Ex Situ Photoluminescent Emissions from Porous...

3284 J. Electrochem. Soc., Vol. 145, No. 9, September 1998 The Electrochemical Society, Inc.

7. 5. R. Kasi, M. Liehr, P. A. Thiry, H. Dallaporta, and M.Offenberg, Appl. Phys. Lett., 59, 108 (1991).

8. K. Saga and T. Hattori, Appi. Phys. Lett., 71, 3670(1997).

9. K. Saga and T. Hattori, J. Elect rochem. Soc., 144, L253(1997).

10. K. Saga and T. Hattori, J. Electrochem. Soc., 144, L250(1997).

11. T. Ohmi, T. Imaoka, T. Kezuka, J. Takano, and M.Kogure, J. Electrochem. Soc., 140, 811 (1993).

12. N. Miki and T. Ohmi, in Ultra Clean Technology, 9,Suppi. 2 (1997).

On the Correlation of Aqueous and Nonaqueous In Situ andEx Situ Photoluminescent Emissions from Porous Silicon

Evidence for Surface-Bound Emitters

James 1. Goe, Frank P. Dudel, and Lenwarci Seals

School of Physics, Georgia Institute of Technology, Atlanta, Georgia 30332-0430, USA

Melissa Reger

School of Chemwal Engrneering and Materials Science, Sarkeys Energy Center University of Oklahoma,Norman, Oklahoma 73019-0628, USA

Paul Kohl*School of Chemical Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0100, USA

Lawrence A. Botfomley*School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta, Georgia 30332-0400, USA

ABSTRACT

Porous silicon samples were prepared by anodizing p-doped Si(100) substrates in both aqueous (HF/H20, HF/CHOH,HF/CH3OH/H20, HF/C2H5OH/H20) and nonaqueous (MeCN/HF) media. The time-dependent porous silicon photolumi-nescence (PL) was monitored during the etch (in situ) and after removal from the etch solution (ex situ). Correlation ofthe ex situ and in situ PL indicates that the composition of the etchant solution plays an extremely important role in theonset, time-dependent intensity, and lifetime of the emission, both in and out of solution. The effect of etchant solutionadditives (ethylene glycol, CH3OH, C2H5OH, NaF, HC1, and NaC1) on the porous silicon PL both during arid following theetching cycle, was also determined. The distinct and different correlations found between aqueous and nonaqueousetchants provide insights into the mechanism of PL. These results, when considered in the context of quantum chemicalmodeling, strongly suggest surface-bound silicon oxyhydride moieties as the source of the porous silicon PL.

IntroductionThe visible photoluminescence (PL) extending from 500

to 800 nm which characterizes electrochemically (EC) pro-duced porous silicon (PS)12 at room temperature has beenthe subject of considerable interest because of its potentialuse in the development of silicon-based optoelectronics,displays, and sensors. While this luminescence is thoughtto occur near the PS surface,34 the source of this visibleemission is controversial as the efficiency and wavelengthrange of the emitted light can be affected by the physicaland electronic structure of the surface, the nature of theetching solution,56 and the nature of the environment intowhich the etched sample is placed.9 With this parameterdependence, it is surprising that, with few exceptions1011PL spectra are generally reported for PS samples (formedin dilute HF solutions) which have already been dried inair or more inert environments after undergoing etch andrinse treatments. These ex situ samples provide spectralinformation but do not indicate the evolution of the PSluminescence. It is therefore of interest to observe the PLas it evolves, in situ, during the etching process and fol—lowing, while the sample is still wet by the etching solu-tion or has been treated by an appropriate rinse. Here weexamine the in situ development of the PS photolumines—cence in several etch solution environments as well as thecontrolled modification of these environments. For certain

* Electrochemical Society Active Member.

select environments, we also consider the net effect ofthese in situ treatments on ex situ behavior.

We have carried out an extensive series of experi—ments5'79'12 in both aqueous (HF/H20, HF/CHOH, HF!CH3OH/HZO, HF/C2H5OH/H20) and nonaqueous (MeCNfHF, anhydrous) etching media. We monitor the time-depen--dent PL, in situ, during the etching cycle and before the PSsample is removed from the etching solution, including astudy of the effect of the additives ethylene glycol, CH3OH,C2H5OH, NaF, HC1, and NaCI on the in situ luminescenceprocess. We also consider the effect of these treatments onthe ex situ behavior of the PL after removal of the PS fromthe etching solution. By correlating the ex situ with the insitu behavior of the PS, we find that the aqueous elec-trolyte composition plays an extremely important role inthe onset, time-dependent intensity, and lifetime (solution)of the PL emitters.7-912 Comparable studies on nonaqueouselectrolytes are found to demonstrate distinctly differentcorrelations between the in situ and ex situ behavior andwe use these resuilts to provide possible insights into themechanism of the PL process. Further, in complement tothe experimental studies, we introduce a quantum chemi-cal modeling which suggests that silicon oxyhydride-likefluorophors9 strongly bound to the PS surface are thesource of the PS PL. The silicon oxyhydride moiety can beviewed as a highly reactive intermediate in the oxidativepathway between the hydrophobic Si—H surface generatedin the HF etching process and the hydrophilic Si02 surface.

.1 Electrochem. Soc., Vol. 145, No. 9, September 1998 The Electrochemical Society, Inc. 3285

The studies reported here complement those of Sailoret al.'3-'6 We consider that the oxyhydride fluorophors arebound to the PS surface and that they are to be distin-guished from surface adsorbates. Sailor et al.'" havereported the hole transfer quenching of the PL from PS byferrocene'3 and the irreversible quenching due to aromatichydrocarbons and redox active solutes,"3 they have alsodemonstrated the reversible quenching of the PS lumines-cence in air due to trace levels of NO and NO2 16 adsorbates.This suggests, as we have also observed, a wide diversity ofabsorbant effects which may well be associated with sur-face interaction. Here we emphasize that the action of anadsorbate on a PS surface must be carefully distinguishedfrom the creation or derivatization of a FL source."'7

The most widely considered hypothesis to explain theluminescence from FS suggests that the FL results from theradiative recombination of quantum-confined electronsand holes in columnar structures or undulating wires."2' Ithas also been postulated that surface-localized states,involving irregularly shaped small crystallites that are notperfectly passivated, produce elementary excitationswhich are first trapped prior to recombination.3"2' Severalresearchers have suggested that the FS luminescenceresults from surface-confined molecular emitters22 includ-ing a molecular complex of silicon, oxygen, and hydrogenknown as siloxene (Si,O3H,). Stutzmann and co—work-ers23-25 have used the optical detection of magnetic reso-nance (ODMR) to establish that the FS "red" emissionresults from a triplet exciton. Identifying the close analogyof both the ODMR and optical excitation (FLE) and FLspectra of PS and "annealed" siloxene, Stutzmann et al.2325have suggested this molecule as the origin of the PS FL. Wehave suggested a more general origin for the observed fea-tures5'92' which focuses on the functional groups associatedwith annealed siloxene and concurs with the observationsof Steckl et al.,27 who have obtained evidence for the sili-con oxyhydrides in strain-etched FS thin films, correlatingtheir observations with crystallinity and FL. Further, in astudy of the thermal oxidation and nitrogen annealing ofluminescent FS, Yan et al.2' have obtained evidence thatresidual hydrogen exists in a thermally oxidized FS film,probably in the form of SiOH.2'

Silicon oxyhydride fluorophors are similar in form tothe localized surface defects (or centers) associated byFrokes et al.2" with the region of the FS surface fromwhich the orange-red FL originates. Frokes et al. haveconsidered and modeled the role of interfacial oxide-relat-ed defects in the form of nonbridging oxygen-hole centers(NBOHC). These luminescent sources have also been sug-gested, more recently, by other researchers.3'

As we have indicated, detailed quantum chemical calcu-lations2' suggest that the FS FL can plausibly be attributedto silicon oxyhydride-like fluorophors strongly bound tothe PS surface. The fluorophor contains an Si=O doublebond and a hydrogen, hydroxide, SiH3, or OSiH3 ligandbound to the silicon. The silanone-based oxyhydrides with-OH or -OSiH3 moieties demonstrate a close correlationbetween their lowest lying excited triplet states and thetriplet exciton identified by Stutzmann.2325 The changes inbonding associated with electronic transitions involvingthe oxyhydride ground electronic singlet states and theirlow-lying triplet states, especially in the SiO-relatedbonds, and a substantial shift to larger internuclear dis-tance associated with these excited electronic states rela-tive to their ground states, correlate well with the observedcharacter of the FL spectra. The excitation of an excitedtriplet state greatly shifted from the ground electronic statewhich undergoes rapid relaxation down the excited statevibrational manifold does much to explain the significantred shift of the FL spectrum (600—800 nm) from the knownabsorption peak wavelength of the excitation spectrum(—350 nm).2' The quantum chemical studies,2' which havealso been used to demonstrate the effects of surface deriva-tization,' also suggest that neither the correspondingsilylenes nor tricoordinated silicon compounds with dan-gling electrons contribute to the observed FL.

The results of our experiments conducted in aqueousand nonaqueous media provide support for a surface-bound molecule-like emitter as the source of the FL fromPS. Several observations lead us to this conclusion,including the early appearance in time (


—Cuvette capwith holes for wires4 Electrical connection to Si sample

4 Pt electrode

wax covering remainder of Sample

—Silicon

Conductive paint holdingwire onto silicon wafer

—Sputtered Ai Ohmic Contact

Fig. 2. Closeup diagram of crucible configuration for etchingsolution and silicon wafer preparation used to study the mercurylamp and laser-induced visible light emission from porous silicon.

The optical train for the PL experiments was arrangedon a heavy-duty optical breadboard as indicated in Fig. 1.The output from an appropriate pulsed laser [Nd:YAG(ytrrium aluminum garnet) pumped dye laser in the range390—320 nm, nitrogen laser at 337.1 nm, and KrF and ArFexcimer lasers at 248 and 193 nm, respectively] or a con-tinuous wave argon ion laser was expanded and sentthrough a mask to the porous silicon surface. In order toexcite the PL process during the etch cycle, the laser out-puts traversed the etching solution and impinged directlyonto the etching silicon wafers. The majority of these insitu studies used nitrogen or KrF excimer laser radiation.Since an anhydrous HF/MeCN solution strongly absorbs248 (KrF) and 193 nm (ArF) radiation, only a nitrogenlaser, mercury lamp, or YAG pumped dye laser could beused for these experiments. In several experiments carriedout after the PS samples were washed and allowed to dryin air, the observed ex situ PL was of sufficient intensity soas to warrant excitation using only the scattered lightfrom the laser sources with no direct "focusing" of thelaser onto the PS surface.

