White luminescence from Si ¿ and C ¿ ion-implanted SiO 2 films

The microstructural and optical analysis of SiO 2 layers emitting white luminescence is reported. These structures have been synthesized by sequential Si 1 and C ion implantation and high-temperature annealing. Their white emission results from the presence of up to three bands in the photoluminescence ~PL! spectra, covering the whole visible spectral range. The microstructural characterization reveals the presence of a complex multilayer structure: Si nanocrystals are only observed outside the main C-implanted peak region, with a lower density closer to the surface, being also smaller in size. This lack of uniformity in their density has been related to the inhibiting role of C in their growth dynamics. These nanocrystals are responsible for the band appearing in the red region of the PL spectrum. The analysis of the thermal evolution of the red PL band and its behavior after hydrogenation shows that carbon implantation also prevents the formation of well passivated Si/SiO2 interfaces. On the other hand, the PL bands appearing at higher energies show the existence of two different characteristics as a function of the implanted dose. For excess atomic concentrations below or equal to 10%, the spectra show a PL band in the blue region. At higher doses, two bands dominate the green–blue spectral region. The evolution of these bands with the implanted dose and annealing time suggests that they are related to the formation of carbon-rich precipitates in the implanted region. Moreover, PL versus depth measurements provide a direct correlation of the green band with the carbon-implanted profile. These PL bands have been assigned to two distinct amorphous phases, with a composition close to elemental graphitic carbon or stoichiometric SiC. © 2003 American Institute of Physics. @DOI: 10.1063/1.1578172 #


I. INTRODUCTION
Semiconductor nanocrystals and related nanostructures have been extensively studied in the last years due to their interesting optical and optoelectronic properties.2][3] This leads to a widening of the band gap, as well as an enhancement of the efficiency of radiative recombination processes.Strong room-temperature photo-and electroluminescent signals have been observed from Sirelated structures, such as porous Si or Si nanocrystals embedded in SiO 2 . 4,5These results have opened extremely interesting perspectives for the development of a Si-based optoelectronic technology.
For the development of light-emitter devices in the visible region, one of the most interesting approaches is the synthesis of nanocrystals embedded in a SiO 2 film.SiO 2 is a well-known material from the standard Si technology, transparent in the infrared ͑IR͒ and visible spectral range, and it forms highly stable and well-passivated interfaces with silicon.
The ion beam synthesis of Si nanocrystals in SiO 2 is a two-step process which involves: ͑i͒ High dose Si ion im-plantation in a previously thermal oxidized Si wafer and ͑ii͒ high-temperature annealing.Thermal treatment of the Si-rich SiO 2 films at temperatures above 900 °C leads to phase separation into nanocrystalline Si and SiO 2 and annealing of the implantation-induced damage. 6,7The complete recovering of the high structural quality of the initial SiO 2 layer and the very stable and defect-free interface with the nanoparticles are the key features which explain the high photoluminescence ͑PL͒ efficiency of these structures. 7,8The same processes have also been recently used for the fabrication of electroluminescent devices. 9he ion beam synthesis of Si nanocrystals in SiO 2 has been the object of extensive investigation including the interpretation of the growth kinetics.1][12] Besides, the dependence of the average size of the nanocrystals on the processing parameters ͑ion dose, annealing temperature and time͒ has also been described in detail.In all of these cases, the emission is located in the red-IR spectral region ͑1.4 -1.8 eV͒ even for the smallest sizes. 13Then, luminescent systems at higher energies require alternative nanoparticles with a different chemical composition.
