Journal of The Electrochemical Society, 154 �5� G110-G116 �2007�G110

Effects of N2-Based Annealing on the Reliability Characteristicsof Tungsten/La2O3/Silicon CapacitorsJoel Molina,z Kiichi Tachi, Kuniyuki Kakushima,* Parhat Ahmet,*Kazuo Tsutsui, Nobuyuki Sugii, Takeo Hattori,* and Hiroshi Iwai

Frontier Collaborative Research Center, Tokyo Institute of Technology, Midori-ku-Yokohama,Kanagawa 226-8502, Japan

In this paper, we report the effects of a N2 postmetallization annealing �PMA� on the reliability and charge-trapping characteristicsof tungsten/La2O3/silicon structures. The samples are stressed with a constant-voltage stressing �CVS� with substrate injection.After the stressing we found that the flatband voltage �VFB� of the samples positively shifts, indicating a net negative chargetrapped in the La2O3 and its interfaces. Samples with PMA exhibit less VFB shift after the stressing as compared to postdepositionannealing La2O3 films. Moreover, La2O3 with PMA at higher temperatures increases the endurance to VFB shift, but the equivalentoxide thickness �EOT� of the metal-insulator-semiconductor capacitors increases; thus, a trade-off between higher endurance tocharge trapping and low EOT is required. The reliability and electrical characteristics of La2O3 films are to some extent sensitiveto the annealing conditions and timing within the process flow.© 2007 The Electrochemical Society. �DOI: 10.1149/1.2712823� All rights reserved.

Manuscript submitted October 26, 2006; revised manuscript received December 13, 2006.Available electronically March 15, 2007.

0013-4651/2007/154�5�/G110/7/$20.00 © The Electrochemical Society

With the continuous scaling down of the complementary metaloxide semiconductor �CMOS� technology, the research of ultrathingate oxide films with higher dielectric constant has been greatlyadvanced. The main motivation behind the use of high-dielectric-constant materials for the gate of MOS transistors is the lower gate-leakage current density levels that can be obtained by using thickeroxide films without compromising their final equivalent oxide thick-ness �EOT� values. In this respect, the hafnium �Hf�-based family ofhigh-k materials has received a lot of interest and has been inten-sively studied during the past years and it is expected to be used inproduction in 2008. The immediate question then arises as to whichmaterial can be considered to replace Hf-based oxides in the future.According to the International Technology Roadmap for Semicon-ductors �ITRS�, lanthanum oxide �La2O3�, which is a member ofrare earth-based oxides, was classified into the next group of poten-tial candidates to succeed Hf-based oxides.1 Besides, the selection ofthe metal electrode material also plays an important role in the finalelectrical characteristics of the devices with high-k dielectrics be-cause poly-Si-based electrodes do not work below 1 nm EOT re-gime due to the gate polydepletion effect, so that introduction of asuitable metal gate is needed. Furthermore, because of the possibleformation and subsequent growth of the low-k interfacial layer at themetal-La2O3 interface after postmetallization annealing �PMA�,even at low annealing temperatures,2,3 the scaling of metal-La2O3stacked structures into a sub 1 nm regime is limited. Therefore, theintroduction of inert gate electrodes with suitable work functions forLa2O3-gated metal oxide semiconductor field-effect transistors�MOSFET� devices is of the utmost importance. Tungsten-gatedLa2O3 film results in lower EOT as compared to aluminum-gatedLa2O3,

4 so that the introduction of this metal atop of high-k La2O3film �already providing an advantage into sub 1 nm scaling� needs tobe evaluated for reliability purposes. Figure 1 shows gate-leakagecurrent Jg vs EOT for some metal/high-k gate stacks.