To achieve reproducibility, the following alignment pro-cedure was adopted. The cuvette was placed into a perma-nent clamp holder connected to a magnetic base x-ymount attached directly to the top of the optical bread-board. After verifying that the cuvette was properly illu-minated by the masked light source of choice for a givenexperiment, a strongly luminescent PS sample from a pre-vious experiment was placed into the cuvette and the posi-tion of the McPherson spectrometer was adjusted to max-imize the PL signal at 650 nm. The direct scattering oflaser light into the spectrometer was minimized. The cali-brating PS sample was then removed and the cuvette filledwith the appropriate etching solution. The silicon sample

of interest (anode) and the Pt wire (cathode) were thenplaced into the cuvette. The sample was adjusted vertical-ly to optimal alignment with the exciting laser source androtated so as to avoid direct laser reflection into the spec-trometer. The leads were attached to a computer-con-trolled home-built constant current source and, after afinal alignment check, the entire system was covered witha heavy black cloth to avoid any contamination of theexperiment by the room lights. Care was also taken toassure that the observed PL was excited in a single photonexcitation process.

The PL was dispersed through the McPherson mono-chromator, which was scanned using a computer-inter-faced stepping motor, and impinged on a Hamamatsu 446phototube. The output from the phototube was sent to anSR400 (Stanford Research Series) photon counter whosegate for most experiments was set to the first 100 p.s of PL.The output from the photon counter was processed usingan IBM-compatible PC. The PL from the PS was studiedboth during the etching cycle, following etching in solu-tion, and upon the removal, washing, and drying of thesample in air. Not only were ex situ samples studied in airbut also under solutions of HC1/H,O, NaF/MeOH,NaC1/MeOH, methanol, ethanol, or ethylene glycol inwhich the washed sample was placed. These samples wereplaced into their original holders and sometimes also intoa cuvette which contained the rinse solutions as outlined.The previously outlined alignment procedure was repeat-ed. A typical scan, first from 490 to 730 nm, in 2.5 nm stepswith 30—40 laser shots per data point for lasers runningwith repetition rates between 20 and 30 Hz, requires—3.5—4 mm. This scan is immediately taken in reverse, fur-nishing a consistent internal check for any possiblechanges which might occur during the scan cycle. Becauseof the changes which can be manifest over the 3.5—4 mmscanning period, several runs were also devoted to timescans at the individual wavelengths 5200, 6200, and some-times 7100 A. Spectral calibration could be accomplishedwith a mercury lamp or the individual laser excitationwavelengths, often in second order.

ResultsExSitu PL Spectra.—PL spectra measured in an ex situ

environment for PS samples formed using four basic elec-trolyte solutions including acetonitrile/2M HF in a nearlyanhydrous environment, methanol/20% HF (vol % using a49% aqueous HF solution), water/20% HF, and methanol!50% HF are depicted in Fig. 3. The PS samples were pre-pared by etching (100) silicon at 4 mA/cm2 for approxi-mately 10 mm, rinsing with MeOH, and drying in air forseveral minutes before the PL was taken. For a given elec-trolyte solution, ex situ spectra were unaffected by a lim-ited variation of etch times (from 6 to 12 mm) and etchcurrent densities (from 4 to 8 mA/cm2).

The PL spectra in Fig. 3 (corrected for phototube wave-length response) clearly show a slight wavelength depen-dence associated with the different electrolytes. The PLfrom the PS formed in the acetonitrile solution, whosewater concentration is of order 0.001 mol/L, 'z was themost red-shifted with a peak wavelength at approximate-ly 7000 A, while the PL from the PS sample etched in 20%HF/water (6 mol/L HF) was the most blue-shifted with apeak at —6500 A. However, this energy difference, whichcorresponds to only 1100 cm' or —3 kcal/mol, mightresult from a number of possible sources not associatedwith differences in electrolyte solution.

The PL emission spectra depicted in Fig. 3 resemble thespectra generally observed for PS formed in HF solutionsand dried in air. However, these air dried PL spectra pro-vide only a glimpse of the PS luminescence in the latterstages of the development of the source of this emissionand provide limited information about the evolution of theemission. It therefore is of interest to monitor the PL as itevolves for a given etch solution throughout the etchingprocess, from the initial stages of preparation when the

Si surface to be etched

4 WIre

J. Electrochem. Soc., Vol. 145, No. 9, September 1998 The Electrochemical Society, Inc. 3287

5000'

5500' '

6000'

6500 '7000 7500

Wavelength (Angstroms)Fig. 3. P1 spectra (ex situ) for PS samples excited by a nitrogen

laser and prepared by etching Si(100) at 4 mA/cm2 for approxi-mately 10 mm in (a) 20% HF in H20, (b) 20% HF in MeOH, (c)anhydrous MeCN/HF, and (d) 50% HF in MeOH. The preparedsamples were rinsed in MeOH and dried in air for several minutesbefore the P1 was taken. The spectra taken at a time gate of0.8—100 ps have been corrected for system response.

sample is first wet by the etching solution through thecommencement of the etch cycle.

Monitoring of "green and orange-red" photoluminescentemissions—To study the evolution of the PS PL, the emis-sion is monitored in situ at two wavelengths, 5200 and6200 A. The 6200 A wavelength represents the best combi-nation of the peak PL wavelength, convolved with thespectrometer + phototube system response, which can beused to monitor the orange-red PL emission commonlycharacterizing UV-pumped (? < 350 nm) porous silicon.5200 A represents the wavelength of choice5 to monitor thePL precursor state, whose signature has been observed bya more limited group of researchers, as it is characterizedby a green spectral emission which appears to transformin time to the final orange-red emitter.5 Figure 4 demon-strates the PL at these two wavelengths where Si(100) isetched in each of the four basic etching solutions consid-ered in this study.

Anhydrous HF/MeCN Etch—The data for the nearlyanhydrous 2 M HF in MeCN etch solution depicted inFig. 4a at higher resolution'2 shows that the 6200 A PL isobserved as early as 10 s into the etching process, longbefore pores can be formed. Figure 4a indicates that the6200 A PL rises linearly during the etching process fol-lowed by a sharp rise in the PL intensity after the EC etch-ing cycle is terminated. At the sensitivity of the detectorused in this study and for the extremely low water contentpresent, during the etch cycle, little or no detectable PLwas observed at 5200 A until after the etching current wasterminated. Here, similarly, a significant rise in PL inten-sity was observed at 5200 A. A gradual decay in the PLwas then observed at each wavelength while the siliconremained in solution. This basic behavior was obtained forseveral etch currents ranging from 4 to 24 mA/cm2; how-ever, it was clear that the slope of the initial linear rise inPL emission intensity decreased with increasing etch cur-rent over this range.7 This dependence, depicted for 11 and24 mA/cm2 etching currents in Fig. 5a, in concert with thesharp rise in PL intensity after the etching current is ter-minated, suggests that the etching process both createsand destroys the PL emitters. This important interplay,7

Time (Minutes)Fig. 4. P1 intensity (arbitrary units) at 5200 and 6200 A vs. time

for Si(100) etched at4 mA/cm2 in (a) 2 mol/1 HF MeCN for 10 mm(this P1 onsets within lOs), (b) 6 mol/1 HF in H20 (20% HF in H20)for 12 mm (this Pt onsets within %

C/)Cci).-C Etch

(20%HF in MeOH

OO10 20 4050 60

10 20 30 40 50 60

3288 J. Electrochem. Soc., Vol. 145, No. 9, September 1998 The Etectrochemicat Society, Inc.

Dilute HF (20%)/water etch—In sharp contrast to thePL from PS formed in a near-anhydrous etch, the rise inthe 5200 A PL is immediately apparent and, at least ini-tially, closely parallels the rise in the 6200 A PL when theetching cycle is carried out in a water/20% HF (6 M HF)etch solution. As Fig. 4b demonstrates, the PL at first risesslowly within the first few tens of seconds of etching as itundergoes a seemingly short gestation period before in-creasing much more rapidly. The PL is clearly observableat both 5200 and 6200 A after 1 mm of etching and is com-parable until the 5200 A feature begins to lag in intensityapproximately 7 mm into the 12 mm EC etching cycle.After the etching current is terminated, the PL intensitycontinues to increase but the 5200 A emission increases ata notably slower rate than does the 6200 A emission. If thePL were displayed as a full spectrum vs. time, it would

Fig. 5. (a, top) Pt intensity (arbitrary units) vs. time for the etch-ing of Si(100) in a nonaqueous solution of 2 mol/L HF in acetoni-trUe (MeCN) for a series of three etch cycles at 24, 11, and24 mA/cm2. For all cycles, the Pt intensity at the end of the etchcycle (620 nm) rises sharply and then drops sharply with the onsetof the etching process. It rises more rapidly (during etch) for the11 mA/cm2 cycle. The time gate for sampling the Pt extended from0.8—100 p.s. (b, bottom) Pt intensity (arbitrary units) vs. time forSi(100) etched in a nonaqeous 2 mol/L HF/MeCN solution beforeand after introduction of 15 p.t of H20 to 2 cm3 of the anhydrousHF/MeCN solution. The stabilizing effect of H20 on the P1. intensi-ty is apparent. Time gate for sampling at 620 nm, 0.8—100 j.is.

