^{1}

^{*}

^{1}

^{1}

^{1}

^{1}

^{1}

^{1}

^{1}

Depositing an antireflection coating on the front surface of solar cells allows a significant reduction in reflection losses. It thus allows an increase in the efficiency of the cells. A modeling of the refractive indices and the thicknesses of an optimal antireflection coating has been proposed. Thus, the average reflective losses can be reduced to less than 8% and less than 2.4% in a large wavelength range respectively for a single-layer and double-layer anti-reflective coating types. However, the difficulty of finding these model materials (materials with the same refractive index) led us to introduce two notions: the refractive index difference and the thickness difference. These two notions allowed us to compare the reflectivity of the antireflection layer
** **in silicon surface. Thus, the lower the refractive index difference is, the more the material resembles to the ideal material (in refractive index), and thus its reflective losses are minimal. SiNx and SiO
_{2}/TiO
_{2} antireflection layers, in the wavelength range between 400 and 1100 nm, have reduced the average reflectivity losses to less than 9% and 2.3% respectively.

Antireflection coatings (ARC) are used in processing solar cells to reduce reflection and to offer better passivation properties [_{0})1/2, where n_{0} is the refractive index of the surrounding medium. Single-layer thin film antireflective coatings are limited by the availability of materials with required refractive indices [

In this study, a numerical optimization of the antireflection proprieties of simple layer and double layers are performed by modeling the refractive index and thickness of adequate materials. Thus, to better evaluate the reflectivity on the surface of a cell coated with a single or double antireflection layer, two new concepts will be introduced: the refractive index difference (RID) and the thickness difference (TD). RID and TD values will help to predict how the performance of a solar cell panels will be affected from reflectivity losses.

The matrix method for calculating spectral coefficients of the first layered media was suggested by F. Abeles (1950) and has been ever since widely employed [_{w} due to the solar spectrum’s long range of wavelength from 400 to 1100 nm.

For a PV cell, it is important to have a minimum of reflection over a whole spectral range. The average residual reflection factor is defined by [

R m = 1 λ max − λ min ∫ λ min λ max R ( λ ) d λ (1)

where λ max and λ min are the maximum and minimum values of the wavelength range respectively. R ( λ ) is the reflection factor. In order to compare the effectiveness of an AR coating on a solar cell, it is important to take the AM 1.5 solar spectrum into account. For this investigation, we have chosen the wavelength range from 400 nm to 1100 nm of the solar spectrum. The reason is that for wavelengths shorter than 400 nm, the spectral power density in the AM 1.5 spectrum is almost zero, while photons with wavelengths longer than 1100 nm are hardly absorbed by the Silicon.

The reflection R ( λ ) can be calculated by the transmission matrices of the ARC layers. It depends on both n ( λ ) and k ( λ ) . For each configuration, a transfer matrix method [

considered as composed of two layers (therefore, three interfaces) on a Si substrate (

First, we describe the characteristic matrix of a single layer. The relationship of matrix defining the problem of single antireflection layer is given by the following relation

( E 0 H 0 ) = M ( E ( S i ) H ( S i ) ) (2)

M is a matrix given by:

M = ( cos φ i ⋅ sin φ n a r c i ⋅ n a r c ⋅ sin φ cos φ ) (3)

The characteristic matrix of a multilayer is a product of corresponding single layer matrices. For a double stack antireflection coating, relation 2 become:

( E 0 H 0 ) = M 1 ⋅ M 2 ( E ( S i ) H ( S i ) ) = M t ( E ( S i ) H ( S i ) ) (4)

where

M t = ( cos φ i ⋅ sin φ 1 n a r c 1 i ⋅ n a r c 1 ⋅ sin φ cos φ 1 ) . ( cos φ 2 i ⋅ sin φ 2 n a r c 2 i ⋅ n a r c 2 ⋅ sin φ 2 cos φ 2 ) (5)

with i 2 = − 1 , n a r c 1 , n a r c 2 represent respectively the refractive index of the top and the bottom antireflection layers. φ 1 and φ 2 are respectively dephasing between the reflected waves of layers k and k + 1.

