《复合材料 Composites》课程教学资源（学习资料）第二章 增强体_Catalytic growth of nanowires：Vapor–liquid–solid, vapor–solid–solid, solution–liquid–solid and solid–liquid–solid growth

Science Direct Current Opinion in Solid state Materials science ELSEVIER Current Opinion in Solid State and Materials Science 10(2006)182-191 Catalytic growth of nanowires: Vapor-liquid-solid vapor-solid-solid, solution-liquid-solid and solid-liquid-solid growth Kurt w. Kolasinski Department of Chemistry, West Chester Unirersity, West Chester, PA 19383, United States Received 12 March 2007: accepted 12 March 2007 Abstract Catalytic growth is a powerful tool to form a variety of wire( whisker) like structures with diameters ranging from just a few nano- metres to the millimetre range. A range of phases(gas, solid, liquid, solution and supercritical fluid) have been used for the feeder phase. i.e. the source of material to be incorporated into the nanowire Solid, liquid, eutectic, alloy and metastable phases have all been invoked to explain the structure of the catalytic particle. Rather than focussing on the differences that lead to the proliferation of an alphabet soup of names for the various growth techniques, this review attempts to focus on the similarities between all of these catalytic growth pro- ses in an attempt to help stimulate a more universal understanding of the phenomenon. The review begins with a precis of the mate- rials from which nanowires have been formed and then proceeds to a discussion of mechanistic aspects e 2007 Elsevier Ltd. All rights reserved 1. Introduction tors have also been gre ImaI nanowires, and are prized for their potential in electronic As a controlled means of growing whiskers and more optoelectronic and sensing applications[6-8]. For more recently nanowires, catalytic growth of solid structures on the potential of these nanostructures in applications traces back to the discovery of Wagner and Ellis [1] that the reader is referred to these recent reviews. Buhro and Si whiskers could be grown by heating a Si substrate in a co-workers [9] have reviewed the formation of semicon- mixture of SiCl4 and H2 with their diameters determined ductor nanowires from solutions and supercritical fluids by the size of Au particles that had been placed on the sur- Here I concentrate on the production of ID nanostructures face prior to growth. Of course, the catalytic growth of car- with the use of vapor phase transport and surface diffusion. bon fibres has long been a recognized problem in the field In the literature we might variously encounter nanowires of catalysis [2]. In this case, such growth must be avoided, (solid core structures with diameters below 100 nm), for instance, in the steam reforming of CH, over Ni cata- nanotubes(single or multi-walled hollow core structures ts, which is the primary industrial source of H with diameters below M100 nm) and whiskers(larger solid The poster child of one-dimensional(ID)nanostruc- core structures). For simplicity, I will use the term nano- tures is the carbon nanotube(CNT) either in single-walled wire generically to describe the structures formed by cata- (SW-CNT) or multiwalled variants (MW-CNT). They are lytic growth unless I specifically want to call attention to valued for a wide range of extreme properties for electronic nanotubes or whiskers. This review does not attempt to applications, for their high thermal conductivity and for be exhaustive. Rather it looks first at a number of materials their high strength [2-5]. A number of other semiconduc- systems that have been grown catalytically in the form of nanowires, nanotubes or whiskers in the past year or two. Then a review of the mechanistic aspects of catalytic E-inail address: kkolasinski(@wcupaedt nanowire growth is made 1359-0286/- see front matter 2007 Elsevier Ltd. All rights reserved doi:10.1016 cossms.2007.03.002

Catalytic growth of nanowires: Vapor–liquid–solid, vapor–solid–solid, solution–liquid–solid and solid–liquid–solid growth Kurt W. Kolasinski Department of Chemistry, West Chester University, West Chester, PA 19383, United States Received 12 March 2007; accepted 12 March 2007 Abstract Catalytic growth is a powerful tool to form a variety of wire (whisker) like structures with diameters ranging from just a few nano￾metres to the millimetre range. A range of phases (gas, solid, liquid, solution and supercritical fluid) have been used for the feeder phase, i.e. the source of material to be incorporated into the nanowire. Solid, liquid, eutectic, alloy and metastable phases have all been invoked to explain the structure of the catalytic particle. Rather than focussing on the differences that lead to the proliferation of an alphabet soup of names for the various growth techniques, this review attempts to focus on the similarities between all of these catalytic growth pro￾cesses in an attempt to help stimulate a more universal understanding of the phenomenon. The review begins with a pre´cis of the mate￾rials from which nanowires have been formed and then proceeds to a discussion of mechanistic aspects. 2007 Elsevier Ltd. All rights reserved. 1. Introduction As a controlled means of growing whiskers and more recently nanowires, catalytic growth of solid structures traces back to the discovery of Wagner and Ellis [1] that Si whiskers could be grown by heating a Si substrate in a mixture of SiCl4 and H2 with their diameters determined by the size of Au particles that had been placed on the sur￾face prior to growth. Of course, the catalytic growth of car￾bon fibres has long been a recognized problem in the field of catalysis [2]. In this case, such growth must be avoided, for instance, in the steam reforming of CH4 over Ni cata￾lysts, which is the primary industrial source of H2. The poster child of one-dimensional (1D) nanostruc￾tures is the carbon nanotube (CNT) either in single-walled (SW-CNT) or multiwalled variants (MW-CNT). They are valued for a wide range of extreme properties for electronic applications, for their high thermal conductivity and for their high strength [2–5]. A number of other semiconduc￾tors have also been grown in 1D structures, primarily nanowires, and are prized for their potential in electronic, optoelectronic and sensing applications [*6–*8]. For more on the potential of these nanostructures in applications, the reader is referred to these recent reviews. Buhro and co-workers [*9] have reviewed the formation of semicon￾ductor nanowires from solutions and supercritical fluids. Here I concentrate on the production of 1D nanostructures with the use of vapor phase transport and surface diffusion. In the literature we might variously encounter nanowires (solid core structures with diameters below 100 nm), nanotubes (single or multi-walled hollow core structures with diameters below 100 nm) and whiskers (larger solid core structures). For simplicity, I will use the term nano￾wire generically to describe the structures formed by cata￾lytic growth unless I specifically want to call attention to nanotubes or whiskers. This review does not attempt to be exhaustive. Rather it looks first at a number of materials systems that have been grown catalytically in the form of nanowires, nanotubes or whiskers in the past year or two. Then a review of the mechanistic aspects of catalytic nanowire growth is made. 1359-0286/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.cossms.2007.03.002 E-mail address: kkolasinski@wcupa.edu Current Opinion in Solid State and Materials Science 10 (2006) 182–191

