Probes for Monitoring Low Dose Silicon Ion Implants forProcess Yield Enhancement

Daniel H. Rosenblatt

Samsung Microwave Semiconductor1530 McCarthy Boulevard

Milpitas, CA 95035408-433-2222

ABSTRACTA newly developed optical densitometry probe,and existing thermal wave, eddy current sheetresistance, and four-point GaAs sheet resistanceprobes were used to measure low dose Si+ ionimplams. The four probes were evaluated fortheir ability to detect implant dose uniformity,their sensitivity to low doses, and theirrepeatability for multiple measurements on asingle wafer, and on several wafers implamedwith the same dose. The new opticaldensitometry probe was found to be highlysensitive (S = 0.5 - 0.6) to low doses (Si+, 100keV, 0.75E12 - 5E12 cm-2). The thermal waveprobe was found to be relatively insensitive (S =0.1 - 0.15) in the same low dose region (Si+, 100keV,2E12- 4E12) and moderately sensitive (S =0.40) at higher doses (Si+, 150 keV, 6E13 -9E13). The eddy current sheet resistanceprobe was found to be about four times moreprecise as the GaAs four point probe. Byproviding precise information about low Si+ doseuniformity, all four techniques can be applied toyield improvement of GaAs devices made withion implantation.

INTRODUCllONOne of the key factors in yield enhancement ofGaAs ICs made on ion implanted active layers isthe ability to monitor the dose uniformity across awafer and the dose repeatability from wafer towafer of the ion implanted layer. Manytechniques used to monitor B+, P+, and As+implants used in Si device applications (e.g.,photoresist-based optical dehsitometry, C-Vprofiling, and DC four point probing) have lackedthe sensitivity and repeatability to monitor lowdose Si+ implants used in GaAs deviceapplications. The purpose of this work is toexplore four probes which are particularlysuitable for GaAs manufacturing technology.Each probe can detect low Si+ doses and assll"ethat the uniformity of implants is sufficient toobtain high device yields.

The optical densitometry and thermal wavetechniques directly monitor the implant and donot require an anneal of the substrate being

implanted. The eddy current sheet resistanceand GaAs four point sheet resistancetechniques require an anneal of a GaAssubstrate after the ion implant and are thereforeindirect measurements. Indirect monitors aremore prone to inaccuracies introduced by thenon-uniformities inherent in GaAs substratesand annealing processes.

EXPERIMENTAL DETAILS28Si+ and 29Si+ implants were done on anEaton 200 MCA ion implanter at SamsungMicrowave and on a Veeco 2400MP implanter atIICO Corp., respectively. The tilt and twist angleswere 10 and 20 degrees, respectively, tominimize ion channeling. For the opticaldensitometry study, polymer coated glasssubstrates were used. For the thermal wavestudy, Si{100) substrates were used. For thesheet resistance studies, semi-insulating GaAs(100) substrates were used. The implantedGaAs substrates were furnace annealed in anarsine ambient for 40 minutes at 850 degrees C.

A NEW OPTICAL DENSITOMETRY PROBE1Until recently, the most advanced opticaldensitometry technique to monitor dose usedphotoresist coated glass wafers. These wafersare light sensitive, have short shelf-life, exhibitserious relaxation after implant, and are relativelyinsensitive at doses below 1E14. In this paper,measurements of Si+ implants by a newlydeveloped instrument (which is nowcommercially available) are presented for the firsttime. In order to read doses in the 1E11 -1E13region with this new instrument, a dyeimpregnated transparent polymer film has beendeveloped that changes its optical density whenimplanted. The polymer film is applied to arecyclable glass wafer. which is then implanted.Optical density is measured at about 350 siteson a 3" diameter wafer. The mean number ofoptical density counts. and the one sigmastandard deviation of those counts are thencalculated.

Figure 1 shows a map of the across-waferuniformityfor a 100 keV, 3E12. 29Si+ implantusing this technique. The standard deviation for

the implant is indicative of the uniformity of theimplant itself, and not convolved with the effectsof the GaAs substrate. the anneal, and the sheetresistance probe, which can add substantial non-uniformity to sheet resistance measurements ofimplants into GaAs.

In order to determine the sensitivity of thetechnique to dose changes in the low 1E12region at 100 keV, 6 different Si+ doses from0.75E12 to 5E12 were implanted in op!icaldensitometry substrates. Three substrates wereimplanted at each dose. Results of theseimplants are shown in Figure 2. This polymersystem has high sensitivity for very low doses.where sensitivity is defined as:

S = Percent change in measured oarameterPercent change in implant dose

A sensitivity of 1.0 would mean that doubling theimplant. dose would double the measured opticaldensity. From the data in Figure 2, the sensitivityof optical densitometry is found to be 0.51 - 0.62over the range 0.75E 12 to 5E12. The figurealso shows the excellent repeatability in meanoptical density counts for the three wafersimplanted at each dose. This sensitivity is highfor the low 1E12 dose region, which is the doseregion chosen for the channel implant in manymicrowave and digitalGaAs IC applications.

