

− 16
≤ N 61
≤ 8
−7
(C38)
− 4≤ N ×
≤ − 2.18 10
••• 62

− 5≤ D 60
≤−4

We then find that the first derivative − 2 / 4( k ) ≤ Π ′2 ( c ntr ) ≤ 4 / 4

( k ) and the linearity is
8
6.25 10 η 2c 1.15 10
7
− × ≤ ≤ × . Back to Eq. C.37 again, the relationship can be quadratic.
ntr

α β( 1) θ( Hkq α θ− 1) ( Hkq • ) (+ − α θ 1) ( Hkq )

2
• − ch
) ((
EV 1 EV 1 EV 1
• •
·( λ l − 2 ) λ l − +4)−( α θ 1) ( Hkq EV 1 ) c( l ( α β − 1) θ( Hkq EV 1 ) 2( λ h − ) λ h
2

T
• •
• •
+ α β( − 1) θ( Hkq EV 1 ) − +4) 2 ( α β − 1) θ( Hkq EV 1 ) c( h( − −α θ 1)

IP
• •
• · ( Hkq EV 1 ) c l ( λ h − 1 ) ( λ l − 1− )(− α θ 1) ( Hkq EV 1 ) 1( λ l − ) ( λ hλ − l ) •

CR
• •
• + 2 ( λ h − 1 ) δ m − 4) 2( + − α θ 1) ( Hkq EV 1 ) c( l ( α β − 1) θ( Hkq EV 1 ) 1( λ h − ) •
N 60 = •
22 22 22
•
·( λ hλ − + l ) ( 2 λ l − 1 ) δ m − +4) α βθ ( Hkq EV 1 ) λ hα θ
− ( Hkq EV 1 ) λ h •
2 2

US
• 2 2 2
•
− αβθ ( Hkq EV 1 ) λ hαθ+ ( Hkq EV 1 ) λ h + −2 ( α 1 ) θ( Hkq EV 1 ) λ l
2 2 2

• •
•
⋅ (2 δ mα β
− ( − 1) θ( Hkq EV 1 ) λ h ) 4+ αβθ ( Hkq EV 1 ) λ δ hm
− 4 αθ ( Hkq EV 1 ) •
•
AN
•
221
• ⋅λ δhm
+ −( α α 1) ( β − 1) θ
(
2
Hkq EV
) 4 λ l
− αβθ ( Hkq EV 1
) +
4 ( θ Hkq EV 1
) •
• •
•
− 4 αβθ ( Hkq EV 1 ) δ m + 4 θ ( Hkq EV 1 ) δ m − 4 δ m
2 •
• •
M

(C.39)

• 2( θ( Hkq EV 1
) (( α − 1) c(2(
l
λ l
•
+ −1) ( αβθ − 1) ( Hkq EV 1
)(λ h
− 1)
• •
·( λ hλ − l ) ) (+ −α β 1) c( h( λ hαθ− ( Hkq EV 1 ) (+ θ Hkq EV 1 ) (+ − α θ 1)
• •
• · ( Hkq EV 1 ) λ l + 2) (+ − α θ 1) ( ( − Hkq EV 1 ) ) 1( λ l − ) λ l + 2) 2( +− α 1) •
ED

N 61
= • • (C.40)
•
· λ lα( β θ( − 1) ( Hkq EV 1 ) λ hδ + m + −1 ) α β( − 1) λ((hα θ − 1) •
• · ( Hkq EV 1 ) λ h + −2 ( λ h 1)δ m − −2)−( α λ α1)β (θl( − 1) ( Hkq EV 1 ) •
2

•
T

•
•
+ 2 δ))m2 ( + θ Hkq EV 1 ) ( αβ − 1) + 4) δm •
• •
EP

•
2
• θ ( Hkq EV 1
) (( α α− β
1)θ( − 1) ( Hkq EV 1
) λ2hα( −β λ − 1) h
N 62
=• • (C.41)
• ⋅ (( α θ− 1) ( Hkq ) λ l + 2) (+ − α λ α1)β(θl ( − 1) ( Hkq ) λ l − 4)) •
C

EV 1 EV 1
AC

2
D 60
= θ ( Hkq EV 1 ) (α β − 1 ) (λ h
− 2 ) λ hα+ −( 1)λ l
− 2 ( α − 1 ) λ lθ) + ( Hkq EV 1 ) ( αβ − 1−)4
(C.42)

Relationship between l c and 2 Π

N 70N +c N 71c c +
2

2 ( )l
72 l
Π = (C.43)
4 kD
l
70

− 4≤ N 70 8 ≤ 20
•• −
• 9.18 10
× ≤ N 71 8
≤ 12
• −
(C.44)
8.24 10
× N
≤D ≤ 4
••• 72

− 5 ≤ 70
≤ −4
 ACCEPTED MANUSCRIPT
−8
We find The first derivative − 4.24 / 4( k ) (≤ ) Π '2 cl ≤ − 9.06 10
× /4 ( k ) and the linearity is
− 25% ≤ η 2c l
≤ − 6.29% . Therefore, the relationship between l c and Π has
2
negative relationship.

α β( θ − 1) ( Hkq • α θ− 1) ( Hkq ) (+ − α θ 1) ( Hkq ) (λl − 2)

2
• ) c((h
EV 1 EV 1 EV 1
• •
•
·λ − + 4) 2 ( α β θ − 1) ( Hkq EV ) c( c ntr ( λ( αθ−
l
( Hkq EV ) (+ θ Hkq EV 1 )
• 1 h h 1

• + −( α θ 1) ( Hkq ) λ + 2) (+ − α θ 1) ( (− Hkq EV 1 ) ) ( λ − 1) λ l + 2) •
•
EV 1 l l
•
• − (− α θ 1) ( Hkq EV 1 ) ( λ − 1)( λ λ − + ) 2( λ − 1) δ m − 4) l • h l h

• •
+ 2 c(2 ( θ Hkq EV ) ( αβ α+β− ( 1) λ − 1) + −( α θ 1) ( Hkq EV 1 )
• ntr 1
• h

• ⋅ (2 λ α−β θ( − 1) ( Hkq EV ) ( λ λ −
2
) )2 + ( (δ mθ Hkq EV 1 ) •

T
•
l 1 h l
•
• · (( − α β − 1) 1( λ − ) λ α−λ − (α λ 1)
2
) + 2)) + c •
2
+ −( 1)

IP
h ntr h l l

N 70 = • •
• ⋅ (θ( Hkq EV ) (( α α− β 1)θ( − 1) ( Hkq EV 1 ) λ2 ( − α β − 1) λ h
2
1
• h

CR
• •
⋅ (( α θ− 1) ( Hkq EV 1 ) λ 2 l + ) (+ − α λ α1)β(θ( − 1) ( Hkq EV 1 ) λ l − − 4)) 4)
• • l

• 2 2 • 2
+ α βθ ( Hkq EV ) λ α −θ ( Hkq EV ) λ αβθ ( Hkq EV 1 ) λ h
22 2
−
2 2 2 2 2

• 1 h
• 1 h

US
• + αθ ( Hkq EV ) λ + −2( α θ 1) ( Hkq EV 1 ) λ
221
2
(2 δ α βm θ− ( − 1) •
•
h l
•
• · ( Hkq EV ) λ ) 4+ αβθ ( Hkq EV ) λ4δ1 h
( − αθ Hkq EV 1 ) λ hδm
hm • 1
AN
• 2 •
• + −( α α β Hkq EV 1 ) λ − 4 αβθ ( Hkq E V ) + 4θ ( Hkq EV 1 ) •
2 2
1) ( − 1 ) θ( l 1

• 2 •
− 4 αβθ ( Hkq EV ) δ m + 4 θ ( Hkq EV 1 ) δ m − 4 δ m
• 1 • ( C.45)
M

• 2( α θ− 1) ( Hkq EV 1
) (−α β
( θ − 1) ( Hkq λ l − 1) • EV 1
)c
(h λ h
− 1)(
• •
N 71
=−
• c ntr ( α β( − 1 ) θ( Hkq EV 1 ) (
λ − 1 ) ( λ λ− − ) ( 2 λ l + 1) ) •
h h l
(C.46)
• •
• +α β
( θ − 1) ( Hkq EV 1
) (λ h
− 1)( λ λ − + ) 2(h
λ − 1) δ m − 4) •l l
ED

N 72
= −( α 1 ) θ( Hkq EV 1
) (α β − 1 ) θ( Hkq EV 1
) (λ h
− 2 ) λ α +β− (
h
1 ) θ( Hkq EV 1 ) − 4 ) (C.47)

) (α β ) λ hα+ −( − 2 ( α − 1 ) λ l θ) + ) ( αβ − 1−)4
2
D = θ ( Hkq EV
− 1 ) (λ − 2 1)λ ( Hkq EV 1
T

70 1 h l

(C.48)
EP

Relationship between h c and 2 Π

C

N 80N +c N 81c c +
2

( )h =
82 h
AC

Π 2
(C.49)
4 hkD 80

− 4≤ N 80 7 ≤ 20
•• −
• 1.56 10
× ≤ NN
81 8
≤ 12
• −
(C.50)
6.66 10
× D
≤ ≤ 4
••• 82

− 5 ≤ 80
≤ −4

−8
We find The first derivative − 4.24 / 4( k ) (≤ ) Π '2 cl ≤ − 9.10 10
× /4 ( k ) and the linearity is
− 24.91% ≤ η 2c h
≤ − 6.25% . Therefore, the relationship between h c and Π has
2
negative

relationship.
 ACCEPTED MANUSCRIPT

( α θ− 1) ( Hkq ) c( l ( α β θ − 1) ( Hkq λ hα+β−θ (

2
•••••••••••••••• EV 1 EV 1
) (λ h
− 2) 1) •••••••••••••••••••••••••

· ( Hkq EV 1
)−+
4) 2( α θ− 1) ( Hkq EV 1
)c
( lc ntr
(2 ( λ l + −1 ) α β( − 1)