φ 1 = 2 π λ ⋅ n a r c 1 ⋅ e a r c 1 , φ 2 = 2 π λ ⋅ n a r c 2 ⋅ e a r c 2 (6)

The detailed derivation of amplitude reflection (r) and transmission (t) coefficients is given in [

r = n 0 ⋅ M 11 + n 0 ⋅ n S i ⋅ M 12 + M 21 − n S i ⋅ M 22 n 0 ⋅ M 11 + n 0 ⋅ n S i ⋅ M 12 + M 21 + n S i ⋅ M 22 (7)

t = 2 n 0 n 0 ⋅ M 11 + n 0 ⋅ n S i ⋅ M 12 + M 21 + n S i ⋅ M 22 (8)

M_{ij} are the elements of the characteristic matrix of the multilayer. The energy coefficients (reflectivity, transmissivity, and absorptance) are given by:

R = | r | 2 (9)

T = n S i n 0 | t | 2 (10)

A = 1 − R − T (11)

with n_{Si} and n_{o} is refractive index of the silicon and vacuum respectively.

In order to reduce reflection, a film with intermediate index of refraction can be applied according to _{Si}) is 1/4 of the wavelength (l), the phase difference becomes π and the two reflected waves cancel out. For complete annihilation, the amplitude of the interfering radiation also has to be identical. Under normal incidence, this requirement is fulfilled when the refractive index of the film is equal to the square root of the refractive index of the substrate [

e c a r = λ 4 n a r c (12)

n c a r = n 0 ⋅ n S i (13)

l is wavelength (m), n_{0} is the air refractive index and it taken equal to 1. n_{car} and n_{Si} represent refractive indices of AR layer and silicon respectively. Equation (12) and Equation (13) permit us to obtain the following figure.

e ( λ ) = 0.144 λ − 11.12 (14)

By replacing this thickness in Equation (2), the optimal refractive index is then given by:

n a r c = λ 4 e c a r = λ 0.576 λ − 44.5

Relation which can be written as follows:

n a r c = A + B λ − C

with A = 1.736, B = 134.12 and C = 77.26

n ( λ ) = 1.725 + 134 λ − 77.3 (15)

We find the Cornu equation [

The refractive index can also be found by extrapolation of the curve 2-5. We obtain the Cauchy equation [

Following the Cauchy equation, the optimal refractive index of a single ARC is given by the following expression:

n ( λ ) = 1.856 + 2.79 × 10 4 λ 2 + 5.73 × 10 9 λ 4 (16)

These results are summarized in

It can be seen from

For a double layer coating for a zero reflection minimum in a large wavelength

Simple layer ARC | Cornu | Cauchy |
---|---|---|

Ref. ind. n a r c = n 0 ⋅ n S i | n ( λ ) = 1.725 + 134 λ − 77.3 | n ( λ ) = 1.856 + 2.79 × 10 4 λ 2 + 5.73 × 10 9 λ 4 |

thickness e a r c = λ 4 n c a r | e ( λ ) = 0.144 λ − 11.12 | e a r c = 0.135 ⋅ λ ⋅ 1 1 + 2.5 × 10 4 λ 2 |

wavelength | l = 500 | l = 600 | l = 700 | l = 800 | l = 900 | l = 1000 |
---|---|---|---|---|---|---|

thickness | 61 | 76 | 90 | 104 | 118 | 132 |

Refractive index | 2.05 | 1.98 | 1.94 | 1.91 | 1.90 | 1.89 |

Reflectivity | 0.0002% | 0.0016% | 0.0008% | 0.006% | 0.0015% | 0.0005% |

range, the refractive index of the two layers have to fulfill these following relations [

n a r c 1 = ( n 0 2 n S i ) 1 / 3 , n a r c 2 = ( n 0 n S i i 2 ) 1 / 3 (17)

And the optimal thicknesses of the top and bottom layers are respectively equal to:

e t o p = λ r e f 4 ⋅ n t o p , e b o t = λ r e f 4 ⋅ n b o t (18)

According to Equation (5) and Equation (6), the refractive index and thicknesses of an optimal double antireflection layer are given respectively by

According to the curves (3) on the right, these thicknesses are governed by the following equations:

e t o p = 0.173 ⋅ λ − 8.8 (19)

e b o t = 0.117 ⋅ λ − 9.74 (20)

The thicknesses and wavelength are in nanometer (nm). From relation (6), the optimal refractive index is given by the following expression:

n t o p = λ 4 ( 0.173 ⋅ λ − 8.8 ) = λ 0.692 ⋅ λ − 35.2 (21)

n b o t = λ 4 ( 0.117 ⋅ λ − 9.74 ) = λ 0.468 ⋅ λ − 38.96 (22)

