.v.
(3.3)
568
C.T. SAH, T.H. NING AND L.L.TSCHOPP
i
I
i
'
i
I
,
ill
N A : 2.2 X 1014cm- 3
'
I
1
't
E s = 2 . 5 x l O 4 volts/cm
qo + QOT+ QSS (crn-2)
q.
t~
4.0 x I0 I0
•
1.0 X I0 II
Z~
II e
•
•
A
Z~.A
•
A
o~
2 . 0 x I0 II
AA • X XX X
x
x
x
X ~ 0
D
× ×AA • OOOo
zL
,o~ IO
Fig. 6.
,
/
~
~
I
I
,
2O 30
9.0 x I0 II
I ~ ~ ,,I
5O
tOO T(°K)
,
I
I
2O0 300
,
5OO
Experimental surface conductivity mobility of electrons as a function of temperature.
Then the charge density fluctuation is 6p (r) = p (r) - pO ( z ) .
(3.4)
If tk°(z) and @~(r) are the electrostatic potentials due to p ° ( z ) and 6 p ( r ) respectively, then (3.1) may be rewritten as H = H ° + H',
(3.5a)
H ° = - (h2/2m *) V 2 - eq~° ( z ) ,
(3.5b)
H ' = -- e a4, ( r ) .
(3.5c)
where and H ° determines the average properties of the electronic states in the channel, but it is H' that scatters electrons in the x - y directions. There are two contributions to @ (r), and hence to H'. First, there is the spatial density fluctuations of the depletion layer ions and the surface
569
SCATTERING OF ELECTRONS
oxide charges; and second, there is the induced electron charge density due to these fluctuations (screening effects). For a weakly inverted and nondegenerate surface, we shall neglect screening. Furthermore, since scattering by surface oxide charges is the dominant mechanism, we shall neglect scattering by the depletion layer ions. With these assumptions, then (.) = e [., (.) - N, (z)],
(3.6)
where n~(r) is the local surface oxide and surface state charge density and N~ (z) is the density averaged over the x-y plane. Let AV/be an elementary volume located at r=R~= (Pl, z~), then the perturbation corresponding to (3.6) may be written as 27)
H' (r)
=
-
(e/O Y, zxv,
( R,)II
- R,I
i
(3.7) = -- (e/~,) ~ A Q i / l r -- R , I , i
where g=½(esi+eox ), AQi=AVI6p(RI), and the sum is over all of the elementary volumes. For an inverted surface in the classical approximation, the induced electrons are concentrated near the interface12,28). If we assume the local electron density to be proportional to e x p ( - e E s z / k B T ), where E s is the surface electric field and eEsz is the electron potential energy, then for the present experimental conditions, where Es~-2.5 × 104 V/cm, most of the electrons are localized within a distance of XcH from the interface, where XcH-~ 10 A at 30°K and XcH_~ 100 A at 300°K. Now the average thermal energy of an electron in the surface channel is kBT, so that its thermal wave length is
~',h = 2~/k = 2re (h2/2m*kBT) ~ . Using m* = 0.323 mo, we obtained 2th (30 ° K ) = 424 A and 2th (300 °K) = 134A. That is, 2th~>XcH, SO that the motion of the electrons perpendicular to the surface may be neglected. Therefore, we shall assume that the induced electrons form a two-dimensional gas at the interface, so that the matrix element between two plane-wave states [k> and Ik'> is
qlzi[ + i p i ' q ] ,
(3.8)
i
where q = ( k ' - k ) , and A is the area of the surface. Now, for a two-dimensional r a n d o m Poisson or Gaussian distribution of the surface oxide and surface state charges, we have
and
<(AQi)Z)av = e2Nl(zl) AVi,
570
c . T . S A H , T. H . N I N G A N D L. L. T S C H O P P
so that <]
= (e2/g) 2 (2n/Aq) 2 ~, N, (zl) AV/exp [ - 2q IzJ] i
(e~l~) ~ (2n/q) 2 (l/A) f N~ (z) exp [ - 2q Izl] dz.
(3.9)
The transition rate is given by the golden rule as
Ck, k, = (2n/h) (l(t,'l H ' Ik)12),v a (Ek -- E~,),
(3.10)
and the momentum relaxation time z is given by
1/~ = y, rk~, (1 - cos 0),
(3.1 l)
k'
where 0 is the angle between k and k', and the sum is over the two-dimensional wave-vector space. Substituting (3.9) and (3.10) into (3.11), we obtain
l/~ (e) = (2~/h) (e2/~) 2 (1/e) ( NI (z) ~ (e, z ) d z ,
(3.12)
where E = Ek= hZ k2 /2m *, and
G(E, z) - t- e x p [ - 4kiz[ sin qS] dq~.
