Combinatorial approach for ferroelectric materiallibraries prepared by liquid source mistedchemical deposition methodKi Woong Kim*, Min Ku Jeon*, Kwang Seok Oh*, Tai Suk Kim*, Yun Seok Kim†, and Seong Ihl Woo*‡
*Department of Chemical and Biomolecular Engineering and Center for Ultramicrochemical Process Systems, and †Department of Materials Science andEngineering, Korea Advanced Institute of Science and Technology, 373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, Republic of Korea
Communicated by Gabor A. Somorjai, University of California, Berkeley, CA, November 24, 2006 (received for review August 24, 2006)
Combinatorial approach for discovering novel functional materialsin the huge diversity of chemical composition and processingconditions has become more important for breakthrough in thinfilm electronic and energy-conversion devices. The efficiency ofcombinatorial method depends on the preparation of a reliablehigh-density composition thin-film library. The physico-chemicalproperties of each sample on the library should be similar to thoseof the corresponding samples prepared by one-by-one conven-tional methods. We successfully developed the combinatorial liq-uid source misted chemical deposition (LSMCD) method and dem-onstrated its validity in screening the chemical composition ofBi3.75LaxCe0.25-xTi3O12 (BLCT) for high remanent polarization (Pr).LSMCD is a cheap promising combinatorial screening tool. It cancontrol the composition up to ppm level and produce homoge-neous multicomponent library. LSMCD method allows us to pre-pare BLCT thin-film library at the variation of 0.4 mol% of La.Maximum 2Pr is 35 �C/cm�2 at x � 0.21. The intensity of (117) XRDpeak is quantitatively related to 2Pr. Newly developed scanningpiezoelectric deformation measurement for nano-sized samplesusing scanning probe microscope (SPM) is also found out to bereliable for determining the relative ranking of Pr value rapidly.
Combinatorial methods consist of library design, preparationof library, evaluation of its functional properties, and its
physicochemical characteristics. These methods are consideredto be very efficient and quickly provide us the quantitativeactivity structure relationship to establish the criteria for de-signing novel functional materials, which have been extensivelyused for the discovery of functional electronic thin film mate-rials, novel polymers, and catalysts since the mid-1990s (1–10).
Thin-film combinatorial libraries have been synthesizedmainly by RF-sputtering, pulsed laser ablation, and chemicalvapor deposition. Thin-film libraries containing various compo-sitions were fabricated by adopting multimask techniques (3–6).Ten binary masks or five quaternary masks should be used toprepare 1,024 (210) different samples on a wafer. These maskoperations complicated the operation of the instrument as wellas creating the contamination problems. These problems weresolved by adopting only two masks, one x–y movable shutter andone shadow mask (7–11). The x–y movable shutter moves at aconstant velocity in a linear or stepwise way to control thedeposition time in one direction of wafer. The shadow maskcontaining regularly spaced holes is located just below themoving shutter, which fabricates a high-density discrete libraryin a very simple manner. Recently, off-axis codeposition methodwas reported to prepare thin-film metal alloy library (12, 13). Bycontrolling the distance between the target guns and the sub-strate, multitarget materials can be deposited on the substratewith concentration gradient simultaneously. This method doesnot need to use a movable shutter or many masks. Only shadowmask is necessary when the fabrication of discrete library is
preferred. However, this method cannot control the thickness ofeach sample on the library and is not suitable when the physi-cochemical properties vary with thickness.
The library fabrication methods discussed above need toprepare single- or multicomponent target materials, which issometimes very expensive, and it can be difficult to obtain auniform composition. It is also very difficult to control thedeposition rate of each component in multicomponent targetarising from the difference in its inherent sputtering property.Sputtering rate depends on many variables, so it is too difficultto control precisely. Therefore, accurate control of compositionwithin �3% in the library cannot be obtained. To overcomethese problems, some research groups applied the wet processincluding spray pyrolysis and ink-jet technique for combinatoriallibrary synthesis. However, these processes did not accomplish afinely compositional tuned thin-film library because of theirinherent weakness (14–17).