·θ( Hkq EV 1
) 1( λ h
− ) ( λ hλ − l
) ) (+ −α β θ 1) ( Hkq EV 1
) 1( λ h
− ) ( λ hλ − l )

+2 ( λ l − 1)δ m
−+
4) 2 (2c(ntr θ Hkq EV 1
) ( αβ α+β− ( 1) 1λ h − )
2
+ −( α θ 1) ( Hkq EV 1
)2 ( λ α−β θ(l
− 1) ( Hkq EV 1
) ( λ λ− h l ) )
+ 2 δ mθ ( ) (( − α β ) λ hα−λ −
(α λ ) + 2 ))
2
( Hkq c EV 1
− 1) 1( λ h
− 1) l
+ −( 1) l

) (( α α− β
1)θ( αβ − 1) λ h
2
N 80 = + (θ( Hkq 1) ( Hkq ) λ2h( −
2
−

T
ntr EV 1 EV 1

⋅ (( α θ− 1) ( Hkq ) λ + 2) (+ − α λ α1)β(θ( − 1) ( Hkq )λ l − −

4)) 4)

IP
EV 1 l l EV 1

2 2 2
+ α βθ ) λ hα −θ ( Hk q EV 1 ) λ hαβθ ) λh
22 2
− ( Hkq
2 2 2 2 2
( Hkq EV 1 EV 1

CR
••+
2
(α θ 1 ) ( ) λ δ( α
2 βm − 1 ) θ( Hkq )λh )
2
αθ ) λh
2
( Hkq EV 1
+ −2 Hkq EV 1 l
− ( EV 1
•+
••⋅ 4 αβθ ( Hkq + −( α α β1 ) ( − 1) θ
2
( Hkq EV 1
) λ δh m − 4 αθ EV 1 ) λ δh m

US
221
••+( Hkq EV
) 4λ l
− αβθ ( Hkq EV 1
) 4+ ( θ Hkq EV 1
) 4− αβθ ( Hkq EV 1
)δ m
2

••
4θ ( Hkq EV 1 ) δ m − 4δ m
AN
(C.51)

•− 2 α( β θ 1) ( Hkq α θ− 1) ( Hkq ) cl ( λ h − 1) (λ l 1) •
M

− EV 1
) (( EV 1
−
• •
•
+λ h ( c ntr ( ( α θ− 1) ( Hkq EV 1
) (− −α θ 1) ( Hkq EV 1
)λ
2l − ) • (C.52)
N 81
=•
+ −( α θ 1) ( Hkq EV 1
) (λ l −1
−)2 ) ( δ m
+− α θ 1) ( Hkq EV 1
) 1( c ntr − )•
•
•
••
• ⋅ ( λ l − 1) 2λ l − c ntr + 2δ m
+ 4) •
ED

N 82
= α β( − 1 ) θ( Hkq EV 1
)( α − 1 ) θ( Hkq EV 1 ) (+ − α 1 ) θ( Hkq EV 1 ) (λ l − 2 )λl − 4) (C.53)

) (α β ) λ hα+ −( − 2 ( α − 1 ) λ l θ) + ) ( αβ − 1−)4
2
D 80
= θ ( Hkq EV 1
− 1) (λ h
− 2 1)λ l
( Hkq EV 1
T

(C.54)
EP

Appendix D. Proof for relationships between parameters and total optimal profits in
Period 3
C

Relationship between θ and 3 Π

AC

2
N 90N + θ + N θ
Π 3 (θ) =
91 92
(D.1)
4k ( − 4+ D 91 θ )

• 0 ≤ N 90 8 ≤ 0
••• −
3.80 10
× ≤ N 91
≤ 24
− 16 − 17
(D.2)
− 1.83 10
× N≤D ≤ 7.01 10
×
••• 92
−7
− 1≤ 91
≤ − 1.35 10
×
 ACCEPTED MANUSCRIPT
−8
We find The first derivative − 6 / 4( k ) (≤ ) Π '3 θ ≤ − 8.06 10
× and the linearity is
− 13
− 5.57% ≤ η≤ 3−θ 1.86 10
× . Therefore, the relationship between θ and 3 Π can be treated as

negative linear.

2
N 90
= − 4( c ntr − δ m ) (D.3)

• 4( Hkq ) (( α αβ
− )c h
2
+ −α β
( • 1) c( h(c λ + 1) (+ − λ 1) δ m − 2)
EV ntr
•
22 h h
•
2
• − αβ λc δ αβ λ δ α+λ δ αcλ δ αβ λ αβ + λ α λ α λ α − c •
2

ntr h m ntr h m
c ntr h m
− c ntr h m ntr h
• 2
•
• + + c ntr h − c ntr h − − 1 ) c l − α λc ntr δ α lλm δ + c ntr l m •
2 2
c ntr h (
• •

T
2
N = + − (α 1 ) c l c( ntr ( λ l + 1 ) (+ − λ l 1 )δ m − + 2 ) c ntr λ lδm − c nt r λ lδm (D.4)
91 • •

IP
• 2 •
− α λc ntr
α λ l + c ntr l + −c ntr λ l c ntr λ lαβ + δ λ c ntr m u
2 2

• •
• + αβλ 2 •
u ( ntr )

CR
− c ntr ( c + δ + −
1 )+( c 1 ) δ + c + αβ δ c αβ
um
− c c
ntr u
•
u m u m
•
•
+ m + 1) ••
2
• − c+ntr αβ c u + 2 αβ c u + αβλ δh αλ m
δ−αλ δ hλmδ δ+ − l m lm

US
• ( Hkq )(α
EV 2 2 2
2
α−β αc λ β α−λ β cαλ
ntr h
βu
22 2 22
− cuh
222
−
22
h
+ 2 α λc ntrβh
22
•
• •
• − 2α λc uβh + 2 α c ntrcu h λ β + 2 α λc uβh + 2 α λc uβh − 2 α c ntrcu h λ β
22 2 2 2 2
•
•
AN
•
+ c u β α+ λ2 β c ntr h − c ntr hλ β α +λ β cαλ β α λ+β2
22 2
−4 c ntr h
2 2 2 2 2 2

• uh h
•
• + 2 c ntr hλ β α +λ 2β αc u h
2 2
−2 c ntrc λ β λ −β α λ β −
2 2
cul
22
+ c u lλ β
22 •
uh
•
h
•
M

• − 2α λ
c uβl α c ntrcu l λ β − 2 c ntrcu l λ β + 2 c u lλ β αβ− (
2
+ − 1) •
2 2 2
2
• •
λ β α −λ 2β αcλu hβ α c ntrcu h λ β α+λ 2β c u l
22 2
+
2

• ⋅ ( c ntr − 1) u
− 2 cuh 2 •
• •
− 2 c u lλ β α+λ2β αc u l c ntrcu l λ β + 2 c ntrcu l λ β − 2 c u lλ β α−λ2λ βc ntr h l
2 2
− 2
• •
• + 2 c ntr hλl λ β α − 2 λ hλl β α +λ λ
2
4 β c ntr h l − 4 c ntr hλl λ β λ +λ β2 •
hl
•
ED

•
• + 2( c ntr − 1)(( c ntr − 1)( α( β − 1) (λ hα+λ− 1) ) l
+ c u (αβ λ+ l •
N 92 = • •
−α β( ( λ λ+ − ) λ β α−λ + −c ntr hλ αλ + 2 α λc ntr h
2 2 2 2 2 2
− 1) 1)) c ntr h
• h l u h •
T

•
− 2 c ntr hλ λ +α−λ
2 2
c ntr
2 2
+ −c ntr
2
λ αλ
2 2
+ 2 α λc ntr
2
− 2 c ntr λλ+
2 2 •
•
h l l l l l l
•
EP

• + 2 α c nt r hλl λ
2
− 2 c ntr hλl
2
λ αλ+ λ2 α λ hλl − 4 c ntr h l + 4 c ntr hλl λ − 2 λ hλl •
• •
− 2 αβλ α β+ ) λ l + 1)
2
− (−β ( αβλ − 1−) −( α
2
• 1) ch ( 1 ) (λ − 2 •
C

u u l

• •
(α − 1) cl
2
( − βλ + βλ ) λ h − 1+)2 •
2
• + −( α 1) cl 2 + −( β 1) (λ − 2
AC

u u h

• •
• ⋅ ( β cλu ( βλ β λ+ 1) λ h
2
l
− 1)( λ u
− 1)
− ( c ntr − 1)(
− u
( l
+ 1) (λ u + − β •
• • (D.5)
·( λ λ − − − 1) λ ) ) 2(
+− β 1) c(h( 1− −α )c ( λ − 1) (λ − 1+) αβ c u
• h l l l h l
•
•
⋅( λ − 1)( λ − 1)
− ( c ntr − 1)(
− αβλ αβλ
+α
2
+ −( 1)( λ − 1) λ l •
•
h u u u l
•
•
+ λ hβα
( −λα
− βλ α α λ+ + +− 1)))) •
• l u l •

− 2) λ l •
2
• ( Hkq D
EV 2
) (α( β − 1) λ2 h( − α β λ −α 1) ( h
+− 1)( λ l
91
=• • (D.6)
− αβλ + 2 αβλ Hkq )
2

• u u
) (− EV 2 •

Relationship between α and 3 Π

 ACCEPTED MANUSCRIPT
2
N 100
N + α + N 102 α
Π 3 (α) =
101
(D.7)
4k D
( D100 + 101
α )

••••• − 4≤ NN
100
≤ 20

− 19.2 ≤ N D101 ≤ 21
− 16 − 16
− 3.77 10
× ≤ D 102
×
≤ 1.42 10 (D.8)
••••
− 5≤ 100
≤− 4

− 1≤ 101
≤1

5.25 / 4( ) (≤ ) Π '3 α ≤ 4.76 / 4( k ) and the linearity is

T
We find The first derivative − k

η 3α

IP
− 5.41% ≤ ≤ 5.46% . Therefore, the relationship between α and 3 Π can be treated as linear.