4-9 Presentresults for W-gated La2O3 stacks are also shown. From theseJg-EOT data, we found a significant difference in the resulting char-acteristics of La2O3 after postdeposition annealing �PDA� or PMA�PDA means that La2O3 was annealed right after its deposition on Siand before metallization�. From the set A samples, La2O3 after PMAproduces the lowest Jg level but its EOT gets slightly increased, sothat the sensitivity of La2O3 electrical characteristics with respect tothe annealing timing and conditions within the process flow can leadto further improvements in its final reliability. It is thought that PMA

* Electrochemical Society Active Member.z E-mail: jmolina@ae.titech.ac.jp

improves the physical quality of both silicon/high-k and metal/high-k interfaces simultaneously so that more reliable characteristics ofmetal insulator-semiconductor �MIS� devices after PMA can be ex-pected. In the present report, we evaluate the PMA effect on thereliability characteristics of MIS capacitors �MISCAPs� withW-gated La2O3 dielectric after constant positive voltage stressing.

Experimental

Two different deposition methods and annealing sequences forLa2O3 were used in order to evaluate the impact of the annealingtiming within the process flow and also the influence of in situmetallization for reliability characterization. The silicon substratedoping for all samples was between 1 and 6 � 1015 cm−3. Only dryN2-based annealing was performed for all the samples and they areschematically depicted in Fig. 2.

La2O3 deposition in vacuum with ex situ tungsten deposition.—Thin films of La2O3 with a physical thickness of 9.7 nm were de-posited on hydrofluoric acid �HF�-last or hydrogen-terminated

Figure 1. �Color online� Jg-EOT plot for metal/high-k stacks.4-9 Set A con-

tains PDA and PMA samples for the same La2O3 film, whereas set B con-tains only PMA samples for a thinner La2O3 film. From set A, the timing ofannealing after La O deposition can significantly change the EOT vs J

2 3 gcharacteristics.

G111Journal of The Electrochemical Society, 154 �5� G110-G116 �2007� G111

n-type silicon substrates by electron-beam evaporation using themolecular beam epitaxy MBE system �ANELVA I� at 300°C. Thepressure in the chamber during the deposition was around 1� 10−7 Pa. PDA and PMA were carried out in N2 ambient at 300°Cduring 5 min by rapid thermal annealing �ULVAC-MILA 3000�.Tungsten was used as the metal gate electrode for both PDA andPMA samples and it was sputter-deposited within 30 min after thedeposition of La2O3 to reduce moisture adsorption and any otherpossible contamination on these films at minimum. Tungsten depo-sition with a thickness of 50 nm was carried out by radio frequency

Figure 2. �Color online� Process flow for the samples used in this work. Twodifferent processing methods were used to evaluate the impact of in situmetallization on W–La2O3 reliability. In the first deposition of La2O3, threedifferent samples were obtained: as-deposited �no annealing�, PDA, andPMA, both at 300°C. In the second �and modified� deposition of La2O3, athinner dielectric film was deposited and three different samples were alsoobtained: as-deposited, PMA at 300°C, and PMA at 500°C.

�rf� magnetron sputtering at room temperature and we used a metalshadow mask with circular patterns for the gate area definition ofMIS capacitors.

La2O3 deposition in O2 ambient with in situ tungsten deposi-tion.— In order to reduce any possible oxygen loss during the La2O3deposition,10 La2O3 film for these samples was deposited at 300°Cwithin an oxygen flow of 1 � 10−5 Pa. Also, in situ sputtering oftungsten was done at 150 W in a contiguous chamber �within thesame MBE system� immediately after the dielectric deposition inorder to reduce the exposure of La2O3 surface to the environment.Because of this, the physical thickness of La2O3 cannot be measuredby conventional methods like ellipsometry but by transmission elec-tron microscopy or any other imaging technique. However, it isestimated that the physical thickness for this film was about 4 nm�more than half the physical thickness used for the first samples�.During the deposition of the metal an argon flow of 1.33 Pa wasused. Patterning of the gate electrodes was done by reactive-ionetching using SF6 plasma with a 30 W power. There wereSiO2-based spacers �200 nm in thickness� used around the gates ofthe devices in order to minimize La2O3 exposition to environment’smoisture. PMA was done in N2 ambient at 300 and 500°C during5 min. For all the samples, capacitance-voltage �C-V�, current-voltage �I-V�, and time-dependent dielectric breakdown �TDDB�characteristics were obtained with a precision LCR meter �Agilent4284A� and semiconductor parameter analyzer �Agilent 4156C�, re-spectively. The experimental procedures for both sets of samples areoutlined in Fig. 2.

