硅基纳米结构的减反特性及其在太阳能电池中的应用.pdf

Vol. 32, No. 8 Journal of Semiconductors August 2011Antireflection properties and solar cell application of silicon nanoscructuresYue Huihui( 岳会会 ), Jia Rui( 贾锐 ) , Chen Chen(陈晨 ), Ding Wuchang(丁武昌 ),Wu Deqi( 武德起 ), and Liu Xinyu( 刘新宇 )Key Laboratory of Microwave Devices and Integrated Circuits, Institute of Microelectronics, Chinese Academy of Sciences,Beijing 100029, ChinaAbstract: Silicon nanowire arrays (SiNWAs) arefabricated on polished pyramids of textured Si using an aqueouschemical etching method. The silicon nanowires themselves or hybrid structures of nanowires and pyramids bothshow strong anti-reflectance abilities in the wavelength region of 300– 1000nm, and reflectances of 2.52% andless than 8% are achieved, respectively. A 12.45% SiNWAs-textured solar cell (SC) with a short circuit current of34.82 mA/cm 2 and open circuit voltage (Voc/ of 594 mV was fabricated on 125 125 mm 2 Si using aconventionalprocess including metal grid printing. It is revealed that passivation is essential for hybrid structure textured SCs,andVoc can be enlarged by 28.6% from 420 V to 560 mV after the passivation layer is deposited. The loss mechanismof SiNWA SC was investigated in detail by systematic comparison of the basic parameters and external quantumefficiency (EQE) of samples with different fabrication processes.It is proved that surface passivation and fabricationof a metal grid are critical for high efficiency SiNWA SC, and the performance of SiNWA SC could be improvedwhen fabricated on a substrate with an initial PN junction.Key words: antireflection properties; silicon nanowires; solar cells; passivationDOI: 10.1088/1674-4926/32/8/084005 PACC: 7865H; 8630J; 6146 EEACC: 2520C; 2550N1. IntroductionSurface antireflection techniques are important for the per-formance enhancement of many optical devices, such as SCsand planar displays ?1 3 . Conventionally, transparent quarterwavelength layers of SiOx , TiO x , or Six Ny with intermedi-ate or gradient refractive indices are used as antireflectioncoatings, although multilayer coatings have also been used tobroaden the spectral width of the antireflection. However, thesecoatings work effectively only in a limited spectral range andfor specific angles of incidence ?4 . An alternative approachto minimize reflectivity utilizes a fine textured surface, com-prising sub-wavelength structural features on the nanometerscale. A sub-wavelength structure surface with a deep and ta-pered profile can suppress the Fresnel reflection substantiallyover a wide spectral bandwidth. The fabrication of the sub-wavelength structures haw been investigated by several meth-ods, including chemical vapor deposition ?5 , laser ablation ?6and annealing in a reactive atmosphere?7 9 . Although thesesynthesis methods are quite acceptable and well controlled,most need either a high synthesis temperature or a long synthe-sis time ?10 due to the limitation of the growth mechanism, andthis increases the production cost. A silver induced aqueous-chemical-etching process conveniently synthesizes SiNWAsat low cost compared to the above methods?10 . However, noanalyses have been undertaken on how SiNWAs could be syn-thesized on pyramid textured Si wafer. Simultaneously, Si NWarray SCs have been reported by many institutes, but lesswork has been done on the loss mechanism from the EQE as-pect?10 12 .In this work, SiNWAs synthesized by a galvanic displace-ment reaction were explored to replace the conventional pyra-mid textured (CPT) structure for crystal Si (C-Si) SCs. SiN-WAs are produced on both polished and CPT Si. It is provedthat SiNWA can beconveniently synthesized on the sidewall ofthe pyramids to form a hybrid structure, which is almost inde-pendent of the crystallographic orientation. The SiNWAs pre-pared on polished Cz-Si show a significantly lower reflectanceof 2.53% in the wavelength region of 300– 1000nm. The low-est reflectance of the hybrid structured Si is 8%, which is alsomuch lower than pyramid textured Si. Also, the reflectanceproperty of the SiNWAs versus etching duration is studiedin detail. SiNWA SC with an efficiency of 12.45% has beenachieved, as the short circuit current is enlarged prominently.The effects of the SiNWAs on the performance of the cell, suchas the external quantum efficiency (EQE), the conversion effi-ciency (á/ , open circuit voltage (Voc), and short circuit current(I sc), have been systemically investigated.2. Experiment sectionSiNWAs were fabricated on 280 m p-type double sidepolished (substrates A) and one side pyramid textured (sub-strates B) Cz-Si (100) substrates with resistivity of 0.9 cm.Various lengths of SiNWAs were obtained by immersing Sisubstrate in aqueous 4.6M hydrofluoric acid (HF) and 0.02Msilver nitrate (AgNO 3/ mixture solution, and soaking the wafer* Project supported by theStateKey Development Program for Basic Researchof China (Nos. 2006CB604904, 2009CB939703), the NationalNatural ScienceFoundation of China (Nos. 60706023, 90401002,60977050, 90607022), andthe ChineseAcademy of Solar Energy ActionPlan (No.YZ0635).Corresponding author. Email: jiarui@ime.ac.cnReceived 15 February 2011 c 2011 Chinese Institute of Electronics084005-1J. Semicond. 