NTATION PAGE 8 No. 07018&
tlecl ~ ~ ~ - "'t , q ot f'O tO ung eme fo¢rc w ftz a¢.I z t l~ n t'N us~v(
RT DATE 3. REPORT TYPE AND OATES COVEREDI.Ju.Ly 1990 .Interim. September 1988 - January 1990
4. TITLE AND SUBTITLI S. FUNDING NUMBER
Total Dielectric Isolation (TDI) of Fully Depleted AFOSR-88-0356Device Structures
6. AUTHOR(S)
Dr Peter L F Hemment
7. PERFORMING ORGANIZATION NAME(51
AND ADORESS(ES) 8. PERFORMING ORGANIZATIONREPORT NUMBER
University of Surrey, Department of Electronic andElectrical Engineering, Guildford, Surrey, GU2 5XH, UK 839, July 1990
9. SPONSORINGIMONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORINGIMONITORINGAGENCY REPORT NUMBERSponsoring/Monitoring: European Office of Aerospace,
Research and Development, Box 14,FPO New York, NY 09510-0200, USA
11. SUPPLEMENTARY NOTES
12a. DISTRIB4ION/AVAILABIUTY STATEMENT 12LECTE 1 l'b. DISTRIBUTION CODE
Approved for public release; , 25 IDistribution unlimitedJ
13. ABSTRACT (Maxmum200 words)
The aim of this programme is to prepare, evaluate and optimise Total DielectricIsolated (TDI) structures in order to develop the technology for the preparation ofsubstrates for circuits which exhibit improved radiation tolerance.)
The work carried out during this reporting period has provided a sound base for thisprojec t.Successful experiments include the preparation of device worthy substrates
> in variously patterned wafers, identification of the microstructure of TDI wafers byjL cross-sectional transmission electron microscopy (XTEM) and the successful
S implementation of a new predictive software package (IRIS). In addition a newC. halogen lamp heated wafer holder (funded under another contract) has been
commissioned and used to prepare substrates with pre-heating and background heating.L' up tu 6800C. ei od ' ,- , ! .
"- 24. SUBJECT TERMS 15. NUMBER OF PAGES
SOI, SIMOX, TDI, Rad Hard, Silicon 16. PRICE CODE
17. SELJRITY CLASSIFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. UIITATION OF ABSTRACTOF REPORT OF THIS PAGE OF ABSTRACT
UNCLASSIFIED -UNCLASSIFIED UNCLASSIFIED
NSN 7540-01-280.SSOO Standard Form 298 (Rev. 2.89)t,,i be4 y AMUI SU. L39-18
AFOSR-88-0356
UNIVERSITY OF SURREY
DEPARTMENT OF ELECTRONIC AND ELECTRICAL ENGINEERING
TOTAL DIELECTRIC ISOLATION (TDI) OF FULLY DEPLETEDDEVICE STRUCTURES
AFOSR-88-0356 Programme Manager:P L F Hemment Dr D Eirug DaviesJuly 1990
INTERIM SCIENTIFIC REPORT
SUMMARY
SIMOX is the most mature SO technology and many laboratories in the USA,
Japan and Europe are developing radiation hard and nigh speed CMOS/SIMOX
circuits. However, circuits fabricated on conventional SIMOX substrates still
require additional processing (eg MESA etching or LOCOS) to achieve lateral
device isolation.
The innovative programme of work undertaken at the University of Surrey is
designed to take SIMOX technology a stage further by achieving both vertical and
lateral isolation during a single 0+ ion implantation stage. The process being
developed, known as Total Dielectric Isolation (TDI), achieves this goal by
implanting into a patterned wafer and thereby synthesising a non-planar buried
layer of Si0 2 , which completely isolates the silicon island from the underlying
substrate. The technique has already been demonstrated and dielectr>,ally
isolated islands have been formed.
The principal achievements during the current period are (i) results of
microstructural analysis of TDI structures by ross-section transmission
electron microscopy (XTEM), (ii) completion of a computer model which permits
the dimensions (tsi, TSi 0 ) to be predicted in both "as implanted" and annealed2
substrates, and (iii) first results of experiments to planarise the top (free)
silicon surface).
The project has benefitted from close collaborative links with staff at
NRL, Washington DC, and IBIS Technology Corp, Danvers, MA.
1
1. REPORT OF WORK COMPLETED
A novel isolation technique (total dielectric isolation - TDI) to replace
MESA etching and LOCOS isolation has been developed at the University of Surrey,
which has twelve years experience in SIMOX technology. The TDI process is an
extension of SIMOX technology and entails 0+ implantation through a patterned
masking layer (SiO 2 ). As the ions lose energy within the mask a non-planar
layer of Si0 2 may be synthesised which, by careful optimisation of the
processing parameters, provides lateral and vertical isolation of .he silicon
device islands. The dimensions of the silicon islands are freely defined across
the wafer by lithography from pm to mm. The islands are delineated by Si/Si0 2
interfaces which are formed by internal oxidation and have a topography quite
different from previous structures and, thus, offer the prospect of forming
novel structures and achieving improved radiation tolerance, which currently is
necessary to meet system requirements. The dielectric isolation will also
ensure circuit operation in high temperature environments, as reported for
conventional SOI/CMOS.
