`
`• • • •• •••• •••• • •••••• ••••• • • •• • •• • ••• •• •••• •••• •••• •••• • • •••• ••• • •• ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• ••• ••••• • • ••• •• • ••• • • •• •• • ••• • • •••• ••• • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
`
`• • • • ••• • • • • •• ••• • • • ••• • ••• • • •• ••• • •• • •••• • •• •• •• • • •• • • •••• •• • • ••• • ••• • •• • •• • • • • • • • • • • • •
`• • • • • ••• • • •• • ••• ••• • • ••• • ••• • • ••• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
`
`ResearchGate
`
`• •••••• • • • • • • ••••••• • •••••• • • •• • ••• ••• • • •• •• •••• •• • • ••• ••• • ••••••••• • • • •••• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
`
`• • •••• • •• • • • •• •• • • • • • • • • • • • • • • • • • • • • • • • • • • • •
`
`• •• • • •• • • • • • • • • • • •
`• • • •
`
`
`
`• •• • •• • ••••••••••••••• ••• • •• • •••••••••••••
`
`• • • • • • • • • •
`• • • • • •
`
`• • •• • •• •• • •• • • • • • • • • • • • • • •
`• • •••• ••• • •• • ••• • • ••• • ••• •••••••••••••••••••••••••••••••••••
`
`
`• • • •••• • • • •• • • •• • • • • • • • • • • • • • •••••••• • • • • •••• •• • • •• • • • • • • • • • • •••••••
`
`•• ••• • • •• • •• •• • • • • • • • • • • • • • •
`• ••• ••••• • ••• • • •• • • ••• • • • • • • • • • • • • • • • • • • • • • •
`
`
`
`• • • • • •••• • • • •• • • •• • • • • • • • • • • • • • •••••••• •• • • • • • • •••• •• • • •• • • • • • • • • • • •••••••
`
`• • • •• • • • •• • • • • • • • • • • • •
`
`• • • •• • • • •• • • • • • • • • • • • •
`
`• •• • •• • • •• • • • • • • • • • •
`• ••• ••••• • ••• • • •• • • ••• • • • • • • • • • • • • • • • • • • • • • •
`
`
`• • • •••• • • • •• • • •• • • • • • • • • • • • • • •••••••• • • • • •••• •• • • •• • • • • • • • • • • •••••••
`
`• • • • •• •• •• ••• • • • • • • • • • • • • •
`• ••• ••••• • ••• • • •• • • ••• • • • • • • • • • • • • • • • • • • • • • •
`
`
`
`• • • • • •••• • • • •• • • •• • • • • • • • • • • • • • •••••••• •• • • • • • • •••• •• • • •• • • • • • • • • • • •••••••
`
`• • • •• • • • •• • • • • • • • • • • • •
`
`• • • •• • • • •• • • • • • • • • • • • •
`
`
`
`
`
`• ••••• • •• • •••• ••• • •• • ••• •••• • • • •• • ••• • •• • • • • •• •••••••••••••••••••••••••••••••••••••••••••••••••••• ••• • • •• • •• •• • • • • • • • • • • • • • ••• • •• • •• • • • • •••• • • • •••••••••••••••••••••
`
`• • • •• •• ••• • •••• • • • ••• • •• • • • • •• • • • ••• •••• • •• • • • •• • • • • ••••• •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••
`
`REC Exhibit 1030, Page 1 of 7
`REC Exhibit 1030, Page 1 of 7
`
`

`

`3100
`
`Journal of The Electrochemical Society, 147 (8) 3100-3105 (2000)
`S0013-4651(99)12-095-0 CCC: $7.00 © The Electrochemical Society, Inc.
