Ferroelectricity in Hf0.5Zr0.5O2 Thin Films: A Microscopic Study of the

Feb 21, 2018 - PFM data combined with the structural analysis of pristine versus trained film by plan-view transmission electron microscopy both speak...
0 downloads 3 Views 14MB Size
Research Article www.acsami.org

Cite This: ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Ferroelectricity in Hf0.5Zr0.5O2 Thin Films: A Microscopic Study of the Polarization Switching Phenomenon and Field-Induced Phase Transformations Anastasia Chouprik,*,† Sergey Zakharchenko,† Maxim Spiridonov,† Sergei Zarubin,† Anna Chernikova,† Roman Kirtaev,† Pratyush Buragohain,‡ Alexei Gruverman,†,‡ Andrei Zenkevich,† and Dmitrii Negrov† †

Moscow Institute of Physics and Technology, 9 Institutskiy Lane, Dolgoprudny, Moscow Region 141700, Russia Department of Physics and Astronomy, University of Nebraska, Lincoln, Nebraska 68588, United States



S Supporting Information *

ABSTRACT: Because of their full compatibility with the modern Si-based technology, the HfO2-based ferroelectric films have recently emerged as viable candidates for application in nonvolatile memory devices. However, despite significant efforts, the mechanism of the polarization switching in this material is still under debate. In this work, we elucidate the microscopic nature of the polarization switching process in functional Hf0.5Zr0.5O2-based ferroelectric capacitors during its operation. In particular, the static domain structure and its switching dynamics following the application of the external electric field have been monitored with the advanced piezoresponse force microscopy (PFM) technique providing a nm resolution. Separate domains with strong built-in electric field have been found. Piezoresponse mapping of pristine Hf0.5Zr0.5O2 films revealed the mixture of polar phase grains and regions with low piezoresponse as well as the continuum of polarization orientations in the grains of polar orthorhombic phase. PFM data combined with the structural analysis of pristine versus trained film by plan-view transmission electron microscopy both speak in support of a monoclinic-to-orthorhombic phase transition in ferroelectric Hf0.5Zr0.5O2 layer during the wake-up process under an electrical stress. KEYWORDS: hafnium oxide, ferroelectric switching, domain structure, polycrystalline ferroelectric films, piezoresponse force microscopy



INTRODUCTION

and from the FeRAM fabrication viewpoint this property of FEHZO films is advantageous. Despite the high interest in FE-HfO2 and prototype memory devices based on it, the exact mechanism of the polarization switching in this material is still not fully understood. The transformation of domain structure in HZO films during alternating current (ac) electric cycling of the FE capacitor (training process, also called “wake-up” effect14,16,18−23) is also debated. It is well established that in the HfO2-based FE films, electric field cycling leads to the increase in the Pr value.21 Several mechanisms have been suggested previously to explain the origin of such effect. In particular, Zhou et al.14 suggested that the wake-up effect originates from the depinning of domains due to the reduction of the defect concentration near the TiN electrode. Alternatively, based on the transmission electron microscopy (TEM) study, Grimley et al.,19 Pešić et al.,16 and Martin et al.18 attributed the wake-up effect to the phase transition from the monoclinic to orthorhombic phase, whereas Lomenzo et al.23 suggested the tetragonal-to-

Few years ago, ferroelectric (FE) properties were discovered in polycrystalline HfO2 thin (∼10 nm) films when doped with La, Si, Zr, Y, Al, or Gd.1−10,16,19 Well before that, the HfO2 films were already used in modern complementary metal-oxide semiconductor and other microelectronics processes as a high-k gate material.11 It is therefore appealing to use HfO2-based FE materials in nonvolatile memory devices, such as 1T−1C ferroelectric random access memory (FeRAM) or flashlike ferroelectric field-effect transistors. Both theoretical12,13 and experimental14−19 studies ascribe the ferroelectricity in polycrystalline HfO2-based thin films to the presence of a metastable noncentrosymmetric orthorhombic HfO2 phase (space group Pbc21), which crystallizes during the annealing of doped HfO2 thin films. Alternatively, it has been demonstrated that alloyed FE-Hf0.5Zr0.5O2 (HZO) films with remnant polarization of Pr ∼ 15 μC/cm2 can be produced during the growth of the TiN top electrode by thermal atomic layer deposition (ALD) at T = 400 °C, without any postdeposition annealing.17 It is worth noting that the temperature required to crystallize HZO in the FE phase is significantly lower as compared to the doped hafnium oxide, © 2018 American Chemical Society

