Jan 06, 2016
Coordination Chemistry Reviews 257 (2013) 3297 3322
Contents lists available at ScienceDirect
Coordination Chemistry Reviews
jo ur nal ho me page: www.elsev ier .com/ locate /ccr
Review
Precursors as enablers of ALD technology: Contributions fromUniversity of Helsinki
Timo Hatanp , Mikko Ritala, Markku LeskelLaboratory of Inorganic Chemistry, Department of Chemistry, University of Helsinki, P.O.Box 55, 00014 University of Helsinki, Finland
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32982. Alkaline earth metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3298
2.1. -Diketonates of alkaline earth metals as precursors for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32992.2. Cyclopentadienyl compounds of alkaline earth metals as precursors for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3300
3. Group 4 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33053.1. Group 4 metal halides as precursors for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33053.2. Group 4 alkoxides as precursors for ALD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33063.3. Group 4 alkylamides as precursors for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33073.4. Group 4 Cyclopentadienyl compounds for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3307
3.5. 4. Group
4.1. 5. Bismu
5.1. 5.2. 5.3. 5.4.
6. Pt gro7. Silver8. Summ
Refer
Abbreviatiomentary metadiglyme, bis(2thickness; FESGST, Germaniu2-methyl-2-prMETHD, methmemory; PCRsition; SBT, stsolid oxide fuetris(3,5-diethy
CorresponE-mail add
0010-8545/$ http://dx.doi.o3.4.1. Bis-cyclopentadienyl compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33073.4.2. Mono-cyclopentadienyl compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3307Other group 4 precursors for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3308
15 and 16 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3309Alkylsilyl compounds of group 15 and 16 elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3309th . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3311Bismuth chloride as precursor for ALD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3312Organometallic compounds of bismuth as precursors for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3312Tris(Bis(Trimethylsilyl)amido)Bismuth(III) as a precursor for ALD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3312Bismuth compounds with oxygen based ligands as precursors for ALD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33135.4.1. -Diketonates of bismuth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33135.4.2. Bismuth alkoxides . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33135.4.3. Bismuth carboxylates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3315
up metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3315 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3315ary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3319
ences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3319
ns: acac, acetylacetonate; ALD, atomic layer deposition; CET, capacitance equivalent thickness; CHT, cycloheptatrienyl; CHD, cyclohexadiene; CMOS, comple-l oxide semiconductor; CN, coordination number; Cp, cyclopentadienide; COD, cyclooctadiene; CVD, chemical vapour deposition; dien, diethylenetriamine;-methoxyethyl) ether; DMAE, dimethylaminoethoxide; DRAM, Dynamic random access memory; EDS, energy dispersive spectrometry; EOT, Equivalent oxideEM, eld emission scanning electron microscope; fod, 1,1,1,2,2,3,3-heptauro-7,7-dimethyl-4,6-octanedione; FRAM, ferroelectric random access memory;m antimony telluride; hfac, hexauoroacetylacetonate; HRTEM, high resolution transmission electron microscopy; me, methoxyethoxy; mmp, 1-methoxy-opanolate; MOCVD, metal organic chemical vapour deposition; MS, mass spectroscopy; MTHD, methoxy-2,2,6,6-tetramethylheptane-3,5-heptanedione, me;oxyethoxy-2,2,6,6-tetramethylheptane-3,5-heptanedione; MOSFET, metal oxide semiconductor eld effect transistor; NVRAM, non-volatile random accessAM, Phase change random access memory; PEALD, plasma enhanced atomic layer deposition; Piv, pivalate; REALD, radical enhanced atomic layer depo-rontium bismuth tantalate; SDTA, single differential thermal analysis; SEM, scanning electron microscope; SIMS, Secondary ion mass spectrometry; SOFC,l cell; TG, thermo gravimetry; TGA, thermo gravimetric analysis; THF, tetrahydrofurane; tmhd, tetramethylheptanedione; Tp, tris(pyrazolyl)borate; TpEt2,lpyrazolyl)borate; trien, triethylenetetraamine; triglyme, triethylene glycol dimethyl ether.ding author. Tel.: +358 919150222, fax: +358 919150198.ress: timo.hatanpaa@helsinki. (T. Hatanp).
see front matter 2013 Elsevier B.V. All rights reserved.rg/10.1016/j.ccr.2013.07.002
3298 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
a r t i c l e i n f o
Article history:Received 4 March 2013Received in reAccepted 3 JulAvailable onlin
Keywords:Atomic layer dALD precursorHigh-K materiPhase change Alkaline earthGroup 4 metalBismuthIridiumSilver
a b s t r a c t
The review focuses on ALD precursors of selected elements such as alkaline earth (Mg, Ca, Sr, Ba), group4 metals, bismuth, silver, iridium, selenium, tellurium and antimony. These elements are needed in dif-
1. Introdu
The succALD (conforcan be enjofor the mateber of requhigh reactivlenges. Volasolids are uconvenienc
The ALDscientic cuemphasis hbetter knowFor examplALD of HfOied TiCl4 [2HfO2 would15 years latlurides and[3], was motor companwith. On theprecursor rplay materiSoon after tsuccessfullythat had gavery high-k
ALD meon their ligamides, alkand even elhave been tands, particdonor funcAnother opsphere of thcomplexes for further t
In this pALD precurvised form 3 July 2013y 2013e 18 July 2013
epositionsalsmemory
metalss
ferent high tech applications but are challenging for ALD. Their precursor design needs careful balancingbetween volatility, thermal stability and reactivitythe key properties of ALD precursors. The extensivestudies showed that cyclopentadienyl based precursors of alkaline earth metals are versatile ALD pre-cursors which react with both water and ozone forming oxide. Group 4 ALD chemistry has been studiedvery widely and many good precursors have been found for the oxide ALD. From a bunch of differentcompound types studied the most promising ALD precursor for bismuth is Bi(OCMe2iPr)3 which showsstable ALD process with water at 150250 C. The success in depositing noble metal lms by ALD canbe attributed more to the reactant rather than the metal precursor. Ru, Pt, Ir, Rh and Os lms can bedeposited from various organometallic and metal organic precursors using O2 as the other precursor.Typically temperatures above 225 C are needed. Using O3 as a reactant lms can be deposited at lowertemperatures. Noble metal oxides are obtained below approx. 200 C and metallic lms above that. Bysupplying both O3 and H2 as consecutive pulses, noble metal lms can be deposited well below 200 C.For silver phosphine stabilized carboxylato and -diketonato complexes are thermally stable enoughenabling hydrogen plasma enhanced ALD of silver metal lms. Alkylsilyl compounds of selenium andtellurium are versatile ALD precursors for metal selenide and telluride lms when combined with metalchloride precursors. The use of alkylsilyl compounds is not limited to group 16 elements but can alsobeen used for group 15.
2013 Elsevier B.V. All rights reserved.
ction
ess of ALD is built on chemistry. The great benets ofmality, uniformity, atomic level thickness control, etc.)yed only when proper precursors have been identiedrial of an interest. The precursors need to meet a num-irements (Table 1) among which the combination ofity and good thermal stability often sets major chal-tility is important too, and though moderately volatile
sed even industrially, liquids and gases are preferred fore.
chemistry has been approached from two directions:riosity and from needs of industry. Over the years theas been shifted toward the latter as the ALD has becomen and more widely used thin lm deposition technique.e, when we in 1993 studied HfCl4 as a precursor for2 [1] this was done for comparison to the earlier stud-], and neither we nor anyone else had a clue that ALD
become a high-k gate oxide in the leading MOSFETser. By contrast, our recent breakthrough in ALD of tel-
selenides with their alkylsilyl compounds as precursorstivated by a request addressed to us from semiconduc-ies via an ALD tool manufacturer that we collaborate
other hand, also in 1990s we did application motivatedesearch but that focused on electroluminescent dis-als rather than the current mainstream semiconductors.he results and knowledge gained in that research were
beneted in developing an ALD process for SrTiO3 [4]ined increasing interest in IC applications because of its
value.tal precursors have traditionally been grouped basedands (Fig. 1): halides, -diketonates, alkoxides, alkylyls, cyclopentadienyls, amidinates, guanidinates, etc.ements [5]. Inside these groups the precursor propertiesuned to meet the ALD requirements by tailoring the lig-ularly by changing the size of the ligands and by addingtionalities to the hydrocarbon groups of the ligands.tion has been to add adduct ligand to the coordinatione metals. A newer trend has involved heteroleptic metalwhere two or more different ligands are used togetheruning of the precursor properties.aper we review some of our own contributions to thesor chemistry over the past 20 years. The aim of the
review is not to present a balanced view of the eld but to focuson studies and discoveries done at the University of Helsinki. Thusreferences to other studies are kept in minimum. A number ofdifferent kinds of precursor groups will be covered. Each chapterbegins with a brief summary of applications of ALD lms containingthe given elements, thereby demonstrating how in ALD precur-sor chemistry basic and applied research are connected seamlessly.Properties (structure, volatility, thermal stability, reactivity) of pre-cursors studied and main results from the lm growth experimentsare reviewed. About the ALD reaction mechanisms only some illus-trative examples will be given since in situ reaction mechanismstudies on ALD processes were reviewed in detail very recently byKnapas and Ritala [6].
2. Alkaline earth metals
Alkaline-earth metals are constituents of many technologicallyimportant materials. SrTiO3 and Ba1xSrxTiO3 are high permit-tivity dielectric materials for future DRAMs [7]. BaTiO3 is notonly a dielectric but also piezoelectric and ferroelectric materialwhich may be used in e.g. ceramic multilayer capacitors and ther-mistor elements [8,9]. CaS, SrS and BaS are host materials forluminescent materials [10,11]. SrBi2Ta2O9 is a ferroelectric mate-rial which may be used in FRAMs [12]. Strontium and bariumare also constituents of the well-known superconducting cupratesLa2xSrxCuO2, YBa2Cu3O7x and Bi2Sr2CaCu2O8+x [13]. Fluorides ofmagnesium and calcium are materials suited for optical coatings[14].
Especially the larger alkaline earth metals are challenging fromthe CVD/ALD precursor chemistry point of view: their compoundshave a tendency to form oligomeric species which have lowvolatility. Thus sufciently volatile, thermally stable and reactivecompounds of alkaline earth metals are not that common.