`
`Spectroscopic Investigations of Borosilicate Glass and Its Application
`as a Dopant Source for Shallow Junctions
`M. Nolan,a T. S. Perova,a,z R. A. Moore,a C. E. Beitia,b J. F. McGilp,b and H. S. Gamblec,*
`aDepartment of Electronic and Electrical Engineering, bDepartment of Physics, University of Dublin, Trinity College,
`Dublin 2, Ireland
`cDepartment of Electronic Engineering, The Queen’s University of Belfast, Belfast, Northern Ireland
`
`Borosilicate glass was investigated as a dopant source for proximity rapid thermal diffusion. A borosilicate gel was spun onto a sil-
`icon wafer and the layer was rapid thermally processed to convert it to a borosilicate glass. Fourier transform infrared spectroscopy,
`spectroscopic ellipsometry, and sheet resistance measurements were used to understand and subsequently optimise the conversion
`of the gel to a borosilicate glass. The optimum conversion step, which avoided any boron loss from the borosilicate glass layer, was
`a curing step of 900⬚C for 45 s. Secondary ion mass spectrometry was used to measure the boron dopant profile of a silicon wafer
`that was doped with the borosilicate glass layer. The wafer had a surface dopant concentration of 4.7 ⫻ 1019 cm⫺3 and a junction
`depth of 65.5 nm. Junction diodes, which were fabricated using the glass layer as a dopant source, displayed excellent character-
`istics, with very low leakage currents and a near ideal forward slope.
`© 2000 The Electrochemical Society. S0013-4651(99)12-095-0. All rights reserved.
`
`Manuscript submitted December 12, 1999; revised manuscript received April 27, 2000.
`
`films.16-19 Detailed investigations were performed on BSG20-25 and
`borophosphosilicate (BPSG)26-34 films, obtained by low pressure
`and atmospheric pressure chemical vapor deposition (LPCVD and
`APCVD). However, as was noted by Becker et al.,28 the properties
`of these films strongly depends on the preparation technique. This
`paper presents the detailed FTIR and ellipsometric analysis of BSG
`films spin-coated on silicon wafers. As deposited, the SOD layer had
`a thickness of 120 nm. This layer is very thin in comparison to the
`BSG layers that were used in previous works. It should also be noted
`that there is no thin capping silicon dioxide layer on the surface of
`the BSG films studied here, since the concentration of boron is not
`high and the films do not absorb moisture on exposure to the atmos-
`phere after the curing process. The measurements were repeated six
`months after the initial measurements were made, and there were no
`changes in the FTIR spectra recorded.
`Experimental
`Proximity rapid thermal diffusion.—Czochralski grown 4 in.
`n-type, <100> oriented, 9-15 ⍀ cm resistivity silicon wafers were
`
`The integrated circuit industry’s continuous drive toward smaller
`and faster devices for the next generation of ultralarge-scale inte-
`grated (ULSI) circuitry puts critical demands on vertical scaling.
`Source/drain junctions with depths <70 nm in 0.18 ␮m technology1
`are required in metal oxide semiconductor field-effect transistors
`(MOSFETs) to reduce short channel effects. The greatest challenge
`arises in the fabrication of boron doped p⫹-n junctions. At present,
`ULSI and very large-scale integrated (VLSI) silicon technologies
`depend primarily on ion implantation for doping. However, there are
`limitations on this technique in the formation of shallow boron junc-
`tions.2-4 Ion implantation generates defects in silicon, and these
`defects must be annealed out at high temperatures after the implant.
`The high diffusivity of boron in silicon5 and transient-enhanced dif-
`fusion of channeling tails during the thermal anneal, makes control
`of shallow junction depths difficult.
`In this paper, proximity rapid thermal diffusion (RTD) is used as
`a technique for fabricating shallow boron junctions.6 A spin-on
`dopant (SOD) deposited onto a silicon wafer was used as a planar
`dopant source during RTD. The wafer configuration during proxim-
`ity RTD is shown in Fig. 1. This technique is very suitable for shal-
`low junction formation for several reasons: (i) the dopant source is
`quick and easy to prepare, (ii) there are no defects introduced into
`the wafers during the diffusion process, and (iii) the dopant diffusion
`is minimized since the wafers are heated to high temperatures for
`short times.
`The SOD is in the form of a borosilicate gel, and is commercial-
`ly available from Filmtronics, USA. On heating, the composition
`and structure of the SOD gel changes to that of a borosilicate glass
`(BSG). The boron supply from the dopant source changes as the
`structure of the SOD gel changes.7 Therefore, in order to use one
`dopant source to dope several wafers repeatably, it is important that
`the borosilicate gel is initially converted into a stable BSG dopant
`source. Fourier transform infrared (FTIR) spectroscopy and spectro-
`scopic ellipsometry are used to determine, and subsequently opti-
`mise, the conversion of the gel to BSG.