THERMALWAVEPROBE1The thermal wave technique measures damagedone to the substrate by the implanted ions. Adifficultyobserved in earlier studies on Si+ intoGaAs is that unimplanted GaAs substratesdisplay a noise level which is two orders ofmagnitude higher than unimplanted Sisubstrates2. Consequently, to obtain anadequate signal to noise level for low doseimplants, Si substrates must be used.

Figure 3 shows a map of the across-waferuniformity in thermal wave units a 100 keV,3E12, 28Si+ implant using this technique. Aswith optical densitometry. the standard deviationis indicative of the uniformityof the implant itself.In order to determine the sensitivity of thetechnique to dose changes in the low 1E12region, 28Si+ implants were done at 100 keVwith a doses of 2E12, 3E12, and 4E12 in toSi(100) substrates. Results of these implantsare shown in Figure 4. The thermal wavetechnique has very low sensitivity (0.1 - 0.15) inthis region. This is about one-fourth thesensitivity of optical densitometry in the samedose and energy range for Si+.

The thermal wave sensitivitywas also explored inthe high 1E13 dose range. 28S;+ implants weredone at 150 keV and doses of 6.75E 13,7.5E13, and 8.25E13. Results of theseimplants are shown in Figure 5. The thermalwave technique has moderate sensitivity (0.4) inthis high dose region.

COMPARISON OF EDDY CURRENT AND

FOUR-POINTSHEET RESISTANCE.PROBES1A commercially available eddy current probe wascompared to a commercially available GaAs fourpoint probe for accuracy and precision. Notethat the GaAs four point probe differs from thestandard four pqint probe used formeasurements of sheet resistance on doped Siwafers in that special AC electronics and adifferent probe head are required for GaAsmeasurements. The precision study was doneby studying the "discriminatingpowerMof the twoinstruments. The discriminating power of aninstrument is its ability to differentiate betweendifferent pieces of product (in this case 24measurement locations on the same wafer) bymaking multiple measurements on each pieceand studying the range of measurement error.Discriminating power is defined in detail in theAT&T Statistical Quality Control Handbook.

In this study, a single GaAs substrate (10# 233x-45), was implanted with 28Si+ at 150 keV to adose of 7.5E13. After anneal, the wafer wasthen measured by both probes at the identical24 locations every day for five consecutive days.A table of X-Bar (average sheet resistance for 5measurements) vs. measurement locationnumber (1-24) was generated for each probe.From this table, plots were derived which looksimilar to standard SPC charts. These plots areshown in Figure 6 (for the eddy current probe)and FIgure 7 (forthe GaAs four point probe).

Each point on the X-Bar chart represents adifferent location on the wafer. or a differentpiece of product, ifone assumes that the wafer isto be made into 24 devices. The control limitsare derived from standard SPC fOffiluiaS.Ifthe X-Bar points stay in control, the measuringinstrument can not see any difference betweenthe measurement locations (pieces of the wafer).A good measurement instrument should havesuch narrow control limitsthat all or most of the X-Bar points are out of control.

In FIgure 6, most of the X-Bar points falloutsidethe control limits, which are very narrow.Consequently. the eddy current probe iscapable of reproducing its results very closelyand can easily distinguish between themeasurement locations on wafer 233x-45. It isvery suitable for precise measurement of sheet

resistance across a wafer, Le. it has excellentdiscriminating power.

In FIgure 7, most of the X-Bar points fall insidethe control limits. This means that the lack ofprecision of the four point measurement is largeenough to make the variations betweenmeasurement locations indistinguishable. TheGaAs four point probe does not have thediscriminating power to adequately characterizethe across-wafer uniformityof wafer 233x-45.

As far as probe accuracy, the eddy current probecan be calibrated with NIST traceable standardsover a wide range of sheet resistance. Thesestandards are used by many fabs and are durablebecause they are n or p type Si substrates.Unfortunately, no equivalent standards exist for

-the GaAs four point probe. The NIST Si waferswill not work on the GaAs version of the fourpoint probe because the probe tips are fiat (tofacilitate measurement of thin GaAs conductinglayers) and do not penetrate the Si wafersenough to make a good contact.

