An electron microscopy study of arsenic segregation in silicon

XX

ELECTRON MICROSCOPY STUDY OF ARSEXIC SEGREGATION IX SILICON* Drpartmrnt

of Zllaterials Enginssring. The Technion. Haifa 3200, Israel

Abstract-.4 T.E.M. investigation was carried out to study the segregation of arsenic in silicon and the resulting substructure. Two doping methods. arsenic implantation and capsule diffusion. were ussd. Xfter annealing at 7oO’C two modes of segregation u’:re observed. In the case of arsenic implantrd silicon. rod-shaped precipitates of .As-Si are formed along / 110: planes. In the case of arsenic diffused silicon the observed results suggest that segregation of arsenic is associat2d tirstly with faulting on \ 111: planes and secondly with the generation of perfect a 2 (110) extrinsic dislocation loops. It is postulated that the difference in segregation mechanism can be attributed to radiation damags centers in implanted silicon. R&sum&-On a etudie par microscopic electronique en transmission la sigrtgation de l’arsenic dans le silicium et la sous-structure qui en resulte. On a utilist deux techniques de dopage, l’implantation d’arsenic et la diffusion en capsule. On a observe deuv modzs d2 st,gr6gation apres un recuit B 700-C. Dans I2 cas de l’implantation ds l’arsenic dans I2 silicium. des p&ipltes d’ As-Si en forme de baguettes apparsissent suivant les plans : 110;. Dans I2 cas de la diffusion de I’arsenic dans I2 silicium. 12s Aultats ohserGs laissent penser que la &gfCgation de l‘arsenic est tout d’ahord associte 6 la formation de fautes situies dans les plans [lil). puns A celle de boucles de dislocations parfaites extrinskques (1.2 (110). On pourrait attribuer la difftrence de mecanisme de segregation aux dommages d’irradiation (dans 12 cas de I’implantation). Zusammenfassung-Mit der Durchstrahlungselektronenmikroskopie wurde die Segregation van Arsen in Silizium und die resultierende Substruktur untersucht. Zwei Dotierungsmethoden, .Arsenimplantation und Kapseldiffusion, wurden angewendet. Nach .4uslagern hei 7OO~Cwurden zwei Segregationsarten gefunden. Im arsenimplantierten Silizium hilden sich stabfijrmige Ausscheidungen As-Si entlang [ 110: -Ebenen. Im Fall des arsendiffundierten Siliziums legen die Beohachtungen nahe. dal3 die Segregation van Arsen verbunden ist erstens mit der Bildung van Stapelfehlern auf : Ill)-Ehenen und zweitens mit der Erzeugung vollstlndiger extrinsischer Versetzungsringe mit Burgersvektor (I Z (110). Es wird postuliert. dal3 der Unterschied in den Segregationsmechanismen den Strahlenschldigungszentren des implantierten Siliziums zugeschrieben werden kann.

INTRODUCTION The techniques of ion implantation and capsule difTusion are used to dope silicon with arsenic. In both techniques there is an excess of arsenic concentration which is not electrically active. Large decreases in conductivity have been observed for arsenic diffused silicon samples during annealing at temperatures lo\v?r than the diffusion temperature [l]. This conductivity loss was interpreted to result from the clustering of arsenic atoms in the form of arsenic atom pairs [I]. These clusters were assumed to be sufficiently displaced from the regular substitutional silicon lattice sites to act as the scattering centers observed by Chou er ~1. [2]. No evidence for the presence of arsenic precipitates has been found [3.4]. However. prismatic dislocation loops with perfect Burgers vectors (f (110)) and “J (ill} faulted loops \v?rz observed in arsenic diffused silicon specix This work was carried out at the Thomas J. N’atson R:scxch Center and the East Fishkill Facility of 1851. S:\i Yor’k. L-.S.-\.

mens [3.4]. The nature of these dislocation loops is not yet clear however. since both vacancy [3] and interstitial [4-J loops have been reported to be present in arsenic diffused silicon. The purpose of this work is to study, by means of transmission electron microscopy (T.E.SI.1. the nature of the substructure of arsenic-doped silicon, which is generated by the segregation of excess arsenic atoms.