K w. Kolasinski Current Opinion in Solid State and Materials Science 10(2006 )182-191 183 2. Materials systems upper limit on the Ge fraction that can be obtained in this mDe Silicon was the system first investigated by Wagner and Silanes do not have to be used as the source material for Ellis [1] and it remains one of the most intensively studied the growth of SiNW and Si- Gex nanowires. Dujardin systems [7, 10-14, 15, 16-19]. Lieber's group["8] has stud- et al. [10] have used a molecular beam epitaxy(MBe) ied silicon extensively including the formation of branched source and a substrate coated with a thin film of Au for this Si nanowires(SINW)[20]. Carbon nanotubes [2, 21, 22]and purpose. The use of an MBE source is quite significant carbon nanofibers [23] are also produced by catalytic because of a basic difference in the dynamics of the interac growth [24, 25]. Heterojunctions between SiNW and CNT tion of an atomic vapor of Si as compared to SiH4 have been formed [26]. Other materials that exhibit cata- Whereas the sticking coefficient of SiH4 is low on a Si lytic growth of nanowires include Sio(a substoichiomet- and practically zero on a H-terminated Si surface, it is ric silicon oxide)[27]: SiO2 [28, 29] Sil- Gex [10, 30] Ge much higher on the surface of a metal catalyst. In contrast, 61, 323: AIN [33] r-Al2O3 [34]: oxide-coated B[35]; the sticking coefficient of Si atoms is unity on a clean or H- CNx [36]: Cdo [37]: Cds [38]: CaSe [9] CdTe [9]: a- terminated Si surface as well as on a metal catalyst particle Fe2O3(hematite), &-Fe2O3 and Fe3 O4(magnetite)[39] Sun et al. [31] have created Ge nanowires( GeNw) by GaAs [15,40, *41, 42, 43, 44]: Gan [18: Ga2O3 [18, 45]; evaporation of Ge powder in flowing Ar at 600C.Au GaP [40, *41, 46] InAs ["41, 471: InN (hexangular struc- nanoparticles act as catalysts and are supplied from a col- tures)[48] InP [9, 41,42]: In2O3 [45]: In2 Se3 [49] LiF lodal solution of thiol-capped Au. The nanowires have [50] SnO2[45, 51, * 52: ZnO nanowires ["7, 8,53and nano- uniform diameters of 30 nm, are up to tens of microme- plates [53]; ZnS [54]: ZnSe [55]: Mn doped Zn2SO4 [56]; and tres in length and have a Au particle on their tip. The ori- ZnTe [57]. Let us now look at the conditions under which ginal size of the colloidal Au particles was 2 nm; hence, catalytic growth has been used to create nanostructures significant aggregation must have occurred. Here we have that we can better understand the range of growth condi- vapor phase transport of atomic Ge to already formed tions that have been used, as well as the similarities and dif- Au catalytic particles ferences in growth characteristics that have been observed Chandrasekaran et al.[32] use a solid-liquid-solid so that we may generalize about some of the important method to produce GeNW. Many um long GeNW are mechanistic characteristics observed to emanate from the same mm diameter Ga cat- An exciting development in the growth of single-walled alyst particle(multiprong root growth, as defined in the carbon nanotubes(SW-CNT)was reported by Takagi et al. next section). In and Sn are also suitable catalysts. A thin [58]. A range of metal catalysts have been shown previ- film of the molten metal is spread on a Ge(1 00) wafer, ously to work for the synthesis of carbon fibres and CNTs which is then exposed to a microwave plasma struck in [2]. Takagi et al. have shown that pyrolysis of ethanol can H2/N2. Ground NaCl is placed around the substrate. This be used in the presence not only of Fe, Co or Ni(the most acts as a source of Cl, which in combination with the common catalysts) or Pt and Pd(which had been previ- plasma acts to transport the Ge. Alternatively, a quartz ously reported) but also the coinage metals(Cu, Ag and substrate can be coated with the metal, and a powdered Au). For the latter three metals to work not only do they mixture of NaCl and Ge is placed around the substrate have to be clean to start with, they must also be smaller Excitation involving a plasma is again used to elicit trans- than 5 nm in diameter for growth to be efficient. Their port and engender GeNw growth explanation is that the metal particles are in a cluster-like A Si wafer rather than a powdered sample has been used structure rather than a crystalline state, and C atoms are as the source material for simultaneous growth of Sio soluble in these clusters. Then, C atoms might precipitate nanowires and SnO, nanobelts by Zhang et al. [27]. SnCl2 formation of a hemispherical cap with a graphitic structure Ar t O2 to 950C for 2 h The Sn catalyst particles, which as the precursor of Sw-CNT growth. They propose that also contain several percent Si and O, form at the base of the essential role of metal particles is to provide a platform the SiOx nanowires rather than the tips. Both the SiOx on which carbon atoms can form a hemispherical cap from nanowires and the SnO2 nanobelts experience multiprong which SW-CNT grow in a self-assembled fashion. root growth Si_ Gex nanowires can be grown with a Au catalyst Zhang et al. [38] also used catalysis over Sn nanoparti much as SiNW can be grown as shown by Lew et al. cles to form branched CdS nanowires. They mixed Cds [30]. Growth was carried out in an isothermal quartz tube SnO2 and graphite powders (in a 1: 1: I ratio) and heated reactor at 325-525C with a total reactor pressure of them to 1200C for 2 h under a constant flow of Ar In this 13 Torr composed of a 10% mixture of SiH4 in H2 and case the Sn particle at the end of the nanowire is signifi- either a 1% or 2% mixture of GeH in H, as source gases. cantly larger than the diameter of the wire that grows from Higher temperatures(>375C) favour the growth of Si rich it. In both of the systems investigated by Zhang et al (x<0.5)Sil_Gex nanowires; however, Ge thin film depo- vapor phase transport is used not only to supply the sition on the outer surface of the wire was observed at growth material, but also to supply the material that forms increased GeH/(SiH4+ GeH,)inlet gas ratios setting an the catalytic particle