Before the interest in ALD, a large variety of -diketonates andalso many other compounds had been used in CVD as precur-sors for alkaline earth metals. In addition a variety of compoundshad been suggested as possible precursors due to their volatil-ity [15]. These compounds include silylamides and phosphides,poly(pyrazaolyl)borates, alkoxides, -ketoiminates, pyrazolates,pyrrolates and cyclopentadienyl compounds with and withoutneutral ancillary Lewis base ligands. Precursors used for ALD ofalkaline-earth metal containing lms are quite few and they are col-lected into Table 2. Basically the precursors have been restricted to
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3299
Fig. 1. Gen n the
-diketonaextensively
2.1. -Dike
The rstsors for ALDvolatile comproperties hcame in coCVD of oxiCa, Sr and Bferent -dithe thd-lig-diketonasynthesis avantages likreactivity. Iproblems wof the -dik[44]. This bthe thd-comdifferent O when differair have beeincorrectly.
The largand only bicoordinativunsaturatioeral classes of metal precursors. Backbones of the ligands are shown and in additiotes and cyclopentadienyl compounds which have been studied at the University of Helsinki too.
tonates of alkaline earth metals as precursors for ALD
alkaline-earth metal compounds introduced as precur- were -diketonates. These are the most well-knownpounds of alkaline earth metals. Their volatilizationave been known for decades [116] but the real boomnnection to high-temperature superconductors. Thede superconductors required volatile compounds fora as well as for rare earths [117]. The number of dif-
ketonato ligands is high but by far the most used isand (thd = 2,2,6,6-tetramethyl-3,5-heptanedione). Thetes have several advantages such as sufciently easynd reasonable thermal stability but also several disad-e limited volatility because of oligomerization and lown thin lm growth experiments there have also beenhich have been associated with some kind of instabilityetonate precursors at their evaporation temperaturesehavior may have been partly related to the quality ofpounds as the actual precursors used may have beenand OH groups containing oligomeric species formedent synthesis methods without exclusion of water andn used, or the compounds have been handled or stored
e size and low charge of alkaline earth metal ionsdentate character of the -diketonato ligands causese unsaturation in M(-diketonato)2 complexes. Then is decreased by oligomerization and oligomers with
four metal if additionaoligomers isuch as pol
Table 1Requirements
Requiremen
Volatility
Thermal sta
Reactivity
No etching osubstrate
Unreactive abyproduct
No dissolutisubstrate
Purity
Inexpensive
Easy to handNon-toxic an
environm ligands may contain donorfuntionalities in the variable Rx parts.centers have been reported for Ba(thd)2 [118] andl aqua, OH or O ligands are coordinated the size ofncreases further [119]. By using neutral adduct ligandsyethers and polyamines it is possible to decrease the
for ALD precursors.
t Comment
Volatility is needed for efcienttransportation. For stable ux liquidsand gases are preferable.
bility Decomposition of precursors is notallowed. Instability would destroy theself-limiting lm growth mechanismof ALD.Aggressive and complete reactionswith co-precursors to ensure fastcompletion of surface reactions andthus short cycle times, and for highlm purity.
f the lm ormaterial
Competing reactions may prevent lmgrowth.
nd volatiles
To avoid corrosion. Readsorption of thebyproducts may slow the lm growth.
on into the lm or Would destroy the self-limiting growthbehavior.Sufciently pure precursors to meetthe requirements specic to eachapplication.
Easy synthesis with good yield, simplecompounds with cheap ligands.
le Stable at ambient air, if possible.d
entally friendlyIf possible.
3300 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
molecule size to monomolecular one [120]. The volatility of thealkaline earth thd complexes increases in series Ba < Sr < Ca < Mgand this is in line with the oligomerization and ion size. Sr(thd)2and Ca(thd)2 form trimers when crystallized in dry organicsolvents anor polyethemetric studthd-complebeing the dthe smalles[21] whichtral ligandsof adductedand Ba sugtion or in t-diketonaintact moletion: diamibelow a cer
Alkalinedeposit theas the co-pSrS and Bahave extenluminescendeposited uTaF5.[1820and toxic He.g. with the(1.6 A/cycle
In contrwith the peratures bof -diketooxide lms.cess has bethough witat 250 C [with Ti(OiPgrown [50,5
Becausetive oxygensecond choicomplexes carbonate deposition involves a rline earth ma constant A third comSrTiO3 havegen plasmano carbonatwide (150[Mg2(thd)4at relatively
2.2. Cyclopprecursors f
In Cp combon atoms iwhile still kincrease thetwo rings tofor most of
number of cyclopentadienyl compounds of strontium and bariumhave been synthesized and characterized quite well [124,125]. TheCp compounds of alkaline earth metals (Mg, Ca, Sr, Ba) are veryimportant in ALD because they enable the growth of oxide lms at
able whilopem thnedtheseere mpo26,1yl coned128]ad beMos
cyclely ifcue int
lmuitabiketer tur wum adiffe
e andetal carge he m
[65,.
st of s a t (x be rehe sfore t eff
d comMR ahere
haves Batrast
resu subls 1, 5
[65,lopeom were re st
whee of
areundse dog poed inper
s iPrC5Hd also need an adduct ligand such as polyaminesrs to be monomolecular [41,121]. The mass spectro-ies show that also in gas phase the alkaline earthxes exist in oligomeric forms with the dimeric speciesominant in the mass spectra [116,122]. Magnesium ast member in the series forms a dimeric thd-complex
again can be turned to monomeric by ancillary neu- [123]. The mass spectrometric and sublimation studies
thd-complexes of the larger alkaline earth metals Srgest that the neutral ligand dissociates upon sublima-he gas phasein this sense adducts with uorinatedto ligands are better because many of them evaporate ascules. The thd-complexes of Mg and Ca make an excep-ne and polyamine adducts stay intact in the gas phasetain dissociation temperature [41,123].
earth metal thd-complexes have been used in ALD to corresponding suldes and uorides using H2S and HFrecursors, respectively [10,35,3741,4447,110]. CaS,S lms doped with different rare earth metal ionssively been studied as phosphor materials for electro-t at panel displays [10,11]. Fluorides have also beensing reactions between the thd-complexes and TiF4 and,36] This method avoids the usage of highly corrosiveF and also much higher growth rates were obtained,
Ca(thd)2 + TiF4 process a four times higher growth rate) compared to the Ca(thd)2 + HF process was achieved.ast to H2S and HF, water does not react effectively-diketonates of larger alkaline earth metals at tem-elow decomposition [44]. This insufcient reactivitynates is the main limitation for their usage in ALD of
However, growth of an oxide by the Cathd2 + H2O pro-en reported [32] and in two studies SrO also grew,h a low rate of 0.1 A/cycle, from Sr(thd)2 and H2O50,55]. Also when Sr(thd)2 + H2O cycles were mixedr)4 + H2O or Ti(OiPr)2(thd)2 cycles SrTiO3 lms could be5].
water has low reactivity with -diketonates more reac- sources have been introduced. Ozone is the commonce for oxygen precursor in ALD of oxide lms. With thd-of the larger alkaline earth metals ozone however formslms [48]. It is possible to get rid of the carbonate by postannealing but it makes an additional process step andisk of defect formation. In contrast with the larger alka-etals, Mg(thd)2 forms quite pure MgO with O3 [22] andgrowth rate of 0.27 A/cycle is observed at 225250 C.mon oxygen source in ALD is oxygen plasma. SrO and
been deposited using Sr(METHD)2, Ti(OiPr)4 and oxy- as precursors and in contrast to the ozone processese was formed. The ALD window was reported also very275 C) [63]. In one study MgO lm was grown from] and H2O2 but with a low rate (0.100.15 A/cycle) and
high temperatures of 325425 C [21].
entadienyl compounds of alkaline earth metals asor ALD
pounds the metal is usually coordinated to all ve car-n the Cp ring (Fig. 2) and this gives shielding and stabilityeeping the reactivity. The Cp rings can be substituted to
bulkiness of the ligand and it is also possible to bridgegether to an ansa-compound. Cp compounds are known
the metals and many of them are volatile. Also a large
reasonsource
Cycing fromentioto use there wenyl coCVD [1tadienmentio1990 [sium h[129]. is thatextremquite dsors arof thinmore sent -dand low
In ostrontitested volatilthe mlized lcover t(Fig. 2)studied
Moused aamouncould using tlost berelevanisolateysis, NIons wwouldadductin conwhichbeforepoundand 19bis-cyc(12), frtures wbases achainsdistancactionscompo(19) thMeltinpreparing temligandBa(iPr3temperatures and growth rates using water as oxygench is very difcult with the alternative thd-precursors.ntadienyl compounds have been widely studied start-e mid-eighties [124] with one of the main motivations, in addition to general curiosity, being the possibility
compounds as CVD precursors. However, before 1999only two actual articles reporting usage of cyclopentadi-unds of heavier alkaline earth metals as precursors for27]. In addition to these articles the usage of cyclopen-mpounds of strontium and barium in CVD has been
in several mostly Japanese patents starting as early as. In contrast, cyclopentadienyl compounds of magne-en used in CVD more widely starting from at least 1989
t obvious reason for the quite low number of reportsopentadienyl compounds of alkaline earth metals aresensitive to air (oxygen) and moisture which makes itlt to use these compounds in CVD where all the precur-roduced into the reactor simultaneously. Also for CVDs containing alkaline earth metals there are many farle compounds than the cyclopentadienyls, like differ-onates and -ketoiminates which have lower reactivityhermal stability, often desired in conventional CVD.ork we synthesized cyclopentadienyl compounds ofnd barium with different cyclopentadienyl ligands andrent well known strategies for making them more
thermally stable i.e. lled the coordination sphere ofenters with multidentate Lewis base molecules, uti-and sterically demanding cyclopentadienyl ligands toetal centers, and utilized donor functionalized ligands
113,130,131]. Table 3 lists the compounds prepared and
the as-prepared compounds contained coordinated THFsolvent in the synthesis. After crystallization a xed= 1 or 2) of THF was attached to each molecule. THFmoved by sublimation, heating under vacuum or byolvent reux method. Thus the loosely bonded THF isor during the evaporation of the compounds and has noect on the thermal properties of the compounds. Thepounds were pure as characterized by elemental anal-
nd mass spectroscopy. All MS showed [ML2]+ species. the neutral solvent or added Lewis base molecule
been coordinated to barium were not seen. For the(Me5C5)2A (A = dien, trien, diglyme, triglyme) this is
with the results of vacuum sublimation experimentslted in sublimates with composition exactly the same asimation. Crystal structures were determined for com-, 7, 10, 11, 13 (Fig. 2), 14, 16, 17 (Fig. 2), 18 (Fig. 2)
113,130132]. For the others, except the propyl bridgedntadienyl compounds 8 and 9, and Ba(Me5C5)2(trien)hich single crystals could not be grown, crystal struc-
already known. All compounds with coordinated Lewisrictly monomeric. Ba(Me5C5)2 is known to form loosere some methyl groups of neighbouring molecules are in
interaction from the barium atom [133]. Similar inter- found in the structure of Ba(tBu3C5H2)2 [130]. In the
Ba(Me2NC2H4C5Me4)2 (18) and Ba(EtOC2H4C5Me4)2nor atoms O and N are coordinated to the metal center.int temperatures of the cyclopentadienyl compounds
general are quite high [113,125]. The lowest melt-atures are observed for the compounds with bulkier3C5H2 and tBu3C5H2-, the lowest being 115 C for2)2(THF)2 [132,134]. Most likely the iPr groups in the
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3301
Table 2Group II precursors used in ALD.
