Annealing encapsulants for InP II: Photoluminescence studies

Thin Sohd Fdm~. 94 (1982) 161-170

161

ELECTRONICS AND OPTICS

ANNEALING ENCAPSULANTS FOR InP II" PHOTOLUMINESCENCE STUDIES J. D. OBERSTAR AND B. G. STREETMAN

Department oJ Electrwal Engmeermg and Coordmated Sctence Laboratory Unwerstty of llhnms at Urbana-Champalgn, Urbana, IL 61801 ( U S A ) (Recewed March 6, 1982, accepted March 29, 1982)

Using low temperature photoluminescence (PL) we examined samples annealed in controlled atmospheres (vacuum, phosphorus vapor or indium vapor) and also samples annealed with SiO2, SiaN 4 or phosphosihcate glass (PSG) encapsulants. PL spectra from samples annealed in the controlled environments indicate that a spectral line at 1.393 eV results from phosphorus-vacancy-related defect luminescence. PSG and Si3N4 caps appear to be comparable in their ability to suppress the formation of substantial 1.393 eV luminescence, but in the spectra of SiO2-capped and annealed samples this feature is more pronounced. High temperature anneals (T ,~ 750 °C) with SiaN 4 result in the emergence of a new peak at 1.378 eV and integrated band intensities greater than those observed in the virgin material. From results discussed in this article and those of the preceding article, it appears that, of the three encapsulants, the PSG cap best preserves the characteristics of the encapsulated InP following furnace anneals.

1

INTRODUCTION

In this article we discuss the results of low temperature photoluminescence (PL) measurements of InP samples annealed in controlled atmospheres and of samples annealed with the three encapsulants analyzed in the preceding article. Low temperature PL is a versatile and non-destructive characterization technique. Because of its extreme sensitivity to lattice perfection it has often been used in the past to study native and processing-related defects in silicon and GaAs. For example, a previous study ofGaAs encapsulants by our group has demonstrated the usefulness of PL in characterizing the effectiveness of encapsulants 1. The sample preparation methods used in this work have been extensively discussed in the preceding article. Except where noted, it may be assumed that identical processing techniques were employed in this work. 2. EXPERIMENTAL DETAILS

The semi-Insulating (SI) iron-doped InP(100) used in this work was grown by the liquid encapsulation Czochralski method at the Naval Research Laboratory2. 0040-6090/82/0000-0000/$02 75

© ElsevierSequota/Printedin The Netherlands

162

J D OBERSTAR, B G. STREETMAN

This material had a resistivity of approximately l 0 7 f2cm and an 56Fe atomic concentration of about 8 x 1016 cm 3 (ref. 3). PL spectra were obtained from samples mounted in a strain-free manner in a gas exchange hquld hehum cryostat. The temperature during photoexcltatlon was maintained at 5 K as monitored by a germanium resistance thermometer embedded In the sample holder. Excitation was provided by the 5145/~ hne of an argon ion laser with an Incident power density of about 103W cm 2 and a focused beam diameter of approximately 40gm. At this wavelength the l/e point for the penetranon depth of the excitation beam is 939/k 4. The front surface luminescence was collected and focused onto the entrance slit of a 0.5 m Spex model 1302 doublegrating spectrometer. The opncal output was detected by a cooled (77K) photomultlpher with S-1 response and was analyzed with conventional lock-in amplifier techniques Throughout the entire series of spectral examinations a single control sample was used to calibrate for possible differences in system response between runs All samples were individually examined prior to processing to determine their optical characteristics Because of system response, only the (nearband-edge) spectral range from 1.44 to 1.30 eV (8606 9535 ,~) was examined The spectral resolution used in this work was 6.5 ~ (1 1 meV) The emission spectra presented are uncorrected for system response Figure 1 shows the PL spectra from virgin samples of the SI iron-doped InP(100) used in these studies. This material exhibits three luminescence bands at 1.414, 1.384 and 1 341 eV. These are attributed respectively to band edge luminescence (BE band), band-to-acceptor and donor-to-acceptor luminescence (BA-DA band) and luminescence from the 1-LO phonon rephca of the B A - D A band A fuller discussion of these features and their origins appears elsewhere 5 Energy (eV) 152

136 r

140 i

144

1414

g >.