2. Materials systems Silicon was the system first investigated by Wagner and Ellis [1] and it remains one of the most intensively studied systems [7,10–14,*15,16–19]. Lieber’s group [*8] has stud￾ied silicon extensively including the formation of branched Si nanowires (SiNW) [20]. Carbon nanotubes [2,21,22] and carbon nanofibers [23] are also produced by catalytic growth [24,25]. Heterojunctions between SiNW and CNT have been formed [26]. Other materials that exhibit cata￾lytic growth of nanowires include SiOx (a substoichiomet￾ric silicon oxide) [27]; SiO2 [28,29]; Si1xGex [10,30]; Ge [31,*32]; AlN [33]; c-Al2O3 [34]; oxide-coated B [*35]; CNx [36]; CdO [37]; CdS [38]; CdSe [*9]; CdTe [*9]; a￾Fe2O3 (hematite), e-Fe2O3 and Fe3O4 (magnetite) [39]; GaAs [15,40,*41,42,*43,44]; GaN [18]; Ga2O3 [18,45]; GaP [40,*41,*46]; InAs [*41,*47]; InN (hexangular struc￾tures) [48]; InP [*9,*41,42]; In2O3 [45]; In2Se3 [49]; LiF [50]; SnO2 [45,51,*52]; ZnO nanowires [*7,*8,53] and nano￾plates [53]; ZnS [54]; ZnSe [55]; Mn doped Zn2SO4 [56]; and ZnTe [57]. Let us now look at the conditions under which catalytic growth has been used to create nanostructures so that we can better understand the range of growth condi￾tions that have been used, as well as the similarities and dif￾ferences in growth characteristics that have been observed so that we may generalize about some of the important mechanistic characteristics. An exciting development in the growth of single-walled carbon nanotubes (SW-CNT) was reported by Takagi et al. [*58]. A range of metal catalysts have been shown previ￾ously to work for the synthesis of carbon fibres and CNTs [2]. Takagi et al. have shown that pyrolysis of ethanol can be used in the presence not only of Fe, Co or Ni (the most common catalysts) or Pt and Pd (which had been previ￾ously reported) but also the coinage metals (Cu, Ag and Au). For the latter three metals to work not only do they have to be clean to start with, they must also be smaller than 5 nm in diameter for growth to be efficient. Their explanation is that the metal particles are in a cluster-like structure rather than a crystalline state, and C atoms are soluble in these clusters. Then, C atoms might precipitate to cover the surface of the nanoparticles, resulting in the formation of a hemispherical cap with a graphitic structure as the precursor of SW-CNT growth. They propose that the essential role of metal particles is to provide a platform on which carbon atoms can form a hemispherical cap from which SW-CNT grow in a self-assembled fashion. Si1xGex nanowires can be grown with a Au catalyst much as SiNW can be grown as shown by Lew et al. [30]. Growth was carried out in an isothermal quartz tube reactor at 325–525 C with a total reactor pressure of 13 Torr composed of a 10% mixture of SiH4 in H2 and either a 1% or 2% mixture of GeH4 in H2 as source gases. Higher temperatures (>375 C) favour the growth of Si rich (x < 0.5) Si1xGex nanowires; however, Ge thin film depo￾sition on the outer surface of the wire was observed at increased GeH4/(SiH4 + GeH4) inlet gas ratios setting an upper limit on the Ge fraction that can be obtained in this temperature range. Silanes do not have to be used as the source material for the growth of SiNW and Si1xGex nanowires. Dujardin et al. [10] have used a molecular beam epitaxy (MBE) source and a substrate coated with a thin film of Au for this purpose. The use of an MBE source is quite significant because of a basic difference in the dynamics of the interac￾tion of an atomic vapor of Si as compared to SiH4. Whereas the sticking coefficient of SiH4 is low on a Si and practically zero on a H-terminated Si surface, it is much higher on the surface of a metal catalyst. In contrast, the sticking coefficient of Si atoms is unity on a clean or H￾terminated Si surface as well as on a metal catalyst particle. Sun et al. [31] have created Ge nanowires (GeNW) by evaporation of Ge powder in flowing Ar at 600 C. Au nanoparticles act as catalysts and are supplied from a col￾loidal solution of thiol-capped Au. The nanowires have uniform diameters of 30 nm, are up to tens of microme￾tres in length and have a Au particle on their tip. The ori￾ginal size of the colloidal Au particles was 2 nm; hence, significant aggregation must have occurred. Here we have vapor phase transport of atomic Ge to already formed Au catalytic particles. Chandrasekaran et al. [*32] use a solid–liquid–solid method to produce GeNW. Many lm long GeNW are observed to emanate from the same mm diameter Ga cat￾alyst particle (multiprong root growth, as defined in the next section). In and Sn are also suitable catalysts. A thin film of the molten metal is spread on a Ge(1 0 0) wafer, which is then exposed to a microwave plasma struck in H2/N2. Ground NaCl is placed around the substrate. This acts as a source of Cl, which in combination with the plasma acts to transport the Ge. Alternatively, a quartz substrate can be coated with the metal, and a powdered mixture of NaCl and Ge is placed around the substrate. Excitation involving a plasma is again used to elicit trans￾port and engender GeNW growth. A Si wafer rather than a powdered sample has been used as the source material for simultaneous growth of SiOx nanowires and SnO2 nanobelts by Zhang et al. [27]. SnCl2 powder was placed upstream from a Si wafer and heated in Ar + O2 to 950 C for 2 h. The Sn catalyst particles, which also contain several percent Si and O, form at the base of the SiOx nanowires rather than the tips. Both the SiOx nanowires and the SnO2 nanobelts experience multiprong root growth. Zhang et al. [38] also used catalysis over Sn nanoparti￾cles to form branched CdS nanowires. They mixed CdS, SnO2 and graphite powders (in a 1:1:1 ratio) and heated them to 1200 C for 2 h under a constant flow of Ar. In this case the Sn particle at the end of the nanowire is signifi- cantly larger than the diameter of the wire that grows from it. In both of the systems investigated by Zhang et al., vapor phase transport is used not only to supply the growth material, but also to supply the material that forms the catalytic particle. K.W. Kolasinski / Current Opinion in Solid State and Materials Science 10 (2006) 182–191 183

K.w. Kolasinski/ Current Opinion in Solid State and Materials Science 10(2006)182-191 Carbothermal reduction, vapor phase transport of in contrast to, for instance, Au catalyzed growth of Si and growth material and reaction with trace amounts of oxygen SiGe nanowires. Also in this system, both the growth and have been utilized by Kuo and Huang in the growth of catalyst phases are transported via the vapor phase Taper Cdo nanowires that are 40-80 nm in diameter and 30- ing to larger diameters is observed in this system, most 50 um in length. Cdo and graphite powders are mixed in likely because the Sn nanoparticles are growing during a 4: 1 ratio. The silicon substrate is coated with a I nm growth. Au film and then placed 25 cm downstream from the reac Xu et al. [53] formed Zno nanowires with a hexagonal tant mixture. The reactants are held at 500C and the cross section. Interestingly no catalytic particle is found substrate at 400C. The Au particle at the end of the at the tip. Rather the catalytic action is provided by a nanowire has a diameter slightly smaller than the nano- ZnBil thin film(a combination of tetragonal ZnI, and wire. Most of the nanowires have a smooth surface but hexagonal Bil3) Mixtures of either Bil3 and Zn powder I an area of the reactor