In this framework, the ion beam synthesis of luminescent structures by sequential Si ϩ and C ϩ ion implantation into the SiO 2 matrix has been proposed. 14 -16In SiO 2 films implanted only with C ϩ ions, it has been early evidenced that a severe drawback consists in the strong outdiffusion of the implanted C during the thermal process.This is caused by the formation of highly mobile CO species.Accordingly, these phenomena could be prevented by annealing under high-vacuum conditions.However, the most interesting approach to stabilize the implanted C profile is the previous existence of a high Si supersaturation, which enhances the interaction of both C and O atoms with the Si ones, and prevents the formation of CuO bonds.In this respect, Sirich SiO 2 layers have been demonstrated to act as efficient barriers for C outdiffusion. 17igh dose sequential implantation of Si ϩ and C ϩ ions into the SiO 2 oxides followed by high-temperature annealing results in the synthesis of structures with intense white ͑to the eye͒ luminescence.This white emission is resulting from the convolution of up to three luminescence bands located in the red-IR, green, and blue spectral regions.They have been related to the formation of different kinds of nanoparticles, such as Si, SiC, and C graphiticlike ones. 18In a previous experiment, even some evidence of the presence of 6H-SiC precipitates was observed. 19However, a further detailed analysis on a wider set of samples has not provided a clear hint of the presence of this specific phase.
][22] There is a general consensus in assigning these bands to the formation of C-related nanoparticles.An additional important feature is their very short lifetime ͑below one microsec-ond͒ compared to the Si nanocrystal-related one, which is of the order of hundreds of milliseconds.Zhao et al. 23 have reported decay times in the picosecond range.These data are quite relevant in view of the development of optoelectronic emitters, where the competitive switching characteristics rely on fast decay rates.All this gives a strong interest in achieving a more mature knowledge on the mechanisms related to these visible luminescence bands as well as on the physicochemical nature of the synthesized phases.
In this framework, this work reports the detailed structural and optical characterization of luminescent systems synthesized by high dose sequential Si ϩ and C ϩ ion implantation.The observed visible PL has been correlated with the implant and annealing conditions and with the microstructure of the processed films.The analysis of the interaction mechanisms between both implanted species in the oxide layers has been carried out by means of a wide range of different techniques, such as transmission electron microscopy ͑TEM͒, Raman scattering, secondary ions mass spectroscopy ͑SIMS͒, and x-ray photoelectron spectroscopy ͑XPS͒.

II. EXPERIMENTAL DETAILS
SiO 2 films, 400 nm thick, were grown on p-type Si ͗100͘ wafers by wet oxidation at 1100 °C.A two-steps Si ϩ implantation was designed by TRIM simulations 24 to obtain a rectangular-shaped Si excess profile at depths between 100 nm and 300 nm in the oxide layer.This was made to enhance C blocking and formation of SiC compounds in the implanted layer.Different doses were selected to obtain an ex-cess atomic content of Si in this region between 1% and 30%.Subsequently, a 60 keV C ϩ ion implantation was performed to a nominal dose as to obtain a peak C concentration equal to the Si excess value.The implant energy was selected to obtain the C-implanted distribution centred in the Si-rich buried region.All the implantations were performed at room temperature.Table I summarizes the implantation conditions for all the different wafers.
Additionally, reference wafers were implanted only with Si.In this case, a 150 keV Si ϩ ion implantation was performed at room temperature to doses ranging from 10 16 to 3ϫ10 17 Si ϩ /cm 2 .The implant parameters were tailored in order to obtain a projected range of about 200 nm and a silicon peak concentration similar to the one in Table I.
The samples were annealed under N 2 atmosphere at 1100 °C in two steps: The initial one was in a rapid thermal processing system ͑1 min͒ to ensure the reproducibility of the process, while the second one was performed in a conventional furnace ͑annealing time up to 16 h͒.This wide range of annealing times allows one to closely follow the Si precipitate evolution through the whole ripening regime, where the nucleation and pure growth are completed after the first minutes.In addition, some selected samples underwent postannealing treatment in forming gas (95% N 2 ϩ5% H 2 ) at 450 °C during 20 min.
