0% found this document useful (0 votes)
378 views24 pagesSemiconductor Device Fundamentals - Pierret - Chapter 2 Solutions
Semiconductor Device Fundamentals - Pierret - Chapter 2 Solutions
Uploaded by
Ronzy Y. GomezCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
378 views24 pagesSemiconductor Device Fundamentals - Pierret - Chapter 2 Solutions
Semiconductor Device Fundamentals - Pierret - Chapter 2 Solutions
Uploaded by
Ronzy Y. GomezCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 24
2
(@/(b) The MATLAB program script yielding both the part (a) and part (b) results is listed
below. A combined plot comparing the part (a) result (solid-line) and part (b) result
(dashed-line) is included before the program script. At T = 300 K the part (a) relationship
yields |Eg(300K) = 1.1245 eV].
12;
300)
1H
MATLAB program script...
2EG Computation (EG versus 7)
close
clear
Parabolic Fit Parameters
BG0=1.170;
a=4.730e-4;
b=636;
Parabolic computation and plot
T=[0:5:600];
2-1
500
‘600 EGEGO-a.*(T.*2)./ (T+) 7
EG300=EG0-a.* (300.*2) ./ (300+b)
plot (T,EG); axis({0 600 1.0 1.2}); grid:
xlabel ("T(K)'); ylabel ("EG (eV) ") +
hold on
Linear computation and plot
EG0=1,205;
a=2.8e-4;
EG=EGO-a.*T;
plot (T, EG, 'b--");
hold off
8T=300K result
£G300
22
(a) See Fig. 2.4(a).
(b) See Fig, 2.4(b) or the left-hand side of Fig. 2.7(b).
(©) Sce the left-hand side of Fig. 2.7(c).
(d) See Fig. 2.10(a).
(e) See Fig. 2.10(b).
2.3
(a) See the right-hand side of Fig. 2.7(b).
(b) See the right-hand side of Fig. 2.7(c).
(c) See Fig. 2.13(a).
(d) See Fig. 2.13(b).
(€) See the extreme left-hand side of either Fig. 2.13(a) or Fig. 2.22(b).
(£) See the extreme left-hand side of Fig. 2.13(b).
(g) See the extreme right-hand side of Fig. 2.16.
(h) See the extreme left-hand side of Fig. 2.18.
(i) See the middle of Fig. 2.18.
() See the extreme right-hand side of Fig. 2.18.
(k) See Fig. 2.19.
(I) See Fig. 2.19.
2-2 24
(a) The removal of the column Hi Ga atom with three valence electrons leaves five dangling
bonds in the vicinity of the vacancy. The removal of the column V As atom with five
valence electrons leaves three dangling bonds in the vicinity of the vacancy.
(b) When a Si atom with four valence electrons is inserted into the missing Ga site, there is
one extra electron that does not fit snugly into the bonding Conversely, when a Si
atom is inserted into the missing As site, there are one too few bonds to complete the
bonding scheme.--There is a hole in the bonding scheme.
Answer-(b)
©) n-type] ... The extra electron noted in part (b) is readily released yielding an increase ir
the electron concentration.
(@) [p-type] ... The missing bond noted in part (b) is readily filled at room temperature
yielding an increase in the hole concentration.
©
@ Gi) Ee
nerocensenenenanereconeccenes Ej
ER
Ey Ey
n-type p-type
2-3 25
As noted in Subsection 2.4.1, g¢(E)dE represents the number of conduction band
states/cm? lying in the energy range between E and E + dE. It follows that the number of
states/om} in the conduction band lying between energies E, and Ec + AT is simply
obtained by integrating g¢(E)dE over the noted range of energies.