2011, 32(8) Yue Huihui et al.Fig. 1. (a) Silver films on substrate A after etching for 2 h; the insetshows the silver particle at the bottom of the NWs. (b) SEM of the Agremoved NWs on substrate etched for 2 h. (c) Top view of substrateA etched for 3 h. (d), (e) Top view of substrate B etched for 2 and 3h. (f) Magnified top view of substrateB etched for 0.5 h.in a hydrothermal bath of 25 ?C for times from 30 to 240 min.Increased etching time results in correspondingly longer wire-length l . During the etching process, Ag particles were pro-duced on the silicon surface and catalyzed subsequent Si etch-ing by acting as local cathodes while the Si below Ag acts asanodes. Concentrated nitric acid (HNO 3 / etching for 1 h hasbeen applied to remove the residual Ag particle, followed by arinse in 5% HF solution for 30 s and DI water three times toeliminate the oxide layer and residual ion, respectively. The Agmass concentration ratio falls below 0.5% in 10 m2 areasatthe bottom position of SiNWAs by energy diffuse spectrometryanalysis. Reflectance and transmittance of various-length SiN-WAs were measured on a 7-SCSpec SC spectral measurementssystem.SCs were fabricated on SiNWAs textured substrates, hy-brid structure of SiNWAs and pyramid structure textured sub-strates and C-Si substrates with PN junctions. Conventionalsilicon cell fabrication processes were used to form a frontphosphorous-diffused emitter, Al back-surface field as well asmetal grid. The basic parameters including á , I sc, Voc, fill fac-tor (FF) and EQE of 125 125 mm 2 cells are measured withcalibrated 1-sun simulators.3. Results and discussion3.1. Observation resultsFigure 1 shows the surface morphology of SiNWAs ob-served by scanning electron microscopy (SEM). The brightarea in Fig. 1(a) is a loose Ag dendritic film on the SiNWA sur-face while the gray area is Si substrates before being cleanedwell by HNO 3. Ag particles on the surface andat the bottom ofFig. 2. Relationship of length of SiNWAs versusetching duration. Thedotted line is the linear fit relationship.the SiNWAs are shown in the inset of Fig. 1(a). SiNWAs withvarious lengths are prepared by changing the etch duration, andasample of about 5 m in length with a diameter of an individ-ual NW ranging from 50 to 250 nm is shown in cross-sectionalSEM in Fig. 1(b). The SiNWAs in sub-wavelength dimensionof uniform length are fabricated on a large Si, as shown in thetop view observation in Fig. 1(c). Etching of substrate B beginsat the top of the pyramids and the pyramids would be mostlyeaten away if the etching duration were long enough, as ob-served in Figs. 1(d) and 1(e). Controlling the etching time foran appropriate value would result in a hybrid textured struc-ture composed of pyramid arrays and SiNWAs synthesized onsubstrate B, as shown in Fig. 1(f). The SiNWAs in the hybridstructure are grown on the sidewalls of the pyramid array andthey arenot vertical to the wafer but to the sidewall of pyramidscertified by SEM observation, i.e. the Si is in [100] while theside wall of the pyramids is in [111]. Therefore, the preferentialcrystallographic orientation of the SiNWAs remains the sameason substrateA, although substrate B hasthe pyramid texture,which is consistent with the analysis of Chen et al.?13 . Theetching rate at the bottom of the pyramids is lower than the topposition, and the top of pyramids are almost eaten away com-pletely when the etching duration is more than 3 h. Accordingto the SEM observation of SiNWAs on polished wafers etchedfor different durations, onecan generally conclude the relation-ship between the length of the SiNWAs versus etching time asdepicted in Fig. 2. The linear fit relationship is also plotted inFig. 2. Such nearly linear etching behavior facilitates the lengthcontrol of SiNWAs on a large scale.3.2. Reflection measurementThe as-synthesized samples seem extremely black, whichreveals their possible excellent optical anti-reflectance ability.A 7-SCSpec solar cell spectral measurements system is usedfor reflectance (R) measurement of the NW arrays. The re-flectance spectra of SiNWAs on a polished surface and a tex-tured surface of varied etching times are shown in Figs. 3(a)and 3(b), respectively. The increasing etching duration de-creases the reflectance of substrate A, and the reflectance isless than 3% in the wavelengths from 300 to 1000 nm when the084005-2J. Semicond. 