Received: November 16, 2017 Accepted: February 21, 2018 Published: February 21, 2018 8818

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces

Figure 1. Schematic view of TiN/HZO/TiN capacitor structures used for the microscopic study of the polarization switching dynamics by the PFM: (a) scheme of electrical wiring and connecting, (b) zoomed view of patterned top TiN electrode upon scanning.

provide an insight into the local switching behavior and should be verified by the direct microscopic studies of the domain structure dynamics upon application of the external voltage. Because ferroelectricity in FE-HfO2 films strongly depends on the interfaces with electrodes,7,19,25 the microscopic study of bare films would be of small value. To reveal the actual local ferroelectric properties, it is important to perform the study of functional capacitors comprising FE-HfO2. In this work, we correlate the information on the domain maps (local piezoresponse) of the HZO film with the structural study by the transmission electron microscopy to explain the wake-up effect in the HZO film. Further, we observe the domain structure evolution to elucidate the microscopic nature of polarization reversal in functional Hf0.5Zr0.5O2-based ferroelectric capacitors during its operation.

orthorhombic phase transition as the origin of the wake-up effect based on the concurrent increase in Pr and the decrease in the dielectric constant. To elucidate the polarization switching dynamics and domain structure transformations in ferroelectric following its electrical cycling, it is important to use a visualization tool providing a nanometer resolution. The inhomogeneity of FE properties in polycrystalline hafnia films is associated with the built-in electric fields due to the charged defects both in the bulk and at the interface.16,19−22 To analyze the evolution of such built-in electric fields as well as the coercive voltage during wake-up effect, the first-order reversal curves (FORCs) technique has been utilized previously.16,19,22,24 Phenomenological models of the wake-up process based on the FORC results22,16 assume that the pristine FE-HfO2 films are characterized by the presence of domains with opposite built-in fields. As shown by Schenk et al.22 and Pešić et al.,16 the ac cycling leads to the disappearance of these built-in fields, which is explained by the redistribution of the trapped charges. Because the FORC technique provides information only about built-in fields averaged across the structure, the proposed models do not



RESULTS AND DISCUSSION The 10 nm thick HZO films with 18 nm thick top and bottom TiN electrodes were grown by atomic layer deposition (ALD) technique on the Si substrate as described in detail elsewhere14 (see also the Supporting Information, Section S1). Crystal8819

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces

Figure 2. SAED patterns of cross-section samples: (a) pristine TiN/HZO/TiN structure; (b) trained TiN/HZO/TiN structure. m, t, o: monoclinic (space group P21/c), tetragonal (P42/nmc), and orthorhombic (Pbc21) ZrO2 phases, respectively. The most intense monoclinic phase reflections, 1̅11 and 111, are absent in the SAED pattern of trained structure. It should be noted that the rings marked as “TiN” and “Pt” can also contain various reflections of m-, t-, and o-phases of HZO (both Pt and TiN layers are left to protect HZO layer during sample preparation with the ion beam milling).