1584 ]

x5

xl

g "s

i-kO BA/DA BE i I i L i I i 9400 9200 9000 88OO 8600 Wovelenglh (~) , ,, Fig 1 P L bpectra of'~lrgln I r o n - d o p e d SI lnP(100) T h e n u m e r i c a l factors are ampllficaUons r e l a n v e to that ubed for the b a n d edge l u m i n e s c e n c e

3

CONTROLLED ATMOSPHERE ANNEALS

In order to interpret the PL spectra of InP annealed with our caps, ~t was first necessary to examine and to characterize the spectra from bare InP annealed in

ANNEALING ENCAPSULANTS FOR I n P : II

163

controlled environments (vacuum, indmm source and phosphorus source). Such heat treatments should favor the creation of annealing defects which might subsequently lead to defect-related luminescence peaks in the PL spectra. Annealing ampoules were fabricated from silica tubing. A small dimple was formed in each ampoule to prevent contact between the samples and the source materials durmg the anneals. The silica tubes were cleaned in standard organic solvents and placed in aqua regia for 24 h. The ampoules were then rinsed in deionized water and electronic grade isopropyl alcohol. Immediately prior to insertion of the samples and source materials, the tubes were outgassed by heating them with an oxyhydrogen flame while they were evacuated with a diffusion pump. The source materials were 99.999% pure indium and 99.9999% pure red phosphorus purchased from Alfa Ventron. After insertion of the samples and source material the ampoules were evacuated to 10- 5 Torr and sealed off to a volume of about 5 cm 3. The amount of red phosphorus used in these anneals was calculated to create a phosphorus vapor pressure of 0.5 atm at 750 °C, which is sufficient, according to published vapor pressure curves of phosphorus over InP 6"7, to ensure an overpressure of phosphorus. For indium source anneals, excess amounts of indium were inserted in the ampoules to yield saturation overpressure during the anneals. After annealing, the ampoules were quenched by inserting the tube end furthest from the sample in cold water. Two-point probe electrical tests of sample surfaces following anneals at 750 °C m any of the three environments revealed that in all cases the originally SI InP had formed a conductive surface layer. Figure 2 exhibits the BA-DA bands from samples annealed in the three environments for the temperature range 450-750 °C. Within the limits of resolution used in this work, no shifts were observed in the peak positions of the BE bands. Changes were noted in the relative intensities of the BE bands but these changes generally varied in accordance with the decreases or increases shown by the BA-DA bands. Because of this and for simplicity, only emission in the region of the BA-DA band is shown m Fig. 2. Three significant features of these spectra are as follows: (1) variations in the BA-DA band intensities with annealing temperature; (2) a new emission peak at 1.378 eV which becomes the dominant emission of the BA-DA band for T i> 650 °C; (3) the appearance of a new small emission peak at 1.393 eV in 450 and 550 °C vacuum and indium source anneals. Comparing the amplification factors of the various BA-DA bands from samples annealed between 450 and 650 °C with that of the BA-DA band from virgin material, it is clear that the annealed samples suffer a loss in luminosity. Presumably these intensity losses result from competition between the usual radiative processes operative in virgin material and new non-radiative processes involving defects created by the anneals. This luminosity loss continues and worsens for samples annealed in indium source ampoules at least up to anneal temperatures of 750 °C (Fig. 2(h)). The general trend is reversed, however, in 750 °C vacuum and phosphorus source anneals (Figs. 2(d) and 2(1)). The BA-DA band intensity in these samples is now greater than that observed in the virgin spectra of Fig. 1. As we have noted elsewhere 5 the increase in BA-DA band integrated intensities above virgin levels is correlated to the increasing dominance of the 1.378 eV transition. This is certainly the case when the featureless BA-DA bands of Figs. 2(d) and 2(1) are compared with the structured BA-DA emission band of Fig. 2(h). We shall discuss possible causes

J D. OBERSTAR, B G. STREETMAN

164

I

1 36 i

i

Energy(eV) 136 1 4 0 (e) In-Source

1 40 i

I

[o) Vocuum 450°C

i

J

i

i

(,)

450°C

1384~ - 1 3 9 3

1384

1 ~ I

i

140 i

P-Source 450°C

1 393 1 393

1384~i I

x20 (b) 550°C

~x20

x20

(f) 550°C

(1) 550 °C 1 393 1

1 384 - / / 1 393

1 393 1 384----

1 3 8 4 --

x40~ (c) 650°C

x4C ,d (k) 650%

(g) 650°C 1 3~4

i 384

1393

,

1

x40

1

384 l

i 39~

378

i 3/×40

i~

×40

×4C

/

(d) 750°C .

