474-rubloff-phys-1987-2379 - VOLUME 58 NUMBER 22 PHYSICAL...

Info iconThis preview shows pages 1–4. Sign up to view the full content.

View Full Document Right Arrow Icon
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 2
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 4
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: VOLUME 58, NUMBER 22 PHYSICAL REVIEW LETTERS 1 JUNE 1987 Defect Microchemistry at the SiOZ/Si Interface G. w. Rubiofi‘, K. Hofmann,(“) M. Liehr, and D. R. Young 0” [BM Thomas J. Watson Research Center, Yorktown Heights, New York 10598 (Received 21 November 1986) The intrinsic oxide decomposition reaction Si+Si02H ZSiOI at the SiOz/Si interface is shown to be nucleated at existing defect sites prior to the growth of physical oxide voids. At lower temperatures than needed for void formation, these defects become electrically active, leading to low—field dielectric break- down unless sufiicient O; is available (low concentrations). The systematics of the required 02 suggests strongly that it reverses the initial decomposition by reoxidizing the SiO product at the interface. PACS numbers: 82.30.Lp, 64.80.Gd, 68.35.-p Thermal oxidation of Si yields thin Si02 layers of ex- tremely high dielectric quality which form the basis for Si field-effect—transistor (FET) technology. Research has yielded significant advances in characterizing defects in such systems from a fundamental] point of view. Our ability to minimize and/or control such defects plays a crucial role in the advance of the technology: For exam- ple, chemical treatment of defects by postoxidation am- bient annealing processes is a prerequisite for producing state—of-the-art FET gate oxides. Although the correla- tion of ambient annealing processes with oxide electrical quality suggests models for the underlying defect chem- istry, the microscopic chemistry of defects in Si/SiOz systems is little understood. Recent work has begun to elucidate the microscopic chemistry of typical bonding arrangements in these sys- teams,2 including those of high-density defects (~1013-1015 CID—2) which can be observed by use of surface-analysis techniques. While these results carry significant insight, it remains a challenge to define how they relate to the lower-density defects which dominate electrical properties, such as charge trapping states (~1010 cm-Z) or dielectric-breakdown sites (~103 cm _2) in high-quality structures. The microscopic chemistry of the oxide decomposition reaction Si+SiOzfi> ZSiOT has recently been shown3 to involve lateral consumption of the oxide: Voids are formed entirely through the oxide to expose the Si sur- face, and the voids grow in diameter by surface self- difiusion of Si atoms to the oxide at the periphery of the void. The lateral inhomogeniety of the reaction suggests the possibility that defects play a role in the process,3 and initial measurements of void density and size distri- bution during the reaction are consistent with a picture of defects as the nucleation site for the reaction.“'5 We report here the detailed kinetics of oxide-void growth induced by vacuum annealing, which demon- strates clearly that oxide decomposition reaction is ini- tiated at defect sites already present in the Si/SiOz struc- ture and thus provides an effective means for decorating microscopic defects. Annealing at ~750—900°C (insufficient for void growth) transforms these defects into an electrically active state which causes low-field dielectric breakdown. Both observations are consistent with enhanced oxide decomposition (Si+Si02H ZSiO) at the defect site. With sufl‘icient 02 (ppm level) present during annealing, low-field breakdown is prevented, pro— viding a means for chemical control of defects; the tem- perature dependence of the 02 concentration needed im— plies that the 02 acts to reoxidize the SiO product at the defect. Dry thermal SiOz layers 500 A thick of device quality were grown on Si(100) at 900-1000°C. Postoxidation annealing at 750—] 175°C was carried out in ultrahigh vacuum,3 in some cases with low partial pressures of 02 purposely present. A scanning Auger microscope was used to obtain both