9
Boise State University ScholarWorks Electrical and Computer Engineering Faculty Publications and Presentations Department of Electrical and Computer Engineering 10-1-2008 Temperature (5.6-300K) Dependence Comparison of Carrier Transport Mechanisms in HfO2/SiO2 and SiO2 MOS Gate Stacks Richard G. Southwick III Boise State University Justin Reed Boise State University Christopher Buu Boise State University Hieu Bui Boise State University Ross Butler Boise State University See next page for additional authors ©2008 IEEE. Personal use of this material is permied. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE. DOI: 10.1109/IRWS.2008.4796084

Temperature (5.6-300K) Dependence Comparison of Carrier

  • Upload
    others

  • View
    7

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Temperature (5.6-300K) Dependence Comparison of Carrier

Boise State UniversityScholarWorksElectrical and Computer Engineering FacultyPublications and Presentations

Department of Electrical and ComputerEngineering

10-1-2008

Temperature (5.6-300K) DependenceComparison of Carrier Transport Mechanisms inHfO2/SiO2 and SiO2 MOS Gate StacksRichard G. Southwick IIIBoise State University

Justin ReedBoise State University

Christopher BuuBoise State University

Hieu BuiBoise State University

Ross ButlerBoise State University

See next page for additional authors

©2008 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposesor for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other worksmust be obtained from the IEEE. DOI: 10.1109/IRWS.2008.4796084

Page 2: Temperature (5.6-300K) Dependence Comparison of Carrier

AuthorsRichard G. Southwick III, Justin Reed, Christopher Buu, Hieu Bui, Ross Butler, G. Bersuker, and William B.Knowlton

This article is available at ScholarWorks: https://scholarworks.boisestate.edu/electrical_facpubs/58

Page 3: Temperature (5.6-300K) Dependence Comparison of Carrier

Temperature (5.6-300K) Dependence Comparison of Carrier Transport Mechanismsin HfO2/SiO2 and SiO2 MOS Gate Stacks

Richard G. Southwick III1, J. Reed', C. Buu', H. Bui', R. Butler', G. Bersuker3, and W.B. Knowlton"2

'Dept. of Electrical and Computer Engineering,2Dept. of Materials Science and Engineering; Boise State University; 1910 University Dr.; Boise, ID 83725 USA

208-426-5705; fax: 208-426-2470; e-mail: bknowltongboisestate.edu3SEMATECH; Austin, TX 78741 USA

ABSTRACT

Temperature dependent measurements have been used to examinetransport mechanisms and energy band structure in MOS devices. Inthis study, a comparison between high-k HfO2 dielectrics andconventional SiO2 dielectrics is made to investigate dielectricspecific thermally activated mechanisms. Temperature dependentmeasurements on large area n/pMOSFETs composed of SiO2 andHfO2/SiO2 gate dielectrics were performed from 5.6K to 300K. Alarge increase in the gate leakage current is observed at the formationof the minority carrier channel. The data indicate that gate leakagecurrent prior to the formation of the minority channel is carrier ratelimited while gate leakage current is tunneling rate limited above thethreshold voltage. Gate leakage current measurements show twodistinct Arrhenius transport regimes for both SiO2 and HfO2 gatedielectrics. The Arrhenius behavior of the gate leakage current ischaracterized by a strong temperature dependent regime and a weaktemperature dependent regime. The activation energy of the strongtemperature regime is found to vary with the applied gate voltage.Frenkel-Poole or other electric field models are able to explain thegate voltage dependence of the gate leakage current for the low-temperature/voltage regime investigated. The data suggest that thevariation of the activation energy for the Arrhenius behavior isweakly electric-field driven and strongly voltage, or Fermi energylevel, driven. The weak electric field and strong voltage dependenceof the thermal characteristics of the gate leakage current may point totrap densities within the HfO2 that vary in energy (hence appliedvoltage) as responsible for the observed activation energies. Trapassisted tunneling (or hopping) could be implicated as the transportmechanism.

INTRODUCTION

Understanding carrier transport is critical to development ofreliability models of metal oxide semiconductor field effecttransistors (MOSFETs). Identifying carrier transport (e.g., Fowler-Nordheim, Frenkel-Poole, etc.) provides reliability models with a

physical basis to enhance the understanding of device limitations andpotential solutions. A detailed knowledge of carrier transport can

give insight to defects that are present in the oxide (a defect mediatedtransport), in which energy is deposited and transport of transientcharges give rise to threshold instabilities. Temperature dependentmeasurements of the gate leakage current have been used extensivelyto understand carrier transport through many different gatedielectrics such as silicon nitride [1, 2], titanium oxide [3], aluminumoxide [4], europium oxide [5], zirconium oxide [6], and others. Withthe emergence of hafnium oxide (HfO2) as the next gate dielectric toreplace silicon dioxide (SiO2), it is not unexpected that temperaturedependent measurements have been made to assess its electricalproperties [7-1 1].

