Properties of amorphous Si-rich silicon nitride prepared by rf-magnetron sputtering

Materials Science and Engineering B71(2000)321–326Properties of amorphous Si-rich silicon nitride prepared byrf-magnetron sputteringM.Vetter,M.Rojahn *Institut fu ¨r Physikalische Elektronik ,Uni 6ersitat Stuttgart ,Pfaffenwaldring 47,70569Stuttgart ,GermanyAbstractThis article investigates the physics underlying field effect passivation of p-type Si surfaces coated with SiN x (:H)of various Sicontents.We find a correlation between the effective lifetime of minority carriers in p-type Si passivated with SiN x ,on the one hand,and the dark dc-conductivity data of corresponding a-SiN x -films with different composition,on the other hand.By increasing the amount of Si in a-SiN x (:H)structures the bandgap decreases from about 5to 2eV.As a result,the dark dc-conductivity rises by several orders of magnitude.The conductivity prefactors and the activation energies as calculated from the conductivity versus temperature data obey the Meyer–Neldel relation with an axio-intercept of about 10−7(V cm)−1and a slope of about 35meV.Thus,Si-rich a-SiN x (:H)films behave like a defect doped n-type semiconductor.The activation energy of the conductance in films with the same composition changes with the hydrogen content.A minimum in the activation energy correlates with a minimum of the surface recombination velocity at the a-SiN x /p-Si interface.We ascribe both effects to a variation of the Fermi level in the films.©2000Elsevier Science S.A.All rights reserved.Keywords :Silicon nitride;Sputtering;Conductivity;Solar cell;Hetero junction;Field-effect passivation /locate /mseb1.IntroductionAmorphous silicon nitride is widely used in the pro-duction of microelectronic devices and also serves as anti-reflection coating for silicon solar cells.In most cases,nearly stoichiometric,hydrogen containing sili-con nitride (a-Si 3N 4(:H))is prepared by using dichlorsi-lane (SiCl 2H 2)or silan (SiH 4)and ammonia (NH 3)as reactants either at high temperatures of about 800°C by chemical vapor deposition (CVD)or at lower tempera-tures of about 300°C by plasma enhanced CVD (PECVD).The CVD silicon nitride contains only a small amount of hydrogen (B 5at%)and exhibits a high defect density (N Def B (1019cm −3)whereas in the PECVD silicon nitride up to 30at%of hydrogen saturates many of the defect states,thereby reducing N Def to the order of 1017cm −3[1].However,it is the high defect density of hydrogenated amorphous silicon nitride (a-SiN x (:H),x =composition ratio N /Si)which is exploited for certain applications.From intense re-search on the defects in stoichiometric a-Si 3N 4(:H)[2,3]it is known that the electrical active defects result from silicon dangling bonds which create a band of elec-tronic states about 3eV above the valence band of a-Si 3N 4(:H).As a result of the high defect density silicon nitride develops a high fixed charge close to the surface when it comes in contact with silicon (Si).This effect is used in the fabrication of oxide nitride oxide semiconductor (ONOS)structures of storage field-effect transistors [4]and also in the antireflection coating of metal insulator semiconductor /inversion layer (MIS /IL)solar cells [5].The common feature of these applica-tions is that the fixed charges in the silicon nitride induce a counter charge in the Si which leads to a space charge region and a strong band bending at the Si surface.In the case of MIS /IL solar cells the a-SiN x (:H)antireflection coating is able to induce a strong band bending resulting in an inversion layer and creating a pn-junction.Recent investigations of MIS /IL solar cells show that the creation of this inversion layer is much improved when non-stoichiometric Si-rich a-SiN x (:H)is prepared [6].The inversion layer leads to a drastically reduced majority carrier concentration resulting in a much lower recombination rate of light generated charge carriers.Through this field effect passivation [7]*Corresponding author.Tel.:+49-711-685-7175.E -mail address :rojahn@ipe.uni-stuttgart.de (M.Rojahn)0921-5107/00/$-see front matter ©2000Elsevier Science S.A.All rights