where there was probably a greater or Bi and Zn powders were heated in flowing Ar to 250- amount of oxygen available a jagged necklace structure 300C. The presence of I changes the growth direction of formed by a lateral growth of the rhombohedral Cdo the nanowires. The O can be supplied by an impurity in nanocrystals over the smooth nanowires. the Ar but no growth is observed in the absence of Bi when Pulsed laser deposition has been used by Morber et al. only Zn or Znl2 powders are used [39]to synthesize FeOy and Mg doped a-Fe2O3 nanorods, Carbothermal reduction can be used to produce ZnO nanowires and nanobelts. The ablation target was a pressed nanowires on a silicon substrate. Moreover, what Yang powder of magnetite(Fe3 O4), which was placed next to a et al. have shown is that growth can be switch away from quartz boat containing polycrystalline alumina wafer sub- Zno nanowires to Mn doped ZnSiO4 if MnCl2 4H2O is strates coated with a 2 nm Au film. Au particles are found added to the reaction mixture. The furnace system was at the ends of the nanowires, therefore it appears that the flushed with high-purity Ar gas to eliminate O2 and heated u film spontaneously breaks up to form the catalytic to 1 100C under a constant flow of Then. th nanoparticles quartz boat was placed in the centre of the furnace and An arc discharge has been used by Li et al. [34] to form held at 1100C under the same Ar flow. After reaction ubic 7-Al2O3 nanorods. The source material is a pressed for 50-60 min at 800-900C, the Si wafers, which are situ- powder mixture of Fe and Al, surprisingly in a 60: 40 ratio. ated downstream of the reaction mixture, were coated with Fe catalyst particles are found at the ends of the nanorods. a layer of nanowires. The majority of this material is com- The discharge is run in a mixture of 0.018 MPa Ar (99.9% posed of a willemite phase(a-Zn2SiO4 with rhombohedral purity) and 0.008 MPa H2(99.99% purity ). The growth structure) forming well-aligned nanorods with lengths of 2- conditions are highly non-equilibrium with the o being 4 um and diameters of 70-150 nm. The Zn catalyst is con- pplied at the trace level as an impurity in the process sumed so there is a severe taper at the end of the wires and gases. The diameters of the 7-Al2O3 nanorods are relatively no metal particle is found uniform, ranging from 20 to 30 nm Simultaneous growth of ZnO and LiF nanowires has Carbothermal reduction of a mixture of an Al com- been reported by Jiang et al. [50]. Zn acts as the cata olex with Fe powder has been used by Jung and Joo [33] ly hen LiF t Zno powders are heated in Ar to to create AIn whiskers. The mixture of the Al complex, 750-850C. The downstream deposition region has a Fe and graphite was calcined at 1200-1500C for 5 h in temperature in the range of 400-500C. Cubic-structured flowing N2. The whiskers often show modulations in their single-crystalline LiF nanowires grew along the (001) diameters along the lengths of the whisker and the shapes and(110) crystallographic directions with diameters of obtained depend on the growth temperature. This system 100-500 nm and lengths of tens of microns. The authors exhibits rather complex chemistry on account of the carbo- propose that there must be a barrier to the incorporation thermal reduction to produce Al, dissociative adsorption of of Lif into the lattice and the Zn particle acts to lower this N2, presumably on the Fe particles, and the formation of barrier and cause growth to be preferential at its ba Fe/Al/N alloy particles of the appropriate size. Very similar conditions can be used to grow Dendritic ZnO nanowires can be grown from Sn catalyst Ga2O3 and SnO, nanowires. Indeed, In,O3 and particles as shown by Gao et al. [59]. They heated ZnO and nanowires can be grown simultaneously without cross con- SnO, powders(1: I ratio) for I h under a pressure of 300- tamination or doping as demonstrated by Johnson et al 400 Torr of Ar carrier gas. The nanostructures grew on [45]. Fifty nanometer Au particles were deposited from the top of the inner alumina wall of the tube furnace in a solution onto the si substrate placed downstream from region located downstream M15 cm away from the source high-purity(6 N)metal reactant species (In, Ga, or Sn) material, which was located in the middle of the furnace placed separately in a quartz boat. The furnace was heated at 1300C, and the local growth temperature was in the to 800-1000C under flowing N2. No oxygen was supplied range of 700-800C. The Sn particle is significantly larger tly to the system other than as an impurity in the n than the Nw diameter. Note that in this system, as in sev- or that which desorbs from the surfaces inside the furnace. eral others, there is a high proportion of the catalyst start- Rectangular In]O3 rods are capped with a rectangular Au ing material compared to the growth material. This stands particle that is slightly smaller than the rod, whereas

Carbothermal reduction, vapor phase transport of growth material and reaction with trace amounts of oxygen have been utilized by Kuo and Huang in the growth of CdO nanowires that are 40–80 nm in diameter and 30– 50 lm in length. CdO and graphite powders are mixed in a 4:1 ratio. The silicon substrate is coated with a 1 nm Au film and then placed 25 cm downstream from the reac￾tant mixture. The reactants are held at 500 C and the substrate at 400 C. The Au particle at the end of the nanowire has a diameter slightly smaller than the nano￾wire. Most of the nanowires have a smooth surface but in an area of the reactor where there was probably a greater amount of oxygen available, a jagged necklace structure formed by a lateral growth of the rhombohedral CdO nanocrystals over the smooth nanowires. Pulsed laser deposition has been used by Morber et al. [39] to synthesize FexOy and Mg doped e-Fe2O3 nanorods, nanowires and nanobelts. The ablation target was a pressed powder of magnetite (Fe3O4), which was placed next to a quartz boat containing polycrystalline alumina wafer sub￾strates coated with a 2 nm Au film. Au particles are found at the ends of the nanowires, therefore it appears that the Au film spontaneously breaks up to form the catalytic nanoparticles. An arc discharge has been used by Li et al. [34] to form cubic c-Al2O3 nanorods. The source material is a pressed powder mixture of Fe and Al, surprisingly in a 60:40 ratio. Fe catalyst particles are found at the ends of the nanorods. The discharge is run in a mixture of 0.018 MPa Ar (99.9% purity) and 0.008 MPa H2 (99.99% purity). The growth conditions are highly non-equilibrium with the O being supplied at the trace level as an impurity in the process gases. The diameters of the c-Al2O3 nanorods are relatively uniform, ranging from 20 to 30 nm. Carbothermal reduction of a mixture of an Al3+ com￾plex with Fe powder has been used by Jung and Joo [33] to create AlN whiskers. The mixture of the Al complex, Fe and graphite was calcined at 1200–1500 C for 5 h in flowing N2. The whiskers often show