EXPERIMENTAL

PROCEDURE

Spcciinen preparation [OOl] silicon wafers were doped with arsenic atoms by using the capsule diffusion technique. The treatment conditions were selected to obtain deep and shallow penetration (see Table 1). After doping. xafers were annealed at 7oO’C for different times. The sheet resistance was measured and the results are given in Table 2. The ion-implantation technique was used as well to dope [Ool] silicon wafers with arsenic. The condi-

YO9

KOXfE1f:

SlO

ARSENIC SEGREGATION

IS SlLICQs

Table 1. Capsule diifusion condi:ions

Diffusion temperature T(‘C1

Penetration treatment

Diffusion time (miq

Junction depth .y,(Clm)

;\rsenic surface concentratiorz Co (atom. cm31

90 135

1.17 0.17

1.8 x 10:’ 1.2 x 10”

I loo

Deep Shallow

IWO

Table 1. The sheet resistance of silicon wafers as function of annealing time at 700-C Sheet resistance (0’2) Annealing time (hr)

t= 0

r=j

r=5

Deep penetration Shallow penetration

5.6 18.1

5.8 l&S

6.7 19.6

and 4 = 1 x 10L” ion/cm?, tions, E = 80KV resulted in the formation of an amorphous surface structure which on annealing for 15 min at 700°C. recrystallized. These ion implantation and annealing conditions were found to be suitable to show arsenic precipitation by T.E.M. without disturbing elects of too high a dislocation density. Specimens for T.E.M. observation were prepared by using the jet polishing technique with a 9: 1 etching solution of I-NO3 and (40Y;)HF. T.E.M. methods

In addition to the conventional methods of analyzing defects and precipitates by T.E.M. (e.g. selected

r=

10

7.1 10.4

t = 20 7.5 -‘11._

area diffraction and dark field) the methods developed by Maher and Eyre [5] and Silcock and Tunstall[6] were used to analyze perfect dislocation and faulted dislocation loops respectively. The convention used by \laher and Eyre to determine the sign of the normal sector to the loop plane has been adopted. In thts convention the loop plane normal. n. is atway pointed upwards towards the electron beam and the FS; RH perfect crystal convention [7] is used to determine the Burgers vector of the dislocation loop as it is viewed from above the specimen. According to the first method [5]$ a dislocation loop will be imaged with outside contrast (e.g. when the minimum of the intensity profile is iocated outside

Fig. I. Arsenic implanted silicon annealed at 7oo’C for 15 min. (al Dark field image of refiectioos I in (c). (b) Dark field image of rctlrction II in (c). (c) Selected arta diffraction pattern. !dt Dark field image of reflection i.220) in [cl.

i;O?clESf:

ARSEWC SEGREG.ATIOY

the dis!ocarion core) when g*b+Sg > 0 and with inside contrast when g.b.Sg < 0 (when g is the di%action Lsctor. b is the Burgers vector and .S+l is the dciiation parameter). By reversing the sign of g {or Sy) one can observe the change in the loop size and it is thzn possible to evaluate the sense of the Burgers vector. i b or -b. According to ths convention used for thz sign of the loop plane normal (II), if n*b > 0. the loop is of interstitial type and if n*b cc 0. the loop is of vacancy type [j]. In this method the angular relationship between b and n is not needed when analyzing dislocation loops with Burgers Lsctor in the safe orientation [j] (i.e. when b is sutticientl>- close to the opposite direction of the electron beam. z and is within an angular domain bounded by the oricntation n,z = 0 and nib = 0. where n, and nj are unit vectors perpendicular to the loop planes. e.g. se2 Fip. 3,. According to the second method [6] a faulted dislocation loop (with h = 4 < 111 ) _ i.e. Frank type) Lvill be out oi contrast when g. b = 2.3. when imaged nit!1 a beam largely deviated from the Bragg reflectinp position (e.g. .l’g*Sg r 1. where ,‘g is the extinction distance). and using the same convention described in the first method mentioned above. By reversinp the sign of a diffraction vector g suitable to this method (e.g. g =X0) one can observe the disappearance of faulted dislocation loops and it is then possible to evaluate the sense of b and the nature of the faulted IOOP.