lization of the as-grown HZO films in the orthorhombic ferroelectric, along with the stable monoclinic, phases occurred during the ALD growth of the top electrode at T = 400 °C. To allow external electric biasing of the samples, the functional ferroelectric capacitor devices 100 × 100 μm2 in area were patterned on the Si chip with contact pads (details of the device fabrication are described in the Supporting Information, Section S1). The schematic view of the samples used for the microscopic study of the polarization switching dynamics by the piezoresponse force microscopy (PFM) technique is shown in Figure 1. To wake up pristine ferroelectric capacitors, they were cycled 104 times by applying voltage rectangular pulses with ±3 V amplitude and 100 μs duration with the pulse generator. Further, we call these capacitors as trained. The structural properties of pristine versus trained HZO films in capacitors were investigated by transmission electron microscopy (TEM). The selected-area electron diffraction (SAED) patterns of TiN/HZO/TiN stack cross sections prepared by focused ion beam were acquired to elucidate possible structural changes (for details see the Supporting Information, Section S2). SAED patterns of TiN/HZO/TiN structures are shown in Figure 2 and confirm the polycrystalline structure of both pristine and trained HZO film. The diffraction rings from polycrystalline TiN layers as well as point reflexes from the monocrystalline Si substrate were used for precise calibration. The presence of hardly distinguishable polycrystalline tetragonal (t) and orthorhombic (o) phases is evident in both pristine and trained structures. However, the important difference is that the reflections on the SAED pattern marked as m1̅11 and m111 in Figure 2a, which can be unambiguously ascribed to the monoclinic (m) phase, are observed only in the pristine structure. We, therefore, conclude that the monoclinic phase is fully converted in o- and/or t-phase following electrical cycling. The obtained results on the field-induced structural transformations corroborate the previously published comprehensive studies of Si-doped HfO2 by Martin et al.18 and18 Richter et al.25 and of Gd-doped HfO2 by Pešić et al.16 and Grimley et al.19

The prepared ferroelectric capacitor devices on the Si chip were subjected to standard electrical characterization before PFM measurements. Pulsed switching measurements of the TiN/HZO/TiN samples were performed using positive-up− negative-down (PUND) technique26,27 with triangular voltage sweeps,17 which allows to separate a current signal associated with the polarization reversal from the normal nonhysteretic displacement current (Supporting Information, Section S3). The sweep rate was 2 × 104 V/s. Figure 3a shows the I−V curves for trained TiN/HZO/TiN capacitors measured by the PUND method during “positive” (P) and “negative” (N) pulses (Supporting Information, Section S3). By integrating the switching current, one gets

Figure 3. (a) Switching P and N I−V curves taken from TiN/HZO/ TiN capacitor (capacitive U- and D-currents are not shown). Inset: the contribution of domains with preferred “up” and “down” polarizations to the I−V curve is sketched. (b) P−V hysteresis curves derived from PUND measurements. 8820

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces

Figure 4. Piezoresponse maps: (a, d) TiN surface morphology, (b, e) PFM amplitude map, and (c, f) PFM phase map for pristine HZO capacitor and the same after single voltage pulse −3 V, respectively.

the value of remnant polarization Pr ∼ 17 μC/cm2. The P−V curves derived from PUND data are shown in Figure 3b. “Relaxed” I−V curves measured during the first PUND voltage train are shifted with respect to “nonrelaxed” I−V curves measured during further cycling irrespective of the polarity of the first pulse (Figure 3a). By nonrelaxed I−V curves, we mean those obtained at further electric field cycling with the second and subsequent PUND voltage trains (Supporting Information, Section S3). The possible reason for such a shift is the relaxation of trapped charges in the polycrystalline HZO layer during some time after the application of PUND voltage pulses, which affects the internal electric field and, thereby, changes the coercive field. We call the devices with no biasing for more than 1 s as relaxed capacitors. During consecutive electric field cycling, the traps are equally filled and appropriate nonrelaxed I−V curves all coincide with each other. Another possible reason of relaxation is the depolarization of the HZO layer induced by unscreened ferroelectric field due to the presence of an isolating dead layer at the HZO/TiN interfaces.28 The coercive voltages for relaxed capacitors are more suitable for comparison with the PFM data. The mean coercive voltages for the relaxed capacitors for up and down polarizations were found to be −1.45 and 1.30 V, respectively. To get further insight into the ferroelectric properties of the HZO films and exclude the effect of interfacial electrochemical mechanisms, the PFM experiments were carried out on the patterned capacitors wired to the contact pads. In these experiments, the 100 × 100 μm2 top electrode pads (Figure 1b) were grounded, whereas the bottom electrode was biased (Figure 1a). This excitation scheme ensures zero potential between the grounded AFM tip and the sample and eliminates the electrostatic forces between the sample surface and the AFM tip. The described scheme also ensures that electrical stimulus does not depend on the quality of probe−sample contact and on the probe position. This helps to avoid misinterpretation and uncertainty in the analysis29−31 of the properties of new materials, such as hafnium-oxide-based ferroelectrics. In particular, all of the contributions to the background associated with the probe−sample interaction are