Therefore, in our study, liquid source misted chemical depo-sition (LSMCD) method was developed to prepare a thin-filmlibrary. Adequate amount of metal precursor compounds wasdissolved in suitable solvent. This mixed solution is equivalent tothe target in the sputtering method. Composition control is moreaccurate with LSMCD than with conventional sputtering, be-cause exact weighing can be obtained even in ppm level. Fur-thermore, the cost of metal precursor solution (a few dollars) isjust the cost of chemicals and much cheaper than target (averagecost for 4-inch target is �$5,000). Hundreds of targets arerequired for the fabrication of a very diverse library. Thematerial cost to fabricate a thin-film library with LSMCD isnegligible compared with conventional sputtering method. Ul-trafine mist is generated with nebulizer (frequency: 1.65 MHz)and transported by Ar flow to the substrate. Suitable thermaltreatments are required to obtain multicomponent metal alloy ormetal oxide thin-film (18–23). The variation of deposition timewith x–y movable shutter will generate the concentration gradi-ent in a similar manner to combinatorial sputtering. It is possibleto fabricate homogeneous liquid-phase combinatorial library byliquid-phase mixing in LSMCD, whereas in sputtering method,each component of library have to be mixed by solid-statediffusion, resulting in the formation of more homogeneousmulticomponent compound with LSMCD. It can be concluded
Author contributions: K.W.K. and S.I.W. designed research; K.W.K. and T.S.K. performedresearch; K.W.K., K.S.O., and Y.S.K. contributed new reagents/analytic tools; K.W.K., M.K.J.,and S.I.W. analyzed data; and K.W.K. and S.I.W. wrote the paper.
The authors declare no conflict of interest.
Freely available online through the PNAS open access option.
that LSMCD enables us to fabricate a thin-film library with moreaccurate control of composition at a low cost. The price of theLSMCD instrument is also much cheaper than a conventionalPVD instrument. These advantages of LSMCD over the sput-tering method will be demonstrated in fabricating a multicomponent ferroelectric oxide thin-film library for the optimizationof its electrical properties as follows.
(24) and (Bi,La)4Ti3O12 (BLT) (25), have drawn an interest forferroelectric random-access memory because of the fatigue-freeproperties over 1010 read/write cycles, unlike Pb(Zr,Ti)O3 (PZT)(26, 27). However, these materials have smaller remanent po-
larization (Pr) than PZT. Numerous attempts have been made tosubstitute lanthanide group atoms for Bi site in Bi4Ti3O12 toincrease Pr. It is well known that Pr depends on the chemicalcomposition (28–32). Woo et al. first presented that the Ce-substituted Bi4Ti3O12 (Bi4-xCexTi3O12, BCT) fabricated byLSMCD exhibited large Pr (2Pr � 20 �Ccm�2 at 15 V, x � 0.75)and superior fatigue endurance over 1010 read/write cycles.Hence, it is quite reasonable to substitute Bi with both La andCe for ferroelectric materials with high Pr. We prepared aBi3.75LaxCe0.25-xTi3O12 (BLCT, 0 � x � 0.25) array with LSMCDto optimize the chemical composition for high Pr. Bi/(La�Ce)was fixed to 3.75/0.25 because our structural refinement result
Table 1. The chemical composition and thickness of BLCT thin-film 8 � 8 samples on a library annealed at 700°C for 1 h in furnaceunder oxygen atmosphere
0/0.25
(0/0.25)
184nm
0.022/0.228
(0.02/0.23)
193nm
0.043/0.207
(0.047/0.203)
196nm
0.065/0.185
