Fluid Imprint and Inertial Switching in Ferroelectric La ...Fluid Imprint and Inertial Switching in Ferroelectric La:HfO2 Capacitors Pratyush Buragohain,† Adam Erickson,‡ Pamenas

Fluid Imprint and Inertial Switching in Ferroelectric La:HfO2CapacitorsPratyush Buragohain,† Adam Erickson,‡ Pamenas Kariuki,§ Terence Mittmann,§ Claudia Richter,§

Patrick D. Lomenzo,§ Haidong Lu,† Tony Schenk,§,∥ Thomas Mikolajick,§ Uwe Schroeder,§

and Alexei Gruverman*,†

†Department of Physics and Astronomy and ‡Department of Mechanical and Materials Engineering, University of Nebraska,Lincoln, Nebraska 68588, United States§NaMLab gGmbH/TU Dresden, Noethnitzer Strasse 64, Dresden 01187, Germany∥Materials Research and Technology Department, Luxembourg Institute of Science and Technology, Belvaux L-4422, Luxembourg

*S Supporting Information

ABSTRACT: Ferroelectric (FE) HfO2-based thin films,which are considered as one of the most promising materialsystems for memory device applications, exhibit an adversetendency for strong imprint. Manifestation of imprint is a shiftof the polarization−voltage (P−V) loops along the voltageaxis due to the development of an internal electric bias, whichcan lead to the failure of the writing and retention functions.Here, to gain insight into the mechanism of the imprint effectin La-doped HfO2 (La:HfO2) capacitors, we combine thepulse switching technique with high-resolution domainimaging by means of piezoresponse force microscopy. This approach allows us to establish a correlation between themacroscopic switching characteristics and domain time−voltage-dependent behavior. It has been shown that the La:HfO2capacitors exhibit a much more pronounced imprint compared to Pb(Zr,Ti)O3-based FE capacitors. Also, in addition toconventional imprint, which evolves with time in the poled capacitors, an easily changeable imprint, termed as “fluid imprint”,with a strong dependence on the switching prehistory and measurement conditions, has been observed. Visualization of thedomain structure reveals a specific signature of fluid imprintcontinuous switching of polarization in the same direction as thepreviously applied field that continues a long time after the field was turned off. This effect, termed as “inertial switching”, isattributed to charge injection and subsequent trapping at defect sites at the film−electrode interface.KEYWORDS: ferroelectric capacitors, HfO2 films, imprint, piezoresponse force microscopy, charge trapping

1. INTRODUCTION

Development of the ferroelectric (FE) HfO2-based thin films1

represents a significant step forward in the quest toward therealization of non-volatile memory technology.2 Thesematerials are also of interest for development of conceptuallynew devices based on the negative capacitance effect.3 Inherentadvantages of FE HfO2 over conventional perovskite filmsinclude compatibility with existing complementary metal-oxide-semiconductor (CMOS) technology, thickness scalabil-ity critical for fabrication of three-dimensional capacitorstructures, as well as a large band gap (>5 eV), which makesthem resilient against electrical breakdown.4 However,integration of HfO2-based films into electronic devices ishampered by serious performance instability associated withthe profound wake-up effect, limited endurance, and strongimprint effect.5−12 Imprint is one of the most seriousdegradation effects in FEs manifested by the shift of ahysteresis loop along the voltage axis with time,13−15 resultingin effective asymmetry of the coercive field, destabilization ofone of the polarization states and, as a consequence, retention

or even write failure. The imprint phenomenon has beenwidely investigated in perovskite FEs, such as BaTiO3 andPb(Zr,Ti)O3 (PZT).

13−18 It is generally attributed to aninternal bias that can have various origins including electricalcharge injection, migration, and trapping at the interface orgrain boundaries, defect dipole alignment, as well as stressgradients.16,19−21 Electron trapping at oxygen vacancies nearthe interface was recently proposed to be a root cause forimprint in the HfO2 devices.

6,7 However, comprehensiveunderstanding of the microscopic mechanism behind thepeculiar imprint behavior of the HfO2-based films requiresmoving beyond the integral electrical testing methods, such asfirst-order reversal curve, used so far.5−7,12 Here, we combinethe pulse switching techniques with high-resolution domainimaging by means of piezoresponse force microscopy (PFM)to gain insight into the imprint effect in La-doped HfO2

Received: June 25, 2019Accepted: August 28, 2019Published: August 28, 2019

Research Article

www.acsami.orgCite This: ACS Appl. Mater. Interfaces 2019, 11, 35115−35121

© 2019 American Chemical Society 35115 DOI: 10.1021/acsami.9b11146ACS Appl. Mater. Interfaces 2019, 11, 35115−35121