`In the past decade, a number of investigations have been per-
`formed to understand how different thermal treatments influence the
`properties of SOD and spin-on glass (SOG) films. Information on
`different properties, such as doping efficiency, density, refractive
`index, and thickness have been found for phosphosilicate glass
`(PSG) films in Ref. 8-12 and for SOG films in Ref. 13-15. Howev-
`er, very few studies have been done on spin-on dopant BSG
`
`* Electrochemical Society Active Member.
`z E-mail: perovat@tcd.ie
`
`Figure 1. Wafer configuration during proximity RTD.
`
`Downloaded 11 Feb 2010 to 134.226.1.229. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
`
`REC Exhibit 1030, Page 2 of 7
`
`

`

`Journal of The Electrochemical Society, 147 (8) 3100-3105 (2000)
`S0013-4651(99)12-095-0 CCC: $7.00 © The Electrochemical Society, Inc.
`
`3101
`
`used throughout this study. As-received wafers were cleaned using
`H2SO4:H2O2 followed by a HF dip. SOD gel was spun onto the
`wafers at 6000 rpm for 15 s. The SOD contains ethanol as a solvent.
`Therefore, following application of the SOD, it was necessary to
`bake the wafers at 200⬚C to evaporate moisture and light organics
`from the dopant layer.6,7 The rapid evaporation of the solvent during
`film deposition (through the spinning) leads to the formation of a
`porous gel film. The film structure in this case13 strongly depends on
`the molecular weight distribution of the oligomers present in the
`solution. Thus, even after baking at 200⬚C, silanol (Si-OH), molecu-
`lar water, and ethanol remain in the gel network to some extent.
`Fully dense films are only produced after heat-treatment at tempera-
`tures between 800 and 1200⬚C.13,32
`Dopant source wafers were cured in a Sitesa rapid thermal
`processor (RTP) at different temperatures in the range 800-1000⬚C
`for 2-60 s (Table I), to convert the SOD gel to a BSG layer prior to
`RTD. FTIR and spectroscopic ellipsometry were used to determine,
`and subsequently optimize, the conversion of the gel to a BSG layer.
`Following the stabilization of the dopant source, the source wafer
`was then stacked in proximity to a silicon product wafer on 0.5 mm
`silicon spacers, Fig. 1. Several silicon product wafers were doped
`using the optimised dopant source. All rapid thermal heat treatments
`were performed in 25% O2:75% N2.
`FTIR Measurements.—The measurements on BSG films were
`performed in the spectral range from 4500 to 500 cm⫺1 with a Fouri-
`er transform Bio-Rad FTS60A spectrometer using a Globar source,
`a KBr beam splitter, and a mercury cadmium telluride (MCT) detec-
`tor. The spectra were collected with 8 cm⫺1 resolution, and 64 scans
`were averaged for each spectrum to improve the signal-to-noise
`ratio. A bare silicon substrate was used as a reference for all of the
`samples, and the analysis of the transmission spectra was performed
`while neglecting reflectance. A clear and detailed analysis of the dif-
`ferent films requires consideration of spectra both in the high fre-
`quency region (from 3000 to 2000 cm⫺1) and in the low frequency
`region (from 1600 to 400 cm⫺1). The spectra are the superposition
`of the spectra of the BSG and the silicon wafer and it is necessary to
`subtract the spectrum of the bare wafer from the measured spectra in
`order to obtain the spectra for the BSG. Hence, all spectra discussed
`below are difference spectra. It should be noted that the influence of
`the substrate absorption becomes especially important if heat-treat-
`ed samples are analyzed, as was shown by Becker et al.28 However,
`we believe that, due to the extremely short thermal treatment in the
`RTP reactor (the maximum time used was 60 s), there is no change
`to the substrate absorption.
`
`Table I. Rapid thermal treatment to convert SOD gel to BSG.