scanning-Auger-microscopy and scanning-electron-microscopy (SEM) images for the void kinetics studies.5 For the electrical measurements, metal-oxide-semiconductor (MOS) capacitors were then formed by evaporation of ZOOO-A-thick A1 dots (0.032- in. diam) through a mask, by use of an rf-heated cruci- ble. Ramped dark current-voltage (I-V) measurements with positive gate bias were employed to characterize low-field breakdown.6 From SEM micrographs of the Si/SiOz surface, it has been possible to determine the oxide—void density and size distribution at various stages of annealing. An ex- ample of this is shown in Fig. l for annealing at 1175 °C, which represents a lateral range of observation of 0.3><0.7 mm2 and a void density ~102 mm ‘2. After an incubation time required to generate voids of sufficient size to be observable, the number of voids present does not increase significantly with time. In addition, the size distribution for most voids remains quite sharp (narrow) as the voids grow in diameter. Both observations demon- strate that a few new voids are formed spontaneously un- der annealing conditions; rather, the voids must be characteristic of sites (defects) which were already present in the structure before the annealing began. This is extremely strong evidence that the void growth process is nucleated at existing defect sites and thus provides a method for decorating (revealing) microscopic defects. It is also evident from Fig. 1 that a considerably small- © 1987 The American Physical Society 2379 VOLUME 58, NUMBER 22 PHYSICAL REVIEW LETTERS VOD NUMBER l l 2 81812141b182022242b “AOQ ANNEALTINIE(min) FIG. I. Size distribution of oxide voids obtained by ab- sorbed current SEM micrographs for a dry thermal oxide an- nealed in vacuum at 1175°C. Voids were observed over an area 0.3 X0.7 mm2 at density ~102 mm ‘2. er number of additional voids are formed somewhat later in time. These “minority species” of voids grow at what appears to be the same rate as the majority species of voids, suggesting that the lateral growth of the void after nucleation is independent of the nature of the nucleating defect. It may well be that the nucleation times of different types of defects differ, thus distinguishing different defects by their different sizes at a given stage of annealing. Nevertheless, the sample represented in Fig. 1 contained mainly one species of defect. We may expect the size distribution of voids to be much broader and/or complicated if several defect types were to be present in comparable numbers. Additional defects asso- ciated with metal impurities7 or preoxidation implanta- tion damage8 produce larger numbers of additional voids. It is important to note that the defect sites which nu- cleate the decomposition reaction to form physical voids in the oxide are normally not electrically active: MOS capacitors fabricated from oxides like those studied here show excellent electrical characteristics, e.g., Z 8—9- MV/cm breakdown fields without low—field breakdown events. However, one would expect some electrical man- ifestation of the evolution/growth of the initial defect be- fore it becomes a physical void in the oxide as a result of annealing. Annealing in sufficiently oxygen-deficient conditions (either vacuum or clean inert gas) indeed re- veals the activation of electrical defects for temperatures 2 750°C. As an example, Fig. 2 shows the I-V characteristics for MOS capacitors produced after 900°C vacuum anneal- ing with different 02 partial pressures present. Without postoxidation annealing (POA), the current remains low and constant until Fowler-Nordheim (field—assisted) tun- neling begins near 6 MV/cm. However, with annealing at low 02 pressures of 5X l0_6 or 5><10_3 Torr, low- field breakdown (LFB) events occur as indicated by ex- cessive current levels at fields <6 MV/cm. (Vacuum annealing, not shown here, is very similar to the 2380 1 JUNE 1987 72 10 I 10’3 7 (Dry oxidation lOOOOC) _ I Annealed 900°C so min in UHV with: NA 10‘1 4 0 53(02): 5 x10_bTorr 7 g A plOz):5x10‘:Torr 2 75 I p (02) = 5 x 107 Torr v 10 7 7 5 <7: No postroxidation anneal é , E E '5 o i 8 10 Oxide Field ( MV/cm ) FIG. 2. Typical ramped [-V curves for MOS structures an- nealed at 900 °C in Oz ambients of various pressures. 