PL measurements have been performed using a 60 mW He-Cd laser for ultraviolet excitation (ϭ325 nm).The luminescent emission was analyzed in backscattering mode by a 0.6 m monochromator and detected with a GaAs photomultiplier.The standard chopper-lock-in technique was used to increase the signal-to-noise ratio.
PL versus depth profiles on selected samples were performed by scanning PL microscopy on the surface of low angle bevelled samples ͑nominal bevel angle of 8°͒.Bevelling was made by mechanical polishing, using a commercial SSM Spreading Resistance Probe system.The sample was excited with a He-Cd laser and the light was recollected with an Olympus metallographic microscope provided with a reflection objective ͑numerical aperture of 0.4͒, which allows one to achieve a lateral resolution of about 1 m.
The detection and analysis of C-related phases was performed by Raman scattering and XPS measurements.Raman scattering measurements were made in backscattering configuration with a Jobin-Yvon T64000 spectrometer coupled with an Olympus metallographic microscope using the blue line of an Ar ϩ laser (ϭ488 nm) as excitation light.The laser power density in the sample was kept below 800 kW/cm 2 to avoid thermal effects.XPS spectra were mea- sured with a Perkin-Elmer PHI-5500 instrument using Al K␣ radiation.XPS versus depth profiles were obtained by measuring the spectra after sputtering the samples to different thicknesses with an Ar ϩ ion beam at 4 keV.
For the microstructural characterization of the wafers, the TEM technique has also been used.Cross-sectional ͑XS͒ TEM was used to identify the different crystalline phases present in the implanted oxides, as well as to characterize the size distribution of the nanocrystal population in the different regions of the samples and their evolution with the dose and annealing time.High-resolution electron microscopy and bright-field observations were performed by using a Philips CM30 microscope at XS samples.Some of these samples have also been studied under dark-field conditions by using a JEOL 200CX microscope for size-distribution measurements only.
Finally, the distribution to the C-implanted atoms has been also assessed by SIMS measurements using an Atomika A-DIDA 3000-30 spectrometer with an O ϩ primary ion source at 6 keV and an electron gun for charge neutralization.

A. Characterization of the C-implanted profile and C-related phases
Figure 1 shows the C depth profiles measured by SIMS in samples implanted with 10% excess, both as implanted and annealed for different durations.The Si-implanted profile, as simulated by TRIM, is also included.Figure 1 also shows the presence of a narrow surface C peak in the samples as implanted and annealed for 1 min.This surface peak vanishes when annealing is performed for longer times ͑more than 10 min͒.It is apparent that the C profile is kept well centered in the Si-rich region, even after annealing for 16 h.This clearly demonstrates the ability of the initial Si ϩ implant to inhibit the formation of CuO phases by stabilizing the C profile.
The Raman spectra measured in the 1200-1700 cm Ϫ1 spectral region at the same set of samples after thermal treatment at 1100 °C for different durations are displayed in Fig. 2. The as-implanted sample shows a broadband contribution, centered at about 1500-1600 cm Ϫ1 .This is similar to that previously measured in amorphous C and SiC films. 25,26This indicates the existence in the as-implanted samples of CuC related vibrational modes from an amorphous phase.After annealing, this band evolves toward two distinct contributions: The E 2g graphite vibrational mode at 1360 cm Ϫ1 and a zone-edge mode at 1590 cm Ϫ1 .This last mode is active as a consequence of the disorder-induced relaxation of the k-conservation rule. 27Therefore, thermal treatment leads to the formation of sp 2 coordinated CuC bonds in an amorphous or highly disordered phase.For annealing times longer than 10 min, these bands disappear in accordance with the vanishing of the surface C layer after oxidation.
The Raman spectra ͑not shown͒ for samples implanted at different doses and annealed during 4 h show an interesting feature: After such a long thermal treatment, only the sample implanted with 20% Si and C excess shows a contribution from sp 2 coordinated vibrational modes, implying the presence of a high amount of carbon in graphitic phases in the implanted region.