RESULTS

AND

DISCUSSIOS

Small rod shaped precipitates were observed in silicon specimens that were implanted with arsenic (E = 80 kV, 4 = 1OL6ios’cm’) and annealed at 7OO’C for ljmin. The rod shaped precipitates can be observed in Figs. l(a) and (b) which are dark field images (exposure time 60 set) of the precipitate refiections marked at I and II in Fig. l(c) respectively. No ring patterns typical of diffused scattering from an amorphous layer. which were observed in earlier annealing stages, could be observed at this stage. The dislocation loops generated due to the radiation damage caused by the implantation can be obserwd in Fig. ltd). which is a dark field micrograph that was imaged with the 120 reflection. 4nalysis of the extra reflections in the diffraction pattern (see Fig. lc) shows that the reflections I and II represent perpendicular sets of parallel rod shaped precipitates (see Figs. la and b). Since the images seen in the micrographs are projections of the rod shape precipitates on the plane of the foil [S] and a strsakin_e of thz precipitate reflections occurs in the ,, 110) directions of the matrix. it is assumed that the precipitates lie along : 110; planes. Analysis in terms of the J-spacing shows that reflections I and II arise from (3031 planes of arsenic-silicon (A-Sil monoclinic structuri’ with (1 = 3.11.A. All other extra reflections

Pi SILICOS

.i;:

Bb assuming roi: si1a~e precipitates ivlth :?-~;L\:JT& dimensions approuimat4y 15 1\ diameter ~2 -5 .i length. and a dzn~it> of the or&r of 1 . 1fJ’of arsenic as p”“?“ate rod Lm3, the concexration was calcuiat?d to be i x 10” atom :m’. Tk total implanted arsenic conccnrration \YX s~aiust:ii :o bs approsimar:l) 1 .< 10” atom cm”. Tlxrckx-2. ihe arsenic as precipitntcd ai this annealing stag: is approsimatel> I”, of the total arsenic concsnrxion. Thmz results indicats that :I posiihl? m&r~ism for arsenic to be el2ctricall~ inactikc: is b> a prcci?;:.tti~9 process.

study of the arsenic diKussd silkor. qrcihas shown no evidence for the presence c< prr-

