SiO2-Photox interface properties studied by photo- and bias-induced charging

Journal of (‘r~stal (,rowth 56 1 988 8

8 ~9

North Holland. Amsterdam

Hg 0.7 Cd ~Te

/ S102-PHOTOX INTERFACE PROPERTIES STUDI ED

BY PHOTO- AND BIAS-INDUCED CHARGING A. KSENDZOV and Fred H. POLLAK Pussies Department. Brook/tn College of the

(ill

Unit croft of \ etc

orA, Brook/in. Sets torI. II ‘/0. 6 5.4

and J.A. WILSON and V.A. COYfON Santa Barbara Research C enter, Go/eta. California 9/Il

.

C ~4

We hase investigated the effects of light illumination in the waselength range 1 pm to 2200 A (with and without DC’ bias) on the Hg0 Cd11 5Te (n-t\pe) Si02 Photox 154 interface at 77 K in the metal insulator semiconductor desire configuration. Illumination of waselength A without bias produced a flatband shift (~Uw)towards positive values. Fhe saturation salue of ~ has a pronounced peak at A 2800 A and is negligible for A > 5100 A. The light induced charge is maintained for at least 8 h at 77 K hut leaks off if M (at or near the interface) withresults an energs distribution centered aboutand 45 discharging eV ,ihose the the device is heated briefly to 300 K. We propose a model to explain these insolving light induced charging of serv slow trap band. states With in thea SiO,-Photox’ oxide salence positive gate voltage charging is extended to waselengths of I pm (limit of our measurements), the resulting charge quickly leaking off This latter effect is related to charging of faster interface states that are apparentls different from thy slot’, states di’,s..usscd above.

I. Introduction Passivation of Hg~ 5Cd 5Te by deposition of TM has beforeign insulator such as Si02-Photox come technologically important due principally to its chemical inertness coupled with the resultant high quality interfaces [1]. It is important to ha~e information about interface traps since they can limit the carrier lifetimes in junction devices or, when charged, alter the effective gate bias of metal semiconductor insulator (MIS) devices. The Hg~ 5Cd5Te interface in the MIS configura-7 tion has been studied in the frequency rangeon I i0 Hz [2 51. Work in the 1 10~ Hz region the Hg~ M interface [5] re‘,Te/Si02-Photox’ vealed5Cd a considerable number of interface traps (> lot cm 2 eV 1) Illumination effects in the very low frequency domain (‘— 10 Hz) on Hg 1 ‘,Cd,Te MIS devices fabricated with a native oxide indicated a total trap density in the oxide as high as 6 x IO~~ cm 2 [61. We have recently re-

ported electroreflectance results at 77 K on a Hg 55 7Cd~Te/SiO2~PhotoxfMMIS structure [71. In order gain information about the states at or to near thefurther Hg~ ‘,Cd M Si02-Photox’ interface we have investigated5Tethe influence of illumination (with and without gate bias) tn the ssavelength (A) range 1 ~zm to 2200 A on the capacitance voltage (C V) characteristics of an MIS device at 77 K. The effect of the light without bias is to shift the flatband voltage (Un,) to positive values. The shift in Uffi (~Ufh)peaks at A 2800 A and charge is negligtble for A>for 5100 A. The light-induced is maintained at least 8h at 77 K but is eliminated by brief heattng at 300 K. The main effect of the bias is to extend iUih to longer wavelengths. However, in this case the charge relatively quickly leaks off. To explain our observations with zero bias we propose a model involving the light-induced charging and dischargtng of very sloss trap states (characteristic frequency <10 s ) in the SiO~PhotoxiM (at

0022-0248 88 $03.50 Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

A. Ksendzov et al.

Hg

55

7Cd0

5Te StO, Photo’c interface properties

or near the interface). The bias-induced effect involves faster states that are apparently different from the very slow states responsible for the zero bias photo-charging. Thus the zero bias model does not account for the bias-dependent phenomenon. At the present time we do not fully understand these bias-dependent experimental results.

835

380

~ 6mm

0 mm

Sm

12mm I3mu,

350

~

340 Hg07Cd03Te/Si02 330 77K X 2800~s

320

2. Experimental details

I

The structure used in this experiment consisted of the same MIS device used in ref. [7]. All lightinduced studies were carried out at 77 K. To provide illumination we used light from a xenon lamp passed through a 0.3 meter grating monochromator (GCA/McPherson model 218). In order to find the photon flux (at A) the power of the monochromatic light was first measured at 5100 A with a light power meter (International Light Power meter model 1L510) and than the calibration was extended to shorter wavelengths using a Si photodiode (EG&G model 4000B) with a known spectral response curve. chargingwas of TM The interface the Hg03Cd07Te/Si02-Photox determined by C V measurements at a frequency of 1 kHz, the amplitude of the ac probe voltage at the gate being 50 mV.

