Pd-cell purified hydrogen for highest purity AIGaAs

grown by MOVPE by Wilson Chu, Marketing Manager, Catalytic Systems Division, Johnson Matthey

& Tom Purcell, Product Manager, HEA

While the semiconductor industry's purity debate continues, we now know one thing for certain- carrier gas hydrogen from Pd-cell purifiers is purer than that made by getter-type purifiers

according to DLTS measurements made in studies of AIGaAs grown by MOVPE, one of the industry's most demanding manufacturing processes. And this research shows that

Pd.cell purified H2 also contributes to devices with superior photoluminescent properties.

How pure is pure ? Few would argue that high-purity source materials and carrier gases are critical to obtaining ultra-high performance electronic and optoe- lectronic devices. And, fabrication facilities are routinely demanding greater than 99.999999% purity, a purity that suppliers today are claim- ing they have little or no problems supplying. The question for the semi- conductor fabricator is whether you are truly getting the ultra-high pu- rities that are claimed and does the quality of the products you are producing reflect this?

This article describes two recent university studies employing widely accepted and reliable methods to systematically investigate the quality of semiconductors manufactured using different carrier gases. Deep level impurities of less than 1 ppm as well as optoelectronic properties of AIGaAs grown by MOVPE (metallor- ganic vapour phase epitaxy) were measured. AIGaAs grown wi th MOVPE is acknowledged by the industry as one of the most stringent material/process combinations. The research team employed two proven and industry accepted methods to measure the quality of the A1GaAs films-deep level transient spectra-

2

e-

~n

,.d

E2 E1

Nanochem ~t~ Li-based q~ ~ ~

S t a n " a r " r . o ' ,

Pd-cell ,~

- - 100 200 300 Temperature (K)

400

Figure 1. DLTS spectra of AIo.2Gao.#,s using H2 as a carrier (Study # 1 , Fig. 1)

scopy (DLTS) and photoluminesence (PL). Both studies were conducted by a research team led by Dr. J.C. Chen, assistant professor of electrical en- gineering at the University of Mary- land Baltimore Campus.

The first study evaluated N2 (vapor-

ized from liquid N2) and H2 purified by a palladium (Pd) cell Model HP-50 purifier (a product of Johnson Mat- they), and purified by a Nanochem L- 300 lithium-getter H2 purifier (a product of Semi-Gas Systems Inc.) Both manufacturers reported impur-

a0o441 I'0 o

Waste Gas Treatment

C L E A N S O R B ° Dry, Room Temperature Conversion to Stable Salts. Outlet < TLV. Unique refill service for used absorber columns.

G CS-GmbH

Headquarters Fraunhoferstrasse 4 D-85732 Ismaning Germany Phone: +49-89-962400-0 Fax: +49-89-962400-22

CS Clean Systems, Inc. 51 Whitney Place Fremont, CA 94539 U.S.A. Phone: (510) 770 9595 Fax: (510) 770 8879

Your primary worldwide source for high purity

B E R Y L L I U M for MBE

99.999+ % purity Available in chunk form

• perfect fit for MBE crucibles

1 and 2 gram pieces • larger quantities also available

~ ATOMERGIC CHEMETALS CORP. 222 Sherwood Ave., Farmingdale NY 11735-1718 USA Tel: 516-694-9000 • Fax: 516-694-9177

Ernail: 717 -1325 (~MCIma i l . com

SEND N O W FOR FURTHER INFORMATION Please send me information detailing

Semiconductor Business Information Service - £745 [PEC63A1/PEC68A1]

Related titles: ~_ GaAs Electronic Materials and Devices

A strategic study of markets, technologies, and companies worldwide [PEC76A1]

III-Vs Rev iew - The leading bi-monthly magazine for the international GaAs and compound semiconductor industries. [PEC65A1]

Name:

Job Title:

I Company: | Address:

State: Pos tCode/Z ip : _ _

Country:

Tel: Fax:

E-maih CEC631

L-__

ELSEVIER ADVANCED TECHNOLOGY

Ii i̧ ~̧ i̧~il

i iii ii iii !ii i ii i i ii iiiii i iii i iiiiiiiiiiiiiiii!iiiiii!i i i!iiiii! i i i !iiii i i iiiiiiiii iiiiiiiii!ii iiiiii!iiiiii iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii!iiiii!ii!i!iiii!i! i ........... ii iiiiiiiiiii!Niiiiiiiiii!ii!ii: i Ni! i ii i i! !! i!! iii !!i!i ! i!i i i iiiii i i

ity concentrat ions in the purified H2 of less than 10 ppb. This initial study compared the carrier gases under both fresh and aged conditions.