Boot ee peter
states/om3 = [ gc(E\dE = Ma 2a ssf YE~Ey dE
ee ww Je,
3 28 3) 9293
* * ct KT * +
2 amivame (gps [2iZ2zt| ere
Eo,
2.6
(a) The probability of electrons occupying states at a given energy under equilibrium
conditions is given by the Fermi function. Here we are told the energy of interest is
E=Ep. Thus
=——1 Ll
Kee) 1l+eGr-Epat 2
(b) The desired probability is again given by the Fermi function. Here we are told Ep = Ey
and the energy of interestis E = E, + kT. Consequently,
Ee +kt) = ——-L____=-_1_ = 9.2
RBe + kT) 1 + ellGetkN)-EMMT 1401 0.269
(c) The problem statement indicates f(Ec+kT) = 1 —f(Ec+kT), or
1 1 Bot kT-ER KT
=1- = = 1
14 eGcthT-EPAT yg EetkT-EPAT 1 4 g(EctkI-ERAT | y @ Ep-Ec-KAT
Thus we must have
Ec +kT-Ep = Ep-(Ee+kT)
Eps Ec+kT|
or
2-4 27
‘The distribution of electrons in the conduction band is given by gc(E/(E); the distribution
of holes in the valence band is given by g()[1 -/(E)]. Working with the electron
distribution we note,
= — LL ___ = e(E-EnaT
AE) = yecenet = © for all E 2 Ey if the semi-
conductor is nondegenerate
Thus
* so E-ErykT
acl YE) = ma 2mx(E-E.) EEN)
me
aK (EE)? eELMT em mim
: ee
‘The extrema points of any function are obtained by taking the derivative of the function anc
setting the derivative equal to zero.
E)] = —* EEN . ¥(p-B,)'2 e(EEDUT
fisteve)) EE Eee)
set
= 0
Clearly
Va Ee
NE peak-Ee AT
or
Epeax - Ee = KTP
and
E peak = Ee + kT/2| .for electrons in the conduction band
‘The development leading to the peak energy of Epeak = Ey — kT/2 for holes in the valence
band is completely analogous.
2-5 28
The electron population at any energy is given by gc(E)(E). Also, since the semiconductor
is nondegenerate
Ey = ——1__ = eB EDIT for all E> E
m® Tre@-Epar = © * .
The electron population at E = E, + 5kT normalized to the peak electron population at
E=E + kT/2 is therefore
BclEct SkTYfiEc+SkT)
Fatlo =F (EqthTDYiEgtkT1D)
= YSKE, g-(Get SkI-ERAT _ 45 2 2
2 eS amapat 7 Ve 3.51 x 10
29
The hole and electron distributions are given respectively by
on*
(electron dist.) = gc(EY(E) = ma Bg YE —E, ee-ER IT
wh
— [ataV me | -eg/sir {E-EykT
= e YE-E; e
wae
and
(hole dist.) = gv(E)[1-fA)] = ma 2p YE, -E e(€-EpkT
wwe
= (2 V 2mp eRBGAET| YES TE e~(Ev- EMIT
en
Note that the approximate (non-degenerate) expressions for the Fermi function established
in Subsection 2.4.2 were employed in writing down the carrier distributions. ‘The required MATLAB program script and resultant plots are presented below.
Computations were performed employing Eg = 1.12 eV, kT = 0.0259 eV, mn*/mo = 1.18
and mp*/mg = 0.81 Table 2.1, and h = 6,63 X 10°34 joule-sec and mg = 9.11 X
103! Ke from the table of physical constants inside back cover).
The plots are clearly consistent with Fig. 2.16 in the text. (Note that the electron
distribution scale is multiplied by 1015 while the hole distribution scale is multiplied by
106.) The distributions at kT/2 from the band edges graphically reconfirming the peak
positions noted in Problem 2.7.
MATLAB program script...
Problem 2.9,..Carrier Distributions
SIntialization
close
clear
Constants
m0=9,11e~31;
mnx=1,18;
mpr=0.81;
hbar=6, 63e-34/ (2*pi):
cl=1.6e-19; joules = cl*ev
221. 0e-6 m*3=¢2*cm*3
Computation
deltaz=linspace (0, S*kT) : deltaE = E-Ec or Ev-E in ev
AemO*sqrt (2*m0) / (pi*2*hbar*3);
An=mnr* (3/2)*A; Ap=mpr* (3/2) *A;
@_dist=cl*c2*An*exp (-EG/ (4*kT)) .*sqrt (cl*deltaE} . *exp (-deltaE/kT);
h_di st=cl*c2*Ap*exp (-3*EG/ (4*kT) ) .*sqrt (cl*deltak) . *exp (-deltaE/kT)
‘Note use of cl and c2 to make distribution units number/cm*3-eV
Plots
subplot (2,1,1), plot (e_dist,deltaB/kT); grid
xlabel ("electron distribution (number/cm*3-eV)");
ylabel (" (E-Ec) /kT')
subplot (2,1,2), plot (h_dist,-deltaE/kT); grid
axis({0,1.4e7,-5,0]) .