Results and Discussion

VFB shift and lifetime projection for W–La2O3-gated MISCAPwith EOT � 1 nm.— From Fig. 1 and from set A samples �bothwith the same as-deposited La2O3 physical thickness�, PMA onW–La2O3 results in an increased EOT with a lower density of leak-age current. This difference in the EOT-Jg characteristic for La2O3after PDA and PMA samples comes after considering that a strongerdensification of the La2O3 film during PDA results in a strongerreduction of La2O3 thickness, thus decreasing the final EOT for thisparticular sample. In previous reports,11,12 the normalized physicalfilm thickness of La2O3 as a function of annealing temperature withdry-N2 ambient has been reported. There, one can find that the nor-malized thickness �normalization is performed with respect to thephysical film thickness of nonannealed film� is minimized at rela-tively low annealing temperatures, in which the densification pro-cess takes place. Figures 3 and 4 show the I-V and C-V character-istics of W–La2O3 stack after the process flow A in Fig. 2. In Fig. 3the gate current density Jg for all samples increases with appliedgate voltage Vg until the oxide film breaks down. La2O3 after PDAor PMA shows better electrical characteristics as compared to thenonannealed sample �as-deposited� in which the extrinsic break-down mode for La2O3 is observed. For both PDA and PMAsamples, different carrier conduction mechanisms can be observed atlow- and high-electric field regimes; however, the conductionmechanism under which the samples reach breakdown was Fowler–Nordheim �FN�. The breakdown strength Ebd of La2O3 is obtainedfrom this “time to zero breakdown” measurement. AnEbd = 4.6 MV/cm was found for W-gated La2O3, which is in agree-ment with the value reported for La2O3 in the literature.

13

From Fig. 4, EOT = 1.45 and 1.62 nm were obtained for PDAand PMA samples, respectively, taking into account the quantummechanical correction.14 W–La2O3 stack without annealing presentshigher hysteresis compared to PDA and PMA samples. Besides, theclockwise hysteresis loop shows that hysteresis is caused by nega-tive charge trapping which occurs during the transition from deple-tion to accumulation regimes. After PDA, an increase in the accu-mulation capacitance Cacc and a slight shift of VFB towards thenegative side occur. The increase in Cacc after PDA is related to thedensification of the La O film. The reduction of the total amor-

2 3

G112 Journal of The Electrochemical Society, 154 �5� G110-G116 �2007�G112

phous dielectric physical thickness from its initial value for the as-deposited stack to thinner values after PDA and PMA �from 9.7 nmfor the as-deposited sample to 9.1 nm after PDA� indicates that an-nealing results in film densification. Some reports also indicate thatpart of the oxygen in La2O3 is consumed to form a silicate at theinterface between La2O3 and silicon.

10 As a result, La2O3 filmwould become oxygen-deficient, especially near/at that interfaciallayer �IL� and with a more negative VFB, because oxygen deficiencyin the oxide film causes a negative VFB shift.

15,16 Now, given that thephysical thickness of La2O3 for those particular samples was rela-tively high �about 10 nm� and that the temperature used during thePDA and PMA treatments was relatively low �300°C for 5 min forboth samples�, it is thought that the observed increase in Cacc shownin Fig. 4 comes mainly by the densification process of bulk La2O3,whereas the silicate formation only affects the lower portion of the

Figure 3. J-V characteristics of W–La2O3-gated MIS capacitors after PDAand PMA. The as-deposited sample shows extrinsic breakdown characteristicas compared to annealed W–La2O3 stacks.

Figure 4. C-V characteristics of W–La2O3-gated MIS capacitors after PDAand PMA. After annealing, lower hysteresis and decrease in EOT are ob-tained as compared to the nonannealed sample.