Dow

nloa

ded

via

UN

IV O

F N

EB

RA

SKA

LIN

CO

LN

on

Sept

embe

r 28

, 201

9 at

23:

15:3

7 (U

TC

).Se

e ht

tps:

//pub

s.ac

s.or

g/sh

arin

ggui

delin

es f

or o

ptio

ns o

n ho

w to

legi

timat

ely

shar

e pu

blis

hed

artic

les.

www.acsami.orghttp://pubs.acs.org/action/showCitFormats?doi=10.1021/acsami.9b11146http://dx.doi.org/10.1021/acsami.9b11146

(La:HfO2) capacitors by establishing a correlation between themacroscopic switching characteristics and the domain time−voltage-dependent behavior. It is shown that La:HfO2capacitors exhibit a much more pronounced imprint thantheir PZT counterparts. This can be explained by the differentphysical properties of the layers. Furthermore, in addition toconventional imprint developing with time in the capacitors setto a specific polarization state, an easily changeable imprint,termed as fluid imprint, with a strong dependence on theswitching prehistory and pulse train sequence, has beenobserved. In addition, domain structure visualization reveals aninertial switching effect, manifested by continued polarizationswitching in the direction of the applied electric field long afterthe termination of the external pulse. These observationshighlight the critical role played by the injected charges andmobile charges/defects in the switching behavior of FE HfO2-based devices.

2. RESULTS AND DISCUSSIONIn this work, the imprint phenomenon has been investigated in20 nm thick La:HfO2 capacitors with the iridium oxide (IrOx)top electrode (TE) and bottom electrode (BE). In thesestudies, the electrical bias was applied to the TE, while the BEwas grounded. The conventional polarization−voltage (P−V)hysteresis loops have been obtained by integrating thetransient switching currents. The pristine state is characterizedby a typical pinched hysteresis loop (Figure 1a, black line) with

very low remanent polarization (Pr). Structural analysis of thecapacitors (see Figure S1) by grazing incidence X-raydiffraction shows a higher volume fraction of non-FEmonoclinic and tetragonal phases in the pristine state,compared to the previously published case with TiNelectrodes.22 PFM imaging of the La:HfO2 capacitors in thepristine state does not reveal any sign of the domain structureafter application of ±6 V voltage pulses and yields a relatively

weak PFM amplitude signal (Figure 1c−f). These PFMobservations are consistent with a low volume fraction of theFE phase in the pristine state. After the capacitors weresubjected to the wake-up process4,23−26 (via application of 105

cycles of alternating square pulses of 7 V in amplitude at 10kHz), the P−V testing yielded a 2Pr value of ∼45 μC/cm2(Figure 1a, red curve), which points to a field-cycling inducedphase transformation from a non-FE to the FE phase.12,23,25

The increase in Pr is accompanied by the increased domainswitchability as seen from PFM imaging (Figure 1g−j). Asignificant difference in the amount of the switched polar-ization by positive and negative pulses might be due to theasymmetric boundary conditions at the top and bottominterfaces, resulting from the different growth conditions of theTE and BE or it could as well be due to the manifestation ofthe field-induced imprint as is discussed next.Investigation of the imprint effect has been carried out using

a pulse train sequence shown in Figure 1b. Before eachmeasurement, the capacitor was cycled 100 times using (±7 V,50 μs) pulses to erase the previous imprinted state (see FigureS2). Then, the capacitor was poled to a certain state byapplication of a set pulse followed by a triangular waveform of100 μs duration to measure the transient switching currents,which were subsequently integrated to obtain the P−Vhysteresis loops. Measurement of the P−V loops as a functionof the delay time td (see Figure 1b) after the set pulseapplication allows testing of the conventional time-dependentimprint. On the other hand, acquisition of the hysteresis loopsas a function of the set pulse parameters (duration andamplitude) provides information on the fluid imprint behavior,which is determined by sample switching prehistory. The term“fluid” is chosen to reflect the ease, with which the sign andmagnitude of imprint changes in the La:HfO2 capacitors uponswitching. The proposed approach allows us to investigate thedynamic behavior of fluid imprint and to differentiate it fromconventional imprint, which take place at different time scales.A time-dependent shift of the P−V loops observed during

investigation of conventional imprint is shown in Figure 2. Thecorresponding transient current−voltage (I−V) curves can befound in Figure S3. The capacitor was switched to a specificpolarization state by applying a set pulse with amplitude in therange from 4 to 6 V with a duration of 5 ms. The set pulseparameters have been chosen to ensure complete poling of thecapacitor, which was verified by PFM observation of theresulting single-domain structure. In Figure 2a, b, it can beclearly seen that the polarity of the set pulse, which switchesthe polarization either to the downward (toward the BE)(Pdown) or upward (Pup) direction, determines the sign of thevoltage shift defined as Vshift = (Vc+ + Vc−)/2, where Vc+ andVc− represent the positive and negative coercive voltages,respectively. A positive set pulse creates a preference for thePdown state leading to the P−V loop shift to the left, while anegative set pulse, which induces the Pup state, causes the shiftto the right.Figure 2c shows the highly nonlinear time dependence of