Temperature dependent measurements on HfO2 have reportedFrenkel-Poole conduction with various trap depths between 0.35eVand 1.5eV below the conduction band [7, 8, 11]. Temperaturedependent charge trapping experiments have been performed andindicate a large presence of charge traps around 0.35eV whichcontribute to threshold instabilities [12]. Band offsets have beenmeasured using Schottky emission [11] and transient and steady statecurrents have been measured and modeled [9, 10], demonstrating theadvantage of temperature dependent measurements.

Despite the large amount of temperature dependent work forunderstanding carrier transport in HfO2, the majority of the work hasfocused on above room temperature measurements with few studiesanalyzing transport below room temperature and even fewer atcryogenic temperatures below 77K.

Although 77K reduces thermal smearing of the Fermi energylevel (Ef) relative to room temperature, it is clear that temperaturesmuch closer to OK provide an appreciably sharper energy probe (Fig.1). This work attempts to provide a clearer understanding of thetrends observed in carrier transport of HfO2 by investigating gateleakage currents in the low temperature regime ranging from 5.6K to300K. Two different gate stacks of TiN/HfO2/SiO2/Si and aTiN/SiO2/Si are measured and both nMOSFETs and pMOSFETs areused. A comparison between the HfO2/SiO2 and SiO2 gate stackallows temperature specific trends of the HfO2 to be identified. Lowtemperatures and finer temperature increments minimize thermalsmearing and can reveal convoluted or competing mechanismsproviding further insight to carrier transport in HfO2/SiO2 gatestacks.

40

30

X 20

10

-0.10 -0.05 0

E-Ef (eV)0.05 0.10

Fig. 1: Thermal broadening function, FT(E) =D02(E), as a function ofE-Ef at 5.6K, 77K and 300K. The energy distribution at 5.6K issignificantly narrower allowing the Ef position to be known muchmore accurately.

978-1-4244-2194-7/08/$25.00 ©2008 IEEE

Temperature5.6

~o 77300

5.6K

o 77K300K

n

48 2008 IIRW FINAL REPORT

Page 4: Temperature (5.6-300K) Dependence Comparison of Carrier

Temperature dependence comparison: carrier transport mechanisms: HfO2ISiO2 & SiO2 MOS gate stacks

This paper is presented as follows. First, the test devices andexperimental procedure are presented. Next, general observations inthe behavior of the gate leakage current, over a broad temperaturerange, are illustrated. The Arrhenius behavior of the gate leakagecurrent is then described followed by an analysis of the activationenergies and approaches used to correct for electric field dependence.

TEST DEVICES AND EXPERIMENTALPROCEDURE

Devices used in this study are MOSFETs fabricated using astandard CMOS process flow including 1000°C/10s dopantactivation and 480°C forming gas anneals. The gate stack of eachMOSFET consists of a titanium nitride (TiN) metal gate and adielectric bi-layer of 3nm of ALD HfO2 on a 1. nm chemicallygrown SiO2 IL. A more detailed fabrication description is presentedin [13]. A control wafer composed of 2nm SiO2 is used as a base tocompare the HfO2 samples. Test devices are large 30ptm/30ptm(width/length) and 50ptm/50ptm MOSFETs to increase the signal tonoise ratio of the gate leakage current. A Janis custom built variablerange probe station (5.6-450K) with actively cooled Kelvin probescombined with a Keithley 4200SCS with remote pre-amps was usedin the measurements.

RESULTS AND DISCUSSION

Gate leakage currents for nMOSFETs and pMOSFETs composedof HfO2/SiO2 and SiO2 are shown in Figs. 2 and 3, respectively. Forboth n- and pMOSFETs, the HfO2/SiO2 gate stack shows a largerincrease in gate leakage current with temperature than the SiO2 gatestacks. For positive gate biases of IV, the high-k gate nMOSFETshows over an order of magnitude increase in gate leakage currentwith the pMOSFET showing over three orders of magnitudeincrease. While the increase in gate leakage current for -1V gatebiases is lower than IV gate biases, it is still higher than SiO2. Sinceequivalent oxide thicknesses, EOTs, are relatively similar (EOThigh-k= 1.3nm and EOTsio2 = 2nm), the HfO2/SiO2 gate leakage currentdata differ from the SiO2 data most likely because of the HfO2defects, phonon modes, and perhaps carrier transport modes.

E01--

_m

io-2black nMOSFET 30m/30pm t = 3nm t - l.1nm

1 o gray nMOSFET 30mni/30prmI = 2nm

I ov4t _E! 11lT~~~~~~~~~~~emp.

0 o-51 o-6 iTemp. Inc. v sio2 g

i o-7

i o-lo-lo1011 F Temp. 5.6Kto 3

-1.0 -0.5 0 0.5

1.Inc

;ion1

1.0

VGate (V)Fig. 2: Gate leakage current for nMOSFETs composed of aHfO2/SiO2 gate dielectric and a SiO2 gate dielectric for temperaturesranging from 5.6K to 300K. The encircled region marks the weakinversion regime of MOS operation and shows a large increase gateleakage current.