La2O3. This is to say, a very thin silicate IL would form between theLa2O3 and silicon and not all of the La2O3 would turn into silicate.This is evidenced by the increase in Cacc after annealing in Fig. 4,because complete reduction of La2O3 into silicate would otherwiseincrease the EOT as compared to nonannealed La2O3 film. Ofcourse, a more complete physical characterization of La2O3 afterannealing is needed to verify IL formation. Nonetheless, there arereports17 in which the reduction of full La2O3 into silicate for eventhicker films can be expected if the annealing temperature is in-creased to levels �900–1050°C� well above the temperatures weused in this report. Additionally, La2O3 is known to be readily af-fected by moisture absorption and the presence of �OH�− ions re-placing O2− sites is another factor for the VFB shift toward the nega-tive side.18 Finally, the PDA sample still presents a significantdegree of hysteresis in the C-V curve, and this is thought to be aconsequence of the plasma-induced damage after tungsten sputter-ing. With PMA, some of the plasma-induced damage is recoveredand La2O3 with better reliability characteristics is obtained as com-pared to PDA samples. The differences in the reliability propertiesafter PDA and PMA for La2O3 are presented later in this article. TheC-V characteristic of the PMA sample shows a VFB recovery towardthe positive side and a slight decrease of Cacc as compared to PDA.During PMA, the dielectric is already covered with the metal elec-trode so that neither the physical thickness nor the oxygen content ofLa2O3 are efficiently reduced during the densification process, andthis causes Cacc to increase slightly above the as-deposited but be-low the PDA samples and recovers VFB as well. The VFB recoveryafter PMA would also come after considering that the formation ofchemical bonds at the interface between the metal electrode and thedielectric during PMA also takes place �Ref. 18-20�. Regardless ofthe physical origin, the net VFB shift is positive after PMA. In brief,the timing of the thermal processing following the deposition ofLa2O3 plays a key role in controlling the threshold voltage and otherimportant electrical characteristics of W-gated La2O3 MISFETs.

Figure 5 shows the dependence of VFB shift on injected chargedensity Qinj for the W–La2O3 stack after positive CVS for both PDAand PMA samples �capacitors with gate area A = 314 � 10−6 cm2

were used�. From this graph it is obvious that La2O3 after PMAproduces a lower amount of VFB shift, even for greater densities ofinjected charge Qinj. Figures 6a and b show the evolution with timeof gate current Ig during a TDDB measurement with positive CVS.In Fig. 6a and b, the initial I is proportional to the CVS applied

Figure 5. VFB shift after positive CVS for W–La2O3 stacked MIS capacitorswith EOT � 1 nm. Here, the stressed samples correspond to set A samplesin Fig. 1 �process flow A in Fig. 2�.

g

G113Journal of The Electrochemical Society, 154 �5� G110-G116 �2007� G113

�more clearly visible for the PMA sample� and so is the time re-quired for both samples to reach breakdown. Ig for the PMA sampleis kept almost constant during the CVS measurement until break-down is reached. This is a significant difference with respect toLa2O3 after PDA, in which Ig decreases almost immediately afterstressing the sample for a few seconds. This fast decrease in gatecurrent with time is attributed to a large density of oxygen vacanciesafter annealing or to plasma damage introduced during the sputter-ing of the metal. Because PMA can anneal out some of the La2O3plasma-related damage introduced during the W deposition, the den-sity of traps for electrons within La2O3 should be reduced and thusthe evolution of Ig with time during CVS becomes more resistant todegradation, as shown in Fig. 6b. In contrast, PDA samples have ahigher density of traps within the oxide that are electron-filled dur-ing the stress, and this can dramatically shift the VFB of the MIScapacitors as shown in Fig. 5. In Fig. 7, the lifetime projection forthe W–La2O3 stack is plotted after TDDB with CVS measurementsby taking into account a 63% mean time to failure. Compared toPDA, a W–La2O3 stack after PMA requires longer times to reachbreakdown even for the same applied gate voltage Vg. Figure 8shows the corresponding charge to breakdown density for the

Figure 6. Evolution of gate-leakage current with time during CVS for La2O3after PDA and PMA. PMA samples reach breakdown at longer times evenfor higher stressing voltages.

