Colloque C5, supplement au nolO, Tome 49, octobre 1988. GRAIN BOUNDARY ELECTRICAL ACTIVITY OF n-TYPE GERMANIUM. N. TABET and C. MONTY.
JOURNAL DE PHYSIQUE Colloque C5, supplement au nolO, Tome 49, octobre 1988
GRAIN BOUNDARY ELECTRICAL ACTIVITY OF n-TYPE GERMANIUM
N. TABET and C. MONTY
CNRS, Laboratoire de Physique des Materiaux, Bellevue, F-92195 Meudon Cedex, France
Rdsum6 : L'observation en mode EBIC du Microscope Electronique h Balayage a Ct6 utilis6e pour caractbriser llactivite Clectronique des joints de grains dans des Cchantillons de germanium (type n, dopes au phosphore ou A l'antimoine). Les joints de grains ont ete Cgalement caract6risiis d'un point de vue structural en utilisant les diagrammes de canalisation dlClectrons (ECP) et la topographie en rCflexion des rayons X. Les joints de faible dksorientation presentent un contrast EBIC qui croit quand la tension d1acc816ration des Qlectrons primaires augmente. Les r6sultats ont Cte analyses en utilisant les modhles existants ; ceux-ci ne permettent pas une description correcte des observations expCrimentales car ils ndgligent la contribution de la zone dCserte au contrast. Abstract : The electron beam induced current mode (EBIC) of the Scanning Electron Microscope has been used to characterize the electrical activity of grain boundaries in germanium samples (n-type, doped with P or Sb). The boundaries have also been characterized from a structural point of view using Electron Channeling Patterns and X-ray topography. Low angle boundaries show an EBIC contrast which increases when the primary electron beam accelerating voltage increases. The results have been analyzed using existing models ; they are not able to well describe the observed behaviour as they neglect the depletion zone contribution to the contrast. I
-
INTRODUCTION
Most semiconductors properties are related to defects : points defects, impurities, dislocations, grain boundaries (GBs). There is a possibility that GBs act as a distribution of recombination centres for electronic carriers and influence the electrical properties of devices. An important illustration is given by solar cells in which the photovoltaic efficiency is limited by the electrical activity of grain boundaries. An important question is to know the origin of the electrical activity of the defects and specially of grain boundaries. It has been shown [ I ] that defects introduce electronic levels in the band gap. These levels can be associated with segregated impurities or dangling bonds in the defect core, the details of such phenomena remain unknown. Several techniques can be used to obain information about the electronic levels introduced by defects : deep level transient spectroscopy (DLTS) is one of them, it provided information on sub-boundaries in Si [2,3] ; electron beam induced current technique (EBIC) is another one, which has been also used mainly on Si [4,51The interpretation of the EBIC contrast shown by grain boundaries C6,71 gives access to the diffusion length and to the recombination velocity of minority carriers on the grain boundary. Nevertheless the models established are not fully valid in all materials and need to be refined.
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1988583
JOURNAL DE PHYSIQUE
C5-648
Despite the number of previous investigations, the correlation between the electrical activity and the atomic and electronic structure of grain boundaries has not been clearly established. We need to investigate new materials, different from Si, varying the doping level and the nature of dopant ; looking at the influence of thermal treatments and trying to increase the quality of the models used to interpret the results. The correlation between several techniques, giving information about the atomic and the electronic structures, is also a way to progress. The aim of this paper is to present the results of investigations by scanning electron microscopy( SEM) using the EBIC mode, the secondary electron imaging and electron channeling patterns (ECP), to characterize the grain boundaries on n-type germanium polycrystals. The electrical activity of grain boundaries is characterized by ths EBIC contrast and correlated to their structural geometric parameters. This study is complementary of a first investigation of Ge by the EBIC technique which has never been used on this material before these studies C81. I1
-
EXPERIMENTAL TECHNIQUES
The samples studied have been prepared from n-type Ge polycrystals doped with Sb. For a first series (a-series) the resi ivity was close to 30 $2. cm (dopant concentration, ND, of the order of 1014 cmPf) and for a second one (b-series) it was of the order of 0.5 $2. cm
.