modulations in their diameters along the lengths of the whisker and the shapes obtained depend on the growth temperature. This system exhibits rather complex chemistry on account of the carbo￾thermal reduction to produce Al, dissociative adsorption of N2, presumably on the Fe particles, and the formation of Fe/Al/N alloy particles of the appropriate size. Dendritic ZnO nanowires can be grown from Sn catalyst particles as shown by Gao et al. [59]. They heated ZnO and SnO2 powders (1:1 ratio) for 1 h under a pressure of 300– 400 Torr of Ar carrier gas. The nanostructures grew on the top of the inner alumina wall of the tube furnace in a region located downstream 15 cm away from the source material, which was located in the middle of the furnace at 1300 C, and the local growth temperature was in the range of 700–800 C. The Sn particle is significantly larger than the NW diameter. Note that in this system, as in sev￾eral others, there is a high proportion of the catalyst start￾ing material compared to the growth material. This stands in contrast to, for instance, Au catalyzed growth of Si and SiGe nanowires. Also in this system, both the growth and catalyst phases are transported via the vapor phase. Taper￾ing to larger diameters is observed in this system, most likely because the Sn nanoparticles are growing during growth. Xu et al. [53] formed ZnO nanowires with a hexagonal cross section. Interestingly no catalytic particle is found at the tip. Rather the catalytic action is provided by a ZnBiIx thin film (a combination of tetragonal ZnI2 and hexagonal BiI3). Mixtures of either BiI3 and Zn powder or Bi and Zn powders were heated in flowing Ar to 250– 300 C. The presence of I changes the growth direction of the nanowires. The O can be supplied by an impurity in the Ar but no growth is observed in the absence of Bi when only Zn or ZnI2 powders are used. Carbothermal reduction can be used to produce ZnO nanowires on a silicon substrate. Moreover, what Yang et al. have shown is that growth can be switch away from ZnO nanowires to Mn doped ZnSiO4 if MnCl2 Æ 4H2O is added to the reaction mixture. The furnace system was flushed with high-purity Ar gas to eliminate O2 and heated to 1100 C under a constant flow of Ar gas. Then, the quartz boat was placed in the centre of the furnace and held at 1100 C under the same Ar flow. After reaction for 50–60 min at 800–900 C, the Si wafers, which are situ￾ated downstream of the reaction mixture, were coated with a layer of nanowires. The majority of this material is com￾posed of a willemite phase (a-Zn2SiO4 with rhombohedral structure) forming well-aligned nanorods with lengths of 2– 4 lm and diameters of 70–150 nm. The Zn catalyst is con￾sumed so there is a severe taper at the end of the wires and no metal particle is found. Simultaneous growth of ZnO and LiF nanowires has been reported by Jiang et al. [50]. Zn acts as the cata￾lyst when LiF + ZnO powders are heated in Ar to 750–850 C. The downstream deposition region has a temperature in the range of 400–500 C. Cubic-structured single-crystalline LiF nanowires grew along the h001i and h110i crystallographic directions with diameters of 100–500 nm and lengths of tens of microns. The authors propose that there must be a barrier to the incorporation of LiF into the lattice and the Zn particle acts to lower this barrier and cause growth to be preferential at its base. Very similar conditions can be used to grow In2O3, Ga2O3 and SnO2 nanowires. Indeed, In2O3 and SnO2 nanowires can be grown simultaneously without cross con￾tamination or doping as demonstrated by Johnson et al. [45]. Fifty nanometer Au particles were deposited from solution onto the Si substrate placed downstream from high-purity (6 N) metal reactant species (In, Ga, or Sn) placed separately in a quartz boat. The furnace was heated to 800–1000 C under flowing N2. No oxygen was supplied directly to the system other than as an impurity in the N2 or that which desorbs from the surfaces inside the furnace. Rectangular In2O3 rods are capped with a rectangular Au particle that is slightly smaller than the rod, whereas 184 K.W. Kolasinski / Current Opinion in Solid State and Materials Science 10 (2006) 182–191

K w. Kolasinski/ Current Opinion in Solid State and Materials Science 10(2006)182-191 185 Ga2O3 nanowires appear to have a nearly spherical particle In, no detectable Se and is slightly larger than the nano- at their wire. When a 30 nm In film is used for catalysis, no In par Mohammad [60] has grown a variety of Ga and In con- ticle is found at the end of grown nanowires. At the le growth taining nanowires including GaN, InAs, InN, In GaAs, temperature either Au or In would be liquids In GaN, In gaasn using self-catalyzed growth involving Chen et al. [36]have grown CsN nanotubes from pyri either liquid Ga or liquid In droplets. Depending on the dine over a Fe-Co catalyst deposited on y-Al2O3. A mix- conditions either multipronged root growth or single- ture of N2 and pyridine passes over the catalyst while pronged float growth (defined in the next section) are heated to 550-950oC. In this case, the catalyst particles observed. A combination of carrier gas(N2 or H2) and have an unusual conical shape. Not only do the appear NH3, if required, flows over the metallic reagents placed to poke into the growing nanotube, they also are attached in Bn boats. One of the reactant metals may also be placed to the substrate rather than floating to the top of the nano- on the substrate. The distance between boats and the sub- tube even though only one tube grows from one particle. A strate, the flow rates and the temperatures of the boats and tapered particle that pokes into the core of a Cnt has also the substrate are all important variables that affect the been noted by Hofmann et al. [22] during plasma enhanced omposition and characteristics of the nanowires. Autocat- growth from C2H, or CH alytic growth shares much in common with and may actu ally be the mechanism behind what is called vapor-solid (VS)growth, in which no catalyst is intentionally added 3. Mechanisms to the system. Nonetheless, if one of the components is a low melting point metal such as Ga, In or Zn, a liquid cat Let us first start out with some generalities. The singular alyst particle may result during growth even if now none is attribute that usually leads to the conclusion that VLS intentionally added growth has occurred is that a metal particle of roughly Yun et al. [35] have heated a mixture of B+40 wt% the same diameter as the nanowire is found at the end of 9203 (which is a liquid at the growth temperature)in vac- the wire. This, of course, does not determine the phase uum over a 5-20 nm film of Au on Si to 600-950C for of the particle during growth and this controversial aspect 30 min B nanowires coated with an oxide layer are formed. will be dealt with further below Interestingly, a root growth mechanism, in which several Nanowires produced by catalytic growth are often nanowires emanate from a single catalytic