(h) 750°C ] • ~3

"W8

I 584

i , 9200

O) 750°C 1 378

,

, 8800

,

, J 920C

~

/

/

,

,

, 8800

i , q200

,

,

i 880@

Fig. 2 PL spectra of the BA-DA band after anneahng SI InP for 60 mm at the mdlcated temperatures in following sealed ampoule environments (a)-(d) vacuum, (e)-(h) lndmm vapor, (1)-(1) phosphorus vapor The numerical factors are amphficatlons relative to the band edge luminescence of Fig l the

for these increases in intensity when the spectra from capped samples are shown in the following secUon. It is perhaps surprising that the vacuum and phosphorus source anneals behave similarly at the high anneal temperature of 750"C. This result is understandable,

ANNEALING ENCAPSULANTS FOR

InP: II

165

however, in view of the physical surface deterioration of the samples. As expected, the samples annealed in the phosphorus source ampoules showed no visible deterioration and retained their mirror-like finish. Samples annealed in the radium source ampoules had the most thermally damaged surfaces of the three sets. We believe this is attributable to the fact that the indium source acts as a large sink for the phosphorus which evaporates from the sample surfaces. Surprisingly, the samples annealed in vacuum exhibited much better surfaces than did the corresponding indium-source-annealed samples. Without excess indium present in the ampoule to absorb the evaporated phosphorus, it appears that the sample surface is eventually capable of achieving phosphorus vapor phase equilibrium within the sealed environment, preventing further surface deterioration. In the 650 °C anneals of Fig. 2, a weak shoulder appears on the BA-DA band at 1.378 eV which at 750 °C becomes the new dominant emission peak of the band. This new spectral peak at 1.378 eV may originate from several sources, including silicon contamination from the silica ampoule walls (silicon is a known contaminant of liquid phase epitaxial InP grown between 550 and 715°C) s, annealing-induced radiative defects or near-surface gettering of impurities from the bulk. At present we cannot identify the specific origin of this spectral feature at 1.378 eV. The new feature of particular interest to us here is the small emission peak at 1.393 eV, which is most pronounced in the 450 and 550°C vacuum and indium source anneals. In comparable phosphorus source anneals this 1.393 eV feature appears only as a weak shoulder on the BA-DA band. Only a few of the many published studies of InP photoluminescence have mentioned emission peaks in this energy region 9-11. To our knowledge Williams e t al. 9 were the first to report a feature at 1.39 eV. Although they did not speculate on the origin of this emission, they noted that their 1.39 eV line increased in intensity when the samples were annealed for 5 min in H2 at 500 °C. They also found that the intensity of the 1.39 eV line increased as the epitaxial layer from vapor phase epitaxial InP was etched off and became strongest near the layer-substrate interface (where thermal damage to the substrate surface occurs prior to growth of the epitaxial layer). More recently, Yamazoe e t al. ~° have reported observing a small emission line at 1.39 eV in 1 h anneals of InP. On the basis of these reports and from the near suppression of the 1.393 eV line in 450 and 550 °C phosphorus source anneals, we believe that the 1.393 eV emission line results from phosphorus-related defect luminescence. Some loss of phosphorus can be expected in the phosphorus source anneals as the atmosphere is equilibrated. To minimize such transients of the phosphorus vapor pressure in the ampoule, others have reported converting red phosphorus to white phosphorus by heating the ampoule prior to annealing12. The annealing behavior of the 1.393 eV line is of interest, since for T > 550 °C the intensity of this emission line decreases dramatically, This reduction could result from radiative recombination competition with the newly emerging line at 1.378 eV. It is also possible, however, that the decreased intensity may result from a reduction in the concentration of the phosphorus-related defects responsible for the 1.393 eV emission at anneal temperatures greater than 550 °C. This would imply that the associated defect is a vacancy complex rather than isolated phosphorus vacancies, since the latter's concentration would be expected to increase with temperature instead of decreasing. Yamazoe e t al. 1° have reported similar decreases in the PL