.\ T.E.\LI. mans

K051EM: ARSEXIC SEGREGATION IN SfLICOS

$12

of the dislocation loops. First the assumption was made that the dislocation loops which were not imaged with fringes typical of faulted loops were perfect dislocations with h = f (110;,. According to the convention selected for this purpose the dislocation loops with Burgers vectors I(1;2[011]. *L2,2[oi I]. Ita/2[101] and iuT2[iOl] are in the safe orientations [5] and all others are in unsafe orientations (see Fig. 3). It is clear from Table 3 that dislocation loops having a Burgers vector in the safe orientation should be visible by using a 220 reflection but should change from visible contrast to invisible (or vice versa) bj using reflections 040 and 300. This is not the case for dislocation loops in the unsafe orientation. In this case, for the 220 reflection, 50”” of them are invisible. Analysis of Fig. 4 in terms of contrast shows that dislocation loops marked as A. B and C are in the safe orientation since they are invisible with the 040 ELECTRON DIFFRACTION REFLECTIONS reflection (see Fig. Ja) and visible with the TOOreflec@- USED FOR PNALISB tion (see Fig. 4b). In addition they are all visible in l - SAFE ORIENTAIONS f (220) reflections (Figs. 4d and e). According to UNSAFE ORIENTATIONS oTable 3 these dislocation loops may have either one or Fig. 3. A stereographic projection which shows the con- of the following Burgers vectors: _~:2[101] ventions used to analyze the nature of the dislocation +a/2[iOl]. However, since dislocation loops A and loops. ni and ni are unit vectors perpendicular to the dislo- B are visible with reflection i3l (which was imaged cation loop plane. z is the unit vector in the opposite direcby tilting the specimen clockwise along 220 axis and tion of the electron beam and b,are Burgers vectors. following the direction of the Kikuchi line map. see Fig. 3), and dislocation loop C is invisible with this cipitates. However, generation of dislocations in reflection. loops A and B have Burgers vector b = kni2[iOf] and loop C has Burger vector specimens deeply penetrated by arsenic (see Tables b = +@[lOl] (se2 Table 3). 1 and 2) and annealed at 700% was observed. Figures Since all micrographs were taken with positive Sg. 2(b) and (c) show the evolution of the dislocation structure in these specimens after annealing at 7CO’C outside contrast [5] (see experimental procedure) should arise when g. b > 0. Dislocation loops A and for 30 min and 20 hr respectively.* No such dislocation loops could be observed in the arsenic diffused B exhibit outside contrast with reflection 220 while dislocation loop C exhibits the same contrast with specimens that had shallow penetration. This is probably due to the fact that the dislocations were reflection 220 (compare Fig. 4d with e). Therefore, annealed out to the free surface from the shallow dislocation loops A and B have Burgers vector: b = a/2[iOl] while dislocation loop C has Burgers doped region (Xj = 0.47 llrn, see Table 1). The T.E.M. methods mentioned above (see experi- vector b = a,K![lOl]. In both cases nab > 0 according to the convention used here (see Fig. 3). This means mental procedure) were applied to analyze the nature that all these loops are of interstitial type. Some of the dislocation loops, (see .A in Figs. 5a *Stereo observations made in this study have shown that these dislocation loops are distributed to a depth and b) exhibit fringes typical to faulted loops. Applying the second T.E.M. method [6] mentioned above. equivalent to the iunction depth. ioo

Table 3. Characterization of g*b products for perfect loops g-b b \

g

k 1.2[011] 21 q-oil] * 1:2[101] k i2[ioi]

fl"[l_lo] ~1.2[110] * 1 ‘I[TlO] -i: i.:z[iio]

030

zoo

220

i3i

22

0

Fl

&Z

fZ

0

e1

il

0

T;z

+1

0

0 *:z

*z TZ

+1 _+7-

FZ +z

52 *2

v2 0

&l 11 I2

0

S?

F_1

e?

Comments

Safe orientation

il Unsafe orientation

I;OlIEM:

ARSESlC

SEGREGATION

IS

SILICO>

<‘T &_

to

ACHYfau!ted loops in diamond cubic crystals. i>C‘ .i ths reeections {___ ?‘?I and i?) should be used. Xnai!sis of Figs. 5(a) and (b) show that while the dislncstion loop .A is visible with a 322 reflection.

Fig. 5. Bright field images taken with S > 0 oi i~~l!:d dlslocation loops generated in arsenic ditfured silicon after heat treatment at 700-C (al g = 223. tb) g = m ___. (cl g = 210.

Fig. 1. Bright field images taken with S > 0 of dislocation in arsenic diffused silicon after heat treatment 2: -UYc la\ g = 040. (b) g = %O. (c) g = (d! g = 33. (e, g = 220.