4( θ( Hkq )c l ) (c l c + 1) (+ − λ l • δ m − 2)
2
Hkq (λ

CR
• − θ ( 1)
EV 2 EV 2 ntr l
• •
• + c ntr (δ θ ( m
( Hkq EV 2
) (λ l − 1 ) λ l + −2 ) ( θ Hkq EV 2 ) ( λ l + 1 ) ) •
N =• (D.9)
100 •
+ c ntr (θ( Hkq ) λ l − 1)
+( Hkq EV 2 ) (− ( δ mθ Hkq EV 2 ) ( λ l − 1) •
2

US
• EV 2
θ
•
+ •δ )) •
m •
AN
M
T ED
C EP
AC
 ACCEPTED MANUSCRIPT
2
• ( Hkq EV 2
) θ( (β− − 1)(( Hkq EV 2
) θ( + Hkq EV 2
)(λ l
− 2) λθl + 4) c h •
• •
• + −2( β 1)(2 ( c ntr − − δ m+ 2 )( Hkq EV 2 )θ λ cl ( h − 1 ) ( λ l − 1) •
• •
+ ( Hkq EV 2 )θ ( c ntr − 1 ) ( λ l − 1 ) λ lλ +δ h (2 m + ( Hkq EV 2 )θ λ( l − 1 )
• •
•
) βθ c u •
2
+ c ntrHkq ) θ( − Hkq EV 2 ) λθl + 2 )) ) 4 c h + β c uHkq +(
2
(( EV 2 EV 2
• •
• + ( Hkq EV 2 )θ c n tr hλ − ( Hkq EV 2 ) βθ λc ntr h
2 2 2 2
+ ( Hkq EV 2 )θλ h
2
•
• •
2
− ( Hkq EV 2 ) βθλ h − 2 ( Hkq EV 2 ) θ λ + 2 ( Hkq EV 2 ) βθ λc ntr h
2 2
• c ntr h •
• 22 •
−4βδ λm h + 4 c ntr mδh λ + ( Hkq EV 2 ) θ λc ntr l + ( Hkq EV 2 ) βθ λc u l
2 2 2 2
c ntr
• •
• •

T
2
+ ( Hkq EV 2 ) θλ l − 2 ( Hkq EV 2 ) θ λc ntr l + 2 ( Hkq EV 2 ) βθ λc u l
2 2

• •

IP
• 2
− 2( Hkq EV 2 ) βθ c ntrc u l λ − 4 c ntr mδl λ β + (( Hkq EV 2 ) θ( c ntr − 1)
2 2 •
• •
2
+ 4 c ntr mδu λ ) 8 + − β c u 4 β c ntr
c u + 4 β δc u m − 4 β λc ntr h + 4 c ntr hλ •
2 2

CR
•
• •
+ 4 β λc ntr h − 4 c ntr hλ + 4 βδ λm h + 4 β δ λm h
c ntr − 4 c ntr mδh λ
• •
N 101 = •
− 4δm λ h − 4 c ntr λ l − 2 ( Hkq EV 2 ) βθ λc u l + 4 c ntr λ l − 2 ( Hkq EV 2 )
2 2 •

US
• •
• ⋅ βθ λc u l + 2( Hkq EV 2 ) βθ c ntrc u λ l + 4 c ntr mδl λ + −4 δ mλ l 2( Hkq EV 2 ) •
• •
⋅θ λ λh l + 2( Hkq EV 2 ) βθ λc λntr h l − 2( Hkq EV 2 )θλ λ + 2( Hkq EV 2 )
2 2

• c ntr •
AN
hl

•
⋅ βθλ λh l + 4( Hkq EV 2 )θ λc ntrλ h l − 4( Hkq EV 2 ) βθ λc λntr h l •
• •
2
• − 2 β((( Hkq EV 2 )θλ l − 2) c ntr + ((( Hkq EV 2 )θ + 2) 2( c u + δ m + 1) •
M

• •
• − ( Hkq EV 2 ) θ ( c u + 2 ) λ l ) c nt r + 2 δ m + c u ( 2 δ m + ( Hkq EV 2 ) θ λ( l − 1 ) ) •
• 2
•
+ ( Hkq EV 2 )θλ λl ) u + c l Hkq ) θ( + Hkq EV 2 ) ( βλ u − 2 βλ β − −(
2

•
(( EV 2 u
1) •
• 22
•
·( λ h − 2 ) λ hθ) − +4) 2 (( c l Hkq ) βθλ u − ( Hkq EV 2 ) βθ λ( l + 1 ) λ u
•
EV
•
• − 2δ m + 2δ λ l − ( Hkq EV 2 ) θλ λh l + ( Hkq EV 2 ) βθλ λh l + ( Hkq EV 2 ) θ
( m •
ED

• •
• ⋅ λ lβ− − ( 1)( λ h − 1) λ) h( − Hkq EV 2 ) βθ c u ( λ l − 1)( λ u − 1) •
• 2 •
+ c ntr (2( λ l + 1) (+ Hkq EV 2 ) θ( βλ − β λu + ( l + 1) (λ uβ + − 1) λ h
T

• •
• ⋅ ( λ hλ − − l− 1) λ l ) ) − 4)) •
EP

• •

(D.10)
C
AC
 ACCEPTED MANUSCRIPT

2
β λc ntrλ h l + β λc ntrλ h l + 2 c ntr hλl λ − 4 c ntr hλl λ
2 2 2
•• θ ( Hkq EV 2
) ( 2− 4 ••

• + 2 β ( c ntr − 1)λ
( u ( c ntr − 1 ) ( β − 1 ) λ hλ + l ) + −c −( β β (
u
1 ) λ hλ − l ))•
• •
• + 2 c( l c β λ u ( l
− 1)( λ u
− 1)
− ( c ntr − 1)(( β − 1) λ( hλ λ h
− − l− 1) λl •
• 2 •
+ β λ( + −1 ) λ uβλ ) ) 2(+ − β 1) c( hc − (λ h − 1) ( λ l − 1−) ( c ntr − 1 )
• l u l
•
• ⋅ ( λ( β λ− − + βλ + 1) (+ − λ l 1) λ lβ−λ ( − 1) λ) uβ λ+ cu ( − 1) •
h l u u h
• •
2
• ·( λ u − 1
−))−( β 1) c h ( β λ− −
( l
2 ) λ lβ+λ− ( u
2 ) λ u − 1) •
• •
2
N 102 = +cl (( β ) λ hβλ− + −2 βλ
+ 1) ( − c ntr ) λ h
2 2 2 2
• − 1) (λ h − 2 u u
β •
•

T
•
2
+ 2 β λ c ntr h + β λc ntr h β λc ntr h + + 2 β c ntrc u h λ
2 2 2 2 2 2 2 2 2
• 2 − 4 − c ntr hλ 2 c ntr hλ •

IP
• 2 222 2 •
c ntrc u h λ c ntrc u h λ + 2 β c n tr uch λ β −λ − 2 β λcu h
2 2
−2 β − 2β cuh
• •
•

CR
• 22 22 2
+ 2 β λc u h + β λ cβu hλ + + β λc u h β λc u h 2 β λc u h − c ntr λ l
2 2 2 2
2 cuh 2 − 2 −
• •
22 2
+ 2 c ntr λ l + c ntrc u l λ c ntrc u l λ β−λ β λc u l + 2 β λc u l •
2 2 2
• 2β − 2β cul − 2
•
22 22 22 2
•
+ 2 β λc uβl λ uβ − βλ + −2 βλ
− λh 2 βλ λh l

US
cu
2 2

• − ( c ntr − 1) − h h •
• 2 •
• + 2 λ hλl λ − l
) •• ( D.11)
•
AN
D 100 = −
θ ( ( Hkq EV 2 ) ) −( θ Hkq EV 2 ) (λ l − 2 )λl − 4 (D.12)

2
D 101 = θ ( Hkq ) ( β − 1 ) ( λ h − 2 ) λ hλ + −( 2 ) λ l βλ
− + 2 βλ ) (D.13)
M

EV 2 l u u

Relationship between β and 3 Π

2
N 110
N + β + N 112 β
Π 3 ( β) =
111
(D.14)
ED

4k D
( D110 + 111
β )

••••• − 4≤ NN
110
≤ 20
T

− 19.2 N ≤ 111
≤ 21
EP

− 31
0 ≤ 112
≤ − 1.10 10
× (D.15)
••••
− 5≤ D ≤−
D 110 4
C

− 1≤ 111
≤1
AC
 ACCEPTED MANUSCRIPT

We find The first derivative − 5.25 / 4( k ) (≤ ) Π '3 β ≤ 4.76 / 4( k ) and the linearity is