reserved.PII:S 0921-5107(99)00399-2M.Vetter,M.Rojahn/Materials Science and Engineering B71(2000)321–326 322surface recombination velocities lower than at the best SiO2/Si interfaces can be achieved.This study is aimed at a better understanding of the physical processes involved in thefield effect passivation of a-SiN x(:H)/Si junction through the analysis of experimental data.By varying the composition ratio of Si-rich a-SiN x(:H) from x=1.33to0the bandgap is continuously reduced from about5eV at nearly stoichiometric composition to less than2eV,the value of a-Si(:H)[8,9].Thereby, a-SiN x(:H)evolves from an insulator to a wide band gap semiconductor causing the SiN x(:H)/Si contact to behave rather like a hetero junction.In our former investigations[8],we have prepared Si-rich a-SiN x(:H)films by rf-magnetron sputtering from a Si target in an Ar/N2plasma,adding H2if required.We have found a minimum of the surface recombination velocity at a certain amount of hydrogen in the plasma.We con-cluded that Si-rich a-SiN x(:H)is in effect a defect doped n-type semiconductor whose doping and there-fore Fermi level is somehow determined by the hydro-gen content[8].In order to understand the complex situation at the Si-rich a-SiN x(:H)/Si junction the role of hydrogen from the plasma and/or the a-SiN x(:H)film in saturating dangling bonds at the Si surface must also be analyzed.This may help to explain the reduc-tion of the surface state density at the Si interface which is a prerequisite for the formation of an inversion layer and a large space charge region[7].We prepare a-SiN x in Ar/NH3/H2plasma on p-Si and on glass substrate in an attempt tofind out whether the use of NH3(10vol% NH3in Ar)instead of N2as nitrogen source leads to any significant changes in the material properties of the a-SiN x(:H).As it turns out,the amount of hydrogen in the plasma resulting from the dissociation of NH3has little influence on the surface recombination velocity, i.e.the amount of hydrogen we have to add to either plasma in order to obtain a specific surface recombina-tion velocity is about the same.In our analysis we take the surface recombination velocity of photogenerated carriers at the p-Si surface as a means for evaluating the band bending at the Si surface.We also determine the temperature dependent dark dc-conductivity of a-SiN x(:H)films on glass.The slope of the conductivity in the Arrhenius plot is used to calculate the activation energy E a in Si-rich a-SiN x(:H)according to[10]|(T)=|0=exp(E a/kT),(1) where|0is the conductivity prefactor.By indicating the position of the Fermi level relative to the conduction band,the activation energy E a serves as a measure of the defect doping level.The correlation between|0and E a is referred to as the Meyer–Neldel relation which is described by[10]asln|0=ln|00+E a/kT m.(2)According to the Meyer–Neldel relation all tempera-ture dependent data yield the same conductivity|00attemperature T m.2.ExperimentalFor the measurement of the effective surface recom-bination velocity of minority carriers S eff we use doublesided polished single crystalline p-type Si wafers(FZ, B100\oriented,boron doped(N D=5×1016cm−3, thickness380m m)covered with a thermal oxide(100nm)one side(called backside in the following).Thesamples are treated by a RCA cleaning procedure fol-lowed by an HF-dip(1%).Thereafter,a SiN x(:H)layerof approximately100nm is deposited on the frontsideof the wafer from an intrinsic Si-target in an Ar/NH3/ H2plasma using a modified rf(13.56MHz)magnetron sputtering device(Varian VT118)[11].The deposition rate at rf-powers of100,200and275W are about12, 40and70nm min−1,respectively.In every run a glass substrate is placed next to the Si wafers so that almost identical SiN xfilms can be tested for their dark dc-con-ductivity.For this purpose,we evaporate aluminium contacts in a double line configuration,5mm in length, 1mm apart,using a shadow mask onto the a-SiN x(:H) on glass layer.In order to determine the effective recombinationvelocity S eff at the Si/SiN x interface,we measure theeffective lifetime~eff of minority carriers photogener-ated in the p-Si wafer using the microwave