The second study compared H2 carrier gas p roduced by a MonoTorr Phase II 3000 getter-based purifier model, a p roduc t of SAES, and by Johnson Matthey's Model HP-50 puri- fier. Data is only for fresh conditions; Dr. Chen and his team are in the process of compiling data for the SAES get ter under aged conditions.

The carrier gas used in the MOVPE process is an important factor w h e n it comes to the puri ty of g rown materials. Carrier gases, of course, carry source materials into the re- ac tor ' s hea t ed zone w h e r e they undergo reactions in the gas phase as well as on the substrate surface. These reactions in turn lead to the decomposi t ion of the compounds; comple te decomposi t ion is critical to obtaining highly pure semiconduc- tor layers and carrier gases play a key role in that decomposi t ion process. On the one hand it serves as an impact par tner by which it influences gas phase reactions; on the other hand, its physical characteristics such as heat conductivity and capacity, viscosity and density have an influ- ence on the concentra t ion and pro- files o f the r e a c t a n t s nea r the substrate surface, and therefore also influence equilibrium reactions [1]. Several makers of commercia l gas purification devices for hydrogen or ni t rogen claim they can p roduce carrier gases with impurities (mainly oxygen or moisture) in concentra- tions of less than 1 ppb. APIMS is cons ide red the industry s tandard when it comes to measuring purity of those carrier gases.

The A1GaAs layer is very sensitive to the purity of source materials used in the MOVPE process. So rather than s imply re ly ing on APIMS or an equivalent analytical method to eval- uate purity, Dr. Chen and his re- search team chose to examine purity from a different perspect ive using the much more sensitive, and per- haps more relevant, DLTS measure- ment . It is well known that the incorporat ion of oxygen or moisture in AIGaAs will create deep-level im- purities that are non-radiative centers [2]. These nonradiative centres will significantly reduce minority-carrier lifetimes and efficiencies [3]. That 's a

0.4

0.3

-m 0.2

o.1

N i t r o g e n as carrier gas

I00

E l

' t

I ,

~ " ! i J

20O 300 400

Temperature (K)

Figure 2. DLTS spectrum of AIo.2Gao.~S using N2 as a carrier (Study # 1 , Fig. 2)

serious problem for minority-carrier devices sucl~ as laser diodes, LEDs and solar cells where long lifetimes are essential for achieving reliable devices with high-performance char- acteristics. Kisker et al. have shown that as little as 1 p p m of oxygen in the AIGaAs growth process resulted in e x t r e m e l y h i g h - g r e a t e r t h a n 1019/ccLoxygen concentrat ions in the epilayer [2]. It is generally be- lieved that ultra-high purity (less than 1 ppm) source materials such as Hz, N2, AsH3 (arsine), TMAI (trimethyla- luminum) or TMGa (trimethylgal- l ium) are c r i t ica l to p r o d u c i n g device-quality AIGaAs. Studies have also shown the deleterious effect of o x y g e n on the PL ef f ic iency of AIGaAs [4].

Since AIGaAs is so sensitive to the impurities in source materials, most impurities are at tracted to it during the growth process. The concentra- tions of impurities in the as-grown AIGaAs epilayers are directly related to those in the source materials. By measuring the quality and impurities in AIGaAs epilayers grown under the same conditions, one can determine the purity of the source materials, or in this case different carrier gases.

. . , . . , , > . . . ~ . ~ . , . ,

Page 4 6 1 ~ I Vol 9 No 3

Experimental procedures The AlxGam_xAS (Study #1- x=0 .2 ; Study # 2 : x = 0 . 2 and 0.35) layers used were grown in a conventional, horizontally configured a tmospher ic pressure MOVPE reactor in operat ion since 1992 and one that routinely p roduced high-purity AIGaAs (n = 1- 10xl015/cc) and high-quality GaAs/ AIGaAs laser devices wi th low thresh- old current density (250 A/cm 2 for SQW lasers). All reactor componen t s (valves, mass flow controller, stainless steel tubing, etc.) w e r e carefull~ cleaned and leak checked to 10 .7 scc /sec using a helium leak detector. Such results point to not only the use of the highest purity source materials available in the market, i.e. Mega-bit grade AsH 3, but to the high-purity of the MOVPE reac tor as well. This quality also assured a clean base line for experiments .