The aim of the DOD programme is to prepare, evaluate and optimise TDI
substrates for circuits with improved radiation hardness and substrates have
been prepared and sent to NRL, Washington DC. The grant supports an experienced
research fellow (Dr U Bussmann) and at the start of the project he was able to
quickly define structures for the first round of experiments. Masked wafers
were prepared by NRL, implanted at Surrey and returned to Washington for
processing. In parallel the research fellow has developed techniques and
expertise to enable him to optimise the TDI structures. This work has entailed
the use of cross-sectional TEM to determine the lateral dimensions and
microstructure of the silicon islands and interfaces, surface analysis
techniques for composition and thickness determindalons and development of a
computer model (IRIS) to simulate the evolution of the oxide layer. The ideal
TDI wafer will have a planar surface and this model, which incorporates r
swelling, sputtering and diffusion, has been used to determine the optimum
values of the processing parameters (energy, dose, mask thickness and etch back) 0
to achieve this end.
The accomplishments and research progress during this period fall under
. ilty Codes
Availado
2 Melt Special
the following headings.
1.1 Preoaration of Substrates
WaLirs were received from NRL with patterned masking layers and were
implanted over a range of O+ doses at 200 keV and substrate temperature of
550*C. These were used for materials science experiments. Further substrates
with FET test patterns etched in the poly-silicon masking layer have been
prepared and returned to NRL for annealing and fabrication of test devices.
Additional wafers with a OCOS patterned masking layer have been implanted
and studied in order to optimise the processing to achieve a planar surface
after 0 + implantation.
1.2 Heated Samole Holder
Extensive testing of a new heated sample holder (funded under a separate
contract) have been successfully completed and the facility is now available for
implants in the temperature range 300*C to 700'C into 3" and 6" wafers.
The temperature has been calibrated and the facility is now used routinely
to prepare substrates.
1.3 Materials Science Exoeriments
The microstructure of both "as implanted" and high temperature annealed
samples have been studied by cross-sectional transmission electron microscopy
(XTEM). A wide range of experimental conditions have beer explored including
implantations into thick and thin masking layers and the use of poly-silicon and
oxide masks. The output from this work has been used to optimise the processing
of wafers for test devices and as input to the computer simulation software.
1.4 Comouter Simulations
The computer programme, IRIS, developed as part of this project, has been
further extended to model the effect of high temperature annealing when the
evolved oxygen depth profile is transformed into a rectangular distribution.
This software is now running and has been used extensively to accurately predict
the structure of device grade substrates.
3
2. FUTURE WORK
Priorities during the next period will be;
(i) descript-,on of the microstructure of TDI structures formed on LOCOS masked
wafers,
(ii) optimisation of the processing conditions,
(iii) direct formation of thin film SIMOX/TDI substrates using lower energy O+
ions and the heated sample holder, and
(iv) refinement of the computer model (IRIS) to extend the working range.
4
PUBLICATIONS
1. U Bussmann and P L F Hemment, "Computer Simulation of High Fluence Oxygen
Profiles in Silicon", Nucl Inst & Meths B47, (1990), 22-28.
2. U Bussmann and P L F Hemment, "Layer Thickness Calculations for SOI
Structures Formed by Oxygen Implantation", Accepted for publication in
Appi Phys Letts.
3 U Bussmann, "Computer Modelling of Oxygen Distributions in SIMOX
Substrates", Proc IV Inter Symp on SOI Technology and Devices,
Ed: Dennis N Schmidt, Electrochem Soc Proc, Vol 90-6, 85, 1990.
4. U Bussmann, A K Robinson and P L F Hemment, "Energy and Dose Dependence of
Si Top Layer and Buried Oxide Layer Thicknesses in SIMOX Substrates".
Accepted by IIT'90, University of Surrey, July 1990.
5. U Bussmann, P L F Hemment, A K Robinson and V Starkov, "Oxygen
Implantation Through Patterned Masks: A MvLhod for Forming Insulating
Silicon Device Islands Maintaining a Planar Surface", Accepted by IIT'90,
University of Surrey, July 1990.
APPENDIX
- Reference 1 (above).
- Progress Reports 1, 2 and 3.
RIPPOAT IAFOSR-88-0356
UNIVERSITY OF SURREY
DEPARTMENT OF ELECTRONIC AND ELECTRICAL ENGINEERING
Total Dielectric Isolation (TDI) of Fully Depleted Device Structures
P L F Hemment24th April 1989
PROGRESS REPORT
1. OBJECTIVE
The aim of this programme is to prepare, evaluate and optimise TDIstructures as quickly as possible in order to develop the technology forthe preparation of substrates which exhibit improved radiation tolerance.
2. TECHNICAL PROGRESS
The Research Fellow (Dr U Bussmann), supported by this contract, wasappointed on time and rapid progress has been made in developingexpertise and carrying out initial experiments. Technical progress willbe reported under separate headings.
2.1 Computer Modelling
It is important to have predictive models to describe the movement of thefree surface and buried interfaces during 0+ ion implantation. Softwarehas been written which enables these movements to be followed in both the"window" and "field" regions of the structures, which are characterisedby different sputtering rates.
The model uses an iterative procedure to simulate the evolution of theimplanted oxygen depth profile. Input data is in the form of theconcentration depth profile, either experimental or calculated usingTRIM, of incremental doses of 0+ ions and the redistribution due todiffusion (if x >2), swelling and sputtering is calculated after theaddition of each increment of value, typically, I x 101 0+ cm-2 . Theiterations are continued until the required dose has accumulated.Figures 1(b), (c) and (d) are calculated depth profiles of I x 101 8 0cm 2 , 1.4 x 1018 0 cm- 2 and 1.8 x 1018 0+ cm-2 implanted into silicon atan energy of 200 keV. Figures 1(f), (g) and (h) are the analogousprofiles when the 0+ ions are implanted through a masking oxide ofthickness 3200 X. The rapid sputter erosion of the mask modifies theshape of the final distribution and increases the value of the criticaldose required to achieve a stoichiometric oxide. These profiles, whichhave been calculated using values of the system variables derived fromexperiment, show good agreement with experimental profiles reported in
the literature. Figures l(a) and 1(e) show the oxygen distributions fora low dose of I x 106 0 + cm- 2 .