detectedphoto conductivity decay(MW-PCD)combined withthe Korona method[12,13].The effective lifetime re-sults from volume and interface recombination and itholds~−1eff=~−1v+~−1s,where~s is the surface minority carrier lifetime and~v the volume lifetime.Before the measurement,the oxidized side of the wafer is charged in a Korona chamber at16kV,resulting in an excellent field effect passivation of the backside of the wafer and a negligible recombination rate.Consequently,only~s at the SiN x/St-interface and~v contributes to~eff.The volume lifetime~v obtained from a separate,double-sided oxidized and Korona-charged wafer using MW-PCD is about~vol:350m s.Thus,~eff is a direct measure of the S eff at the SiN x/Si-interface.In order to compute S eff we useS eff='D n(~vol−~eff)vol efftan W'~vol−~effD n vol eff,(3)where D n is the minority diffusion constant(28cm2s−1 in0.35V cm FZ p-Si)and W the wafer thickness.In the experimental setup(Phoenicon MRM[12,13]) excess charge carriers are generated in the SiN x(:H) coated Si wafer by a20ns laser puls(1047nm).In order to simulate solar illumination we use a bias light of about150mW cm−2(halogen lamp).The transientM .Vetter ,M .Rojahn /Materials Science and Engineering B 71(2000)321–326323of the photoconductivity after the laser puls is detected by the microwave reflection of the wafer using a Gunn diode (10GHz)as the microwave source.~eff is ob-tained from an exponential fit of the decay of the transient.We measure the temperature dependent dc-conduc-tivity (Keithley 617Electrometer)of SiN x (:H)films on glass by applying 100V to the A1contacts and varying the temperature between 333and 213K.At tempera-tures above 350K the conductivity versus temperature curves are irreversibly shifted to lower conductivities.We attribute this mainly to an out-diffusion of hydro-gen and an annealing of defects in the films at higher temperatures.As a measure of the composition of the a-SiN x (:H)films,we determine the quasi-static refrac-tive index n of all films on the p-Si wafers using an ellipsometer with a rotating analyzer (SE400,Sentec Instruments,u =632.8nm).The thickness of the films is measured independently with a thickness profiler (Sloan,DekTak 3030)which enables us to use theabsorbance of Si-rich films as a free parameter in the fitting routine of the ellipsometer.3.ResultsThe composition of a-SiN x (:H)films can be varied from nearly stoichiometric (x =1.33,n =1.97)to a-Si(:H)(x =0,n :3,3)by adapting plasma parameters like gas partial pressures and rf input power [11].By changing the composition from stoichiometric a-Si 3N 4(:H)to a-Si(:H)the optical bandgap decreases from about 5to 2eV [8,9].Si-rich a-SiN x (:H)films with a refractive index n \2.1and an optical bandgap of about 3eV result in a good electronic passivation of p-Si surfaces which can be ascribed to field-effect passi-vation caused by the creation of an inversion layer at the silicon surface.However,the deposition tempera-ture and the rf input power into the plasma as well as the hydrogen content in the films must be carefully balanced in order to achieve values of S eff lower than 100cm s −1[6,8].For several input power values the effective minority carrier lifetime ~eff in p-Si reaches a maximum (resp.S eff a minimum)at deposition temper-atures of about 170–200°C,as illustrated in Fig. 1.Here,for the Si wafers with the Korona passivated backside,the effective minority carrier lifetime is di-rectly related to S eff at the Si surface.Similarly,Fig.2shows that for a variety of deposition temperatures a maximum of the effective lifetime ~eff is obtained at a certain amount of H 2in the plasma which presumably yields to a specific hydrogen content of the a-SiN x (:H)layers.As proposed in [8],the field-effect passivation of Si by a-SiN x (:H)is correlated to the passivation of Si dangling bonds at the Si surface and to the position of the Fermi levels in Si and a-SiN x (:H).Based on the model of a hetero junction,the band bending in the p-type Si is strongest as the Fermi level in a-SiN x (:H)gets close to the conduction band of a-SiN x (:H)leading to a maximum conductivity in the material.From the dark dc-conductivity of a-SiN x (:H)-films with different composition and hydrogen content we infer the posi-tion of the Fermi level,according to Eq.