E1-f

a,

1 o-2- black pMOSFET 30ni/30Lm tgHfl = 3nm t=So 1 Inmmgray pMOSFET 30pm/3Onm tSiO = 2nm

10-3

I0 weak-\&6XlX / /~inversion

1 o-5 Ev

10 8Temp. Inc.

I O\1I

I o-Temp. 5.6K to 300K

1-,I-1.0 -0.5 0

Gate (V)

Temp. Inc.

0.5 1.0

Fig. 3: Gate leakage current for pMOSFETs composed of aHfO2/SiO2 gate dielectric and a SiO2 gate dielectric for temperaturesranging from 5.6K to 300K. The encircled region marks the weakinversion regime. For accumulation (positive voltages), the gateleakage current decreases dramatically at very low temperatures-25K.

General Temperature Observations

The gate current versus gate voltage (IG- VG) temperaturedependent curves for both n- and pMOSFETs show similar trends inthree different regimes. These regimes are: 1) weak inversion - seeencircled region in Figs. 2 and 3; 2) relatively large positive gatebiases (0.7V to IV); and 3) relatively large negative biases (--1V).The following subsections propose qualitative explanations for thetemperature increases for the three regimes.

Weak Inversion - For both n- and pMOSFETs composed ofHfO2/SiO2 and SiO2 gate dielectrics, a large increase in the gateleakage current, Jgateleak, is observed during weak inversion, theencircled region in Figs. 2 and 3. As the temperature increases, theslope of the sudden increase in the Jgate-leak decreases followingtemperature dependent trends in sub-threshold slope [14]. Applying a3kBT bias between the source and drain [14] to measure the presenceof inversion charge reveals the sudden increase in Jgate-leakcorresponds to the sudden increases of minority carriers in thechannel (Fig. 4).

The data of Figs. 4 and 5 suggest carrier-limited and tunneling-limited regimes. Consider Fig 5, the data indicate the first carriersthat enter the channel from the source and drain do not flow from thesource to the drain (i.e., source and drain current are asymmetricacross the voltage axis), but from the source and drain through thegate thereby increasing the Jgate-leak. Prior to the threshold voltage,depletion/weak inversion, the gate leakage current is carrier-limited,due to the absence of minority carriers, labeled in Fig. 4. It appearsthat the Jgateleak then saturates even as the inversion charge continuesto increase exponentially before losing its exponential dependencewith the gate bias. Since carrier concentration is increasing butcarrier transport (i.e., tunneling) is saturating, tunneling is self-limiting indicating a tunneling-limited mechanism, labeled in Fig. 4.As temperatures increase, the transition between tunneling-limitedand carrier-limited tunneling occurs at lower voltages due to theincreased number of minority carriers as temperatures increase.Above the threshold voltage however, enough minority carriers are

2008 IIRW FINAL REPORT 49

Southwick et al.

49

Page 5: Temperature (5.6-300K) Dependence Comparison of Carrier

Temperature dependence comparison: carrier transport mechanisms: HfO2ISiO2 & SiO2 MOS gate stacks

present so that the tunneling current is tunneling-limited over thetemperature range investigated.

< i 0-5,.,'- * ~~~~~~~~~~~~Gate

110- Source

1 o- VD 3KT

81 o-10

10-1 0 Tunnelingrate limited

Cr 1 0.pMOSFET 3o0fm/30~m cuarrier

cCu103t.His= 3nm rate limitedt1 Inm (chemical)- 14

-1.25 -1.00 -0.75 -0.50

VGateFig. 4: A comparison of source current, a measure of the inversioncharge, to the gate leakage current. The slope in the increase of thegate leakage current matches the sub-threshold slope in the sourcecurrent. Current is tunneling rate limited above the threshold voltage(VTH) and carrier rate limited below -VTH at low temperatures.

4.0x10-9pMOSFET 3OWWm30Qim

5.'2Tem.15.6

0

2-2.0x10-9

-4.Ox10V, -3KT.

Temp. 5.6K

-1.25 -1.00 -0.75

VGate (V)Fig. 5: A plot of all four terminal currents in a HfO2/SiO2 pMOSFET.Tunneling current (i.e., gate current) is carrier starved as observed bythe increase in the gate leakage current corresponding to the presenceof minority carriers flowing from the source and drain to the gate.

Relatively Large Positive Gate Biases: VGate 0.7V to lV - Asthe gate bias increases, the Fermi energy level in the silicon movescloser to the conduction band of the HfO2, Fig. 6a. Nevertheless, atlow temperatures (i.e., minimal thermal broadening - Fig. 1) and lowvoltages, the Fermi level is not close enough to access defect statesnear the HfO2 conduction band. As the temperature increases, thethermal energy distribution broadens (Fig. 1) allowing electrons tooccupy higher energy levels in the silicon that have a greaterprobability to coincide with defect levels in the HfO2, Fig. 6a. Thecombined qualitative result is the temperature dependence of the gateleakage current increases as the gate voltage increases.