5X10_°—Torr result.) These LFB events are indicative of a deleterious defect in the MOSA structure, which renders it useless for a device. In contrast, if a sufficiently large concentration of 02 is present in the annealing ambient, low-field breakdown does not occur. This is demonstrated by the characteris— tic for 5><10T2 Torr O; in Fig. 2. Note that in this case a characteristic essentially identical to that of the control (no POA) is obtained. Thus the presence of sufifcz'ent oxygen during annealing prevents the formation of the electrical defect. It is natural to assume that the electri- cal defects shown here evolve from the same microscopic defect (originally inactive electrically) which nucleates the decomposition reaction and leads to the formation of oxide voids. From such measurements we have established the sys- tematics of the 02 concentration required to prevent low-field breakdown. In the form of an Arrhenius plot, Fig. 3 shows as squares the various combinations of Oz partial pressure and annealing temperature (during POA) for which we determined the presence or absence of LFB events. Solid symbols indicate that no LFB was observed, open symbols the opposite. Note that the pa- rameter space for the prevention of LFB is rather sharp- ly separated from that where LFB occurs, as represented by the line pcm. Ultrahigh-vacuum annealing also yields LFB. Thus a minimum 02 partial pressure is required at a given temperature to prevent LFB. Furthermore, re- sults are also shown in Fig. 3 for annealing in low partial pressures of N2 and Ar. Since pressures above pm. do not prevent LFB, we conclude that 02, not N2 or Ar, is required. (Other oxgen-containing species like H20 VOLUME 58, NUMBER 22 PHYSICAL REVIEW LETTERS 1 JUNE 1987 900 850 800 750°C 2 | l | | ‘0 I I I 1 Low Field Breakdown: 1O 0 Ultrahigh Vacuum _ A Dry Oxidation Iooo°c N2 °’ A’ 1 — ‘3' 02 L No Low Field Breakdown: -1 _ O 10 g 2 :1 8 ‘— m _ 5 U) (I) e _ a. pent E a 7 p30 10’7 v i 8 to o o o _ o o o o o o 10 9 _____J,_.__;_l__;_4.— 8.0 8.5 9.0 9.5 10.0 10.5 11.0 1/T (10'4/°I<I FIG. 3. Effect of 0; pressure on low-field breakdown in SiOz films (500 A) annealed for 60 min at 750 to 900°C. Solid and open symbols represent annealing conditions without and with breakdown degradation, respectively. The critical 0; pressure for suppression of low-field breakdown is indicated by the solid line. Also shown is the equilibrium pressure for SiO (dashed line). might also suffice, a possibility yet to be explored.) Figure 3 also shows as the line labeled pSio the temperature-dependent vapor pressure of SiO, the prod— uct of the decomposition reaction. It is significant that the “phase boundary” pcm has the same slope (activation energy) as the slope of SiO desorption. This reveals that electrical defects leading to LFB are prevented by the supplying of 02 faster than the release of SiO, assuming this to be the rate-limiting step in transforming defects into an electrically active state (for LFB). This suggests that the 02 acts to reoxidize the SiO product to a non— deleterious state (e.g., $02) by the reaction ZSiO+Oz~+ ZSiOZ. To provide an 02 pressure at the interface (where the reaction begins) sufficient to bal- ance the SiO vapor pressure may require a higher 02 pressure over the oxide surface; this may explain the vertical displacement of the pan from the pSio lines in Fig. 3. The results in Fig. 3 demonstrate clearly that low con- centrations of 02 are required during annealing in order to prevent low-field breakdown. In principle, this pro- vides a method for defect control, since it is unlikely that technology would ever be able to prevent defects entirely. In view of these results, it might be surprising that high- quality FET oxides are ever produced, because inert-gas ambient annealing in these temperature ranges are rou- tine steps to device fabrication. The answer lies in the actual concentrations needed: E.g., in a typical l-atm POA furnace, the 10 ‘4 Torr 02 needed at 750°C would correspond to ~01 ppm. Such pure furnace environ- ments are