− 5.57% ≤ η 3β ≤ 5.31% . Therefore, the relationship between β and 3 Π can be treated as linear.
2
) •
22 2 22 2 2
• − αθ ) + 2α θ ) c ntr hλ αθ+ ( Hkq
2 2 2
( Hkq EV
c ntr hλ ( Hkq EV EV 2
•
2 2
•
2
− 2 αθ ) c ntr hλ αθ+ c hHkq ) (θ αθ
− ) •
2 2 2 2 2
• ⋅c
ntr h
λ ( Hkq EV 2
( Hkq EV 2
) (( EV 2
• •
− (−α θ 1) ( Hkq ) λ + −4)−( α θ 1) ( Hkq ) cl
2
• EV 2
) 2( λ l
− l EV 2 •
• •
⋅ ( αθ ( Hkq )(λ − 2) λ hαθ+ ( Hkq ) + 4) 2+ ( αθ Hkq )ch
• EV 2 h EV 2 EV 2
•
• ⋅ (( α θ− 1) ( Hkq EV 2
) (c l λ h
− 1)( λ l
− 1)
+ λ h
( c ntr (( α θ− 1) ( Hkq EV 2
) •
• •
•
− (−α θ 1) ( Hkq ) λ l − +2)−( α θ 1) ( Hkq )(λ − 1)
− 2) δm •

T
EV 2 EV 2 l

• + −( α θ 1) ( Hkq ) ( c ntr − 1)( λ − 1) 2λ − c ntr + 2δ + 4) •

IP
EV 2 m
•
l l
•
2
• + −2( α θ 1) ( Hkq EV 2
) (λ lαθ ( Hkq EV 2
) ( c ntr − 1) λ h •
N 110 = • •

CR
• + 2 ( c ntr ( − c+ntr δ m
+1) + δ m ) ) 2(+ − α θ 1) ( Hkq EV 2
)cl •
• •
⋅ ( c ntr ( αθ ( Hkq )(λ − 1)( λ hλ − + l ) 2( λ l + 1))
• EV 2 h
•
•

US
αθ
• +− ( ( Hkq Hkq
EV 2
)) ( λ h
− 1)( λ hλ − + l ) 2( λ l
− 1) δ m − 4)
•
2
•
• + 4 αθ ( EV 2
) c ntr hλm δ − 4 αθ ( Hkq EV 2 ) c ntr hλm δ •
• •
+ 4 αθ ( Hkq ) c ntr hλ ) c ntr hλ α − −( 1 ) θ( Hkq ) •
AN
− 4 αθ
2

• EV 2
( Hkq EV 2 EV 2

• •
⋅ λ lαθ( ( Hkq − 1)
+ 4 ) 4− ( θ Hkq ) c ntr
2 2
) ( c ntr c ntr mδ
•
EV 2 EV 2
•
• 22 2 2 2 • (D.16)
M

+ 8 c ntr mδ αθ ) λ αθ+ ( Hkq ) λh

2 2 2
− 4 c ntr − ( Hkq EV EV 2
•
2 h
•
•
− 4 αθ ( Hkq ) λ δh m + 4θ ( Hkq )+ 4θ ( Hkq )δ − 4δ m
2 •
• EV 2 EV 2 EV 2 m •
T ED
C EP
AC
 ACCEPTED MANUSCRIPT

αθ (
•••••• Hkq EV 2
) (−2 ((c h α θ− 1) ( Hkq EV 2
) (c l λ h
− 1)( λ l − 1) ••••••

− c ntr ( λ( h2
αθ(− Hkq EV 2
) (+ θ Hkq EV 2
) (+ θ Hkq EV 2
) (( α λ− 1) l

+ αλ λ lαλ− αλ
2

u
)+ 2) ( +θ Hkq EV 2
) (−( −α 1)( λ l
− 1) u
+ u
) + 2)

− αθ ( Hkq EV 2
) c u hλ αθ+ ( Hkq EV 2 ) λ u ( c u ( λ h − 1+)+ λ h 1)
• •
• + αθ ( Hkq EV 2 )cu + −( 1 2 α θ) ( Hkq EV • 2
) λ hαθ+ ( Hkq EV 2 ) λ hλl
• 2
•
− θ ( Hkq EV 2 ) λ hλl − 2 λ hδmαθ − ( Hkq EV 2 ) λ lαθ+ ( Hkq EV 2 ) λ l
• •
• 2 •
+ θ ( Hkq EV 2 ) λ lθ − ( Hkq EV 2 ) λ lαθ ( Hkq EV 2 ) 2 λ u + δ m + 4)
2
−
• •
• + 2 c( αθ ( −( Hkq ) ) ( c ntr − 1) λ hαθ+
2
( Hkq EV 2 ) ( c ntr − 1) λ h •

T
u EV 2
• •
λ uα((θ − 1) ( Hkq EV 2 ) ( c ntr − 1) λ l + c ntr (θ( Hkq EV 2 ) + 2)

IP
• ⋅ ( λ u + −1) •
N 111 = • •
+− θ ( ( Hkq EV 2 ) ) + 2 δ) m( + − α θ 1) ( Hkq EV 2 ) 1( c ntr − ) ( λ l − 1 ) λ l
•

CR
•
• + −( α θ 1) ( Hkq ) c
( λ − 1)( λ − 1)
− 2 2 c + δ + 4) •
EV 2 l l u ntr m
• •
• ⋅ (2 c ntr ( θ( Hkq EV 2 ) ((1 −2 ) α λ h + −2( α λ λ1) l + − u ) 2 δm •

US
• •
·( λ hλ + u − 1 +)2)
− 2( α θ− 1) ( Hkq EV 2 ) c l c( ntr − 1 ) ( λ hλ − λ + l− 1)
•
u
•
•
+ c ntr ( ( 2 α1 θ− ) ( Hkq EV 2 ) λ h − θ ( Hkq EV 2 ) ( 2 α − 1 ) λ lλ + − u ) 4 ) •
2

• •
AN
•
+ −( α θ 1) ( Hkq EV 2 ) (c l λ hλ + − u+
2
2) ( θ Hkq c EV 2 ) ((2 α 1)λ− •
h
• •
• − 2( α λ−λ1) − + u ) 4 δ) m + h ((2 1) α θ(− Hkq EV 2 ) (+ θ Hkq EV 2 ) •
2
(D.17)
M

• •
• · (( α − 1) 2( λ l − ) λ lαλ+ − u 2 αλ u ) − + 4) ( ( c u θ Hkq EV 2 )
2 2
•
• •
• ••
• ⋅ ( αλ h − 2 αλ αh − −( 1)( λ l − 2) λ ) (lθ + Hkq EV 2 ) + 4))
2

22 2
αθ ( −( Hkq ) λ hλ − − u ) ( + ( c h λ u − 1))
− 1)
2
N 112
= EV 2
) (( c ntr − 1)( cu λ h
(D.18)
ED

( − ( Hkq ) ) λ h + 2 αθ ( Hkq ) λ hθ − ( Hkq )•

2
• αθ
EV 2 EV 2 EV 2
D 110 =• • (D.19)
• + −( α θ 1) ( Hkq ) (λ l − 2 ) λl − 4 •
• EV 2 •
T

2
)λ h ) λ hαθ− ) λ u + 2 αθ ( Hkq )λu
2
D 111 = αθ ( Hkq − 2 αθ ( Hkq ( Hkq (D.20)
EP

EV 2 EV 2 EV 2 EV 2

Relationship between l λ and 3 Π

C
AC

2
N 120
N + λ + N 122 λ l
Π 3
( λ)l =
121 l

2
(D.21)
4 ( D120 +
kD 121
λ l
+ D 122 λ l )

•••••• − 4≤ N 120
≤ 20
−8
− 16
≤ N 121
≤ − 9.09 10
×
−8
9.10 10
× ≤ N 122 ≤ 4
(D.22)
− 5≤ D 120 8 ≤ − 4
•••••
−
9.03 10
× ≤ D 121 ≤ 2
−8
− 1≤ D 122 ≤ − 9.28 10
×
 ACCEPTED MANUSCRIPT
−8
We find The first derivative 9.08 10
× /4 ( k ) (≤ )Π '3 λ l ≤ 2.33 / 4( k ) and the linearity is

0.05% ≤ η 3λ l
≤ 24.99% . Therefore, the relationship between l λ and 3 Π can be treated as positive
linear.

• − ( Hkq ) αβθ
2 2222
cu + ( Hkq
2

) αβ θc u • 22
− ( Hkq
2
) α λ θh
222
EV EV EV 2
•
2 2
•
• − ( Hkq ) α β λ
2222
θ
2
− ( Hkq ) α λ
22
c θ
22
− ( Hkq )
2
α
2
β
22
•
EV EV ntr h EV 2
•
2 h 2
•
+ ( Hkq EV ) α λc ntrθh + 2( Hkq EV 2 ) α β λ cθntr h
22 22 22 22
• ⋅ c ntr hλ θ •
2 2 2 2

• 2 2
• 2

• − ( Hkq EV ) αβ λc ntr θh − ( Hkq EV 2 ) α β λ θc u h + ( Hk q EV 2 ) α β

22 2 22222 2

2
•
• •

T
+ 2( Hkq EV ) α βλ θ − ( Hkq EV 2 ) αβλ θh
222 22 22 22 22
• ⋅ c u hλ θ + ( Hkq EV ) αλ θ
2 2

2 h •2 h

IP
• 2 2 •2
+ 2 ( Hkq EV ) α λcθntr h + 2 ( Hkq EV ) α β c ntr hλ θ − 2 ( Hkq EV 2 ) α
22 2 22 22

• 2 2 •
• •

CR
− 4( Hkq EV ) α β λ cθntr h + 2( Hkq EV 2 ) αβ λc ntr θh
22 22 22 22
⋅ c ntr hλ θ
2