W–La2O3 stack after PDA and PMA. By linearly extrapolating thetime to breakdown data up to 10 years, we found that the maximumgate voltage necessary for W–La2O3 to reach breakdown isVg = 2.1 and 2.5 V for PDA and PMA samples, respectively. Inbrief, PMA provides more reliable devices because of a lower VFBshift after stress and its longer lifetime before breakdown, yet theslight increase in EOT after PMA is a major disadvantage. We pointout that these lifetime predictions are only a rough approximationbecause once the exponential law of voltage dependence reachesvery low voltages �compared to those used during operation condi-tions�, the linear model tends to approach a more power-law-basedmodel.21,22 In any case, a 10 year operation is guaranteed by usingthe very simple linear Vg or reciprocal 1/Vg model because of thehigh electric field used during the stressing of these samples. At highelectric field, the linear and reciprocal models converge and that is

Figure 7. �Color online� Lifetime projection for W–La2O3-gated MISCAPafter annealing. By extrapolating time to breakdown to lower Vg with a linearmodel, both PDA and PMA samples are projected to survive 10 years ofcontinuous operation if Vg � 2 V.

Figure 8. Charge to breakdown density Qbd for W–La2O3 stack after anneal-ing. Longer lifetimes are obtained for PMA samples even at higher V .

g

G114 Journal of The Electrochemical Society, 154 �5� G110-G116 �2007�G114

why both models can be used here to project, at least roughly, thelifetime of La2O3. Figure 8 shows that for W–La2O3 after PMA,higher densities of injected charge are required to reach breakdownas expected. At a same gate voltage of Vg = 3.6 V, Qbd density forLa2O3 after PMA is almost three orders of magnitude higher thanthat for La2O3 after PDA, so that longer lifetimes are ensured forLa2O3 after PMA.

VFB shift for W–La2O3 gated MISCAP with EOT � 1 nm.— Inthis section we present some reliability results of a very thin La2O3film deposited within an O2 flow �thus minimizing O2 deficiencyafter the deposition� and in situ metallization so that less exposure ofLa2O3 to the environment was obtained in order to reduce any pos-sible contamination or degradation on its electrical properties, thesesamples correspond to set B of the process flow shown in Fig. 2.Because of the in situ metallization, it was not possible to measurethe physical thickness of La2O3 after its deposition. From C-V andJ-V data, however, we could obtain both EOT and gate-leakagecurrent density, respectively �see the two highlighted data points inthe upper left corner of Fig. 1�. EOT = 0.58 and 0.74 nm were ob-tained for the W–La2O3 stack after PMA at 300 and 500°C, respec-tively. Figures 9 and 10 show the fresh J-V and C-V characteristicsfor these samples together with the as-deposited condition. It isshown that there is no significant difference between the as-deposited La2O3 and the La2O3 after PMA at 300°C in both J-V andC-V characteristics. Also, VFB shifts after annealing were reduced�compare Fig. 4 and 10�. La2O3 after PMA at 500°C decreasesgate-leakage current and Cacc as well, and this comes as a result ofthe SiO2-based IL formation that is developed at the La2O3–siliconinterface. This SiO2 IL was detected after X-ray photoelectron spec-troscopy �XPS� measurements as seen in Fig. 11. It is thought thatbetter interface trap properties �improvement in the quality of theinterface immediately above the silicon substrate or reducedinterface-state density Dit compared to the as-deposited sample� be-tween La2O3 and silicon can be achieved because of this IL forma-tion; however, Dit after PMA still needs to be evaluated because ofthe low quality �i.e., high defect density� of the SiO2-based IL for-mation. Besides, forming gas annealing the samples would be nec-essary to passivate the defects at this newly formed Si–SiO2 inter-face. A higher PMA temperature further reduces any possibleplasma-related damage after the tungsten electrode deposition. Allof these effects combined would suggest a decrease in the density oftrapped charge at both the La O –silicon interface and the bulk of

Figure 9. J-V characteristics for W–La2O3 �EOT � 1 nm� after in situ met-allization. Higher PMA temperature reduces Jg.

2 3

La2O3. Nonetheless, a higher reliability for this W–La2O3 stackcomes along with an increase in EOT after PMA at higher tempera-tures. By suppressing the IL formation, reduced EOTs can be ob-tained but their reliability can be degraded, so an important issue tobe solved for La2O3 with high-temperature annealing is to engineerEOT vs IL formation for better reliability.