These relations can be written in the form of the dispersion equation giving the Cornu refractive index:

n b o t = 2.137 + 177.9 λ − 83.25 (23)

Besides by extrapolation of the left

These results are summarized in

In order to better search for the ideal materials for an antireflection layer on the silicon surface, we introduced the notions of refractive index difference (RID) and thickness difference (TD). Thus, the refractive index difference of a material at the wavelength λ is called the spectral refractive index difference (SRID). Over the whole range of the spectrum considered, we speak of difference of total refractive index difference, which is the sum of all the difference of spectral refractive index on the spectrum [400 - 1100 nm].

It is defined as the difference between the refractive index of the antireflection

ARCs | Exact valeurs | Cornu model | Cauchy model |
---|---|---|---|

Top ARC | n t o p = ( n 0 2 n S i ) 1 / 3 | n t o p = 1.445 + 73.5 λ − 50.87 | n t o p = 1.507 + 1.86 × 10 4 λ 2 + 2.2 × 10 9 λ 4 |

Bottom ARC | n b o t = ( n 0 n S i 2 ) 1 / 3 | n b o t = 2.137 + 177.9 λ − 83.25 | n b o t = 2.259 + 6.345 × 10 4 λ 2 + 6.49 × 10 9 λ 4 |

Thickness | e t o p = λ r e f 4 ⋅ n s u p e b o t = λ r e f 4 ⋅ n s u p | e t o p = 0.173 ⋅ λ − 8.8 e b o t = 0.117 ⋅ λ − 9.74 | e t o p = 0.166 ⋅ λ 1 1 + 1.23 × 10 4 λ 2 + 1.46 × 10 9 λ 4 e b o t = 0.111 ⋅ λ 1 1 + 2.8 × 10 4 λ 2 + 2.87 × 10 9 λ 4 |

layer n(λ) and the optimal one, given by the phase condition at the wavelength λ.

Δ n ( λ ) = | ( n 0 n S i ) 1 / 2 − n A R C ( λ ) | (25)

with ( n 0 n S i ) 1 / 2 = n ( λ ) = 1.856 + 2.79 × 10 4 λ 2 + 5.73 × 10 9 λ 4 = 1.725 + 134 λ − 77.3

Δ n ( λ ) can be given by the following relation:

Δ n ( λ ) = | 1.856 + 2.79 × 10 4 λ 2 + 5.73 × 10 9 λ 4 − n A R C ( λ ) | (26)

Relation (15) is the spectral refractive index difference (SRID).

To obtain the total refractive index difference, we sum on all the wavelengths of the spectrum considered. The expression of the total refractive index difference (or refractive index difference) is then given by the following relation:

Δ n = | ( ∫ 400 1100 1.856 + 2.79 × 10 4 λ 2 + 5.73 × 10 9 λ 4 − n A R C ( λ ) ) d λ | (27)

Let Δ n 1 ( λ ) be the refractive index difference between the upper layer of refractive index n 1 ( λ ) and that of optimal refractive index for a layer of the same position and Δ n 2 ( λ ) the difference between the refractive index of the bottom layer ARC and that optimal for a layer of the same position (described by relation 13). Then, the difference in refractive index for a double AR layer at a wavelength is given by the average of Δ n 1 and Δ n 2 .

Δ n 1 ( λ ) = | n t o p ( λ ) − n A R C 1 ( λ ) | (28)

Δ n 2 ( λ ) = | n b o t ( λ ) − n A R C 2 ( λ ) | (29)

Δ n ( λ ) = Δ n 1 ( λ ) + Δ n 2 ( λ ) 2 (30)

n t o p ( λ ) and n b o t ( λ ) represent respectively the optimal refractive indices of the top and the bottom antireflective layers. And n 1 ( λ ) , n 2 ( λ ) represent respectively refractive indices of the chosen antireflective coatings of top and bottom layers

Taking into account relations 20 and 21, it comes:

Δ n 1 ( λ ) = | 1.507 + 1.86 × 10 4 λ 2 + 2.2 × 10 9 λ 4 − n A R C 1 ( λ ) | (31)

Δ n 2 ( λ ) = | 2.259 + 6.345 × 10 4 λ 2 + 6.49 × 10 9 λ 4 − n A R C 2 ( λ ) | (32)

Relation (19) is the spectral refractive index difference, which depending only on the wavelength and the refractive index of the bottom ( n 2 ( λ ) ) and top ( n 1 ( λ ) ) layers. Over the entire solar spectrum, the refractive index difference (RID) is defined by the sum of all differences of spectral refractive index.