APPLICATION OF THE EDDY CURRENTPROBETO MMICYIELDENHANCEMENTClearly, the eddy current probe is the moreprecise and accurate sheet resistance probe. Itis more suitable for yield enhancementapplications where a variation in implantiondose,substrate donor or acceptor level, annealfurnace temperature, or other process variationmust be detected to improve yield. In fact, theeddy current probe was used to measureblanket (unpatterned) channel implants intosubstrates from various suppliers. The probedetermined that substrates manufactured bySupplier A produced a factor of two betteracross-wafer uniformityin sheet resistance thansubstrates manufactured by Supplier B3. A totalof 191 MMIC wafers were then fabricated onwafers from both suppliers. Each MMICwaferhad ninevan der Pauwstructures locatedacrossit for the determination of the channel layersheet resistance and its uniformity. The sheetresistance for this channel layer is approximately600 ohms/square. The standard deviation of thechannel layer sheet resistance (for 9 van derPauw structures) as a function of supplier isshown in Figure 8. The median standarddeviation .was 14 and 31 ohms/square forSupplier A and B substrates, respectively. MMICyield on Supplier A substrates was higher, inagreement with the eddy current probe and vander Pauw structure measurements.

REFERENCES1. For the identity of the manufacturers of eachprobe, contact the author.

2. Dennis Kamenitsa, Eaton Corp., privatecommunication.3. For the identity of Suppliers A and B, contactthe author.

~.

i?~

".- ~

/:/r::~::~~~:~~\ .r ------------------- - - -- - - - - - - - - - - - - - - - --

FIgure 1. Wafer map of optical density counts fora 29Si+, 100 keV, 3E12 implant. The meancount level is 11479 and the standard deviationis 0.9%. The contour interval is 1.0%.

16000

>-

en 12000c: enClJ-

0 §ro 0.~ 0 80000.0

4000O.OOE+OO 2.00E+12 4.00E+12 6.00E+12

29Si+ Dose (cm-2)

Figure 2. Optical density counts vs. dose for29Si+, 100 keV implants. 3 wafers weremeasured at each of the 6 dose levels.

.t

..

.1

0 0 0 0 - - - - - . - - - - - - - ".- - - - - - - - - - - - - - - --------------

- - - - - - - - - - - - - -- - - - - - - - - - - - - - ----------... --0-0---0-0---

\++. 000--0 0-0-

ce---cco-c-q.coogcoco-a-dcg\J;Jg!-::Ioo-:::-g

go-cce-a..- c c - - 0 C - - - c

c -::I - C - a - 0

~~ ~~ c -..- .... C .. C 0 C 0 C 0 ~ - "'- Q..-C'"'o

"""""""".""+~c1:c/"..""0"0"0000-+0/

c" .. -+.....

FIQufe 3. Wafer map of thermal wave counts for a28Si+, . 100 keY, 3E13 implant into a Sisubstrate. The mean count level is 705 and thestandard deviation is 0.20,'0 The contour intervalis 1.0%.

'"c::J0

Uc:.>>~

$:(;jE~

1=

750 ]740]7301720 I710 -700 t690 ]680

1670

1.50E+12 3.50E+12 4.50E+122.50E+ 12

28SI+ Dose (cm'2)

Figure 4. Thermal wave counts vs. dose for28Si+, 100 keY implants. 3 wafers weremeasured at each of the 3 dose levels.

2200

S.ooE.13 7.ooE+13 8.00E.13 9.00E+ 13

~r Dose(cm"1

Figure 5, Thermal wave counts vs. dose for28Si+, 150 keY implants. 3 wafers weremeasured at each of the 3 dose levels,

102.00

!

.~ '~ 100.00 ,2 nm .-a 96 00 "~ .

j ~:~~ I ~::~:::.::::C:;:;::',":;;7<::~::::::: ::::;::.::_--x 90.00 I88.00.,.

0 10 15 20 25

Measurement locat1on Number

[ . X-Bar X.Dauble.Bar UCl "...let.

Figure 6. Plot of X-Bar of Resistance vs.Measurement Location Number for wafer 233x-45 as measured by the eddy current sheetresistance probe.

<0 101.00 Tv 100.00 +c: I2 99.00 -'- .~ 98.00 -;- ,u' --- --.~ -'-Q: ,"0 97.00 t~ 96.00 tx 95.00 T

94.00 ~

0 10 20 2515

Measurement location Number

( . X.Ber X-Double.Bar UCl lCl!

Figure 7. Plot of X-Bar of Resistance vs.Measurement Location for wafer 233x-45 asmeasured by the GaAs four-point sheetresistance probe.

v :::... -'""$:gi ;;.c~ , "

..=='".cU...0

s..<i5

co.. ~'

':. .. ":, ", ,, . ,,' ,

,::::' :'. .::,.t'l:ili:;jkppIIorA kppIW.

Subm-ate SuppUer

Rgure 8. Across wafer sigma of channel layersheet resistance VS. substrate supplier for vander Pauw stI1.JCtu"eson MMIC Wafers, Each datapoint represents sigma forone wafer,

CI) 2450'E:J

24000()c:.>

2350>

$: 2300(;jE 2250...Q)

.t::

- . --,-

:. -

J..