VG= 0.8V

VG= IV

G = -IV

= 0.8

ad0

a) n b)

Fig. 6: Energy band diagram of a HfO2/SiO2 nMOSFET at fourdifferent gate biases, a) 0.8V and lV and in b) -0.8V and IV. Asindicated in a), the HfO2 conduction band becomes closer to thesilicon Fermi energy with increasing positive gate bias. Conductionelectrons in the silicon are moved even closer in energy to the HfO2conduction band with temperature for positive gate bias, part a). In b)as the voltage becomes more negative, the silicon Fermi energy levelis further away from defects near the HfO2 conduction band. Createdusing [15].

Examining the relative increase in Jgatejeak, (A/gate,J5.6K,gate), inFig. 7., it is observed that as the gate voltage increases so does therelative Jgate.leak. The relative Jgate.leak increase at 300K for a gate biasof lV is 6 fold. This increase in gate leakage current is much higherthan the theoretical increase based on the work of Ling-Feng Maowhich shows only a 0.2 fold increase at a gate bias of lV [10]. InMao's work, changes in tunneling currents due to changes ineffective mass are calculated for various temperatures. When appliedto this work, a large gate leakage current composed of defectmediated leakage current may explain the higher relative gateleakage current increase in Fig. 7. The relative increase in Jgate.leak forthe SiO2 samples is not as large as the high-k samples but is alsohigher than what is predicted in [10]. It is important to note that thework done by Mao was for HfSiO not HfO2 which also may accountfor some differences.

Relatively Large Negative gate Biases: VGate, -lV - For high-ksamples, when the gate bias increases in the negative voltagedirection, the Fermi energy level in the silicon moves further awayfrom defects near the HfO2 conduction band, Fig. 6b. This results inthe temperature dependence of the relative gate leakage currentdecreasing as the voltage decreases. This is observed in Figs. 2 and 3near VGate = -IV; as the gate voltage decreases the spread of the gateleakage current decreases. The relative gate leakage current for thenegative bias regime (not shown) shows the opposite trend forpositive gate biases, Fig. 7. Comparing to the theoretical work byMao, the relative gate leakage current should not decrease withincreasing voltage. This may further point to defect mediatedtunneling as the primary component for the temperature dependenceof the gate leakage current. As the gate voltage is decreased fornegative voltages, fewer defects are being accessed which contributeless to the temperature dependence. As was seen in the positive biasregime, the negative bias regime also shows a significantly smallertemperature dependent increase for SiO2 compared to HfO2. Thus,increases in the direct tunneling current due to thermal broadening ofcarriers (Fig. 1) in the TiN metal gate and silicon substrate arethought to have a smaller impact on the large increase in thetemperature dependence of the HfO2 gate leakage current.

As the gate voltage continues to increase in the negative directionbeyond -lV, the Fermi energy level in the TiN metal gate becomescloser to the defects near the HfO2 conduction band. It can be

0 2008 IIRW FINAL REPORT

Southwick et al.

50

Page 6: Temperature (5.6-300K) Dependence Comparison of Carrier

Temperature dependence comparison: carrier transport mechanisms: HfO2ISiO2 & SiO2 MOS gate stacks

expected then, in a gate leakage current dominated by defectmediated tunneling, that the relative increase in the gate leakagecurrent should increase at these higher voltages. Figure 8. shows thegate leakage current for -IV and -2V. As indicated, the relative gateleakage current increases from 1.21 and -IV to 2.55 at -2V consistentwith the proposed mechanism.

7closed symbols nMOSFET 30pm/30pm

6 tox,2=3nm t =1mmopen symbols nMOSFET 30pm/30pm

t =2nm55 H;fO/S12 2 V-1 (V) VG

t

Inc.

/ 0.9->LO3 / 0.8 *

high-k

SiO2 VGat, Inc.

0 Lf ff=

0 50 100 150 200 250 300T(K)

Fig. 7: Plot of the relative gate leakage current increase over thetemperature range of 5.6K to 300K for gate voltages of 0.8, 0.9, and1V. The relative increase in gate leakage current for HfO2/SiO2 gatedielectric is much higher than SiO2 and the theoretical analysis donein [10]. Defect mediated tunneling in the HfO2 devices may accountfor the difference.

-20 Larger change in gate leakage

current due to temperature

Smaller change in gate leakage

-2. -1. 1.io-~~~~~~~-.

VGATE (V)

Fig. 8: Plot of the gate leakage current for voltages from -1V to -2V.As the temperature increases at these higher voltages, the change inJgate increases indicating an increased temperature dependence. Anincrease in Jgate with temperature is consistent with electrons fromthe TiN metal gate accessing defect levels in near the HfO2conduction band.