`
`Sample
`
`Curing temperature (⬚C)
`
`Curing time (s)
`
`11
`12
`13
`14
`15
`16
`17
`18
`19
`10
`11
`12
`13
`14
`15
`16
`17
`18
`19
`21
`
`1800
`1800
`1800
`1800
`1850
`1850
`1850
`1850
`1900
`1900
`1900
`1900
`1950
`1950
`1950
`1950
`1000
`1000
`1000
`1000
`
`10
`25
`45
`60
`10
`25
`45
`60
`10
`25
`45
`60
`15
`10
`25
`45
`12
`15
`25
`45
`
`Spectroscopic ellipsometry measurements.—The refractive index
`and the thickness of the sample were obtained by means of spectro-
`scopic ellipsometric measurements. This technique has been widely
`used for the characterization and study of the physical properties of
`thin films.35-37
`The samples were measured at room temperature with a rotating
`polarizer spectroscopic ellipsometer (SOPRA GESP5) in the visible
`to near ultraviolet wavelength range (225–880 nm). For each spec-
`trum, 250 points were measured, giving a wavelength resolution of
`3 nm. The data were collected using the current tracking mode for
`the position of the analyzer in order to improve the accuracy of the
`measurements. In order to improve the accuracy in determining the
`physical parameters of the samples, each spectrum was measured at
`two different angles of incidence (65 and 75⬚). Several different
`models were used to fit the data measured (tan ⌿ and cos ⌬) simul-
`taneously at the two angles of incidence. A suitable three-layer
`model (air, oxide layer, silicon), using a Cauchy law fit for the oxide
`layer, was chosen based on the accuracy of the results obtained (a
`detailed description of this model will be presented in a separate
`paper). This paper presents the results obtained from this model and
`compares them to results obtained from FTIR.
`Secondary ion mass spectrometry (SIMS) and sheet resistance
`measurements.—The sheet resistance of the wafers was measured
`using a Jandel four-point probe. The boron concentration profiles of
`the product wafers were measured using a CAMECA IMS 3F sec-
`ondary ion microscope in the National Microelectronics Research
`Centre (NMRC), Ireland.
`Device fabrication and characterization.—Having optimized the
`conversion of the borosilicate gel to a BSG, diodes were fabricated
`using BSG as a dopant source in proximity RTD. A four-mask process
`was used to produce p⫹-n diodes. The substrate was wet oxidized at
`1000⬚C to produce a masking oxide layer of thickness approximately
`0.4 ␮m. This oxide was patterned and a p-type diffusion from a boron
`nitride disk was carried out at 1000⬚C for 20 min, followed by a 60
`min drive-in at 1000⬚C, to form deep p-type junctions. A second mask
`was used to open the windows for the very shallow proximity RTD p-
`type regions. An LPCVD oxide was deposited from a tetraethoxy-
`orthosilane (TEOS) source at 720⬚C. Opening contact windows in this
`oxide layer, followed by aluminum metallization, completed manu-
`facturing of the device. Finally, the forward and reverse current-volt-
`age (I-V) characteristics of the p⫹-n diodes were measured.
`Results and Discussion
`FTIR.—IR spectroscopy is known to be a fast and nondestructive
`method of determining a number of important properties of dielec-
`tric films. Those properties are boron and phosphorous concentra-
`tion, absorbed water content, chemical bonds structure, thickness,
`and density.20-22,40 The ratio of the B–O band peak intensity at a fre-
`quency near 1370 cm⫺1 to the Si-O-Si absorbance maximum near
`1075 cm⫺1 is widely used to determine the boron content for glass
`films of thickness up to 2.5 ␮m.20-25
`FTIR was used to investigate the structure and properties of the
`SOD layer after the rapid thermal curing (RTC) process step. The
`wafers were cured at 800, 850, 900, 950, and 1000⬚C for 2–60 s,
`Table I. The FTIR spectra for the wafers treated at 900 and 1000⬚C
`are shown in Fig. 2 and 3. These spectra clearly show the change in
`structure with variation in temperature and time. In particular, as the
`heating time is increased, the Si-O-Si asymmetric stretching vibra-
`tion (AS) band at ⬃1075 cm⫺1 shifts to a higher frequency and the
`absorption intensity increases. This shows that the silicon dioxide
`thickness has increased and the SOD layer has densified.15,20-22
`For a given temperature, the B-O stretching vibration band in the
`region 1370-1440 cm⫺1 shifts to a lower frequency as the heating
`time increases (Fig. 2 and 3). This is a characteristic change, which
`occurs on densification of borosilicate glass.20,21 The absorption
`intensity and the contour of this band also changes significantly as
`the heating time is increased. These changes were observed for all of
`the (RTC) temperatures investigated. The intensity of the broad band
`
`Downloaded 11 Feb 2010 to 134.226.1.229. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
`
`REC Exhibit 1030, Page 3 of 7
`
`

`

`3102
`
`Journal of The Electrochemical Society, 147 (8) 3100-3105 (2000)
`S0013-4651(99)12-095-0 CCC: $7.00 © The Electrochemical Society, Inc.