2011, 32(8) Yue Huihui et al.Fig. 3. Reflectance of substrates(a) A and (b) B etched for differenttimes.Fig. 4. (a) Transmission andabsorptanceof substrateA etched for 4 hand CPT C-Si. (b) Reflectance of C-Si NWs on substrate A in thewavelengths 300– 1600nm.etching duration is longer than 2 h, as shown in Fig. 3(a). Thereflectance of substrate A etched for 4 h is about 20% less thanthe CPT surface in the whole wavelength region and thus im-parts a large reduction in reflectance. By the equation R. ) D1 A./ T. ), where is the optical wavelength, and R, T ,A are the wavelength dependent reflectance, transmission, andabsorption of the SiNWAs, respectively, the reduced reflectiv-ity by a SiNW array can significantly increase the absorptionability. The absorption of substrate A etched for 4 h derivedfrom transmission and reflectance is plotted in Fig. 4(a), inwhich the CPT C-Si is also plotted as a reference. As the fig-ure shows, the absorption is larger than 95% and shows a morethan 10% increase compared with CPT C-Si in the wavelengthregion of 400– 1000nm.This observed reduction in the reflection for the etchedsamples can be explained by scattering theory. Compared tothe diameter of the pyramids, which is of the order of m,the diameter of Si NWs, e.g. from 50 to 300 nm, becomescomparable to the wavelength of the incident light. The in-cident light will be dominantly scattered according to sub-wavelength scattering theory and therefore prolongs the opticalpath length. Also, the SiNWAs introduce a possible porositygradient, which also implies a change in refractive index withdepth?10 . The above factors result in enhanced light absorp-tion.The reflectance of substrate B is more complex after theSiNWAs are fabricated, as shown in Fig. 3(b). The variationsin diameter of the hybrid structure change the reflection prop-erties in the whole wavelengths region. In short wavelengths,the wavelengths of the incident light are similar to the averagediameter of Si NWs. Consequently, the scattering effect wouldbe dominant, and thus significantly suppressed the transmis-sion and decreasedthe reflection. However, in the long wave-lengths, the SiNWAs would be much more appropriate for anti-reflection purposes, whereas the micrometer scaled pyramidsdeteriorate the anti-reflection properties. Therefore, there is atrade-off between the influence of both the pyramid array andthe NW array on the reflection property for the long wave-length region. Accordingly, the reflection is firstly reduced andthen enhancedin the reflectance spectra asthe etching time in-creases. Simultaneously, we can conclude that the reflectionreduction using the hybrid texturing structure is not prominentcompared to that using the Si NW array only by Figs. 3(a) and3(b). Thus the main mechanism of reflection reduction shouldbe due to the involvement of Si NWs, which eventually en-hancesthe absorption of SiNWA textured SCs.For both substrates A and B, the reflectance increasessharply in the wavelength longer than 1000 nm, as shown inFig. 4(b). According to the equation R D R R./ dRd , the av-erage reflectance of substrate A is 2.52% in the wavelengths300– 1000nm, where is the wavelength of the optical wave,and R./ is the wavelength-dependent reflection coefficient. Itis also noteworthy that the reflectance of our samples is com-parable or much lower compared with those reported in the lit-erature. For example, Peng et al.?10 obtained reflectance of1.4% over the wavelength ranging from 300 to 600 nm forSiNWAs prepared on monocrystalline Si substrates. Rappichet al.?14 obtained reflectance values for nanoporous siliconranging from 2%– 30 % and 2%– 20% in an anodizednanowire-like structure in the 300– 850nm wavelength range.3.3. C-Si NW solar cellsSCs were fabricated on hybrid structure textured substrateswithout a SiN passivation layer (samples C1– C2)and with aSiN layer (samples D1– D2).SCs with SiNWAs were also man-ufactured on C-Si substrates where initial PN junctions werealready formed (samples F1– F2)or not (samples E1– E3).Weuse a conventional silicon SC fabrication technique processincluding phosphorous diffusion to form a p–n junction, andscreen-printing to form an aluminum back-surface-field andmetal grid. Current – voltagecurves of 125 125 mm 2 cellsaremeasured under calibrated 1-sun simulators. Our best SiN-WAs SC is made on C-Si substrate with an etching time of 2 h,and they have an efficiency of 12.45% with open-circuit volt-age of 594 mV, short-circuit current of 34.82 mA/cm 2 and fillfactor of 67%. Although the efficiency is not as high as com-mercial silicon solar cells, it is much better than the 9.31% SCwith a Voc of 548.5 mV reported by Peng et al. ?15 . The bestSC with a nanostructure as the anti-reflectance coating ever re-ported was made by Yuan et al.?16 with Voc, J sc, and FF of612 mV, 34.1 mA/cm 2, and 80.6%, respectively. Compared084005-3J. Semicond. 2011, 32(8) Yue Huihui et al.Fig. 5. Comparison of performance for different SC samples.(a) Comparison of conversion efficiency, Voc, I sc and FF. (b) EQE of samplesCand D. (c) EQE of samplesE and F. (d) Schematic diagram of low shunt resistanceeffect to sample D.084005-4J. Semicond. 2011, 32(8) Yue Huihui et al.with their results, the I sc of our best sample is improved by0.71 mA/cm 2. Their as-grown thermal oxide passivation layerand as-evaporated metal grid matched their SC co