Arrhenius Behavior

The Arrhenius behavior of the Jgate-ieai& reveals two differenttransport regimes for both HfO2/SiO2 and SiO2 MOSFETs (Fig. 9)with the exception of HfO2 pMOSFETs in accumulation. The twodifferent transport regimes can be characterized by a weaktemperature sensitive regime (5.6K to 50K) witht=I eV activationenergies and a strong temperature regime(f450K to at least 300K)with a very wide temperature transition regime. Two transportregimes were also observed by Compagnoni et al. [91. They reportactivation energiesd87meV for temperatures higher than 165K and5meV for the lower temperature regime in experiments performed toa minimum temperature of 77K. In this study, a lower activationenergy of -30pteV is found, indicating a very weak temperature

dependence. The inset of Fig. 9 shows the data of this study in atemperature regime down to which Compagnoni et al. investigated(i.e., 77K). In this regime, an activation energy of 4.5meV isdetermined, which is consistent with the activation energy reportedby Compagnoni et al. (Fig. 9 inset). The much larger activationenergy of 4.5meV is explained by realizing that the range oftemperature used down to 77K to calculate the activation energy is ina transition region which is not linear (Fig. 9).

T(K)7.1 6.3 5.6 5.02 -- 50.0 25.0 16.7 12.5 10.0 8.3

10- p

E

Cu,-ml 1 0-3

Q)

0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20

1IT(Kl)Fig. 9: Arrhenius plot of the gate leakage current density forHfO2/SiO2 and SiO2 MOSFETs showing two different transportregimes, a weak temperature dependent -30pteV regime and a highlytemperature dependent regime. For gate leakage currents from=125K to 77K which appears linear (inset), the weak temperatureregime has and activation energy of 4.5meV. As the measurementscontinue to 5.6K, =125K to 77K is in a transition region and notlinear.

Four different transport regimes can be seen in HfO2/SiO2pMOSFETs in accumulation (Fig. 10). Transport regimes labeled 1and 2 in Fig. 10 are similar in activation energy and transitiontemperature as HfO2/SiO2 nMOSFETs. In addition, the pMOSFETsshow transports regimes labeled 3 and 4. The activation energycalculated for regime 3 is approximately -4meV. A transitionbetween 3 and 2 occurs at temperature of about 25K. Transportregime 4 exists below -9K and is weakly temperature dependent. Itis unclear at this point what might cause such a response. The Fermienergy level in the silicon for similar biases in HfO2/SiO2nMOSFETs and pMOSFETs are aligned to relatively the sameenergy level in the HfO2 so it is unlikely that a different energy rangeis accessed in the pMOSFET.

The slopes of the transport regimes (i.e., activation energy, EA)change with gate voltage for both HfO2/SiO2 and SiO2, Fig. 11. TheSiO2 range of activation energies are lower than the high-k range ofEAs, and vary from 10meV to 40meV. The EAs for the high-k samplesrange from 20meV to 90meV depending on the gate bias. The high-kEAS are somewhat low as compared to [12]. This may be because theEAs have not been corrected for the electric field effect on the trap[12, 16] or that the EA extracted from [12] was using thresholdvoltage shift data. Rough correction calculations using [16], showthat the EAs would increase by -200meV towards that of [12]. Boththe HfO2/SiO2 and SiO2 samples exhibit a transport regime that is

2008 IIRW FINAL REPORT 51

Southwick et al.

,1

Page 7: Temperature (5.6-300K) Dependence Comparison of Carrier

Temperature dependence comparison: carrier transport mechanisms: HfO2ISiO2 & SiO2 MOS gate stacks

nearly independent of temperature in the low temperature rangeprobably due to a low phonon density.

T(K)

io-2

0o-3

E0

00) 1 0-

10-611T(K-1)

Fig. 10: Arrhenius plot of HfO2/SiO2 pMOSFET in accumulation(+VGate). Four different transport regimes are observed and indicatedby arrows and numbers. The lines are only used to show trends forregions 1I

0.08

,-.W

>%0.060)W 0.04

0

g 0.02

:

-4.

SiO2 /High-k/0! /-*r-nMOSFET

./*-./ pMOSFET

L ~~~-1f2 S 0ie

.0 pMOS Acc.?\ \ ~~~~~~~nMVOSInv.

nMOS Acc. SlQ\)o_ *

pMOS Inv.

-1.0

Fig. 11: Plot of the EA foHfO2/SiO2 and SiO2 dielecHfO2/SiO2 dielectric. Boththe applied gate voltage.

220

<1W 200L.

'S 180

E0 160EF- 140

.2 120

' 100.80F-80

-1.0 -0.5 0 0.5 1.0

VGate (V)Fig. 12: Plot of the center transition temperature for HfO2/SiO2 andSiO2 MOSFETs. SiO2 MOSFETs have a higher transitiontemperature than HfO2/SiO2 MOSFETs. The transition temperaturevaries with gate voltage.