xlabel (‘hole distribution (number/cm*3-eVv)');
ylabel (* (E-Ev) /kT')
2-7 (EEQKT
4 8 19 12 14
electron distribution (numberlcn*3-eV)
4 6. 8 10 42 14
hole distribution (numberdemn*3-2V) to
(a) Utilizing Eq. (2.6a), the approximate (nondegenerate) expression for the Fermi function
established in Subsection 2.4.2, Eq. (2.13a), and Eq, (2.16a), one obtains
acl EYE) (mav Ima /n23) YE-Ee e-E EDT
selERE) 2
(normalized dist.) =
2mternnn) ee EOMT
= 2VEHEe 9 -(e-EKT
Ya(RT pe
(b) A plot of the normalized electron distribution verus energy for three different
temperatures and the MATLAB program script yielding the plot are given below. From the
plot one observes that the peak energy, which occurs at &T/2, moves to progressively
higher energies with increasing T. More significantly, the distribution becomes less peaked
in nature and the height of the peak decreases with increasing temperature,
v
co MATLAB program script...
eProblem 2.10...Normalized Electron Distribution as a function of T
BIntialization
close
clear
Computation and plot
k=8.617¢-5;
T=[300 600 1200];
kik. *T}
B_Ec=linspace (0, 0.4);
for i=1:3,
dist=2*sqrt (E_Ec) / (sqrt (pi) *kT (i)* (3/2)) .*exp(-E_Eo/kIE (i) 7
yi, :)=dist;
nd
e
plot (E£¢,y); grid
axis ([0,0.4,0,20))
label ("B-Ec (eV)'); ylabel({*normalized distribution (1/eV)')+
text (.005, 12, '300K", 'Color', yellow");
text (01, 7, 600K", *Color*, 'magenta');
text (.015, 3, 112004", ‘Color’, 'cyan')
normalized distrbution ({ieV)
9 ar
02 025 O38 025 O04
EEc(ev) 2A
Substituting the Eq.(2.6b) expression for gy(E) and the Ea: (2.7) expression for f(E) into
Eq.(2.8b), one obtains
j-—_L__. - -_1.__
1a eE-EOkT | 4 e(Gr- BAT
and
Ey
_mpv2ing YE, -E dE
TA Sesonam 1 + € OF EVAE
Now letting
Ey-E
"= "UT
Ey-E
ny =o
Epotiom > —°°
yields
mgvimg KEP? ("Pan
ew Jp 1+ em
Recognizing
a
Fini) = dT peemieDirac integral of order 1/2
jp Laem
and defining
mg kT P?
Ny = 2) 2B
Qn
one obtains
(2.9)
(2.10a)'
(2.10b)'
(2.10c)'
(2.41)
(2.12)'
(2.13b) p=Nv # Fin () @.14b)
If the semiconductor is nondegenerate, such that Ep 2 Ey + 3KT, then ny $-3. Since
120 in the Fermi-Dirac integral, exp(7}-My) 2 exp(3) for all 7. Thus one obtains
Fin) = [ nite Oran = & e(Eu- ExykT
0
Substituting this approximate relationship into Eq.(2.14b) finally yields
p = Ny e(Ev- ERVaT (2.16b)
242
@ electron distribution = go(EY(E) = (Nc/KI) e°*E ERAT
where use has been made of the fact that the semiconductor is nondegenerate (Ep
This yields
=no{ —@1_
one | oe
The integral here can be performed in closed-form.
n= Ne[n-Ini+e™ = Ne[ne + In(ite®)]
‘The relationship analogous to Eqs.(2.14a) is therefore
n = Ne [n+ In(l+e%)]
If the semiconductor is nondegenerate, jc $ -3. Thus
In(l+e™®) = Ine" (1+e% )] = -nNe + In(1+e7™)
= -Ne+e% ...since exp(mc) << 1 and In(1+x) sx ifx << 1.