AFOSR-88-0356
2.2 Determination of Microstructure
The Research Fellow has access to a JEOL 200CX transmission electronmicroscope and an Ion Tech Fast Atom Beam Miller for detailed cross-sectional TEM (XTEM) studies of the TDI structures. A procedure forsample preparation has been developed and proven using TDI samples usedin ezrlier experiments. The preparation of XTEM samples from the firstbatch of wafers will commence immediately after their return to Surreyafter high temperature (13000C, five hours) annealing at NRL.
The seventeen stage preparation procedure, resulting in samples of twod-fferent materials, is as follows:
Period Action Working Time(hours)
Day 1 1 Cleaving (15 - 20 pieces, 3 mm) 1.5
2 Cleaning - Acetone - Methanol + Toluene 1
3 Gluing - epoxy resin 1.5
Day 2 4 Mount between glass on stubs 1.5
5 Coarse Polishing 2
6 Fine Polish (SiC) 1
7 First Diamond Polish (1 Um) 1
8 Optical Check 0.5
9 Reverse Sample Mount, Between Cover Slips 1.5
Day 3 10 Coarse Polishing 0.5
11 Fine Polish (SiC) 1
12 Fine Polish (1 pm) 1
13 Optical Check 0.5
14 Mount First Copper Grid I
Day 4 15 Remove from Stub 1.5
16 Mount Second Copper Grid 1.5
TOTAL 18.5
Day 5 17 Commence Ion Milling (10 - 20 hours for each sample)
-- x--Select sample for analysis
AFOSR-88-0356
2.3 Implantation of Masked Wafers
A set of twelve 3" wafers has been received from the Naval ResearchLaboratory 'NRL) which have patterned polysilicon masking layers ofthickness 0.25 pm and 1.5 pm. Areas are defined which are suitable foranalysis by Rutherford backscattering, SEM and TEM. The first batch offour wafers have been implanted at -5500C with 200 keV 0+ ions over thedose range 0.3 x 1010 0+ cm- 2 to 2.4 x 10WO o+ cm-2 . Two of these
planted wafers have been sent to the NRL for capping and hightemperature annealing and will be returned to Surrey for analysis.
2.4 Materials Science
A novel feature of the patterned TDI structures is that the formation ofthe synthesised dielectric layer can be dominated by different mechanismsin the window and field regions. If the sputter rate is high (eg Si0 2masking layer) then the final oxygen depth distribution will be a"sputter limited profile" which may not achieve an adequately high volumeconcentration to directly form the stoichiometric silicon dioxide. It isimportant, therefore, to determine the dependence of the volumeconcentration upon sputter rate as this information is input data for thecomputer model (Section 2.1).
Figures 3(a) and (b) are the backscatter energy spectra of 1.5 MeV He +
from a SIMOX sample (1.8 x 1018 0+ cm- 2 ) and a masked sample (3200 4 ofSi0 2 ) implanted with a dose of 1.7 x 1018 0+ cm 2 , respectively. Theoxygen depth distributions and the composition of the buried oxide (SiOx)can be inferred from the yield deficiency between channels 150 and 250,which is due to the included oxygen. These preliminary results suggestthat the oxide formed beneath the mask (Figure 3(b)) is sub-stoichiometric, which is consistent with previous SIMS analyses. Furtheranalyses will be carried out to improve the statistical significance ofthe results.
2.5 Heated Sample Holder
This TDI proect draws upon the expertise of staff at Surrey and requircsaccess to the ion implantation facilities. New equipment, which hasrecently been ordered and will be used by Dr Bussmann, includes a lampheated sample holder. This unit will enable the wafer to be pre-heatedto 5000C - 6000C and will provide background heating to maintain thewafer at a constant temperature during the three hours to five hours ofimplantation time. Dr Bussmann has participated in design and technicaldiscussions with the manufacturer.
3. FUTURE WORKPLAN (APRIL - JUNE 1989)
The initial implantations have been carried out on time and during thenext reporting period (April 1989 - June 1989) it will be possible toevaluate the TDI structures. This will entail the following activities.
I6AFOSR-88-0356
FIGURE CAPTIONS
Figure 1: Calculated depth profiles of 1 x 10'a 0 + cm- , 1.4 x 1018 0+
cm-2 and 1.8 x 1018 0+ implanted at 200 keV into bulk silicon[(D), (c) and (d)] and into silicon through an S'0 2 mask ofthickness 3200 A [(f), (g) and (h)], respectively. Figures 1(a)and 1(e) show the low dose depth profiles.
Figure 2: Backscatter energy spectra of 1.5 MeV He+ ions from (a) a SIMOXsample (1.8 x 1018 0' cm- 2 , 200 keV) and (b) a masked sample(3200 X, SiO2 ) implanted with a dose of 1.7 x 1018 0 + c M - 2 ,respectively. The residual oxide (-500 A) on the surface ofthe masked sample was removed before carrying out theanalysis.
0 MASK K
-(a) 0C- L e)
0 Ole 0
0 200 400 600 800 1000 0 200 400 600 600 1000
2- 2 --
MASK
1(b) M f
0- 0
0 200 400 600 800 1000 0 200 400 600 800 1g
2- 2-
0C 0()
0 0
o 200 400 600 800 1000 0 200 40C 600 600 1000
2 2
(d) (h)
1 d) I
Ch
0 101 1
0 200 400 600 800 1000 0 200 400 600 800 1000
depth (nm) FIGURE I depth (nm)
4600
* FIGURE 2(b)!N3450 ---
.'- .''. .
