T.H. Wang, T.F. Ciszek, M. Page, Y. Yan, R. Bauer, Q. Wang, J. Casey
R. Reedy, R. Matson, R. Ahrenkiel, and M. M. AI-Jassim
National Renewable Energy Laboratory
1617 Cole Blvd., Golden, CO 80401,USA


(silicon tetra-iodide): Sis+41g+Si14g (Tp6OooCJ., Source
silicon pre-doped with boron to about 5 ~ 1 0cm'
' ~ IS placed
at the bottom of a vertical open-tube-type reactor that is
continuously purged with hydrogen (or helium), and heated
from underneath. It reacts with iodine to form a reservoir of
silicon tetra-iodide at the lower section of the reactor. 2.
Supersaturating the transport agent with silicon:
(Tp1200"C). In the presence of
additional source silicon, the silicon tetra-iodide turns into
silicon di-iodide, which is highly unstable. All of the silicon
iodide species are prevented from escaping by
condensation near the top of the tube because of a topcold thermal gradient and the gravity effect. 3. Depositing
Si: 2Sil~~+Sis+Sil4~
(-900OC). A substrate is introduced
into the silicon iodides reservoir at a temperature of about
9OO"C, and deposition of silicon starts. In short, a
disproportionation reaction between silicon di-iodide and
silicon tetra-iodide transports silicon from the source to the

Fast (3 pmlmin) and direct deposition of large-grain
(-20 pm) polycrystalline silicon layers on foreign
substrates at intermediate temperatures (-900°C) is
achieved by an atmospheric pressure iodine vapor
transport technique. A hole Hall mobility of 51 cm2N.sat a
doping density of 5 . 5 ~ 1 0 cm"
' ~ was measured in an asdeposited material. After hydrogen passivation, it
increased to 76 cm2N.s. Crystallographic defects mostly
consist of less detrimental stacking faults and twins, with a
few dislocations. Diagnostic PV devices using an N'-a-Si/ia-Si/APIVT-Si(absorber)/P'-CZ-Si structure demonstrated
an open-circuit voltage of 0.48 V and 0.58 V at one and 13
suns, respectively. Highly [ I IO]-oriented silicon layers
were attained.
The pursuit of thin-layer silicon solar cells has branched
into three distinct approaches in the last few years [I],
namely, nanocrystal thin films, polycrystal thin layers, and
single-crystal layer transfer. Whereas progress in the first
and last categories has been significant, efforts in the
second approach have been limited by grain size and
deposition rate dictated by the commonly used
(ch1oro)silane-basedchemical vapor deposition technique.
However, the recently developed atmospheric pressure
iodine vapor transport (APIVT) [2] may provide a muchneeded means for depositing the active silicon layer in this
area. In addition to being an open deposition system with
potential for continuous deposition, low capital cost, and
no need for expensive effluent treatment, APlVT most
importantly has the advantages of fast (3 pmlmin)
deposition and large grain sizes (-20 pm) at 900°C on
foreign substrates.


This paper will report various characterization results on
the materials deposited by APIVT. In particular, preferred
growth in the [IIO] orientation will be of special interest.

Silicon Source
Fig. 1. The process of the atmospheric pressure iodine
vapor transport.


We chose Corning LGA-139@glass-ceramic and mullite
as the substrates because of their excellent thermal
expansion match with silicon. Corning Vycor@ hightemperature glass was used for X-ray diffraction
determination of film texturing. Heavily 6-doped CZ-Si

The details of the atmospheric pressure iodine vapor
transport of silicon are described elsewhere [2]. It basically
involves three steps (Fig. 1). 1. Forming a transport agent