‘We therefore conclude: .
n= Ncet = Neer Be kT
“The above holds for a nondegenerate semiconductor and is the desired relationship
analogous to Eq.(2.16a). Actually, the relationship has turned out to be identical to
Eq.(2.16a). 213
(a) Rewriting Eqs.(2.13), one obtains
2.
ve = mal
32
My = nol
where.
sf
No = ra
° | nn?
Using the numbers cited in the problem statement, the -value given on the inside back
cover, and remembering to convert from eV to joules, one calculates
No = ie (9.109 x 10°! )(8.617 x 10°5 (300)(1:602 x 10°19 i.
= 2] 220.109 x 107" )(8.617 x 10° (300)(1.602 x 10°19
(6.625 x 1034
= 2.510 x 1025/m3 = 2.510 x 101%em3
and therefore
Nov = (2.510 x 1019/em3) (m*/mg)3/2
(b) Semiconductor Nc (cm3) Ny (cnr)
Si 3.22 x 1019 1.83 x 1019
Ge 1.02 x 1019 5.42 x 1018
GaAs 4.26 x 1017 9.41 x 1018 214
(a) Referring to Fig. 2.20, one concludes:
(® ni(Si) = ni(Ge, 300K) at T = 430K.
Gi) n(GaAs) = (Ge, 300K) at T = 600K.
(b) With the differences in the effective masses neglected,
mia _ e PONT _ o(Ecp-EcaVvauT
Te = Sr eV0.0518 = 2.42 x 108
oa
2.45
The MATLAB program script implementing the requested nj(Ge) vs. T computation is
reproduced below along with sample numerical results. As must be the case since the same
computational equation was used in both cases, the numerical results are found to be
consistent with the values displayed in Fig. 2.20.
MATLAB program script...
sProblem 2.15...ni versus T for Ge
%Initialization
close; clear
Computation
k=8.617e-5;
P=[2252252475] 7
naz (1.76e16) .* (T.*1.5) exp (-0.392./(K.*T))7
apisplay result on screen
jelength (T);
Eprintg (*\n\nT (K) ni (Ge) \n") >
for 4i=1:4,
fprinté (*%-10.£%-10. 3e\n" , T (ii) -mi (4) 7
end
TK) ~—ni(Ge)
225 -9,841e+10
250 8.705e+11
275 -5,25le+12
300 2.375e+13
325 8.597e+13
350 2.61 1e+14
375 6. 888e+14
400 1.620e+15
425, 3.463e+15
450 — 6.838e+15
475, 1.263e+16
2-14 2.16 :
(a) AsT- 0, n-¥0 and p-+0, (Sce the discussion in Subsection 2.5.7.)
(b) Since N >> nj, one would have
n= Np and p = nj/Np_...if adonor
p=Nqa and nt = n2/Naq_ ...if an acceptor
We are told n = N and p = nj2/N. Clearly the impurity isa donor.
(c) Here we are given the minority carrier concentration, n = 105/em3, As long as the Siis
nondegenerate, one can always write
np = ne
Thus »
p= nitin =o? = 1015/em3
Note: From previous problems we recognize that the above carrier concentrations do
indeed correspond to a nondegenerate semiconductor.
(d) Given Ep-Ej = 0.259eV and 7 =300K,
n = n,eEF-EMET = (4910) 00.259/0.0259 — 2.29 x 10!4/em3
p= np eGi-EDAT ~ (1910) -0.259/0.0259 . 4.54 x 105/em3
{e) Employing the np product relationship,
np = 122 = nP
n=VInj = 1.414 X 10'3/em3
‘Next employing the charge neutrality relationship,
p-n+Np-Na = nl-n+Np = 0
Np = n/2 = nF = 0.707 X 103/em3 217
(a) Atroom temperature in Si, ni = 10!%cm3. Thus here Np >>Na, Np >> ni and
n= Np =10!5/cm?
p= n2/Np = 105/em3
(b) Since Na >>Np and Na >> rir
p= Na = 1016/em
n= n@iNaq = 104/cm3
(c) Here we must retain both Na and Np, but Np-Na >> ni.
n= Np-Na = 1015/em3
p = n2/QVp-Na) = 105/em3
(d) We deduce from Fig. 2.20 that, at 450K, nj(Si) = 5 x 10!3/cm3, Clearly, nj is
comparable to Np and we must use Eq.(2.29a).