The Group III p recursors w e r e TMAI and TMGa; the Group V pre- cursor was pure arsine. No get ters we re used to purify any sources. Materials were grown at 730°C with a precursor V/III ratio of 40, and a growth rate of 40 ~tm/hr. A 2 ~m th ick u n d o p e d AIGaAs layer was

For Precise Wafer Processing

P J 5 auto-lap

~ Li~GITECH Logitech Ltd

Erskine Ferry Road, Old Kilpatrick Glasgow G60 5EU, Scotland, UK

Telephone: + 44 1389 8754,14 Telefax: + 44 1389 890956

Telex: 777673 Logtec G

.~q :

grown on n + G a A s (1018/CC) sub- strates. Total carrier flow was ap- proximately 10 l/m.

Because the experiment 's MOVPE reac to r was not load-locked nor located in a cleanroom, the interac- tion be tween the internal surfaces of the reactor and the laboratory ambi- ent, i.e. humidity, could differ de- pending on the season, i.e. summer or winter. In addition, the purity of source materials could change with time according to some studies [5]. To compensate for these conditions, the calibration run used Pd-cell pur- ified H2 as a reference (labelled a standard run) before the other car- rier gases were sampled. Such an approach provided a reliable compar- ison of material quality.

In o rder to avoid any residual impurities in the new purifier when changing from one carrier gas to another, a two-hour purge at 6 1/m was p e r f o r m e d b e f o r e g r o w i n g further samples. A detailed growth condition is similar to the one pub- lished in reference [6].

Measuring purity The free carrier concentrat ion was measu red by capac i tance-vol tage (C-V) measurements.

Composit ion of AI and Ga was determined by PL and double crystal x-ray rocking curve measurements. The compos i t ion calculat ion was based on DCC software provided by Bede Scientific Instruments Ltd. Sur- face morphology was studied with a Nomarski interference microscope.

Low-temperature PL was employed to characterize the optical properties of material. The PL setup included an Ar+-ion laser (using 514.5 nm line), a Spex monochromator , a liquid Nz- cooled Ge de tec tor and an Si CCD array detector.

To quantitatively measure the deep impurities in the AIGaAs (both Alo.2. Ga0.sAs and Alo.35Gao.65As ), deep level transient spectroscopy (DLTS) was used to characterize layers. Schottky contacts, 700 ~tm in diameter, were made by evaporating Au through a metal mask onto the AIGaAs surface immediately after etching in a freshly prepared H2SO4:HxO2:H20 (2:1:10) solution for 15 sec. Ohmic contacts were made by evaporating AuGe/Au on the n + GaAs substrate, and alloy- ing in forming gas (95% N2 + 5% Hz) atmosphere at 420°C for 2 minutes.

14

12

10

¢¢ 8

.q 6 e..

4

i 685

! T = I S K , s l i t w l d t h = l m m ! ! ~ ~ #480 ~m, Vd~eU), l~er power V=5.8 mW

#4~/(H2, Li-based purifier,, P=6.5 mW

/ #481 no peak found (N2 carrier)

690 695 700 705 710 715 720 725

W a v e l e n g t h (nm)

Figure 3. PL spectra of AIo.2Gao.sAs (Study #1, Fig. 3)

The DLTS setup was made by Sula Technologies and consisted of DLTS and a d o u b l e c o r r e l a t o r DLTS (DDLTS). The t e m p e r a t u r e scan range was from 30 to 380K. Data was taken automatically by a compu- ter th rough a high-speed analog- digital convertor board.

Pd-cell vs. Li-getter and N2 carrer gas All samples showed excellent surface morphology no matter what type of carrier gases were used. The C-V results showed that the undoped AIGaAs layers grown in H2 always had n-type background concentra- tions from 3xlO 15 tO 10 x 1015/cc. The typical DLTS spectrum for an undoped Alo.zGao.sAs layer of a stan- dard run, which had a carrier con- centration of 5.0x10~5/cc, is shown in Fig. 1. Three main deep electron traps, El, E2 and E3, were observed with activation energies of 0.70, 0.53 and 0.30 eV respectively. These traps are commonly observed in MOVPE- grown AIGaAs layers [7,8]. The 0.70 (El) and 0.53 (E2) eV traps were associated with oxygen-related impu- rities that may come from source materials, the reactor, the carrier gas or a combination of all three. The 0.30 (E3) eV trap was attributed to Ge contamination in source material

Page 48 Vol 9 No 3

and was not related to carrier gas [7]. Therefore, the very low concentra- tion (-1013) of this E3 trap in the undoped A1GaAs indicated that the source materials were in general purer than those repor ted by other groups which had E3 density around 1014/cc [7].This assured that a clean base line was established for the experiment. A shallow electron trap, E4, was also observed with an activa- tion energy of 0.08 eV, which could be a shallow impurity level.