I

0

2

4

6

8 0 12 U90, volts)

4

6

18

20

Fig 1 (‘apacitiince voltage curves at 77 K for different time exposures of 2800

A light at

zero bias.

effect of all preceding illumination periods. After U0, reached a value of 20 V (or at the end of the day, whichever came first) the sample was heated to 300 K which restored it to the initial, uncharged state. This was checked by cooling the sample down to 77 K and taking C V measurements after the 30 mm heating procedure. In fig. we present set of by C V curves from our I sample aftera typical illumination 2800 A light for different periods of time. The value of U 0, in each measurement was determined to be the voltage corresponding to a capacitance of 375 pF [8].

3. Experimental results

_________________

20

20

3.1. Light illumination at zero bias voltage

5

280o~

The sample (in the open circuit configuration) was first illuminated by light at A from the monochromator for a given amount of time (I). The C V measurements were then taken in the dark. this procedure a new illumination period After was started. lished that if the sample was kept at 77 K in the From the C V measurement we have estabdark after the above illumination procedure, the light-induced charge remained unchanged for at least 8 h. This demonstrates that the trap mechanism is very slow with a characteristic frequency < 10 s ~. Thus after every new illumination period we have actually measured the cumulative

-~

5

0

0 5

0

Hg0 7Cd0

3

4

5

3Te/S 02

6 7 t6IcrnO

PlIO

: 2

3

4

P (I0’8/cm2) Fig. 2 Photo-induced flatband shift. ~U 2, P, for light of wavelength 0,, as a function of the total of2200 photons per inset cm shows an expanded version 3300, number 2800 and A. The of the 2800 A results. The solid lines are a least-squares fit to eqs. (5) (7) and (10).

.4 Ksend—os’ et al.

8 ~6

Hg,

C 1,, Te SiC) Photos interim

Plotted in fig. 2 is the light-induced flathand shift, ~Ufh, as a function of the total number of incident photons per unit area, P(hw). for three different wavelengths. The inset shows an expanded version of the 2800 A data. Note that for all three wavelengths ~Ufh tends to saturate as P increases. The saturation level for ~Ufh peaks at A 2800 A while for A> 5100 A we have found no significant light-induced charging.

3.2. Light illumination with gate bias Application of a negative bias to the gate pro duced no appreciable change in the charging pieture described above. However, the application of a positive bias made it possible to charge the system by utilizing A> 5100 A. The result of illuminating the structure by A I ~sm light is presented in fig. 3. Curve (1) is the C V measurements with no light exposure and no bias; this curve is the same as the t 0 data of fig. 1. The C V results after exposure to 1 ~tm light for one hour and a gate bias of + 10 V is shown in curve (2.). The light and the DC bias were then shut off and the C V evaluated after 10 mm in the dark as displayed in curve (3). The results of fig. 3 indicate that I h of illumination under + 10 V bias has increased UIh from approximately I to 4 V. Fig. 3 shows that the charge leaks off relatively quickly since USh after only 10 mm toofabout relaxation at 774 K the dark was reduced 2 V from V.inThis is in sharp contrast with the case of charging by

390

~.

~ a o

370 ~

\f~/7

330

~ t3 V

~

illumination only and suggests a different charging mechanism when bias is applied.

4. Discussion 4.1. Charging ht’ light illumination

nit/i

zero bias

to//age

The results of figs. I and 2 can he explained by the presence of very slow electron traps at or near tM interface. The the Hg 557Cd5~3Te/SiO2-Photox fact that the photo-induced flatband shift has a maximum for A 2800 A suggests that ts~omechanisms are operative; i.e., both charging and discharging. For 2800
310 290 270

properties

Hg0 7040 3Te Sic2 ‘

-,

scae. As shown in the figure two processes can oc-

gate

cur: (1) Charging. An electron can be photo-excited from the SiO, valence band into a trap state, some of the resultant holes moving into the Hg55 5Cd557Te

Fig. 3. Capacitance soltage curves at 77 K for (1) no light exposure. (2) after exposure to 1 pm light for 1 h and a bias of + 10 V and (3) 10 mm after the light and bias were shut off,

region. This process can take place for photon energies (ha) less than or equal to E5 2’ We de-

01’

~ U

volts)

A. Ksendzov et al.

Hg

0 7CcI0 5Te S02 Photox interface properties

837

to both (I) and (2). We can write: N1 ( h ~)

ffh~D(E) dE, ~ 1E52

E9,2

{

h~D(E)dE

hw < Eg2/2,

(Ia)

hw>E62/2.