1150"
50 11,, 174 236 300
Channel Number
a 0 0 C 0 0 1 0 0 0 0 2 0 0 0 0 0 0S 0 0 a a 0 0 0 0 0 0 0 0 0 0 a
42l 0'l 0'l 0'l I'l 2252 022 42 * 40 4022 222 4245 2026 4042 4063 2IM 2223 2400 2024 2240 .521 I2ll 2037 2622 I7.0 246 2 1 232 2'24 2575 2232 2425 2M2 4 M. )$2 53 225 725 2460 3222 253230 255 2352 2222 255l 2222 I'll 2620 25l2 2622 2623 2427 257 26 761 52 23 55 25 52 25
200 2572 M 2232 2 327 203 2453 2223 240 23. I742 a., 2353 202 226 2I22 2323 2323 232 122 2 2020 22'l222" "04 , 20 I", 235 2 22 20 2 24*2 2232 2623 2525 242' 2544 262t 2522 2226220 2250 22*2 222 221 2232 2222 222* 2232 220 227 ' '' 2 1'2a2 2253 M34 1.11 2234 2222 222 1220 17.2 23222(. 3 I744 270 263. 1222 35, 43 132 122 3203 2722 1 242 2022 2222 3051 25 1 "1 4 20
.0 220", 36. 1. I'll, 23 22 223. 42.3 'l2 30 342 *3 ,2 424 , 2 ,2 . .2.,,,, , * 4
222 .2*5 3256 27 22 ,2 5 :0. , , .2 2 24 4 522 352 7 1 52 3523
I11 141 . 1 i I I I .
230 314 532 222 12 265 324 22 4 56 4 20 23 2 1 2 M 0 l0
222 2 2 0 2 0 3 2 2 3 7 2 0 2 17 2 12 2 2222 3 0 2 7 1 13 3 2 0 1 1 1 1213 3 1 3
260 2 7 31 2 31 7 2 3. I7 12 0 0 7 3 ' ' ') 3 0 3 3 6 3* I 7 0 0 0 a 2 1 0 0 0 2 a
430 l 0 7 294 21) 15 7 5 101 714 012 715 711 7It 0~l 0I1 0 21 034, *€5 7I 2
240 7 7 0 2 3 4 2 a 0 2 7 3 0 2 0 2 7 0 4
46 o o a a 7 2 0 0 0 4 3 2 a 0 0 2 7 02*0 52 7 7 0 3 5 7 lot 1 14 1 2 0 21 2 5 3 2* 5 1 515 2 11 . M4 5
573 532 23'62 526I 26l 32 1 1 4 13I 110 Il' Il1 I4 tl 2 7 4| 7 79 |1 JI 11 11
University of Surreyb/Department of Electronic and Electrical Engineering IT ? Z
Solid State Devices and Ion Beam Technology 0' fIon Beam Analysis and Implantation C
x25 lalSokoidMIon Apr 17 12:45:31 1989
Lchargo 10.0470 uG Chopper Count.i 138657LTirne 1075.8 secs RTime 1091.0 5005I
4092
FIGURE 2(a)
3069
C
0~t '' NN
2046Qz~~ >''~~4''' ~N &
\ ~"~"V.4~4n~~rx~vV 'a '
0 I 0150 14 18 24 30
XChannel Numbe
50 1114 3 0 7 * 0 1 27382 2424 71 306
20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
20 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 040 0 0 0 3)0 103 3071 7707 7727 307 7047 7044 4007 7077 304 4007 3044 307 2 7633 744 33
40 34 7023 760) 7 327 3*74 33 3 36 3 37 370 3471 3372 3716 3547 7404 7144 1 0 759 t370 507 743 370 7460
60 7060 3354 3525 M 32572 3270 3702 7247 3 270 323 4 727)4 31 41 3 27 7 207 3 2717 3344 0 32 124 3247 330400 7277 7171 3407 3720 7431 1 317 3302 I402 100; 3426 104 3574 310 71:2 73 747l 34710 3514, 7141 I5l3120 34 750 333 700 30 3742 3724 32 317 332 3327 3265 3.1 3766 3244 7 2 724 300 10 104
04 22 77 23 35 04 242 2040 3025 3041 704 7046 240 7221 2362 27345 2741 2711 2772 3443 746140 2754 " 260 2 144 221 2021 Il 244 24 5 1 2324 331 2303 211$1 2443 2510 2443 2747 2272 227. 1 222 21 27
14000 M 13 104 762 1620 2 716 2403)1 14140 227 120 00 120 1070 13 0 7 3 4 6 0200 014 034 076 03 024 022 654 464 £2 113 00 913 4 6 63 02 042 074 443 605 047
220 66 64 06 02 4 0 3 65 6 64 100 672 671 60 3 7 6 2 6 6 1 0 46 02260 702 2.061 00 14 1006 122 114 2247 122 1265, 2344 23310 2370 2360 1#743 270 1410 2400 2370 2405
140 1 431262 130 0470 1465 2442 1400 1447 1410 2472 1207 1271 1202 464 4461 443 262 242 120, 10660 70 40 30 43 34 25 22 12 10 6 I $ 4 I I 4I
7 00 2 2 3 1 2 7 3 3 1 3 3 3 3 3 3 1 3 3 41I2 4 1 0 0 1 1 2 1 0 0 2 1 4 0 I 1 0 I
360 0 1 1 2 0 1 0 3 2 1 0 2 0 0 0 0 0 0 I 1
30 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 2 2 2
340 0 1 1 0 1 1 1 0 0 0 4 0 0 0 0000 0 0 0 0 1 0 I 0 0 1 0 0 0 1 0 0 0 0 1420 0 0 2 1 0 0 0 1 1 0 0 2 1 0 0 0
44 0 0 0 0 0 020 0 0 0 0 0 044 2 0 2 0 6 0 0 0 0 0 0 0 0 1 0
460 0 0 0 0 0 01 0 0 2 27) 334 3706 I230737 10012 2413 537 7'N 30 3 2 4 1 0 4 3 2 7 2 3 3
2AFOSR-88-0356
UNIVERSITY OF SURREY
DEPARTMENT OF ELECTRONIC AND ELECTRICAL ENGINEERING
Total Dielectric Isolation (TDI) of Fully Depleted Device Structures
P L F HemmentJuly 1989
SECOND PROGRESS REPORT
ABSTRACT
The initial experiments described in the First Progress Report haveprovided a sound base for this project and further good progress has beenmade during this second reporting period in the areas of preparation of
substrates, cross-sectional TEM analysis, computer modelling andmaterials science experiments. Delivery from the manufacturers of theheated sample holder is a few weeks behind schedule due to late deliveryof components from a sub-contractor.