The concentrat ions of oxygen-re- lated traps (Ec-0.53 and 0.7 eV) are 0.2 to 9 xlO 13 and 3.4-5x1014/cc for Pd- and Li-purified H2, respectively. The deep impurity level of AIGaAs grown using Pd-purified H2 is con- sistently lower than that of A1GaAs using the Li-based purifier, the differ- ence being one to two orders of magnitude.

Equally important, the lifetime of a Pd-cell is longer than that of an Li- based purifier. In 1964, Johnson Matthey invented and patented the Pd alloy dlltusion membrane technol- ogy that permit ted for the first time, cost-efficient, high-capacity produc- tion of ultrapure H2 (99.999999%). The purifier incorporates a barrier membrane that permits only H2 to pass through. On the other hand, the Li-based purif ier relies on highly reactive metal compounds such as

0.15~

0.1£

o 0

oq

.J

0.05

Undoped A1GaAs x=20%

I 4OO

Oxygen

~ ' ~ E1 SAESSampte 796 t ~ 3 x ~ 10 ,~ cm-3

%% E2

_1013 cm -3

Sample 791 Pd cell

I I I I 200 240 280 320 360

T e m p e r a t u r e (K)

Figure 4. DLTS spectra of AIo.2Gaso.eAs using H2 as a carrier (Study # 2 , Fig. 1)

lithium hydride to snag impurities as the H2 flows through. Eventually, the compounds break down and allow impurities to pass through as well. In the first study, the Pd-cell was used for more than three years, while the Li-based purifier became saturated after 10 cylinders of H 2 (volume ~213 ft3/cylinder, 99.99% p u r e ) w e r e used.

Figure 2 shows the DLTS spectrum of AlGaAs grown using N2 (without any getter) as the carrier gas. This r epresen t s a p- type (background concent ra t ion , p -6x l0~6 / cc ) with possible oxygen-related traps. The concen t ra t ion of these traps are more than one order of magnitude higher than that of AlGaAs using Pd- cell purified H 2.

Table 1 shows the deep impurity concentrations in different samples. The memory effect due to impurities left in the reac to r and /o r o the r source materials was also studied. As can be seen in Table 1, the E1 trap densities of standard runs 480 and 482 are 2 and 3 x 1012 /cc . The difference be tween them is very small, which suggests that the mem- ory effect of impure carrier gas in source materials may be negligible as

long as the residence time in source is not very long. However, the long- term memory effect remains un- known.

Fig. 3 shows the low temperature (15 K) PL spectra for three samples using Pd-ceU and Li-based purified H 2 a n d N2 as the carrier. The samples using the H 2 have similar PL spectra. However, the intensity of the near band edge peak (~, = 698nm) is stron- ger if a Pd-cell purified H2 is used. This indicates that Pd-cell purified,Hz is purer than that from an Li-based

purifier, which means devices with superior optical properties. This data is in good agreement with the DLTS results.

Note with the N 2 sample that no near band edge peak can be detected. However, the study uncovered a s trong peak located at a longer wavelength (~= 1063 nm) that was observed for the first time in AlGaAs samples. It suggests that some un- known impurities were introduced w h e n N2 was used. The radiative nature of this deep level impurity is interesting and not being reported, but research is underway to help understand its origin.

Pd-cell vs. SAES getter All the A1GaAs layers grown had similar background carrier (n-type) concentrat ions be tween l x l 0 ]5 and 2x1016/cc, which indicated the con- centration of deep level impurities was below background carrier con- centrations.

All four AlGaAs layers studied can produce strong PL under low-tem- perature (14K) measurements. While all layers had very good quality, again the PL intensity of layers grown by using Pd-cell purified H 2 a r e consis- tently stronger than that of layers grown using the SAES get ter purifier. However, it should be noted that the PL measurement is sensitive to sur- face preparation and other issues, which suggests that the low-tem- perature PL measurement only pro- vides a qualitative comparison.