U.S. Government work not protected by U S . copyright


However. 3) on non-passivated thinlayer silicon are difficult to interpret. These values are about 36% (as-deposited) and 55% (hot-wire hydrogenated) of the Hall mobility of a similarly doped single-crystal silicon. the minority-carrier diffusion length is expected to be greater than 30 prn. ’ ’ ______ 2 3 4 5 6 7 8 - 4 0 I (mrroaeccmds) Fig. Note that the bottom horizontal straight line is the electron beam path and the EBlC signal baseline. and a hydrogen pressure of 1 torr for 3 hours. the silicon layer shows a mixture of largely random nucleations and a very small fraction of epitaxy. Deposition rates up to 5 pm/min have been achieved. Even when deposited on a single-crystal silicon substrate. A hole density of 5.5~10”~ m means ‘ ~ that dopant boron was transported at essentially the same efficiency as silicon.o*oL. 4 seems to indicate very minimal recombination activity at the grain boundaries. a room-temperature Hall mobility of 51 cm2N. hot wire temperature of 1900°C. we found that the Hall mobility has been increased to 76 cm2N.s for the majority-carrier holes was measured for a polished 56-pm-thick silicon layer deposited on Corning LGA-139@. Subjecting the same sample to hot-wire hydrogenation at a substrate temperature of 300°C. 2. The electron-beam-induced current (EBIC) line scan in Fig. The top curve is the corresponding EBlC signal. The apparent decay 139 Grain-boundary recombination is of great concern to solar cell performance. Typical grain structure obtained by APIVT. The minority-carrier lifetime Tbulk=l/(l/ T=-I/T~)in the layer is expected to be greater than 90 ns.! - . time ‘se ranges from the surface-dominated lifetime r. distance of 2 cm. RF-PCD decay signal of a 56-pm thick nonpassivated silicon layer on glass-ceramics. with better films obtained at about 3 pmlmin. 3.wafers were used as a conductive substrate for diagnostic devices. ELECTRONIC PROPERTIES With a van der Pauw configuration. as .s. With an inferred electron mobility of 140 cm2N. Open operation of this technique with purging gas overcomes the usual difficulty in a conventional chemical vapor transport or deposition svstem that is either a closed system or an nent. actual diffusion lengths measured by internal quantum efficiency after device processing are still below 10 pm for now.=90 ns to the trapping dominated decay of >9 ps (typical of polycrystalline silicon). regardless of the substrate materials used. Lifetime measurements (Fig. 2). open Ii Fig.s (35% of electron Hall mobility in single-crystal silicon). SUBSTRATEEFFECTS The grain structures are almost the same under similar deposition conditions. Grain sizes as large as 20 pm are easily attainable (see Fig. lo.

Stresses in the material are moderate. it is very possible to achieve highly [IIOI-orientated films. A potentially beneficial effect is grain-boundary passivation. (a) Conformal grain boundaries. both polycrystalline silicon layers show a downshift of -3 cm-’ and a half-width broadenning by the small amount of 1 cm”. It is widely believed that twins and stacking faults in silicon have minimal impact on solar cell performance. 7b). as epitaxy is not our task at hand and the polycrystal layers deposited on heavily boron-doped silicon permit us to realistically gauge the material performance. A pump beam size of about 0. All the peaks were normalized to the 1I ] peak. 7.E+15 0 1 Depth (micon) 0.5 500 510 520 530 540 Raman Shift (llcm) Fig. Polycrystalline 140 . both deposited on high-temperature glasses. at the lower left. SIMS depth profile of iodine in IVT-Si Fig. As mentioned earlier. 6. But this is of no concern.1 mm x 5 mm was used to cover hundreds of grains. On a glass substrate. on the other hand. Fig.5 1. 5a also shows that small grains near the interface are quickly overcome by the neighboring larger ones. The solid line is for a sample that was standard [I deposited under normal conditions. In the bulk. Compared to a stress-free silicon single-crystal reference with a nominal Raman shift at 521 cm”. Such a build-up of silicon iodide at the interface is confirmed by secondary-ion mass spectrometry (SIMS) mapping of iodine in Fig.E+17 I. Raman shifts of silicon layers on CZ-Si substrates grown under hydrogen and helium ambient.silicon layers grown on single-crystal silicon substrates using hydrogen and helium purge gases were studied by Raman spectroscopy (Fig.E+16 I. The high concentration of H was introduced by the growth ambient as a purging gas. A few microns away from the interface. 8. 5c. Stacking faults and twins are the major defects (Fig. as shown in Fig. the iodine level drops to below 10l6 ~ m -Possible ~. This clearly shows that this silicon layer is highly [ I 101-oriented. the grains become much larger. as evidenced by the small contrast variations in Fig. The dashed line.E+19 l. CRYSTALLOGRAPHY Transmission electron microscopy (TEM) images indicate that the grain boundaries are conformal (Fig. -1000428 - H (H2) -Si-reference 1500 l. a preheated silicon substrate appears to be needed. our most recent experiments indicate that by modifying the deposition conditions. is a sample deposited under a different condition. X-ray diffraction shows near-random orientations of crystallites (further discussion later). in agreement with the TEM study. Fig. leading to very fast grain evolution (growth) away from the interface. in comparison to standard silicon powder peaks. This indicates that there are some low-level stresses in the silicon layers. This sample demonstrates over 435% the [220] intensity of the standard peak. m I. with occasional dislocations. with slight [ I IO] texturing. but with similar grain size and thickness. (b) High resolution image of twins and stacking faults. 8). silicon layers deposited under normal conditions exhibit almost random orientations.E+18 . To avoid such a silicon iodide build-up and to achieve epitaxial growth. Similar grain size and surface morphology were obtained on all other substrates. 7a).c. -I \ -E (a) (b) Fig. 7a. The high iodine content near the top surface (within 1 pm) is expected from the adsoption of the reactor ambient after growth. However. and comparison studies with helium (Hall mobility and Raman shift) as the purging gas showed that this hydrogen may be in molecular form and does not effectively passivate defects. iodine effects on silicon bulk material remain to be studied. IODINE INCORPORATION Some residual iodine is seen in the silicon layers. This phenomenon may be caused by a build-up of silicon iodide at the interface prior to deposition because of an initially cold substrate brought in from outside the growth region. No significant difference was seen in the Raman spectrum between the layers grown with hydrogen and helium purge gases. 9 shows the Xray diffraction peaks of two samples. but structural imperfections are minimal. 6. which showed almost random orientations.