(Ib)

Ni2(hw) 0,

~Jhw

Hga 7T

D1(E)dE,

ht~
(2a)

hw>Eg2/2

(2b)

In fig. 4, the dark region corresponds to N1 while the shaded region represents N1 2’ In order to develop our model more fully we define the following quantities: n1(n12) Fig. 4. Proposed model of the Hg07Cd03Te(n-type) SiC)2PhotoxTM mnterfacial region including a distribution of electron traps in the SiC)2 centered around E55 with full width at half maximum of tE55. For the sake of clarity the energy gaps of the Hg(5 7Cd0 5Te ( 250 meV) and SiC)2 ( 9 eV) are not drawn to scale. The dark region represents the states per unit area accessed by the charging only process while the shaded region indicates the region of charging and discharging.

n

number of traps per unit area filled in the N~(N12) part of the trap distribution;

total number of filled traps per unit area.

Therefore, for process (1) we can write’

dn~

“~7

Qi ( P/t) (N1

n~

(3)

while for processes (I) and (2): note the quantum efficiency of this process as Q1. (2) Discharging. For photon energies greater than E62/2 not only does charging take place as discussed above but, in addition, electrons in the trap states can be photo-excited into the conduction band of the SiO2. Some of these electrons can be transferred into the conduction band of the Hg55 7Cd0 ~Te. The quantum efficiency of this effeet is denoted as Q2. Note, that in order for process (I) to cause charging, the photoexcited hole must move into the Hg~Cd07Te against the action of electric field whereas in process (2) the excited electron which produces discharging is pulled towards the Hg55 5Cd07Te by the electric field. Thus we expect process (2) to have a higher probability than process (1), i.e. Q2 >> Q~. We denote as N1 the number of states per unit area available to mechanism (1) but not mechanism (2) and as N1 2 the number of states available

dn~,2 Q1(P/t) (N12

n12)

—

Q2(P/t) n~2.

dt (4) The solutions to eqs. (3) and (4) can be written as: n1(P) — N1 [I exp( — Q~P)j. (5)

nt2(P) —

N~2Q1 ~

~1 +

{1

exp[

(Qt

+

Q2)P]

}, (6)

and ,~[P( ha)]

n1 [P( hta )I

+

ni~2[P( hw)]

.

(7)

Eqs. (5), (6) and (7) show that n[P(hca)] saturates as P —s ~ We denote the saturation value of n[P(hw)] as n5(hsa), such that from eqs. (5), (6) and (7): ,~(.hw)

—

N1(hta)

+

Ni2(hta) ~, Qi +

~,

(8)

4 K setid:ot et al

538

Hg,, C d

0

Te .S io Photos ittterfw e propertir

It is also convenient to express n[P( ho.’)] in terms of ii ( ho.’) as follows:

Hg

7Cd0 3Jv S

2

“‘K

n[P(hw)J LU

Expi

‘

• Theory ii

[1

exp(

x{exp(

QtP)]+\’l~Q~Q

Q~P) exp[

(Q1+Q

O8~

)~1}

,03

06

1’

(9) Comparison can thus he made hetsseen the experimental results of fig. 2 and our model. Assuming for simplicity that the charging takes place near the insulator semiconductor interface, the photo-induced flatband shift. ~U10.can he related to the number of photo-filled traps n[P( hw)] hs the relation ~lh

[ P( ho.’ )j

e n [P( hw )1

C

.