Because of the very small La2O3 thickness, a high gate-leakagecurrent density is always observed in these samples, so the time tobreakdown data necessary to project the lifetime of this ultrathinLa2O3 film cannot be so easily extracted. A conventional TDDBmeasurement under constant voltage stress for thicker oxide filmsprovides good reproducibility when considering lower variability ofthe time to breakdown data. For ultrathin oxides, however, it is verydifficult to extract time to breakdown data beacause the detection ofhard-breakdown is no longer straightforward. While in thick oxidesthe breakdown event is easily detected as a sudden increase in leak-age current, in ultrathin oxides, the problem is the large backgroundtunneling current that can mask the localized current flow. There-

Figure 10. C-V characteristics for W–La2O3 �EOT � 1 nm� after in situmetallization. A more stable VFB for all samples is obtained.

Figure 11. �Color online� Annealing temperature dependence of Si 1s XPSspectrum of La2O3 deposited within oxygen flow. A La-silicate IL formationis present for all samples whereas an additional SiO2-based IL formationappears for PMA at 500°C.

G115Journal of The Electrochemical Society, 154 �5� G110-G116 �2007� G115

fore, the soft breakdown mechanism becomes the dominant effectbehind ultrathin oxide degradation.23-25 One method to detect theoxide degradation on the electrical characteristics of these ultrathinoxides is by measuring changes in the stress-induced leakage current�SILC�, which corresponds with anomalous increases in the gate-leakage current that are proportional to the stressing time and whichare related to increases in Dit. Figure 12 shows that for low electricfields, an anomalous increase of the SILC for W–La2O3 stack oc-curs, whereas at high fields the FN current remains unchanged. Forultrathin oxides and low-bias stressing, SILC and soft-breakdown�SBD� mechanisms play a fundamental role in the degradation of thedevices. Figures 13a-c show the C-V characteristics before and afterCVS for the ultrathin La2O3 films. As with thicker La2O3, a positiveVFB shift which is proportional to the stressing time is observed forall samples. The La2O3 film with PMA at 500°C shows less VFBshift after CVS. A high Dit �which is present in these samples�makes it difficult to calculate the real VFB of the devices. For thatreason we also obtained �and compared to calculated values� VFB bydouble differentiating �1/�C/Cacc��

2 vs Vg curve from the experi-mental data in which a sharply peaked curve was obtained. Theposition of the peak was then considered as the experimental VFB.

26

The general trend for all the stressed samples was to shift their VFB

Figure 12. �Color online� J-V characteristics for W–La2O3 �PMA at 500°C�before and after CVS stress. The huge increase in SILC currents after stress�which are related to Dit� poses a serious concern for these thinner La2O3films.

Figure 13. �Color online� C-V characteristics of W–La2O3 �EOT � 1 nm�shown. Even though a high D is present, the lowest shift in V after stres

it FB

to the more positive side as a consequence of electron trapping �asthe VFB shift is more severe for samples annealed at lower PMAtemperature�. One important difference between these samples andthose obtained by standard vacuum deposition of La2O3 �set Asamples of Fig. 2� is that less VFB shift, less hysteresis, and thus amore reliable La2O3 film is obtained by the in situ deposition oftungsten immediately after La2O3 deposition within O2 ambient.

Figure 14 shows how hysteresis for these W–La2O3 MIS capaci-tors increases proportionally to stressing time. The as-deposited andPMA at 300°C samples show the largest hysteresis increase for theW–La2O3 stack. PMA at 500°C shows the lower change in hyster-esis with stressing time and this trend is kept almost constant around80 mV. In this respect, the reason for the lower hysteresis is a physi-cally thinner stack, which has a lower number of defects �assuminga uniform density cm−3; as it has a lower volume, the number ofdefects in the film is reduced�, and consequently there is less chargetrapped. Figure 15 shows the dependence of the VFB shift on injectedcharge density �Qinj� for La2O3 films with EOT � 1 nm after posi-tive CVS. The La2O3 films after PMA at 500°C presents the lowestVFB shift after stress. These data are taken from the C-V character-istics shown in Fig. 13a-c. Compared to Fig. 5, a higher reduction inVFB shift after CVS for the thinner La2O3 film is obtained. This isrelated in part to the high quality of both the film itself and its

and after positive CVS. Set B samples �after process flow B in Fig. 2� aretained for the sample with PMA at 500°C.