As with the first study, here again the DLTS technique (see Table 2) proves to be the most sensitive in measuring the deep level impurities

Table 1. Deep impurity concentrations in AIGaAs using H2 and N2 as carrier gases (Study .#1)

Gas Sample # Type Oxygen-related trap density (1/cc)

Pd-H2 OM368 n 6.6 xl 013

Pd-H2 OM374 n 8 xl 013

Pd-H2 OM377 n 9 x1013

Li-H2 OM384 n 3.4 xl0 TM

Li-H2 OM404 n 5 x10 TM

Pd-H2 OM480 n 2.0xl 012

N2 OM481 p 1.8 xl0 TM

Pd-H2 OM482 n 3 x1012

Pa0e491 IvogNo3

Table 2. Deep impurity concentrations in AIGaAs using H2 as a carrier gas (Study #2)

Gas Sample # Type

OM7911 Pd-H2

SAES-H2 OM7961 n

Pd-H2 OM7952

SAES-H2 OM7972 n

1 AIo.2Gao.~s

2 A o.3sGao.ssAs

n . . . . . .

3 xl013

n -1 xl0 ~3

5 xl013

Table 3.Overall Comparison of Carrier Gases (Studies I and 2)

Purifier DLTS Results~ x ~01 ~ PE ,ntensit~ Rat!ng Pd-Oe ; ; 0 2 t o l

SAES getter : 3 t o 5

L i~ge t te r . . . . . . . ; Weaker; ; F a i r

N i t rogen ;; ; ;18 ; ;Undetectable Poor

Contacts: Wilson Chu,

Johnson Matthey, Catalytic Systems Division, 460 East Swedesford Road,

Wayne, PA 19087-1880

U.S.A Telephone: [1] (610) 971-3105

Fax: [1] (610) 293-1284 Email: CHUW@Matthey.com

Tom Purcell, HEA,

Bessemer Drive, Stevenage,

Herts, SG1 2DX, UK

Tel~fax: [44] (0) 1438 310091/354 745.

quantitatively. The results show that the AIGaAs grown with Pd-purified H2 contain 3 to 5 times less deep level impurites than that grown with SAES get ter purified H2. The results are also shown in Fig. 4. The lowest trap level is less than lx1013/cc, which was achieved by the Pd-cell purified H2.

Pd-celh proven purity With the first study, by measuring the deep impuri ty densi ty in AIGaAs grown by MOVPE, it was found that the lowest oxygen-related impurity (2x1012/cc) was achieved by using Pd-ceU purified H 2 as a carrier. In addition, low-temperature PL mea- surements showed the Pd-purified H2 produced devices with superior op- tical properties. AIGaAs with poor optical quality and high impuri ty density was obtained when N2 was used as a carrier. Finally, the lifetime of a Pd-cell is longer than that of an Li-based purifier. Pd-ceU purifiers do not wear out with time like getter- type purifiers which eventually get used up.

In the second study, Dr. Chen and his research team observed that while the AIGaAs layers grown using H2 car r ie r gas suppl ied by two different purifiers all showed reason-

able quality, the sample grown by the Pd-cell H2 always had lower oxygen- related deep traps and be t t e r PI efficiency than those grown by the SAES get ter purifier. The lowest trap dens i ty p r o v e d to be less than lx l013 /cc .

Table 3 provides an overall com- parison for the carrier gases evalu- ated in both studies.

References [1] H. Hardtdegen and P. Giannoules, An Outstanding Innovation in LP-MOVPE: Use of Nitrogen as the Carrier Gas. Ill- Vs Review, Vol. 8 n.3, pp. 34-39. [2] G.B. Stringfellow, Organometallic Vapor-Phase Epitaxy: Theory and Prac- tice, p. 298-300 (Academic, Boston, MA 1989) [3] M.J. Tsai et. al., J. of Electronic Material, 13, 437 (1984). [4] D. Kisker et. al. Appl. Phys. Lett. 40,614 (1982). [5] H. Terao and H. Sunakawa, J. Cryst. Growth, 68, 157 (1984). [6] Z.C. Huang et. al. AppL Phys. Lett 66, 2745 (1995). [7] R.T. Green and W.I. Lee, jr.. of Electro, nic Material 20, 583 (1991). [8] P.K. Bhattacharya, S. Subramanian and M.J. Ludowise, J. AppL Phys., 55, 3664 (1984).

Pa0e 01 Ivo0No3