Академический Документы
Профессиональный Документы
Культура Документы
given by and respectively in the P and N sections. In This second-order differential equation is not twice integrable,
what follows, we suppose that all dopant and carrier however, by defining = √ ⁄√ we may obtain the first
concentrations are normalized to the intrinsic carrier density order approximate equation (for a detailed justification please
. Hence, for a symmetric junction we may define the c.f. Appendix A)
dimensionless density = ⁄ = ⁄ . Similarly, ′( )
electron and hole carrier densities are normalized accordingly = [ ( ) − ].
( )
as ( ) = ( )⁄ and ( ) = ( )⁄ . This allows us to (8)
write ( ) ( ) = 1 under equilibrium. Further substitution of (6) in (8) gives the first-order
differential equation
′( )
= [ ( ) − ].
P N ( )+4
(9)
This equation can now be integrated exactly. In order to do so,
=0 it should be first rearranged as ⁄( − ) + 4 on the
Fig. 1. Illustration of PN junction. left-hand-side and on the other. This will yield
4
1− ( )− ( ) nx in units of ni
( )= , < 0,
1+ ( )− ( )
(14) 105
in which
⁄ℓ
( )= . 100
+2
(15)
Now, electron and hole densities for < 0 may be easily 0.1
found from (3) and (6) as
1 104
( )= ( )+4+ ( ) ,
2
1 xL
( )= ( )+4− ( ) . 30 20 10 0 10 20 30
2 Fig. 4. Concentration of electron carriers for different doping levels:
(16)
= 10 or = = 10 , = 40.1 (black: analytical,
Because of symmetry, we may use (+ ) = (− ) to obtain
dotted: numerical), = 10 or = = 10 , = 12.7
solutions for > 0 as well. The results based on (16) and (14) (dashed: analytical, dot-dashed: numerical).
will guarantee smooth transition of hole and electrons across
the boundary at = 0. The continuity of derivatives are also Qx in units of qni
1 107
supported by symmetry, and easily follow (16), by noting that
(0) = 0. To observe this, it is sufficient to take derivatives of
(16) and simplify expressions for ′(0) and ′(0). This will
readily give ′(0) = − ′(0) which is the desired result. This 5 106
2.5
5 106
2.0
1 107
1.5 Fig. 5. Total space charge across the junction for different doping
levels: = 10 or = = 10 , = 40.1 (black:
analytical, dotted: numerical), = 10 or = = 10 ,
1.0 = 12.7 (dashed: analytical, dot-dashed: numerical).
xL
30 20 10 0 10 20 30 ( )= , < 0.
Fig. 3. Concentration of hole carriers for different doping levels: 1− csch
2 ℓ
= 10 or = = 10 , = 40.1 (black: analytical, (17)
dotted: numerical), = 10 or = = 10 , = 12.7 Now, one may integrate (17) as
(dashed: analytical, dot-dashed: numerical).
4
( )
2 −2 ( )= ln = ln .
ℰ( ) = − ℓ tanh ( )
(26)
2 + tanh Inserting (17) into (16), and using the above, we obtain an
+ tanh 2ℓ . explicit equation for the distribution of electrostatic field. For
this purpose, we first extend (17) into both sides of the
(19) junction as
Using the trigonometric identity tanh + tanh =
tanh [( + )⁄1 + ], the above expression may be ( ) = −sgn( ) ,
further simplified as 1 + 2 csch ℓ
(27)
2 1 + tanh where sgn(∙) is the sign function. A plot of the electric field is
ℰ( ) = − ℓ tanh 2ℓ . shown in Fig. 6b. The tails of the potential variation at the
2+ + ( − 2)tanh
2ℓ borders of the depletion area are somewhat similar to
(20) exponential functions [20].
Evidently, the electric field ℰ( ) must be an even function so
that we have the following full expression, being valid Ex kVcm
xL
everywhere both for < 0 and > 0 30 20 10 10 20 30
20
2 1 − tanh
ℰ( ) = − ℓ tanh 2ℓ . 40
2+ − ( − 2)tanh
2ℓ 60
(21)
A typical sketch of the electric field is shown in Fig. 6a. 80
Hence, the maximum electric field ℰ = |ℰ(0)| at the
100
middle of the junction is given by
120
2
ℰ = ℓ tanh . 140
2+
(22) (a)
The expression within the brackets is logarithmically Voltage V
divergent and using tanh (1 − )~ ln(2⁄ ) for small (and 0.8
positive) , and taking 1 − = ⁄( + 2) we obtain
= 1− 1−( )
≈ 2 ⁄( + 2) ≈ 2⁄ , where we have
0.6
used the binomial expansion first, as well as the fact ≫1
next. After some algebra we get
0.4
ℰ ≅ ln .