A systematic XPS analysis has been carried out on the processed structures.A typical concentration profile in an FIG. 2. Raman spectra of the 10 at.% exc.SiϩC coimplanted and annealed wafer.For more clarity, the spectra have been vertically shifted.The small feature appearing in some spectra at about 1500 cm Ϫ1 is an artifact from the detector system.annealed sample is shown in Fig. 3.The C atomic content at the surface is up to 40% in this case, and it extends up to about 10-20 nm in depth.The corresponding XPS band is centered at 284.3 eV, and is characteristic of carbon bonded to four C atoms (CuC 4 ).This gives additional support to the assignment of the graphite signal from the Raman spectra to this C-rich surface region.The origin of the high carbon content in this surface layer has to be related to a contamination of the beam during the implantation.The thermal stability of the implanted profile is well demonstrated also by the XPS profiles of annealed samples.
We checked that the whole C-implanted profile is proportional to the implanted dose, and it does not show any sign of broadening when increasing the implanted dose or annealing time.The C signal in the implanted region has a maximum content at a depth of about 150 nm.The position of the C XPS peak always ranges between those positions characteristic of graphite ͑284.3 eV, CuC 4 ) and SiC ͑282.8 eV, CuSi 4 ).In principle, this variation could be related to the presence of C-rich SiC complexes in the implanted region, or to the coexistence of two different C-related phases, as graphite and Si-rich SiC.
In Fig. 4, the energy position of the C 1s band versus depth is represented for the samples implanted to the highest doses ͑corresponding to 20% and 30% of implanted excess atoms͒.In the other cases, their intensity is too low to allow a reliable analysis.As reference, we have also indicated the positions corresponding to both CuC 4 and CuSi 4 bonding configurations.The relative position of the experimental band in relation to these references provides an indication of the relative amount of C in the respective phases.It is apparent that the predominance of the graphite phase in the sample implanted with 20% excess atoms, while for the sample with 30% excess atoms there is a significant shift toward the SiC phase.A fitting of the spectra with two Gaussian contributions centered at the reference positions allows one to estimate the relative weight of each phase.For the sample implanted with 20% excess atoms, about 75% of implanted C is in the graphite phase and only about 25% of the atoms are bonded to Si in a SiC phase.On the other hand, these fractions are of 40% and 60%, respectively, in the other sample ͑30% excess atoms͒.These data agree well with the Raman measurements.As previously indicated, only for the samples implanted with 20% excess atoms, there is a remaining detectable contribution in the Raman spectra from graphite vibrational modes after a long annealing time.
In the XPS Si 2p spectrum the SiuC ͑at 100.2 eV͒ and SiuSi ͑at 99.6 eV͒ components have not been resolved, due to their small energy separation.However, in the regions outside the C-implanted peak, the Si-related XPS band presents a well pronounced shoulder on the low-energy side, corresponding to the signal characteristic of SiuSi 4 bonding configuration at 99.6 eV.This implies the existence of pure Si aggregates in these portions of the implanted layer.

B. Detection and characterization of Si nanocrystallites
The XS TEM analysis of the implanted and annealed samples has revealed the existence of a complex multilayer structure, as can be seen in Fig. 5.This corresponds to the sample implanted with a 30% excess of Si and C and annealed during 16 h.Figure 5 shows the presence of a buried layer with a darker contrast where no crystalline precipitates are detected ͑labeled as region 2 in the TEM image͒.Moreover, in the regions next to the buried layer ͑regions 1 and 3, respectively͒, a population of small nanocrystals has been observed.The analysis of the nanodiffraction patterns from these regions has allowed us to identify them as Si nanocrystals.The inset in Fig. 5 shows a high-resolution TEM image of one of these nanocrystals from region 1.Finally, region 4 corresponds to the remaining SiO 2 layer, without precipitates.