I I

-e I t..c

--.it.-f r520 ore PL Signal

620 nor P1 Signal

-I—' ' I''I• —L • ' 1 ' I-30 0 30 60 90 120 150 10

MinutesFig. 6. The effect of adding 40 pi of H20 to 2 cm3 of a 2 mol/L

anhydrous HF/MeCN solution. The Pt intensity (arbitrary units) at520 and 620 nm excited by a nitrogen laser for Si(1 00) etched at6 mA/cm2 followed by the periods of (i) Pt growth, (u) decreaseimmediately following water addition, and (iii) subsequentincrease with wafer stabilization. Time gate for sampling0.8-100 p.s.

appear as though the PL peak wavelength intensity wasshifting to the red and growing more intense with time.The peaking of the 5200 A (--18 mm) and 6200 A (--23 mm)PL in time is found to be consistent over a broad spectrumof etching cycles with etching currents varying from 4 to12 mA/cm2. The PL intensity for both the 5200 and 6200 Aregions decreases to a negligible value within 50—55 mm ifthe PS sample remains in the water/20% HF solution.After an initial period, the orange-red PL always domi-nates the green PL in this water/HF solution. The conver-

240

200

160

.' 120(I)

8040

0

4000 5000 6000 7000 8000

Wavelength (A)Fig. 7. P1 spectra obtained with a KrF excimer laser in air as a

function of time after a PS sample was etched in near-anhydrous2 M HF/MeCN for 10 mm at 8 mA/cm2, removed from the etchsolution, rinsed in methanol, and dried in air. An initially weakorange-red Pt increases significantly with time. Spectra are cor-rected for system response. Time gate for sampling 0.8—100 p.s.

__Add 40sl Water to 2ccetching solution

L) Illa '/HI.. I

ca)C-J

4dC00C00a-

100

Minutes

4 Etch Cycles

Add 15 iL of Waterto etch solution

25000

20000

co 15000000-c0

StartEtch

Time (hours)


sion of the green PL can, however, be impeded through theintroduction of ethylene glycol.

Dilute HF/water/methanol etch.—When using a 6 M HFin methanol solution, we find that both the 5200 and6200 A FL (Fig. 4c) are easily observed very early in theetch process as the FL intensities, which parallel eachother, rise rapidly and are clearly apparent after only a fewseconds. Further into the EC etching process (—1—4 mm),the orange-red photoluminescent intensity begins to in-crease at a more rapid rate than does the 5200 A PL, abehavior similar to that displayed for the HF/H20 etchsolution (Fig. 4b). After the etching cycle is terminated, the6200 A PL continues to increase while the 5200 A FL peaksand begins to slowly decay.

The cycling behavior observed upon etching in thesemethanol/HF solutions is also distinct from that character-izing the anhydrous MeCN/HF solutions. Figure Ba and bdemonstrates the observed behavior for concurrent 12 mmetch cycles of 8 or 24 mA/cm2 where the PL is allowed todecay before each new EC etching cycle. The FL at 6200 Arises quite sharply for all the etch cycles; however, the

3.0E4

2.5E4

o 25 50 75 100 125 150Relative Time (Minutes)

Fig. 8. Cycling behavior for a 6 M HF in MeOH(oq) (20% HF inMeOH) etching solution. The observed behavior is demonstratedfor several etch cycles in (a, top) and in closeup (b, bottom) for con-current 12 mm etch cycles of 8 and 24 mA/cm2 where the P1 isallowed to decay before each new etching cycle. Several trendscan be noted.

behavior observed as a pristine wafer, first placed in theetching solution for 2—3 mm, undergoes EC etching is dis-tinct. We observe a slope rise for an 8 mA/cm2 etch whichcan be seen to exceed that of following 8 mA/cm2 cyclesand a subsequent dip and slight rebound in the PL intensi-ty before the conclusion of the EC etch cycle. Following ECetches on the same sample are found to display a slightlydiminished initial slope increase in the PL intensity, fol-lowed by a clear slowing of the monitored rate of emitterformation; however, the PL intensity now consistentlypeaks at the conclusion of the EC etch. In all cases, the ter-mination of the etch process is followed by a dip in the PLintensity In addition, the PL is seen to quickly rebound.There is a clear trend to an increased rebound intensity asthe 8 mA/cm2 etching cycles are repeated on a given sam-ple. The 24 mA/cm2 etch following the third 8 mA/cm2 etchdisplayed in Fig. 8 leads to an expected increase in PLintensity; however, this increase considerably lags behindthe increase in etch current density and is followed by aconsiderably lower rebound intensity Similar behavior isobserved at 5200 A, although the dip in FL intensity fol-lowing the EC etch is considerably muted. Finally, weobserve a rapid decrease and quenching of the PL in theHF/MeOH solution. As Fig. 4c demonstrates, the PL inten-sity is quenched within 25 mm. Additional experimentsinvolving the introduction of small quantities of methanoland ethanol into an HF/H20 etch solution suggest that thismay result from a methanol/HF interaction to produceintermediates which enhance the dissolution of the PS sur-face. It is to be noted, however, that the introduction of eth-ylene glycol5 to this HF/methanol etch solution again slowsthe transformation of the green precursor state to the finalorange-red PL emitter while stabilizing the FL for extend-ed periods.5

Finally we observe that the effects of multiple etchingcycles are reflected also in the ex situ FL observed aftertreatment in HF/MeOH. Figure 9 displays the ex situ PLobserved as a function of etch time in solution for a PSsample prepared in HF/MeOH at an etch current of10 mA/cm2, rinsed in methanol, and dried in air. Here theex situ FL intensity develops more rapidly than does thatfor the anhydrous etching process, and we observe a clearred shift of the FL with increased etch times.

50% HF/methanol etch.—The etching of FS in a solutionat considerably higher HF concentration in MeOH producesa significantly different FL response with time. Figure 4d

Fig. 9. Ex situ P1. emission excited by a KrF laser os a function ofetch time in solution for a PS sample prepared in a 20% HF inMeOH(aq) etch at a current density of 10 mA/cm'. Etch times areindicated. The PS samples were rinsed in methanol and dried in airafter each etch. Spectra, which were scanned from 4000 to 8000A over a time frame of 4 mm, are corrected for system response.The error l?ounds for system response correction at wavelengths A> 7600 A are significant. The time gate for sampling was0.8—100 p.s.

TerminateEtch

TerminateEtch

TerminateEtch

Ce

C0C-)C000CeC0)C

C')

C

00C00.c0>1U)Ca,4-C

50 100 150 200 250 300

Time (Minutes)

2.0E4

1.5E4

1.0E4

5.0E3

0.OEO

0

3.0E4

2.5E4

2.0E4

1.5E4

1.0E4

5.0E3

0.OEO

24

20U,CD

U,C0C

— 0 minutes- - 2.5 minutes— 5.0 minutes- - - 7.5 minutes

10.0 minutes

12

8

4

04000 5000 6000 7000 8000

Wavelength (A)

3290 J. Electrochem. Soc., Vol. 145, No.9, September 1998 The Electrochemical Society, Inc.

depicts the time dependence of the PL at 6200 A for a p-Si(100) wafer etched in 14 mol/L HF in MeOH. No PL wasobserved during the etching process. After the etch was ter-rnrnated, a gestation period of approximately 24 mm pro-ceeded the observation of any visible PL. Clearly, the surfaceof the PS is undergoing significant chemical changes duringthis extended time period to produce the luminescence.

in situ PL spectra as a function of etch solution.—Photoluminescent spectra from PS, formed and allowed toremain in the etching solution for limited periods, aredepicted in Figure 10. For the three solutions considered,the silicon was etched for 12 mm at 8 mA/cm2. The PLdata were recorded while the samples remained in solu-tion, 2—3 mm after the etching current was terminated forthose samples placed in a near-anhydrous MeCN/HF solu-tion and in a 20% HF in MeOH solution and —24 mm aftertermination for the 50% HF in methanol solution. Theratio of the PL at 6200 A to that at 5200 A for theMeCN/HF etching solution can be clearly assessed fromFig. 10. These spectra, obtained for the wetted PS samples,are clearly blue-shifted relative to the spectra depicted inFig. 3 and obtained for similarly prepared PS samplesdried in air. In other words, the drying of the PS samplecaused the peak spectral emission wavelength to shift byseveral hundred angstroms to longer wavelengths for eachof the basic solutions used in this study The red shift inthe PL emission with increased exposure to air must beintimately tied to the transformation of the emitting medi-urn on the PS surface.

Effect of etch current density—Figure 11 exemplifies theeffect of etch current density on the PL observed during theetch cycle. Here the silicon was etched in a 6 M HF/H20solution at a current density of 10 mA/cm2 for 70 mm. The6200 A PL clearly increased in intensity during the etchingprocess. As the current density was increased to 16 mA/cm2,a sudden increase in the 6200 A PL was observed, followedby a continuous albeit less pronounced increase in the PLintensity with etch time over the next 60 mm. 'When the etchcurrent was increased to 30 mA/cm2, the PL intensityincreased quickly followed by a rapid decay over the next70 mm, suggesting the onset of electropolishing. A similar

In Situ P1 observed after etching fri aJ 50%HF ri MeOH,b) MeCN /HF, and c) 20%HF in MeOH

Fig. 10. Pt. spectra (in situ) For porous silicon exdteci by a nitro-gen laser after etching Si(100) for 12 mm at 8 mA/cm2 in (a)14 mol/L HF in MeOH(aq) (50% HF in MeOH); (b) near-anhydrous2 M I-IF in MeCt.1; and (c) 6 md/I HF in MeOH(aq) (20% HF inMeOH). These in situ spectra taken at a time gate for sampling of0.8—100 p.s should be compared with the ex situ spectra in Fig. 3.

Time (Minutes)150 180 210

Fig. 11. P1 emission intensity as a function of etch current andtime for the etching of Si( 100) in a 6 M HF/H20 solution at the indi-cated etch currents.

behavior, although at slightly higher etching currents, isalso observed for the 6 M HF/H20/MeOH solution.

The influence of additives on the in situ PL fromPS—Post etch treatments.—The varied nature of the in situPL characteristic of the basic etching solutions which wehave outlined has also focused our attention on the effectof additives to the etching solution and hence on post-ECetch treatments. The stabilizing effect of small quantitiesof water on the anhydrous etching process,7 as well as theeffect of ethylene glycol5 on the green and orange-red PLfrom a sample etched in HF/H20 and HF/H,O/CH;3OHsolutions are exemplary.

