Arthur Sherman - Chemical Vapor Deposition for Microelectronics (779637), страница 24
Текст из файла (страница 24)
Since no theoretical description of such a complex system isavailable, reactor uniformity must be obtained by trial and error, or "characterization." As experience with such systems grows, it should eventually bepossible to model them and make theoretical predictions of their performance.5.3 SILICON DIOXIDE AND OXYNITRIDESThe only other plasma-enhanced CVD film that has seen wide use in integrated circuit manufacture is the plasma oxide film.
We say "so-called" becauseit is not truly Si0 2, but rather SiOxNyH z. In fact, it is just this ability to modifythe film stoichiometry that makes these films so valuable. Many of the filmcharacteristics change depending on this stoichiometry, so it allows a freedomto alter film characteristics that is not possible with thermally-grown films.Plasma oxide can be grown from a number of oxidizers plus SiH 4 . Amongthese are N20, 02, CO 2 and even TEOS (tetraethoxysilane). Generally, O 2is not used as it too often leads to homogeneous nucleation. The preferredreactants have proven to be Si H4 and N20, so we will restrict our discussionto these. Films grown in both cold-wall and hot-wall reactors will be considered.A very thorough study of plasma oxides deposited in a parallel-plate coldwall system has been reported by van de Ven.
8 The operating frequency usedwas 57 kHz, the pressure was 400 mTorr, and the wafers were held at 300°C.The influence of the N20/SiH 4 ratio, and the power level are shown in Figure12. As expected, deposition rates increase with power level. The etch rate inbuffered oxide etch (BOE) also increases, suggesting a less dense film with thehigher deposition rate. Higher N2 0/SiH 4 ratios give a lower index of refrac-5001{N2~ NH3 seeMc'E~Cl>400200400'Ci)01- J Vi-.J~Ij•Ii\~100b2 '8ar----- I 200"Xl -,,"CD3n'Q.)!"~~ h300('):::J"'"IrA~-~ t:-~-~ ~-":-,:4.-i200II100Q.)'"CIoIR.I~205o75199'"C752?O75200S,H4 (SCCM):?OO75.600300~400225430060010Radial Position (inches)<I"OH3"02.7001412108r~~oc:IoQ.I75~'Ci)I------ 1 9'!I400I'VII500'';;1--: sot---~6'E~Cl>____4C19975W1I_ _ ).., -,I600R I~ 19350II_ \ -.J~ ~ IJI.~ ~'~l~ ~-J~ _1r--......;~t-:::::a..~-~!.j600300SlH 4 (SCCM)!r00Cl>~I[~~~ ~cr:'';:iJ-300 f--JOOcoc0)/1+-Iaf1~11/VICl)oen;:+'1214Radial Position (inches)0'::J-ho~600c:'E~CQ)+-I500300I~ ......
"'~b4.......'Ci)00-Cl>0jE~(~ ~r-r~IIII200I1006i"02.600II400I~ICl>+-IIdCO:EI"OH35,H4 (SCeM)R I30050193•500300501 984300600501 91~100I800I81050)I12IICl>502791001 89I14Radial Position (inches)Figure 11: Deposition rates versus reactive gas flows.!y~~j~~L.-t~ II. V---- .-J~_~''''-IttI~~~~• I"02' 1700 SCCMI NH3IIIoICl)(")_:::J262I10I715+---+-1------/None:III10Radial Position (inches)12on'enI25i~r-+R I_+-----+--------...~~vV' II20000-0IS.H 4 ISeCM)t----+---+-__...C0_-j-!~300(")iI )t----+-----+-~+---___+____+-+__-+__i___+________i_:c:::III,~#VIk!~,/-~~c:'';:III400a::0!Ico~500!14Plasma-Enhanced CVD133tion, which approaches the value appropriate for thermal oxide of 1.48.
Ingeneral, higher values of N 2 0/SiH 4 and lower power levels are preferred tominimize gas phase reaction and nucleation.rideposition800Aimin 70060050011.0012001OOO"lm·n, 800rateAirJCx/,x~x _ _ )(_11.00600200ratei7°1800x//etch rate( B.O.EJAir.1400min1000Index1.70/~1000min/e1chdeposition rate1.656001.60165refractiveindex1n 1.55 -1.60n1.55145T0,20!/.0!80N O/SiH4602III100 120 1/.0ratio --.1.50 1./.8145100200 3001.00 500 6CO wattsFigure 12: Film characteristics for plasma oxide deposition versus gas composition and power level. 8Results obtained for a typical process condition are shown in Table 2,where plasma oxide and nitride films are compared.
In spite of the powerlevel being a third that used for plasma nitride, the plasma oxide depositionrate is almost twice as high ("'"'600 A/min). This is just another indication ofthe ease with which silane can be oxidized compared to its being nitrided.This data indicates only a small quantity of N in the films.