166

J 1) OBERSTAR, B (, STREETMAN

intensities of other InP anneahng-related defects for T > 500 C In addition, their deep level transient spectroscopy measurements of four annealing-related deep level traps indicate that all four traps achieve their maximum concentrations at 500 C anneal temperatures 4

ENCAPSULAN F ANNEALS

The B A - D A bands from samples encapsulated and annealed with our three encapsulants are shown in Fig. 3 All encapsulant layers were removed from the annealed samples with hydrofluoric acid prior to PL measurements. Two-point probe electrical tests of sample surfaces following 750 C anneals with phosphoslhcate glass (PSG) or SI3N 4 caps Indicated that conductive surface layers had formed during the anneal Similar effects following anneals with PSG and SI3N,~ caps have also been reported by others 13 14 On the basis of the appearance of the 1.393 eV hne in Figs. 3(a) 3(d) it is clear that phosphorus-vacancy-related defects are created m all our SiO2-capped samples. It may also be observed that at temperatures between 600 and 6 5 0 C the 1 378 eV emission is of comparable intensity with the 1.384 line present in virgin material (The spectrum of Fig. 3(d) was obtained by analyzing regions away from the cracks which normally develop in our S102 caps at this temperature ) Although not evident in Figs. 3(a)-3(d), for samples annealed at T < 650 "C. we often observe the BA DA band peak at 1.381 eV instead of at the usual location of 1.384 eV. Local variations in the single crystal or in the cap may account for the inconsistent position of the B A - D A band peak. We believe that this 1.381 eV peak is related to the 1.384 eV line and that the shift in energy results from the overlap of the 1 384 eV peak with the emerging 1.378 eV peak. By examining the spectra of our PSG-capped and annealed samples shown in Figs. 3(1)-3(1} it appears, because of the presence of shoulders at 1.393 eV. that phosphorus vacancy defects are also created to some extent in these samples. This could be due to pinholes or to insufficient phosphorus content in our PSG to suppress phosphorus out-diffusion totally. Further work with various types of PSG films will be required to clarify this Issue Comparing Fig. 3(1) with Figs. 3(d) and 3(hi it can be concluded that these PSG caps are much more effective than SIO2 o r S I 3 N 4 in preventing the formation of the 1 378 eV emission feature Judging by the intensity of the 1.393 eV emission, o u r SI3N 4 caps seem comparable with the PSG caps In their ability to prevent substantial phosphorus out-diffusion (Figs 3(e)-3(h)} It should be noted that for anneal temperatures T >~ 650 C the 1 378eV emission comes to dominate the BA DA band of our Si3N.wcapped and annealed samples Furthermore, the luminosity of the BA-DA band in the 750 C anneal (Fig. 3(h)) is now significantly greater than that found m virgin samples. Also, as previously discussed with the S102 spectra, we sometimes find the BA DA band peak in Sl3N.,-capped samples annealed at T < 650 C at 1.381 eV rather than at 1 384eV It was mentioned m the previous section that the emission at 1 378 eV may originate from silicon in-diffusion, annealing-induced radiative defects or nearsurface getterlng of impurities from the bulk A comparable downward energy shift of 6 meV by the BA-DA band peak has also been recently reported by Shanabrook

ANNEALINGENCAPSULANTSFOR InP: II 136

140

(o) S,02 Cap

450"C

1595~

1584~]

(b) 550°C

Energy(eV) 140 1~ 140 (e) Si3N4Cop (i) PSGCap 4500C 450"C 1595~ I 1584~J 156

(f) 550°c

1595 i 378~~[~

167

%) 550°c

1395 1378--~)

1384~ ~

1393

(g) 650°C 15781 395

(k) 650°C

>~

(c) 600°C 1393 i384~ II 1378

i~x20 ~

,10

(h) 1378 750°C

(d) 650°C

(I) 750"C i3931 1384

]_5841595{ 1578 I I

J 92

O~l

x20 i

9000

i

J

8800

x20 9200 L9 0 0, 0 I 8 8 0 0 Wovelength (~,)