(1).In order to separate the influence of differing Si to N ratios (as expressed by the varying refractive index n )from the influence of changing hydrogen content in the samples,we prepare one series of samples with zero hydrogen content but various refractive index and a number of series of samples with similar composition ratios but varying hydrogen contents.We expect a change of the composition ratio (Si /N)to have the following consequences on the activation energy of a-SiN x (:H).It is known from photoemission measure-ments [14]that the smaller bandgap when preparing Si-rich a-SiN x (:H)is mostly caused by a lower conduc-tion band level.Given that the defect band of Si-dan-Fig.1.The effective life time ~eff depends on the deposition tempera-ture and rf-input power into the plasma.The maximum value of ~eff is reached at a deposition temperature of about 200°C.Fig.2.In order to reach a maximum effective lifetime ~eff of photo induced carriers H 2must be added to the NH 3/Ar plasma.M .Vetter ,M .Rojahn /Materials Science and Engineering B 71(2000)321–326324Fig.3.The dark dc-conductivity of Si-rich SiN x prepared without hydrogen changes strongly with the composition,here indicated by the different refractive indexes of the samples.quence,we fixed the upper temperature limit to about 60°C.The hydrogen flow in the plasma not only determines the effective lifetime,as demonstrated in Fig.2,but also effects the defect doping and therefore the Fermi level in the SiN x (:H)-films,as shown in Fig.4.The activation energy E a changes by more than 200meV and reaches a minimum at the same hydrogen flow which leads to a maximum effective lifetime ~eff (see Fig.2),as does the conductivity prefactor |0.The actual dark dc-conductivity shows a maximum at that point.In contrast to the just described response of the activation energy to different hydrogen content in the plasma,the activation energy of the samples shown in Fig.1remains almost constant throughout the varying deposition temperatures.A scatter of less than 100meV is caused by slightly differing refractive indexes,result-ing from minor changes in the composition of the samples.Fig.5summarizes the activation energy data which we obtained for a large number of samples.The overall trend shows hydrogenated SiN x (:H)samples with higher refractive index than those samples prepared without hydrogen.It can also be observed that the activation energy E a tends to be higher for films con-taining hydrogen.To be specific,of samples which show the same refractive index those prepared with hydrogen will have a significantly higher activation energy E a The large scattering of the activation energy of about 400meV for a specific refractive index mainly results from the different hydrogen content of the sam-ples.Similarly,the conductivity prefactor |0as calcu-lated from Eq.(1)changes with the refractive index and the hydrogen content of the SiN x (:H)films.The con-Fig.4.The comparison with Fig.2shows that the activation energy E a and the extrapolated conductivity prefactor |0reach their mini-mum when the effective life time ~eff has the maximum value.Fig.5.The activation energy of a-SiN x films without hydrogen is lower than that of hydrogen containing samples.The scatter in the activation energies of different hydrogen containing SiN x (:H)films mainly results from varying hydrogen content in the films.The lines are for orientation and show a decreasing activation energy as the silicon content rises.gling bonds in the middle of the bandgap determines the Fermi level,a change of the composition from stochiometric to a-Si should therefore reduce the activa-tion energy E a .This is demonstrated in Fig.3where the Arrhenius plots of the zero hydrogen series of films is shown.A lower activation energy clearly correlates with a higher refractive index n ,indicating the more a-Si-like composition of the material.Also,a decreasing slope of the plots at lower temperatures can be ob-served which may hint a change in the transport mech-anism