ℓ
(23)
The built-in voltage of the junction is exactly given by the 0.2
initial difference in Fermi levels prior to the contact, being
=2 ln . xL
20 10 10 20
(24) (b)
Hence, we may define an effective width of the depletion
region as Fig. 6. Electrostatic field and potential across the junction for
difference doping levels: = 10 or = = 10 , =
2 40.1 (black: analytical, dotted: numerical), = 10 or =
≡ = 4ℓ. = 10 , = 12.7 (dashed: analytical, dot-dashed:
ℰ
(25) numerical).
5
From (23) and (25) we may readily obtain the accurate scaling
laws 1.00
ℰ ∝ √ ln , 0.50
(28)
for peak electric field and
0.10
1 0.05
∝ ln ,
(25) are indeed in exact agreement with the known Fig. 8. Dependence of effective depletion width in a symmetric
expressions from depletion layer analysis. Silicon junction versus dopant concentration: accurate (solid) and
simple depletion approximation (dashed).
This result shows that the effective width of the depletion
layer actually is inversely proportional to the doping Error in Emax
concentration. As a matter of fact, calculations of Fig. 3 also
confirm and are in complete agreement with this finding, as by 5.5
increasing the doping density from = 10 to = 10 , the
5.0
peak electric field is also increased from 41.6kV/cm to
4.5
142kV/cm by a factor of 10 which is roughly equal to 3.4.
4.0
Figures 7 and 8 compare the exact maximum electric field
3.5
from Appendix C and effective width of the depletion layer
due to our analysis (solid lines) versus depletion layer 3.0
approximation (dashed lines). Figures 9 and 10 present the N in units of ni
relative errors of analytical solution versus the depletion layer 105 106 107 108 109
approximation given in percent. It is notable that by Fig. 9. Relative error in calculation of maximum electric field in
examining the depletion layer approximation versus our depletion layer approximation.
analysis here, it is found that the estimations for the width of
Error in W
depletion layer as well as the maximum electric field coincide
5.5
fairly well, just within a few percent.
5.0
C. Highly Asymmetric Junctions
4.5
Equation (10) still holds but has to be integrated differently, III. NON-ZERO BIAS
since (0) is no longer zero. In fact, we may make use of the In this section, we study the case of non-equilibrium under
fact that the whole depletion layer is effectively shifted into external bias. From (D.5), and after numerical tests for large
the P domain because of the zero curvature of the electric field , we may write down the further approximation
in the N domain, so that one may have (0) = < 0. This
will result in ( ) ≈ −2sinh[ln tanh( ⁄ℒ )],
(38)
− ( ) −
= ⁄ℓ
, < 0. where ℒ = ℓln√ . Hence, we readily obtain by (3) the
( )+4+ ( ) +4+ approximate expressions for equilibrium carrier densities as
(31)
By going through a few algebraic steps similar to what was ( ) ≈ exp[−ln tanh( ⁄ℒ )],
done with (11) to obtain (12), the equation (31) would yield ( ) ≈ exp[+ln tanh( ⁄ℒ )].
(39)
4 These expressions consistently satisfy the boundary and initial
1− ( )− ( )
( )= , < 0, conditions at origin and infinities. The electric field under
1+ ( )− ( ) equilibrium is given by (1) as
( )
2( − ) ℰ( ) = = − ln sech ( ⁄ℒ ),
( )= ⁄ℓ
. ( ) ℒ
+4+ (40)
(32) resulting in a relation similar to (23) for the maximum electric
If we define ( ) = 0, then can be found from by field. We assume that under positive external bias, (39) are
setting the numerator of ( ) equal to zero and ignoring the perturbed as
middle term. This will result in −Δ
( ; ) ≈ exp −ln tanh ,
ℒ
2( − ) +Δ
⁄ℓ
≈ ±1, ( ; ) ≈ exp +ln tanh ,
+4+ ℒ
(41)
(33)
with Δ being a voltage-dependent constant to be determined.
out of which the positive sign only may lead to physically
This will result in
meaningful solutions. Hence, we obtain
−Δ
2( − ) ( ; ) ( ; ) = exp −ln tanh
= −ℓ ln , ℒ
+4+ +Δ
− tanh .
(34) ℒ
where by the strong asymmetry we have = (0) − (0), (42)
and (0) ≈ ⁄ and (0) ≈ ⁄ . Hence, ≈− ⁄ . Within the depletion region, at = 0 we should have
Now, by noting ≈ = ⁄ and ≫ we get
( ; ) ( ; ) = exp .
( + ) (43)
≈ −ℓ ln ≈ ℓ ln . This will curiously yield the simple result
+
(35)
Then (32) may be put into the similar accurate form of (17) as Δ = ℒtanh = ℒtanh .