In Fig. 6, the Si nanocrystal mean diameter and standard deviation from both regions 1 and 3 are represented as a function of the excess of implanted atoms.These data have been measured for the samples implanted with 10%, 20%, and 30% excess and annealed for 16 h.Data obtained from Si-implanted reference samples have been added.All of these values have been obtained from the corresponding size histograms, which were measured by detailed TEM observations under dark-field imaging conditions at a significant number of precipitates ͑higher than 50͒.For the samples implanted to lower doses, the number of precipitates observed was not high enough to obtain reliable statistics.The measurements were performed using dark-field conditions for the imaging of Si nanocrystals in SiO 2 , as described in Ref. 14.
A comparison of Fig. 5 with the Si ϩ -and C ϩ -implanted profiles reveals the strong dependence of the microstructure in each region with the relative C to Si ratio: Region 2 corresponds to the maximum of the C-implanted profile, where both Si and C excess contents are similar.The high C content in this region strongly inhibits the precipitation of Si in the nanocrystalline form.Most of the Si excess atoms are prob-ably trapped in SiC-related complexes which are not resolved in the TEM image but detected by XPS.The formation of the nanocrystals is only possible in regions 1 and 3, where the concentration of C atoms is well below the Siimplanted one.Furthermore, a noticeable decrease is apparent in the average size of the Si nanocrystals of region 1 in respect to those observed in region 3 ͑i.e., 1.1 nm against 1.4 nm for 10% Si excess͒.The implanted species profile versus depth in Figs. 1 and 3 reveals that the ratio between C-and Si-implanted atoms is higher in region 1 than in region 3.Moreover, the average size of the Si nanocrystals is always larger for the reference samples not containing carbon.Thus, the different density and size distribution in regions 1 and 3 is likely due to a preferential bonding of Si with C atoms.

C. Optical analysis: Photoluminescence behavior
As already mentioned, the oxides coimplanted with Si ϩ and C ϩ ions show, after annealing, an intense luminescence signal visible to the naked eye.For Si-and C-implanted contents of 10% or higher, this luminescence appears as an intense white emission.This results from the simultaneous emission of up to three bands covering the whole visible spectral range from the red to the blue region.Fig. 7 shows the PL spectra measured from the samples implanted with the different excess concentrations after annealing for 2 h.The spectra always present a band in the red-IR region, as plotted in Fig. 7 ͑left-hand side͒.In the following, it will be labeled as the R band.The samples implanted to doses leading to excess Si and C contents up to 10% show a additional band in the blue-green spectral region, labeled as the G band ͑see Fig. 7, right-hand side͒.Finally, the samples implanted to higher doses show a third luminescence band in the blue wavelength range, labeled as the B band.
The peak energy of these three bands as a function of the atomic excess concentration is plotted in Fig. 8, together with data from the reference samples.The spectral position of the R band shows the same dependence on the implanted dose as the red-IR emission characteristic of Si nanocrystals.This allows the assignment of this band to these nanoprecipitates.As shown in a previous work, 8 for doses leading to Si excess content below 10%, the average size of the nanocrystals does not depend on the implanted dose, and the corresponding PL emission is centered at about 1.7-1.8eV.For higher doses, interaction effects between the Si nanocrystals lead to an increase in the average size ͑as shown in Fig. 6͒, and it determines a redshift of the PL band, according to the quantum confinement effect.The other PL bands show a similar shift as the implanted dose increases.This is apparent in particular for the G band, which evolves from the blue region for a carbon content below 10% to the green one for the other cases.In this case, such a large redshift has to be attributed to a variation in the composition and/or atomic configuration of the carbon aggregates, rather than to quantum confinement effects characterized by the size increase of the particles with the implanted dose.