12
Mp « [Mol +n] = 1.21% 10'4Yem3
2 13y2
path = @210%F - sorsroen
121 x 10!
(e) We conclude from Fig. 2.20 that, at 650K, nj ~ 10!6/cm3. Here nj >>Np. Thus
i = 10!6/orn’
Pp = 1016/cm3
2-16 2.18
(i As established in the text [Eq.(2.36)],
= Fed ee 5 3 er inteng*iatg")
Taking mp"/imp* to be temperature independent and employing the values listed in
Table 2.1, one concludes
Ej displacement
part TK) kT (eV) idgap (eV)
(a-c) 300 0.0259 —0.0073
(d) 450 0.0388 —0,0109
(e) 650 0.0560 -0.0158
Alternatively, the rip*/mg and tq*/mo versus T fit-relationships cited in Exercise 2.4 may
be used to compute the mp*/nin* ratio. One finds
Kj displacement
pat =I) = mp*ima* kT) i
(ac) 300 1.680 0.0259 —0.0075
(d) 450 0.703 0.0388 ~0.0103
©) 650 0.719 0.0560 —0.0139
Gi) Ep - Ej is computed using the appropriate version of Eq.(2.37) or (2.38).
(a) Eg- Ej = kT In(Np/nj) = 0.0259 In(1015/10!9) = 0.298 eV
(b) Ej- Ex = kT In(Na/nj) = 0.0259 In(1016 101) = 0.358 eV
(c) Eg - Ej = kT In[(WVp-Na)/ni} = 0.0259 1n(1015/10!0) = 0.298 eV
(@) Ep- Ej = kT In(wini) = 0.0388 In(1.21 x 1014/5 x 1013) = 0.034 eV
(©) Ep- Ej = kT In(n/nj) = 0 snl ni)
Eo — Ec
ER
- &;
Ey
(a) and (c)
tw
3 ——_————— Fe
1.015 Ep=Ey
0.49
Ey
)
(a) A sample MATLAB program that computes 1, DP, and Ep-Ej given T, Np, and Na is
listed below. The program incorporates the 1(T) computation given in the solution to
Exercise 2.4a.
MATLAB program script...
calculation of n,p and EF-Ei (nondegenerate, fully lonized)
aInitialization
clear: close
‘Specification of basic parameters
Teinput ("Please input the temperature, T, in Kelvin...T="];
NA=input ("Please input NA(cm-3)...NA=");
ND=input ("Please input ND(cm-3)...ND=")?
k=8.617e-5;
Nnet=ND-NA;
3ni computation (from Exercise 2.4a solution)
Constants and T-range
A=2.510e19;
Eex=0 00747 gvalue was adjust to match S&G ni (300K) value
8
‘Band Gap vs. T
EGO=1.177
a=4.7300-43
b=636;
EG=EGO-a.* (T.*2)./ (T+D) +
%
sEffective mass ratio (mnr=mn*/m0, mpr=mp*/m0)
mnr=1.028 + (6.11e-4) .*T - (3.09e~7) .*T.*2;
mpr=0.610 + (7.83e-4) .*T ~ (4.46e-7) .*T.*27
%
@Actual ni calculation
nic. * ((P./300) .*(1.5)) .* ((mnz.*mpr) .* (0.75) } .*exp (~ (EG-Eex) «/ (2
AKT %Computation of n, p, and EF-Ei
if Nnet=0,
n
p=ni;
EFi=0;
elseif Nnet>0,
n=Nnet/2+sart ((Nnet /2)*2+ni*2) ;
peni2/nz
EFixk*T*log(n/ni):
else
p=-Nnet /2+sqxt ( (Nnet/2)*2+ni*2) :
i*2/p:
P
EFi=-k*T*1log (p/ni) +
end
’Printout of results
format. compact:
n
BP
EFL
(b) Results obtained employing the part (a) pro; are tabulated below. The Problem
2.17/2.18 part (d) and (¢) results are slightly different because of inaccuracies in teading
the elevated temperature values of nj from Fig. 2.20.