Metal precursor Other precursor Material deposited Reference
Mg Te MgTe [16,17]
Mg(thd)2 TiF4 TaF5 MgF2 [18,19]H2O2 MgF2 [20]O3 MgO [21]
MgO [22]
Mg(C5H5)2 H2O MgO [2330]AlMe3, O3 MgAl2O4 [30]
Mg(EtC5H4)2 H2O MgO [31]Ca(thd)2 H2O CaO [32]
O3 CaO [33,34]O3, (MeO)3PO, H2O Ca10(PO4)6(OH)2 [34]HF CaF2 [35]TiF4, TaF5 [18,36]H2S CaS [10,3741]
Ca(iPr3C5H2)2 H2O CaO [42]HfCl4, H2O CaO-HfO2
Ca(Tp)2 H2O CaB2O4 [43]
Sr(thd)2 HF SrF2 [35]H2S SrS [10,38,4447]O3 SrO(SrCO3) [48]O3, Ti(OiPr)4, H2O SrTiO3 [48,49]H2O, Ti(OiPr)4 SrTiO3 [50,51]Ti(OiPr)4, O2 plasma SrTiO3 [52]H2O SrO [5254]H2O, Ti(OiPr)2(thd)2 SrTiO3 [55,58]La(thd)3, Fe(thd)3, O3 La1xSrxFeO3 [59]Ti(OiPr)4, H2O plasma SrTiO3 [60,61]H2S, Se SrS1xSex [62]
Sr(METHD)2 O2 plasma SrO, SrTiO3 [63]Sr(MTHD)2 O3 SrO [64]Sr(Me5C5)2 H2S SrS [65]
Sr(nPrMe4C5)2 H2O SrO [66]La(Me4C5H)3, Mn(Me4C5H)2, H2O La2O3SrOMnO [67]
Sr(iPr3C5H2)2 H2O SrO [6872]O2 SrO [73]O2, RuO4 SrRuO3 [73]O3 SrO [71,74]O2 plasma SrO [72]Ti(OiPr)4, H2O SrTiO3 [4,70,7577]H2O, Ti(OiPr)2(thd)2, O3 SrTiO3 [78]Ti(NEtMe)4, O3 SrTiO3 [79,80](Me5C5)Ti(OMe)3, O3 SrTiO3 [74,81](MeC5H4)2Hf(OMe)Me, O2 plasma SrHfO3 [72]H2S SrS [65,82]
Sr(iPr3C5H2)2(DME) (Me5C5)Ti(OMe)3, O2 plasma SrTiO3 [83,85]H2O SrO [71]O3 SrO [71](MeCp)Ti(OMe)3, O3 SrTiO3 [71](Me5C5)Ti(OMe)3, O3 SrTiO3 [71]
Sr(tBu3C5H2)2 H2O SrO [86,87]Ti(OMe)4, H2O SrTiO3 [86,99]Ti(OMe)4, O3 SrTiO3 [100,101]Ti(OiPr)4, O2 plasma SrTiO3 [102](Me5C5)Ti(OMe)3, O3 SrTiO3 [87](MeC5H4)2Hf(OMe)Me, H2O SrO-HfO2 [72]
Sr(Tp)2 H2O SrB2O4 [103]Sr(tButAm2imidazolato)2 O3 SrO [104]
Sr[Ta(OEt)5(me)]2 O2 plasma SrTa2O6 [105]H2O [106,107]
3302 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
Table 2 (Continued )
Metal precursor Other precursor Material deposited Reference
Sr[Ta(OEt)5(dmae)]2 H2O SrTa2O6 [106,107]O2 plasma [108]Bi(N(SiMe3)2)3, H2O SrBiTaO [109]
Ba(thd)2 H2S BaS [10,110]
Ba(Me5C5)2 Ti(OiPr)4, H2O BaTiO3 [4,75]H2S BaS [65,82]
Ba(nPrMe4C5)2 Zr(NMe2)4, Y(MeC5H4)3, H2O BaOZrO2Y2O3 [111]
Ba(tBu3C5H2)2 H2O BaO [75,112]Ti(OiPr)4, H2O BaTiO3 [113]Ti(OMe)4, H2O BaTiO3 [112,114]
Ba(TpEt2)2 H2O BaB2O4 [115]Ba(tButAm2imidazolato)2 O3 BaO [104]
Fig. 2. Molecustability [113,1
iPr3C5H2 ling temperaand compoably high m
In orderbetween th
Table 3Cyclopentadiesity of Helsink
1 2 3 4 5 6 7 8 9 10 1114 15 16 17 18 19 20 lar structures of Ba(Me5C5)2(diglyme), Ba(tBu3C5H2)2 and Ba(Me4C5C2H4NMe)2 repre30].
igand have freedom to bend and rotate so that the melt-ture is lowered. Adducts with multidentate Lewis bases
unds with donor functionalized Cp ligands had remark-elting temperatures well above 200 C.
to study the differences in stability and volatilitye compounds thermogravimetric analyses (TGA) were
nyl compounds studied at Laboratory of inorganic Chemistry, Univer-i.
Compound Reference
Sr(C5Me5)2(THF)2 [65]Sr(iPr3C5H2)2(THF) [65]Sr(iPr3C5H2)2 [132]Sr(tBu3C5H2)2(THF) [130]Sr(tBu3C5H2)2 [130]Sr(C9H7)2(THF)x [131,132]Sr(Me7C9)2(THF)2 [131,132]SrC3H6(iPrC5H3)2 [131,132]SrC3H6(tBuC5H3)2 [131,132]Ba(Me5C5H)2(THF)x [65,113]Ba(C5Me5)2(A) A = dien, trien, diglyme, triglyme [113]Ba(iPr3C5H2)2(THF)2 [132]Ba(tBu3C5H2)2(THF) [113]Ba(tBu3C5H2)2 [130]Ba(Me2NC2H4C5Me4)2 [113]Ba(EtOC2H4C5Me4)2 [113]Ca(iPr3C5H2)2(THF)2 [42]
carried outmon practia dynamic sample is inof the sampA TG curve
Fig. 3. TG curatmosphere, 1senting the three different ways to pursue volatility and (thermal)
under nitrogen atmosphere (Figs. 35) as is the com-ce while evaluating ALD and CVD precursors. TypicallyTGA measurement is done i.e. the temperature of thecreased with a constant heating rate while the weightle is followed as a function of time and temperature.
showing a weight loss in a single step and residue
ves of cyclopentadienyl compounds of strontium. Flowing 1 atm N20 C/min, 10 mg samples.
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3303
Fig. 4. TG curves of cyclopentadienyl compounds of barium with donor function-alized ligands and Me5C5 ligands + ancillary Lewis base ligands. Flowing 1 atm N2atmosphere, 10 C/min, 10 mg samples.
close to zero indicates a good volatility and thermal stability. TGAgives information of thermal stability of a compound studied onlyin the condensed state, not in the gas phase. Therefore when thecompound studied is highly volatile and there is no or very littleresidue after the measurement it can only be said that the com-pound is thevaporationposition ofdecomposittral adduct of the mainthe precurscomplexes like iPr3C5Horate and leand Sr(Me5tively, indicmentioned with a loss plexes, i.e. tLewis base l(18, 19) or a
Fig. 5. TGABa(tBu3C5H2)2Sr(tBu3C5H2)210 mg sampl
TG measurements leaving 3070% residues. To make any conclu-sions what are the decomposition paths is quite impossible. Clearlyadding multidentate Lewis bases to Ba(Me5C5)2 (1114) did notenhance but lowered the volatility as compared to the Lewis basefree Ba(Me5C5)2 or the THF adduct (Fig. 4). Also the thermal stabilitywas not enhanced. This was also conrmed by vacuum sublimationexperiments: vacuum sublimation of the adducts 1114 requiredhigher temperatures (180260 C) than pure Ba(Me5C5)2 and alsothe sublimation yields were low (for 11 63% and for 1214 3239%)though it seems that the compounds sublimed intact as the subli-mates were found to be identical with the unsublimed compounds.Noticeable is that TGA showed Ba(Me2NC2H4C5Me4)2 (18) to bestable up to ca 290 C above which it decomposes rapidly. Undervacuum Ba(Me2NC2H4C5Me4)2 (18), however, showed very goodsublimation behavior at 200260 C.
In conclusion the best approach for making volatile and ther-mally stable cyclopentadienyl compounds of Sr and Ba wasdenitely the usage of large and sterically demanding cyclopenta-dienyl ligands. The iPr3C5H2 and tBu3C5H2 ligands were found tomake a good match with Sr and Ba. With similar ligands the thermalproperties of Sr and Ba compounds are very similar. Inset in Fig. 5shows TG curves measured for Sr(tBu3C5H2)2 and Ba(tBu3C5H2)2.It seems likely that iPr4C5H2 and (Me3Si)3C5H2 ligands couldbe equally good [135,136] but extremely crowded ligand iPr5C5
[137] may the reactiv
ntaisicateting zed cities ed ifenereen o su2). Thiket
thers are
Ba, t Ba co
and C5H2]. Groown ermally stable at least up to the temperature where the process was complete. Larger residues indicate decom-
the samples. Multiple steps indicate dissociation orion of the compounds studied, like dissociation of neu-ligands at lower temperature followed by evaporation
compound at higher temperature, or decomposition ofor compound into a stable and non-volatile solid. Thosewhich have protecting, sterically demanding ligands,2 and tBu3C5H2 (25, 1517) were found to evap-
ave residues below 8%. Compounds Sr(Me5C5)2(THF)xC5)2(THF)x (x = 02) left 18 and 37% residues, respec-ating some decomposition. Under vacuum the abovecompounds 15, 10 and 1517 sublimed at 110220 Cof possibly coordinated THF solvent. All the other com-hose having pentamethyl substituted Cp + multidentateigand (1114), indenyl (6, 7), N and O functionalized Cpnsa ligands (8, 9) seemed mainly to decompose in the
and pecomplAdductionaliVolatilenhanc
In ghave bbut als(Table the -dhigherdienylSr andSr and
SrSSr(iPr3[65,82are sh curves of Ba(Me5C5)2(THF)2 Ba(iPr3C5H2)2(THF)2 and(THF). Inset shows TGA curves measured for Lewis base freeand Ba(tBu3C5H2)2. Flowing 1 atm N2 atmosphere, 10 C/min,
es.