1. OBJECTIVE
The aim of this programme is to prepare, evaluate and optimise TDIstructures as quickly as possible in order to develop the technology for
the preparation of substrates for circuits which exhibit improvedradiation tolerance. TDI structures incorporating N+ implantation will
additionally be examined.
2. TECHNICAL PROGRESS (April 1989 - June 1989)
2.1 Preparation of Substrates
A further two patterned 3" (100) silicon substrates have been success-fully implanted. These have TDI FET test patterns etched in thepoly-silicon masks of 0.25 Vm (transparent to 200 keV O+ ions) and 1.5 Pm(opaque) thicknesses. These substrates have been returned to NRL,Washington, for capping, annealing and fabrication of test devices.
New materials science experiments have been started with the aim ofevaluating TDI structures formed by the synthesis of a Si 3N4 layer. Thisis supported oy 200 keV N+ implants into Si0 2 masked wafers. Thepatterned oxide layers are 0.31 lim and 0.47 um thick and the implanteddoses are 0.4 x 1011 N+ cM 2 , 1.0 x 101 6 N' cm- 2 and 1.4 x 1018 N+ cm-2.
2.2 Microstructure of "as Implanted TDI Structures"
Cross-sectional samples prepared from an "as implanted" (no postimplantation anneal) wafer (NRL Wafer Number 5) have been examined in aJEOL 200CX TEM. This wafer has a patterned poly-silicon mask of nominalthickness 2500 X above a Si3N, buffer layer of thickness 320 . These
were deposited on a screen oxide of nominal thickness 210 .
3AFOSR-88-0356
Figure I shows a schematic of the actual structure together with across-sectional micrograph from which the layer thicknesses weredetermined. The edge of the mask is non-ideal having a slope of .450 inthe lower part but is steeper near the top surface. The Si3N . has beenincompletely etched and extends about 0.45 pm into the window region.
An unannealed TDI structure formed by a dose of 1.8 x 101a 0/cm 2 is shownin Figure 2, which includes a schematic of a typical structure and TEMmicrographs from representative areas. Surprisingly, within the windowregion the silicon surface is =500 a below the plane of the originalsurface which is still preserved and can be seen under the field mask.Within the field region the structure consists of 1400 X of polysilicon,about 600 X of buffer layer, 1400 X of defective but crystalline silicon,2900 X of amorphous oxide and 2000 1 of highly defective silicon abovethe silicon substrate. As a negligible amount of oxide is deposited inthe polysilicon the change in thickness is due solely to sputtering whichgives a sputtering coefficient for 200 keV 0+ ions at 550 0 C of 0.22atoms/0 atom. Within the window region the structure consists of 3500 aof defective single crystal silicon, 2900 A of amorphous oxide, about2000 X of damage silicon above the silicon substrate. As predicted theamorphous oxide layer has the same thickness (2900 a) in both the fieldand window regions.
The edge of the window shows a smooth transition between the geometriesof the ficld and window structures. It is noted that sputtering of t!'rmask has reduced the slope of the mask edge. Finger like protrusions ofdefective silicon penetrate into the synthesised layer of SiO 2. Asilicon rich region has been reported previously ( ) in wafers with anoxide mask and this phenomena is associated with the lateral movement ofthe window edge due to sputtering. The effect is less pronounced in thispolysilicon masked wafer and it is anticipated that oxygen segregation
during high temperature annealing will achieve good dielectric isolationbetween the silicon device island and substrate.
Conclusions which may be drawn from these experiments include:
(a) desirability of a thicker polysilicon mask (3000 A) to achieve afield oxide extending up to the silicon surface,
(b) sputter coefficient for 200 keV 0+ in polysilicon (5500C) is 0.22atoms per ion (the measured rate for SiO2 is 1.1 atom per ion withtheoretical (TRIM) values of 0.18 dnd 0.38, for Si and Si0 2 ,respectively),
(c) the improved e,,ge structure may be due to the lower sputtering rateof polysilicon compared to Si0 2,
(d) it is desirable to have vertical mask edges, and
AFOSR-88-0356
(e) the Si N,/SiO 2 buffer layer maintains its integrity during 0+
implantation.
Additional specimens implanted with a higher dose (2.4 x 1018 0 cm- 2 )
have been thinned and are r -dy for examination in the TEM.