(3.13)
I/
0
The electron mobility is given by
tq = (e/m,,) ( E r ) / ( E ) ,
(3.14)
where m s is the mobility effective-mass and (A)-
t'Aexp(-E/k"T)
dE
(3.15)
0
is the average value of A for a two-dimensional nondegenerate gas. We have evaluated/~ for three assumed distributions of oxide charges:
(i)
N,(z) = N I '0(Z)
;
(ii) N l ( z ) = ( N J ~ ) exp (--z/s); (iii) Nl(z ) = (Niz/fl 2) exp ( - z/fl) ; where z is measured from the interface into the oxide. The results are plotted in fig. 7. Case (i) is of particular interest. It corresponds to the case where all the surface oxide charges lie in the oxide at the oxide-semiconductor interface. The mobility in this case takes a simple form of I~ = (e/mu) (~/e2) 2 (2hk.T/n2NO • (3.16)
SCATTERINGOFELECTRONS 103
I
I
~
....
I
~
~ ~ ~ I
571
~
I
(2) (I)
.....
I./
.,'
/"
J ~
/'--'~74
.,,./4
/
/
_%/ Nz = 9.0 x IOllcm-2 :
2
~Oz'
m*: 0.323 % ....
/'
m/z =U.tz)~m o
/
/q
/'/
•v," ,,q,.. /
:8.o
,,
/
//
~"./,,L<-,; / ~/.',; ~ / ~
.
.. / ¢ ,
/-.:
,C;U< /
,.3
E i "> o 102 v
..
/ # / /y
1
I
/ -I |
/
i
i
I.I //.7 iI
5
/./
I0
II
A 20
I /2 / I //./// ,j,.~///
I 30
X
/
I
50
~ k
~1 I00 T(°K)
z/
"~,u.:wT
~
I 200
I 500
500
Fig. 7. Theoretical two-dimensional electron mobility due to scattering by surface oxide charges for three different charge distributions. (1) Ni(z)~ Nil(z); (2) Ni(z)= - (NUa) exp ( -- z/a); (3) N1 (z) -- (Niz/• 2) exp ( -- z/fl).
The same result could also be obtained from the two-dimensional scattering cross-section in the Born approximation given by Stern and Howard:8). The dependence p ~ o z T / N i agrees with the experimental observations (figs. 5 and 6). However, using the bulk values of mu=0.259 mo and ~=8.0, we found that p~ given by (3.16) is about 20 times smaller than the observed values at low temperatures. The discrepancy could be due to one or all of the following factors. (i) There is some screening effect. (ii) The surface oxide charge has a finite distribution into the oxide layer (see fig. 7). (iii) Surface layers of Si and SiO2 may be more polarizable than bulk Si and bulk SiO2, so that the effective value of ~ is larger than its bulk value. (iv) The distribution of the surface oxide charges may be correlated instead of random. Correlation has the effect of reducing scattering and hence increasing the mobility which will be discussed in detail in a future paper. It appears that (ii) and (iv) are the important factors. We note that eq. (3.16) differs from the two-dimensional Rutherford formula 8) of fq = (3 x/2/4\/~) (g/e) ~/(kBT/m*)/N,,
572
C . T . SAH, T . H . N I N G
AND L.L. TSCHOPP
which is inversely proportional to the coupling constant, (e2/g), and has a x/T temperature dependence, while (3.16) gives p~-loc(eZ/~)z and has a T dependence.
4. Scattering by surface phonons At high temperatures and/or for low surface oxide charge densities, scattering by lattice vibrations cannot be neglected. This is evidenced by the measured mobility deviating from its linear dependence on T (fig. 6). A n estimate of lattice scattering mobility, /~L, may be extracted from experimental data by using the approximation 29). 1/pL = l/,Uexp- l//ui.
(4.1)
Here Pi is assumed to be proportional to TIN I and fitted to the low-temperature data. The resulting #L is shown in fig. 8. The theoretical curve, which we shall derive below, is based on a two-dimensional model. For T < 100°K,
105~_
I
'
I i i bi l
i
I
I
/
I ( Q o + Q +Qss )
•
I x I011 crn-2
×
2 x IOIIcm~2
D
4 x l O I I c m~2
q_
9 x I011 crn-2
5 ~
I
\
4 x IolOcrn-2
~.