Figure 14. �Color online� Increase in C-V hysteresis H after stress.W–La2O3 after PMA at 500°C shows the lowest increase in H after stressingtime.

befores is ob

G116 Journal of The Electrochemical Society, 154 �5� G110-G116 �2007�G116

interface with silicon substrate. A reduction in the density of oxygenvacancies in the film should decrease the density of trap sites inLa2O3, and because of a SiO2-based interfacial layer formation,trapping of charge at the interface with silicon substrate is thought todecrease as well. Moreover, a very thin La2O3 film increases thecontribution of direct tunneling �DT� leakage current, so the densityof trapped charge within La2O3 is decreased because DT throughvery thin oxide layers can be a much less destructive transportmechanism than FN injection for thicker oxides.27,28

Conclusions

The electrical and reliability characterization of W–La2O3–nSistacks with EOT = 1.5 nm and less than 1 nm after annealing wascarried out. Compared to PDA samples, PMA samples show betterreliability characteristics so that important parameters like higherendurance to VFB shift, longer lifetimes, and higher charge densitiesto breakdown can be obtained. Because of the high voltages appliedduring stressing, a 10 year lifetime projection for the W–La2O3–nSisystem is guaranteed by using the very simple Vg linear model. It isthought that La2O3 deposition within an oxygen flow can reduceoxygen deficiency in the dielectric film and in situ deposition oftungsten on La2O3 would reduce exposure of the La2O3 surface toenvironment contamination, so that an improved interface betweenLa2O3 and silicon can be obtained. In an effort to minimize theeffects of oxygen vacancies and to stabilize the reactivity of themetal oxide, the deposition of La2O3 under process B �Fig. 2� can bethought of as being done under PDA in O2 ambient, so that La2O3 isoxidized during the deposition itself, creating a silicate layer. Al-though detrimental to the permittivity, the resulting structure shouldbe more stable and less apt to contain oxygen deficiencies.29 Ahigher PMA temperature for La2O3 enhances the formation of aSiO2-based interfacial layer, and this higher annealing temperaturefurther reduces possible plasma-related damage during the tungstenelectrode deposition at the same time. All these procedures com-

Figure 15. VFB shift after CVS for W–La2O3 with EOT � 1 nm. The lowestVFB shift is obtained for the sample with LaSixOy–SiO2 stacked structurewhich is formed after PMA at 500°C.

bined with a very thin physical thickness for La2O3 result in a morereliable W–La2O3 stack. It is thought that the formation of a verythin SiO2-based interfacial layer could help to improve the reliabil-ity characteristics of metal/high-k stacked devices. However, the in-crease in EOT by lower-k interfacial layer formation is an issue thatmust be addressed, because MOS devices with small EOT less than1 nm require very thin dielectrics, the controllability and quality ofthis interfacial layer becomes crucial. Finally, the lowest VFB shiftafter stress corresponds to the LaSixOy-SiO2 stack structure formedafter PMA at 500°C.

AcknowledgmentsThis work was partially supported by the Semiconductor Tech-

nology Academic Research Center �STARC� and Special Coordina-tion Funds for Promoting Science and Technology by Ministry ofEducation, Culture, Sports, Science and Technology, Japan.

Tokyo Institute of Technology assisted in meeting the publication costs ofthis article.

References1. G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys., 89, 5243 �2001�.2. J. A. Ng, Y. Kuroki, N. Sugii, K. Kakushima, S. Ohmi, K. Tsutsui, T. Hattori, H.