excluded. It should be noted that the minimal distinguishable size of a domain is defined by the thickness of the top TiN electrode.32 The piezoresponse mapping of pristine and trained TiN/ HZO/TiN structure was performed by resonance-enhanced dual ac amplitude resonance tracking (DART) PFM and combined band-excitation (BE) PFM33,34 and atomic force acoustic microscopy (BE AFAM) techniques, respectively, both in off-field mode. The choice is dictated by the characteristic features of these techniques. The BE PFM allows to acquire full spectra of the piezoresponse at each point of scan. For large vector averaging of the acquired spectra, the possible contribution of the AFM background in piezoresponse spectra can be detected. The fitting of the acquired spectra was performed by the model of harmonic oscillator with an additional shift linearly dependent on the frequency (Supporting Information, Section S4.1.3). The last contribution is intended to compensate for a possible AFM background and ultimately allows to obtain the most accurate values of piezoresponse amplitude and phase. The additional accuracy of measuring the piezoresponse amplitude is achieved by eliminating the effect of local contact stiffness, which is the topographical crosstalk, by normalizing the measured piezoresponse to the pure contact mechanical response (Supporting Information, Section S4.1.4). All of these precautions are nonexcess when studying properties of rather new materials such as hafnium-oxide-based ferroelectrics. For the reasons described above, the combined BE PFM/AFAM technique was chosen for most of the measurements. The only disadvantage of BE PFM is the slow scanning speed due to a prolonged averaging of the acquired piezoresponse spectra in each point of the scan. This is necessary due to the spreading of the exciting voltage power along multiple frequency harmonics of the spectrum, which results in a weak absolute piezoresponse. Therefore, for the study of materials with weak piezoelectric properties, an extremely prolonged scanning is required and the drift of the sample in a nonideal thermostatic box becomes critical. This problem can be overcome by DART PFM technique, in which the exciting voltage power is concentrated in just two frequency harmonics 8821

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces

Figure 5. Static domain structure of trained HZO film at the intermediate state of capacitor (−1.4 V) and mechanical contact response maps: (a) TiN surface morphology; (b) amplitude−frequency characteristics corresponding to points 1, 2, and 3 in (e); (c) BE PFM phase hysteresis loop; maps: (d) BE PFM amplitude, (e) BE PFM phase, (f) BE PFM parameter Q, (g) BE AFAM amplitude, (h) BE AFAM phase, (i) BE PFM contact resonance frequency, (j) normalized BE PFM/AFAM amplitude, (k) normalized BE PFM/AFAM phase, and (l) overlapping of normalized BE PFM/AFAM amplitude and phase. Image contrast of amplitude maps is improved by conventional quantile segmentation of images.

the same excitation and data acquisition settings was performed. The amplitude map (Figure 4e) reveals the significant increase in piezoresponse all over the area accompanied by a lock-in overload at some regions. The phase map (Figure 4f) demonstrates that the structure is fully downward polarized. Artifacts at the amplitude map due to a lock-in overload do not allow to perform careful analysis of data, but both the polarization reversal in ferroelectric regions and the transition of nonpolar regions in ferroelectric phase during the very first voltage pulse are visible unambiguously. Details of the cycle-by-cycle evolution of the HZO domain structure during wake-up process will be reported elsewhere. However, the significant increase in piezoresponse is evident even after the application of the single voltage pulse. Therefore, the PFM data reveal the presence of mixture of polar and nonpiezoelectric phases in the as-prepared polycrystalline HZO film, thus corroborating the TEM analysis. A microscopic study of polarization reversal was performed for trained functional HZO capacitors, whose ferroelectric properties remain stable during further cycling. A detailed analysis of the domain structure of the trained HZO film and its evolution was performed by the resonance-enhanced combined band-excitation (BE) PFM and atomic force acoustic microscopy (BE AFAM) techniques, which are homeimplemented in commercially available AFM Ntegra (NTMDT) using a digital signal processor with a wide-band voltage generator, analog-to-digital and digital-to-analog front ends (Nanoscan Technologies) (details of implementation of BE techniques are described in the Supporting Information,