The kinetic evolution of the intensity of the R and G bands is plotted in Fig. 9.The R band shows a strong increase with the annealing time.However, in almost all of the samples, a saturation is not reached even after the longer annealing time of 16 h.As described in previous studies, the evolution of the PL intensity with the annealing time of Si nanocrystals ion beam synthesized in SiO 2 is characterized by the existence of two distinct regimes; ͑i͒ an initial fast increase ͑transient stage͒ that bends over after 3-4 h of annealing and ͑ii͒ a saturation regime for longer annealing times up to 16 h.Once the nanocrystal population has reached the asymptotic Ostwald ripening stage ͑after about a few minutes by annealing at 1100 °C͒, any further increase in PL efficiency is mainly related to the progressive passivation of interfacial nonradiative defects, such as P b centers.This has been established by observing an inverse correlation between the PL intensity and the density of P b defects measured by the electron spin resonance technique for annealing times longer than 20 min. 7,8he observation of a different thermal evolution of the R band ͑no saturation in intensity͒ under the actual experimental conditions suggests that the presence of C atoms slows down the kinetics of formation of the Si nanocrystals and/or their interface passivation.To further investigate this issue, additional thermal treatments have been performed in form-ing gas (N 2 ϭ95%,H 2 ϭ5%) ambient.Previous studies on Si-implanted SiO 2 have demonstrated that hydrogenation at moderate temperatures considerably enhances the PL intensity and lifetime of the Si nanocrystal-related band, by passivating P b interfacial defects. 8he red emission from the sample implanted with 30% excess of Si ϩ and C ϩ ions and annealed at 1100 °C during 20 min increases by a factor of 5 after forming gas annealing ͑450 °C, 20 min͒, while keeping the same spectral shape ͑not shown͒.For the higher-energy bands, no significant changes are observable.A similar behavior has also been obtained in the sample implanted with the same dose and annealed during 8 h, leading to the conclusion that even after the longest annealing times a significant density of nonradiative defects is still present at the Si/SiO 2 interface.This observation suggests an active role of the implanted C atoms in inhibiting the formation of well-passivated Si/SiO 2 interfaces.An alternative explanation which, in principle, cannot be ruled out, considers the damage accumulation in the implanted layer as a consequence of the extremely high doses used in the present experiment.A high density of point defects in such structures could account for the reduction in optical efficiency of the Si nanostructures, either after standard hightemperature annealing in inert atmosphere or after hydrogenation.For the G band, the data on the PL intensity versus annealing time plotted in Fig. 9 show the existence of a more complex behavior.In fact, two different trends are observed depending on the implanted dose: ͑i͒ For doses leading to implanted excess atom concentrations below 10%, the G band slowly vanishes for increasing annealing duration, while ͑ii͒ for doses exceeding this value, the intensity of the band increases and shows a fast transient stage toward saturation for annealing times between 10 and 20 min.The 10% excess case shows an intermediate behavior between both cases.
For the higher implanted doses, the G band shifts from the blue to the green spectral region, while the PL intensity presents an evolution characteristic of the growth kinetics of particles in the SiO 2 matrix.As a function of the implanted dose, the PL intensity at the saturation regime shows a nonmonothonic behavior, with a maximum emission for a 10%-20% excess.The intensity of the B band, only detected for the higher implanted doses ͑not shown in Fig. 9͒, presents a similar behavior.
Figure 10 shows the results of the indepth resolved scanning PL measurements.The intensity of both R and G bands ͑versus depth͒ has been measured at different points from the low angle bevelled surface of the sample implanted with 20% excess atoms and annealed for 10 h.The XS TEM image of the sample has been added in the background and, for more clarity, the y axis for the normalized intensity of the G band has been inverted.The evolution of the B band is hindered by a parasitic component in the experimental setup, which did not allow its indepth resolved analysis.Figure 10 shows that the intensity of the R band decreases when the laser spot passes through the shallower ͑region 1͒ and deeper ͑region 3͒ layers containing Si nanocrystals, while it remains constant through the region 2 ͑without Si nanocrystals͒.Such direct experimental evidence provides a definitive identification of this band as related to the Si nanocrystals.On the other hand, the intensity of the G band closely follows the C-implanted profile.It decreases when scanning region 1 and mostly region 2, while the signal becomes negligible once the laser spot has reached region 3.Such a direct correlation of the G band with the C-implanted profile ensures that the observed optical features in the green spectral range are in-deed related to carbon clusters mainly located in region 2.