Δ n 1 = | ∫ 400 1100 ( 1.507 + 1.86 × 10 4 λ 2 + 2.2 × 10 9 λ 4 − n A R C 1 ( λ ) ) d λ | (33)

Δ n 2 = | ∫ 400 1100 ( 2.259 + 6.345 × 10 4 λ 2 + 6.49 × 10 9 λ 4 − n A R C 2 ( λ ) ) d λ | (34)

Δ n = Δ n 1 + Δ n 2 2 (35)

If the number of layers is greater than 3, the refractive index difference will be defined in another way. Consider a stack of N layers deposited follow that order N , ( N − 1 ) , ⋯ , 2 , 1 with refractive indices n N , n N − 1 , ⋯ , N 1 , respectively. The spectral refractive index difference for a multilayer coating is defined by the following equation:

Δ n = | n 1 ( λ ) − n 0 | + | n 2 − n 1 ( λ ) | + ⋯ + | n S i ( λ ) − n N ( λ ) | = Δ 1 + Δ 2 + ⋯ + Δ n (36)

In the case where the study is done on all the wavelengths, refractive index difference (RID) is defined by the following relation:

Δ n = ∫ 400 1100 Δ m ( λ ) d λ (37)

Δ n = ∫ 400 1100 ( Δ n 1 , 0 + Δ n 2 , 1 + ⋯ + Δ n S i , N ) d λ = Δ n 1 , 0 + Δ n 2 , 1 + ⋯ + Δ n S i , N

It is defined as the difference between the optimal thickness ( e a r c ) of an antireflection layer for silicon and that given by the amplitude condition.

Δ e ( λ ) = e a r c − λ 4 ⋅ n a r c ( λ ) (38)

Following relations (14) and (38), thickness difference (TD), is related to the wavelength l by the following expression:

Δ e ( λ ) = − 0.144 ⋅ λ + 11.12 + e a r c (39)

Let e_{1} and e_{2} be the thicknesses of the respective lower and upper layers forming the AR double layer stack. Noting e_{arc}_{1} and e_{arc}_{2} those optimal for a double-layer antireflection coating, the thickness difference (TD) of a double antireflection layer is defined by the following relations:

Δ e 1 = e a r c 1 − λ r e f 4 n a r c 1 (40)

Δ e 2 = e a r c 2 − λ r e f 4 n a r c 2 (41)

n_{car}_{1} and n_{car}_{2} respectively represent the optimal refractive index of the inner and the upper antireflection layers. According to relations (19) and (20), these expressions can be written as following relations:

Δ e 1 = e a r c 1 − 0.173 ⋅ λ − 8.8 (42)

Δ e 2 = e a r c 2 − 0.117 ⋅ λ − 9.74 (43)

Thickness difference (TD) of a double layers is presented as follows: Δ e = Δ e 1 / Δ e 2 , l = 600nm is chosen as wavelength of reference in this paper.

_{av}) Depending on Refractive Index Difference (RID)

1) Case of Simple Layer Antireflection Coating.

_{x} corresponding to a low refractive index difference of 61; and a refractive index difference of 445 corresponds to a large reflectivity of 15%. These results are explained by the fact that this refractive index difference is even bottom; the refractive index of the AR layer is close to that of the optimal antireflection layer (given by the phase condition). Using _{av} on the silicon coated with a single antireflection layer can be modeled as a function of the total refractive index difference (Dn) by the following relation:

R a v = − 5.133 × 10 − 5 ⋅ Δ n 2 + 0.049 ⋅ Δ n + 3.36 (43)

Antireflection layer | Refractive index at l = 600 nm | RID (Dn_{t}) | Average reflectance R (%) |
---|---|---|---|

SiN_{x} | 1.86 | 67 | 9.1 |

Al_{2}O_{3} | 1.77 | 127 | 9.2 |

ZrO_{2} | 2.16 | 142 | 9.4 |

SiN_{x} | 2.2 | 161 | 10.8 |

SiN_{x} | 2.4 | 279 | 13.1 |

SiO_{2} | 1.46 | 344 | 14.3 |

TiO_{2} | 2.60 | 445 | 15 |

2) Case of Double Layer Antireflection Coatings.