Electric Field Correction Factors

As was alluded to in the previous section, a common correctionfactor to EAs is the electric field. A typical electric field correctionfactor is Frenkel-Poole [17]. In Frenkel-Poole conduction, theamount of energy required for an electron to move from the localizedtrap to the conduction band is reduced in the presence of an electricfield (EOX) [18-20]. The higher the electric field, the less energy isneeded for the electron to escape. This results in lower activationenergies at higher electric fields. The equation describing Frenkel-Poole conduction (JFP) is given as:

(B-8FPEOX )JFP PE=°Xe kBT

wherejjlF q

(1)

-0.5 0 0.5 1.0 KB iS IoLLizmann-s constalntev/dV, oqB iS mne Irap cpLn Irom tne-0.5 0 05conduction band, q is the charge of an electron, eo is the permittivityVGate (V) of free space, and k is the dielectric constant. From (1), the

r transport regime above 150K for both dependence on the electric field is seen in the pre-exponential and astrics The for SiO is lower than the a modifier to the numerator of the exponent, or the activation energy.theSiOc and HfOr/SiO2 E 'S depend on It is therefore important to accurately extract the electric field. Below

are several methods of calculating the electric field.

Consider the range of temperatures in which a transition betweencarrier transport regimes occurs. Within a given transition range oftemperatures, there exists a midpoint temperature which we shalldefine as the "center transition temperature". The center transitiontemperature for the transition regime between the highly dependenttemperature regime (T > -200K) and the weakly temperaturedependent regime (T < -50K) is plotted as a function of Vgate in Fig.12. The HfO2/SiO2 MOSFETs on average show a lower centertransition temperature than the SiO2 MOSFETs. Although HfO2 andSiO2 both display two carrier transport regimes, the higher centertransition temperature coupled with the lower activation energy inSiO2, may explain the small gate leakage current temperaturedependence observed in Figs. 2 and 3. The smaller EA and highercenter transition temperature in SiO2 indicates a significantdifference in the carrier transport in SiO2 compared to HfO2.

E =VGate VFB

ox

ox

Fox _fO= VGate VFB COX SO2Eox,H oft CSCOX,HfO2 ox,HfO2 + Cox,SiO2

VGate - VFB - 0, (VGate )

(2)

(3)

(4)-oxH x,f2 t C Cox,HfO2 ox,HfO2 ox,S102

where VFB is the flat-band voltage, Co, SiO2 is the SiO2 layercapacitance, Cox,HfO2 is the HfO2 layer capacitance, to, is the dielectricthickness, and Os is the surface potential in the silicon which is afunction of gate voltage. In (2) the voltage across the oxide isapproximated as the difference in the gate voltage and flat-band

52 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~2008IIRWFINAL REPORT

Southwick et al.

l,_ ;,. __-+,,_+ t_XTIV. IA- ;,. +'k- +,,,- A--+II, V,__ +11,_

E---,,,,

52

Page 8: Temperature (5.6-300K) Dependence Comparison of Carrier

Temperature dependence comparison: carrier transport mechanisms: HfO2ISiO2 & SiO2 MOS gate stacks

voltage and is converted to an electric field through the dielectricthickness. For a two dielectric gate stack and assuming the traps arelocated in the HfO2 dielectric, the voltage across the HfO2 dielectricis modified by the capacitances of each dielectric layer as shown (3).

At low voltages, the surface potential in the silicon accounts for alarge part of the total gate voltage drop and cannot be ignored.Accounting for the surface potential drop in the silicon, qs(VGa,e),which reduces the voltage across the HfO2 is done in (4). UsingSILVACO, the electric field in the HfO2 layer for various voltageswas simulated for temperatures ranging from 10K to 300K. Resultsindicate that the electric field changes little with temperature and areclosely approximated with [15] (Fig. 13). A plot of the four waysdiscussed to calculate the electric field is shown in Fig. 13. Asignificant difference in the manner the electric field through theHfO2 is calculated is observed, particularly in inversion. Substantialerror can occur when performing an electric field correction if aninsufficient method is used to calculate the electric field through theHfO2.

nMOSFET

2.5 - HfO 3nm

E t 1.Inm (chemical)

2.0 =E (VG -"y> ~~~~~oxGate FBox,phsclS

_ 1.5 Simulation:Silvaco

V . 0 BandDia

L.

-0.5 o . xH, , I Gat I) Io

-1.0 -0.5 0 0.5 1.0

Gate (V)Fig. 13: Plot of the electric field calculated three different ways usingequations (2) - (4). The extracted electric field at low voltagesdepends greatly on the chosen method. The electric field wascalculated for temperatures ranging from 1OK-300K usingSILVACO and are observed to vary little. EOX,Hfo2 can beapproximated to SILVACO using the band diagram program [15].