`
`spectively, with increasing temperatures. The rate of loss depends on
`the RTC process temperature. Enhancement of band intensity with
`temperature is explained by incorporation of the B–OH groups38,39
`into the borosilicate phase, during which Si–O–B bonds are pro-
`gressively formed. The observed minimum in the intensity of B–O
`band at certain temperatures and times occurs during the transfor-
`mation of tetraborate to borate and then to a borosilicate glass. At the
`intermediate stage, some boron may exist in ionic forms that are not
`IR active. The minimum observed for B–O peak intensity may be
`explained by the presence of IR inactive intermediate products that
`are formed during reconstruction of the layer rather than by boron
`loss from the layer (Fig. 2 and 3). However, the shift of the B–O
`band at ⬃1370 cm⫺1 to the high-frequency side seen in Fig. 3 after
`45 s treatment at 1000⬚C is due to the boron diffusion into the under-
`lying silicon.
`The band at 910 cm⫺1 can also be used to determine when the
`BSG forms. This band belongs to the bending vibration of B–O–Si
`units, and is a characteristic band, which appears on the formation of
`BSG.20-22 The band intensity increases as the RTC time increases
`(Fig. 2, 3). Figure 3, in particular, shows the increase in intensity and
`a stabilization of the peak at 25 s.
`Analysis of the ratio of two vibrational bands B–O/Si–O (B–O at
`1370 cm⫺1 and Si–O–Si at 1075 cm⫺1) allows the concentration of
`the boron in the dopant layer to be determined.20-25 Figure 4 shows
`the dependence of the B–O/Si–O ratio on RTC time. For 800⬚C,
`there is a very little small change in the ratio between 25 and 45 s,
`which signifies a very slow rate of conversion of the gel to a glass.
`For 900⬚C, the ratio reaches a maximum value of 0.2 after 45 s cure
`and remains the same after a 60 s cure. The 1000⬚C profile reaches
`a maximum at a shorter time of 25 s.
`Spectroscopic ellipsometry.—The refractive index and the thick-
`ness of the SOD layers have been determined from spectroscopic
`ellipsometry measurements. Figure 5 shows the refractive index val-
`ues as a function of the curing time for different temperatures. All of
`the samples have a high value of refractive index for short curing
`times, ⱕ10 s. These high initial values are due to light organic resid-
`uals in the SOD layer, the existence of which are clearly shown in
`Fig. 2 and 3 by the presence of the 3000-3500 cm⫺1 band.
`There is a significant decrease in refractive indices for curing
`times >10 s, following the evaporation loss of the organic residuals.
`A similar behavior was observed for SOG films due to the loss of
`hydroxyl and ethoxy groups.14 It is known that the presence of such
`molecular functional groups in BSG increases the refractive index,
`and that the loss of these groups reduces the refractive index.14,42
`This observation is also consistent with the FTIR results (Fig. 2 and
`3), which show a reduction in the OH band intensity in the 3000-
`3500 cm⫺1 region as hydroxyl groups are lost. The refractive indices
`decrease to values between 1.47 and 1.55, which are typical values
`for BSG.42
`
`Figure 2. IR absorbance spectra of SOD layers that were annealed at 900⬚C
`for various times: 10 s (thin solid line), 25 s (dashed line), 45 s (dotted line),
`60 s (heavy solid line).