Fig. 6. Growthture. At 120 Cdistribution ac14 (2002) 193be even too bulky for Sr and Ba in the sense thatity with H2O is lowered. However, making the tetraopropyl substituted cyclopentadienyl ligands is mored compared to the trisubstituted ligands [135,137].the cyclopentadienyl compounds or using donorfunc-yclopentadienyl ligands was found not to be benecial.or thermal stabilities of adduct compounds were not
compared to the Lewis base free compounds.al cyclopentadienyl compounds of alkaline earth metalsused especially in ALD of oxides of Mg, Ca, Sr and Baldes of strontium and barium have been depositede advantages of the cyclopentadienyl compounds over
onates are high reactivity against oxygen precursors andmal stability. Oxygen precursors used with cyclopenta-H2O, O3 and oxygen plasma. Unlike the -diketonates ofhe cyclopentadienyl compounds form sufciently purentaining lms with all three oxygen precursors.
BaS were deposited using Sr(Me2C5)2, Ba(Me5C5)2 and)2 as metal precursors and H2S as a sulfur precursorwth rates of the processes at different temperaturesin Fig. 6. Typical ALD window was observed for all
rate of SrS and BaS thin lms as a function of the deposition tempera-, a pulse length of 1.5 s was used with complex 1 to ensure its uniformross the substrate [65]. Reprinted with permission from Chem. Mater.7. Copyright 2002 American Chemical Society.
3304 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
three processes. As can be seen from the gure, the temperaturewhere the growth rate starts to increase is lower when Sr(Me5C5)2(>350 C) and Ba(Me5C5)2 (>300 C) are used as precursors ascompared to Sr(iPr3C5H2)2 (>380 C) [82]. The increase of thegrowth ratetion of the punder the ssors followexpected: Bligand thanformed molm. The surity levels i0.30.5 at.%and BaS prof the Cp clower evapoalso resultetemperaturgrown usin
We haveand ternarmaterials dBa1xSrxTiOcursor. Watlarger alkalhydroxide [of dehydraprocess temcharacteristimportant ttion/dehydrof the other
The binadue to theirand BaO groabsorption used were known thatH2O and forof Ba oxidechemistry otion. The ampurge timesBa(OH)2. Athas a slightto be muchrates were and Sr(tBu3crystalline problems wand H2O. AFinally, MgCgrowth rateprocess at 2
For SrTiTi(OiPr)4, HWith the The growthmostly amowhich is alSrTiO3 lming approxthen in prein Sr(OH)2substrates
crystalline whereas on amorphous Al2O3 they were amorphous[70]. When SrTiO3 was deposited on an MBE-grown seed layerat 325 C the lm grew epitaxially and no interface between theMBE and ALD layers could be seen with HRTEM (Fig. 7). To deposit
omeriateh th
3C5H Sr:T
Sr:Ti as fo
tion w rateO cysitio
in Alectrpaci)4 an. A 49O anng inigh, ositi
3C5)213], ulted
of 27 at 3servpositte. Atced hater
ratelengtsor sraturoundC wm. Wed toA/cydes d EO
alsocompere d
e)4 anio the.30:1res wers se curr thscalereactdepo)4 an
the obssitioubst
Sr M M. The at high temperatures is associated to a decomposi-recursors. The temperature independent growth rateself-limiting conditions observed for the three precur-
the order Ba(Me5C5)2 > Sr(Me5C5)2 > Sr(iPr3C5H2)2 asa is a larger cation than Sr, and iPr3C5H2 is a larger
Me5C5 requiring larger surface area in the saturativelynolayer of the precursor on the surface of the growinglde lms deposited were all polycrystalline. Impu-
n the lms deposited at 300 C were low (C 0.1 at.%, H, O 0.30.6 at.%). When compared with the previous SrSocesses using the thd compounds and H2S, the usageompounds allowed lower deposition temperatures asration temperatures could be used. The Cp compoundsd in higher or similar growth rates, depending on thee, as the thd compounds. Roughnesses of the SrS lmsg the Cp compounds of strontium were also lower.
studied cyclopentadienyl compounds for both binaryy oxides of the larger alkaline earth metals. Oxideeposited have been CaO, SrO, BaO, SrTiO3, BaTiO3 and3. In all the processes H2O was used as the oxygen pre-er is a difcult precursor for the binary oxides of theine earth metals because of the easy formation of metal112]. The hydroxide formation and the reverse reactiontion back to the oxide are strongly dependent on theperature and can greatly complicate the ALD growthics of the binary oxides. Fortunately in the far moreernary oxides, such as SrTiO3 and BaTiO3, the hydra-ation processes are effectively stopped by the presence
metal oxide [70,112].ry oxides of Ca, Sr and Ba are not important materials
high reactivity in ambient atmosphere. However, SrOwth experiments were carried out in order to study theof H2O in the growing lms [70,112]. The precursorsBa(tBu3C5H2)2, Sr(iPr3C5H2)2 and Sr(tBu3C5H2)2. It is
especially BaO is very keen in absorbing large amountsming Ba(OH)2(H2O)n which could interfere the growth
materials. It was found that at 250340 C the growthf BaO is signicantly inuenced by the H2O absorp-ount of H2O absorbed decreases with increasing H2O
and growth temperature. Up to 290 C the lms were 340 C there was still some Ba(OH)2 in the lms. SrOly lower reactivity towards H2O than BaO and it proved
easier to grow uniform SrO lms. At 300 C the growth0.42 and 0.28 A/cycle for processes using Sr(iPr3C5H2)2C5H2)2, respectively. At this temperature the lms were(1 1 1) oriented SrO. With calcium there is much lessith hydration. CaO was deposited using Ca(iPr3C5H2)2t 300 C the growth rate was around 1 A/cycle [42].p2 is a good precursor for ALD of MgO lms. Constant
of 1.16 A/cycle could be achieved with the MgCp2H2O00300 C as reported by Putkonen et al. [26].O3 two sets of precursors were used: Sr(iPr3C5H2)2,2O [4,76], and Sr(tBu3C5H2)2, Ti(OMe)4 and H2O [70].rst set temperature range of 250325 C was studied.
was the most ideal at 250 C giving highly uniform butrphous lms with a growth rate of 0.5 A/cycle. At 325 Cready above the decomposition limit of Ti(OiPr)4 thes began to crystallize during the growth when exceed-. 50 nm thickness, rst with random orientation andferred (1 0 0) orientation. High water doses resultingformation enhanced the crystallization and also thehad an effect. On crystalline platinum the lms were
stoichiapprop
WitSr(tButhe 1:1and a ratio wsaturagrowththe Srcompofeature
DieMIM caTi(OiPr500 Cand H2resultifairly h
DepBa(MeH2O [1set resaturesgrownrate obdecommediaenhanafter wgrowthpulse precurtempewere fat 340the limprovto 0.50electro280 an
ALDferent lms wTi(OMing ratfrom 1structuBST layleakag
Aftesmall larger While Ti(OiPrdue toreadilycompoglass sTi andby ASwaferstric SrTiO3 a cycling ratio of 5:6 (Sr/Ti) was found to be.e second set of thermally more stable precursors2)2 and Ti(OMe)4, SrTiO3 was grown at 300 C [70]. Withi cycling ratio a self-limiting growth rate of 0.4 A/cycleratio of 3:7 in the lm were achieved. The 2:1 cyclingund to result in slightly Sr rich lms but reaching theas found to be more complicated and an increase in the
was observed. The average surface composition duringcle clearly inuences the growth behaviour [70]. Surfacen effecting the growth behaviour seems to be a generalLD of ternary compounds.ic properties of the SrTiO3 lms prepared were tested intors. A 490 nm lm grown at 325 C using Sr(iPr3C5H2)2,d H2O had a permittivity of 130 after annealing in air at
nm lm grown at 300 C using Sr(tBu3C5H2)2, Ti(OMe)4d annealed at 750 C had a permittivity of 90 at 1 V,
EOT of 2.2 nm. Leakage currents of
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3305
Fig. 7. HRTEM SiO2 lepitaxial SrTiO
H2O was these proceintegration
The reachigh and froprecursors tion may caespecially fmade indussor. STO lSr-rich ALDage (107 ASr(tBu3C5Hannealing aRu/RuOx/Srrutile TiO2.ALD of STOminimizingthe Ti conteoxygen vacboth proces
3. Group 4
Oxides orials group their good dhigh-k gatedielectric min DRAMs aing ZrO2Athin lm prhigh dielecas a dielectcurrents. Ticoatings, annanolaminathe electrolcrystal strulm deposittion of the the anatasewhen a hig
nal a appl
rece proy, prwithomp
give
oup
oridee voe ALrs [1Cl4 apatibthatausinle is ecurhe oible
chlo images showing the inuence of an epitaxial MBE seed layer on ALD growth. A 3 [70].
rst scaled up on 200 mm silicon wafers [114]. Latersses have been extensively studied at IMEC for device
[8699].tivity of the Sr and Ba Cp compounds with water ism that point of view there is no need for other oxygen[4,86]. But as noted, with water the hydroxide forma-use complications. In addition, slow purging of waterrom large batch reactors with high surface areas hastry to prefer ozone as an alternative oxygen precur-ms can be grown with ozone but the state-of-the-art
STO lms which show EOT < 0.5 nm with low leak-/cm2) at 1 V have been made with water using the2)2H2OTi(OMe)4H2O process at 250 C followed byt 600 C [88,92]. The structure of the MIM capacitor was
rich STO/TiN. The RuOx/STO interface may also contain Both RuOx and rutile TiO2 are benecial surfaces for. They help the crystallization of the STO lm thereby
the dead-layer in STO. In addition TiO2 may increasent in STO lms and RuOx can heal traps associated toancies by releasing oxygen during annealing with theses decreasing the leakage current density.
elements
tetragohigh-k
Thedesiredstabilitibility latest cligands
3.1. Gr
Chlthey archlorid20 yeaand HfincomSrTiO3sibly cexampALD prFrom tas poss
The
f the group 4 metals form one of the most studied mate-in ALD. The interest towards ZrO2 and HfO2 stems fromielectric properties. ALD HfO2 is used industrially as a
oxide in MOSFET transistors [138] and ALD ZrO2 as aaterial in DRAM capacitors [139]. The high aspect ratiosnd also the need to decrease the leakage current by mak-l2O3ZrO2 nanolaminates call for the use of ALD in theocessing. TiO2 is a versatile material which also has atric constant especially in rutile form but is not usedric material because of the problems of high leakageO2 lms nd applications as photocatalysts, protectived as a component in titanates and mixed oxide andtes such as AlxTiyOz (ATO) that is used commercially inuminescent displays. Group 4 metal oxides have severalctures with properties of their own. Therefore in thinion besides the composition and purity also the forma-desired crystal structure is important. In case of TiO2
form is preferred in photocatalysts, and the rutile formh-k value is needed. ZrO2 and HfO2 have monoclinic,
from room 750 C. As mhigh-k applmost stablearound 20 3040. The210 C, tetrmonoclinicthe lms rstructure [1mainly mo[145]. Withphase increnanolamina
Metal cha binary oxthe two prebyproduct arate oxygeayer which forms during the processing is seen between Si and the
nd cubic forms from which the last ones are desired inications.nt development in Ti precursors for ALD aims at theperties mentioned above: high reactivity, good thermaleferably liquid at room temperature, and good compat-
Sr(R3C5H2)2 precursors for deposition of SrTiO3. Thelexes are heteroleptic following the idea that one of thes reactivity and the other thermal stability.