However, a 2 to 3%level of hydrogen is also mentioned. 8 Other experiments carried out at 13.56M Hz and 1 Torr pressure 9 show hydrogen content as high as 8% for some"films. It would appear that the higher degree of ionization (and dissociation)achievable at the higher frequency stimulates hydrogen incorporation. On theother hand, the stress apparently remains compressive even at the higher frequency.9Plasma oxide has found utility in high-frequency applications for dual-layerisolation,8 because of its low dielectric constant and high breakdown voltage.Also, it is in compression when deposited, so that it can be used as the dielectric when thick films (2 to 5 microns) are needed. Such thick films when deposited by thermal CVD (which is deposited in tension) tend to crack. Onefinal advantage to the use of plasma oxide rather than plasma nitride is that134Chemical Vapor Deposition for MicroelectronicsTable 2: Plasma Oxide and Nitride CharaeteristicsGases0/oSiH48Silicon DioxideSilicon NitrideSiH4 + N202%SiH4 + NH3 + N29%Ok N20, NH3 resp.98%45%RF Power Density0.05 W/cm 20.17 W/cm 2RF Frequency571{Hz57 kHzOperating Pressure53 Pa33 PaSubstrate Temperature300°C300°CDeposition Rate60 nm/min38 nm/minFilm Uniformity:!:Film CompositionSi0 1 .
9 No ' 15Refractive Index1.542.02Film Density2.38 g/cm 32.75 g/cm 3Etch Rate (B.O.E.)130 nm/min20 nm/minEtch Rate (CF4 + 02 plasma)10 nm/min150 nm/min50/0±4 %Sb.l N. (H)the hydrogen content can be much lower (i.e., 2 to 3% versus 20 to 30%).For applications where evolution of H 2 from a plasma nitride layer during laterhigh-temperature processing would be deleterious, the plasma oxide is preferred.As noted earlier, the stoichiometry of plasma oxide films can be adjustedby changing deposition conditions. The electrical behavior of a composite filmconsisting of a thin thermal oxide covered by a thin, silicon-rich, plasma oxidehas been studied. 1o The sil icon-rich fil m actually consists of sil icon crystalsinterspersed within the plasma oxide. It is deposited in a 13.56 MHz, parallelplate, cold-wall reactor operated at 600 mTorr with the wafers at 350°C.
Theratio of N 2 0 flow to SiH 4 flow was varied during the experiments fromto150 to alter the stoichiometry. As just one illustration of film behavior (composite film in this case), we show the dielectric constant as a function of N 2 0/Si H4 flow in Figure 13. 10 Increasing the Si H 4 flow and thus increasing thesil icon excess leads to a substantial increase in dielectric constant.
Apparently,the high dielectric constant of the Si-rich plasma oxide films is used for dualdielectric storage capacitors in dynamic memories.If desired, plasma oxide films can be doped much as the plasma nitride filmwe discussed earlier. In fact, doping with boron and phosphorus has been carriedout as an alternative to the standard atmospheric-pressure thermal CVD processfor BPSG. 11 ,12 The latter process has the drawbacks of high defect density andpoor thickness uniformity, so it was hoped that plasma BPSG would be an improvement. However, there are differences in the films in terms of H 2 and N 2content, and their effect on reflow temperature, intrinsic stress and passivationeffectiveness had to be exam ined.aPlasma-Enhanced CVD13516I-Z•<{l-•4' •til 12Z08w-J.....c••• IiULaJ4"SOOA 200A Si-Rich OxidetJ.U....a:....uas depositedo annealedb.4~88i iii 0l!Jo0.11,10!A6J 6, 00GAS PHASE RATIO ([N 2 0]/[SiH4])Figure 13: Dielectric constant as a function of flow ratio (N 2 0/SiH 4 ).10Depositions were done in a hot-wall tubular plasma reactor at a frequencyof 410kHz.
Deposition temperatures ranged from 300° to 400°C, and pressuresfrom 830 mTorr to 1.5 Torr. Gases used were SiH 4 , N2 0, O 2 , PH 3 and B2 H6 ,with the latter two diluted by argon. The boron and phosphorus concentrationswere adjusted by changing the reactant gas rnixture.Stress in the as-deposited films was compressive, as expected. After densification (800°C for 1 hour in steam), the stress was reduced, and in some casesmade slightly tensile. Comparison of the plasma oxide films with thermal oxidefilms is shown in Table 3. Given the compressive stress found in the plasmaoxide films, we would expect improved crack resistance.
Tests were carriedout!! on a 1-micron plasma BPSG film {P = 3.4 wt %, B = 3.2 wt %) over 1micron thick aluminum lines by thermal stressing at 460°C. There were nomicrocracks, although thermal atmospheric pressure BPSG would crack underthe same conditions. No major differences were found in step coverage or reflow characteristir~, as compared to the thermal films.The ability of the plasma BPSG film to passivate against sodium penetration was compared to the thermal film. Evidence was found of some sodiumpenetration in the plasma films, and none in the thermal ones.Since one major motivation behind the use of plasma BPSG was to providean improved passivation barrier, the better crack resistance is an advantage,but the greater sodium penetration is a negative. Therefore, it is not clear if itwould be advantageous to make this replacement l l for a final passivation film.Its use as an inter-metallic dielectric may be more useful.136Chemical Vapor Deposition for MicroelectronicsTable 3: Comparison of Stress in Plasma and Thermal Oxide Films 1!Glass TypeStress ( x 10-9 dynes cm- 2 )As-DepositedDensifiedPlasma Glass(Non-Doped)-3.0Plasma PSG(10%)-1.66 (C)Plasma BPSG-(1-2) (C)Thermal Oxide-(3-4) (C)LPCVD PSGHigh Temp.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