•
2
•
• − 2 Hkq α β λ θc u h
22 22 2
+ 2 ( Hk q EV ) α β λ cθu h
22
+ 2 ( Hkq EV 2 ) α
2 2 • 2 2
( EV 2 )
•
2
•

US
• ⋅β c c
2

ntr u h
λ
22
θ − 2( Hkq EV
) α
22
β c c
ntr u h
λ θ
22
2
+ 2( Hkq EV 2
) α
2222
β λ θc uh
2
•
•
2 2
• 2

• − 2 ( Hkq EV ) α β λ cθu h + 2 ( Hkq EV ) α β λ θc u h − 2 ( Hkq EV 2 ) α β

22 2
•
2 2 2 2

2 2

• •
AN
− 2( Hkq EV ) α β c ntrcu h λ θ + 2( Hkq EV 2 ) α β c ntrcu h λ θ
2 222 22 22

• ⋅ c u hλ θ 2 •
• •
+ 4 ( Hkq EV ) αβ θc + 4 ( Hkq EV ) α δ c ntrλmθh − 4 ( Hkq EV ) αβ δc ntr λ mθh
2 2 2

• 2 u 2
• 2
M

• + ( Hkq ) αβ
(4 c ntr mδ − ( Hkq EV ) ( αβ θ− 1) ( c ntr − 1) ) λ θ + 4( Hkq EV 2 )θ
2 • 2

•
EV 2 2 u
•
• − 4 ( Hkq EV ) c ntr θ + 8 ( Hkq EV ) αβ θc
2
− 4 ( Hkq EV ) αβ c ntrcu θ + 4 ( Hkq EV 2 ) •
2 u 2

• •
= • ⋅ δ mθ + 4( Hkq EV )αβ δc uθm + 4( Hkq EV 2 )α λc ntrθh − 4( Hkq EV 2 )αβ λc ntr θh
2
N 120
2

2 •
• •
− 4 ( Hkq EV ) α λ θh
c ntr + 4 ( Hkq EV ) αβ λc ntr θh − 4 ( Hkq EV 2 ) αδ λm hθ
• 2 2
•
ED

• + 4 ( Hkq ) αβδ λm θ − 4 ( Hkq EV ) α δ c ntrλmθh + 4 ( Hkq EV 2 ) αβ δc ntr λ mθh •

EV h
•
2 2
•
Hkq EV ) α ( β θλ− 1) + 2) c ntr + ((( Hkq EV 2 ) ( αβ θ− 1)
2
• + 2( Hkq EV ) αβ ((( 2 2 h
•
•
T

•
•
− 2) 2c − u
( δ m + −1 )( Hkq EV 2 ) α( β θ − 1) ( c + 2 ) λ ) c2 ntr − δ m • u h
EP

•
+ c Hkq (( ) (1 − αβ θ) δ − 2 ) m( + Hkq EV 2 ) α ( β θ − 1) ( c + 1) λ) h
•
•
u EV 2
• u

• − ( Hkq
EV
)α( β − 1) c((Hkq
2
2
EV
) (α( β − 1) h
+ +1) ( θ 2
Hkq EV 2 ) αβ ( λ − 2) λ u • u
C

• •
Hkq EV ) α ( β − 1) c( Hkq (− )αβθ ( c ntr − 1 ) λ + ( Hkq EV 2 )αβθ
2
• +− 4) 2(θ 2 EV h • 2 u
AC

• •
⋅ ( c ntr − 1)( λ + −1) 2λ c ntr + −2 δ m ( Hkq EV 2 ) ( α θ− λ1) ( c − 1)
• h u h
• l

• •
− (( Hkq EV ) ( − βα + −α θ 1) + ( ( Hkq EV 2 ) ( α β − 1+)1 ) θ + 2 ) c ntr + 2 δ mλ ) h
• 2
•
• •
− ( Hkq EV )αβθ λc ( ( ( θ + Hkq EV ) ( α − 1) c( ( − Hkq EV 2 )αθ
2
− 1) (λ −1 +)4)
• 2 u h u
• 2 l

•
+ ( Hkq EV )α βλ + 2 βλ + −( β 1) 2 λ − ) λ θ) − 4) 2( Hkq EV 2 ) ( α − 1) •
2
(− θ −
•
2 u u
• h h

• ⋅ c (2( δ + 2) (+ Hkq EV 2 ) αθ ( βλ
2
− βλ β − − ( 1)( λ − 1) λ β +λ c ( u − 1)) •
• (D.23)
l m u u h h u
•
•
ntr (
• + c Hkq( )αθ βλ ( − βλ + 1) 1( λ − ) λ ) − 2 )
2
) 4θ − c ntr
2

EV 2
+ −( β
•
u u h h
•
• − 4 δ 2 + 8c δ •
•
m ntr m
•
 ACCEPTED MANUSCRIPT
2
• − 2( α θ− 1) ( Hkq EV 2
) (α( β − 1)θ( Hkq EV 2
) c h + c l c( ntr
( αθ ( Hkq EV 2
) •
• •
• ·( − βλ λh βλ
+ αθ
h
+ u
) +− ( ( Hkq EV 2
) ) − −2) αθ ( Hkq EV 2
) •
• •
·( − β ( c u + λ h ) (+ β cu + 1 ) λ uλ +αθ )+ ( Hkq ) − 2 δ) m
• h EV 2 •
• •
+ α β( − 1)θ( Hkq ) c hc( ( λ h − 1−) ( c ntr − 1) (λ h + 1) )
N = +•
EV 2 l
• (D.24)
121
• •
αβθ ( Hkq ) c ntr hλ αβθ c nt r hλ αθ− ) c ntr hλ
2 2

2 )
− 2 ( Hkq ( Hkq
•
EV 2 EV EV 2
•
• + 2 αθ ( Hkq ) c ntr hλ αβθ
− ( Hkq ) ( c ntr − 1) λ( u c ntr − −c u 1 )2 •
EV 2 EV 2
• •
2
• + αβθ ( Hkq EV 2 ) c ntrc u − αβθ ( Hkq EV 2 ) cu − αβθ ( Hkq EV 2 ) c u − c ntr mδ •
•
•
••

T
+ ) 2+ + αβθ ) λ hαθ− Hkq )λ h )
2

• − 2( c ntr δ m
c ntr ( Hkq EV 2
( EV 2 •

IP
• ( α θ− 1) ( Hkq ) (α( β θ − 1) ( Hkq α β θ − 1) ( Hkq )ch •
2
EV 2 EV 2
) c2h ( − EV 2
• 2 •
⋅ ( c ntr − 1)
− ) (−2 (β + −c+ntr β cu )

CR
2
N 122
=
•
αθ ( Hkq EV 2
c ntr − 1) (c u 1)
• (D.25)
• − 4 c ntr mδ ) •
• •

US
) ( β − 1 ) ( λ h − 2 ) λ hβλ βλ θ ) + − ( ( Hkq
2
D 120 = αθ ( Hkq EV 2
− u
+ 2 u EV 2 )) − 4 (D.26)

D 121 = − 2 ( α − 1 ) θ( Hkq EV 2 ) (D.27)

AN
D 122 = −( α 1 )θ
( Hkq EV 2 ) (D.28)

λ and 3
M

Relationship between h Π

2
N 130
N + λ + N 132 λ h
Π 3
(λ h ) =
131 h
(D.29)
2
4 ( D130 +
kD 131
λ h
+ D 132 λ h )
ED

•••••• − 4≤ N 130
≤ 20
−8
− 16
≤ N 131
≤ − 9.09 10
×
T

−8
5.20 10
× ≤ N 132 ≤ 4
(D.30)
EP

− 5≤ D 130 7 ≤ − 4
•••••
−
1.49 10
× ≤ D 131 ≤ 2
−8
− 1≤ D 132 ≤ − 9.29 10
×
C
AC

−8
We find The first derivative 9.08 10
× /4 ( k ) ≤ Π '3 ( λ h ) ≤ 2.33 / 4( k ) and the linearity is

− 31.14% ≤ η 3λ h
≤ 25.04% . Therefore, the relationship between h λ and 3 Π has positive
 ACCEPTED MANUSCRIPT

) (α( α( −β αβλ αβλ

2
− 1)
+ 1)( β − 1) (c h − −β −2
2 2
•• ( Hkq EV 2
1)( u u
••

αβ λc (
2
• − (−α 1) 2( λ l
− ) λ l ) c2(h +− β 1)(( α − 1) 1c l • ( λ l − −) u u
− 1)
• •
• − ( c ntr − 1−) ( αβλ αβλ +α 1) 1( λ − ) λ ) ) c( − −α
2
1) c l
2
+ −( u u •
l l h

• •
− αβ c + −β cα λ + −c ntr l λ αβ λ β λ c uαλ
22 22 22 2
+ cu l −
2 2 2 2 2
c ntr l
• u u l l •
•
+ 2α λ c ntr − 2 c ntr l λ αβ
2
− λ2 cul + 2 β cλu l + 2 αβ c ntrcu l λ
2 2 2 • 2

•
l
•
• − 2 β c c λ λ +β−αβ
ntr u l (
2
− 1 ) ( c
l
2

ntr
− 1 )
22
λ − 2 ( α − 1 ) β λ
c lu
2
• u
• •
• + 2 αβ λc uβl λ − 2 c u l 1 ) c lcntr l λ αβ + 2λ
2
cu l
2
− 2 ( α − 1 ) c λ + −2 ( α l l
•
• •

T
− 2 β cλu αβ − 2 c ntrcu l λ + 2 β c ntrcu l λ + −2 ( α 1 ) β c l cu ( λ l − 1 )
• l •