and there is no need for high averaging. For this reason, we used the DART PFM to analyze pristine TiN/HZO/TiN structure. In general, the elimination of the topographical crosstalk by the combined BE PFM/AFAM technique increases the accuracy of the analysis of the PFM data. However, most qualitative conclusions, such as the consequences of wake-up process, can be drawn by the PFM analysis alone. For the piezoresponse mapping of pristine TiN/HZO/TiN structure, we used the DART PFM technique implemented in a commercially available AFM MFP-3D (Asylum Research). The driving voltage frequency was about 340 kHz and the amplitude was 0.6 V. The comparison between the amplitude and phase maps (Figure 4b,c) reveals that there are two types of regions in pristine HZO film: (i) isolated regions with a high amplitude and a binary PFM phase and (ii) surrounding regions with an almost zero amplitude and noisy phase. The first type of regions are likely attributed to the grains of polar o-phase with a high vertical component and both orientations of polarization vector, whereas latter ones can be associated with nonpiezoelectric phases or, at least, those with lower piezomodule d33. We can assume that these passive phases can be associated with the nonpolar m-phase, as well as with the t-phase and misaligned o-phase. Areas with a nonzero piezoresponse have a continuum of amplitude values (Figure 4b) and binary PFM phase (Figure 4c), thus revealing the continuum of polarization orientations in the o-phase grains of the pristine film. Further, the single voltage pulse with an amplitude −3 V and duration 400 μs was applied and piezoresponse mapping with 8822

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces

Figure 6. Domain structure dynamics of the trained TiN/HZO/TiN structure: (a−m) after the application of voltage pulse with different amplitude (blue arrows indicate sequence of voltage pulse applied in accordance with Figure S6); (n) ratio of the areas occupied by the domains with opposite polarization direction; (o) maximal normalized BE PFM amplitude. Black arrows indicate the location of domain seeds with the opposite direction of the polarization vector; white dashed and dotted lines indicate the localization of domains with preferred up and down polarization, respectively. Amplitude and phase scale bars are common to all of the maps.

pulse −1.4 V, 100 ms), we conclude that it is the topography that mainly contributes to the BE AFAM amplitude and phase maps. Variations in the contact area of the topographic hills and pits during scanning result in the contact resonance frequency variations (Figure 5i). Due to the inclination of the cantilever to the horizontal plane, there are different local contact areas on the different slopes of TiN crystallites depending on the orientation of the tip axis with respect to the local surface. This also affects the local contact resonance frequency, the resonance amplitude, and the phase.

Section S4). The following parameters were used: the central frequency near the contact resonance frequency was ∼670 kHz, the bandwidth was 97.7 kHz with 1000 frequency bins, and the peak-to-peak value of the exciting voltage was 1.0 V and 6 mV in the BE PFM and BE AFAM, respectively. The amplitude and phase of PFM and AFAM response in each point of scan were determined from the fitting experimental data with a vector fitting technique (Supporting Information, Section S4.1.3). Comparing BE PFM (Figure 5d−f), BE AFAM maps (Figure 5g,h) and TiN surface morphology (Figure 5a) for trained HZO capacitors at the intermediate state (after single voltage 8823