We have found that samples prepared in any of the fourbasic etching solutions, once cleaned in a combination ofdoubly deionized water and methanol or ethanol andallowed to dry, yield a significant ex situ PL. If a cleanedsample is immediately placed in doubly deionized water,we observe a slow rise in the in situ PL intensity at both5200 and 6200 A. Similarly, if such a cleaned sample isplaced in an ultrahigh-purity methanol or ethanol solu-tion, the in situ PL slowly increases and gradually dimin-ishes on the time scale of several hours. However, thisbehavior is in sharp contrast to the effect of immersing thePS in a combination of the alcohols with HE

Post etch treatment with ethanol—The PL induced by anaqueous HF etch [e.g., HF (20%) in H2OJ can be readilydepressed through the postetch addition of ethanol to thissolution. The 6200 A PL is most affected, showing a defi-nite decay upon addition of ethanol to the etch solution.However, after displaying an initial dip in intensity, the5200 A PL is observed to recover to virtually its initiallevel. The effect of ethanol can, however, be more diverse.We exemplify with the results for a silicon sample firstetched in a solution of 20% HF in water for 10 mm i ol-towed by a subsequent etch after the addition of ethanol tothis solution. The 6200 A PL intensity in the HF/H20 solu-tion increased after the etch was terminated as exemplifiedin Fig. 4b. Depicted in Fig. 12 is the 6200 A PL as a func-tion of time when 700 pLof ethanol was added to 2 mL ofthe 20% HF/H20 etching solution and this solution wasused to etch the silicon sample for a second time. Althoughthe 6200 A PL increases more rapidly during the etch peri-od, after the cessation of the EC etch, the PL intensitydecreases much more rapidly and is almost completely

PL at 6200 Angstroms with continuous etching

(I).4-'C0

C-)

C0.4-'0a-

10 mA/cm2(4

3OmA/cm2

03 60 90 120

C

ft

b

C

5000 doo 'oooWavelength (Angstroms)

7d00 75Q0


HF in H20 with 700.tl EtOH added

20% HF in H20

6200200

20 30 40 00 60

Fig. 12. Pt emission as a function of time showing the effect ofadding ethanol to a 20% HF/H20 etching solution. Here 700 p.1 ofEtCH was added to 2 ml of the HF/H20 solution previously used toetch a Si(100) sample and the subsequent sqlution was used toetch the silicon again. We compare the 6200 A Pt emission inten-sity to that depicted in Fig. 4b.

quenched within 30 mm. This behavior, while not as pro-nounced, closely parallels that observed for a silicon waferinitially placed in an HF/H20/MeOH solution (Fig. 4c).

Postetch treatment with methanol—Figure 13 demon-strates the diverse quenching effect which the simple alco-hols can have on the PS PL. The data depicted in the figurewere obtained using a concentrated HF (48%) etch upon theaddition of methanol. A PS sample was produced from a14 M HFJH2O solution etched for 12 mm at a current densi-ty of 8 mA/cm2. The PL decay was then monitored after theEC etching process was complete and while the PS sampleremained in the etching solution. After approximately50 mm of exposure to the HF solution, 100 p.L of methanolwas added to 2 mL of the concentrated HF solution and animmediate loss (Fig. 13) in PL (6200 A) intensity wasobserved. At approximately 68 mm after the termination ofthe EC etch, a second 500 jiL aliquot of methanol was addedto the etching solution and an extremely rapid quenching ofthe PL was observed. At 5200 A, a similar albeit somewhatslower quenching was monitored.

C

00C00-c0

It is important to emphasize that if we allow a PS sam-ple to remain in the HF/H20/MeOH etching solution afterthe EC etching cycle is complete, the PL will be quenchedcompletely within 15 mm. We find that the quenching cycleappears to follow the reverse of the development of the PL(Fig. 4) as the orange-red PL quenches much more rapidlythan the green PL. Only after the orange-red PL is com-pletely quenched do we observe the complete quenching ofthe green FL5 in an HF/MeOH-based solution. We have alsoobserved PL from an HC1/H20 solution following an etchcycle. In sharp contrast, the addition of methanol to thisHC1/H20 etching solution does not appear to quench theFL but may in fact provide some enhancement. In furthercontrast, we find that a strongly photolummnescent samplecleaned of HF and placed in MeOH will exhibit a signif i-cant FL for several days. If we add a single drop of HFetching solution to a previously etched PS sample as itsoaks in MeOH, the PL is rapidly quenched. The results inan HF/MeOH environment, obtained when methanol isintroduced into an aqueous HF etching environment orvice versa, are to be contrasted with a virtually completelack of quenching when MeOH is introduced into a non-aqueous HF/MeCN etching solution. It seems apparent thatthe interaction between methanol and to a lesser extentethanol and HF in an aqueous solution leads either to theformation of a solution phase compound which very effec-tively quenches the triplet exciton excited state giving riseto the PL from PS or accelerates the PS surface oxidationor dissolution process.

Postetch treatment of PS samples in methanol, H20—NaF.—The importance of fluoride ion to the dissolutionprocess is clearly outlined in Fig. 14. Here we indicate theeffect of the introduction of NaF prepared in a saturatedmethanol solution (—0.4 M), to PS samples photoluminesc-ing in air and a methanol solution. As Fig. 14a and b indi-cate, the postetch FL from a porous silicon sample placed inmethanol solution increases somewhat more rapidily thandoes that for a sample placed in air. The introduction of NaFimmediately quenches this FL on a time scale of seconds.Fluoride ion is clearly a very effective quencher. The con-sideration of diffusion effects in the etching solution sug-gests that the quenching process results from the extremelyrapid dissolution of the top few surface layers which con-tain surface-bound emitters.

Post etch treatment with HC1—PL enhancement and stabi-lization—Figure 15 depicts the in situ effect of water andHC1 on the orange-red PL from a PS sample previouslydried in air. The PS sample was first etched in a methanol!20% HF [6 mol/L HF in MeOH(aq)] solution for 10 mm,washed with methanol, dried in air, and then placed first ina solution of doubly deionized water and then —6 M HC1.The intensity of the FL increased somewhat upon removalfrom the HF solution and again increased more rapidly asthe sample was placed in water.78 However, after addingHC1,78 the orange-red FL intensity increased significantly(Fig. 15) and remained virtually constant over the period(up to 3.5 h) in which the sample was present in the HC1solution. Within the time frame of these described process-es, the wavelength-dependent spectral profile of the PLemission spectrum was only slightly altered from the time itwas removed from the HF etching solution and driedthrough the period in which the sample remained in the HC1solution. The introduction of HC1 clearly has a profoundstabilizing effect on the in situ PS FL.

We have also examined the influence of the heavieracids HBr and HI on the nitrogen laser excited FL from aPS surface.8 In contrast to the apparent stabilizing effectof hydrochloric acid on a PS surface photoluminescing indoubly deionized water, HI almost completely quenchesthe PL. The effect of an HBr solution is intermediate.

We have also examined the effect of HC1 introductioninto an HF/H20 etching sOlution in the early stages fol-lowing the etching process. Again, the enhancement andstabilization of the 6200 A orange-red PL is apparent;however, we find that the effect of HC1 on the 5200 A flu-

0 1

PL at 6200 Angstroms

lOOpL MeOH

60Minutes in 48% HF

Fig. 13. Plot of 6200 A Pt intensity as a function of time whenmethanol is added too sample prepared by etching silicon (100) ina concenfrated HF solution. The methanol enhanced decay of theemission is apparent. See text for discussion.

3292 1 Electrochem. Soc., Vol. 145, No. 9, September 1998 The Electrochemica! Society, Inc.

200000C00

1000

orescence, which rises much more slowly, is far less pro-nounced. HC1 appears to preferentially stabilize theorange-red emitter.

Postetch treatment of PS samples in methanol—NaCI.—The intriguing stabilization of the PS FL in an HO solu-tion focuses our attention on the effect which chloride ionhas on the FL process. As the data in Fig. l6a and bdemonstrate, a stabilization of the PL emission rate froman etched PS surface, albeit at a notably lower photoncount, accompanies the addition of NaC1 to a luminescingPS sample. In fact, an intriguing behavior is monitored. Aswe have noted, in the absence of Mad, the 6200 A PL fromPS placed in a methanol solution slowly rises, eventuallypeaks, and then more gradually decreases in intensity. IfNaC1 is added to this solution before the FL has reachedits maximum intensity in methanol, the PL slowly increas-es to a maximum intensity and then plateaus. As Fig. iSademonstrates for a PS sample introduced to a NaC1-satu-rated solution in methanol, this plateau is reached over atime scale of several hours (compared to 30 mm for HC1).If, however, the saturated NaCl solution is introduced to a

Effect of Water and25000 HCI on 620 nm PL

20000

Equal volume of HCI added to water

Time (Hours)

Fig. 15. The effect of an HCl solution on the stabilization of the PSP1 for a sample previously dried in oir and placed in doubly deion-ized water. The wavelength monitored was 620 nm and the datawere taken using nitrogen laser excitation. Here the Pt rises slowlyin air and water and then increases precipitously in HCl solution.

methanol solution in which the PS sample is previouslyplaced, as indicated in Fig. 16b, after the PL has peaked inthe methanol solution, the PL is stabilized at an intensitycorresponding closely to that at the time of the NaCl intro-duction. This is certainly an intriguing result which por-tends means to stabilize the PL from porous silicon withappropriate chloride solutions.

Trends in ex situ PL spectra.—From the data in Fig. 3 and9, it should be clear that the ax situ PL spectra (i) red shiftin air relative to the observed in situ luminescence and (ii)red shift with the duration of the overall etching processand are affected by a previous etch cycle. We observe that asample, prepared in a near-anhydrous in situ environment(MeCN/HF), after it is washed (methanol), dried, and placedin air, displays a considerably weaker initial luminescencethan does a sample prepared in an aqueous HF solution(20% HF/H20, 20% HF/MeOH). Samples prepared in aque-ous HF generally display a clear luminescence immediatelyafter ex situ preparation [washing (methanol), drying inairl. With time, however, the PL from the sample generatedin the anhydrous MeCN/HF increases and may eventuallysurpass that from the sample generated in aqueous etch. Infact, the most intense ex situ PL which we have observedresults after etching in a near-anhydrous solution. Theex situ spectra which we have observed do not appear toshift with increasing laser power. In fact, their intensityscales linearly with incident laser photon flux, demonstrat-ing that the observed behavior does not result from multi-photon processes. Finally, we observe that the ex situ FLspectrum in air, in time will generally equalize regardless ofhow the sample is prepared. The ex situ PL graduallyincreases in intensity over the course of several hours andthen plateaus in intensity as the spectrum red shifts.