(10)

where C is the capacitance of the insulator (SiO,) layer. Thus iL~>can be used a direct measure of the number of filled traps. n. From eqs. (5) (7) and (10) the dependence of ~b51, on P(hw) can he fit using as adjustable parameters N1. N1 and Q~.Shown by the solid lines in fig. 2 is a least-squares fit of the experimental data to these equations for the three different wavelengths. The quantities N1, ~ ~. Q~.and Q~obtained h\ the above fitting procedure hase a fairly large standard deviation. However, the parameter n,( h0) can he obtained with a high degree of reliability using eqs. (9) and (10). Shown in fig. ~ by the open circles is n5(hw) determined from experinient as a function of ho.’. Note that there is a pronounced peak at about 4.7 eV and that njhw) 0 for h~< 2.4 eV. The photon energy dependence of n5( ho.’) can he accounted for on the basis of the model in fio 4 and eq. (8). From the abose considerations (least-squares fit in fig. 2) we find that Q~ 10 Q1 and hence for our case eq. (8) becomes n,(hw)~N1(hw)+1~5A1~(hw)

(11)

To calculate N1 and N1.’ (see eqs. (1) and (2)). we have assumed a Gaussian distribution for D1(E) having E55 4.5 eV and 6E55 0.5 eV. For E2 we take 9 eV which is close to the hand gap of SiO,. Our value of E55 corresponds to the energy of traps

02I

C

~

‘

~

4 0

‘

I r

Fig

‘~

5 0

‘

Erergy

e

I’sperimental (open circles) and theism cOral (closed circles) salucs ot n (/iL I

in thermal Si02 produced h\ P and In Impurities [9].Shown by the closed circles tn fig. 5 are the salues of ii5(ha) calculated using the ahove procedure. There is good agreement hetsseen expertment (open circles) and theory (closed circles). 4.2. Charging hi’ light t//ununarion

ott/i posttts

e hta.s

As we mentioned before, application of positise gate bias extends photocharging effect towards lower energies of incident light ( 1.2 eV was the lowest energy of our measurements). The accu mulated charge leaks off in a matter of minutes rather quickly in comparison with the preceding case. The lower light energy limit of the effect suggests that the carrier generation h~ lmghi in Hg55 Cd(5 Te substrate is insolved. Also, the faster charge leak-off points to a different nature for interface states that hold the charge. At the present time we are not able to propose a detailed model explaining this effect.

5.Summar~ We have investigated the effects of illumination at 77 K, with and without DC bias, on the C V characteristics of a Hg55.’Cd551Te SiO~~Photoxi’~ MIS device. The light-induced charge for zero bias is maintained for at least 8 h indicating sers slow

S S S S

A. Ksendzou et al.

Hg

0 ~Cd555Te SiC), Photox interface properties

839

rates having a characteristic frequency < 10 ~, The saturation value of z~U0, for zero bias

search Project Agency contract MDA-903-83-C108.

exhibits a pronounced peak at A 280(1 A (4.7 eV). These results can be explained by a model

References

involving light-induced charging and discharging of very slow trap states in the Si02 near the

[1] J.A. Wilson and V.A. Cotton, J. Vacuum Sci Techno

interface. These trap states have an energy distribution centered about 4.5 eV above the Si02 Valence band. With a positive gate voltage the charging mechanism is extended to longer wavelengths, the resuiting charge quickly leaking off. At the present time we do not have a detailed model to explain the bias-induced phenomenon.

Acknowledgements

(1985) 199.

[21 W.F.

I A3

Leonard and M. Michael, J. Appl. Phys. 50 (1979)

1450. [3] Y. Nemmrovsky and 1. Kmdron, Solid State Electron. 22 (1979) 831. [41 MA. Kineh, in: Semiconductors and Semimetals, Vol. 18. Eds. Wmllardson and A. Beer (Academic Press. New York. 1981) p. 313. [5] G.H. Tsau, Z. Sher, M. Madou, J.A. Wilson, V.A. Cotton and C,E. Jones, J. AppI. Phys. 59 (1986) 1238. [61 RB. Shoolar, B.K. Janousek, R,L. Alt, R.C. Carlscallen, M.J. Daughiery and A.A. Fote, J. Vacuum Sci. Technol. 21 (1982) 164, [7] A. Ksendzov, F.H. Pollak, J.A. Wilson and V.A Cotton.

A.K. and F.H.P. acknowledge the support of Battelle, Columbus, OH, the Night Vision Electro-Optics Center, Ft. Belvoir, VA, and the Army Research Office. J.A.W. and V.A.C. acknowledge the support of Defense Advanced Re-

AppI. Phys. Letters 49 (1986) 648. [8] E.H. Nmeollman and A. Goetzberger. Bell System Tech. J. 46 (1967) 1055. [9] R.F. DeKeersmaeeker, D.J DiMarma and S.T. Pantelides, in: The Physics of Si02 and Its Interfaces, Ed. S.T Pan telides (Pergamon, Oxford, 1978) p. 189.

5