(0.065/0.185)
198nm
0.087/0.163
(0.09/0.16)
194nm
0.108/0.142
(0.112/0.148)
193nm
0.13/0.12
(0.13/0.12)
190nm
0.152/0.098
(0.153/0.097)
187nm
0.014/0.236
(0.014/0.236)
186nm
0.036/0.214
(0.036/0.214)
195nm
0.057/0.193
(0.057/0.193)
200nm
0.079/0.171
(0.079/0.171)
201nm
0.101/0.149
(0.103/0.147)
196nm
0.122/0.128
(0.118/0.132)
196nm
0.144/0.106
(0.146/0.104)
194nm
0.166/0.084
(0.165/0.085)
192nm
0.028/0.222
(0.028/0.222)
192nm
0.05/0.2
(0.052/0.198)
196nm
0.071/0.179
(0.076/0.174)
200nm
0.093/0.157
(0.093/0.157)
206nm
0.115/0.135
(0.115/0.135)
200nm
0.136/0.114
(0.136/0.114)
197nm
0.158/0.092
(0.158/0.092)
192nm
0.18/0.07
(0.183/0.067)
190nm
0.042/0.208
(0.042/0.208)
194nm
0.064/0.186
(0.068/0.182)
200nm
0.085/0.165
(0.088/0.162)
204nm
0.107/0.143
(0.111/0.139)
201nm
0.129/0.121
(0.129/0.121)
202nm
0.15/0.1
(0.151/0.099)
197nm
0.172/0.078
(0.172/0.078)
195nm
0.194/0.056
(0.195/0.055)
197nm
0.056/0.194
(0.056/0.194)
195nm
0.078/0.172
(0.078/0.172)
200nm
0.099/0.151
(0.095/0.155)
205nm
0.121/0.129
(0.121/0.129)
207nm
0.143/0.107
(0.143/0.107)
203nm
0.164/0.086
(0.162/0.088)
197nm
0.186/0.064
(0.189/0.061)
196nm
0.208/0.042
(0.201/0.049)
194nm
0.07/0.18
(0.07/0.18)
195nm
0.092/0.158
(0.089/0.161)
201nm
0.114/0.136
(0.114/0.136)
208nm
0.135/0.115
(0.133/0.117)
204nm
0.157/0.093
(0.159/0.091)
202nm
0.178/0.072
(0.176/0.074)
196nm
0.2/0.05
(0.2/0.05)
194nm
0.222/0.028
(0.223/0.027)
191nm
0.084/0.166
(0.084/0.166)
196nm
0.106/0.144
(0.109/0.141)
200nm
0.128/0.122
(0.128/0.122)
205nm
0.149/0.101
(0.147/0.103)
200nm
0.171/0.079
(0.171/0.079)
200nm
0.193/0.057
(0.194/0.056)
195nm
0.214/0.036
(0.215/0.035)
191nm
0.236/0.014
(0.231/0.019)
192nm
0.098/0.152
(0.098/0.152)
188nm
0.12/0.13
(0.12/0.13)
193nm
0.142/0.108
(0.143/0.107)
194nm
0.163/0.087
(0.16/0.09)
196nm
0.185/0.065
(0.184/0.066)
195nm
0.207/0.043
(0.207/0.043)
192nm
0.228/0.022
(0.228/0.022)
189nm
0.25/0
(0.25/0)
188nm
The top number is the chemical composition Ce�La designed to prepare. The blanket value is chemical composition (Ce�La) measured by wavelength dispersivespectroscopy, and the bottom number is the thickness measured by SEM cross-section image. The location of 8 � 8 samples on the wafer is equivalent to thelocation of the square in the table.
Kim et al. PNAS � January 23, 2007 � vol. 104 � no. 4 � 1135
CHEM
ISTR
Y
Dow
nloa
ded
by g
uest
on
Sep
tem
ber
18, 2
020
with BLT powder showed that 2Pr was maximum when the Bi/Lawas 3.75/0.25 (33–35).
Fast and reliable measurement of remanent polarization isrequired for a continuous-gradient library and �m-sized sam-ples, because the conventional capacitance–voltage measure-ment method is slow and requires electronic wirings. A scanningprobe microscope can apply the electric field into the sample onthe library using a sharpened tip as movable electrode andinduced the local deformation of samples. The correlationbetween the deformation length and Pr will be discussed.
Results and DiscussionA BLCT (0 � x � 0.25) thin-film library was prepared byLSMCD equipped automated shutter (see Materials and Meth-ods).
The chemical compositions of BLCT samples on the librarymeasured by wavelength dispersive spectroscopy and thicknessof each sample on the library measured by SEM cross-sectionimage are shown in Table 1. The real location of each sample isthe same as the location of Table 1; 0.4 mole % of compositioncontrol was demonstrated, as was the uniform thickness, 200nm � 10 nm.