DiscussionWe have characterized several environments in which

FL can be generated on an etched PS surface and haveestablished conditions under which this FL can be stabi-lized, enhanced, or quenched. The experimental resultswhich we have outlined are summarized in Table I. Weargue in this discussion that the origin of the visible FLemission from PS is reasonably associated with a surface-bound chemical entity and that the emission correlateswith a transition originating from a triplet exciton, asestablished by Stutzmann and co-workers.2320 We suggestthat this surface-bound source can be viewed as asilanone-based silicon oxyhydride emitter bound to the PSsurface. In this discussion, we evaluate the formation ofthe photoluminescent sources and their subsequent loss byeither their oxidation to Si02 or dissolution in the HF-based etching solutions investigated in this study.

Sample inAir

Add NaF(MeOH(5.5g NaE/3Dcc MeOH)

5000

4000

3000

2000

1000

3000Plcl,ed in 20%I-I Fi MeOH

Sample inAir

Sample placed inMeOH

Add NaF

C00C000

0 15 30 45 60 75 0 30 60 90 120 150

Time (minutes) Time (minutes)

Fig. 14. Effect of introducing NaF (fluoride) to the PS etch solution (a, left) for o sample etched in 6 M HF in MeOH(aq), rinsed in methanol,and dried in air for - 12 mm and then placed in a saturated NaF/MeOH solution, and (b, right) for a sample etched in 6 M HF in MeOH(aq),placed in air for 10 mm and then in methanol for —90 mm, followed by addition of NaF. P1 intensity increases after removal from theHF/MeOH etch solution in either air or methanol and is quenched immediately upon introduction of fluoride ion.

U)C00C04-0-Ca-

15000

10000

5000

II 0.5 1.0 1.5 2.0 2.5 3.0 3.5

It has been established that PS is formed by an oxida-tion-dissolution mechanism where silicon oxides, currentdensity, and crystallographic structure play a pivotalrole231 and the rate of growth of the oxides and porosity ofthe structure are functions of solvent composition. WhereasPS fabricated from an aqueous electrolyte consists of high-

Table I. Summary of in situ and ex situ experiments.PS Samplein...

Change(Addition)

5200Aintemity

ZOOAintensity

Comments

AithydrousHF/MOCN richsolution

Etch WrIter than6200A

Figs.rr 4).)Sun also Figures7,10

MCCN/HF I5ct HO2cc solution

Sigmtcusitstabilization

Figure 5(b)

MrCN/HF Etch curronidensity

Ralu ofiocreasr/decrease withinceeaeud 1

Figure 5(a)

MeCN/HF 40p/ (lO2w sOlution

Swall change butiticresan

Initial dropoffand thenrebound toslightly highermccl

Figure 6

MeCS1/HF Icc HnO2cc solution

Spectea showgenes do,n.n.aenlost few minutes— decays

Orange tooge-tndspevosl shift

E.haaootowsi of FL in sun(sOn alan Pigeon 7)

MeCN)HF 12cc) +

20p1 ItsoMrOH Slight

eshoacewentSlabdieaaon owhigher nwi.uis.bert

Eaiance..eat of Ft sir(san also Pigeon 7)

HF (20%)nwater - etch

_______Comparable to6200A

Figton 4(b)

HF (-20%) inwateni?s4EOH etch

Cnanp.enble to6280A

Fsstee flue aed decay thaaHF)1150 - Pigeon 4(c) - Etchcycle behavior Pigores 8.9.to

ofF (-50%) inMeOX-etch

rdrgligtblndensity

Obeonned long gnstaaon penod foe FL— Fiss 4(d). 10

HF (20%) inwater—etch

Currentdensity

bocteo.so in Pttsp *030etA/crot

Oeiset of elecmropolisbingPigeon II

rIP (20%) in HcO(2cc)

500salethylenegl)nol

Stahi)inatson ofemission

Stokilie.entoe ofemission

Te.risfonssalion penn—vos.ogn.rrdPtissloweslcnesidne.hly (Ref. 5)

HF (20%) in HsO(2cc)

700.gethoowI

Iwoot denpoffand thee slightrscnnei — notsentsg effect

Dropoff in PtintensIty

Sswpk ha. sheeter Itfesoneoften footer nine in intensity —Pt is less t.tcMe in,ubmqarnt etches - Firm 17

HP cosicenleased(40% by volume)

_________Methanol Quenches Quenching of

PtMeOH sad HF iscnnnibiealinn act to quench PL(Fig (3)

HF (20%) in HnO.E1OH added

Elch withadded EtOK

Significantintoitsitydecressn

Significantintensitydecrease

Air HtO Slew von Slow den See Figures (4(u), 4(b). (5

HiO HC) Slow riseeventuallypluicous

Rapid nonecrntoollyplateaus

Son Figure 5

AirAir

litsF/MrOHMnOH

Raged quenchIttenotlyincreases toplateau - Izior

Firm 14(a)Figaen 4(b)

MethanolMethanol

HopHoC)

Raged qiuenchStabilization stlower photon)rcnl

Pigeon (4(a)Addition of MaCI beforepeaking of P1 in methanolFigure (6(o)

Mnthusn( MaCI Stahilicasion atFL level ofnlrodaction

Addi005 of M.d oboepeaking of FL in niedtantnl!re 16(b)

8sF (20%) in HiO2cc)

I

Reverse etchcunrtnn

pwlunty

Oncresae ofintensity andJuralion of Pt

I4gorc 18

ly nanoporous branched structures, PS fabricated from anonaqueous electrolyte on p-type silicon in the absence oflight (required for n-type silicon) is comprised of open andaccessible macroporous structures with deep, wide, andwell-ordered channels.32 Random nanopores, with poten-tially far less accessible photoluminescent PS structuresare generally formed in the aqueous etching process.

Emission from the nonaqueous etched material estab-lishes that very fine nanoporous structures are not a pre-requisite to induce PL in PS. In fact, we have observed PLfrom both aqueous and nonaqueous etched samples priorto the formation of any pores.'2 We find that much lessthan 100 nm of silicon need to be removed before the lumi-nescence centers are formed. The topographical changes inthe surface are consistent with the formation of emitting,constrained surface species. As the open structure of PSprepared by nonaqueous etching results from an oxida-tion-dissolution process, small quantities of water arerequired for the onset of PL. This suggests that con-strained emitting species may also be formed on the wallsof the macroporous structure and that the oxyhydrides27'28may play an important role.

Quantum confinement vs. surface-based chemicalchange—As the morphology of PS depends on etching con-ditions,33 and the morphological changes appear to parallelchanges in PL intensity,34 the source of the observed PL hasbeen linked to a physical change in the silicon microstruc-ture. Quantum confinement suggests that the PL originatesfrom nanosize particles of silicon formed during the ECetching cycle.'182° Silicon particles with diameters less than50 A have been reported.35 Further, correlations of the wave-length of maximum luminescence with porosity36 also sug-gest the possibility of size confinement and nanocrystallinesilicon particles37-39 produced by techniques other than ECetching have exhibited properties similar to porous silicon.38While enticing, the quantum confinement hypothesis ischallenged by other experimental observations. For exam-ple, size confinement cannot account for the observed redshifts as a function of etch time (see Fig. 9) which are mon-itored upon the drying of the PS sample and under someoxidative conditions.4° In these experiments, the size of sur-face silicon particles must become smaller yet, counter tothe expected manifestation of this behavior, a red shift inthe PL is observed. Further, chemical change must be occur-ring which contributes, at least heavily to (dominantlyinfluences) the observed behavior.

Recent studies of the PL emission as a function of exci-tation energy in our laboratory indicate that for a selectionof laser excitations over the wavelength range extendingfrom 193 to 400 nm,9 an almost identical time-dependent

620 nm P1 in NaCI Solution

3293J. Electrochem. Soc., Vol. 145, No. 9, September 1998 The Electrochemical Society, Inc.

2000 8000

1750 7000

isoo 6000

.t— 1250 — 5000C C

8 1000 8 4000

c 750 c 300002 500 2 2000a-

250 1000

0

Add NoCI

Etched sample inMeOH

60 120 180 240Time (minutes)

360 0 120

Fig. 16. Effect of NaCI introduction on a luminescig PS sample in methanol. (a, left) The NaCI prepared as a saturated solution (—0.2M) is added to a sample photoluminescing at 6200 A in MeOH before the PL has reached its peak intensity. The solution stabilizes at thepeak PL intensity although at a much longer time scale than,that for HCI (Fig. 15). (b, right) The NaCI prepared as a saturated solution (0.2M) is added to a sample photoluminescing in MeOH (6200 A) after the Pt intensity in methanol has peaked. The solution is stabilized at theintensity upon addition.

Time (minutes)

240

3294 J Electrochem. Soc., Vol. 145, No. 9, September 1998 The Electrochemical Society, Inc.

PL emission feature results. This behavior is inconsistentwith quantum confinement and suggests the pumping ofthe excited state manifold of a molecule-like species, fol-lowed by rapid, nonradiative relaxation through the excit-ed state vibrational manifold and the subsequent emissionof radiation at much longer wavelength emanating fromconsiderably lower excited state levels.

Our studies of the FL emission as a function of excitationenergy support the conclusions of Kanemitsu et al.,41Frokes et al.,2'3° and Macauley et al.42 Collectively, thesestudies demonstrate that no obvious relationship exists be-tween the PL spectrum and either the PLE spectrum or theparticle size distribution. Kanemitsu et aL4' have used opti-cal absorption, Raman spectroscopy, and transmissionelectron microscopy to assess the relationship between par-ticle size and PL. They found a blue shift in the optical ab-sorption spectrum for particles decreasing in size from 9 to2 nm but observed no change in the corresponding FL peakenergy. This result clearly demonstrates that the PL peakemission energy is not correlated with particle size. Ourtime-dependent studies, over the 0—100 p.s time frame,9suggest the manifestation of a pumping sequence betweenthe ground state and strongly shifted excited state5'9 of asurface-bound fluorophor as the pumped excited state lev-els, high in the excited state manifold, subsequently un-dergo rapid nonradiative relaxation. Thus, a cascadeprocess can lead to a red shift of the PL emission spectrumas a function of time following optical excitation at muchshorter wavelength.43 This in turn suggests9 that the emis-sion observed over a 0—100 p.s time gate may not representthat from the lowest vibrational—rotational levels of thestrongly shifted excited state. With this caveat we alsoestablish that the time-dependent evolution of the FL in airis quite similar regardless of the manner in which a sampleis prepared. The combination of these studies, in concertwith the work of Prokes et al.2'2° and Macauley et al,,42 pro-vide strong evidence suggesting that a quantum confine-ment process does not represent the dominant source of theobserved room temperature FS FL.