the voltage shift for both polarities of set pulses. Nearlysymmetric P−V loops have been observed right after the accycling. Subsequently, the imprint starts to evolve with Vshiftreaching a value of 1.8 V (0.9 MV/cm) within 24 h. Thisbehavior is similar to that reported for PZT-basedcapacitors14,15 except that the magnitude of imprint,normalized to the coercive voltage, is nearly 3−4 times largerin the La:HfO2 capacitors so that their total voltage shift could

Figure 1. (a) P−V loops for the La:HfO2 capacitor in the pristinestate (black) and after the wake-up process (red). (b) Pulse trainsequence used for investigation of the imprint behavior. τ denotes theduration of the set pulse while td denotes the time delay between theset pulse and the triangular waveform. (c−j) PFM amplitude (c,e,g,i)and PFM phase (d,f,h,j) images of the La:HfO2 capacitor in thepristine state (c−f) and after wake-up (g−j). All measurements weredone on a 160 μm-diameter capacitor.

ACS Applied Materials & Interfaces Research Article

DOI: 10.1021/acsami.9b11146ACS Appl. Mater. Interfaces 2019, 11, 35115−35121

35116

http://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://dx.doi.org/10.1021/acsami.9b11146

be as large as the coercive voltage. (A tentative explanation ofthis effect based on the difference in the depolarizing fields ofthe HfO2 and PZT capacitors is proposed in the SupportingInformation ,Section S6 and Figure S6). It is interesting tonote that the set pulse amplitude has only a slight impact on

the voltage shift as is shown in Figure S4. (In addition, we notethat imprint in the HfO2-based capacitors could be reduced byreplacing the IrOx electrodes with the TiN ones as shown inFigure S5. The effect of the electrode material on the imprintbehavior is a subject of separate studies.)Previously, to explain the time-dependent behavior of

imprint in PZT, Grossmann et al.14,15 proposed an interfacescreening model (ISM), according to which the presence of aninterfacial passive layer16 between the FE thin film and theelectrode results in incomplete screening of the polarizationcharges by the charges on the electrodes. When such acapacitor is poled, space separation between the boundpolarization charges and the electrode charges leads to astrong electric field across the passive layer. This fieldpromotes charge transport into the layer27 and followed bycharge accumulation and trapping at the FE/passive layerinterface.28 If the de-trapping time constant of the accumulatedcharges is much longer than the polarization switching time, aninternal bias field due to these charges will manifest as avoltage shift in the P−V loop.29 Tagantsev and Gerra16 thenput forward an analytical treatment of the problem along thelines of the ISM based on charge injection into the passivelayer from the electrode. They found a universal logarithmic-type time dependence of the voltage shift for an exponentialinjection current into the passive layer and for weakpolarization screening. This expression, given by Vshift = V0ln(1 + t/τ0), where V0 and τ0 are the logarithmic slope and thecrossover time between the linear and logarithmic chargerelaxation, respectively, gives an excellent fit to the nonlineartime dependence of Vshift observed in La:HfO2 capacitors(Figure 2c), highlighting a critical role played by the passivelayer in development of imprint.

Figure 2. (a−c) Development of conventional imprint in the La:HfO2capacitors as a function of the delay time td between the set pulse andthe measurements. (a,b) P−V loops after (+6 V, 5 ms) (a) and (−6 V,5 ms) (b) set pulse application. The arrows indicate the direction ofthe voltage shift. (c) Voltage shift as a function of time [calculatedfrom the loops in (a,b)] for both polarities. The blue curves representfit of the experimental data by the logarithmic law Vshift = V0 ln(1 + t/τ0). (d) Effect of the testing pulse sequence on the C−V loopssuggesting that the observed voltage shift can be a result of themeasurement process itself. The measurements were performed onthe 90 μm-diameter capacitors.