4 metal halides as precursors for ALD
s of group 4 metals are very important precursors sincelatile, stable, cheap and highly reactive with water. TheD process for all these oxides has been known for over,2,140]. Chlorides have certain drawbacks such as ZrCl4re solids, the reaction by-product is corrosive HCl, andility with some other lm constituents like strontium in
readily forms SrCl2. The problem of being solid and pos-g particle problems has been overcome and HfCl4, forused in industrial scale [141]. TiCl4 is extensively usedsor when cheap, easily delivered precursor is needed [5].ther halides iodides have been in some extent studiedprecursors for the TiO2, ZrO2 and HfO2 lms [142,143].ridewater process works at a wide temperature range:
temperature (with the highly volatile TiCl4) to at leastentioned above, in the growth of ZrO2 lms for the
ications the crystal structure formed is important. The phase is the monoclinic one but its k-value is onlywhile the tetragonal and cubic phases have values of
ZrCl4H2O process produces amorphous lms belowagonal at 210300 C, and a mixture of tetragonal and
at higher temperatures. Interestingly the k-values ofemained rather low (around 20) despite the crystal44]. HfCl4H2O process on the other hand, produces
noclinic phase which also shows k-value close to 20 decreasing lm thickness the fraction of the tetragonalases, however, as observed in for example HfO2Ta2O5tes [145].lorides react also with metal alkoxides producing eitheride or mixed oxide depending on whether the metals incursors are the same or different, respectively [146]. Theof this reaction is alkylchloride [147]. Because no sep-n precursor is used, and oxygen is bound to a metal
3306 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
Fig. 8. Thermequipment wiposition was dfragment [149
with high ounderlying
3.2. Group
Metal alIn ALD simpwidely studZr and Hf, bof their lowsmaller alkodepends onother handOMe > OEt >nium (Fig. 8with iso-pr
TiO2 witides and wdecompositthe titaniummethoxide position onthat the lence of lm[150,151]. Tthe precursthat gives a
Polycryswater at 20the temperhand, the and longerHowever, tindicating textremely ause in ALD
The poohas been trifunctionalizin homolepclusion thatbe achievedimproves sheterolepti
rystal structure of [Zr(OtBu)2(dmae)2]2 dimer [154]. Reproduced with aion from J. Non-Cryst. Solids 303 (2002) 24. Copyright 2002, Elsevier.
dmae (dmae = dimethylaminoethoxide) (Fig. 9) allowed toe slightly the deposition temperature but thermal decompo-f the metal precursor was anyway seen in the experiments
The situation was similar when OtBu ligand was changed tor a complete homoleptic dimethylaminoethoxide complex,
ae)4, was used [155]. The monomeric heteroleptic alkoxideu)2(mmp)2 with tertiary mmp ligand (mmp = 1-methoxy-2-l-2-propanolate) (Fig. 10) [156] showed rather good thermaly up to 275 C but at these temperatures the reactions withremae conC bun in A/cyly [1ogethwith
at ly coer, wwthal decomposition of Ti(OMe)4, Ti(OEt)4, Ti(OiPr)4 and Ti(OtBu)4. ALDth in situ QMS was used to produce the data. The amount of decom-etermined by dividing the main fragment of the ligand by a precursor].
xygen afnity, the process oxidises the substrate orlm less than the normal oxide ALD processes.
4 alkoxides as precursors for ALD
koxides are common precursors in CVD of oxide lms.le titanium alkoxides (OMe, OEt, iOPr, tOBu) have beenied and used with water to deposit TiO2. Alkoxides ofy contrast, have been much less used in ALD becauseer thermal stability compared to the Ti alkoxides. Thexides tend to oligomerize and the oligomerization also
the metal increasing in the order Ti < Zr Hf. On the, the stability of the alkoxides decreases in a series
OiPr > OtBu which has been best conrmed with tita-) [148,149]. Monomolecular complexes are formed onlyopoxy and t-butoxy ligands.h anatase structure can be grown from titanium alkox-ater between 225 and 350 C. Because of thermalion of the alkoxides, the maximum temperatures with
ethoxide and i-propoxide are about 325 C while theallows the use of 350 C [148,149]. Close to the decom-set temperature however the decomposition is so slowms are still reasonably uniform and a linear depend-
thickness on the number of deposition cycles is validhe growth rate per cycle follows inversely the size ofor being the highest with the smallest methoxide ligand
Fig. 9. Cpermiss
ligandincreassition o[154]. OiPr oZr(dmHf(OtBmethystabilitwater residuat 240be seeto 1.48similarused trange urationseverehowevthe gro rate of 0.65 A/cycle at 350 C.talline ZrO2 lms can be grown from Zr(OtBu)4 and0300 C. The growth rate decreased strongly whenature was increased from 200 to 300 C. On the othergrowth rate increased with higher precursor doses
pulse times indicating decomposition of Zr(OtBu)4.he polycrystalline lms showed a high-k value of 32etragonal or cubic structure [152]. Hf(OtBu)4 is also anir and moisture sensitive compound and as difcult to[153].r thermal stability of zirconium and hafnium alkoxidesed to be improved by using ligands that have been donored by amine or ether groups. These have been used bothtic and heteroleptic complexes but with a general con-
thermal stability and high reactivity with water cannot at the same time because coordinative saturation thattability also decreases reactivity. The use of a dimericc precursor [Zr(OtBu)2(dmae)2]2 with primary alkoxy
Fig. 10. Crystapermission froand Sons.in incomplete resulting in low growth rate and hightents [157]. Fairly homogeneous lms could be grownt at 340 C some decomposition occurred which couldthe increase of the growth rate from 0.58 A at 240 Ccle at 340 C. The homoleptic Hf(mmp)4 behaved quite58]. A homoleptic hydroxylamide Hf(ONEt2)4, whener with water, showed a narrow growth temperaturethe optimum being 300 C where reasonably good sat-0.5 A/cycle was achieved [159]. The lms were quitentaminated with hydrogen (11 at.%) and carbon (6 at.%),hile the nitrogen content was below 0.1 at.%. At 250 C
was weak, showing that the bidentate hydroxylamidel structure of M(OtBu)2(mmp)2 (M = Zr, Hf) [156]. Reproduced with am Chem. Vap. Deposition 8 (2002) 163. Copyright 2002, John Wiley
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3307
ligands shield hafnium more effectively than the related monoden-tate NR2 and OR ligands.
3.3. Group 4 alkylamides as precursors for ALD
Alkylamgroup 4 meusually conthe nitrogewidely usedtages: easymolecules rlms alkylabined in thewith ammo
Thermalreverse ordstable thanThermal de180 C [160the growth strongly tomuch higheimum depoand Zr(NEt2and Hf(NEtdecompositour reactorstemperaturdifference bperatures h[164].
Good reZr(NMeEt)4values close[165]. This cand Hf(NMdeposited fstructure anto be moreHfO2 can bearth ions. 7% of yttriunal or cubicstrontium a[170].
3.4. Group
3.4.1. Bis-cyBis-cyclo
organometametal oxideand the othgroups. Theboth waterThe thermaof alkoxidesare achievahave only s
The ALDferent bis-Cprecursors (Fig. 11). Ththe Cp2Zr(Cof Cp group
GrowHfMefMe2
istry.
ns with thline mixelinic
compfers les tprecon inlowe
(Cpy as omppouctio
are Me l
showrmit
Monoescrut modest growth rate and tendency to form lms with ae of oxide phases. Zirconium alkylamides, on the other handhigher volatility and growth rate but suffer from thermallity. Films grown from alkylamides show in addition betteroutcome where the high permittivity tetragonal and cubic
are dominating. In mixed ligand complexes (CpR)Zr(NMe2)3Me or Et) the benets of the both precursor types can beed [177]. Pure, conformal lms (Fig. 13) were grown by an
process with a growth rate close to 1 A/cycle. No signs of pre- decomposition were seen at 300 C. A CET value of 0.7 nmrgeted low leakage current density was measured from a
apacitor with 6 nm ZrO2 lm. When the similar cyclopenta- and alkylamido ligands were applied to hafnium the resultsot as good as with zirconium. HfO2 lms grew in thef(NMe2)3O3 process with a rate of 0.8 A/cycle but the lmsed mixed phases with the monoclinic phase being dominant
The as-deposited thin lms (1 nm.
Cp try premolecuin the variatiligand studiedstabilittuted cCp comThe reagroupsthe Cpmentsand pe
3.4.2. As d
bility bmixturshow instabiphase phases(R = H, combinozone cursorwith taMIM cdienylwere n(CpR)Hcontain[178]. after alms sth rate of HfO2 lms as a function of deposition temperature. Precur-2 and water. The inset shows the growth rate at 350 as a function ofpulse length [173]. Reproduced by permission of The Royal Society
ith the surface OH groups (Fig. 12) [174]. This is ine expected bonding strength of the ligands. The poly-lms of HfO2 show monoclinic structure while ZrO2d phases. Thus, the presence of the low permittivity
phase in ZrO2 has an effect on CET values which were
ounds of the group 4 metals are usually solids. Indus-liquid source materials and the tuning of the precursoro liquid at room temperature is one of the driving forcesursor development. In bis-Cp compounds such a small
structure as a change of one methyl ligand to a methoxyred the melting point below room temperature. TheMe)2Zr(OMe)Me compound showed similar thermalthe bis-methyl compounds [175]. The methoxy substi-ound improved slightly also the other weak property ofnds, viz. low growth rate which increased by 0.1 A/cycle.n mechanism studies showed that methyl and methoxymainly removed during the metal precursor pulse andigand during water/ozone pulse [176]. XRD measure-ed that the ZrO2 lms formed contained mixed phases
tivity value remained quite low, however.