0040 W/L = 30pm/30pm pMOSFET circle000t =3nm

0.035 t s,o = 1.lmm (chemical)

0.030

0.0250015 inMOSFET square

> 0.020

6.55meV0.015

0.010

0.005-

0.000 V

-0.005 II0 200 400 600 800 1000

E1/ oxHfO, (MV/cm)Fig. 14: Frenkel-Poole analysis of the gate leakage current in thehigh temperature transport regime (T< 150K). The activation energyincreases as the electric field in the HfO2 increases which isnonsensical.

Following Frenkel-Poole analysis and determining the activationenergy of JGate/Eox,HfO2, at different electric fields results in Fig. 14.Using a linear fit, the trap depth at flat-band is near OeV and,depending on where the linear fit is taken, may be negative whichdoesn't make sense. Also observed in Fig. 14 is an increase in theactivation energy as the electric field increases. The increase in EA asEox,HfO2 increases does not follow Frenkel-Poole theory which statesthat EA should decrease with electric field. Possible reasons for whyFrenkel-Poole analysis fails here, but has been successful inexplaining the temperature dependence in the work of many others,may be due to the low voltage and low temperature regime discussedhere. That is, it is reasonable to expect that traps in the HfO2 aremore easily accessed at high voltages and temperatures. Frenkel-Poole transport has been shown in similar samples for higher gatevoltages from I to 2V [21].

Examining EA as a function of gate voltage, Fig. 11, the activationenergy decreases and increases, as the gate voltage increases. A moreplausible explanation to the varying EAs maybe due to the position ofthe Fermi energy level and different temperature dependent transporttraps being accessed as the gate voltage changes. We suspect that asthe Fermi energy level begins to coincide with a high density ofdefect states that the EA should decrease. Conversely, the further inenergy the Fermi energy level is from traps, the great EA should be.

SUMMARY

Carrier transport mechanisms at low temperatures and lowvoltages are not well understood in HfO2 gate dielectrics. Bycomparing the temperature dependence of the gate leakage current inHfO2 to SiO2, a better understanding of the transport mechanismspecific to HfO2 can be realized. Gate leakage current measurementson HfO2/SiO2 gate dielectrics and SiO2 gate dielectrics were madefor temperatures ranging from 5.6K to 300K. Results indicate astrong increase in the gate leakage current for both HfO2/SiO2 andSiO2 dielectrics in the weak inversion regime which is attributed tothe increase in minority carriers in the channel. HfO2/SiO2 dielectricsshow a large temperature dependence compared to SiO2. InHfO2/SiO2 for positive voltage -1V, the temperature dependenceincreases with gate voltage, this is thought to occur due to increasedinteraction of electrons with defects near the HfO2 conduction band.For HfO2/SiO2 operating near -1V, the temperature dependencedecreases as the gate voltage becomes more negative, and isattributed to traps nears in HfO2 the conduction band becominginaccessible. Both HfO2/SiO2 and SiO2 dielectrics show twotransport regimes on an Arrhenius graph with pMOSFET HfO2/SiO2dielectrics showing four in accumulation. The activation energy forthe transport regime for T > -150K varies with gate voltage andcannot be modeled with a Frenkel-Poole or electric field approach.Care must be taken when calculating the electric field at low gatevoltages (e.g., close to use conditions) as band bending in the siliconconsumes a significant portion of the applied gate voltage. Variationsin the activation energy with respect to gate voltage indicate a weakelectric field dependence and are probably more dependent on theFermi energy level alignment with defects in the HfO2.

ACKNOWLEDGMENTS

The authors would like to thank the contributions of fellowmembers of the Knowlton Research Group at Boise State Universityand the support of Dr. Haitao Liu and Dr. Steve Groothuis of MicronTechnology, Inc. for assistance with the temperature dependentelectric field calculations using SILVACO. Funding for this projectwas made possible through the Idaho SBoE-HERC, Micron PhDFellowship. Equipment and supplies were provided in part by

2008 IIRW FINAL REPORT

Southwick et al.

53

Page 9: Temperature (5.6-300K) Dependence Comparison of Carrier

Temperature dependence comparison: carrier transport mechanisms: HfO2ISiO2 & SiO2 MOS gate stacks

DARPA Contract #N66001-01-C-80345 and NIH INBRE#P20RR16454.

REFERENCES

[1] H. Tanaka, "Limitation current in Si3N4/SiO2 stacked dielectricfilms," Applied Surface Science, vol. 147, pp. 222-227, 1999.

[2] Y. Takahashi and K. Ohnishi, "Estimation of insulation layerconductance in MNOS structure," Electron Devices, IEEETransactions on, vol. 40, pp. 2006-2010, 1993.

[3] J. C. Tinoco, M. Estrada, B. Ifniguez, and A. Cerdeira,"Conduction mechanisms of silicon oxide/titanium oxide MOSstack structures," Microelectronics Reliability, vol. 48, pp. 370-381, 2008.