I

I

I

92010 9000 8800

Fig. 3 PL spectraof the BA-DAband after anneahngSI InP for 60 mln at the mdlcated temperatures with the followingencapsulants:(a)-(d) $102; (e)-(h) SiaN4;(1)-(1)PSG The numencal factors are amphficatlons relative to the band edge luminescence of Fig. 1. e t al. x 5

for Si3N4-capped and annealed SI InP(Fe). From temperature-dependent PL measurements they attribute the new emission to donor-acceptor pair recombination, with the implication that the donor concentration has increased. They conclude that such donors could be provided by silicon in-diffusion from the

168

J 1) OBERSTAR, B G STREETMAN

SI3N~ cap or from accumulation and redistribution of donors from the bulk to the surface. From the appearance of the spectra in Figs. 3(h) and 3(1) and from the secondarTy ion mass spectrometry [SIMS) 28S1 profiles shown in Figs. 4(b) and 4(c) of the preceding article ~t certainly appears likely that silicon could play a role in the 1 378 eV peak. It is not clear at this point, however, why the 1.378 eV peak does not similarly dominate the BA DA band of SlOe-capped and annealed samples when the SIMS 2ssi profiles from these samples showed the deepest ln-&ffuslon.

[al "e

Y

(bl

~,OC'

600 L~nrl~c~l X~mDer l~ur~

~0'2 ,~C,

qr,~

F'tg 4 Inlegraled lnlenslDe~ of (a) lhe BE band and (b) the B A - D A band lrom SI lnP(Fe)(100) unprocessed (~ - I or capped with either SHN., { ) or PSG t I and annealed for 60 mln at the indicated temper,ttures

In our high temperature ampoule anneals [Figs. 2(d) and 20)) and again in our 750 ' C S13N 4 anneal (Fig. 3(h)) we observed B A - D A band and BE band intensities greater than those found in the virgin material Such increases in band edge luminosity have been reported in the past for annealed SI GaAs(Cr)16. Recent reports have also described increases in the band edge emission of InP samples capped with S 1 3 N 4 and annealed 1-~ 17. We have already noted here and elsewhere 5 that these increases in luminosity are correlated to the emergence of the 1.378 eV feature. Such increases in luminosity may result from the annealing out of native defects in bulk InP, from surface accumulation of impurities with greater radiative efficiency than those normally present or from a reduction in the surface space charge layer due to the presence of the Impurity and/or defect Involved in the 1 378 eV emission. Work is under way to ascertain the origin of the increased levels of luminosity as well as that of the related 1.378 eV emission Figures 4(a) and 4(b) plot the integrated intensities of the BE and B A - D A bands for S 1 3 N 4- and PSG-capped and annealed samples. Before pointing out the significance of these curves we must comment on a feature not discussed in this article which affects our SI3N 4 data. We have observed that the spectra from samples annealed with Si3N4 at T>~ 650~C (and also from some high temperature

ANNEALING ENCAPSULANTS FOR I n P : II

169

phosphorus-source-annealed samples) contain a discernible tail of luminescence on the low energy side of the BA-DA band (Figs. 2(1), 3(g) and 3(h)). This broad tail extends from approximately the low energy side of the BA-DA band to the 1-LO phonon band and has been discussed in another article 5. The effect of this luminescence tail has not been included in the measurements of the Si3N 4 BA-DA integrated band intensities plotted m Fig. 4. Considering the Si3N 4 spectra of Fig. 3 together with the results of Fig. 4 it becomes clear that the BE and BA-DA intensities m the Si3N4-capped and annealed samples exceed virgin levels when the 1.378 eV emission dominates the BA-DA band. More importantly, it can be concluded from Fig. 4 that our PSG caps maintain the PL properties of the underlying InP samples with much higher integrity than do our Si3N 4 caps. 5