loops gwated

i31,

it is invisible with a ? reflection. Since the g*b products, when b = (I 3 ( 111j. yield values of =I! 3 [6] it arises from the analysis that this is a Frank type dislocation loop. Analysis of Fig. 5(c) show that loop A is also invisible with the 220 reflection. The onlv Burgers sector that obeys the invisibility criterion both with g. b equal to 2. 3 and 0 is b = 1 3[nl] (see Table 11. Since the micrographs in Fig. 5 were taken uith reflections belonging to the EZ! zone asis then II. h is positive if the convention mentioned above is being used (see Fig. 3). This means that the faulted dislocation loop A in Fig. 5 is of the interstitial type. Dislocation loop B (Fig. 5) can be characterized identically- to loop A even though it is not as mvisibls with r:t%ction 227 (see Fig. 5b) as loop ,A. In summary. annealing of arsenic diflused silicon at temperatures louver than the doping temperature causes the generation of both faulted and perfect interstitial dislocation loops without farming arsenic precipitates. These interstitial loops can accommodars the excess arsenic that segregates at lowr rznperatures. The presence of faulted dislocation loops sug-

Jll

KOMEM:

ARSENIC SEGREG4TIOX

Table 4. Characrertiation

g

IS SILICOX

of E. b preducrs for faulrzd loops

b \

117 __-

q-_-

23 -0

*1/3[111] 5 1;3[Iil] + 1/3[Ql] 1!‘3[111] 1;3[1 ii]

&9 --+‘3 12 3 -_,,2*3

t2 12 3 +13 2 ‘3 - ’- .i.

0 k4.3 TJ’3 0

gests that a coherent segregation of the excess arsenic

atoms takes place first on [ 111; planes, which result in the formation of Frank dislocation loops. During the annealing, these Frank loops are unfaulted by the nucleation of two Shockley partials with b = n/6 thus forming the perfect dislocation (112j loops [4,9]. By assuming that all the dislocations have a loop shape with measured average diameter of approximately 8SO.A and a density of the order 5 x 10” loops per cm3, the concentration of arsenic in the extra plane of the dislocation loops is of the order of 5 x 10” atom/cm3. The value is 3”b of the total arsenic concentration (Co 2 1.8 x 10” atom/cm3, see Table 1). and can represent the percent of electrically inactive arsenic at this state of annealing. CONCLUSIONS T.E.M. observations suggest that following arsenic implantation as one technique, or capsule diffusion of arsenic into silicon as another, two different mechanisms account for the segregation of arsenic after annealing at 700’C. In the case of arsenic implantation, the arsenic segregates as rod shaped As-Si precipitates along [ 110: planes. In the case of capsule diffusion the arsenic segregates on (11l] planes. As a result. faulted interstitial loops with b = u/3 (111) are formed. These unfault by the generation of perfect interstitial

0

loops with b = u:2 (1 lo/. The ditierence in the segrepation mechanisms could arise due to the fast that in the arsenic implanted silicon. precipitation is promoted by the presence of radiation damage centers. These will not be present in the case of arsenic diffused silicon.

dislocation

-i~l;no~~/erlgenl~nrs-The author wishes to express

his appreciation to Drs. S. Mader, H. Rupprecht and L. Glowinsky for verv helpful comments and discussions. Likewise. the author rvishes to thank Dr. T. B. Light for his helpful assistance. Special appreciation is expressed to Drs. R. Rosenberg and A. Reisman for their continuous encouragement and support. In addition. the author wishes to thank IBM for their permission to publish the work.

REFERENCES I. R. 0. Schwenker. E. S. Pan and R. F. Levsr. J. appl. Phj~ 12. 3195 (19711. 3. S. Chou. L. .A. Davidson and J. F. Gibbons. Apyl. Ph,w. Lett. 17. 13 (19701. 3. S. Dash and IV. L. Joshi, IBM JI Rw. Dec. 14. 453 (19iO). 4. G. Das. Pror. 3rd Int. Con& on H.U..Lf., p. 277. Oxford (1972). 5. D. M. Maher and 8. L. Eyre. Phil. Mug. 23. 409 (1971). 6. J. M. Silcock and W. S. Tunstall, PM. Mug. 10. 361 (1964). 7. A. Howie and IV. J. Whelan. Proc. R. Sot. (A) 263. 217 (1962,. 8. Y. Komem and S. Rezek, Mecall. Trans. (A) 6A. 549 ( 1975). 9. R. S. Barnes, Discuss. Fwad. Sot. 31, 38 (1961).