Symp. II. XRD peaks of silicon films deposited with different conditions. and under 13X concentration. 258.” Recent Res. The Electrochemical Society. 1 (1999):241-256.. “Growth of Large-Grain Silicon Layers by Atmospheric Iodine Vapor Transport. “A More Than 18% Efficiency HIT Structure a-Silc-Si Solar Cell Using Artificially Constructed Junction (ACJ). Soc. 9. Baba.G. Takahama. [ I ] R.” J. Devel. Bergmann. it reached 580 mV for 2mm x 2-mm cells. S. Nakamura. respectively.l11 I00 Voltage (V) 2 theta Fig.. T.. . Scott Ward. our active layer is a p-type poly-Si. IO. “HighEfficiency a-Silc-Si Heterojunction Solar Cell. T. and S. Silicon layers deposited by atmospheric pressure iodine vapor transport show promising physical.. Miguel Contreras. “Crystalline Si Films on Foreign Substrates for Electronic Applications.. Haku. * 20 40 60 AM1. Res. [3] T. Tsuge.5 LVLLJ . Crystal Growth Res.trapping and short diffusion lengths from internal quantum efficiency measurements. Isomura.B. M. REFERENCES Fig. I-V curve of a HIT device structure. Short-circuit current is limited due to lack of light [4] Y . T. K. and S. ACKNOWLEDGEMENTS We would like to thank Marc Landry. and crystallographic properties for solar cells. H. This work is supported by the U. Ast of Cornel University for providing the glassceramics substrates.F. Vol. chemical. DIAGNOSTIC DEVICES [2] T. Functioning pln junction devices have been made from APIVT Si layers on heavily doped silicon substrates. Fig. M.1219. Nishiwaki. Nakano.857( 1992). H. 11. instead of an n-type [4]. Vol. Wakisaka. Sawada. Nishikuni. and Yong Zhang for assisting in the experiments and measurements. S. Anna Duda... Two of the perfectly [ I IO] aligned grains actually allow one to see a high-resolution image of the grain boundary with dislocation cores (Fig.” 1994 /€€€ First World Conference on Photovoltaic Energy Conversion. S. Terada. DEdC3699G010337 to the National Renewable Energy Laboratory. 11 shows the I-V curve of a HIT cell with a fill factor of 65%. A HIT (Heterojunction with Intrinsic Thin-layer) structure [3] was used. Wang and T. large grain sizes at intermediate temperatures. Takahama. The active layer deposited by APIVT was not subjected to any intentional treatment nor does it have significant [I101 texturing. Kuwano. M. the results of the first HIT structure devices are promising: V. N. High-resolution image of two perfectly [110]aligned grains. 300 . However. Ciszek. The control cell using CZ-Si with the same HIT structure has 598 mV and 660 mV. under AM1. T. TEM observation of the same sample indicates that over 90% of the grains are within about 5” of the 4 1 0 > axes. p. Proc. p.5 is 480 mV. 147(5) 1945-1949 (2000). Matsuyama. Tanaka.S. [ZOI 250 A 200 150 1000830 -1000831 Standard : [I111 :normalized 1000517 2mm x Zmm 1 -*. High deposition rate. Nevertheless. We also thank Prof. 141 . Department of Energy under Contract No. Tsuda. IO). N.Tsuda. The intrinsic layer and n+ a-Si emitter were deposited at a low substrate temperature (Te200”C) by hot-wire chemical vapor deposition. SUMMARY Fig. 80 J A[5. D.H.” Mat. and highly [IIO] preferential orientation may be the right combination to lead to practical and efficient thin-film Si solar cells. Nakano.