IP
• •
− 2 ( β αβ − 1) c2u uλ( + β α β − 1) c ntrcu u λ + −2( α β λ1)λ c ntr l u
2
N 130 = ⋅ ( λ − 1)
• u
•
• − 2 (α − 1 ) β c c λ λ λ λ β c c + 2 •

CR
l ntr l u
+ −2 ( α 1 ) (
β c + −
1 ) lu
2 ( α − 1 ) ntr ( u u
)
• •
• ⋅ λ lλu + −2( α β λ1) ( c − (λ− α β λ 1) ( c λ + − α β 1) c l
2
+ 1) l l
− 2) 2( u
• l u u

• •
λ − 1 ) λ θ) 4( + Hkq EV 2 ) (( α αβ ) c + −α β(
2
·c 1) ( c ntr − δ −m 2 ) c h •
2

• ntr (

US
u u
− h

• •
− (−α 1 ) c + −αβ αc δ λ c ntr m l + c ntr mδl λ αβ+δ λ c ntr m u 2 αβ c u
2 2
−+
2 2 2
c ntr
• l u
•
• − αβ c c + αβ δc uδm α λ+ − m + c ntr l λ α+λ− c ntr l
2
c ntr l λ αδ
2
+ λ ml •
AN
ntr u
c ntr l
• •
• +α δ c ntrλm l − c ntr mδl λ δ λ − α m l + −( 1 ) c ( δ mλ ( − 1+) c ntr ( λ l + −1)2 ) •
relationship. l l

• •
• + αβ λc ntr αβu λ −αβ c ntr u − cntrcu u λ αβδ − λ mu
M

2
•
• •
• − αβ δ λ mαβ
c ntr u
δ λ− cu m u + 1) ••
•
(D.31)

• − 2 α( β θ − 1) ( Hkq EV 2
) (c(hHkq
(θ • EV 2
) (( α − 1) c( l λ l − 1)
+ c ntr
•
ED

•
⋅ ( α( β −α +αλ−
) λ αβλ + l + + (− α β
− 1) ( c u + 1) λ uβ + cu + λ l )
• l u
•
• 2 •
− λ−l ) 2 c ntr + θ ( Hkq EV 2 ) (α( β − 1) + −1) 2 ) ( δ m + − α θ 1) ( Hkq EV 2 ) c l
• •
T

2
• −− (α 1 ) θ( Hkq EV 2 ) c ( l c ntr − 1 ) ( λ l + 1 ) + αθ ( Hkq EV 2 ) c ntr λ lαθ − 2 •
N = ⋅• • (D.32)
EP

131 2
• ( Hkq EV 2 ) c ntr λ lθ− ( Hkq EV 2 ) c ntr λ lθ+ 2 ( Hkq EV 2 ) c ntr λ lαβθ − •
• •
⋅ ( Hkq EV 2 ) ( c ntr − 1) λ( u c ntr − − c u+ 1) αβθ ( Hkq EV 2 ) c ntrc u
• •
C

• − αβθ ( Hkq c u − αβθ ( Hkq EV 2 ) c u − 2 c ntr mδ

2
− 2 ( c ntr + δ m ) •
EV 2 )
• •
AC

•
+ 2 c
2
+ αθ ( Hkq ) λ θ − ( Hkq ) λ ) •
• ntr EV 2 l EV 2 l •

α β( θ − 1) ( (− Hkq ) ) (( α θ− 1) ( (− Hkq α θ 1) •
2
•
EV 2 EV 2
) ) c2(l +−
• •
)c
( lc ) ((α( β
2
N 132
= ⋅• ( Hkq EV 2 ntr
− 1)
+( θ Hkq EV 2
− 1)
+ 1)( c ntr − 1) • (D.33)
• − 2 αβ ( c ntr − 1) c u + αβ c u
2
) 4+ c ntr mδ ) •
• •
2
D 130 =−
θ ( ( Hkq EV 2 )) +
( θ Hkq EV 2 ) ( α − 1 ) ( λ l − 2 ) λ l αβλ
− u
+ 2 αβλ u )− 4 (D.34)

D 131 = − 2 α β( − 1 ) θ( Hkq EV 2 ) (D.35)

D 132 = α β( − 1 ) θ( Hkq EV 2 ) (D.36)

 ACCEPTED MANUSCRIPT

Relationship between u λ and 3 Π

2
N 140
N + λ + N 142 λ u
Π 3
(λ u ) =
141 u
(D.37)
2
4 ( D140 +
kD 141
λ u
+ D 142 λ u )

− 4≤ N 140
≤ 20
−8
− 16
≤ N 141
≤ − 9.18 10
×
−8
− 9.09 10
× ≤ N 142 ≤ 4
(D.38)
− 5≤ D 140 8 ≤ − 4

T
−
− 9.17 10
× ≤ D 141 ≤ 2

IP
−8
− ≤1
••••••••••• ×
D 142 ≤ − 9.10 10

CR
−8
We find The first derivative 6.01 10
× Π' ≤ 3 ( λ u ) ≤ 2.33 / 4( k ) and the linearity is

0.02% ≤ η 3λ u
≤ 24.97% . Therefore, the relationship between u λ and 3 Π cannot be treated as

US
22 22 222 22 22
α α −β α cλu β α λ− β αλ
c ntrβh + 2 α λc ntr
βh •
2 2
• ( Hkq EV 2 2 2
)( − cuh − h
•
22 2
•
• − 2 α λc uβh + c ntrc u h λ β + 2 α λc uβh + 2 α λc uβh − 2 α c ntrc u h λ β
2 2 2
2α •
AN
• 22 2 •
+ c u β α+λ2β − c ntr hλ β α λ+ β αλ
c u hβ α λ β
+ − 4 c ntr h
2 2 2 2 2 2
c ntr h 2
• h
•
• + 2 c ntr hλ β α λ+ β2 α
2
cuh
2
−2 c ntr c u hλ β λ β− α λh β α− λ2β
2 2 2
cuh − 2 cu h •
• •
M

2
+ 2 α c ntrc u h λ β α −λ + −c ntr hλ αλ + 2 α λc ntr h 2 c ntr hλ λ + •
2 2 2 2 2 2 2
• c ntr h h
− h
• 2
•
− (−α 1)(( c ntr − 1)
− 2( c uc − 1)
+ β λc uα) 1) c l
2 2 2

• β ntr l
+ −( •
•
⋅ (( β − 1)( λ − 2) λ − 1)
+ 2( α − 1)( ( β c u − c+ntr+ − −c u 1) ( β 1) •
h h
•••••••••••
2 2
· 1c ntr − ) λ λ)β( l − − h
1) c h ( α β( − 1)
− −( α 1) 2( λ l
− ) λ l + 1)
N 140 = (
ED

+ −2 (α 1 ) c( l c( ntr
− 1 ) ( λ β− −( l
1 ) λ( hλ λ h
− − l+ 1 ) ) c u ( β βλ
− l ))

+ −2( β 1) c( h − αβ λc u ( h
− 1)
− −( α 1) c( l λ l − 1)( λ h − 1)
+ ( c ntr − 1)
T

·( λ hα (β − 1)
+ −( α λ 1) 1l ( + − )− ) λ θ) ) ) 4( + Hkq )
2
( α 1) 1( λ − EV 2
EP

l l

2
⋅ (( α αβ
− )c h + −α β
( 1)( ( δ m
λh − 1)
+ c ntr ( λ h
+ −1) 2) ch
2
− (α − −α 1) c ntr mδl λ − +c ntr+ δ mα β + −β +( + c ntr δm 2) c u
2 2
− 1) (c l ( cu
C

•• + −
(β 1) λ hδ ( + c ntr ( − c+ntr− δ mδ λ − ))−)(
+1 α 1)( 1( c ntr − ) c ntr
AC

m mh
•−
positive linear.
•• − ( c ntr + 1) δ mλ )α( l
+− 1) c( l ( δ mλ l
− 1)
+ c ntr ( λ l + 1)
− +2) 1)
2 ••••••••••••••••
4( c ntr − •δ m
) •

(D.39)

2
2 αβθ ) ( (α β − 1) λ h • 1)(( c ntr − 1)( α
( β
2 2
• ( Hkq EV 2
− 1) (c h + − c ntr
• •
+ −( α λ 1) ) l + −c u ( αβ α β ( ( − 1) λ hλ +λ l ) + − l− 1 )) ( α β − 1) c h •
•
• 2 •
N 141 = ·( λ h ( c ntr − −c u+ +1−) + −c ntr cu 1 )( α 1) (c l − −α 1) c l (D.40)
• •
• ·( λ l ( c ntr − −c u+ +1−) + c ntr cu 1 )) 2 ( αβθ δ Hkq EV 2 ) •
• •
• · (2 c ntr ( c ntr − −
c u− 1 )2 ( c ntr + c u + 1 ) ) •
• m •
 ACCEPTED MANUSCRIPT

) (α( β θ − 1) ( (− Hkq ) ) c2 ( + − α β θ 1) ( Hkq )ch •

2
• αβθ ( Hkq EV 2 EV 2 h EV 2
• •
α 1) c( c
2
N 142
=
•
⋅ ( c ntr − 1)
+( θ Hkq EV 2
) (−( −α cl + −
1) 2( l ntr
− 1) • (D.41)
• − ( αβ − 1)( c ntr −
2
1) ) 4+ c ntr mδ ) •
• •
2
D 140 = θ ( Hkq EV 2 ) (α β − 1) (λ h
− 2 ) λ hα+ −( 1)λ l
− 2 ( α − 1 ) λ l θ) + − ( ( Hkq EV 2 )) − 4
(D.42)

D 141 = 2 αβθ ( Hkq EV 2 ) (D.43)

D 142 = αβθ ( − ( Hkq EV 2 )) (D.44)