A number of researchers have recorded a blue shift in thePS photoluminescent emission spectrum that parallels in-creasing etch effectiveness. They have attributed this behav-ior to the creation of quantum confined silicon particulateswhich decrease in size with increasing etch time. However,one might also attribute the observed changes to the role ofsurface curvature (tension) and structural strain as it influ-ences the PL process by creating local force gradients. Infact, Unagami44 has measured the intrinsic stress in poroussilicon layers formed in an aqueous HF etch, finding amacroscopic stress which extends from compressive to ten-sile. Friedersdorf et al.45 have observed the influence ofstress on the FL of PS structures, correlating the blue shiftin PL peak energy with surface microstructure, and con-cluding that the characteristic cellular structure occurringin high-porosity films is due to high surface stresses. Thiscellular structure is not formed during the etching processitself but occurs during electrolyte evaporation after re-moval of the FS from solution. Enhanced surface tensionsassociated with increased curvature appear to be operative.Finally, in close agreement, Kanecko et al.46 have veryrecently noted that while an enhanced etching cycle leads toa blue shift in the PL spectrum, the higher porosity structurecreated is clearly strained. While these authors note that theobserved blue shift suggests quantum effects, the surfacestate of the porous layer also is suggested to play an impor-tant role. The changes observed in the FL emission spectradepicted in Fig. 3 and 10 as a function of etch solution canalso be readily attributed to similar surface-based effects.

Finally, we note that Kanemitsu et al.41" provide someindication that the FL intensity does increase significant-ly with decreasing particle size. This trend, which mightbe expected on the basis of quantum confinement,47 mightalso be explained by a decrease in the PL intensity result-ing from an increase in the energy transfer quenchingwithincreased effective particle size to which an emitting sur-face-bound fluorophor is bound. Further, it might be

argued that the PS samples generated in anhydrous mediaare possessed of distinctly different effective structuralsizes, suggesting that differences in the PL intensity forthose samples prepared in aqueous and anhydrous mediaare not due entirely to surface chemical changes but to sig-nificant changes in the absorption process. However, it canalso be argued that the microstructure of the PS samplesgenerated in anhydrous media is accompanied by a signif-icant surface roughening which can create confinedregions and which results, in part, from the presence ofsmall quantities of water which enhances the FL. Further,it would also be surprising to find that a significantchange in the absorption process does not produce a sub-stantial change in the spectral distribution of the FL.

The characteristics which we observe as a function ofetch conditions and the outlined discrepancies with thequantum confinement model naturally cause us to consid-er the possibility that the source of the FL is chemical inorigin and surface localized. Here, the diversity of behav-iors which we observe as a function of etch solution andthe extremely rapid quenching of the PL which can beinduced by NaF (Fig. 14) through dissolution and HI8 byrapid heavy ion quenching are certainly also suggestive ofa chemistry dominated by interactions at the silicon sur-face—etch solution interface.

The triplet exciton emission which characterizes poroussilicon has been equated to the very similar emission spec-trum of "annealed" siloxene.22 However; this is a very com-plex species whose synthesis is tedious.23-25 Other re-searchers48 have shown that the luminescence is affected bythe silicon hydride and silicon oxide coverage of the sur-face. These observations, while seemingly counter to eachother, suggest that the oxygen and hydrogen functionalityon the silicon surface is important to the luminescenceprocess. We suggest that the surface functionality of"annealed" siloxene is best described as resulting from thesilicon oxyhydrides.

A chemical link.—The idea that a surface-confined moi-ety is the source of luminescence fits well with the FL dataobtained in the initial stages and throughout the process ofetching. The oxidation, complexation, and dissolution ofsilicon as it transforms from the zero valent state to thefully complexed, tetravalent state (either in the solutionsoluble or insoluble form) proceeds through a series ofelectron-transfer and chemical reactions. These electron-transfer and chemical reactions should occur sequentially,the rate of each being dependent upon the chemical envi-ronment, temperature, and electrochemical potential (inthe case of electron transfer). An appropriate reactionsequence involving two intermediate (metastable) states,A and B, can be represented by the following pathways

The exact chemical nature and rate of reaction of A andB is dependent on the activity of the chemical speciesavailable where X can represent a variety of ligands,including F, HF, H20, OH, and oxygen. The formal oxi-dation state of species A and B must vary between zeroand four. The species A and B might be viewed as differentluminescent centers. The effect of the different electrolytesand the experimentally generated potentials (i.e., currentdensity) is to change the rate of formation and reaction ofA and B.

For example, in the aqueous solutions, a high concen-tration of H30 is available which competes with F andHF in complexing the oxidized forms of silicon. At high-er fluoride concentrations, or under anhydrous conditions,the fluoride form of the silicon complex is favored as com-pared to the oxide or hydroxide. While it is impossible toquantitatively determine the intermediates and their rateconstants from the results in this study, the sequential

Si+y1H+ z1X

—x2e x3eA —- B -- Si(IV)+y2H +y3H+z2X +r3X


nature of the monitored reaction paths can be used as aframework for understanding the role of complexingagents and potential.

The initial rise of the green and orange-red FL withinthe first minutes of etching must occur concurrently withthe roughening of the surface and the formation of reac-tion intermediates at the silicon—solution interface. Whileon a short time scale the surface roughening may lead tothe formation of surface topological features consistentwith quantum confinement, the formation and dissolutionof unique chemical species is also possible. We contendthat the most likely chemical entity which is the source ofluminescence is a silicon oxyhydride.5'9 The rise in FLintensity with etch time for the basic solutions consideredin this study (e.g., Fig. 4) parallels the expected increase insurface coverage of silicon reaction intermediates, possi-bly suboxides. Fourier transform infrared (FTIR) studies'°show that during and after the electrochemical etchingreaction, suboxides in the form of Si(II) are indeed formedat the silicon surface. After the etching current has beenterminated, any silicon dioxide and, to a slightly lesserextent, silicon-based suboxide species, formed during theetching process, should be dissolved chemically by the HFsolution. Because oxide dissolution is expected to occur asthe FL decays in time, the behavior of the luminescenceshortly after the EC etch termination (Fig. 4) suggests thatthe luminescence is associated with the presence of someform of silicon oxyhydride on the FS surface.

The complete oxidation of the FS surface to Si02 resultsin a loss of the green and orange-red luminescence. Si02fluoresces on a much shorter time scale49 in the region near4500 A and thus cannot be the source of the emission. Infact, the results obtained by Henglein et al.,56'51 in theirstudy of luminescent colloidal silicon particles preparedfrom oxide-coated crystalline silicon formed in the com-bustion of silane, are relevant to our discussion. Henglein etal. found that the orange-red FL can be activated by theaqueous HF etching of colloidal silicon particles suspendedin a 1:1 cyclohexane—propanol-2 solution in the presence ofair. In contrast, there is no luminescence in the absence ofair (under argon). The required presence of oxygen (or smallconcentrations of water) indicates that the oxide layer cre-ated in the combustion of silane does not promote lumines-cence. Rather, as Heinglein et al. note, the development ofthe orange-red FL in nonpolar cyclohexane suggests thatthis original oxide layer must first be removed by HE Thesilicon particles, with a nonpolar surface, are created in thecyclohexane phase. Here an equilibrium is established be-tween their surface oxidation and reduction by 02 (H20)and HF, respectively. Near-pristine colloidal particles areformed whose oxidation, to a much lower level than theoriginal oxide-coated particles, produces the luminescencein a process which must compete with an eventual dissolu-tion through SiF4 (Si(OH)4) formation.7'3252'53

Based upon a study5° of the quenching of the lumines-cence in polar solvents, trimethylamine, ammonia, andsulfuric acid and the apparent stabilizing effects of non-polar solvents,7'32 Henglein et al.6° have suggested theimportance of a protolytic equilibria between the threesurface structures

with the neutral hydroxyl structure corresponding to thatspecies which can be expected to produce the orange-redluminescence. Thus, these authors suggest that theorange-red FL occurs when colloidal silicon particlescarry only a limited component of oxidized centers andthat the protonation state of these centers strongly affectsthe luminescence. This result is quite consistent with theformation of a silicon oxyhydride constituency encom-passing a required OH or OR ligand.

Fast quenching and quantum confinement—The quan-tum confinement theory is not consistent with theobserved decays as a function of etch solution (Fig. 4,5,8)and the fast quenching of the FL due to the introductionof NaF and HI9 to the etching solution. If the source of theFL is a hydrogen-passivated nanocrystalline solid, onewould expect only a minimal decay of the FL in the etchsolution as HF only chemically dissolves silicon oxidesand not silicon hydrides. As it has been demonstrated54that the surface of fabricated porous silicon is mostlyhydride-coated, consisting of (111) faces, the probabilitythat any nanoparticles on the surface are hydride-coatedis high. HF solutions react primarily with silicon—oxygenbonds to dissolve oxides. Therefore, those silicon nanopar-tides terminated with hydrogen should be only minimallyaffected by the HF solution. Further, any nanocrystallinesilicon particle terminated by an oxide coating would beexpected to decrease in size as the dissolution process con-cludes with a hydride surface termination. As a result, aslight decay in FL and not the complete elimination of FLwould be expected if the FL was associated with nano-crystalline surface structures.

The luminescence decays depicted in Fig. 4 and 5 canalso be demonstrated to be consistent with a surface-con-fined moiety as the source of the FS PL.5'7'9 A plausibleexplanation for the growth and decay of the green andorange-red FL is based upon the formation and oxidativechemical transformation of the silicon oxyhydrides.5Experimental observations of the time-dependent behav-ior of the in situ FL depicted in Fig. 4b and c combinedwith the quantum chemical modeling of the low-lyingelectronic states of the silanone-based silicon oxyhy-drides7'6 provides strong support for the hypothesis thatthe long-lived green FL5 and its subsequent transforma-tion to a final orange-red emission are associated with thechemical transformation of surface-bound oxyhydrides.