particle, occurs found to have a uniform diameter. The wires are not at low temperature and float growth, in which one catalyst always round but might also exhibit other crystallograph particle sits atop each nanowire, occurs at high tempera- cally defined shapes, such as hexagonal ZnO or rectangular ture. The catalyst particle is a Au-B eutectic and dissolu- In,O3. Under some conditions, particularly for long tion of not only B into the Au particles but also the growth times, tapering of diameter to smaller (or less com- interaction of the eutectic with liquid B2O3 may be impor- monly larger) values is found. Often growth requires a bit tant in describing the growth dynamics of bundles of nano- of an induction period before uniform nanowires begin to bes and the switching between root growth and float grow. These considerations are represented schematically growth Laser ablation has been used by Jia et al. [55]to produce The initial period before uniform growth commences is ZnSe nanowires. Contrary to previous experiments involv- associated with any of a number of processes. In some g nanowire growth during laser ablation [61-63], the cases, the catalytic particles must be formed by vapor phase wafer target was not placed in a furnace. The evidence and/or surface diffusion transport or else their surfaces for self-catalytic VLS growth, as suggested by the authors, have to be cleansed of impurities(oxides or terminating thi- is not conclusive ols). The particles may be deposited directly, for instance, ZnTe nanowires with an average diameter of 30 nm from the evaporation of a colloidal solution with a well- and lengths >l um can be grown under the influence of a defined size. Alternatively, a thin film of metal can be evap- Au catalyst. Janik et al. [57] used a MBE source and a 3- orated directly onto a substrate and if the metal does not 20 A film of Au coated onto a GaAs substrate. The nano- wet the substrate, it will ball up into islands either immedi- wires, which are inclined about 55 to the(100)substrate ately as the result of Volmer-Weber growth [64] or else normal, have a zincblende crystal structure and their subsequently when the system is annealed, the onset of ost- growth axis is ( 111. The growth temperature is at slightly wald ripening [65] will lead to a distribution of island sizes above350°C The particles might also result from an evaporation and In2Se] nanowires have been grown either with an Au growth process that occurs during the initial stage, such talyst or with In acting as a self-catalyst by Sun et al. as when carbothermal reduction is used to generate a vol [49]. In2Se3 powder was held at 900-950C and placed atile metal that proceeds to condense elsewhere in the reac- upstream in flowing Ar from a Au or In coated silicon tor. Indeed, catalytic particles form readily under a variety wafer held at 650-700C. The nanowires are 40-80 nm in of conditions from any number of high vapor pressure, low diameter and up to 100 um in length. The spherical Au par- melting point metals. Instead of being the exception, ticle found at the tip of the nanowire contains less than 5% rather seems to be the case that the onus is on an investiga

Ga2O3 nanowires appear to have a nearly spherical particle at their tip. Mohammad [60] has grown a variety of Ga and In con￾taining nanowires including GaN, InAs, InN, InGaAs, InGaN, InGaAsN using self-catalyzed growth involving either liquid Ga or liquid In droplets. Depending on the conditions either multipronged root growth or single￾pronged float growth (defined in the next section) are observed. A combination of carrier gas (N2 or H2) and NH3, if required, flows over the metallic reagents placed in BN boats. One of the reactant metals may also be placed on the substrate. The distance between boats and the sub￾strate, the flow rates and the temperatures of the boats and the substrate are all important variables that affect the composition and characteristics of the nanowires. Autocat￾alytic growth shares much in common with and may actu￾ally be the mechanism behind what is called vapor–solid (VS) growth, in which no catalyst is intentionally added to the system. Nonetheless, if one of the components is a low melting point metal such as Ga, In or Zn, a liquid cat￾alyst particle may result during growth even if now none is intentionally added. Yun et al. [*35] have heated a mixture of B + 40 wt% B2O3 (which is a liquid at the growth temperature) in vac￾uum over a 5–20 nm film of Au on Si to 600–950 C for 30 min. B nanowires coated with an oxide layer are formed. Interestingly, a root growth mechanism, in which several nanowires emanate from a single catalytic particle, occurs at low temperature and float growth, in which one catalyst particle sits atop each nanowire, occurs at high tempera￾ture. The catalyst particle is a Au–B eutectic and dissolu￾tion of not only B into the Au particles but also the interaction of the eutectic with liquid B2O3 may be impor￾tant in describing the growth dynamics of bundles of nano￾tubes and the switching between root growth and float growth. Laser ablation has been used by Jia et al. [55] to produce ZnSe nanowires. Contrary to previous experiments involv￾ing nanowire growth during laser ablation [61–63], the wafer target was not placed in a furnace. The evidence for self-catalytic VLS growth, as suggested by the authors, is not conclusive. ZnTe nanowires with an average diameter of 30 nm and lengths >1 lm can be grown under the influence of a Au catalyst. Janik et al. [57] used a MBE source and a 3– 20 A˚ film of Au coated onto a GaAs substrate. The nano￾wires, which are inclined about 55 to the (1 0 0) substrate normal, have a zincblende crystal structure and their growth axis is h111i. The growth temperature is at slightly above 350 C. In2Se3 nanowires have been grown either with an Au catalyst or with In acting as a self-catalyst by Sun et al. [49]. In2Se3 powder was held at 900–950 C and placed upstream in flowing Ar from a Au or In coated silicon wafer held at 650–700 C. The nanowires are 40–80 nm in diameter and up to 100 lm in length. The spherical Au par￾ticle found at the tip of the nanowire contains less than 5% In, no detectable Se and is slightly larger than the nano￾wire. When a 30 nm In film is used for catalysis, no In par￾ticle is found at the end of grown nanowires. At the growth temperature either Au or In would be liquids. Chen et al. [36] have grown C5N nanotubes from pyri￾dine over a Fe–Co catalyst deposited on c-Al2O3. A mix￾ture of N2 and pyridine passes over the catalyst while heated to 550–950C. In this case, the catalyst particles have an unusual conical shape. Not only do the appear to poke into the growing nanotube, they also are attached to the substrate rather than floating to the top of the nano￾tube even though only one tube grows from one particle. A tapered particle that pokes into the core of a CNT has also been noted by Hofmann et al. [22] during plasma enhanced growth from C2H2 or CH4. 