SUMMARIZING REMARKS

The results and conclusions of this study can be summarized as follows. (1) PL spectra from samples annealed in controlled environments (vacuum, indium source and phosphorus source) indicate that a spectral line at 1.393 eV results from phosphorus-vacancy-related defect luminescence. (2) The presence of the 1.393 eV line suggests that, at least up to anneal temperatures of 650 °C, our SiO2 caps allow phosphorus loss from the sample surface. A weak feature at 1.393 eV is also evident in the spectra of PSG-capped and annealed samples, suggesting some phosphorus loss. This may occur because of pinholes or because of inadequate phosphorus in these PSG films. Our Si3N 4 caps seem comparable with the PSG layers in their ability to prevent substantial phosphorus out-diffusion. (3) High temperature anneals (T ~ 750 °C) with S i 3 N 4 result in the emergence of a new peak at 1.378 eV which dominates the BA-DA emission. This emission at 1.378eV may originate from silicon m-diffusion, annealing-induced radiative defects or near-surface gettering of impurities from the bulk. Associated with this are BA-DA band and BE band integrated intensities greater than those found in virgin material. (4) In comparison with SiO2 or Si3N 4, PL measurements indicate that PSG caps best preserve the spectral features and the luminosity levels of InP after annealing. ACKNOWLEDGMENTS

This work was supported by the Office of Naval Research under Contract N00014-76-C-0806 and by the Joint Services Electronics Program (U.S. Army, U.S. Navy and U.S. Air Force) under Contract N00014-79-C-0424. We wish to thank R. L. Henry and E. M. Swiggard of the Naval Research Laboratory for the InP substrates. We would also like to acknowledge the technical assistance of S. Modestl and S. S. Chan. REFERENCES 1 K V Valdyanathan, M J Helix, D J Wolford, B G Streetman, R J BlattnerandC A Evans, J E l e c t r o c h e m S o c , 124 (1977) 1781

170

2 3 4 5 6 7 8 9 10 I1 12 13 14 15

16 17

J 1) ()BERSTAR. B ( ,

SIVREETMAN

R L Hcnr,, and E M S,slggard. Ga//tum 4J~cnMc and RchttedL'ompoumA 19"6. m In~t Pins Con/ ~;e; 33h(1977128 J D Obcrstar. B G StreetmamJ E BakerandP WiIhamb. J Lh.~l;o~hcm Sm ,12S(1981)1814 B O Seraphm and H F Bennett. m R K Wdlard,,on and A C Beer (cd,,). Semt~onsht~lors am[ 5"emtmetaA. Vol 3. Academic Press New "York. 1968. p 499 J D Obcr,,tarandB G Streetman. J 4ppl Pln~ 53(198215154 N N Sirota.. In R K Wdlard~on and A C Bcct (cds). Seml~omh. tots and Semtmetal~. Vol 4 Aeadcm,c Pro,,,,. New York, 1968. p 80 K J Bachmann and F Buchlcr J Eh,~m)dwm Sm . 121(1974) 835 G G Baumann. K W B e n e a n d M H Pdkuhn. J Electro, hem % ~ . 1 2 3 ( 1 9 7 6 ) 1232 E W Wllham.,. W Eldm. M G Astle,,. M Webb. I B Mulhn B Straughan and P I Tufton. J Eh,¢tpothem Sos 120(197"q 1741 Y Yamazoe, 5' S,isal, T Nl',hlno and ~ Hamakawa, ,lpn ,I 4ppl Ptl~ ~ , 20 ( 1981 ) 347 J I I Fist.hbach,(; Benz. N S t a t h a n d M H Pdkuhn, S'ohd-,Slalc('onlmun 11(1972) 725 K T ' , u b a k l a n d K S u g o a m a lira 1 4ppl Ph~s 19(1980) 1185 J P Donnelle) a n d C E Iturwltz. 4ppl Phl~ Lell 31(1977) 418 J Kasahara, l F G~bbons. T J M a g e e a n d J Peng, J 4ppl Pins 5l(19801419 B V Shanabrook. P B Klein, P G Slebenmann, H B Dlctnch and S G Bishop E',temh'd 4hst;a~Is o / tits' Eh'{l;o{henm a/ So, tell Fall gle~'l , Dent e;. 19,~1. Electrochemical Socmty. Pennmgton, NJ. 1981. p 841 W H Ko',chel. S G Bishop. B D Mc('ombe, W "Y Lure and H H Winder Ibm htt S l m p on (;a 4s aml Related Coml~oun¢l.s /976. in h, st Pin ~ (on[ Set 33a (1977t p 98 G S Pomrenke. Y S P a r k a n d R L Hengehold. J 4ppl Phts 52(1981)969