-cyclopentadienyl compoundsibed above, the Cp-precursors offer good thermal sta-
3308 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
Fig. 12. Surfacbased on in simethyl than Cgroups (b). Thereleasedconvesion from Lang
Mixed lithermally htivity againused. Initiawindow wiRose et al. [1slowly fromat 350 C. Thsignicantl350 C. SrTi(Me5C5)Ti(Ousing SrO:TSr/Ti atomilized by anhad much l
3.5. Other g
Low thecern especithat the pounds aime reactions in the Cp2ZrMe2/D2O ALD process at 350 C as proposedtu studies with QMS and QCM. Upon adsorption of Cp2ZrMe2 morep ligands are released in exchange reactions with the surface OD
following D2O pulse releases the rest of the ligands as MeD and CpDrts the surface back to OD-terminated [174]. Reprinted with permis-muir 21 (2005) 7321. Copyright 2005 American Chemical Society.
gand complex (Me5C5)Ti(OMe)3 has been found to beighly stable[179181] but unfortunately they lack reac-st water and more powerful oxidizers like O3 have to bel ALD investigations revealed a very wide ALD processth deposition temperatures up to 375400 C [179,180].81] found that at 260330 C the growth rate increased
0.22 to 0.29 A/cycle and then raised up to 0.50 A/cyclee density of lms was the highest at 330 C and dropped
y at 350 C. However, the growth was still linear atO3 was deposited at 250 and 300 C using Sr(iPr3C5H2)2,Me)3 and O3 as precursors [81]. Films grown at 300 C
iO2 cycle sequences of 10:10 and 5:5 both possessed ac ratio of 0.90. The lms grown at 300 C and crystal-nealing had high leakage while lms grown at 250 Cower leakage.
roup 4 precursors for ALD
rmal stability of titanium complexes has been a con-ally in deposition of SrTiO3 lms where it is desiredlms grow directly into the crystalline form. Com-ed to solve this have been e.g. the above mentioned
Fig. 13. Crossprocess at 275part of the tring is 115 nmReproduced b
(Me5C5)Ti(OInterestinglalkoxo andthermal stshow low tTi(NMe2)2(in a control
Yet anoa combin(CHT = C7H7been synthstable and The oxide g0.70.8 A/cytivity phase-sectional SEM images of a ZrO2 lm grown by the CpZr(NMe2)3/O3C into 60:1 aspect ratio trenches. The upper image shows the top
ench, those below the middle and bottom parts. The trench open- and depth 6.75 mm. Labels denote the ZrO2 lm thicknesses [177].y permission of The Royal Society of Chemistry.
Me)3 and Ti(iOPr)2(thd)2 [5557,71,79,81,8385,87].y the recent approach to use heteroleptic alkylamido-
acetamidinato-alkoxo complexes also improved theability although the homoleptic parent compoundshermal stability [182]. TiO2 lms can be grown fromOiPr)2 and Ti(iPrNCHNiPr)(OiPr)3 together with waterled way up to 350 C with a rates of 0.75 and 0.45 A/cycle.ther approach to use heteroleptic complexes involvesation of cyclopentadienyl and cycloheptatrienyl) ligands (Fig. 14) [183]. These complexes have
esized for Ti and Zr. The compounds are thermally verygrowth temperatures exceeding 400 C are possible.rowth requires the use of ozone and growth rates ofcle have been recorded. In ZrO2 lms the high permit-s are dominating and accordingly low CET values were
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3309
Fig. 14. Molecfor CpTi(C7H7)Mater. 24 (201
reported. Tat low temp(400 C), an
4. Group 1
Metal seuously incresolar cell abchange randfor future ngration denrequires thnanometerssmall areasmaterial inthe thin lmnique for rethe ALD cherial Ge2Sb2Tof three eleall consider
For non-clean liganmany nonmwithout thicautions be
Molecular structure of (tBuMe2Si)2Te [3]. Reproduced with permission fromem. Soc. 131 (2009) 3478. Copyright 2009 American Chemical Society.
l level. By contrast, the hydrides of selenium and telluriumch more dangerous and would require extensive safety sys-nd exhaust gas treatment. Alkyl compounds of selenium andm have been quite widely used in CVD but in ALD they do
ovide efcient exchange reactions with the common metalsors. In group 15 there is similar trend: NH3 can be used withriate precautions while PH3, AsH3 and SbH3 are much moreous
kylsi
reaktes, wre idFig. 15. J. Am. Ch
harmfuare mutems atelluriunot prprecurappropdanger
4.1. Al
A bselenidSe) weular structure of (MeCp)Zr(C7H7) (above) and TG curves measured and (MeCp)Zr(C7H7) [183]. Reprinted with permission from Chem.2) 2002. Copyright 2012 American Chemical Society.
iO2 lms on the other hand showed anatase structureerature (250 C), rutile structure at high temperaturesd mixed phases in between.
5 and 16 elements
lenide and telluride thin lms have been gaining contin-asing interest. Two well-known examples are CuInSe2sorber and Ge2Sb2Te5 phase change material. Phase-om-access memory (PCRAM) is a promising technologyon-volatile data storage [184186]. Reaching high inte-sities together with low power consumption in PCRAMsat memory cells are scaled into dimensions of a few
only. Connement of the phase change material into and volumes is most efciently done by depositing theto tiny vias. This requires excellent conformality from
deposition method and makes ALD an attractive tech-placing the so far dominant sputtering methods. Frommistry point of view the common phase change mate-e5 (GST) was seen a major challenge because it is made
ments that had not been studied much earlier and wereed difcult to ALD.metals hydrides are often an ideal choice as they allowd elimination by protonation. However, hydrides ofetals are highly toxic. In group 16 water of course iss concern, and H2S can be used with appropriate pre-cause its odor threshold limit is much lower than the
thermal staized that alproperties [and Fig. 16 lyl compouthe group oration behVolatility oweight. ComtemperaturResidues seAs < Sb < Bi.is accompadecompositapparently
Fig. 16. TG cuTe [188].gases.
lyl compounds of group 15 and 16 elements
hrough in ALD of GST and related materials, includingas made when alkylsilyl compounds (R3Si)2M (M = Te,
entied as efcient ALD precursors with good volatility,bility and high reactivity [3]. Soon after, it was also real-kylsilyl compounds of group 15 elements have similar187,188]. Fig. 15 shows the structure of (tBuMe2Si)2TeTG curves measured for a set of group 15 and 16 alkylsi-nds. TGA measurements reveal that all compounds of16 elements are volatile and they exhibit pure evap-avior. Their thermal stability also seems to be good.f the compounds depends strongly on their molecular
pounds of group 15 elements lose weight at higheres compared to the group 16 Se and Te compounds.en at the end of the measurements get larger in the order
DTA (not shown) revealed that the step of weight lossnied by heat release (exotherm) which indicates someion. However (Et3Si)3As leaves practically no residuebecause the decomposition products are volatile too. Allrves measured for different alkylsilyl compounds of As, Sb, Bi, Se and
3310 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
Fig. 17. Tempdeposited with[3,187,190].
compoundswith largerpounds stuwhile those
Compouious metal hselenides. Fby ALD at rSbCl3 and Gdioxane add
The alkystraightforwmetal tellurthe metal p
(R3Si)2Te(g
In situ microbalandechlorosilexchange aduring bothThe favorabsoft acidbLewis acid, Lewis basespair. Upon ebonded to ais a soft Lewuct of the eacidbase p
Both Sb2behavior aslong enougperatures tpulse timespeculiar feagrowth of Sas low temprapidly (Figthe lowest tion comingof 70 C [3,rapidly andlargest concvolatility of
GeSloride
denc for ason
butthanwn asor anary of Geed wTe [n in , thepos
atedc comnd Sng dernam anre tw, thome
and adju
well dechlorosilylation reaction was later extended to also ele-erature dependencies of ALD growth rates of various materials metal chlorides and alkylsilyl precursors of tellurium and antimony
studied are sensitive to air and moisture. Compounds alkylsilyl groups are less sensitive. Most of the com-died i.e. those with Me3Si and Et3Si ligands are liquids
with tBuMe2Si are low melting solids.nds of tellurium and selenium react efciently with var-alides forming the corresponding metal tellurides and
or example, Sb2Te3, GeTe and GST lms were depositedemarkably low temperature of 90 C using (Et3Si)2Te,eCl2C4H8O2 as precursors, where C4H8O2 is a neutraluct ligand.lsilyl compounds of tellurium and selenium offer aard exchange reaction route to the deposition of theide and efcient elimination of the chloride ligands ofrecursor, the overall reaction being:
) + MCl2(g) MTe(s) + 2R3SiCl(g) (1)reaction mechanism studies with a quartz crystalce and quadrupole mass spectrometer veried theylation reaction mechanism and showed that the ligandnd formation of R3SiCl as a byproduct occur stepwise
metal chloride and alkylsilyl tellurium pulses [189].ility of this reaction is supported by the Lewis hard-
ase concept: in (R3Si)2Se and (R3Si)2Te silicon, a hardis bonded to the heavy group 16 elements which are soft, thus forming an unfavorable hard-soft Lewis acidbasexchange reaction with metal chlorides, silicon becomes
harder base. Further, if the metal of the metal chlorideis acid (large cation with a low charge), the other prod-
Fig. 18. metal ch
depenlookedyet. Recussionrather be groprecur
Tercycles obtainand SbpositioIdeallythe comand relometriGeTe ano stroin the tmaniuexposuAnyhostoichiGeTecan beline as
The
xchange reaction will form a favorable softsoft Lewisair (MSe or MTe).Te3 and GeTe binary processes showed the self-limiting
expected in ALD, i.e. the growth rates saturated withh precursor pulses. Despite the low deposition tem-he saturation level was reached with reasonably short
of only 1 s. Temperature-wise the processes showedtures compared to the ALD processes in general. Theb2Te3 occurred with a good rate of about 0.6 A/cycle aterature as 60 C but above that the growth rate dropped. 17). Also with GeTe the highest rate was achieved atpossible deposition temperature of 90 C, the limita-
from GeCl2C4H8O2 that needs a source temperature190]. Above 90 C the GeTe deposition rate decreased
dropped to zero already at 150 C. Consequently theern related to these processes arises from the modest
GeCl2C4H8O2 combined with the strong temperature
mental antthose of gerSbCl3 and (Ea germaniution of 82 apoints to a ligand exchstitution is has no stoiction. Anyhoof Ge82Sb18two extremcycles werea wide rang
Despite from the gelow impuribTe phase diagram showing compositions reached so far with thealkylsilyl tellurium/antimony based ALD processes.