[4] J. Kolodzey, E. A. Chowdhury, T. N. Adam, Q. Guohua, I. Rau,J. 0. Olowolafe, J. S. Suehle, and C. Yuan, "Electricalconduction and dielectric breakdown in aluminum oxideinsulators on silicon," Electron Devices, IEEE Transactions on,vol. 47, pp. 121-128, 2000.

[5] A. A. Dakhel, "Poole-Frenkel electrical conduction in europiumoxide films deposited on Si(100)," Crystal Research andTechnology, vol. 38, pp. 968-973, 2003.

[6] F.-C. Chiu, Z.-H. Lin, C.-W. Chang, C.-C. Wang, K.-F. Chuang,C.-Y. Huang, J. Y.-m. Lee, and H.-L. Hwang, "Electronconduction mechanism and band diagram of sputter-depositedAl/ZrO2/Si structure," Journal ofApplied Physics, vol. 97, pp.034506-4, 2005.

[7] T. P. Ma, H. M. Bu, X. W. Wang, L. Y. Song, W. He, and M. M.Wang, "Special reliability features for Hf-based high-k gatedielectrics," IEEE TDMR, vol. 5, pp. 36-44, 2005.

[8] G. Ribes, S. Bruyere, D. Roy, C. Parthasarthy, M. Muller, M.Denais, V. Huard, T. Skotnicki, and G. Ghibaudo, "Physicalorigin of Vt instabilities in high-k dielectrics and processoptimisation," Integrated Reliability Workshop Final Report,2005 IEEE International, pp. 4 pp., 2005.

[9] C. M. Compagnoni, A. S. Spinelli, A. Bianchini, A. L. Lacaita,S. Spiga, G. Scarel, C. Wiemer, and M. Fanciulli, "Temperaturedependence of transient and steady-state gate currents in HfO2capacitors," Applied Physics Letters, vol. 89, pp. 103504, 2006.

[10] L.-F. Mao, "Modeling of temperature dependence of the leakagecurrent through a hafnium silicate gate dielectric in a MOSdevice," Semiconductor Science and Technology, vol. 22, pp.1203-1208, 2007.

[11] W. Zhu, T. P. Ma, T. Tamagawa, Y. Di, J. Kim, R. Carruthers,M. Gibson, and T. Furukawa, "HfO2 and HfAlO for CMOS:thermal stability and current transport," presented atInternational Electron Device Meeting, pp. 20.4.1-4, 2001.

[12] G. Bersuker, J. Sim, P. Chang Seo, C. Young, S. Nadkarni, C.Rino, and L. Byoung Hun, "Mechanism of Electron Trappingand Characteristics of Traps in HfO2 Gate Stacks," Device andMaterials Reliability, IEEE Transactions on, vol. 7, pp. 138-145,2007.

[13] G. Bersuker, J. Peterson, J. Barnett, A. Korkin, J. H. Sim, R.Choi, B. H. Lee, J. Greer, P. Lysaght, and H. R. Huff,"Properties of the interfacial layer in the high-k gate stack andtransistor performance," presented at Proc. of ECS SpringMeeting, pp. 141, 2005.

[14] Y. Taur and T. H. Ning, Fundamentals ofmodern VLSI devices.New York: Cambridge University Press, 1998.

[15] R. G. Southwick and W. B. Knowlton, "Stacked dual oxideMOS energy band diagram visual representation program (IRWstudent paper)," IEEE Transactions on Device and MaterialsReliability, vol. 6, pp. 136-145, 2006.

[16] J. W. McPherson, J. Kim, A. Shanware, H. Mogul, and J.Rodriguez, "Trends in the ultimate breakdown strength of highdielectric-constant materials," IEEE Transactions on ElectronDevices, vol. 50, 2003.

[17] J. Frenkel, "On Pre-Breakdown Phenomena in Insulators andElectronic Semi-Conductors," Physical Review, vol. 54, pp. 647,1938.

[18] J. Frenkel, "On Pre-Breakdown Phenomena in Insulators andElectronic Semi-Conductors," Physical Review, vol. 54, pp.647-648, 1938.

[19] A. K. Jonscher, "Electronic properties of amorphous dielectricfilms," Thin Solid Films, vol. 1, pp. 213-234, 1967.

[20] J. G. Simmons, "Poole-Frenkel conduction in amorphoussolids," Philosophical Magazine, vol. 23, pp. 59-86, 1971.

[21] R. G. Southwick, J. Reed, G. Bersuker, and W. Knowlton,"Preliminary Study of Temperature Dependence of TiN/3nmHfO2/1.Inm SiO2/Si Gated MOSFETs," presented at IIRW, pp.,2007.

QtuESTIONS AND ANSW1ERS

Q: Does your electric field equation include the effects of chargetrapping?A: No, the electric field equation assumes no charge trapping.

Q: Have you ever tried to solve for electric field using the continuity ofconductance?A: No, we have not. All of the electric field calculations used so farassume no current transport through the dielectrics.

4 2008 IIRW FINAL REPORT

Southwick el: al.

,54