After mechanical polishing by diamond paste (particle sizes : 8 pm, 2 pm and 0.5 pm) they were chemically polished using ffCP4ft (5HF + 25 H NO + 15 CH.,COOH + 0.3 Br by volume) during 4 min at room temperature and cleanec? with distille -$ water. Schottky contacts have been obtained by evaporating under vacuum ( < 10 Torr) a gold film (10-30 nm thick) through a tantalum mask (diameter % 1.5 mm). The barrier height deduced from current voltage curves were of the order of 0.54 eV assuming a thermoionic mechanism [8]. Ohmic contacts were prepared by rubbing the back face of the samples with an aluminium rod wetted with gallium. 111
-
ELECTRICAL AND STRUCTURAL CHARACTERIZATION OF GRAIN BOUNDARIES
EBIC images have been obtained on both a and b-series of Ge samples and compared to secondary electron images of the same areas. The geometrical relationships between the lattices of grains separated by the studied grain boundary, were determined using ECP. Three main classes of boundaries can be distinguished corresponding to 1/ small misorientations (sub-boundaries) ; 2 / misorientations close to a coincidence ("special boundariestt); 31 large misorientations, far from a coincidence ("general boundaries"). Examples of EBIC images of specified boundaries are shown on figure 1, 2 and 3. Figure lb shows the ECP obtained, outside the Schottky junction, for a non electrically active boundary (large misorientation) and an active one (sub-boundary). Figure 2b shows the secondary electron image of the area studied by EBIC in figure 2a. The following general trends can be deduced from the observations
:
- Sub-boundaries (class 1) are the defects showing the highest activity but parts of a same boundary can be active, while other parts are not visible on the EBIC image (fig. la or 2a). This observation can be related to changes in the dislocation structure of the boundary. Indeed, dfAnterroches et al. C91 have observed several types of dislocations by electron microscopy in Ge sub-boundaries (tilt 3" around c0111). Nevertheless the main role is probably to be attributed to impurities whose segregation, or precipitation, is correlated to the core structure of the extended defects.
- Boundaries of class 2 such as a E 3 (figure 3) are generally non active but localized activity appears in some places, which can be associated with precipitates. The: crystallographic structure of the boundary, specially the GB dislocations, is a related parameter. Studies of metals have demonstrated the importance of extrinsic dislocations on the segregation phenomena [lo]. It has been shown that thermal treatment can modify the electrical activity of C 25 tilt boundaries in Si [11]. Calculations of the electronic structure of tilt boundaries in Si C121 ( C 3 ( 1 1 2 ) ) are in agreement with the idea that there are no dangling bonds in these boundaries. The correlation structure-segregation in coincidence boundaries seems,
to be due to extrinsic dislocations in semiconductors too.
- Many general boundaries pertaining to class 3 are not electrically active in samples doped to the higher level (b-series). As one would expect, these boundaries should contain more recombination centers than the others, the result is surprising. Impurities should easily segregate to these boundaries but they could as well increase the electrical activity or passivate the boundary, depending on their charge state i.e.their nature. Analytical electron microscopy is strongly needed to have a better understanding of the observed behaviour. IV
-
QUANTITATIVE EBIC ANALYSIS
The EBIC intensity, I , has been recorded as a function of position, x, accross active boundaries. The figure 4 shows a series of profiles h- I = I*(x) (IB is the EBIC current in the bulk, far from the grain boundary) measured on a sub-boundary of the b-series as a function of the primary electron beam accelerating voltage Eo; the primary beam intensity, 10, remaining constant (10 r~
.
Marekts model [6] relates the halfwidth of the 1*(x) profiles to the diffusion length, L , of minority carriers near the GB. Values of L have been deduced from the profiles of fig.4 and compared to values deduced from the EBIC current measured near the diode edge (Ioannou and Davidson [16]) resulting in L % 15 Urn; Marek's method gives lower values by a factor 'L 10
.
The Donolato model C131 can *be used to obtain information on the grain boundary electrical activity. From I (x)/I contrast profiles area, A , and variance, u , (IB is the EBIC current in the bul! far from the grain boundary), it is possible to deduce L/R and v .LID, where v is the recombination velocity of carriers to the grain boundary: R , is the maxigum depth of the generating volume of the primary electron beam (depending on E ) , D is the diffusion coefficient of carriers. Figure 5 shows values of A and c? measured on I (x) profiles shown in figure 4.