The details of our modeling study have been given else-where,5'926 however, the most relevant results are present-ed for the silanone-based silicon oxyhydrides in Tables IIand III. Further calculations were also carried out on theground and lowest lying electronic states of the silylenesand dangling electron SiXYZ compounds which varythrough the series from SiH2 to Si(OH)2. In Table II wesummarize the SiO bond lengths calculated at theMF2/DZF5'26 level for the ground state singlets and lowestlying excited state triplets of several model silanone-basedoxyhydride compounds which we have considered.6'9'26 Thecalculations on the SiXYZ compounds from SiH2 toSi(OH)2 26 demonstrate that these molecules possess excit-ed states whose lowest lying transitions all lie far into theultraviolet region and are therefore inaccessible to thelight sources used in this study.

The calculations on the silanones demonstrate a signifi-cant change accompanying the transition to a low-lyingexcited triplet state which produces a considerable length-ening of the SlO bond.66'56 In contrast, calculations on thecorresponding silylene isomers demonstrate much smallerchanges in the bond lengths and bond angles of the low-

Si(O)H2Si(O)H(OH)6

1.5451.537

1.7001.709

0.1550.172

Si(O)(OHl2 1.536 1.709 0.173Si(O)H(OS1H2)Si(O)SiH3(OSiH3lSi(Ol(OSjH3)2Si(O)H(SiH3lSi(O)(OH)(SiH2)Si(O)(SiH5)5

1.5371.5431.5371.5531.5431.560

1.7081.7121.7041.6951.7121.681

0.1710.1690.1670.1420.1690.121

a Bond length for the excited singlet ii 1.705 A.Bond lengths for HO—Si—QH silylene are 1.670 A for the

ground-state iinglet and 1.680 A for the exciied triplet.

-WW

+W

- Si-0-H -Si-0+H

Table II. Si=O bond lengths (A) for silanones at the MP2/DZP level.

Moleculer(SiOlSinglet

r(SiOlTriplet Ar(SiO)


Table Ill. Ground-state singlet—excited energy separation forsilanones (silylenes).

AE(S..T) IIE(S-T) XAdlabatj.Molecule (kcal/ntol) (cv) (nm)

Si(O)H5 60.1 2.60 475Si(O)H(OH) 70.9 3.07 403Si(O)(OH)2 71 3.08 402Si(rnH(0S1H5) 70.3 3.05 406Si(O)SiH5(OSiH9) 69.6 3.02 411Si(O)(OSiH3)5Si(O)(OH)(S1H5)Si(O)H(SiH2)Si(O)(SiH3)2

67.171.257.353.9

2.913.092.482.34

426401499530

HSiOH 38.4 1.66 744HOSiOH 64.2 2.78 445SiH3OSiOH 67.4 2.92 425

lying excited states which can he accessed via excitationspectroscopy in the appropriate spectral regions. The data inTable III indicate the calculated singlet-triplet separationsfor the model compounds of interest. It is noteworthy thatthe calculated locations of the silanone-based silicon oxy-hydride excited triplet states for Si(O)H(OH), Si(O)(OH)5,Si(O)H(OSiH5), Si(O)SiH5(OSiH5), and Si(O)OH(SiH5) con-taining an OH or OSiH5 moiety and the known peak wave-length of the porous silicon excitation spectrum (—350 nm)29bear a clear resemblance to the known singlet-triplet split-tings of the low-lying silicon monoxide intercombinationband systems.55 The large change in the SiO bond lengthsindicated in Table II for all the silanones in turn produces alarge shift in the excited state potentials relative to theground state. The relative location of the ground and excit-ed state potentials, especially for the OH— or OSiH5—bondedsilanones, promotes optical pumping high up the excitedstate potential and a significant difference in the peak of thePLE spectrum (—350 nm) and the observed PL emissionrange [—500—550 nm (green), —600—800 nm (orange)]. Whilethe experimental data are consistent with a most efficientexcitation pumping high up the excited triplet state mani-fold5'9 for these silanone-based compounds, there also is evi-dence for subsequent rapid nonradiative vibrational relax-ation through this excited state manifold before theemission9'26 of radiation.

There are additional striking aspects to the data inTables II and iii. First, from Table II we note that whenev-er an OH or OSiH5 group (Table II, lines 2, 4, 5, and 8)isbound to the silicon—oxygen bond, the calculated change, 4r(SiO), in transition is consistently a substantial 0.17 A.Further, an additional OH group does little to affect this dif-ferential change in the bond length accompanying the sin-glet—triplet transition. In the absence of an -OH or -ORgroup, the change is notably smaller, decreasing from 0.155A for Si(O)H5 to 0.121 A for Si(O)(S1H5). These are distinctdifferences associated with -OR vs. -R group bonding to thesilicon which can have a considerable affect on the relativeenergies of PLE excitation and PL emission spectra. Fur-ther, we note the very similar adiabatic energy differences(AE 3.05 e11) which characterize those singlet—triplet tran-sitions where an -OH or -OR group is bound to the siliconoxide bond. The shift of the excited triplet state relative tothe ground state singlet suggests that a peak maximum at—3.5 eV should be observed in the PLE excitation spectrumassociated with a silanone-based oxyhydride emitter con-taining an -OH or -OR group. While the virtually identicalenergy increments for those silanones containing an -ORgroup can be contrasted with the much lower requiredpumping energies associated with the series Si(O)H2 (0.155),Si(O)H(SiH5) (0.142), and Si(O)(SiH5)2 (0.121), these lattersilanones display a smaller change, Ar (SiO), in the SiO bondlength, which is expected to lead to a considerably smallerred shift of the PL emission feature relative to the PLEwavelength.5

The results of the outlined quantum chemical modelingsuggest that the unsaturated silanones with bound -OR hg-ands produce a much larger red shift of the PL emissionspectrum relative to the peak of the PLE spectrum than will

silanones with only -R group ligands. Excitation of theselatter silanones should be more efficient at longer wave-lengths, consistent with the experimental observations of astrong green in situ PL for Hg lamp vs. KrF and ArF opti-cal pumping. It is also to be noted that the unsaturatedbonding which we consider would appear necessary inorder that the electronic transitions of interest for these Si,0, and H compounds fall within the visible and UV regions.

The green—to—orange-red PL transformation —Thesources of the "orange-red" PL emission which corre-spond to -Si(O)OH or -Si(O)OR constituencies would beexpected to undergo dissolution at a more rapid rate thanthe -H or -R (vs. -OH or -OR) group terminated oxyhy-drides. This certainly follows the experimentally observedvery long time decay of the PL which can be followed visu-ally to observe that the "orange-red" PL will first decaycompletely followed by the subsequent loss of any residual"green" PL emission.

The green and orange-red emissions have been previous-ly assigned to interlocking exciton fluorescence and phos-phorescence emissions, respectively.19 The experimentaldata," however, favor a chemical transformation processinvolving two distinct surface species whose photogener-ated excited states have comparable emission lifetimes.'Although the overlap of the spectral features associatedwith the green and orange-red emitters prevents a readyevaluation of the radiative lifetime associated with thegreen emitter, its radiative decay routes do not appear tobe compatible with a fast fluorescence process and are farless rapid than that of the short-lived Si05 emission at450 nm,49 Our observations suggest a chemical conversionbetween two moderately long-lived (several microsec-onds)5'9 surface-bound constituencies and establish that anintersystem crossing process involving a short-lived greenand long-lived orange emitter, as suggested by others,1925seems far less likely Based on the results of our quantummechanical modeling, we assign the green and orange-redemissions to the silanone-based oxyhydrides in variousstages of oxidation.9 The growth of the FL with time dur-ing and for some period after the etch cycle is consistentwith an increasing surface coverage of the FL emitters astheir formation competes with dissolution.9'12 As the oxidedissolution occurs and the FL decays in time, the oxidizedsilanone-based oxyhydrides terminated with -OH or -ORgroups (the orange-red emitters) undergo dissolution at amore rapid rate than those -H or -R group terminated oxy-hydrides (the green emitters)."

Current density dependence and the silicon suboxides.—Current density also plays an important role in the etchingprocess. With the aqueous HF etching solutions, two etch-ing regimes exist and are defined by a critical current den-sity At j > j,, electropolishing of the silicon sur-face occurs. At j < j,,, pores are formed. At low currentdensities the electro-oxidation reaction ultimately resultsin dissolution of SiF4, whereas at high current densitiesdissolution occurs by formation of both S1F4 and S102 orSi(OH)4. Even at low current densities, the reaction of sil-icon with water (producing surface hydroxyl groups) issignificant. We have argued7 that the formation of hydrox-yl groups on the surface facilitates the formation of pho-toluminescence centers. Two reactive modes of water withthe silicon surface can be envisioned: (i) the formation ofsilicon hydride and silicon hydroxide bonds across adja-cent silicon surface atoms (Fig. 17), and (ii) the concertedreaction of a water molecule at a silicon atom site to formthe oxide,'° with the subsequent release of hydrogen or the

a The evaluation of a radiative lifetime for the green-emittingprecursor state can be complicated both by its continual forma-tion and by its chemical transformation during the course of anemission decay measurement.

Reports of enhanced orange—red FL emission after RTDOwhere the OH content is believed minimal might be deemedioconsistent with a surface-bound silanone structure. However,Yan et al.28 have certainly obtained evidence that ressdual hydro-gen exists in a thermally oxidized PS film, probably in the formof SiOH.

J Electrochem. Soc., Vol. 145, No. 9, September 1998 The Electrochemical Society, Inc. 3297

insertion of the oxygen into a silicon—silicon surface bond.At j the rate of further oxidation of the surfaceintermediates increases and the rate of dissolution as Si02or Si(OH)4 increases proportionately. This leads to thereduction and eventual quenching of the PL. This mech-anism is completely consistent with the reaction modesoutlined and the stabilization and enhancement of the PLfrom the nonaqueous HF/MeCN solution containing smallquantities of water7 (Fig. 5b).

The outlined mechanistic routes are also in concert withthe observations of Steckl et al.27 and Yan et al.23 and focusattention on the importance of silanol-bond formation andon hydroxyl anion participation during the etch process.Similarly, these reaction paths are consistent with theobservations of Prokes and co-workers,57 who have moni-tored a very intense PL when a porous silicon sample isprepared in concentrated hydrogen peroxide (H202). In ananodic etching process, the hydroxyl anion, if formed atthe silicon surface—solution interface or in the near prox-imity of the silicon surface, is expected to diffuse to thissurface, forming a hydroxyl bond. The importance of thehydroxyl anion should be manifest if the polarity of thesilicon surface is changed. Indeed, Fig. 18 demonstrates ananticipated drop in the PL intensity for the reverse biasetching of the PS surface. We suggest that this diminishingof the time-dependent PL signal results as hydrogencations rather than hydroxyl anions diffuse to the reactingsilicon surface.