Microbeam XRD (Bruker D8Discover GADDS CS, CuK�
radiation) was used to analyze the crystallographic phase and thepreferred orientation of as-prepared thin-film samples on thelibrary. As shown in Fig. 1A, regardless of La content, all sampleson the library did not contain any peaks of Bi2O3 (2� � 28°) aswell as the other mono- or binary oxides. The intensities of (00n)peaks arising from the extent of crystal growth in the c-axis didnot change much with La content. However, the intensity of(117) diffraction peak normalized by that of Pt (111) peakincreased with the increase of La content and reached themaximum at x � 0.21. In the region of x � 0.21, there was anabrupt decrease in this value (Fig. 1B). The crystal growth in thea and b axes is related to the intensity of the (117) peak.
It was reported that Ps is only 4�Ccm�2 along the c axis and50�Ccm�2 along the a or b axis measured with single crystalBi4Ti3O12 (36, 37). It was also reported that BLT thin film grownepitaxially in the a axis shows a Ps value of 50�Ccm�2 (39).
Fig. 2 shows the grain size as a function of La content. Asshown in Fig. 2, grain size increased with the increase of Lacontent. This can improve the ferroelectricity and piezoelectric-ity of thin film. However, when the grain size is larger than 0.5�m, the leakage current density increased from 10�6 Acm�2 to10�4 Acm�2 in the region of La content greater than x � 0.2[supporting information (SI) Fig. 7].
2Pr values of 64 samples on the library were measured by thepolarization-electric field (P–E) hysteresis loop at the appliedvoltage of 10 V as shown in Fig. 3. Twenty Pt top electrodes(diameter: 200 �m) were deposited on each sample with athickness of 100 nm to form a capacitor structure. As shown inFig. 3, the remanent polarization increased with La content (x).Pr increased slightly between x � 0 and x � 0.10, but increasedrapidly between x � 0.1 and 0.21 and decreased abruptly whenx � 0.23. The maximum 2Pr (35�Ccm�2) was obtained at x �0.21. Comparing Fig. 3 with Fig. 1B, it could be suggested thatthe remanent polarization of BLCT thin film (x � 0.25) bequantitatively related to the intensity of normalized (117) peakof BLCT. As shown in Fig. 4, BLCT shows excellent fatigueendurance at 5 V over 1010 cycles similar to other bismuth-layered ferroelectrics. This is due to the increase of structuralstability induced by decrease in oxygen vacancies (27). Substi-tution of La and Ce by Bi decreased oxygen vacancies.
Four different BLCT thin films (2 cm � 2 cm) were fabricatedat the same deposition condition of combinatorial experimentsdiscussed above by varying the La content (x � 0.1, 0.12, 0.18,and 0.194) to validate the results obtained by 8 � 8 thin filmsamples on the combinatorial library. As shown in Fig. 5A, the
change in 2Pr of BLCT individually prepared shows a similartrend to that of BLCT prepared combinatorially, even though2Pr values of individual BLCT samples are slightly lower thanthose of combinatorial array. Such an offset was due to thethickness difference between samples on the library (200 nm)and individual (240 nm) samples. Even though we need theidentical deposition condition for samples on the library andindividual samples, it was found out that the deposition rate forsamples on the library is 20% lower than that for individualsamples, because the shadow mask and moving mask used forsamples on the library hindered mist from approaching to thesubstrate. As shown in Fig. 5 B and C, the thicker individualferroelectric samples have larger coercive field strength. As thecoercive field strength increases, higher electric field strength isnecessary to switch the dipole moment. Therefore, individualsample shows a little bit lower remanent polarization thansample on the library at the same chemical composition underthe same electric field (500 kV/cm).
This result indicated that the preparation of combinatorialarray by LSMCD is reliable for the rapid screening of variousthin films having various compositions.
A high-density library (2,500 samples per wafer) can beprepared with existing technology, but the evaluation of their
Fig. 1. Structural analysis by using microbeam XRD. (A) The microbeam XRDspectra of BLCT thin-film samples on the library as a function of La content, x.These spectra were scanned in �-2� type in the range of 2� between 17 and 53°under Cu K� radiation (� � 1.5405 Å) operated at 40 mA and 40 kV. (B) Theratio of instensity of (117) diffraction peak to that of Pt (111) diffraction peakwas plotted as a function of La content.