Reflectance losses values and refractive index difference (RID) corresponding for each of double stack ARCs are presented in

_{2}/TiO_{2} AR double layer is lower, corresponding to an index difference of 84. This reflection is greater for the double SiO_{2} /Al_{2}O_{3} antireflection layer (10.8%) corresponding to a refractive index difference of 278. However, the refraction index difference for a single antireflection layer is not comparable to that coated with a double antireflection layer. In other words, the reflection on a cell coated with a double AR layer may be greater than that coated with a single AR layer while the latter has a difference in refractive index greater than the first. The average reflectance (R_{av}) on the silicon surface coated with a double antireflection layer is a function of the refractive index difference (Δn) and can be modeled as follows:

R a v = 1.82 × 10 − 4 ⋅ Δ n 2 − 0.022 ⋅ Δ n + 2.86 (44)

3) Case of multilayer antireflection coatings

Consider these different configurations of deposition of the AR layers on silicon with the same materials:

➢ A. TiO_{2}/Al_{2}O_{3}/ZrO_{2}/SiO_{2} ARC on silicon (degrowth of refractive indices).

➢ B. SiO_{2}/Al2O_{3}/ZrO_{2}/TiO_{2} ARC on silicon (growth of the refractive indices).

➢ C. ZrO_{2}/Al_{2}O_{3}/TiO_{2}/SiO_{2} ARC on silicon (refractive indices neither growth nor degrowth).

➢ D. Al_{2}O_{3}/ZrO_{2}/SiO_{2}/TiO_{2} ARC on silicon (refractive indices neither growth nor degrowth).

Couches AR | Dn (600 nm) | Dn (800 nm) | Dn_{1} | Dn_{2} | Dn | R_{av} (%) |
---|---|---|---|---|---|---|

SiO_{2}/TiO_{2} | 0.12 | 0.12 | 76 | 92 | 84 | 2.3 |

SiO_{2}/SiN_{x} (n = 2.4) | 0.09 | 0.10 | 76 | 73 | 75 | 2.4 |

SiO_{2}/SiN_{x} (n = 2.2) | 0.19 | 0.17 | 76 | 192 | 134 | 3.3 |

SiO_{2}/ZrO_{2} | 0.22 | 0.17 | 76 | 211 | 144 | 3.6 |

SiO_{2}/SiN_{x} (n = 2) | 0.28 | 0.23 | 76 | 302 | 189 | 5.2 |

SiO_{2}/SiN_{x} (n = 1.8) | 0.37 | 0.31 | 76 | 420 | 248 | 8.4 |

SiO_{2}/Al_{2}O_{3} | 0.42 | 0.35 | 76 | 480 | 278 | 10.8 |

multi stack | Dn_{1,0} | Dn_{2,1} | Dn_{3,2} | Dn_{4,3} | Dn_{Si,4} | Dn | R_{av} (%) |
---|---|---|---|---|---|---|---|

A | 319 | 216 | 269 | 304 | 884 | 398 | 3.97 |

D | 535 | 269 | 485 | 789 | 884 | 592 | 20.11 |

B | 1109 | 304 | 269 | 216 | 1673 | 714 | 33.5 |

C | 805 | 269 | 573 | 789 | 1673 | 822 | 42.7 |

With Δ n j , k Refractive index difference (RID) between layer number j and layer number k: then Δ n 1 , 0 represents RID between the n_{0} (air refractive index) and n_{1} (refractive index of upper layer).

_{Si}) leads to a lower reflectance (less than 4%) over the entire spectrum wavelength range of solar. These results were expected. Indeed, the high reflectivity on the surface of silicon solar cells is due to the large discontinuity between the refractive index that exists at the interface (air-cell). The deposition of a stack where the refractive indices grow, makes a sudden change in this index is replaced by a continuous transition from a low refractive index material to a high refractive index material. Thus, for a minimization of the reflectivity with a multilayer stack, the refractive index difference must be minimal. Consequently, the refractive indices of the different layers composing this AR stack must grow from the refractive index of air to that of silicon.