2.3 Materials Science Experiments
2.3.1 Stoichiometry
Previous analyses by SIMS( ') of oxides formed beneath an oxide mask have
suggested that the layer is not stoichiometric SiO 2 but rather SiOx wherex = 1.7. Preliminary RBS analysis, recorded in the first report, seemed
to support the SIMS data. During this reporting period very detailed RBS
analyses have been carried out to achieve a more reliable measure of thevalue of x and to support the computer modelling (Section 2.4). It is
important to note that this new data clearly shows the oxide isstoichiometric (x = 2.00 ± 0.05) to within experimental errors. The
reason for the discrepancy between SIMS and RBS analyses is not obvious
but as ion yields are known to be influenced by the matrix it is
reasonable to accept the new RBS data described below.
Figure 3 shows 1.5 MeV He+ RBS spectra from a thick polysilicon layer(1.5 11m), a standard SIMOX sample implanted with 1.8 x 1018 0+ cm- 2 at200 keV and a TDI test structure in which 1.8 x 1018 0+ cm-2 were
implanted through an SiO 2 mask of thickness 3200 . Figure 4 shows RBSsimulation spectra for (i) bulk silicon, (ii) 3000 silicon on 3000 A
SiO 2 on silicon, and (iii) 2000 silicon on 3000 'A Si0 2 on silicon. Thedata is consistent and by taking ratios of yields it is concluded thatthe oxide in the TDI sample has a stoichiometric composition (±2.5%).
2.3.2 Implantation into Polysilicon
Figure 5 shows 1.5 MeV He+ RBS spectra from a wafer with a thinpolysilicon mask (0.25 lm) and a Si3 NSi0 2 buffer layer. The regions
are labelled in the figure. The buffer layer has been used as a shallow
buried marker and enables the sputtering coefficient of 200 keV 0+ into
polysilicon to be estimated (0.22 atoms per ion).
Figure 6 confirms the formation of a buried layer of SiO2 in the thickpolysilicon film (1.5 pm) after implantation of 1.8 x 108 0
+ cm1 .
2.4 Computer Modelling
Further refinements have been made to the computer model and simulations
have been run to determine the sensitivity of the structure to the
diffusion of excess oxygen in the buried oxide.
Figure 7 shows the evolution of the oxygen depth profile in bulk silicon
over the dose range 0.6 x 1018 0+ cm- 2 to 2.6 x 1018 0+ cm- . The
sputter rate is 0.2 and excess oxygen is allowed to move to the upper(near surface) Si/SiO2 interface. As expected this interface advances
towards the surface and rapidly becomes more abrupt.
AFOSR-88-0356
Figure 8(a), (b) and (c) show the evolution of the oxide profiles insamples with an oxide mask of thickness 3200 A. The sputtering
coefficient is 1.1 atoms per ion. Figures 8(a), (b) and (c) showprofiles when excess oxygen diffuses to the lower interface, distributesequally and diffuses to the upper interface, respectively.
The following conclusions may be drawn:
(a) in SIMOX structures the oxygen peak moves away from the surfacewith increasing dose,
(b) for TDI samples the oxygen peak moves towards the free (upper)surface, the effect is independent of the direction of oxygendiffusion being dominated by sputtering of the mask, ana
(c) the critical dose just to achieve a stoiclometric oxide is1.35 x 1018 0+ cm- 2 for SIMOX and 1.57 x 101 O+ cm- 2 for TDI
structures.
2.5 Heated Sample Holder
Delivery of the heated sample holder (funded under another contract) hasbeen delayed due to late delivery of components from a sub-contractor.Acceptance tests will be carried out during early August 1989 and it isplanned to implant wafers using preheating during the next reportingperiod.
3. FUTURE WORKPLAN
Priorities during the next period will be:
(a) cross-sectional TEM analysis of annealed TDI structures,
(b) implantation of additional wafers for the fabrication of CMOS teststructures at NRL,
(c) refinements to the computer model, to include thermal segregation,
(d) materials science, further measurements to quantify the effects ofswelling and sputtering.
REFERENCES
(1) P L F Hemment et al, Nucl Inst Meths B37/38, 766 - 769, 1989.
6AFOSR-88-0356
Activity July August j Sept
XTEM analysis of the first batch ofwafers annealed at Naval Research ILaboratories.
0+ ion implantation. I-I I-I
Sample preparation and XTEM analysis. I II
Computer modelling. I
Definition of new structures. i
Commission heated sample holder. I-I
Materials science experiments (eg RBS) I- I I-I
Electrical evaluation (NRL) I-I
Discussions will be held with our sponsors/collaborators to defineimproved structures and to agree upon future experiments to assess theelectrical characteristics and radiation tolerance of the structures.
P L F Hemment
July 1989
7
AFOSR-88-0356
FIGURE CAPTIONS
1. XTEM micrograph showing the edge of the thin polysilicon mask.
2. XTEM micrograph shown as-implanted structure (D = 1.8 x 1018 0+
cm - , E = 200 keV, Ti = 5500C) in mask, transition and windowregion.
3- RBS profiles after implantation (D - 1.8 x 1018 0+ cm-2 ,
E = 200 keV, Ti = 5500C) into silicon and through a 320 nm thickSiO 2 mask (remainder of oxide mask removed). The profile of thethick polysilicon mask is taken as a reference.
4. Theoretically expected RBS profiles for the layered structuresshown in the schematics.
5. RBS profiles of the thin polysilicon sample (see schematic) beforeand after implantation (D = 1.8 x 1018 0+ cm-2, E = 200 keV,Ti = 5500C).
6. RBS profiles of the thick polysilicon sample (1.5 )jm) before andafter implantation (D = 1.8 x 108 0+ cm- 2, E = 200 keV,Ti = 5500C) in comparison to an implantation into bulk silicon.