e of the growth rate. Various alternatives have beenGeCl2C4H8O2 but no better precursor has been founds for these temperature dependences are still under dis-
they seem to be related to these specic lm materials this chemistry in general because zinc telluride couldt a high temperature of 400 C using the same telluriumnd ZnCl2 [3].GST lms were deposited by mixing the binary ALDTe and Sb2Te3. A growth rate of about 0.30 A/cycle washere the cycle refers to a sum of binary cycles of GeTe3,191]. A common way for controlling the ternary com-ALD is to vary the cycle ratio of the binary constituents.
lm should adopt the stoichiometry corresponding toition of the stable ternary compound but because GST
ternary materials are more alloy-like and many stoichi-positions exist along the pseudo-binary line between
b2Te3 (Ge1Sb2Te4, Ge2Sb2Te5, Ge3Sb2Te6, etc.) there isriving force towards a particular stoichiometry. Indeed,ry process the antimony content increased and the ger-d tellurium contents decreased with increasing SbCl3imes without approaching any specic composition.e GST composition could still be adjusted close to thetric Ge2Sb2Te5 by tuning the ratio of the binary cyclesSbTe. On the other hand, if so desired, the compositionsted to other stoichiometries on the GeTeSb2Te3 tie
by just changing the binary cycle ratio.imony (Fig. 17) and various metal antimonides, alsomanium [187]. Elemental Sb lms were deposited fromt3Si)3Sb, while GeCl2C4H8O2 and (Et3Si)3Sb depositedmantimony alloy with a germanium rich composi-t.% Ge and 18 at.% Sb. The germanium rich compositionreaction mechanism more complicated than the simpleange; most likely some antimony to germanium sub-involved, too. Contrary to the GeSbTe system, GeSbhiometric compound and tends to elemental segrega-w, by mixing the ALD cycles of elemental Sb with those, the lm composition could be tuned in between thesees. Furthermore, when also GeTe and/or Sb2Te3 ALD
added, the GeSbTe composition could be tuned overe of compositions in the ternary phase diagram (Fig. 18).the low deposition temperature, the GST lms depositedrmanium and antimony chlorides contained reasonablyty contents of about 2.4 at.% oxygen, 1.0 at.% hydrogen,
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3311
0.7 at.% carbon, and 0.6 at.% chlorine as analyzed by Time-of-FlightEnergy Recoil Detection Analysis (TOF-ERDA) [3,191]. In elemen-tal Sb lms no chlorine could be detected with ERDA [187]. Thesilicon residues were more difcult to analyze because the lmswere on Si substrate, but the silicon content appeared to be below1 at. %. Low levels of silicon and carbon in the GST lms havealso been veried with Secondary Ion Mass Spectrometry (SIMS)[191].
The binary Sb2Te3 and elemental Sb lms were found to growin polycrystalline form whereas with GeTe both amorphous andpolycrystalline lms have been obtained. The ternary GST hasusually been amorphous in the as-deposited state which can beconsidered as a fortunate consequence of the low deposition tem-perature because strong crystallization during the growth couldprevent complete lling of the three-dimensional PCRAM memorycell structures. With the amorphous GST and the GeTe lms com-plete lling of sub-100 nm holes in actual PCRAM device structureswas successfully demonstrated (Fig. 19).
Phase change properties of the ALD GST lms have been studiedwith high temperature X-ray diffraction and resistivity measure-ments and compared to the sputtered GST lms. In general, thephase change properties exhibited by the ALD GST lms have beensimilar to their sputtered counterparts though in the ALD lmsthe phase change temperatures and resistivities were somewhathigher. The small amounts of impurities, oxygen in particular, in theALD GST lms likely explain these differences. Laser pulse exper-iments have veried that the ALD GST lms can be repeatedlyswitched between amorphous and crystalline states.
While thdone on lmsubstrates, wafers witThis also aldemonstratswitched bcycles with
5. Bismuth
Many aplms contasensors and
Fig. 19. Transmission electron microscopy image of a GeTe lm ling a31 nm 84 nm via indicated by the arrow [190]. Reproduced with permission fromJ. Electrochem. Soc. 158 (2011) D694. Copyright 2011, The Electrochemical Society.
fuel cells) [xide
e var, holas sees, latinglluridle inike foum aiCl3, h teoligofcie
Table 4Compounds of
Precursor Material References
BiCl3 Bi2Se3 [3]Bi2Te3 [3]Bi2S3 [200]
Bi(Ph)3 BiTiO pyrochlore [201]Bi4Ti3O12 [202]SrBi2Ta2O9 [203]
Bi(CH2SiMe3 BiSiO [204]
Bi(N(SiMe3) BiTaO [109]BiTiO pyrochlore [205]Bi4Ti3O12 [205]SrBi2Ta2O9 [109]
Bi(thd)3 Bi2O3 [206,208]Bi(OCMe2iPr Bi2O3 [206]
Bi(mmp)3 + H2O Bi4Ti3O12 [209]BiTiO pyrochlore [210]BiTiSiO [211,212]BiTiAlO [213]e early research on ALD of GST reviewed above wass made in small research reactor with 50 mm 50 mmlater the GST ALD process was scaled up to 200 mmh less than 3% 1 thickness nonuniformity [192].lowed fabrication of integrated PCRAM devices whiched that the memory cells with ALD GST could beetween the low and high resistivity states up to 106
a very stable resistance for both states.
plications have been suggested in literature for thinining bismuth. Binary bismuth oxide can be used in
solid electrolytes (ionic conductors in solid state oxide
metal oincludopticallysts, gbattericonducand teexamp
Unlstronti, like Bbismuta low the su
bismuth used as precursors in ALD.
Other precursors
(Et3Si)2Se (Et3Si)2Te H2S
Ti(OiPr)3 + H2O O3 + Ti(OiPr)4 + H2O Bi(Ph)3 and SrTa2(OEt)10me2 in n-butylacetate + O2 plasma
)3 O3
2)3 Ta(OEt)5 + H2O Ti(OMe)4 + H2O Ti(OMe)4 + H2O SrTa2(OEt)10(dmae)2 + H2O
H2O )3 H2O
Bi(mmp)3 and Ti(mmp)4 in ethylcyclohexane + O3/O2 Ti(OiPr)5Ti(OiPr)4 + Si(OEt)4 + H2O Bi(mmp)3, Ti(mmp)4 and Si(OEt)4 in ethylcyclohexane + O3Bi(mmp)3 and Ti(mmp)4 in ethylcyclohexane, AlMe3 + H2O 193196]. Ternary and quaternary bismuth containings have a wide spectrum of potential applications. Theseious data storage devices (DRAM, FRAM, magnetic,ographic), electro-optical modulators, oxidation cata-nsors, solid oxide fuel cells, photocatalysts, lithium ionsers, optical amplier, superconductors, and oxygen
solid electrolytes [193]. Bismuth suldes, selenideses are other materials of interest. They nd use for
thermoelectrics [197].r example in the case of the larger alkaline earth metalsnd barium even the simplest compounds of bismuthBiMe3 and Bi(OtBu)3 are quite volatile. This is becausends to form compounds that are monomeric or havemerization degree which in turn may be explained byntly high electronegativity (2.02) and charge to ionic
3312 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
radius ratio (Bi3+, 2.91(CN = 6)) of bismuth [198,199]. So volatilityis not the main issue while trying to nd suitable ALD precursorsfor bismuth. The issues are more like insufcient reactivity and lowthermal stability.
ALD procchalcogenidSuccessful BiCl3, BiPh3Bi(OtBu)3 a
5.1. Bismut
BiCl3 is apressure oftures in ALDand (Et3Si)formed [3]fur precurschalcogenidthe soft Lew(S2, Se2, pure oxideoxychloride[215].
5.2. Organo
Bi(Ph)3organometaTrialkyl andthey have compoundsexplosion hform any were achievtitanium tewas difcurated into tpost-deposO3 Bi(Ph)3 fcycles weregrown [202bismuth tanthe solutionprecursors coating deepeting for SBT compo[203].
BismuthBi(CH2SiMe[204]. Fig. Bi(CH2SiMeparison. In a precursorof 0.4 A/cyctime above350 C showof the precdeposition the ratio wbut annealilms showsome other
Above: Molecular structure of Bi(CH2SiMe3)3. Below: TG curves of Bi(iPr)3H2SiMe3)3 [204].
is(Bis(Trimethylsilyl)amido)Bismuth(III) as a precursor for
n the simplest alkylamides of bismuth like Bi(NMe2)3 are to be monomeric and quite volatile [217], and they haveentioned as possible MOCVD precursors though actual
has not been reported. Down sides of these compoundsering ALD usage are their light sensitivity, extremely highvity to air and moisture, and thermal instability. Comparedle alkylamides the silylamide Bi(N(SiMe3)2)3 is more stables sensitive [109]. TGA under normal pressure N2 atmosphere
a single step weight loss of 98% at 220270 C (Fig. 21). Thissuggests that Bi(N(SiMe3)2)3 is stable even up to 270 C, butowth experiments implied that decomposition takes placey at temperatures exceeding 200 C. The compound melts at37 C and was a solid at 120 C ALD source temperature used.owth of BiOx was achieved at very narrow temperature win-tween 190 and 200 C using Bi(N(SiMe3)2)3 and H2O as thesors. Uniform amorphous BiOx thin lms could be depositedrowth rates between 0.08 and 0.23 A/cycle, but the repro-lity of the process was poor [109]. The irreproducibility inary oxide growth was thought to be due to partial reductionuth to metallic form, leading to termination of the growth
Other peculiarities in the BiOx growth were also observed. revealed that the grown amorphous BiOx lms were granu-ile amorphous lms are usually smooth and featureless. Theesses have been reported for several bismuth oxide ande materials. These processes are collected into Table 4.ALD has been reported for eight bismuth precursors:, Bi(CH2SiMe3)3, Bi(N(SiMe3)2)3, Bi(thd)3, Bi(mmp)2,nd Bi(OCMe2iPr)3.