As shown in

than optimal, in other words, as one moves away from the zero thickness difference (Δe = 0), the reflectivity grow rapidly. ZrO_{2} ARC is an example where a zero refractive index difference (Δe = 0 then e = 72 nm) corresponds to an average reflectance of 9.4% while thickness difference of Δe = 10 (e = 82 nm) and Δe = −10 (e = 62) correspond to mean reflectivity of 10% and 10.4%, respectively. That is an increase in reflection of more than 6% in each case.

At the reference wavelength λ_{ref} = 600 nm, the thickness value of each layer given by the phase condition, (relation 8) is compared with that optimal (making the minimum reflectance) for different types of antireflection layer. The thickness difference (TD) and average reflectance values then obtained with each of antireflection layer) are summarized in

In

Figures 8(a)-(f) show variation of reflectance losses with thickness difference (TD) for each of stack double layer ARC on silicon.

As can be seen, the reflectivity on a double layer AR strongly depending on the respective thicknesses of the different stacks constituting it. For each type of AR layer, there is a couple of optimal thickness minimizing the average reflectance of the light rays arriving on the surface. As can be seen, this pair of thickness differs little from that given by the relationships (25) which would lead to a

zero thickness difference ( Δ e 1 = Δ e 2 = 0 ) of the curves 8, only the one representing the AR double-layer stack SiO_{2}/SiN_{x} (n = 1.8) leads to a minimal reflectance outside the zone where the thickness differences are zero. Indeed, in

Antireflection coating on Si | e = λ r e f 4 ⋅ n | Optimal thickness (nm) | Thickness Difference (TD) | Average Reflectance (%) |
---|---|---|---|---|

SiO_{2} | 103 | 106 | 3 | 14.3 |

ZrO_{2} | 69 | 72 | 3 | 9.4 |

TiO_{2} | 58 | 58 | 0 | 15 |

Al_{2}O_{3} | 85 | 88 | 3 | 9.2 |

SiN_{x} (n = 1.8) | 81 | 85 | 4 | 9.1 |

SiN_{x} (n = 2.2) | 68 | 72 | 4 | 10.8 |

SiO_{2}/TiO_{2} | 103/58 | 103/58 | 0/0 | 2.3 |

SiO_{2}/ZrO_{2} | 103/69 | 103/69 | 0/0 | 3.6 |

SiO_{2}/Al_{2}O_{3} | 103/85 | 95/78 | −8/−7 | 9.0 |

SiO_{2}/SiN_{x} (n = 1.8) | 103/81 | 110/89 | 7/8 | 2.0 |

SiO_{2}/SiN_{x} (n = 2) | 103/74 | 105/77 | 2/3 | 2.0 |

SiO_{2}/SiN_{x} (n = 2.2) | 103/68 | 100/66 | −3/−2 | 2.2 |

this case the reflectivity is minimal (2% on average) if Δe (SiN_{x}) is between −15 and 0; and Δe (SiO_{2}) is between −20 and −10.

We have modeled the optimal refractive index for single layer and multilayer antireflection coatings for minimization of reflectivity losses at the silicon surface. Silicon nitride (SiN_{x}, n = 1.8) single layer ARC reduces consequently reflectance but It was found that the antireflection effect of SiO_{2}/ARC2 double-layer ARC is better than that of single layer. For SiO_{2}/SiN_{x} double-layer ARC, the optimal antireflection effect is obtained with refractive indices of 1.46 and 2 for the top and the bottom layer, respectively. Refraction Index Difference (RID) and thickness Difference (TD) have allowed us to better understand the mechanisms of photon losses at the surface of silicon solar cells coated antireflection layer: Reflectivity is even lower than these refractive index difference and thickness difference are low. The results reported in this study can be used as a significant tool for efficiency improvement in thin film silicon solar cells. However, for an ideal antireflection layer, the absorption loss especially in the wavelength range [400 - 1100 nm], must be low.

Diop, M.M., Diaw, A., Mbengue, N., Ba, O., Diagne, M., Niasse, O.A., Ba, B. and Sarr, J. (2018) Optimization and Modeling of Antireflective Layers for Silicon Solar Cells: In Search of Optimal Materials. Materials Sciences and Applications, 9, 705-722. https://doi.org/10.4236/msa.2018.98051