`
`that appears at ⬃1430 cm⫺1 (peak position of B–O stretching vibra-
`tions20-22) decreases rapidly with time and the peak shifts to a lower
`frequency. Then the intensity of this band increases again and reach-
`es a saturation value with a peak position at ⬃1370 cm⫺1. This com-
`plicated behavior can be explained by the influence of a number of
`different processes: (i) the decrease in film thickness as a result of the
`layer shrinking with temperature, (ii) the evaporation of the organic
`residuals from boron and silicate oligomers, and (iii) the formation
`of the borosilicate network through the bridging oxygen atoms.
`The existence of organic residuals (attached to the Si and B
`atoms38-40) after baking at 200⬚C and even after very short RTC
`treatment can be verified by the observation of the vibrational bands
`belonging to hydroxyl and ethoxy groups, in the high frequency
`region (2400–3600 cm⫺1) and at frequencies 1200 cm⫺1 (OH) and
`at 1635 cm⫺1 (CH3).6,7,22,32,41 The dramatic decrease in the absorp-
`tion intensity of the OH stretching vibration band (shown as insets in
`Fig. 2 and 3) verifies that the progressive loss of the OH groups in
`the form of H2O and C2H5OH due to the chemical reactions de-
`scribed in Ref. 38, 40. This result indicates that during thermal treat-
`ment, the hydrolysis and condensation reactions are continuing, re-
`leasing water and ethanol, and producing to Si–O–Si, B–O–B, and
`Si–O–B bonds34,38-40 at the expense of the Si–OH, B–OH, Si–OR,
`and B–OR groups. This is also consistent with the growth of the
`Si–O–Si and Si–O–B vibrational bands at 1075 and 910 cm⫺1, re-
`
`Figure 3. IR absorbance spectra of SOD layers that were annealed at 1000⬚C
`for various times: 2 s (thin solid line), 5 s (dashed line), 10 s (dotted line), 25
`s (dashed-dotted line), 60 s (heavy solid line).
`
`Figure 4. The dependence of the B–O/Si–O infrared peak ratio on curing
`time for different temperatures.
`
`Downloaded 11 Feb 2010 to 134.226.1.229. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
`
`REC Exhibit 1030, Page 4 of 7
`
`

`

`Journal of The Electrochemical Society, 147 (8) 3100-3105 (2000)
`S0013-4651(99)12-095-0 CCC: $7.00 © The Electrochemical Society, Inc.
`
`3103
`
`increase in reaction rate at higher temperatures is evident in Fig. 6
`by the more rapid increase in layer thickness. The measured refrac-
`tive index value at 950⬚C is highest after a 25 s cure. This value is
`not as high as the maximum value measured for 900⬚C. However, the
`rapid rate of change occurring at 950⬚C means that the true maxi-
`mum may not have been measured. After processing for 45 s, the
`refractive index decreases by over 0.03 from the 25 s value. In this
`case, however, both oxide layer growth and boron out-diffusion may
`be contributing to the decrease. In Fig. 3, the maximum B–O fre-
`quency position for 1000⬚C increases from 1370 to 1390 cm⫺1
`between the 25 and 45 s cures. This change in frequency is consis-
`tent with a decrease of boron content in the BSG layer. Sheet resis-
`tance measurements also confirm that boron diffuses from the oxide
`layer into the underlying silicon. These trends are more pronounced
`for RTC at 1000⬚C but, due to the increase in the conversion rate at
`high temperatures, additional data points would be required to deter-
`mine the detailed refractive index behavior.