7. Calculated 8-profiles for the implantation of different doses intosilicon (E - 200 keV) with diffusion directed to upper interface.
8. Calculated e-profiles for the implantation of different dosesthrough a 320 rm thick SiO 2 mask. The diffusion is directed to(a) lower interface, (b) both interfaces, and (c) upper interface.
2l0nm poly-Si30 nm Si 3N4 , _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
25 nmSiO2 -S-Si
FIGURE I
l4Onm poly-Si
/ / / 35Onm Si
29090nm ~
FIGURE 2
Im lrtaion through* - oxide mask
thick poly-Sireference,
standardSIMOX
Ch.-IlM~~
FIGURE 3
(a) 30n Si(b) 200 nm Si300 nm Si02 300 nm SiO 2
bulk Si bulk Si
FIGURE 4
210 nm poly-Si
30 nm Si3N4
25 nm SiO 2
bulk Si
afterI / zpntaton
Ibefore
% l i / mplantatio n pl yN! At mask
(buffer)
bufferburied layeroxide
FIGURE 5
implantationinto poly-Si
°', ., .w 4-- standard
*a; STMOX poly-Si"Z reference
FIGURE 6
2-
o2.6 x 10'a 0'cm 2
1.
0-.
0 200 400 600 800
depth (nm)
FIGURE 7
(a)2 T.
26 101 e*ocm2218
06
0
0 200 400 600 800
22
0 1 81
1~ 14
1 0
/ 06
0-
0 200 400 600 800
FIGURE 8
AFOSR-88-0356
UNIVERSITY OF SURREY
DEPARTMENT OF ELECTRONIC AND ELECTRICAL ENGINEERING
Total Dielectric Isolation (TDI) of Fully Depleted Device Structures
U Bussmann
June 1990
PROGRESS REPORT
1. REPORT FOR THE LAST PERIOD
Extensive testing of the heated sample holder, which waspurchased under a separate contract has been carried out. 3"
to 6" wafers can now be implanted in the temperature range
200 - 700'C (calibrated). The system is now in routine use for
substrate preparation.
1.1 Heated Samole Holder
- Extensive testing: ok.- Temperature calibration.
- Now in routine use for substrate preparation.
1.2 Substrate Preparation
A new set of twenty-one masked and patterned substrates has
been fabricated by the NRL for materials science experiments:LOCOS oxide masks of 1920, 2565 and 3760 A; SiO, masks of 2370,3685 and 4389 A; poly-Si masks of 2450 A.
When the oxide mask is formed i- a LOCOS process 55% of thismasking layer is located above and 45% below the silicon
surface level. As shown by simulations, a planar surface canthen be achieved after implantation (see schematic inFigure 1). Also, the transition between the buried oxidelayers in the mask and window regions will be smoother.Therefore, a structural improvement of the critical transitionregion is expected.
A different approach uses relatively low implantation energiesto avoid steep steps between the oxides in the transitionregion. The lower flank of the implantation profile itself inthe mask region as well as in the window roqaie-n will hPsteeper due to th Ou2 Lraggle of the ion distribution.This i4aL iead to the absence of silicon inclusions at thelower interface, which are always found under standardimplantation conditions. The necessary wafer temperature cannow be maintained by the background heating using the newsample holder. An additional advantage is the formation of athinner silicon overlayer for fully depleted device areas.
2
AFOSR-88-0356
A wafer with a LOCOS oxide pattern of 3760 A has beenimplanted with 0.4, 1.8 and 2.2 x 10" 0* cm-' at the standardenergy of 200 keV/atom at the elevated temperature of 700'C,with background heating from an array of halogen lamps.
Wafers with a 2370 A SiO2 pattern and a 2450 A SiO. patternwere implanted at a relatively low energy of 100 keY/atom with0.3, 1.2 and 1.5 x 10" O cm-.
The wafers were sent to the NRL for capping and annealing.
1.3 Materials Science Results
Microstructure of annealed samples (poly-Si masks)
- "thin" poly-Si mask(energy 200 keY, annealing 1310°C for five hours)
Polycrystalline silicon masks have been successfully used toform continuous non-planar buried oxide layers. As an exampleFigure 2 shows the result of a 1.8 x 10" O cm-2 implantationthrough a 250 nm mask. As expected, after annealingdislocations are found in the silicon top layer and siliconinclusions in the oxide close to the lower interface. Inaddition a vertical chain of inclusions appears in thetransition region. These inclusions might be avoided bydecreasing the slope of the mask edge. A thin superficialsilicon overlayer is left underneath the mask, which can beused as an etch stop when removing the remainder of themasking material.
- "thick" poly-Si mask(energy 200 keY, annealing 1310'C for five hours)
Figure 3 shows the transition region of a sample implantedwith 1.8 x 10" O cm-2 after annealing. Silicon islands areformed, which are disconnectel from the bulk material,although many silicon inclusions still exist in the transitionregion. In the mask region all the oxygen is stopped in theupper half of the thick poly-Si layer forming a continuousoxide. After removal of the thick poly-Si (including its"buried" oxide layer) such silicon islands could be used forbuilding devices.
- LOCOS oxide mask(energy 200 keY, annealing 13000C for four hours, oxide cap)
A wafer with a LOCOS oxide mask was obtained, underalternative funding, for a test implantation. Figure 4 showsfirst results after implantation of 1.8 x 10" 0* cm-',
subsequent capping and annealing. Compared to the previousimplantations using poly-Si masks less Si-inclusions are foundon that "vertical chain". However, the edge of the mask regioncontaii.a hi-h number of inclusions, although probably thesewould not be detrimental oiL zhu pdrforn . .in the silicon islands. Apart from the surface dip in thetransition region itself a nearly planar surface has been
3
AFOSR-88-0356
achieved. A more detailed study of the resulting structures ison the way using masking patterns designed and prepared incollaboration with the NRL.
