Professional Documents
Culture Documents
1-s2.0-0038110180901690-main
1-s2.0-0038110180901690-main
63-71
Q paluaon Press L~I...1980. Rid i Great Btitain
G. BACCARANI
and M. RUDAN
Istitutodi Ekttronica, FacolP di Iagcgmia, Univcrsitf~di Bologaa, Bologna,Italii
G. SPADM
CNB-L&oratorio
Lamcl,viade Castagnoli1, Bologna,Italia
and
Abduct-The well-known C-V tchiquc for dctcmining the doping profile in s&conductors is reexamined. A
simpliticd m&l is dcvcloped, which takes into account the e&t on the C-V characteristics of the diffusion
polentielia~~~~oftbesemiconductor.Thissrodelkdsannl~ytothesunecondusion
inferred by Kennedyet rrl.frqma few numericalexampks, i.e. the zmwbias majority carrier concentration.instead
of the impurity proflk. is provided by the expression (C3/qes). (dad V)-‘. Parametric formulae for the C-V
cm are propsed in excelkntagreementwith some results of a numerical analysis.
mayarisewhenthe impurityprofilechangesrapidlyover
the extrinsicDebye length.All the above papersrely on
a few numericalexamples.
Finally, Verjans and Van Overstraeten (VVO) con-
sidered the effect of the Fusion voltage arising from
the Merencc in doping level between the bulk of the
semiconductor and the edge of the spaccchafge
layer[l3], and determined parametric formulae for the
capacitance-voltagerelation.
It is shown in.Appendix A that. in principle, an exact
determination of the doping profile can be obtained by
solving a 3rd order pa&i differential equation with
suitable boundary conditions. From the Poisson equa-
tion, the doping profile is then expressed in terms of the
2nd derivative of the electric potential. Because of the
large errors usually associated to the evaluation of the
2nd derivativeof a numericallydeterminedfunction,
however, such a procedureis not warranted,and we
shallnot attemptsuch a methodof dution.
Instead,we deriveanalyticallyKMKs result by means
of a simpMed model which, whik assuming an abrupt
space-charge edge, takes into account the effect of the
quasi-neutralregion on the potential dishibution within
the semiconductor So w we show that the effect of I
devices). Equations (4) imply an abruptspace charge DilTerencingqns (5) and (7) with respectto x, yields
edtje(ASCFJbut do not involvethe ident&ationof n(x)
and N&x) in the quasi-neutralregion. From the dQ, = q&A dxt (8)
assump& in eqns (4) it follows that the surfacecharge
0, is given by
and
0, = 4 f NJ(x) dx + 9 I,: U%(x)- no(x)1
ti (5)
The space charge capacitance C, is thus given by Ng = 10" cm-q No=9.9X10’6cm-3 and o=JOOA
depending on the surface potential, and between
c, +!p, -0.97 k?Iq and -1.19krrlq for (~=xKKIA. For an
(10) exponential doping profile
s I
I0
n&x)dx =
I
Y
bow - rxll dx (13)
WhCEtheyStlltCthUthCfESttamOftbtigltth8DdSidC
-Ix;
x *bow- &e&l1 q. (14) (q/e) *x#*ND&) dxr gives the variation for 4, due to a
further depleting of the space charge region while the
term (q/e,)x,[(ND(k)- n&)1 dx* gives the contribution
One can easily prove that for a homogeneously doped to d4, of the variation of the diffusionpotential. So the
semiconductorthis contrii yields - Wq rcg8rdless parametric expressions derived by (VVO)contain twice
of the applied voltage or impurity concentration[l4]. the contriition of the d&&m potential and therefore
For spatially varying protiles we computed this con- will yield ns vahres for the majority carrier con-
tribution using a program which solves the Poisson centration as will be shown in the next section.
equation. Stat&g from a given imp&y profile we From qns (9) and (10)immediatelyfollows that
obtained both ndx) and R&X) for di&rent values of
the surface potential.
(19)
By using qn (13)we determmedx, and used this value
to evaluate A4, in qn (14).This was done for d&rent
shapes of the impurity profile. only for very steep which corresponds to the conclusion reached by (KMK)
pro&s will the variation of A4, with 4, be important on a numericalbasis. From qn (19)it is straightforward
and will eqn (10)not be correct anymore. to derive the relationship
For example, for Gaussian doping profiles as used in
Section 3 of this paper where 09)
ND(X)= NB+ Noexp ( - x’12a3 (15)
where Vistheappliedvoltageandwhichisvalidfor
A4S varies between -1.05kZJq and -1.36k7lq for both !Ichottky barrier diodes and MOS capaWrsr5l.
68 G.BACCARAM &al.
3. CAPACITANCE-VOLT- BELATIONSHP
4s = -~M9)[~xJLD)I
t (iVdNB)(UL~)*[l -(I+ XL) exp(- xdL)l
-(x&J .[I + WdNd - expbLll-’
W(O)= &*
&6(L)= 0. (B3)
for x<x, @Ia) The increment of the semiconductor charge following the
change &$, in surface potential is now given by
and
@lb)
which is obtakd from eqn (12) of the text by sttbsbtutii ND(X) =-q*ns
to Ao(x).In all the prackal situ&ins investigakd in this work,
expWWkT) W(x) dr W)
(Bl) has msidted an excekt zemimkapproxktionofthe
ekctricpoMtial,sothatafewiteratkmsmmnaUysut7icedto and finally. the semicbnductor m is
reach convmgetmy.
InordertogetaroughestimakoftheaccuracyoftheEnal
C*=+$ CBS)
SOhlththCilllpOdSWfpcepOtCll~~,~ScompsndWiththe I
integral
With the above pro&ure, the semiconductor capacitance is
-fde+Ndx) - ne *exp(cp$(xMml*x dx dctmmcd in tams of dilkrential equations rather than by
differencmg of space-charge solutions at different surface poten-
tials.
SO doing round4 error are greatly reduced, and an
and the deviatkms were always less than 0.1%. accuracy of the order of 0.1% is expected.