Pant| T (K)| Na(cnv3) (cov3) | Ex-Ei (eV)
@| 300 [0 10!5_[/1.00x1010 | 1.00x10!5 0.298
(b) set 1916 0 1.00x1010 | 1.00x104 0.357
300
450
nj (cov) n (cm)
(©) 9x15 | 10!6 | 1.00x1010 | 1.00x10!5 | 1.00x105 [0.298
@) 0 10! | 4.71x1013 |_1.19x10!4 [1.87108 [0.0359
)| 650 [0 10'4 | 1.146x10!6] 1.151x1016] 11411016] 2.44x10-4
2.20
‘There is more than one way to work this problem, with alternative approaches likely to
yield slightly different answers, The most straightforward approach is recorded here.
At the onset of degeneracy
Ep = E,—3kT «M-type semiconductor
Eg = Ey+3kT _...p-type semiconductor
and the maximum nondegenerate carrier concentrations are therefore
2-19 n = NceG@r-EOKT = Noe
p = Nye@- ENT 2 Nye3
However
n=Np w-type Si at room T
peNa -»p-type Si at room T
Thus
Npimax = Ne e3 = (3.22 10!) e-3 = 1.60 x 10!8/em3
Naimax = Nv e-3 = (1.83 1019) e-3 = 9.11 x 10!7/om3
Nc and Ny were computed using the expression Nov = (2.51 x 10!/em3)(7mn,p"/mn9)3?2
and the effective mass values in Table 2.1. The cited computational relationship is given in
the text below Eq.(2.14b). Also see Problem 2.13.
2.21
‘The MATLAB program listed below computes Ep — Ej vs. Na or Np up to the
nondegenerate limit (Va or Np = 1018/cm3) and yields results very similar to Fig. 2.21.
MATLAB program script...
%EF-Ei versus NA or ND (nondegenerate, fully ionized, 300K)
clear; close
%speci fication of basic parameters
0el0;
NB=logspace (13,18); | $NB-ND or NA
SComputation of EF-Ei versus doping
EFiD=kT.*log (NB. /ni);
Plot out Fermi level positioning
semilogx(NB, EFiD, NB, EFiA);
axis({1.0e13,1.0e18,-0.56 0.561);
grid; xlabel('ND or NA‘); ylabel('EF-Ei');
text (1.0e14,0.30, *Donort); text (1.0e24,-0.30, ‘Acceptor');
2-20 2.22
(a) Because mn” << mp” in GaAs, the density of states in the conduction band is
considerably smaller than the density of states in the valence band at a given energy
displacement from the respective band edges. To be precise, using the effective mass
values from Table 2.1,
fo(BetAE) _ al =(20667 - oss
BEE) (mg) 1052
The required sketch is shown below.
Be(E) >
mo
By 8 v(E) >
(b) Ej will lie [ABOVE] midgap. Because the density of states is smaller in the conduction
band, the Fermi level must be displaced from the middle of the band gap toward the
conduction band edge to achieve an equal number of filled states in the two bands.
(©) Employing £T = 0.0259 eV and the effective mass values from Table 2.1, we find
Ej = tothe + 3K inting* nn”) ..Eq.(2.36)
= Fee Fy . 30.0259) 1n(0.52/0.066)
or
Bj = S$ + oosoev
&j lies approximately 0.04 eV above midgap in GaAs at room temperature. (Note
that the displacement from midgap is considerably larger for GaAs compared to Si. How-
ever, the room temperature displacement of 0.04 eV is still only 0.04/1.42 = 2.8% of the
total band gap.) (d) Using the results of Problems 2.13 and 2.20, we know
No = 4.26x 10!7/cm3
..GaAs
Ny = 9.41 x 10!8/em3 at 300K
and
Npmmax = Nc @3 = (4.26% 10!7) 3 2.12 x 1016/cm3
Natmax = Nv e-3 = (9.41 x 1018) e3 = 4.68 x 10!7/em3
Please note that n-type GaAs becomes degenerate at relatively low donor dopings. This
fact is very important in the modeling of certain GaAs devices; constructive use is made of
this fact in other GaAs devices.
2.23
(a) Attemperatures T — OK there are essentially no filled levels above EF and all levels
below Erare filled. (The foregoing was pointed out in the Fermi function discussion.)
Moreover, as T — 0 K, freeze-out occurs and p — 0; all the acceptors are emptied of
electrons. Now, Ep can't be below Ey because that would mean p # 0. Likewise, Er
can't be above Ea because that would mean all of the acceptor are filled with electrons. We
therefore conclude Ey