extremely difficult to achieve, and LFB prob- lems as described here should be expected only in some future generation of processing equipment. Together with previous results concerning the micro- scopic chemistry of oxide decomposition in SI/SIO2 struc- tures,3 one obtains a picture of three stages of defect evolution. Initially a microscopic defect which exists may well be inactive with respect to some electrical characteristic (e.g., LFB). With annealing, the decom- position reaction which dominates oxide-void growth also occurs preferentially at that defect site in order to nu- cleate the formation of the oxide void. It is not surpris- ing that the activation energy is somewhat lowered at a defect site (cf. an ideal bonding site), since the local chemical bonding there is certainly disturbed from the ideal in some way. During the evolution of the defect by annealing, the defect becomes electrically active, e.g., as in low-field breakdown. At a later stage the reaction around the de- fect proceeds far enough to produce a microvoid sufficient to permit escape of the SiO decomposition product. If an additional reactant (02) is supplied at sufficient rate to the defect, the chemistry may be driven toward stability of the structure by the reoxidizing of the SiO decomposition product to $02, thus reforming the interface which was degraded by the decomposition reac- tion. In this work we have studied the microchemistry of defects associated with low-field breakdown, which occur at very low density (perhaps ~lO3 cm_2). We have also found9 that hole-trapping rates increase under simi- lar oxygen-deficient conditions but decrease with sufficient 02 concentrations during annealing. Since such trapping states occur at higher density (perhaps ~10“) cm—Z), related versions of this defect micro- chemistry may afTect other kinds of defect bonds in Si/Si02 structures than those which dominate break- down. Further studies of other gaseous species and to- ward the range of higher total pressures are planned. This work underscores the potential of research which couples a high degree of control of chemical processes to the electrical figures of merit for the thin film and/or in- terface of interest. We are grateful to M. Bradley and S. I. Raider for valuable discussions. We thank .I. E. Lewis for coopera- tion in the SEM studies and J. Calise for growing the ox- 2381 VOLUME 58, NUMBER 22 ides and depositing the Al contacts. This work is spon- sored in part by the US. Olfice of Naval Research. (“Permanent address: Forsehungsinstitut, Allgemeine EleklricitaI-Gesellschaft Telcfunkcn, D-7900 Ulm. Federal Republic of Germany. (b)Permanent address: Fairchild Laboratory. Lehigh Uni- versity. Bethlehem. PA [8015. 1G. J. Geradi. E. H. Poindexter, P. J. Caplan. and N. M. Johnson. Appl. Phys. Lett. 49. 348 (l986); P. J. Caplan, E. H. Poindexter, B. E. Deal. and R. R. Razouk, J. Appl. Phys. 50, 5847 (1979). 2G. Hollinger and F. J. Himpsel, Appl. Phys. Lett. 44. 93 (1984); M. Sobolewski and C. R. Helms. J. Vac. Sci. Technol. A 3, 1300 (1985); F. J. Grunthaner. B. F. Lewis, J. Maserjian, 2382 PHYSICAL REVIEW LETTERS 1 JUNE1987 and A. Madhukar. J. Vac. Sci. Technol. 20, 747 (1982). 3R. Tromp, G. W. Rubloff. P. Balk. F. K. LeGoues, and E. J. van Loenen, Phys. Rev. Lett. 55, 2332 U985). 4K. Hofmann, G. W. RublofT. and R. A. McCorkle. Appl. Phys. Lett. (10 be published). 5M. Liehr, J. E. Lewis, and G. W. Rublofi". J. Vac. Sci. Technol. A (to be published). 6K. Hofmann. G. W. Rublofl“. and D. R. Young, to be pub- fished. 7M. Kobayashi, T. Ogawa, and K. Wada. in Extended Abstract: of The Electrochemical Society, Canada, [085 (Electrochemical Society. Pennington, NJ. 1985). Vol. 85-]. No. 66. p. 94. 85. l. Raider, K. Hofmann, and G. W_ Rublofl", to be pub— lished. 9K. Hofmann. D. R. Young. and G. W. Rubloff. to be pub- fished. ...
View Full Document

This note was uploaded on 12/10/2009 for the course IF PFIS1200 taught by Professor Antonio during the Spring '09 term at Universidade Federal do Rio de Janeiro.

Page1 / 4

474-rubloff-phys-1987-2379 - VOLUME 58 NUMBER 22 PHYSICAL...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online