2ln
(44)
( )= , < 0. We now assume symmetric hole and electron mobilities, and
−
1− csch write the equations for the current density = 2 = 2 as
2 ℓ
(36)
Hence, the expression for the electric field should be = [ℰ( ; ) ( ; ) − ( ; )],
modified as
= [ℰ( ; ) ( ; ) + ( ; )].
ℰ( ) (45)
− Combining the above gives
2 1 + tanh
=− ℓ tanh 2ℓ
− ( ; )+ ( ; )
2+ + ( − 2)tanh
2ℓ ℰ( ; ) = ,
( ; )− ( ; )
1 + tanh 2ℓ [ ( ; ) ( ; )]′
−tanh . =2 .
2+ + ( − 2)tanh ( ; )− ( ; )
2ℓ
(37) (46)
7
Using (41) and plugging in (46) we arrive at the non- and diverges at = , approaches zero under strong reverse
equilibrium electric field expression given by bias, similar to the well-known behavior of depletion
capacitance [1-4].
ℰ( ; )
−Δ +Δ B. I-V Characteristics
sech ( ; ) − sech ( ; )
=− ln ℒ ℒ .
ℒ ( ; )− ( ; ) Returning to the equation (45) for the current density, if we try
(47) to directly plug-in the quasi-equilibrium carrier densities (41),
one would obtain zero current. Hence, one would need first to
Hence, by employing hyperbolic-trigonometric identities, and estimate the change in electric field due to bias from another
within the quasi-equilibrium approximation, the electric field method. For this purpose, we start from the equation
at = 0 under forward bias is
Δ ℰ( ; ) = ℰ( ; 0) + [ ( ; ) − ( ; 0)] ,
ℰ(0; ) = − ln sech = − ln 1− .
ℒ ℒ ℒ
(48) (55)
This expression may be easily generalized for both negative After plugging in (49), this may be re-written as
and positive voltages and non-zero current density as
2
| | = [ ( ; ) − ( ; 0)] .
ℰ(0; ) = − ln 1− + exp − .
ℒ 2 2
(49) (56)
Evaluation of integral and simplifying results in the expression
A. Junction Capacitance
2 ℒ
( )= ( ) ,
The small-signal capacitance per-unit area of the junction may
be found as (57)
where the function ( ) is given as
= ,
( ) = 2Shi cosh − Shi 1+
(50) 2 2 2
where is the per-unit area charge difference on either side of
the junction due to the application of the voltage. It may be − Shi 1− .
2
evaluated as (59)
Here, Shi( )is defined in terms of the exponential integral as
(60)
= ( ; ) ,
The dimensionless function ( ) = ( ) is plotted in
(52) Fig. 11.
which after using (42), (44), and some simplification results in I V
ℒ
= sinh . 2.5 1014
| | 2
1−
2.0 1014
(53)
Please notice that (53) represents the total capacitive effects of 1.5 1014
diffusion kinetics and depletion layer combined at once. It is a
finite value at zero bias 1.0 1014
5.0 1013
ℒ
| = sinh ,
2 Vin units of VT
40 20 20
(54)
Fig. 11. Normalized approximate I-V characteristics of junction.
8
= + 4[4 + − ( − )] + +4 − 4
−2sinh( ⁄ℓ) −
+ = + 4[4 + ] + 4 + ( )= ,
−2sinh( ⁄ℓ) +
= +4+4 + (D.4)
= +4+4+ . and subsequently as
4
(B.4) 1−
Hence, we have = , and thus (9) is the true derivative of ( )= .
(10) as asserted above. 1 + 2 csch( ⁄ℓ)
(D.5)
APPENDIX C All remains is to ignore 4⁄ for large in the numerator to
The equations (1) and (2) for < 0 may be combined to obtain (17).
obtain
℘′′( ) = [2sinh℘( ) − ],
(C.1) REFERENCES
where ℘( ) = ln ( ) or ℰ( ) = ℘′( ). Solution of this [1] B. G. Streetman, S. K. Banerjee, Solid State Electronic Device, 6th ed.
Englewood Cliffs, NJ: Prentice-Hall, 2009.
nonlinear equation usually leads to numerically unstable
[2] M. Razeghi, Fundamentals of Solid State Engineering, New York:
results. But it can be analytically integrated once on (−∞, ) Kluwer, 2002.
by multiplying both sides by ℘′( ), and noting that ℘′( ) < [3] W. Shockley, Electrons and Holes in Semiconductors, Princeton,
NJ:Van Nostrand, 1950.
0, ℘′(−∞) = 0, and ℘(−∞) = ln . This will give the firs- [4] J. Lindmayer and C. Y. Wrigley, Fundamentals of Semiconductor
order logarithmic differential equation Devices, Princeton, NJ: Van Nostran, 1965.