-
~
Q
5
+ °~ J,
2
°
103 I
20
50
,
I
50
i i i i I
I00 T(°K)
i
I
200
iX
500
Fig. 8. Surface phonon scattering mobility of electrons as a function of temperature. The theoretical curve is given by eq. (4.8) of text, with Pae = 7.4 × lO~/Tcm2/V-sec, he)a/k~ = 650°K, and ZR/ZA = 2.3.
SCATTERINGOF ELECTRONS
573
where only surface acoustical phonons are important, the data can be fitted to pL=7.4xlOS/TcmZ/V-sec. For T>150°K, it can be fitted to PL= 108/T2 cm2/V-sec. The additional high-temperature mobility drop can be interpreted in terms of surface optical or intervalley phonon scattering. In the two-dimensional deformation potential theory of surface phonon scattering, the electron-phonon coupling matrix elements have the same form as that of the three-dimensional bulk case. Thus, for scattering by surface acoustical phonons we have 3°) I(k'i Hac Ik _ q ) [ 2 = (ZEhe)q/ZpAu 2) (Nq q- ½ ~ ½) (~k',k±q,
(4.2)
where he)q is the acoustical phonon energy with_ wave vector q, Nq is the phonon occupation number, p is the areal mass density of the semiconductor sheet, ut is the longitudinal sound velocity, A is the area of the surface, and Z A is the acoustical surface deformation potential in eV. Eq. (4.2) is identical to the form used by Kawaji 2a). The relaxation time and the mobility derived from (4.2) are 2a) l/'rac = (m*Z2kBT/h3pu2), (4.3) and ,uac = (e/m,) (Ezac)/(E) = (eh3pu~/m*muZ2kBT), (4.4) where ( ) has the same meaning as in (3.15). Thus, at low temperatures the mobility due to surface acoustical phonon scattering is proportional to T-1. It is noted that (4.4) also holds, within a factor of order unity, for scattering in a three-dimensional thin film 22). At higher temperatures, optical or intervalley phonons cannot be neglected. The corresponding matrix elements may be written as 31) I(k'l Hop rk _+ q)l 2 = (Z~he)R/2pAu~) (NR + ½ -T- ½) 6k, k±q,
(4.5)
where he)R is the phonon energy which is assumed to be independent to q, NR is the phonon occupation number, and Z R is the electron-optical or intervalley phonon coupling constant in eV. The relaxation time is
l/zop = (m*Z2he)R/2h3pu 2) INR + (NR + 1) u (E - he)R)],
(4.6)
where u(x) is the unit step function, u ( x < 0 ) = 0 and u ( x > 0 ) = 1. For combined acoustical and optical phonon scattering, the relaxation time is given by a°) 1/'CL = 1/Za¢ + l / t o p = (1/rat) {I + (Za/ZA) 2 (he)R/2kaT) [Na + (NR + l) U (E -- he)R)]}. (4.7) The mobility for this two-dimensional case can be obtained explicitly since
574
C . T . SAH, T . H. N1NG AND L. L. TSCHOP P
the density of state is independent of energy. This is given by
PL = (e/m,)
1 -- [1 + X0R]exp(-- X0R) [1 + XR °] exp(-- xR°) -1 +--I~[Z,/ZA] - - - - ~ 5 . ~ 5x,N, ~ 7 + 1 + ½[ZR/ZA] 2 XOR[2NR + i] f '
(4.8)
where x°=he)R/kB T. Using (4.8), p.~=7.4x 105/T cm2/V-sec, he)R/ka= =650+50°K32), and ZR/ZA =-2.3, a good fit to the experimental data was obtained as shown in fig. 8. The physical significance of this good fit may be explored by comparing the theoretical P,c of (4.4) with the low temperature data of pat=7.4 X 105/TcmZ/V-sec. Using the bulk values of m*=0.323 too, m=0.259 mo, u t = 9 . 0 x 105 cm/sec, and ZA=7.65 eVa3), we obtain p = (30 A) x (2.33 g/ /cm 3 )= (30 A) x Pb,~k, where /)bulk=2.33 g/cm 3 is the mass density of bulk silicon. This gives an effective thickness of the surface layer of 30 A, which is similar to the thickness of the electron layer of Xcn=kBT/eEs= 10-100 A. This agreement further substantiates the two-dimensional model of electronacoustical phonon interaction involving surface sound waves. 5. Discussion and conclusions
We have demonstrated that on a weakly inverted silicon surface covered with thermally grown oxide, electron mobility is determined mainly by surface oxide charge scattering and by surface phonon scattering. These scattering mechanisms can be described by the classical two-dimensional models. The effects of the surface oxide charge should be less important for a heavily inverted and degenerate surface because of screening, but then the surface mobility becomes relatively independent of temperature, giving us less information on the scattering mechanisms. Our analysis of ion scattering emphasizes the spatial density fluctuation of the scattering centers. The two-dimensional theory accounts well for the temperature and density dependences of the surface conductivity mobility. Such spatial density fluctuations would also cause fluctuations in the surface potential and the quantum energy levels in the surface channel which will be presented in a future publication. We have extended the two-dimensional theory of surface acoustical phonon scattering to include optical or intervalley phonons. Comparison with experiment indicates that for T> 150°K, the effects of the optical or intervalley surface phonons are important.