1.4 Commuter Simulation
The programme IRIS (U Bussmann and P L F Hemment, Nucl Inst &Meths B47, 22, 1990) has now been further developed to modelthe effect of very high temperature (>1300°C) annealing whenthe implantation profile is transformed into a rectangularprofile. It has been shown, that oxygen transport occurs upthe macroscopic concentration gradient as the system moves toreduce the total energy through the growth of large SiO,precipitates at the expense of small precipitates andeventually reaches thermal equilibrium with the segregation ofall of the implanted oxygen at the buried oxide layer. Thislayer, thus, may be considered as a planar precipitate ofinfinite radius. The phenomenological approach used here doesnot attempt to simulate processes on an atomic scale, whichlead to the thermal segregation. In our calculations all theimplanted oxygen (ie the total fluence minus the small amountof oxygen, which is deposited close to the surface andsubsequently lost by surface erosion) is bound within astoichiorietric SiO2 layer. The justification for neglectingout-diffusion is that there is gocd agreement betweenpredicted and measured layer thicknesses in structures whichhave been well characterised.
Profiles have been calculated over a wide range of fluences.Figures 5 and 6 summarize the resulting layer thicknesses for200 keY and 150 keY ions. Experimental layer thicknesses, asdetermined from XTEM micrographs or SIMS profiles from theliterature, have been included for comparison. Taking intoconsideration the scatter of the experimental data, there is
good agreement between experiment and simulation.
An important irethod for producing low defect density silicontop layers involves sequential implantation and annealing. Theprogramme IRIS has been used to make predictions of layerthicknesses after multiple step implantations in comparisonwith single step implantations. The implantation of 1.8 x 10"8
0' cm-2 in three equal steps will be considered as an example.Figure 7a shows the implantation profile for a fluence of0.6 x 10" 0' cm-2 together with the rectangular profile afterthermal segregation during high temperature annealing. Thislayer structure is the target for the second implantationstage with the same fluence, resulting in the distributionsshown in Figure 7b (before and after annealing), while thefinal oxygen distributions after the third implantation andannealing are shown if Figure 7c. For comnpr-' ., tneimplantation of 1.8 x 10 ' 0' em-2 single step is includedas Figure !A. Uziatj multiple implantation and annealing leadsto steeper flanks in as-implanted profiles due to the previous
thermal segregation. The simulation also shows that the buriedoxide layer is shifted 35 nm towards the surface. This shiftis due to the early existence of a stoichiometric layer, whichleads to an increase of the amount of oxygen being
4
AFOSR-88-0356
redistributed by diffusion.
2. FUTURE WORKPLAN
Priorities during the next period will be:
(a) Cross-sectional TEM analysis of TDI structures formed byimplants through LOCOS oxide masks.
(b) Fabrication of additi)nal wafers for the optimisation ofthe structural properties.
(c) Lower energy (100 keY) implants using the heated sampleholder.
(d) Refinements of the computer model to extend the energyrange to 50 - 300 keY (especially important for thesimulations of the planned lower energy implantations).
3. FIGURE CAPTIONS
Figure 1: LOCOS oxide mask (schematic): Expected structure ofthe periphery of the window before and afterimplantation and annealing.
Figure 2: "Thin" poly-Si mask: Resulting structure of thetransition region after implantation (1.8 x 10" 0*cm-') and annealing (1310'C for five hours).
Figure 3: "Thick" poly-Si mask: Resulting structure of thetransition region after implantation (1.8 x 10" 0*cm-2 ) and annealing (1310'C for five hours).
Figure 4: LOCOS oxide mask: Resulting structure of thetransition region after implantation (1.8 x 108 Ocm-2 ) and annealing (13000C for four hours).
Figure 5: Calculated depth of the maximum oxygen concentration(Rr) and interface positions for 200 keYimplantations into silicon in comparison toexperimental data from the literature (symbols).
Figure 6: Calculated depth of the maximum oxygen concentration(R ) and interface positions for 150 keYimplantations into silicon in comparison toexneri-mntal data from the literature (zcyri-1
Figure 7: Calculated oxygen concentration profiles afterimplantation (histograms) and high temperatureannealing (rectangular profiles):
(a) 0.6 x 10" 0* cm-'(b) 2 x (0.6 x 10" O cm-2 )(c) 3 x (0.6 x 10" 0* cm-')(d) 1.8 x 10" O' cm", single implantation stage
5
original \ \\ \ ",surface
'2" .. " new surface
Si ", ,\ . \Si02, .. ,:, SiO' ,
before implantation after implantation
Figure I
poly-Si mask
Si top layer
buried oxide
Figure 2
thick poly-Si maskwith implanted oxid(
Si island
buried oxide ~
Figure 3
capping oxide capping oxide
+ LOCOS oxideSi island '+ implanted oxide
Figure 4
700-A1el
4200 Ae
0 I I I
0 0.5 1.0 1.5 20 .
LFigu0e
600
500-
-300-cu
200-
100-
0 0.5 1.0 1.5 2.0 2.5FIluence (10 cm
Figure 6
Z.)
" ' annealed
o3-cq
as-inplanted
000
0
b).- 6
0
0
0
I
0
0
c) 6-
0
I= 3-0
0 200 400 600 80
dept (m Fiur 7
• 3 -
000
00- I
d)det61nm)Figrem