but has yet to be verified through further research.The slope of the Arrhenius plots similarly increases with rising temperature for the hydrogen con-taining samples.As already mentioned,at temperatures above approximately 100°C the electronic properties of these samples are irreversibly changed.As a conse-M.Vetter,M.Rojahn/Materials Science and Engineering B71(2000)321–326325Fig. 6.Conductivity prefactor|0versus activation energy E a for SiN x(:H)films.As a reference,the Meyer–Neldel relation for doped and undoped a-Si(:H)[10]is also shown by dashed line.tent in SiN x(:H)results in a much enhanced conductiv-ity of the samples.This behavior is explained by a change of the electronical structure of a-SiN x(:H). Amorphous semiconductors do not exhibit a sharp bandgap but have large bandtails extending into the bandgap leading to located states[15].Also,as opposed to crystalline material,defects with similar origin do not generate the same discrete defect energy level.In-stead,resulting from the disorder in the material,they create a distribution of defects leading to a defect band in the bandgap.In a-Si3N4(:H)this defect band is created by Si dangling bonds and lies about3eV above the valence band.The nitrogen dangling bonds couple with the lone pair electrons of the N-atoms leaving rather less active electrical states in and near above the valence bandtail[16].In nearly stoichiometric a-Si3N4(:H),an insulator with a bandgap of about5eV, tunneling processes in the defect band are the dominant conduction mechanism[17].Photoemission studies show that changing a-SiN x(:H)to more Si-rich compo-sitions by substituting N-atoms for Si-atoms in the random network structure does not much influence the valence bandtail.However,the introduced Si dangling bonds lead to an extension of the defect band.In addition,non-binding states result in a more extended conduction bandtail.This change of the conduction bandtail causes a decrease of the optical bandgap whereas the increasing overlap between the conduction bandtail and the defect band results in a lower activa-tion energy E a and a higher dark dc-conductivity,as shown in Figs.3and5.Thus,Si-rich a-SiN x(:H)can be considered a defect doped n-type semiconductor.In the same way,the generally lower conductivity and larger activation energy E a of hydrogen containing films can be understood.By saturating the Si dangling bonds,the hydrogen reduces the extension of the con-duction bandtail and the defect band resulting in a smaller overlap of both bands.However,this theory of a rising activation energy with increasing hydrogen content in thefilms clashes with the observation of a minimum of the activation energy for a specific hydro-genflow into the plasma,as shown in Fig. 4.Pre-sumably,the defect state density is not homogeneously reduced through the hydrogen incorporation.Further research has yet to solve this question.In any case,an even further increased hydrogen content in thefilms will lead to even more dangling bonds being saturated, thereby lowering the defect dominated conductivity.A clear correlation between the minimum of the surface recombination velocity S eff and the minimum of the activation energy E a is apparent from Figs.2and4. Thus,we are lead to the assumption that an increasing Fermi level due to a rising hydrogen content results in a stronger band bending at the p-Si surface.This in turn leads to a inversion which causes the minimum of the surface recombination velocity S eff.ductivity prefactor|0exhibits a similar dependence on the refractive index and the hydrogen content of the samples as the activation energy E a which is expressed in the Meyer–Neldel-relation,as shown in the following.Our temperature dependent conductivity data are now used to verify the Meyer–Neldel-relation for the SiN x(:H)samples.The Meyer–Neldel-relation,as given by Eq.