The effect of methanol and ethanol—We now turn ourattention to a consideration of the effect of methanol and

B

(a) ( ,H ,SI _Ht F- e- s( ,F,Si ,Sç F' ,Si\ ,Si ',Si /Si\,S\ H ,sç H

E\ , ' / \ /OH OH ,Si ,Si ,OHH '' •2 HON Si St OH- S Si/ \

OH ' s( 'OH ' ,s( HOH

Aqueous etch of Si in HF - H20 - M.

Fig. 17. Reaction mechanisms for dissolution and silicon fluorideand silicon hydroxide (oxyhydride) bond formation in nonoqueousand aqueous HF etch solutions.

8.0E40C

6.0E40C0

4.0E40-

I 2.0E4

Fig. 18. P1 intensity at 6200 A as a function of polarity. In (a) weplot P1 intensity as a function of time for a Si(100) sample anodi-cally etched at 18 mA/cm2. In (b) we observe the effect of reversebiasing the current so as to form a cathodic etch of the silicon. TheP1 intensity drops precipitously (as does the current density) in acathodic etch. The etching solution was 6 M HF in MeOH(aq).

ethanol as they are introduced to the HF etch solution. Itis apparent that the systems HF/MeOH/H20 and HF/EtOH/H20 are complex. Yet the behavior of a silicon waferin these solutions leads to PL features which are observedconsistently. First, the rate of increase of both the 6200 Aorange-red and 5200 A green luminescence at the begin-ning of the etch cycle clearly increases at a much fasterrate (Fig. 4, 8, and 12) upon introduction of these alcoholsto the etch solution. Further; a similar effect accompaniesthe introduction of ethylene glycol.5 The rapid decay(Fig. 4, 8, and 12) of the PL is also apparent in the etchingsolutions to which the alcohols have been introduced to anHF-based environment. Once the FL is quenched, we havefound no chemical treatment which can again initiate theluminescence process. Only a new electrochemical etchcan produce further PL. In sharp contrast, adding signif i-cant quantities of methanol (up to —i mL) to -2 mL of anear—anhydrous MeCN/HF etching solution does little toaffect the PL. We suggest that these observations are bestexplained by considering that MeOH and EtOH mightinteract differently at distinctly different times in theetching cycle. At the beginning of the etch cycle, MeOHand EtOH contribute to the formation of PL emitters,whereas at the end of the etch cycle they or compoundswhich they form through interaction with the aqueousetch constituencies promote dissolution and the quenchingof the PL. We suggest that these alcohols may interactslowly with HF in solution to create more effective agentsto hasten the dissolution process.

The variation of the HF, water, and methanol concentra-tion in the etch solution produces a dramatically differenttime-dependent emission. This again supports chemical

Reversed Bias(18 mA)1.0E5

Etchl8rnA 1 '—HH48.0E4

Etthing Curent00 Dropped to —10 mAC /2 6.0E40

Tt..'4.OE4 /.0,C

S 2.0E4

LI,0.OEO0 0.1 0.2 0.3 04

Relative Time (Hours)

0.OEO

(100) Silicon Surface Reactions

A C

0.1 0.2 0.3Relative Time (Hours)

F E 0

FN ,F 2HF' - e- s( ,F5j Si*S1 Si 'TSi Si/ ' Al ' " ' ' r I \ / + -

F F F HF

Non-aqueous etch of Si in HF - MeCN.

A B C

/SI\ / H H s( Ht.oH- - s( ,OH,Si ,Sç oir L OH'S. ,Sç,sç H H s•¼, H

- e-IOH-

F U


transformation and a chemical entity as the source of thePL on the PS surface. We note that while CH30H isthought to dissociate,58 (as a weak base in solution), toCH30 + W, it may also be relevant59 that the 1120 disso-ciation energy greatly exceeds the MeOH dissociationenergy to form the hydroxyl radical

1120 -÷H + OH AE = 119 kcal/mol [1]CH3OH - CH3 + OH E = 92.35 kcal/mol [2]

These energetics suggest that methanol can, on one hand,be a source of a inethoxyl group and on the other hand,might be a source of a hydroxyl group bound to the siliconsurface.

CH3CNis a very weak base (vs. even CH3OH) and cyanideis a very poor leaving group.6° If C113011 is introduced to anear-anhychous CH3CN/HF etching solution, an HF reac-tion with methanol to produce water via reaction 2 61 mightreadily occur

HF + CH3OH -1120 + CH3F [3]Here an increased water concentration would be very sig-nificant and, as Fig. Sb demonstrates, should greatly en-hance the "orange-red" PL from PS. The formation of astrongly reactive agent at the silicon—solution interfacemay coincidentally contribute to the dissolution processand the eventual decay of the 6200 A FL; however, in thenear-anhydrous environment, the strongly enhancingeffect of water and the potential dissolution and/orquenching processes may well balance each other.

Recently, Warntjes et al.'1 have studied the EC methoxy-lation of a PS surface. In order to better understand theirobservations, we have expanded our quantum chemicalmodeling studies926 to interpret6 their experimental re-sults. Warntjes et al. have used the partial anodic dissolu-tion of porous silicon in anhydrous methanol to electro-chemically graft methoxy groups onto a hydrogenatedporous silicon surface at room temperature. These authorsmake two key observations. (i) they find that the visiblePL spectrum blue shifts upon methylation by close to10 nm from that of hydrogenated PS, and (ii) they obtainevidence for methyl [v(CH3)] and (SiOC) infrared bands,suggesting the grafting of OCH3 units and the formation ofSiOCH3 constituencies.

An extension of the quantum chemical modeling studieswhich we have presented in Thbles II and Ill is presented inFig. 19 where we summarize the adiabatic excited statetriplet—ground state singlet wavelength onsets for a numberof substituted silanones. The calculated structures in Fig. 19indicate the changes to be expected as one converts the

0 = Si — OH and 0 = Si — OSiH3tetravalent silicon constituencies which we have previous-ly considered9'26 to an

0 = i - OCH,fluorophor. We have determined6 that the predicted changein adiabatic energy commensurate with a blue shift in thevisible PL spectrum is not readily associated with the con-version of a hydroxyl group (OH) to an -OCH3, functionalgroup. This change is most likely associated with the con-version of an -OSiH3 unit to -OCH3, in excellent agreementwith the observations of Warntjes et al.'1 Further, any con-version of an -OH or -OSiH3 to a methyl (CH3) constituen-cy will likely result in a substantial red shift in the PLspectrum.6

Warntjes et al. have noted, in analogy to the mechanismfor HF etching, that silicon, as it is oxidized in methanol,should undergo spontaneous dissolution to form solubletetramethoxysilane. This process is expected to proceedthrough the successive grafting of methoxy units (e.g.,analog of F atoms in HF, see the following discussion) tothe silicon atoms. However, dependent upon the experi-mental conditions, it might also be possible to graft methyl

Si Si

X = 399 nm

OCH3 SIH3

0

CH, OH CH, OCH3 CHa OSiH3X = 398 nm

0

Si

OSIF-I3 Oct-I,= 411 nm

Fig. 19. Calculated wavelengths corresponding to the adiabaticexcitation energies for the lowest triplet states of several silanonebased silicon oxyhydrides. See also Ref. 6.

groups onto a hydrogenated silicon surface to substitutethe silicon—hydrogen bonds, eventually forming the moreeasily removed tetramethylsilane. Of course, many thingschange with methanol concentration including the distri-bution of HF species, the double layer, and the solutiondielectric constant. However, the above considerationscause us to consider the possible evolution of the PS dis-solution cycle with change in experimental conditions.

The etching and PL process—Several researchers haveconsidered the reactions which describe the dissolution ofsilicon in an HF-containing electrolyte. The formation of asilicon oxyhydride emitter5'69 would appear to be consis-tent with the comparative and competitive mechanismswhich have been suggested for silicon dissolution in aque-ous and nonaqueous media.32'5253 Figure 17a outlines thecycle of suggested conversion steps in a nonaqueous etch-ing process which provide the dissolution of siliconthrough the eventual formation of an 5114 leaving group.This process can be inhibited through the introduction ofwater52 and, in an aqueous medium, the competitive mech-anism indicated in Fig. 17b is thought to proceed throughthe formation of silicon hydroxide bonds which might wellresult in the formation of surface-bound PL oxyhy-drides.5'69 It is appropriate to evaluate these competingprocesses as they occur in an aqueous medium and to con-sider the potential modifications of these basic mechan-isms when the alcohols CH3OH and C2H5OH are present.

The nonaqueous etching process (Fig. ha) in HF/MeCNsolution is thought to form an initial dihydride surface (A)where ab initio calculations62 have shown that kineticrather than purely thermodynamic factors are responsiblefor hydride passivation.63 The initial dihydride surface,which can be considered either a (100) surface or the kink

0 0

//Si SI

0

/SIN NXa,=403 nm X401 nm

OH H OCH3 H OSIH,= 406 nm

00 0

/Sift,

H

Si

SiftX=401

0

Si

OSIH3

X = 410.9 nm

0

= 400 rim X = 411 rim


site of a (111) surface, is thought to be stable in solution.32This surface does not appear to be photolumineseent59consistent with our quantum chemical calcu1ations2' andthe experimental evaluations64 of SiH excited states. Thereaction sequence with HF is thought to be hole-initiated(B), producing a proton and a silicon radical.32'52'53 Afterelectron ejection into the silicon conduction band, facili-tated by the ready oxidation of silicon, the positivelycharged silicon is thought to be complexed by a fluorideion (C). This process is thought to continue as the highelectronegativity of fluoride in the silicon fluoride bonddestabilizes the adjacent Si—H bond, which is then moresusceptible to oxidation through electron injection or holeconsumption (D). In the absence of water, a fourth oxida-tive step (E) can take place as the addition of fluoride tothe oxidized silicon further destabilizes the Si—Si bonds.The polarity of the two Si—F bonds can facilitate the fur-ther addition of HF and the final dissolution of SiF4 whichdissolves as the silicon hexafluoride.

As Gerischer et al. have suggested52 and we have previ-ously o

Ex Situ Photoluminescent Emissions from Porous...

Documents

Transcript of Ex Situ Photoluminescent Emissions from Porous...