`The most favorable RTC process step would be at a temperature
`that does not result in boron loss from the BSG layer, since the layer
`is used as a boron dopant source, but that also converts the SOD gel
`to BSG in the shortest time possible. It is possible to determine the
`optimum conditions by combining the FTIR, spectroscopic ellip-
`sometry, and sheet resistance results. We know from sheet resistance
`measurements, and from spectroscopic investigations, that boron
`diffuses from the BSG layer at temperatures greater than 900⬚C. As
`the rate of conversion increases with RTC temperature, a 900⬚C cur-
`ing step will be more time efficient than thermal treatments at 800 or
`850⬚C. Figures 4 and 5 shows that the BSG layer is in its most sta-
`ble condition after a 45 s cure at 900⬚C. These are the optimum RTC
`conditions for the conversion of SOD gel to BSG.
`The suitability of the converted BSG layer as a dopant source was
`determined by, first, doping a bare silicon wafer, and, second, by
`using the technique as a process step in the fabrication of a p⫹-n
`diode. The SOD gel was heated to 900⬚C for 45 s and then stacked in
`proximity to a silicon wafer and heated to 1050⬚C for 5 s. Figure 7
`shows the SIMS concentration depth profile of the product wafer. The
`wafer has a surface concentration of 4.7 ⫻ 1019 cm⫺3; the concen-
`tration is 1018 cm⫺3 at a depth of 65.5 nm. Three wafers were doped
`with a single BSG layer, and sheet resistance measurements confirm
`that all of the wafers were equally doped, with a sheet resistance of
`approximately 530 ⍀/m2. For comparison, one product wafer was
`doped with a SOD gel that had not been converted to BSG. The boron
`concentration profile is also shown in Fig. 7. The wafer has a higher
`surface concentration and a deeper junction depth than the wafer that
`was doped with the BSG layer. Sheet resistance measurements show
`that the boron profile of wafers that were subsequently doped with
`this SOD gel continually change until the SOD gel has converted into
`
`Figure 6. Variation in thickness of SOD layers, obtained from the spectro-
`scopic ellipsometry measurements, with changing RTC temperature and time.
`
`Figure 5. The dependence of the refractive index of SOD layers on RTA tem-
`perature and time.
`
`Initially, an increase in the refractive index could be expected due
`to the densification of the gel to a BSG structure.14 However, the
`decrease caused by the evaporation of the organic residuals (0.25-
`0.15) is one order of magnitude bigger than the expected decrease on
`BSG densification (0.025).14,15
`Oxide growth at the underlying Si/BSG interface might also be
`expected because the thermal processing is taking place under an
`oxygen-containing atmosphere. Sol-gel layers can retain a structure
`of interconnected pores, even after high temperature thermal treat-
`ment. We have estimated the porosity of the BSG layer to be about
`3%.43 Hence, the growth of SiO2 beneath the BSG layer, via the
`interconnected pores of the layer, should also be considered. How-
`ever, it is very difficult to observe changes in the global refractive
`index due to any densification or oxidation process while the organ-
`ic residuals influence the results.
`For curing times >10 s, the refractive index continues to vary
`with temperature and time. There are several factors that must be
`taken into consideration: (i) the refractive index will decrease if the
`underlying Si oxidizes, because of the decrease in the overall B2O3/
`SiO2 ratio, (ii) the refractive index will increase as the layer becomes
`more dense, and (iii) any diffusion of boron from the BSG layer will
`also cause a decrease in the refractive index.
`At 800⬚C, there is no significant change in refractive index for
`curing times >20 s. This is in good agreement with the B–O/Si–O
`dependence at 800⬚C measured by FTIR and further verifies that any
`change is very slow at this temperature. The refractive index for 850
`and 900⬚C increases by a small amount and reaches a maximum
`value at 45 s. This refractive index behavior is consistent with den-
`sification of the BSG layer, evidence for which also comes from the
`characteristic decrease in B–O frequency (1400 cm⫺1 region in
`Fig. 2 and 3). Next, the refractive index for the 850⬚C and 900⬚C
`samples decreases by 0.03 between the 45 and 60 s cures. Oxidation
`of the underlying Si or boron out-diffusion from the BSG layer
`would produce such a decrease. The film thickness determined from
`the spectroscopic ellipsometry calculations (Fig. 6), is in excellent
`agreement with the Si–O FTIR peak intensity, and shows that the
`film thickness increases with curing time. In contrast, there was no
`difference in the sheet resistance of the silicon source wafer before
`and after the 850 and 900⬚C RTC steps, confirming that the boron
`did not diffuse into the silicon source wafers, in significant quanti-
`ties, at these process temperatures. The decrease of 0.03 in refractive
`index must arise from the oxidation of the underlying silicon. The
`oxidation rate is a little higher than that of a normal rapid thermal
`oxidation process. This is reasonable, as these layers are significant-
`ly thinner than normal thermal oxide layers.