(2),states an exponential relationship between the conductivity prefactor|0and the temperature with a pre-conductivity prefactor|00.This relationship holds for a-Si(:H)and also m c-Si[19,20],|00ranging between 0.1(V cm)−1and0.01(V cm)−1.In Fig.6prefactors|0 are plotted against the activation energy E a Clearly,the Meyer–Neldel-relation holds also for our SiN x(:H) samples with|00of about10−7(V cm)−1and kT m of about35meV,respectively T m=400K.4.DiscussionIn order tofind out how passivation of the surface of a silicon wafer at the p-Si/a-SiN x-interface can be ex-plained,we prepare series of SiN x(:H)films with vary-ing composition ratios and hydrogen content at different deposition temperatures.We obtain the effec-tive surface recombination velocity S eff using the mi-crowave detected photo conductivity decay combined with the Korona method as well as the temperature dependent conductivity of ourfilms and their refractive index.Let us now explain the strongfield-effect passi-vation of the silicon surface in terms of the energy of the bandgap,the position of the Fermi level and the conduction mechanisms in Si-rich a-SiN x(:H)films.The decreasing activation energy E a with increasing Si-con-M.Vetter,M.Rojahn/Materials Science and Engineering B71(2000)321–326 326However,the above model cannot explain the depen-dence of the photocarrier lifetime in p-Si on the SiN x deposition temperature.This is because the activation energy of the conductance of SiN x(:H)-films does not depend on this deposition temperature.Thus,the maxi-mum of~eff at200°C in Fig.1is not explained by optimumfield effect passivation but by the maximum reduction of surface states at this deposition tempera-ture[18].5.ConclusionsWe have compared the effective lifetime of minority carriers in p-type Si passivated with SiN x with the dark conductivity data of corresponding a-SiN x-films of var-ious composition.We explain the decreasing activation energy of the conductance with increasing Si content in the SiN x-films by a change of the conduction bandtail and an increasing overlap between the conduction bandtail and the defect band.Our SiN x-films obey the Meyer–Neldel-relation;thus,Si-rich a-SiN x can be con-sidered a defect-doped n-type semiconductor.The max-imum lifetime of minority carriers in p-Si at the p-Si/SiN x-interface correlates with the minimum of the activation energy of the conductance of the SiN x-films. From our experiments with SiN x-films of various hy-drogen content we conclude that the Fermi level in-creases with a rising hydrogen content.Thus,the formation of an inversion causes the minimum of the surface recombination velocity. AcknowledgementsWe like to thank Wacker Siltronic AG,Burghausen for the support with Si wafers and B.Winter for oxidation of Si wafer.We thank I.Ko¨tschau and U. Rau for valuable discussions.References[1]J.Robertson,M.J.Powell,Appl.Phys.Lett.44(1984)415.[2]J.Robertson,W.L.Warren,J.Kanicki,J.Non-Crystal.Solids187(1995)297.[3]W.L.Warren, C.H.Seager,J.Robertson,J.Kanicki, E.H.Poindexter,J.Electrochem.Soc.143(1996)3685.[4]H.S.Momose,T.Morimoto,Y.Ozawa,K.Yamabe,H.Iwai,IEEE Trans.Electr.Dev.ED42(1995)704.[5]A.G.Aberle,R.Hezel,Progr.Photovolt.:Res.Appl.5(1997)29.[6]uinger,A.G.Aberle,R.Hezel,Proc.14.Europ.Photo-voltaic Solar Energy Conf.,vol.1,H.S.Stephens&Associates, Bedford,1997,p.853.[7]A.G.Aberle,S.Glunz,W.Warta,J.Appl.Phys.71(1992)4422.[8]M.Vetter,Thin Solid Films337(1999)118.[9]H.R.Philipp,J.Electrochem.Soc.120(1973)295.[10]R.A.Street,Hydrogenated amorphous silicon,Cambridge Uni-versity Press,Cambridge,1991,p.227.[11]M.Vetter,Dissertation,University of Stuttgart,Shaker Verlag,Aachen,1999.[12]M.Schfthaler,U.Rau,W.Fssel,J.H.Werner,Proc.23rd IEEEPhotovolt.Spec.Conf.[13]M.Schfthaler,Dissertation,University of Stuttgart,Shaker Ver-lag,Aachen,1995.[14]R.Krcher,L.Ley,R.L.Johnson,Phys.Rev.B30(1896)1984.[15]R.A.Street,Hydrogenated Amorphous Silicon,Cambridge Uni-versity Press,Cambridge,1991,p.135.[16]J.Robertson,Philos.Mag.B63(1991)47.[17]S.M.Sze,J.Appl.Phys.38(1967)295.[18]M.Vetter,M.Jetter,Proc.14th Europ.PVSEC,vol.2,H.S.Stephens&Associates,Bedford,1997,p.2473.[19]R.Bru¨ggemann,M.Rojahn,M.Ro¨sch,Normal and InvertedMeyer-Neldel Rule in Hot-Wire CVD Deposited Nanocrystalline Silicon,Rapid Research Notes,Phys.Stat.Sol.(A)166,R11 (1998).[20]R.Bru¨ggemann,A.Hierzenberger,P.Reinig,et al.,Electronicaland optical properties of hotwire deposited microcrystalline sili-con,J.Non-Cryst.Solids227–230(1998)982..。