3. Mechanisms Let us first start out with some generalities. The singular attribute that usually leads to the conclusion that VLS growth has occurred is that a metal particle of roughly the same diameter as the nanowire is found at the end of the wire. This, of course, does not determine the phase of the particle during growth and this controversial aspect will be dealt with further below. Nanowires produced by catalytic growth are often found to have a uniform diameter. The wires are not always round but might also exhibit other crystallographi￾cally defined shapes, such as hexagonal ZnO or rectangular In2O3. Under some conditions, particularly for long growth times, tapering of diameter to smaller (or less com￾monly larger) values is found. Often growth requires a bit of an induction period before uniform nanowires begin to grow. These considerations are represented schematically in Fig. 1. The initial period before uniform growth commences is associated with any of a number of processes. In some cases, the catalytic particles must be formed by vapor phase and/or surface diffusion transport or else their surfaces have to be cleansed of impurities (oxides or terminating thi￾ols). The particles may be deposited directly, for instance, from the evaporation of a colloidal solution with a well￾defined size. Alternatively, a thin film of metal can be evap￾orated directly onto a substrate and if the metal does not wet the substrate, it will ball up into islands either immedi￾ately as the result of Volmer–Weber growth [64] or else subsequently when the system is annealed, the onset of Ost￾wald ripening [65] will lead to a distribution of island sizes. The particles might also result from an evaporation and growth process that occurs during the initial stage, such as when carbothermal reduction is used to generate a vol￾atile metal that proceeds to condense elsewhere in the reac￾tor. Indeed, catalytic particles form readily under a variety of conditions from any number of high vapor pressure, low melting point metals. Instead of being the exception, it rather seems to be the case that the onus is on an investiga￾K.W. Kolasinski / Current Opinion in Solid State and Materials Science 10 (2006) 182–191 185

K.w. Kolasinski/ Current Opinion in Solid State and Materials Science 10(2006)182-191 Fig. 2 illustrates several other aspects of growth that Initiation Steady State Termination must be considered. Will the nanowire be produced from deposition I transport to drdt<0 root growth in which the catalyst particle is found at the nucleation base of the nanowire, Fig. 2a, or by float growth, in which the particle is located at the tip of the nanowire? will multi drdt>0 of sidewalls ple prong growth ensue, Fig. 2c, in which more than one dndt=o nanowire emanates from each particle or will single-prong growth occur, Fig. 2d? A rather widely reported misunderstanding [1, 66, 67] is that VLS growth occurs because the sticking coeffic Fig. 1. General considerations on the different regimes that occur during on a liquid is unity and must be higher than the sticking talytic growth of es and nanotubes coefficient on the solid. This has also been mistakenly repeated in other fields of structure formation involving growth [68]. For growth of SiNWs from silanes, which tor to show that catalytic particle are not involved in the have a much higher dissociative sticking coefficient on growth process the particle than on the substrate or sidewalls, the reason Once the catalytic particles are formed or deposited, for the greater rate of dissociation is because of the cata- they may still need to be primed for the growth of nano- lytic action of the metal in the particle not the fact that it wires. For instance, the pure metal catalytic particle might is liquid. There is no general evidence for the assertion not be that active for nanowire formation. Instead, an that the sticking coefficient must be larger on a liquid than admixture of the growth compound and the metal might a metal. Second, the assertion does not even apply gener be required to form an (unstable or stable)alloy, a true ally to VLS growth. The success of mBe in semiconduc- eutectic or some other solid/liquid solution. In this case, tor processing relies largely on the fact that the sticking saturation of the catalytic particle with the growth material coefficient of numerous evaporated semiconductor materi or the formation of the proper composition may lead to an als is virtually unity on a solid substrate regardless of its induction period before growth. Note that the incorpora- composition. VLS growth in a MBE configuration, in tion of a significant amount of growth material into the which the growth material is supplied by evaporation catalytic particle is expected to change the volume and, from a crucible unto the substrate, has been demonstrated therefore, the diameter of the particle from its initial value. for nanowires composed of, e.g. Si, SiGe and Ill-V com- Hence Ostwald ripening and incorporation of growth pounds. a sticking coefficient difference alone cannot material can both conspire to change the size of the cata- account for the formation of nanowires. Other factors lytic particles must also be considered that allow the nanowire/catalytic The growth of nanowires with a uniform radius is asso- particle interface to act as a sink for the incorporation of ciated with a steady state growth in which material is trans- new material into the nanowire at a greater rate than the ported to the particle/nanowire interface. If the nanowire growth of the sidewalls or the thickness of the substrate radius is related to the particle radius and the particle in between particle sites. For instance, the catalytic parti radius is constant, a natural explanation for a uniform cle can lower the barrier that is present for the incorpora nanowire diameter is that the particles have reached a tion of new material at the growth interface as compared steady state and their diameter is not evolving in this to the nucleation of an island on a sidewall or the region. If the incorporation of material is directed solely substrate. y the particle and incorporation directly into the sidewalls Fig. 2a and b also illustrate several dynamical process is suppressed, a constant particle diameter leads to a con- that can affect growth. Adsorption occurs from the fluid stant nanowire diameter. The particle might be able to (whether gaseous, liquid or supercritical) phase. Adsorp- affect the size of the nanowire either by direct matching tion might be molecular or dissociative and may either of the size of the nanowire to the size of the particle or else occur (vi) on the nanowire(viii) on the particle, (ix)on by some mechanism involving the curvature of the particle the substrate. A natural way for the catalytic particle to in which strain and lattice matching play a role. direct material to the growth interface is if the sticking Finally, a tapering to smaller diameters and cessation of coefficient(probability of adsorption) is