R has been estimated using the formula of Kanaya and Okayama C171 for Ge : R ~ ~ l - ~ ~ ( k e vL) .values deduced are close to 5 pm (nearly independent of E ) , again lower than those measured by the Ioannou and Davidson method. 46 lo8 pm2s-1 [18], vs is of the order of 5.10~ cms-l. Using a vzlue of D of Using an estimated density of minority carriers in excess : p % 1012 ~ m - ~ and , a of the order of 10-~cm,leads to a carrier source on grain boundary thickness, 6 ~ s - ~ on . the relaxation time T R of the the GB : a h = vs.p/6= 5 ~ ~ ~ ~ c m -Depending recombination centers on the grain boundary, their density, p R , can be estimated With -rR as small as 2 loT3s [2], p R should be of the order of OR TR 1022,,-3 (1015,,-2 with 6 = a particularly high value close to one trap per atomic site. The density of levels seen by deep level transient spectroscopy (DLTS) in pure tilt subboundaries in Ge [2] is of the order of l ~ l ~ c m - ~ . ( ~ m )= 1.72
It should be emphasized that the recombination centres density calculations based on the EBIC and cathodoluminescence contrasts lead usually to values higher than those determined by other techniques. In the present case, the carriers excess density p near the GB can't be estimated with a great accuracy. Moreover the models used to interpret EBIC contrast are not adapted to materials in which the main part of the EBIC is produced in the depletion zone of the Schottky dioge. This probably explains the non reliable values of L and V obtained from the I (x) profiles. B. Sieber et al. [15] have considered thg case sf dislocations in CdTe and have taken into account the effect of the reduction of the efficiency of the defect due to the drift of carriers by the electric field of the junction. A model to interpret G.B. contrast in such situations is in progress.
C5 -65 0
JOURNAL DE PHYSIQUE
Fig.la EBIC image on a Ge polycristalline sample (b-series). Strong differences appear in contrast depending on boundaries. A same boundary can have active parts and no active ones. The background is not uniform. This is probably du to variation of the dislocation density.
Fig.2a Other EBIC image. Comparison with fig.2b shows that some boundaries are electrically active and others are not
Fig.lb ECP corresponding to the unactive boundary crossing the junction seen on Fig-la. The misorientation is large (class 3 ) .
Fig.2b. Secondary electron image by SEM corresponding to the area seen on fig.2a. The boundaries seen are chemically etched.
Fig.3 EBIC -
image on a twin (boundary
T3).
Fig.4 EBIC intensity profile of a sub-grain as a function of electron beam accelerating voltage Eo, at constant 10 ( 10-~OA).
Fig.5 Area, A , and variance, -
0
the profiles seen on fig.4, as a function of Eo.
,
of
JOURNAL DE PHYSIQUE
C5-652 REFERENCES
11 23 33 43 53 61 73 81 93 [lo]
[ [ [ [ [ [ [ [ [
Taylor W.E., Ode11 N.H. and Fan H.Y., Phys. Rev. 88 (1952) 867. Broniatowski A., J. de Physique, C4, 9, 44 (1985)T39. Werner J., Jantsch W., Queisser H.J., Solid State Comm. 42 (1982) 415. Leamy H.J., J. Appl. Phys., 53, 6 (1982) R51. Hanoka J.I., Solar Cells 1 (1980) 123. Marek J . , J. Appl. Phys. 53 (1982) 1454. Donolato C., J. Appl. Phys. 54, 3 (1983) 1314. Tabet N., Monty C., Phil. Mag-B, 57, 6, (1988) 763 dfAnterroches C., Bourret A., Phil. Mag. 49, 6' (198b) 723. L. Priester, in "Les Joints de Grains dans les mat&riaux", Ed. de Physique
(1985). [ll] Battistella F., Dr. Ing. Thesis, Toulouse (1985). [12] Mauger A., Bourgoin J.C., Allan G., Lannoo M., Bourret A., Billard L., Phys. Rev. B, 35, 3 (1987) 1267. [131 Donolato C. in "Polycristalline Semiconductors : physical properties and [14] [15] [I61 [17] C181
applications". Ed. Harbekec, 1985. Klein S.A., Phys. Lett. 24A (1967) 513. Sieber B., Phil. Mag B 55, 5 (1987) 585. Ioannou D.E., Davidson S.M., J. Phys. D : Appl. Phys. 12 (1979) 1339. Kanaya K., Okayama S. J. Phys. D : Appl. Phys. 5 (1972) 43. Sze S.1.I. "Physics of Semiconductors Devices" 2nd Edition (1981) John Wiley, New York.