Figure 3. (a) Pulse train sequence used for investigation of the imprint effects. (b,c) Transient I−V curves for increasing τ but fixed td = 10 μs (b),and for fixed τ = 100 ns but increasing td (c). The amplitude of the preset and the set pulses was 6 V. The preset pulse duration was 5 ms. For thetriangular waveform the amplitude was 7 V with a 100 μs period. The arrows in (b,c) indicate the direction of the initial voltage change. Thevertical red lines in (b,c) serve as a visual reference to compare the shift in the switching current peaks upon changing τ (b) and td (c). (d)Illustration of the fluid imprint mechanism depicting the charge distribution after application of the set pulses of different duration and constant 10μs delay and showing the stabilization of the switched polarization due to the charge injection and entrapment at the dielectric/FE layer interface.(e) Illustration of the conventional imprint mechanism depicting the charge distribution at the interfaces after application of the set pulses of fixed100 ns duration and different delay times and showing stabilization of opposite polarization state due to the redistribution of charges across the FElayer. The measurements were performed on the 90 μm-diameter capacitors.



35117

http://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://dx.doi.org/10.1021/acsami.9b11146

The small signal C−V measurements provided a first glimpseof the fluid imprint as explained below. Figure 2d shows ahorizontal shift of the C−V loops, which depends on thesequence of the testing pulses, or switching prehistory. It canbe seen that although the starting and the ending points of theloops at 0 V are the same for both pulse train sequences, theC−V loop shifts to the left if the testing waveform ends with anegative bias and to the right if it ends with a positive bias. Theformer suggests that the capacitor structure is quite symmetric,while the latter is a strong indication that the voltage shift ofthe C−V loops occurs only during testing, that is, is a result ofthe measurement process itself.To gain a deeper insight into the dynamic behavior of the

fluid imprint, the I−V measurements have been carried out as afunction of the set pulse duration τ (with the delay time tdfixed at 10 μs). For these studies, the pulse train was slightlymodified in comparison to the one shown in Figure 1b: afterthe ac cycling, a preset pulse of sufficiently high amplitude andduration is applied to induce complete poling of the capacitorfollowed by application of the set pulse (Figure 3a). Results ofthe fluid imprint testing are shown in Figure 3b (note that thetriangular waveform starts with the negative bias). It can beseen that an increase in the positive set pulse duration resultsin the shift of the positive switching current peak to the leftalong with the appearance of a negative current peak, which isindicative of the gradual development of the preference for thePdown state.For comparison, results of the I−V measurements as a

function of the delay time td between the set pulse and thetriangular waveform, which tests the conventional imprintbehavior, are shown in Figure 3c (the set pulse duration isfixed at 100 ns). It can be seen that, similar to the effect of theset pulse duration, increase in the delay time td shifts thenegative current peak to the left. However, in addition to this,there is a split of the positive current peak into two peaks,which move further apart upon increase in the delay time. Thisbehavior can be explained in the following way. First,application of the negative preset pulse switches the polar-ization to the upward direction, which gets stabilized due tothe injected charges, leading to an imprinted Pup state. Thisstate is almost unperturbed by the shortest positive set pulse of50 ns (Figure 3d) as is indicated by the asymmetric I−V curve(Figure 3b)there is only a positive current peak resultingfrom the switching of the Pup to Pdown state under a positivehalf of the triangular waveform. Next, an increase in the setpulse duration to 120 ns leads to partial polarization reversaland formation of the polydomain structure (Figure 3d). As aresult, the current peaks start to appear under a negative half ofthe triangular waveform due to the switching from the Pdown toPup state (Figure 3b). Charge injection during application ofthe positive set pulse creates an internal electric field pointingtoward the BE, thereby favoring the downward polarization. Asa result, we observe a shift of the switching current peaks to theleft (toward negative voltage). Note that if we use a delay timemuch longer than 10 μs used in Figure 3b to allow relaxationof the injected charge, then redistribution of the free chargecarriers due to the induced polydomain state would screen andstabilize domains of both polarities (Figure 3e). Becausecharge redistribution/de-trapping is much slower than theduration of the triangular waveform, the I−V testing itself doesnot disturb the screening field configuration of the polydomainstate (Figure 3e). This means that the screening fields ofalternating directions (oriented upward for the Pup domains

and downward for the Pdown domains in Figure 3e) should leadto the splitting of the positive current peak: the regions withthe downward internal field will switch at lower bias than theregions with the upward field. This is exactly what is observedin Figure 3c. The longer the delay time, the larger is theinternal field (Figure 3e), which implies that the two positivecurrent peaks will move further apart. This split-up tendency isalready visible in Figure 3b where the width of the set-pulsewas varied at the lowest delay time. This proves that the chargeinjection effect is also impacted by the polarity of the domains.Note that the current peak splitting is consistent with theobservation of the multiple current peaks using field cyclingwith nonsaturating pulse trains (sub-cycling).5,6 If long setpulses (>300 ns) were used so that all domains would beswitched, then the internal field would be of the same directionand no current peak splitting would occur.Thus, the La:HfO2 capacitors exhibit two distinctly different