h chloride as precursor for ALD
volatile and thermally stable compound with a vapor 0.1 mbar at 173 C [214]. Reported source tempera-
have been 140 C [3]. In the reactions with (Et3Si)2Se2Te selenides and tellurides of bismuth are easily. Also Bi2S3 can be deposited using H2S as the sul-or [200]. With these reactants formation of bismuthes is favorable and this may be partially explained byis acid Bi3+ coming together with the soft Lewis base
Te2). The problem with BiCl3 is its inability to form lms. With water, O2 and O3 the lms formed are
BiOCl or otherwise heavily contaminated by chlorine
metallic compounds of bismuth as precursors for ALD
[201203] and Bi(CH2SiMe3)3 [204] are the only truellic bismuth compounds that have been used for ALD.
triaryl compounds of bismuth are well volatile andbeen used largely in MOCVD. Down sides of these
is their lack of reactivity against H2O, their potentialazard and harmfulness [216]. Though Bi(Ph)3 does notlm with H2O, BiTiO lms with Bi/Ti ratios up to 0.6ed when Bi(Ph)3H2O reaction cycles were mixed withtraisopropoxideH2O cycles. Further Bi incorporationlt, however [201]. In addition, bismuth was incorpo-he lms in a metallic form rather than as an oxide, andition annealing was needed to oxidize bismuth. Withorms dark non-uniform lms [202]. When Bi(Ph)3 + O3
mixed with Ti(OiPr)4 + H2O cycles BiTiO could be]. In liquid injection plasma enhanced ALD of strontiumtalate (SBT) Bi(Ph)3 has been used as a component in
[203]. Simultaneous injection of two or more metalcan, however, lead to composition nonuniformity whilep trench structures because the precursors are com-the same chemisorption sites. Nevertheless, uniformsition over a 5-in. wafer was reported by Shin et al.
silicates, BixSiyOz have been deposited from3)3 and O3 at temperatures between 200 and 450 C20 shows the structure and TGA curve measured for3)3. TG curve for Bi(iPr)3 is included for volatility com-the Bi(CH2SiMe3)3O3 process Bi(CH2SiMe3)3 acted as
for both bismuth and silicon. Constant growth ratele independent of the Bi(CH2SiMe3)3 precursor pulse
1 s was observed at 250350 C. Films deposited aboveed a clear thickness prole indicating decompositionursor. The Si/Bi ratio increased with the increasingtemperature but in the lms deposited at 250350 Cas around 2. The as-deposited lms were amorphousng either in N2 or O2 atmosphere resulted in crystallineing the presence of orthorhombic Bi2SiO5 phase and
bismuth silicates.
Fig. 20. and Bi(C
5.3. TrALD
Eveknownbeen musage considsensitito simpand lesshowsresult ALD gralread1331ALD Grdow beprecurwith gducibithe binof bism[205]. FESEMlar wh
T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322 3313
Fig. 21. AbovBi(N(SiMe3)2)3
granular mobismuth to aby its reoxidno metallic
Depositicomponentdeposition.other metaBiTaO, Smade by mSrTa2(OEt)1growth wasby the thermticomponenvarying the
When Bicycle ratio the Bi/Ta rawere deposSrTa2(OEt)1strontium dthe Sr/Ta raelectric layewas realizealways som
. 22.
rem SBTing Batio :1 cyith srriedation/Si-s
in limFig
70 and190 nm
Mixin 1:1 r0.33. 7lms wwas capolarizPt/SiO2[70].
Mae: Crystal structure of Bi(N(SiMe3)2)3. Below: TG curve of(N2, 1 atm, 10 C/min, 10 mg sample) [109].
rphology was thought to originate from a reduction ofn elemental form at some point of the growth, followedation during the further deposition cycles as there was
bismuth in the lms [205].on of tantalum and titanium containing multi-
lms was found more reproducible than BiOx It is likely that mixing the BiOx growth cycles with thel oxide cycles stabilizes the growth. MulticomponentrBiTaO [109] and Bi4Ti3O12 [205] lms wereixing Bi(N(SiMe3)2)3H2O cycles with Ta(OEt)5H2O,0(dmae)2H2O and Ti(OMe)4H2O cycles. The lm
similarly found to be limited to 190200 C and belowal stability of Bi(N(SiMe3)2)3. In the case of all the mul-t oxides the metal component ratios could be varied by
binary oxide cycle ratios.TaO was deposited at 190 C using a xed BiOx/Ta2O5of 2:1 BiTaO the growth rate was 0.33 A/cycle andtio was around 1.1. Amorphous SrBiTaO (SBT) lmsited by mixing two Bi(N(SiMe3)2)3H2O cycles with one0(dmae)2H2O cycle. The lms deposited were foundecient with a stoichiometry of Sr0.8Bi2.1Ta2Ox thoughtio in the precursor was xed to 1:2. The desired ferro-red perovskite, i.e. the so-called Aurivillius phase SBT
d after careful annealing treatments but still there wase pyrochlore phase present in the lms. Permittivity of
thermal stathat togethof 190200muth oxidelikely that w
5.4. Bismutfor ALD
Several known andcompoundsalkoxides,
5.4.1. -DikSeveral
volatile, suwith H2O. Iforms loosemonomericshows 97%evaporationat temperawindow wa0.1 A/cycle,the ligands
5.4.2. BismBismuth
cursors. Thhave very pBi(OtBu)3 m[206,219,22tiary alkoxyor secondawhich showmary or seSawyer-Tower QV loop of a Pt/SBT/Pt capacitor structure [70].
anent polarization of 0.75 C/cm2 were measured for lm (Fig. 22) [70].i(N(SiMe3)2)3H2O cycles with Ti(OMe)4H2O cycles
resulted in smooth amorphous lms with a Bi/Ti ratio ofcle ratio was needed for obtaining phase-pure Bi4Ti3O12ufcient Bi content. Crystallization of the BiTiO lms
out at 500750 C. Ferroelectric lms with remanents up to 0.5 C/cm2 (51 nm lm) were prepared onubstrates by annealing rst at 500 C and then at 600 C
itations of the Bi(N(SiMe3)2)3 precursor are its lowbility and the high evaporation temperature neededer limit the growth temperature to very narrow rangeC. The tendency of the precursor to deposit binary bis-
unreliably is a further limitation. In addition, it is alsoith O3 there would be silicon incorporation in the lms.
h compounds with oxygen based ligands as precursors
bismuth compounds with oxygen based ligands are their thermal properties have been studied. Volatile
can be found in all the common compound groups i.e.-diketonates and carboxylates.
etonates of bismuth-diketonates of bismuth are known [218]. They arefciently thermally stable compounds and they reactn ALD only Bi(thd)3 has been used [206208]. Bi(thd)3
dimers in the solid state (Fig. 23) and is most likely in the gas phase. Bi(thd)3 melts at 130134 C and TGA
weight loss in a single step at 160300 C indicating (Fig. 24). Bi2O3 was deposited from Bi(thd)3 and H2Otures between 200 and 350 C. The ALD temperature
s found at 270300 C but the growth rate was low,
and believed to be due to the large steric hindrance of.
uth alkoxides alkoxides form an interesting group of bismuth pre-e simplest alkoxides Bi(OMe)3, Bi(OEt)3, Bi(OiPr)3oor volatility but even the smallest tertiary alkoxideay be sublimed cleanly already at 80100 C/0.05 mbar0]. It seems that in general bismuth alkoxides with ter-
ligands have good volatility while those with primaryry alkoxy ligands do not. There are several examples
that the structures of bismuth alkoxides with pri-condary alkoxy ligands are oligomeric while the ones
3314 T. Hatanp et al. / Coordination Chemistry Reviews 257 (2013) 3297 3322
Fig. 23. Molec
Fig. 24. TG curylates Bi(O2Ct
Society of Che
with tertiarof Bi(OtBu)muth alkoxdecrease wivolatile as cstability of a compounbe more stalower the treactive wi
Bi(OtBu)of Bi2O3 usibut the pre(mp. 139 Cimpossible.
Bi(OCMedetail [206]40 C) at thethe Bi(OCMa growth raat this temdecomposit240 C the ggrown at 1ular structure of Bi(thd)3. Loose dimers are formed in solid state.ves of different bismuth alkoxides, -diketonate Bi(thd)3 and carbox-Bu)3 and Bi(O2CMe)3 [206]. Reproduced by permission of The Royalmistry.
y alkoxy ligands are monomeric [219,221]. Structures3 and Bi(OiPr3)3 are shown in Fig. 25. Among the bis-ides with tertiary alkoxy ligands the volatility seems toth increasing molecular weight i.e. Bi(OtBu)3 is the mostan be seen from TG curves shown in Fig. 24. ThermalBi(OtBu)3 however is not good and it was found thatd with a bit larger alkoxy ligandOCMe2iPr seems toble. Using even larger tertiary alkoxy ligands seems tohermal stability. Bismuth alkoxides are in general wellth water though the reaction is not vigorous.
has been only tested in a preliminary manner for ALDng H2O as the oxygen precursor [206]. Bi2O3 lm grewcursor, solid at the evaporation temperature of 70 C), seemed to suffer from ageing making repeatable ALD
2iPr)3 is more stable alkoxide and was studied in more
. Bi(OCMe2iPr)3 has the advantage of being a liquid (mp. temperature used for its evaporation (85 C). At 150 Ce2iPr)3H2O process was found to produce Bi2O3 withte of 0.38 A/cycle. ALD type growth was veried fullyperature but it was also conrmed that no signicantion took place even at temperatures of 200250 C. Atrowth rate was found to be 0.30 A/cycle [200,222]. Films50 C were amorphous as-deposited while at growth
Fig. 25. Molcuomitted for cla
temperaturmuth oxidethe lms wotherwise v
Bi(OCMeSrBi2Ta2O9were Ti(OMTi(OMe)4Hlm. Bi4Ti3after anneaprecursors H2O.
Bi(mmpdonor functity and candeposit BixTIn three stuof a solventcycles wereand the ALreported atprocess revat 250 C [2heated sourlar structures of Bi(OtBu)3 (above) and Bi(OCiPr3)3 (hydrogen atomsrity) [206].
es above 200 C a thus far unidentied crystalline bis- phase was observed. TOF-ERDA results revealed thatere oxygen rich, probably due to OH residues, and hadery low impurity levels.2iPr)3 has also been studied in ALD of Bi4Ti3O12 and
at 240 C [200,222]. For Bi4Ti3O12 the other precursorse)4 and H2O. Mixing Bi(OCMe2iPr)3H2O cycles with2O cycles in a 3:1 ratio gave a Bi/Ti ratio of 1.47 in the
O12, an orthorhombic ferroelectric phase was observedling the lms in air at 700 C. For SrBi2Ta2O9 the otherused were Ta(OEt)5, Sr(iPrC5H2)2 or Sr(tBu3C5H2)2, and
)3 i.e. Bi(OCMe2CH2OMe)3 is an alkoxide with a tertiaryionalized ligand. Th