The presence of two well resolved bands ͑G and B͒ in the case of samples implanted to the highest doses can be correlated with the existence of two distinct phases involving the C-implanted atoms, according to the interpretation of the XPS experiments: Graphitic carbon and SiC.The detailed TEM observation of region 2 ͑including high-resolution imaging and electron nanodiffraction͒ has not shown any signal from a crystalline phase, implying the probable amorphous nature of these precipitates.We have to notice that the detailed composition of these precipitates is not well established.Although the XPS data are explained in terms of both graphitic elemental carbon and stoichiometric SiC, this could alternatively be determined by Si x C y ͑or even Si x C y O z ) domains with different composition, still close to elemental carbon and stoichiometric SiC.This has been confirmed by electron energy loss spectroscopy measurements on the same set of samples, which detected the presence of Si x C y O z complexes ͑data not yet published͒.
The assignment of the G and B bands from the samples implanted at the higher doses to the C-rich and SiC-related precipitates is in accordance with the structural analysis.The samples with the highest relative contribution of the G band in the PL spectrum are those implanted with 20% excess atoms.These are also the samples with a highest relative amount of carbon bonded in the graphitic phase.The same set is also the only one that shows a remaining graphitic contribution in the Raman spectra after a long annealing time, as previously pointed out.Increasing the Si ϩ -and C ϩ -implanted dose leads to an enhancement of the SiCrelated phase, and this is also accompanied by an increase in the relative intensity of the B band in respect to the G one.
The proposed interpretation for the G and B bands is also supported by the data reported in literature.Hayashi et al. 28 observed a broad luminescence band centered at about 2.2 eV from C-rich SiO 2 films prepared by rf cosputtering, and the position of this band shifts toward higher energies when the C content decreases.This behavior, together with the Raman and optical absorption data, has led to the interpretation that this band is related to carbon clusters in the oxide.A yellow-green luminescence band ͑2.0-2.2 eV͒ was also observed in C ion-implanted SiO 2 layers.In this case, the intensity of the PL band was well correlated with the contribution of carbon-related vibrational modes in the Raman spectra. 14A luminescence band at higher energies, in the range from 2.5 eV to 2.95 eV, has also been reported from carbon graphite microcrystals embedded in SiO 2 layers synthesized either by ion implantation 22,23 or by sputtering deposition of C-rich oxides 29 followed by thermal annealing.It is worth remarking upon the strong dependence of the PL energy on the conditions of fabrication of the samples, which determines the carbon cluster structure and size.
1][32][33][34][35] Besides, some authors have analyzed the PL emission from Si ϩ and C ϩ ion-implanted SiO 2 . 20,21Rebohle et al. 21have reported the appearance of two bands close to the G and B ones, at about 2.1 eV and 2.9 eV, in a Si ϩ and C ϩ sequential implantation process with excess atomic concentration in the range of 5% to 10%. 21The absence of the R band related to the Si nanocrystals was due to the fact that both Si and C implant conditions were chosen to obtain similar overlapping excess Si and C profiles.TEM observations in these samples corroborated the absence of nanocrystals in the implanted layer.They attributed the highenergy band ͑2.7-2.8 eV͒ to the formation of amorphous clusters of Si y C 1Ϫy O x complexes.Quite interestingly, they point out that a different origin of the blue emission, for instance from implantation-induced point defects, can be ruled out.This is a particularly important point to discuss in the present interpretation of the results for the low-implanted doses ͑1% and 5% excess͒, where, apparently, the G band shows two anomalies: The PL peak appears at very high energies ͑2.8 -3.0 eV͒ and it shows a lower thermal stability.The formation of oxygen-vacancy-related defects in SiO 2 has been previously claimed to interpret the strong blue luminescent signal from Si ϩ , Ge ϩ , and Sn ϩ ion-implanted SiO 2 specimens. 36,37These layers show a strong PL signal peaked at energies between 2.7 eV and 3.2 eV, together with an UV emission around 4.3 eV.The intensity of these bands strongly depends on the implanted dose and ion mass ͑ruling the defect production negligible in the case of carbon͒, and on the annealing temperature ͑defect thermal stability͒.In the case of silicon implantation, the oxygen-vacancy centers are completely annealed out at temperatures above 900-1000 °C and, in fact, no sign of their PL emission has been found in our reference samples.