imprint effects in response to the change in the delay time td(conventional imprint) and in the set pulse duration τ (fluidimprint). A simplified band diagram illustrating these twoeffects in the FE capacitor with the interface passive layers isshown in Figure 4 (see also in Figure S7). As explained above,

separation between the bound polarization charges and theelectrode charges leads to an electric field across the passivelayer. At the same time, there is an oppositely aligneddepolarizing field formed across the FE layer. At V = 0, boththe interfaces and the FE therefore appear tilted. This can giverise to two different routes of imprint: (1) electron injectioninto the passive layer due to the interface potential barrierlowering during the electric field application and subsequententrapment at the FE/passive layer interface. This processaccounts for fluid imprint and could occur through Schottkyemission or direct tunneling into the first trap state duringpulse application;17 (2) electron migration across the FE layervia Poole−Frenkel emission or trap-assisted tunneling30 andentrapment near the FE/passive layer interface or at the defectsites within the interfacial layers or grain boundaries within theFE film causing conventional imprint. This effect can beaccelerated if the opposite bias is applied to the capacitor aslong it is lower than the threshold switching voltage (FigureS7). This mechanism explains why the electric field cyclingcould give rise to the local imprint of different polarities.5

The soundness of the injected charge mechanism and itsimpact has been particularly evident when investigating thedomain structure evolution as a function of time after pulseapplication, which reveals an interesting effect (Figure 5). The

Figure 4. Simplified band diagram illustrating mechanisms of fluidand conventional imprint. CB, VB, EF and Egap stand for theconduction band, valence band, Fermi level and bandgap energy,respectively. See details in the text.



35118

http://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://dx.doi.org/10.1021/acsami.9b11146

capacitor was subjected to the same preliminary treatment asfor the imprint measurements (Figure 3). Then, the capacitorwas imaged by PFM: first, after application of the positivepreset pulse (6 V, 5 ms), which yielded complete poling intothe Pdown state (Figure 5a,b), and, next, after the negative setpulse (−3.5 V, 1 ms) application, which resulted in partialswitching to the Pup state (Figure 5c,d). The capacitor wasthen left in this state for 36 h with no additional bias applied,while its polydomain structure was monitored by PFM. It wasfound that the domains continued to switch to the Pupdirection long after the set pulse was applied (Figure 5e,f).Histogram analysis of the PFM phase images (Figure 5g)shows that additional 20% of the polarization has switched tothe Pup state during that period in the absence of the externalfield. This continuous switching in the same direction as thepreviously applied field is reminiscent of the inertial motion ofdomain walls and, hence, is termed as “inertial switching”. Theinertial switching has been also observed after the positive setpulse application; in this case, to the Pdown state (see FigureS8).That this is indeed switching and not back-switching due to

the pre-existing internal field is confirmed by the P−V loopmeasurements performed right before and 36 h afterapplication of the negative set pulse (Figure 5h). The initialloop (blue curve in Figure 5h) is fairly centered (similar to theP−V loops for short delay times in Figure 2a) indicating almostno imprint, while the loop acquired 36 h after the negative setpulse is shifted to the right (red curve in Figure 5h)asignature of the Pup imprint. Note that stabilization of thepolydomain structure due to conventional imprint wouldtypically lead to a pinched P−V loop.31 Finally, to make sure

that the observed behavior is not an artifact due to the PFMimaging process, several continuous PFM scans have beentaken. No significant change in the domain structure wasdetected, confirming that the data in Figure 5c−f obtained by atime-dependent process was not affected by PFM imaging.Based on the collection of these data, it can be concluded thatthe inertial switching is one of the manifestations of theinjected charge entrapment.

3. CONCLUSIONS

In conclusion, a combination of the pulse switching techniqueand high-resolution domain imaging by means of PFM hasbeen used to investigate the time-dependent and switching-history-induced imprint in FE La:HfO2 capacitors. Thisapproach allowed us to establish correlation between develop-ment of imprint and domain switchability. It has been shownthat the La:HfO2 capacitors exhibit a much more pronouncedimprint than the conventional PZT-based FE capacitors. Also,in addition to conventional time-dependent imprint occurringwith time in the poled capacitors, an easily changeable fluidimprint with a strong dependence on the switching prehistoryhas been observed. Our studies suggest that while theredistribution of the mobile charges under the depolarizingfield and subsequent entrapment at the film−electrodeinterface is the root cause of the conventional imprint, chargeinjection into and transport across the interface dielectric layeris the main mechanism for the fluid imprint, which occursduring the external voltage application. An interestingconsequence of this effect is inertial switchinga very slowcontinuation of domain switching in the direction of thepreviously applied field even after the field was turned off.From an application point of view, imprint can be a seriouslimitation factor for the memory device performance. Thus, anoptimization of the interface structural properties is required tominimize or eliminate its detrimental effectsimilar to what isnecessary for alleviation of the wake-up effect and retentionloss.