T
Relationship between ntr c and 3 Π

IP
CR
N 150
N c+ + N 152
c
2
ntr
( c ntr ) =
151
Π 3
(D.45)
4 kD
ntr
150

US
− 4≤ NN
150
≤ 20

− 16
≤ N 151
≤ 8
•••• (D.46)
AN
−7
− 4≤ ≤ − 3.34 10
×
••• 152

− 5≤ D 150 ≤ − 4
M

We find The first derivative − 2 / 4( k ) ≤ Π '3 ( c ntr ) ≤ 4 / 4( k ) and the linearity is

c and 3
7
1.24 10 η 3c 1.82 10
7
− × ≤ ≤ × . Therefore, the relationship between n Π cannot be treated as
ntr

linear.
22 222 22 22 2
α α −β α cλu β αλ−β − 2 α λc uβh + 2 α λc uβh •
2 2
• ( Hkq EV 2
)( cuh − h
• •
ED

22 22
+ 2 α λc uβh + β α+λ βcαλ β α λ+β2λ β αh λ β+λ 2β αcλu hβ − cu l
2 2 2 2 2

• cu uh
− h •
• 22 2 •
+ c ul + 2 c u lλ β αβ λ)β αu λ β− α2λ βc u h − 2 cu h
2 2 2
− 2 cul + −( 1
• •
•
T

• 2
+ 2 α λc uβl − 2 c u lλ β α+λ 2β c u lλ β αλ− 2 λ β λ +λ β
2
cul − 2 hl
2 hl
• •
EP

• + 2 (α β( − 1 ) λ α + −( 1 ) λ + −c +( αβ α β ( − 1 ) λ α + −( 1 ) λ l + 1) ) •
h l u h
• •
• ⋅λ β αλ− λ αλhλ + −
2 2 2
+
2
+ 2 αλ λh l − 2 λ hλl β − −( 1) c( h αβλ
2 2
•
u h l l u
•
C

•
2
• − 2 αβλ αu β+ − 1)
− −( ) λ + 1) (+ − α 1) c( l βλ + 2 βλ
2
( α 1) 2( λ l
− l
− u u •
N 150 =
AC

• 2
•
•
+ −( β 1) 2( λ h
− )λ h
− 1)
+ 2( α − 1) c( l − βλ βu λ+ ( l
+ 1 ) λ u + −( β 1) λ h •
• 2 •
·( λ hλ − − l− 1 ) λ l + β λc u ( ) (λ u − 1
+)) 2( 1) c( h( αλ λl αλ
−−l
2
− 1 β −
• l l •
•
− αλ λh lλ λ+λ αβλ αλ
+ −αβλ λα
2
+− − − h− 1) cl ( λ h − 1) •
hl
•
l u h h
•
• ·( λ l − 1+) αβ λc u ( − 1 ) ( λ u − 1+ ) αβ λ( + 1 ) λ uθ) ) 4( +
2
Hkq ) •
h h EV 2
• •
2
α αβ )c h + −α β 1)( ( δ m
λh − 1)
− 2) ( − −α + αβ + 2 αβ c u
2 2
• ⋅ (( − ( ch 1) c l cu •
• •
•
+ αβ δcδu mαδ +λ −αβδ
m
λα mh
+ mh
+ −( 1 ) c ( δ mλ (
l l
−1
− )2+ ) αδ λm l •
• •
• − δ mλ l αβ− ( cu + 1) δ mλ u + 1) •

(D.47)
 ACCEPTED MANUSCRIPT

( c( u αβ α −β λ hλ +
λ •
2
βλ ) +−
2
• 2 αθ ( Hkq )( (( − 1) 1)
EV 2 u l l
• •
• − 2 α( β λh −
− 1) 2( α λ− 1) ) l( − −α 1) c((l β − 1) λ hλ (λ h − − l 1 ) •
• 2 •
− λ+lβ λ ( l + −1 ) λ uβλ ) (− −β 1) c( h( (λ hα β α −αλ+ λ− ) + l
• u l •
•
+ αβλ u − 1) + −( α 1) 1( λ l − ) λ lαβλ − αβλ uαβ+λ
2
)+
2
cu h
2 •
•
u
•
• − αβ λ αβ
2
c u hλ αβ− λ α c uh
2
+ c u h + β λc uαβ
2
λ −β λ β λ
c − c
2
+ c •
l ul ul ul
N 151 = • • (D.48)
22 2
+ αβ λ h − 2 αβλ αλ βλ
+ λ h +−+h 2 αβλ λhαλ λ− 2 − 2 βλ λh l
2 2 2
• h h l hl •
• •
+ 2 λ hλl αλ+λ β αβ − +l + ( θ Hkq EV 2 ) (α( β − 1) c h ( λ h + 1 ) •
2 2 2
( − 1) λ u ) 4
• l

• •
+ −( α 1) c l ( λ l + −1 ) αβλ u ( c u + δ m + −1 ) αβ c u + −αβλ αλ

T
•
h h
•
•

IP
2 2
• − αβλ δh αβλ δ αλ δ αλ
+ δ αλ
+ −λ hm − αλ δl mαλ δ+
2

m hm hm
+− l l lm
• •
•
+ λ lδm λ δ −αβδl mλ +
2 2 1)
− •
• •

CR
mu

2
• θ ( −( Hkq EV 2
) ) (α( β θ − 1) ( Hkq − 1) λ h •
EV 2
) (α( β − 1)
+ 1) 2 λ( h
− αβ
• •
⋅ (( α θ− 1) ( ( − Hkq EV 2 ) ) λ lαβθ
+ ( Hkq EV 2 ) λ u − +
2)−( α λ 1) l •

US
•
N =• (D.49)
152
⋅ ( αθ ( Hkq EV 2 ) λ l + −4) 2 αβλ αu θ(( − 1) ( Hkq EV 2 ) λ l + 2) •
• •
• 2
+ αβθ ( Hkq EV 2 ) ( αβ − 1) λ u ) 4− •
• •
AN
− 2) λ l •
2
• θ ( Hkq D
EV 2
) (α( β − 1) λ2 h( − αβ − 1) (λ hα + − 1)( λ l
150
=• • (D.50)
− αβλ + 2 αβλ θu ) (+(− Hkq ) ) 4−
2

• •
M

u EV 2

Relationship between l c and 3 Π

N 160
N c+N c161c
2
+
Π 3 ( )l =
l 162 l
(D.51)
4 kD 160
ED

••••••• − 4≤ N 160 8 ≤ 20
−
9.08 10 ≤ 12
T

× ≤ N 161 8

−
(D.52)
EP

8.28 10
× N
≤D 162
≤ 4

− 5≤ 160
≤−4
C

−8
We find The first derivative − 4.23 / 4( k ) (≤ ) Π '3 cl ≤ − 9.13 10
× /4 ( k ) and the linearity is
AC

− 24.99% ≤ η 3c l
≤ − 6.25% . Therefore, the relationship between l c and 3 Π cannot be treated as
negative linear.
 ACCEPTED MANUSCRIPT

1) c( h αβ λc u λ u − 1) •
2
− ( c ntr − 1)
2
• αθ ( Hkq EV 2
) (2( β − ( h
− 1)(
•
2
•
• ⋅ ( λ( h(α β α −αλ+ λ−
) αβλ l
+ l
+ u
− 1)
+ −( α 1)( λ l
− 1) λ lαβλ
− u •
•••••••••••• 2
+ αβλ u
t ) ) 2+ β ( c ntr − 1 ) c u ( α( β − 1) λ hα β
− ( − 1) λ1 hλ( ( u
+ ) + −α 1)
2
·( λ l − 1) λ lλ +αβ αλ
u
(λ − l
+ − l− 1 ) ( c ntr − 1 )(( β − 1)( α( β − 1)
2 2
+ 1) λ h
− 2( β − 1) λ hαλ( −λ +αβλ
+l l u
) + −( α λ 1) 2( l − α βλ λ
− 1) lu

2
+ β αβ ) (− −β 1) c( h( α β − 1)
− −( ) λ αβλ
+
2 2
( − 1) λ u
α 1) 2( λ l
− l u

2
N 160 = − 2 αβλ + 1) + βc Hkq( − αβ
+ −α αβ( λ ) 2 h( αβ 1) (λ α− − 1) 2( λ l −
2

u u
+− h
)

T
2
· λ l + 1)) 4+ ( α αβ ) c hc − ( αβ λc u ( + 1 ) + −α β( 1) λ hδm
2
θ EV 2
) (( − ntr u

IP
2
− α β( − 1) λ hδ ( m
+ 1 ) + ( α λ−λ1) (( − 1 )δ −)
−1 αβδ λm u
•• + l l m

CR
2
αβ δ( m
+ 1 ) λ u + 1) + c ntr ( ( α αβ
− λαλ
) ( αβλ
h
− α− β 1) l
+ u )+− ( 1)
••⋅
2
c hc( ntr
(λ h
+ 1) (+ − λ h
1) δ m
− −2) αβ δc λu mαβ
u
δ αβ
+ cum + cu
•+

US
••• − 2 αβ αβλ
c u +δ αλ δ αλδ
hm
λδ
− δ αβδ
hm
λ+− lm lm
+− m mu
+ 1)
2
4( c ntr − •δ m ) ••••••••••••••••••••
• (D.53)
AN
α αθ λ l − 1) c ntr − 1) •
2
) ( ( − −β 1) c( h λ − (
2
• 2( − 1) ( Hkq EV 2 h
− 1)(
• 2 •
•
⋅ (( β − 1) λ( hλ λ h
− − l− +1) λ lβ λ ( l
+ −1) λ uβλ u
) •
N =• (D.54)
M