1136 � www.pnas.org�cgi�doi�10.1073�pnas.0610146104 Kim et al.
functional properties and physico-chemical characterization can-not be performed rapidly as long as conventional procedures areused. It is quite desirable to develop new methods to measure thefunctional and physico-chemical properties of thin film sampleson the library in a very short time. P–E hysteresis characteristicsof BLCT thin film samples on the library were measured withRT66A ferroelectric tester one by one for each sample in thisstudy, which is not suitable for high-density library because ittakes long time. Here, we report the rapid indirect measurementof remanent polarization by measuring the piezo-electricity witha scanning probe microscope (SPM) (38–41). A metallic atomicforce microscope (AFM) tip was used as a movable top elec-trode. The electric field was applied between the AFM tip andbottom electrode to measure the topographic deformation. Alaser beam was focused on the upper side of the cantilever andphoto-diode to detect an angle of reflection from the upper side
of the cantilever as shown in Fig. 6A. This enables us to detectthe topographic deformation up to 0.01 nm. Local piezo-response on the grain or domain of each samples on the librarycan be detected nondestructively just by scanning the AFM tipusing XY manipulator, which will decrease the measurementtime extensively compared with conventional measurement ofthe P–E hysteresis loop. Tip bias was continuously changed from�10 V to �10 V, and its topographic deformation of typicalBLCT film library is shown as a curve of butterfly shape as shownin Fig. 6B. Fig. 6C shows the relationship between topographicdeformation (H � H1 � H2) and remanent polarization (2Pr)in BLCT array as a function of La content (x). H1 and H2 arelift heights with a tip bias, �10 V and �10 V, respectively.Piezoelectricity and ferroelectricity show the similar tendency asa function of La content. Maximum H was 0.9 nm at x � 0.2,indicating that relative ranking of Pr can be estimated rapidly onthe basis of topographic deformation. This method can be usedreliably for measuring Pr of discrete or continuous high-densityferroelectric combinatorial array indirectly.
In conclusion, we fabricated a BLCT thin-film library by usingLSMCD equipped with an automated shutter, and mapped thestructure and remanent polarization as a function of La content.This library-synthetic method allows us to control the chemicalcomposition accurately (ppm level in the library). Structuralanalysis by microbeam XRD clearly shows that (117) crystallinegrowth has an important role in improving the remanent polar-ization. Newly developed scanning piezoelectric deformationmeasurement is found out to be reliable for determining therelative ranking of Pr value rapidly.
Materials and MethodsPreparation of Precursor Solutions. For the preparation of BLT(Bi/La � 3.75/0.25, Bi 20% excess) solution, 2.465 g ofBi(NO3)3�6H2O and 0.122 g of La(NO3)3�6H2O were dissolved in10 ml of 2-methoxyethanol (Solution 1). One milliliter of Tiisopropoxide was dissolved in 9 ml of 2-methoxyethanol (Solu-tion 2). Solution 1 and 2 were mixed. After the solution mixing,10 ml of 2-methoxyethanol and 1 ml of 2-ethylhexanoic acid wereadded and stirred for 30 min. In preparing BCT (Bi/Ce �3.75/0.25, Bi 20% excess), Ce nitrate was added instead of Lanitrate, and the other conditions were the same as those ofpreparing BLT solution.
Preparation of BLCT Thin Film Library by LSMCD Technique. Schematicdiagrams of LSMCD are shown in SI Fig. 8A.
We generated the mists of BLT and BCT metal precursor
Fig. 2. The surface morphology of BLCT thin films as a function of La content.x � 0 (A), x � 0.08 (B), x � 0.17 (C), and x � 0.25 (D) in BLCT combinatorial array,where x is La content.
Fig. 3. The characterization of P–E characteristic using a ferroelectric tester(RT66A). The plot of 2Pr of BLCT thin-film samples on the library as a functionof La content, x. Pt top electrodes of 200 �m diameter were deposited on thethin-film library by using RF-sputtering system to fabricate the capacitorstructure for measuring the remanent polarization. The thickness of topelectrode was �100 nm.