4. EXPERIMENTAL SECTION4.1. Sample Preparation. For electrical characterization, the

La:HfO2 films were integrated in a capacitor structure with the 25 nmthick iridium oxide (IrOx) BE and TE sputtered in a BESTECphysical vapor deposition tool at room temperature. IrOx wasdeposited on silicon substrates with 200 nm thermally grown SiO2.Electrodes were fabricated using the process at a pressure of 1 μbar,and 20:5 Ar/O2 ratio to obtain a typical resistivity value of about 100μΩ/cm2. The capacitor areas were defined by evaporating IrOxthrough a shadow mask. Lanthanum-doped hafnium oxide filmswere deposited in an Oxford Instruments OpAL ALD tool by atomiclayer deposition (ALD) as described elsewhere.21 The HfO2 filmswere doped by replacing every tenth HfO2 cycles by La2O3 cyclesresulting in a La content of about 10 mol %. The IrO2 BE and thecomplete IrO2/La:HfO2/IrO2 was annealed in O2 atmosphere at 600°C for 10 min to ensure high oxygen content in all layers.

4.2. Characterization. A microscopic external probe in contactwith the TE was used to apply voltage using a Keysight 33621Aarbitrary waveform generator. A positive voltage leads to a Pdown state,while a negative voltage resulted in the Pup state. The switchingcurrent signals were recorded by a Tektronix TDS 3014Boscilloscope. PFM imaging has been carried out using a commercialatomic force microscope system (MFP-3D, Asylum Research) withsingle-crystalline diamond tips (D80, K-Tek Nanotechnology)employing a 330 kHz AC voltage with 0.65 V amplitude.Capacitance−voltage (C−V) measurements were performed using a

Figure 5. Observation of inertial switching in the La:HfO2 capacitor.PFM amplitude (a,c,e) and phase (b,d,f) images after application ofthe (6 V, 5 ms) preset pulse (a,b), after application of the (−3.5 V, 1ms) set pulse (c,d), and 36 h later (e,f). It can be seen that thecapacitor continues to switch even though there was no additionalexternal bias applied after the set pulse. The dark (bright) regionscorrespond to Pup (Pdown). (g) Histogram analysis of the phase imagesin (d,f) showing evolution of the polarization toward Pup afterapplication of the set pulse. (h) P−V loops acquired beforeapplication of the (−3.5 V, 1 ms) set pulse (blue) and 36 h later(red), which show development of a strong preference for the Pupstate.



35119

http://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfhttp://dx.doi.org/10.1021/acsami.9b11146

commercial Radiant Precision SC tester. All measurements wereperformed at room temperature.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acsami.9b11146.

Structural characterization of La:HfO2 capacitors withIrOx electrodes; erasure of imprint in La:HfO2capacitors by electrical cycling; effect of delay time oncurrent−voltage characteristics in La:HfO2 capacitorswith IrOx electrodes; effect of set pulse amplitude onconventional imprint; comparison of the wake-upprocess and imprint behavior in La:HfO2 capacitorswith IrOx and TiN electrodes; comparative analysis ofthe depolarizing fields in HfO2 and PZT capacitors;development of fluid and conventional imprints; andinertial switching to the Pdown state (PDF)

■ AUTHOR INFORMATIONCorresponding Author*Email: [email protected] Schroeder: 0000-0002-6824-2386Alexei Gruverman: 0000-0003-0492-2750NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSP.B. and A.G. acknowledge support by the National ScienceFoundation through Materials Research Science and Engineer-ing Center under grant no. DMR-0820521. C.R. receivedfunding from the European Union’s Horizon 2020 researchand innovation program under grant agreement no. 780302(project 3εFerro). P.D.L. is funded by the German Ministry ofEconomic Affairs and Energy (BMWi) project (16IPCEI310).T.S. gratefully acknowledges the German Research Foundation(Deutsche Forschungsgemeinschaft) for funding part of thisresearch at NaMLab in the frame of the “Inferox” project (MI1247/11-2).