) •
161
+ β λc u ( l
− 1) (λ u − 1
+)) 4( α θ− 1) ( Hkq EV 2
) ( c ntr λ
( l + 1
• •
•
+ −( λ 1)δ − 2) •
• l m •

• ( α θ− 1) ( Hkq EV 2
) ( αθ ( Hkq EV 2
) (( β − 1)( λ h
− 2) λ h •
N =• • (D.55)
βλ αθ
162
Hkq )) − 4)
2

• − βλ + 2 ) (+(− •
ED

u u EV 2

− 2) λ l •
2
• θ ( Hkq D
EV 2
) (α( β − 1) λ2 h( − αβ − 1) (λ hα + − 1)( λ l
160
=• • (D.56)
− αβλ + 2 αβλ θu ) (+(− Hkq ) ) 4−
2
T

• u EV 2 •
EP

Relationship between h c and 3 Π

C

N 170
N c+N c171c
2
+
Π 3 ( )h =
172 h
(D.57)
4 hkD 170
AC

••••••• − 4≤ N 170 8 ≤ 20
−
8.57 10
× ≤ N 171 8
≤ 12
−
(D.58)
6.89 10
× N
≤D 172
≤ 4

− 5≤ 170
≤−4

−8
We find The first derivative − 4.23 / 4( k ) (≤ ) Π '3 cl ≤ − 8.43 10
× /4 ( k ) and the linearity is
− 24.91% ≤ η 3c h
≤ − 6.26% . Therefore, the relationship between l c and 3 Π cannot be treated as
negative linear.
 ACCEPTED MANUSCRIPT

α − 1) c( l (c β • λ λ − 1)
− ( c ntr − 1)
2 2
• αθ ( Hkq EV 2 ) (2( u l
− 1)( u
• •
• · (( β − 1) λ hλ (λ h − − − +1 ) λ β λ ( + −1 ) λ βλ
2
l u )
) 2+ β • l l u

• •
• ·( c ntr − 1 ) c (α( β − 1) λ α −β ( − 1) λ1λ(( 1) 1( λ − ) λ l •
2
u
+ ) +−α h h u l

• • 2

• + λ αβ ( αλ λ c ntr − 1 )(( β − 1)( α( β − 1) + 1) λ h

2
u
− + − − 1 )) ( l l •
• •
− 2( β − 1) λ αλ ( −λ +αβλ + −( α λ 1) − 2( α βλ − 1)λ
2

•
+ h u ) l l lu • l

• •
) λ βλ− + −2 βλ u
2
+ β αβ 1) (c ( β − 1) 2( λ − 1)
2 2
( − 1) λ ) (+ − α
• u
•
l h h u

• •
( − αβ
+ −α αβ ( λ )2( +− α β 1) (λ α − − 1)( λ − 2) λ l + 1))
2
N 170 = + β c Hkq
2

• u h
• h l

T
• + 4θ( ) ( − c ( αβ λ
c ( + 1) ( + α
− β 1) λ δ
2
α β− ( − 1) λ •
hm h

IP
EV 2 ntr u u
• •
( − α λ λ1) (( − αβδ λm αβ δ+ ( m + 1) λ u + 1) •
2
• ⋅ ( δ m + 1) + − 1) δ m − 1) u l l

• •

CR
+ c ntr ( ( α αβ − λ α) 1 ) λ αβλ+ − − ) ( α 1 ) c + −( α 1 ) c l
2
− (−
2

• h • l u l

• •
2) αβ δc uλm αβ δ +αβ c u m +
2
⋅ ( c ntr ( λ + 1) + (−λ 1) δ m − − cu
• l l u
•

US
• + 2 αβ αβλ c +δ αλ δh mαλδ−λδ δh αβδ
u m
+λ − lm lm
+ − m mu
+ 1) •
•
2
•
• − 4( c ntr − δ m ) •
• • (D.59)
•
AN
•

2
• 4α β( − 1 ) θ( Hkq EV 2 ) ( c ntr ( λ h + 1 ) (+ − λ h 1)δ m
− −2 )2 α β( − 1)θ •
• 22
•
M

• ⋅( Hkq ) (( α − 1) c( l λ − 1)( λ − 1)
+ ( c ntr − 1)( λ( (hα β α − + ) •
N 171 = • EV h l
(D.60)
2 •
− αλ
+ +λl αβλ l − 1)
+ −( α 1)( λ − 1) λ lαβλ
− αβλ u + )
• u l u
•
•
− αβ λc u ( − 1)( λ u − 1)) •
• h •

• α β( θ − 1) ( (− Hkq EV 2
) ) (θ( Hkq EV 2
) (α( β − 1)
+ 1) •
N =• • (D.61)
ED

2
λ l αβλ
+ − αβλu
172

• + θ ( Hkq EV 2
) (−( −α 1)( λ l
− 2) 2 u
) + 4) •

− 2) λ l •
2
• θ ( Hkq D
EV 2
) (α( β − 1) λ2 h( − αβ − 1) (λ hα + − 1)( λ l
T

170
=• • (D.62)
− αβλ + 2 αβλ θu ) (+(− Hkq ) ) 4−
2

• u EV 2 •
EP

Relationship between u c and 3 Π

C

N 180
N c+N c181c
2
+
( )u
AC

182 u
Π 3
= (D.63)
4 ukD 180

••••••• − 4≤ N 180 8 ≤ 20
−
1.82 10
× ≤ NN
181 8
≤ 12
−
(D.64)
9.09 10
× D
≤ 182
≤ 4

− 5≤ 180
≤−4

−8
We find The first derivative − 4.24 / 4( k ) (≤ )Π '3 cu ≤ − 9.08 10
× /4 ( k ) and the linearity is
− 33.33% ≤ η 3c u
≤ − 6.29% . Therefore, the relationship between u c and 3 Π cannot be treated as

negative linear.
 ACCEPTED MANUSCRIPT

λ l − 1) •
2
1) c((h α − 1) c( l λ + ( c ntr − 1)
2
• αθ ( Hkq EV 2
) ( 2(
− β − h
− 1)(
• •
• ·( λ hα (β α( αλ
− +λ−
)αβλ l
+ l
+ u
−1
+ )(− α 1) 1( λ l
− )λl •
• 2 •
− αβλ αβλ+ ) ) 2(− α − 1) c l c( c ntr − 1 )(( β − 1) λ hλ (λ − − l− 1 ) λl
• u u h •
• 2 2 •
+ β λ( ) λ βλ ) 1 )(( 1)( α( β − 1)
+ 1) λh
2

• l
+ −1 u u
− ( ntr
− β −
•
•
1 ) λ( hαλ −λ +
αβλ
2 •
− 2 (β − +l u ) + (α − 1 )λ − 2 ( α − 1 ) βλ λl u
•
l l
•
• + β αβ
( − 1) λ
2
) (− −β 1) c( h( α β
2
− 1)
− −( α 1) 2( λ − ) λ αβλ
+
2
•
u l l u
• •
2
• − 2 αβλ + 1) (+ − α 1) c((l λ hβλ− βλ
2
β − 1)( λ − 2) + −2 u
1)) •
N 180 =
u h u

• •

T
2
•
+ 4θ( Hkq EV 2
) (( α αβ
− )c h + −α β
( 1) c( h(c ntr
λ h
+ 1) (+ − λ h
1) δ m •

IP
• 2 2 •
− 2)
− αβ λc δntr αβ λ δ α+λ δ αc ntr h m + − c nt r hλmδ αβ −λ c ntr h
2
c ntr h m
•
hm
•
• + αβ λc αntr λh α λ+α
2
c ntr h c ntr h −−
( 1) c l
2
α λc ntr
δ α lλm δ +
2
c ntr •

CR
− − lm
• •
2
• + −( α 1 ) c c( l ntr ( λ l + 1 ) (+ − λ l 1 )δ m
− +2 ) c ntr λ lδm − c ntr λ lδm •
•
2
•
α λc ntr
αλl + λ lαβ
+ δ λ c ntr m u
2 2
+ −c ntr λ

US
• − c ntr l l
c ntr •
• •
+ αβλ ( ( c ntr − 1 ) c ntr − ( c ntr + 1 ) δ ) − +c ntr αβλ δh αλ δ−
• u m m hm •
•
+ αλ l δ λ δ− δ + + 1) •
AN
• m lm m • ( D.65)

2
2 αβθ c ntr 1)( α( β λ hα β 1) λ( hλ + 1) •
2 2
• ( Hkq EV 2
) (( − − 1) − ( − u
• •
M

•
+ −( α 1) 1( λ l
− ) λ λ +αβ αλ( λ
l u
− l
+ − l+ 1 ) ( αβ − 1) ch ( λ h − 1) •
N =• (D.66)
181
·( λ u − 1
+ )(− α 1) c l (λ l − 1 − )) 4 ( αβθ
) (λ u − 1 Hkq ) •
EV 2
• •
• δ m−
·( c ntr ( λ u + 1 ) + − λ( 1 )2 ) •
• u •

•
2
• αβθ ( Hkq EV 2
) (θ( Hkq EV 2
) (−α β
( − 1)( λ h
− 2) (λ hα−λ − 1) l
N =• • (D.67)
ED

182
• 2(
+− α λ 1) ) l ( + Hkq EV 2
) (θ αβθ
− ) + 4) •

− 2) λ l •
2
• θ ( Hkq D ) (α( β − 1) λ2 h( − αβ − 1) (λ hα + − 1)( λ
T

EV 2 l
180
=• • (D.68)
− αβλ + 2 αβλ θu ) (+(− Hkq ) ) 4−
2

• •
EP

u EV 2
C
AC