Iwai, and H. Wong, Microelectron. Eng., 80, 206 �2005�.3. Y. Kuroki, J. Ng, K. Kakushima, N. Sugii, K. Tsutsui, and H. Iwai, ECS Trans.,

1�5�, 239 �2006�.4. J. Molina, K. Kakushima, P. Ahmet, S. Nobuyuki, K. Tsutsui, and H. Iwai, ECS

Trans., 1�5�, 757 �2006�.5. Y. H. Wu, M. Y. Yang, A. Chin, W. J. Chen, and C. M. Kwei, IEEE Electron

Device Lett., 21, 341 �2000�.6. H. Yamada, T. Shimizu, A. Kurokawa, K. Ishii, and E. Suzuki, J. Electrochem.

Soc., 150, 429 �2003�.7. J. Sanghun, F. J. Walker, C. A. Billman, R. A. McKee, and H. Hyunsang, Tech.

Dig. - Int. Electron Devices Meet., 2002, 955.8. S. J. Rhee et al., J. VLSI Signal Proc. Syst. Signal, Image, Video Technol., 2005,

168.9. H. Iwai et al., Tech. Dig. - Int. Electron Devices Meet., 2002, 625.

10. J. P. Maria, D. Wicaksana, A. I. Kingon, B. Busch, H. Schulte, E. Garfunkel, and T.Gustafsson, J. Appl. Phys., 90, 3476 �2001�.

11. Y. Kim, S. Ohmi, K. Tsutsui, and H. Iwai, Jpn. J. Appl. Phys., Part 1, 44, 4032�2005�.

12. Y. Kim, S. Ohmi, K. Tsutsui, and H. Iwai, Solid-State Electron., 49, 825 �2005�.13. J. W. McPherson, K. Jinyoung, A. Shanware, H. Mogul, and J. Rodriguez, IEEE

Trans. Electron Devices, 50, 1771 �2003�.14. J. Hauser, CVC NCSU Software, Version 3.0., NCSU �1996�.15. B. E. Deal, M. Sklar, A. S. Grove, and E. H. Snow, J. Electrochem. Soc., 114, 266

�1967�.16. S. Ohmi, C. Kobayashi, I. Kashiwagi, C. Ohshima, H. Ishiwara, and H. Iwai, J.

Electrochem. Soc., 150, F134 �2003�.17. T. Gougosi, M. J. Kelly, D. B. Terry, and G. N. Parsonset, J. Appl. Phys., 93, 1691

�2003�.18. A. Kuriyama, S. Ohmi, K. Tsutsui, and H. Iwai, Jpn. J. Appl. Phys., Part 1, 44,

1045 �2005�.19. M. Kadoshima, Mater. Res. Soc. Symp. Proc., 786, E6-27 �2003�.20. D. A. Buchanan, Tech. Dig. - Int. Electron Devices Meet., 2000, 223.21. J. H. Stathis, IBM J. Res. Dev., 46, 265 �2002�.22. E. Y. Wu, E. J. Nowak, A. Vayshenker, W. L. Lai, and D. L. Harmon, IBM J. Res.

Dev., 46, 287 �2002�.23. M. Depas, T. Nigam, and M. Heyns, IEEE Trans. Electron Devices, 43, 1499

�1996�.24. E. Miranda, J. Sune, R. Rodriguez, M. Nafria, X. Aymerich, L. Fonseca, and F.

Campabadal, IEEE Trans. Electron Devices, 47, 82 �2000�.25. J. S. Suehle, IEEE Trans. Electron Devices, 49, 958 �2002�.26. R. J. Hillard, J. M. Heddleson, D. A. Zier, P. R. Choudhury, and D. K. Schroder, in

Diagnostic Techniques for Semiconductor Materials and Devices, J. L. Benton, G.N. Maracas, and P. Rai-Choudhury, Editors, PV 92-2, p. 261, The ElectrochemicalSociety Proceedings Series, Pennington, NJ �1992�.

27. I. V. Grekhov, A. F. Shulekin, and M. I. Vexler, IEEE Trans. Electron Devices, 44,2307 �1997�.

28. H. Sasaki, M. Ono, T. Yoshitomi, T. Ohguro, S. Nakamura, M. Saito, and H. Iwai,IEEE Trans. Electron Devices, 43, 1233 �1996�.

29. M. Copel, E. Cartier, and F. M. Ross, Appl. Phys. Lett., 78, 1607 �2001�.