■ REFERENCES(1) Böscke, T. S.; Müller, J.; Braühaus, D.; Schröder, U.; Böttger, U.Ferroelectricity in Hafnium Oxide Thin Films. Appl. Phys. Lett. 2011,99, 102903.(2) Scott, J. F. Ferroelectric Memories; Springer: Berlin, 2000.(3) Iñ́iguez, J.; Zubko, P.; Luk’yanchuk, I.; Cano, A. FerroelectricNegative Capacitance. Nat. Rev. Mater. 2019, 4, 243−256.(4) Park, M. H.; Lee, Y. H.; Kim, H. J.; Kim, Y. J.; Moon, T.; Kim, K.D.; Müller, J.; Kersch, A.; Schroeder, U.; Mikolajick, T.; Hwang, C. S.Ferroelectricity and Antiferroelectricity of Doped Thin HfO2-BasedFilms. Adv. Mater. 2015, 27, 1811−1831.(5) Schenk, T.; Schroeder, U.; Pesǐc,́ M.; Popovici, M.; Pershin, Y.V.; Mikolajick, T. Electric Field Cycling Behavior of FerroelectricHafnium Oxide. ACS Appl. Mater. Interfaces 2014, 6, 19744−19751.(6) Schenk, T.; Hoffmann, M.; Ocker, J.; Pesǐc,́ M.; Mikolajick, T.;Schroeder, U. Complex Internal Bias Fields in Ferroelectric HafniumOxide. ACS Appl. Mater. Interfaces 2015, 7, 20224−20233.(7) Fengler, F. P. G.; Pesǐc,́ M.; Starschich, S.; Schneller, T.;Künneth, C.; Böttger, U.; Mulaosmanovic, H.; Schenk, T.; Park, M.H.; Nigon, R.; Muralt, P.; Mikolajick, T.; Schroeder, U. DomainPinning: Comparison of Hafnia and PZT Based Ferroelectrics. Adv.Electron. Mater. 2017, 3, 1600505.

(8) Fengler, F. P. G.; Hoffmann, M.; Slesazeck, S.; Mikolajick, T.;Schroeder, U. On the Relationship Between Field Cycling andImprint in Ferroelectric Hf0.5Zr0.5O2. J. Appl. Phys. 2018, 123, 204101.(9) Carl, K.; Hardtl, K. H. Electrical After-Effects in Pb(Ti,Zr)O3Ceramics. Ferroelectrics 1977, 17, 473−486.(10) Arlt, G.; Neumann, H. Internal Bias in Ferroelectric Ceramics:Origin and Time Dependence. Ferroelectrics 1988, 87, 109−120.(11) Lohkam̈per, R.; Neumann, H.; Arlt, G. Internal Bias inAcceptor-Doped BaTiO3 ceramics: Numerical evaluation of increaseand decrease. J. Appl. Phys. 1990, 68, 4220.(12) Pesǐc,́ M.; Fengler, F. P. G.; Larcher, L.; Padovani, A.; Schenk,T.; Grimley, E. D.; Sang, X.; LeBeau, J. M.; Slesazeck, S.; Schroeder,U.; Mikolajick, T. Physical Mechanisms behind the Field-CyclingBehavior of HfO2-Based Ferroelectric Capacitors. Adv. Funct. Mater.2016, 26, 4601−4612.(13) Warren, W. L.; Dimos, D.; Pike, G. E.; Tuttle, B. A.; Raymond,M. V.; Ramesh, R.; Evans, J. T. Voltage shifts and imprint inferroelectric capacitors. Appl. Phys. Lett. 1995, 67, 866.(14) Grossmann, M.; Lohse, O.; Bolten, D.; Boettger, U.; Schneller,T.; Waser, R. The Interface Screening Model as Origin of Imprint inPbZrxTi1‑xO3 Thin Films. I. Dopant, Illumination, and Bias Depend-ence. J. Appl. Phys. 2002, 92, 2680.(15) Grossmann, M.; Lohse, O.; Bolten, D.; Boettger, U.; Waser, R.The Interface Screening Model as Origin of Imprint in PbZrxTi1‑xO3thin films. II. Numerical Simulation and Verification. J. Appl. Phys.2002, 92, 2688.(16) Tagantsev, A. K.; Gerra, G. Interface-Induced Phenomena inPolarization Response of Ferroelectric Thin Films. J. Appl. Phys. 2006,100, 051607.(17) Tagantsev, A. K.; Stolichnov, I.; Setter, N.; Cross, J. S. Natureof Nonlinear Imprint in Ferroelectric Films and Long-TermPrediction of Polarization Loss in Ferroelectric Memories. J. Appl.Phys. 2004, 96, 6616.(18) Tan, Z.; Tian, J.; Fan, Z.; Lu, Z.; Zhang, L.; Zheng, D.; Wang,Y.; Chen, D.; Qin, M.; Zeng, M.; Lu, X.; Gao, X.; Liu, J.-M.Polarization Imprint Effects in the Photovoltaic Effect in Pb(Zr,Ti)O3Thin Films. Appl. Phys. Lett. 2018, 112, 152905.(19) Warren, W. L.; Pike, G. E.; Vanheusden, K.; Dimos, D.; Tuttle,B. A.; Robertson, J. Defect-Dipole Alignment and Tetragonal Strain inFerroelectrics. J. Appl. Phys. 1996, 79, 9250.(20) Dimos, D.; Warren, W. L.; Sinclair, M. B.; Tuttle, B. A.;Schwartz, R. W. Photoinduced Hysteresis Changes and OpticalStorage in (Pb,La)(Zr,Ti)O3 Thin Films and Ceramics. J. Appl. Phys.1994, 76, 4305.(21) Lee, D.; Yoon, A.; Jang, S. Y.; Yoon, J.-G.; Chung, J.-S.; Kim,M.; Scott, J. F.; Noh, T. W. Giant Flexoelectric Effect in FerroelectricEpitaxial Thin Films. Phys. Rev. Lett. 2011, 107, 057602.(22) Schroeder, U.; Richter, C.; Park, M. H.; Schenk, T.; Pesǐc,́ M.;Hoffmann, M.; Fengler, F. P. G.; Pohl, D.; Rellinghaus, B.; Zhou, C.;Chung, C.-C.; Jones, J. L.; Mikolajick, T. Lanthanum-Doped HafniumOxide: A Robust Ferroelectric Material. Inorg. Chem. 2018, 57, 2752−2765.(23) Mueller, S.; Adelmann, C.; Singh, A.; Van Elshocht, S.;Schroeder, U.; Mikolajick, T. Ferroelectricity in Gd-Doped HfO2Thin Films. ECS J. Solid State Sci. Technol. 2012, 1, N123.(24) Zhou, D.; Xu, J.; Li, Q.; Guan, Y.; Cao, F.; Dong, X.; Müller, J.;Schenk, T.; Schröder, U. Wake-up Effects in Si-doped Hafnium OxideFerroelectric Thin Films. Appl. Phys. Lett. 2013, 103, 192904.(25) Park, M. H.; Kim, H. J.; Kim, Y. J.; Lee, Y. H.; Moon, T.; Kim,K. D.; Hyun, S. D.; Hwang, C. S. Study on the Size Effect inHf0.5Zr0.5O2 Films Thinner than 8 Nm Before and After Wake-UpField Cycling. Appl. Phys. Lett. 2015, 107, 192907.(26) Grimley, E. D.; Schenk, T.; Sang, X.; Pesǐc,́ M.; Schroeder, U.;Mikolajick, T.; LeBeau, J. M. Structural Changes Underlying Field-Cycling Phenomena in Ferroelectric HfO2 Thin Films. Adv. Electron.Mater. 2016, 2, 1600173.(27) Tagantsev, A. K.; Stolichnov, I. A. Injection-Controlled SizeEffect on Switching of Ferroelectric Thin Films. Appl. Phys. Lett. 1999,74, 1326.