higher on the par growth will occur if either the particle enters a phase in ticle and vanishingly small elsewhere. Diffusion of adatoms which it is consumed, if the growth material is no longer will occur (i)across the substrate(if the sticking probability supplied to the system, or if the temperature is reduced is not negligible), (ii) across the particle and (iii) along the below a critical value. The temperature will play a role in sidewalls. Diffusion across the substrate and along the side- umerous processes, e.g. dissociative adsorption, surface walls must be rapid and cannot lead to nucleation events diffusion, bulk diffusion through the particle, in determin- Nucleation of the nanowire anywhere other than on the ng the composition of the particle by affecting solubilities particle must be suppressed so that growth only occurs at and thermodynamic stability of certain phases as well as the particle/ nanowire interface and so that sidewalls do diffusion of metal atoms away from the particle not grow independently of the axial growth. There may

tor to show that catalytic particle are not involved in the growth process. Once the catalytic particles are formed or deposited, they may still need to be primed for the growth of nano￾wires. For instance, the pure metal catalytic particle might not be that active for nanowire formation. Instead, an admixture of the growth compound and the metal might be required to form an (unstable or stable) alloy, a true eutectic or some other solid/liquid solution. In this case, saturation of the catalytic particle with the growth material or the formation of the proper composition may lead to an induction period before growth. Note that the incorpora￾tion of a significant amount of growth material into the catalytic particle is expected to change the volume and, therefore, the diameter of the particle from its initial value. Hence Ostwald ripening and incorporation of growth material can both conspire to change the size of the cata￾lytic particles. The growth of nanowires with a uniform radius is asso￾ciated with a steady state growth in which material is trans￾ported to the particle/nanowire interface. If the nanowire radius is related to the particle radius and the particle radius is constant, a natural explanation for a uniform nanowire diameter is that the particles have reached a steady state and their diameter is not evolving in this region. If the incorporation of material is directed solely by the particle and incorporation directly into the sidewalls is suppressed, a constant particle diameter leads to a con￾stant nanowire diameter. The particle might be able to affect the size of the nanowire either by direct matching of the size of the nanowire to the size of the particle or else by some mechanism involving the curvature of the particle in which strain and lattice matching play a role. Finally, a tapering to smaller diameters and cessation of growth will occur if either the particle enters a phase in which it is consumed, if the growth material is no longer supplied to the system, or if the temperature is reduced below a critical value. The temperature will play a role in numerous processes, e.g. dissociative adsorption, surface diffusion, bulk diffusion through the particle, in determin￾ing the composition of the particle by affecting solubilities and thermodynamic stability of certain phases as well as diffusion of metal atoms away from the particle. Fig. 2 illustrates several other aspects of growth that must be considered. Will the nanowire be produced from root growth in which the catalyst particle is found at the base of the nanowire, Fig. 2a, or by float growth, in which the particle is located at the tip of the nanowire? Will multi￾ple prong growth ensue, Fig. 2c, in which more than one nanowire emanates from each particle or will single-prong growth occur, Fig. 2d? A rather widely reported misunderstanding [1,66,67] is that VLS growth occurs because the sticking coefficient on a liquid is unity and must be higher than the sticking coefficient on the solid. This has also been mistakenly repeated in other fields of structure formation involving growth [68]. For growth of SiNWs from silanes, which have a much higher dissociative sticking coefficient on the particle than on the substrate or sidewalls, the reason for the greater rate of dissociation is because of the cata￾lytic action of the metal in the particle not the fact that it is liquid. There is no general evidence for the assertion that the sticking coefficient must be larger on a liquid than a metal. Second, the assertion does not even apply gener￾ally to VLS growth. The success of MBE in semiconduc￾tor processing relies largely on the fact that the sticking coefficient of numerous evaporated semiconductor materi￾als is virtually unity on a solid substrate regardless of its composition. VLS growth in a MBE configuration, in which the growth material is supplied by evaporation from a crucible unto the substrate, has been demonstrated for nanowires composed of, e.g. Si, SiGe and III–V com￾pounds. A sticking coefficient difference alone cannot account for the formation of nanowires. Other factors must also be considered that allow the nanowire/catalytic particle interface to act as a sink for the incorporation of new material into the nanowire at a greater rate than the growth of the sidewalls or the thickness of the substrate in between particle sites. For instance, the catalytic parti￾cle can lower the barrier that is present for the incorpora￾tion of new material at the growth interface as compared to the nucleation of an island on a sidewall or the substrate. Fig. 2a and b also illustrate several dynamical process that can affect growth. Adsorption occurs from the fluid (whether gaseous, liquid or supercritical) phase. Adsorp￾tion might be molecular or dissociative and may either occur (vii) on the nanowire (viii) on the particle, (ix) on the substrate. A natural way for the catalytic particle to direct material to the growth interface is if the sticking coefficient (probability of adsorption) is higher on the par￾ticle and vanishingly small elsewhere. Diffusion of adatoms will occur (i) across the substrate (if the sticking probability is not negligible), (ii) across the particle and (iii) along the sidewalls. Diffusion across the substrate and along the side￾walls must be rapid and cannot lead to nucleation events. Nucleation of the nanowire anywhere other than on the particle must be suppressed so that growth only occurs at the particle/nanowire interface and so that sidewalls do not grow independently of the axial growth. There may time Initiation deposition nucleation saturation dr/dt > 0 Steady State transport to growth interface passivation of sidewalls dr/dt = 0 Termination dr/dt < 0 Fig. 1. General considerations on the different regimes that occur during catalytic growth of nanowires and nanotubes. 186 K.W. Kolasinski / Current Opinion in Solid State and Materials Science 10 (2006) 182–191