[5] E. M. Buturla. P. E Cottrell, B. M. Grossman, and K.A. Salsburg,
“Finite-element analysis of semiconductor devices: The FIELDAY
℘′( ) = √2 2cosh℘( ) − 2cosh(ln ) − [℘( ) − ln ], program,” IBM J. Res. Develop., vol. 25, no.4, pp. 218-231, 1981.
(C.2) [6] S. E. Laux and K. Hess, “Revisiting the Analytic Theory of p-n
which can now be numerically integrated stably subject to the Junction Impedance: Improvements Guided by Computer Simulation
Leading to a New Equivalent Circuit,” IEEE Trans. Electron Devices,
boundary condition ℘(0) = 0 for < 0. Hence, the vol. 46, no 2, pp. 396-412, Feb 1999.
maximum electric field is exactly given by [7] H. R. Pota, “A New Derivation of the Law of the junction” IEEE
Trans. Education, vol. 47, no.4, pp. 497-499, Nov. 2004.
[8] A. Zareian Jahromi and S. Khorasani, “Analysis of p-n Junctions
ℰ = √2 2 − 2cosh(ln ) + ln . Without Depletion Layer Approximation,” 14th Iranian Conference on
(C.3) Electrical Engineering, Tehran (2006).
[9] B. Gharekhanlou and S. Khorasani, “Current-Voltage Characteristics of
Graphane p-n Junctions,” IEEE Trans. Electron Dev., vol. 57, no. 1, pp.
APPENDIX D 209-214 (2010).
[10] B. Gharekhanlou, S. Khorasani, and R. Sarvari, “Two-dimensional
Referring to the equations (14) and (15) we have for <0 Bipolar Junction Transistors,” Mat. Res. Exp., vol. 1, no. 1, 015604
(2014).
4 [11] B. Gharekhanlou, S. Khorasani, “Generation and recombination in two-
⁄ℓ ⁄ℓ
1− +2 − +2
dimensional bipolar transistors,” Appl. Phys. A, vol. 115, no. 3, pp. 737-
( )= . 740 (2014).
⁄ℓ ⁄ℓ [12] Yu. G. Peisakhovich, A. A. Shtygashev , L. A. Borynyak, N. Yu.
1+ −
+2 +2 Petrov, “Electric field and charge density in the plane of a quasi-
(D.1) equilibrium asymmetric 2D p–n junction with zero current,” Technical
⁄ℓ Physics, vol. 60, no. 10, pp. 1494-1500 (2015).
Dividing the numerator and denominator by we find [13] R. Corkisha and M. A. Green, “Junction recombination current in
abrupt junction diodes under forward bias,” J. Appl. Phys., vol. 80, no.
4 5, pp. 3083-3090 (1996).
⁄ℓ ⁄ℓ
− − [14] J. H. Ning, J. S. Yuan and J. Song, “Effects of base and emitter doping
( )= +2 +2 . gradients on the electrical performance of heterojunction bipolar
⁄ℓ ⁄ℓ transistors,” Solid State Electron., vol. 41, no. 9, pp. 1263-1268 (1997).
+ −
+2 +2 [15] A. A. Barybin, and E. J. P. Santos, “Unified Approach to the Large-
(D.2) Signal and High-Frequency Theory of p−n-Junctions,” Semicond. Sci.
Technol., vol. 22, no. 11, pp. 1225 (2007).
Now, we make use of the approximations +2 ≈ for [16] H. Zandi, M. Bavafa, M. Chamanzar, and S. Khorasani, “Harmonic
large , and find Content and Relaxation Resonant Frequency of a Modulated Laser
Diode,” Scientia Iranica, vol. 16, no. 2, pp. 145-156 (2009).
[17] S. Khorasani and B. Cabon, “Theory of Optimal Mixing in Directly
⁄ℓ 4 ⁄ℓ
− − Modulated Laser Diodes,” Scientia Iranica, vol. 16, no. 2, pp. 157-162
( )= ⁄ℓ ⁄ℓ
, (2009).
+ − [18] X. Yang and D. K. Schroder, “Some Semiconductor Device Physics
(D.3) Considerations and Clarifications,” IEEE Trans. Electron Dev., vol. 59,
which by can be simplified using hyperbolic-trigonometric no. 7, pp. 1993-1996 (2012).
[19] A. Boukredimi, K. Benchouk, “New improved capacitance–voltage
identities as model for symmetrical step junction: a way to a unified model for
realistic junctions,” J. Comput. Electron., vol. 13, pp. 971–982 (2014).
[20] D. J. Roulston, An Introduction to the Physics of Semiconductor
Devices, vol. 1, Oxford University Press, 1999.