35120

http://pubs.acs.orghttp://pubs.acs.org/doi/abs/10.1021/acsami.9b11146http://pubs.acs.org/doi/suppl/10.1021/acsami.9b11146/suppl_file/am9b11146_si_001.pdfmailto:[email protected]://orcid.org/0000-0002-6824-2386http://orcid.org/0000-0003-0492-2750http://dx.doi.org/10.1021/acsami.9b11146

(28) Li, P.; Huang, Z.; Fan, Z.; Fan, H.; Luo, Q.; Chen, C.; Chen,D.; Zeng, M.; Qin, M.; Zhang, Z.; Lu, X.; Gao, X.; Liu, J.-M. AnUnusual Mechanism for Negative Differential Resistance in Ferro-electric Nanocapacitors: Polarization Switching-Induced ChargeInjection Followed by Charge Trapping. ACS Appl. Mater. Interfaces2017, 9, 27120−27126.(29) Schorn, P.; Ellerkmann, U.; Bolten, D.; Boettger, U.; Waser, R.Non-linear Imprint Behavior of PZT Thin Films. Integr. Ferroelectr.2003, 53, 361−369.(30) Fan, Z.; Xiao, J.; Wang, J.; Zhang, L.; Deng, J.; Liu, Z.; Dong,Z.; Wang, J.; Chen, J. Ferroelectricity and Ferroelectric ResistiveSwitching in Sputtered Hf0.5Zr0.5O2 Thin Films. Appl. Phys. Lett. 2016,108, 232905.(31) Schenk, T.; Yurchuk, E.; Mueller, S.; Schroeder, U.; Starschich,S.; Böttger, U.; Mikolajick, T. About the Deformation of FerroelectricHystereses. Appl. Phys. Rev. 2014, 1, 041103.



35121
http://dx.doi.org/10.1021/acsami.9b11146

Documents

Fluid Imprint and Inertial Switching in Ferroelectric La ...Fluid Imprint and Inertial Switching in Ferroelectric La:HfO2 Capacitors Pratyush Buragohain,† Adam Erickson,‡ Pamenas