SCATTERING OFELECTRONS
575
References 1) O. Leistiko, A. S. Grove and C. T. Sah, IEEE Trans. Electron Devices ED-12 (1965) 248. 2) F. F. Fang and W. E. Howard, Phys. Rev. Letters 16 (1966) 797. 3) J. Grosvalet, C. Jund, C. Motsch and R. Poirier, Surface Sci. 5 (1966) 49. 4) E. Arnold and G. Abowitz, Appl. Phys. Letters 9 (1966) 344. 5) F. F. Fang and A. B. Fowler, Phys. Rev. 169 (1968) 619. 6) T. I. Kamins and N. C. MacDonald, Phys. Rev. 167 (1968) 754. 7) N. St. J. Murphy, F. Berz and I. Flinn, Solid-State Electron. 12 (1969) 775. 8) V. N. Dobrovol'skii, Yu. S. Zharkikh and L. N. Abessonova, Fiz. i Tekhn. Poluprovodnikov 5 (1971) 723 [Soviet Phys.-Semiconductors 5 (1971) 633]. 9) T. Sato, Y. Takeishi, H. Hara and Y. Okamota, Phys. Rev. B 4 (1971) 1950. 10) C. T. Sah, J. R. Edwards and T. H. Ning, Phys. Status Solidi (a) l0 (1972) 153. l l ) A. Ohwada, H. Maeda and K. Tanaka, Japan. J. Appl. Phys. 8 (1969). 629. 12) F. Stern, in: Proc. Tenth Intern. Conf on the Physics of Semiconductors (U.S. Atomic Energy Commission, Division of Technical Information, Washington, D. C., 1970) p. 451. 13) R. F. Greene, in: Solid State Surface Science, Vol. l, Ed. M. Green (Marcel Dekker, New York, 1969) p. 87. 14) J. R. Schrieffer, Phys. Rev. 97 (1955) 641. 15) R. F. Greene, D. R. Frankl and J. Zemel, Phys. Rev. 118 (1960) 967. 16) R. F. Pierret and C. T. Sah, Solid-State Electron. 11 (1968) 279. 17) R. F. Greene and R. W. O'Donnell, Phys. Rev. 147 (1966) 599. 18) F. Stern and W. E. Howard, Phys. Rev. 163 (1967) 816. 19) C. B. Duke, Phys. Rev. 168 (1968) 618. 20) E. D. Siggia and P. C. Kwok, Phys. Rev. B 2 (1970) 1024. 21) A. V. Chaplik, Zh. Eksperim. Teor. Fiz. 59 (1970) 2110 [Soviet Phys.-JETP 32 (1971) 1143]. 22) V. Ya. Demikhovskii and B. A. Tavger, Fiz. Tverd. Tela 6 (1964) 960 [Soviet Phys.Solid State 6 (1964) 743]. 23) S. Kawaji, J. Phys. Soc. Japan 27 (1969) 906. 24) H. S. Fu and C. T. Sah, IEEE Trans. Electron Devices ED-19 (1972) 273. 25) M. Luong and A. W. Shaw, Phys. Rev. B 4 (197t) 2436. 26) R. F. Greene, D. Bixler and R. N. Lee, J. Vacuum Sci. Technol. 8 (1971) 75. 27) See, e.g., J. D. Jackson, Classical Electrodynamics (Wiley, New York, 1962), p. 111. 28) D. R. Frankl, ElectricaIProperties of Semiconductor Surfaces (Pergamon, New York, 1967) p. 21. 29) P. P. Debye and E. M. Conwell, Phys. Rev. 93 (1954) 693. 30) E. M. Conwell, High Field Transport in Semiconductors, Solid State Physics Supplement 9 (Academic Press, New York, 1967) p. 108. 31) Ref. 25, p. 150. 32) G. Dolling, in : Inelastic Scattering of Neutrons in Solids and Liquids (International Atomic Energy Agency, Vienna, 1965) p. 249. 33) C. T. Sah, unpublished.