`At 950⬚C, the refractive index shows the same trend as the 850
`and 900⬚C plots, except that the changes occur more rapidly. The
`
`Downloaded 11 Feb 2010 to 134.226.1.229. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp
`
`REC Exhibit 1030, Page 5 of 7
`
`

`

`3104
`
`Journal of The Electrochemical Society, 147 (8) 3100-3105 (2000)
`S0013-4651(99)12-095-0 CCC: $7.00 © The Electrochemical Society, Inc.
`
`Figure 7. SIMS profiles of junctions that were formed by diffusion of boron
`from uncured and cured SOD.
`
`a stable BSG dopant source. These results highlight the need to con-
`vert the borosilicate gel to a stable BSG layer.
`Diodes were fabricated using proximity RTD to form shallow
`boron-doped junctions. Figure 8 shows the I-V characteristics for the
`diode. The forward bias turn-on voltage was 0.6 V, and the measured
`ideality factor was 1.07 over a 350 mV range. Under reverse bias, the
`diode has a sharp avalanche breakdown at ⫺29 V. The leakage cur-
`rent observed is very low, rising to 50 pA as the diode approaches
`breakdown. A more detailed description of these devices are pre-
`sented in a later paper.
`
`Conclusions
`The thermal conversion of a SOD borosilicate gel layer to a sta-
`ble BSG dopant source layer has been achieved by combining the
`results of FTIR, spectroscopic ellipsometry, and sheet resistance
`measurments. The layers were cured in an RTP in the temperature
`range 800-1000⬚C for 2-60 s in an atmosphere of 25% O2:75% N2.
`The changes in the properties of the layer can be summarized as
`follows
`1. The thickness of the layer increases with increasing temperature
`and time. This increase in thickness is attributed to oxide growth at the
`Si-BSG interface during the heat-treatment in an oxidizing ambient.
`2. Initially, the refractive index decreases following the loss of
`organic residuals from the layer. The refractive index then increases
`as the BSG layer densifies, and subsequently decreases again as the
`underlying silicon oxidizes. For the 950⬚C and 1000⬚C treatments,
`the refractive index may decrease again due to boron diffusion from
`the BSG layer.
`Heating at 900⬚C for 45 s was the optimum RTC process step
`required to convert the gel to a BSG, since there was no detectable
`boron loss from the BSG dopant source at this temperature. Very shal-
`low junctions, with profiles suitable for MOSFET technology, were
`fabricated by RTD of boron from the stabilized BSG layer. Diodes
`were successfully manufactured, and displayed excellent electrical
`characteristics. From the forward I-V characteristics a value of 1.07
`for the ideality factor was deduced and the reverse characteristic had
`a very low leakage current with a sharp breakdown at ⫺29 V.
`Acknowledgments
`Intel Ireland is gratefully acknowledged for financial contribu-
`tions toward this research project. The authors would also like to
`thank Professor J. K. Vij for use of his FTIR spectrometer..
`References
`1. The National Technology Roadmap for Semiconductors, p. 74, Semiconductor
`Industry Association, San Jose, CA (1997).
`2. W. Eichhammer, M. Hage-Ali, R. Stuck, and P. Siffert, Appl. Phys. A, 50, 405 (1990).
`3. C. Park, K. M. Klein, Al. F. Tasch, R. B. Simonton, and G. E. Lux, Tech. Dig. Int.
`Electron. Devices Meet., 67 (1991).
`4. S. Hong, G. A. Ruggles, J. J. Wortman, and M. C. Ozturk, IEEE Trans. Electron.
`Devices, 38, 476 (1991).
`
`Figure 8. I-V characteristics of a p⫹-n diode: (a) forward bias characteristics
`(b) reverse bias characteristics.