Fig. 4. Fatigue endurance behavior of BLCT thin film samples (x � 0.014,0.172, 0.194). The frequency was 1 MHz, and applied voltage was � 5 V. Onlythese compositions were tested because of long test time.
Kim et al. PNAS � January 23, 2007 � vol. 104 � no. 4 � 1137
solution by using an ultrasonic nebulizer. First, BCT mist wastransported to the top of the computer-operated shutter by an Arflow of 500 sccm. Below the shutter, shadow mask containing
8 � 8 holes of 3 mm in diameter was intimately contacted on topof the wafer. BLT mist was deposited on the wafer in a stepwisefashion (deposition time: 20 sec/step) by moving shutter inx-direction. Linear variation of deposition time generated alinear concentration gradient of BLT in x-direction. Then, BLT
Fig. 5. Confirmation of the reliability of combinatorial experimentation byusing LSMCD process. (A) The comparison of 2Pr of BLCT samples deposited bycombinatorial experimentation with that of BLCT samples deposited individ-ually (one sample at one time). (B) P–E hysteresis loop of BLCT samples chosenamong the BLCT library. (C) P–E hysteresis loop of BLCT fabricated individuallyat the same condition.
Fig. 6. Mapping of topographic deformation using the SPM. (A) The sche-matic drawing of measuring topographic deformation by using SPM. (B) Thetopographic deformation of samples detected by SPM. Detection time was 1 sper point and DC voltage was changed continuously from �10 V and 10 V. A2 �m � 2 �m region of each library was scanned with a gold-coated AFM tip.(C) The plot of topographical deformation (H � H1 � H2) and 2Pr of BLCTthin-film samples on the library as a function of La content.
1138 � www.pnas.org�cgi�doi�10.1073�pnas.0610146104 Kim et al.
Dow
nloa
ded
by g
uest
on
Sep
tem
ber
18, 2
020
mist was deposited in the same manner except for the movementof shutter in the opposite direction. After this procedure, thesubstrate holder was rotated by 90°; BLT and BCT mists weredeposited in the same manner except for the deposition time (13sec/step) as explained in SI Fig. 8B. The above methods completea single cycle of deposition. We were then able to generate theliquid phase library of same volume but with different Ce/Laratio. After 3 cycles, we prepared the BLCT thin-film library with200 nm thickness after thermal treatments at 230°C for 2 min, at400°C for 10 min in air, and at 700°C for 1 h under O2atmosphere.
Characterization of BLCT Thin Film Library Prepared by LSMCD Tech-nique. To analyze the new crystallographic phase and preferredorientation of as-prepared thin-film library, we can use themicrobeam XRD (D8 DISCOVER with GADDS for combina-torial screening by Bruker-AXS) with small beam size (�500�m) and xyz manipulator controlled by computer software. Theangle between detector and substrate was 20° and the anglebetween x-ray gun and substrate was 15°. The XRD mapping wascarried out in �-2� scans under Cu K� radiation (� � 1.5405 Å)operated at 40 mA and 40 kV.
The chemical composition of prepared thin-film library wasobtained by wavelength dispersive spectroscopy (Microspec3-PC). Surface morphology of thin-film library was analyzed bySEM (Philips 533M).
To measure the electrical properties, Pt electrodes with 200�m diameter were deposited on the thin-film library by using theRF-magnetron sputtering system. P–E hysteresis was measuredby using a Radiant Technology RT66A ferroelectric tester at 10V. The leakage current characteristics were measured by usingprogrammable Kethley 617 electrometer with the condition of0.05 V of step voltage and 0.1-sec delay time.
A scanning probe microscope (SPA 400 SPI 3800N, SeikoInstrument Korea, Seoul, South Korea) was used to measure orscan the topographical change and surface roughness of thin filmarray. When measuring the topographical change, the detectiontime was 1 sec per point, and voltage was changed from �10 Vand 10 V. The scan region was 2 �m � 2 �m, and a gold-coatedAFM tip (SI-DF3-A, C � 2.3 N/m) was used as a movable topelectrode. The measurement was carried out �20 times perpoint, and the induced topographical change was averaged.
This research was funded by Center for Ultramicrochemical Process Sys-tems (CUPS) sponsored by Korea Science and Engineering Foundation.