Moreover, we found out that the lifetime decays of the B and G bands are in the nanosecond range for the whole set of annealed samples ͑not shown͒.This agrees with the available reports on time-resolved measurements.Intense blue PL bands have been reported from thermal SiO 2 films coimplanted with Si ϩ and C ϩ ions by Zhao et al., 20 related to the formation of SiC and/or Si, C, and O related complexes.They reported lifetime constants ranging from 50 to 800 ps.Lifetime decays from 100 ps up to few ns have also been measured in porous SiC, 30,34 providing an additional indication of the likely relationship of the luminescent B band with SiC-related secondary phases in the SiO 2 matrix.

IV. CONCLUSIONS
The detailed microstructural and optical analysis of the Si ϩ and C ϩ sequentially implanted and annealed SiO 2 layers has allowed us to characterize the different bands arising in the PL spectra, which are responsible for the intense white emission observed at the higher implanted doses ͑for atomic excess concentrations higher than 10%͒.These bands have been assigned to the emission from Si nanocrystals, and C-related amorphous nanoparticles with a composition close to graphitic carbon and SiC, respectively.At lower implanted doses, the PL spectra only show one band, in addition to the well-known red one related to the Si nanocrystals.While still being related to carbon-containing clusters, it shows a lower thermal stability.Carbon implantation has also been observed to strongly inhibit the formation of Si nanocrystals in the region of the C-implanted profile, affecting also the growth kinetics of the Si nanocrystals and the passivation of the Si/SiO 2 interfaces.As a result, a proper choice of the implantation and annealing conditions allows one to tune the color and wavelength of the emission, by tailoring the various contributions in the PL spectra.

FIG. 1 .
FIG.1.SIMS profile of the C concentration in the 10 at.% exc.SiϩC coimplanted wafer, before and after annealing at 1100 °C.The Si-implanted profile has been added, as simulated by TRIM.

FIG. 3 .
FIG. 3. Concentration vs depth profiles for the implanted species as obtained by XPS measurements on the 10 at.% exc.SiϩC coimplanted and annealed sample.

FIG. 5 .
FIG. 5. XS TEM image of the 30 at. % exc.SiϩC implanted and annealed ͑16 h͒ sample.The image shows the presence of a buried layer with darker contrast and free of crystalline precipitates ͑region 2͒.Comparison with the implanted profiles indicates that this region corresponds to the maximum of the C-implanted profile.Si nanocrystals are observed in regions 1 and 3, which are located above and below this buried layer.Finally, region 4 corresponds to the remaining SiO 2 layer without precipitates.The inset shows a high-resolution image of a Si nanocrystal from the region 1, above the buried C layer.

FIG. 8 .
FIG. 8. PL energy position of the R, G, and B peaks as a function of the implanted Si and C atomic excess concentration.

FIG. 10 .
FIG. 10.Normalized PL intensity of the R and G peaks vs depth profiles from the sample implanted with 20% excess atoms and annealed for 10 h.The labelling indicates the regions with different C content.

TABLE I .
Implantation parameters for the wafers implanted with Si ϩ and C ϩ ions.