1. Jandeleit B, Schaefer DJ, Powers TS, Turner HW, Weinberg WH (1999) AngewChem Int Ed 38:2494–2532.
2. Yoo YK, Xiang XD (2002) J Phys Condens Matter 14:R49–R78.3. Xiang X-D, Sun A, Briceno G., Lou Y, Wang KA, Chang H, Wallace-Freeman
12. Takeuchi I, Famodu OO, Read JC, Aronova MA, Chang K-S, Craciunescu C,Lofland SE, Wuttig M, Wellstood FC, Knauss L, et al. (2003) Nat Mater2:180–184.
13. Smith RC, Hoilien N, Roberts J, Campbell SA, Gladfelter WL (2002) ChemMater 14:474–476.
14. Luo ZL, Geng B, Bao J, Gao C (2006) J Comb Chem 7:942–946.15. Wang J, Evans JRG (2005) J Mater Res 20:2733–2740.16. Okamura S, Takeuchi R, Shiosaki T (2002) Jpn J Appl Phys 41: 6714–6717.17. Li GZ, Yu M, Wang ZL, Lin J, Wang RS, Fang J (2006) J Nanosci Nanotechnol
6:1416–1422.18. Hoffman MM (1995) Integr Ferroelectr 10:39–53.19. Chung HJ, Kim JH, Moon WS, Park SB, Hwang CS, Lee MY, Woo SI (1997)
Integr Ferroelectr 12:185–197.
20. Woo SI (2001) Materials 12:4–13.21. Chung HJ, Kim JH, Woo SI (2001) Chem Mater 13:1441–1443.22. Kim KW, Woo SI, Choi KH, Han KS, Park YJ (2003) Solid State Ionics
159:25–34.23. Kim KW, Lee SW, Han KS, Chung HJ, Woo SI (2003) Electrochim Acta
48:4223–4231.24. Araujo CA, Cuchiaro JD, McMillan LD, Scott MC, Scott JF (1995) Nature
374:627–629.25. Park BH, Kang BS, Bu SD, Noh NW, Jo W (1999) Nature 401:682–684.26. Scott JF, Araujo CA (1989) Science 246:1400–1405.27. Vijay DP, Desu SB (1993) J Electrochem Soc 140:2640–2645.28. Kojima T, Sakai T, Watanabe T, Funakubo H, Saito K, Osada M (2002) Appl
Phys Lett 80:2746–2748.29. Chon U, Kim K-B, Jang HM, Yi G-C (2001) Appl Phys Lett 79:3137–3139.30. Watanabe T, Funakubo H, Osada M, Noguchi Y, Miyayama M (2002) Appl
Phys Lett 80:100–102.31. Uchida H, Yoshikawa H, Okada I, Matsuda H, Iijima T, Watanabe T, Kojima
T, Funakubo H (2002) Appl Phys Lett 81:2229–2231.32. Kim KT, Kim CI, Kang DH, Shim IW (2002) Thin Solid Films 422:230–234.33. Jeon MK, Kim Y-I, Sohn JM, Woo SI (2004) J Phys D 37:2588–2592.34. Jeon MK, Kim Y-I, Nahm S-H, Woo SI (2004) J Kor Phys Soc 45:1240–
1243.35. Jeon MK, Kim Y-I, Nahm S-H, Woo SI (2005) J Phys Chem B 109:968–972.36. Xu Y (1991) in Ferroelectric Materials and Their Applications (Elsevier, Los
Angeles), pp 308–320.37. Lee HN, Hesse D, Zakharov N, Gosele U (2002) Science 296:2006–2009.38. Ahn CH, Tybell T, Antognazza L, Char K, Hammond RH, Beasley MR, Fischer
O, Triscone J-M (1997) Science 276:1100–1103.39. Paruch P, Tybell T, Triscone J-M (2001) Appl Phys Lett 79:530–532.40. Hong JI, Song HW, Hong SB, Shin HJ, No KS (2002) J Appl Phys 92:7434–7441.41. Ahn CH, Rabe KM, Triscone JM (2004) Science 303:488–491.
Kim et al. PNAS � January 23, 2007 � vol. 104 � no. 4 � 1139