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces

relaxed structure measured by PUND (Figure 3). At the first glance, the difference from PUND results is that the visual BE PFM/AFAM phase changes start upon voltage pulses of −1.3 and 1.2 V (Figure 6b,h) corresponding to the near middle of the coercive voltages range revealed by PUND. However, the amplitude changes occur at lower voltages. For example, the dark areas (marked by black arrows) on the amplitude maps in Figure 6b,c,h become visible, although the BE PFM/AFAM phase did not change. The BE PFM/AFAM phase switching occurs upon later voltage pulses (Figure 6c,i). This is due to the fact that vertical and lateral sizes of the seed of the opposite domain remains smaller compared to the neighboring large domains. In this case, the BE PFM/AFAM phase of the vibrations is caused by neighbors, whereas the BE PFM/AFAM amplitude senses the emergence of the seed.35 The BE PFM/AFAM maps in Figure 6 do not contain any regions without the piezoresponse. Local decreases in the piezoresponse signal on the fully polarized HZO film (Figure 6a,g) can be associated with the o-phase seeds with opposite polarization vector direction, horizontally oriented domains, antiferroelectric t-phase grains, and passive m-phase grains less than 100 nm in size. In the FE capacitor due to mechanical coupling, electromechanical response with the same PFM phase should be registered on such regions. Another reason of the inhomogeneity of the amplitude is the process of the nucleation of domains with the opposite direction of the polarization vector, which becomes visible in the phase in Figure 6b,h,i. When comparing these data with the PFM data for pristine structures in Figure 4, it is evident that during the wake-up process, the amount of the nonferroelectric phase decreased dramatically. Taking into account the TEM analysis in Figure 2, we conclude that during the wake up of the TiN/HZO/TiN capacitors, the m- and t-phases undergo almost complete phase transition to the ferroelectric o-phase. The same effect was revealed by TEM for Si-doped HfO216,25 and Gd-doped HfO2.16,19 The maximum polarization in the HZO orthorhombic phase predicted from the DFT calculations is Pr ∼ 50 μC/cm2.12 Assuming a dispersion in the orientations of the polarization vector from the perpendicular direction, the difference between the maximum and the experimentally obtained polarization value of Pr ∼ 17 μC/cm2 is reasonable. Further, it should be noted that the maximal normalized BE PFM/AFAM amplitude is higher for the fully polarized HZO film (Figure 6o). Smaller amplitude for the intermediate domain structures is associated with the sum of the local electromechanical responses of small antiparallel domains. One should keep in mind that the passive 18 nm thick top TiN layer also increases this mechanical coupling. The maximal normalized PFM amplitude did not reach a saturation, especially with positive switching pulses. This means that the polarization reversal process did not end. There are significant number of domain residues with downward polarization. The minimum of the curves is achieved at the coercive voltage with an accuracy of 0.1 V, which corresponds to an equal number of domains and strongest mechanical coupling. Thus, for upward polarization, the coercive voltage is lower and simultaneously the built-in fields that prevent the polarization reversal are larger. In addition, the maps of the normalized amplitude for the fully polarized HZO film (Figure 6a,g) are nonuniform despite the in-phase and antiphase piezoresponse all over the area. It could be associated with the domain residues mechanically coupled with neighboring domains, small grains of the nonpolar phase, misaligned o-phase grains, and variation

The continuum distribution of the phase values in the BE AFAM (in the range from 70 to 120°) is much narrower than the phase difference in the BE PFM (Figure 5h,e). Therefore, topography cannot have a significant effect on the visualization of the piezoresponse phase. Subtraction of the BE AFAM phase from a BE PFM phase during normalization of the complex amplitudes results in the shift of the phase on the normalized maps (Figure 5k). In general, the BE PFM amplitude may contain information not only about the domain structure but also about the TiN topography (Supporting Information, Section S4.1.4). Analysis of the phase map (Figure 5e) and map of parameter Q (Figure 5f) allows us to separate the domain walls from the TiN crystallite boundaries. Parameter Q were determined during the fitting of spectra and characterized the width of the contact resonance spectra (Supporting Information, Section S4.1.3). For well-defined piezoresponse spectra (points 1 and 2 in Figure 5b), the parameter Q is associated with a dissipation of cantilever oscillation and called the quality factor. In the case of contact resonance, a dissipation should depend on local mechanical properties (viscosity or Young’s modulus) and the local topography and determines the resonance enhancement of the local piezoresponse by cantilever. Due to opposite strains on domain boundaries, there is a very weak piezoresponse (below noise level) in point 3 in Figure 5b. In this case, the fitting of a predominately noisy (background) spectrum by a spike-like sharp Lorenz curve results in extremely high values of Q, which appears as white edges on the map of parameter Q along the domain walls. Therefore, the map of parameter Q helps to visually separate the domain walls from the TiN crystallite boundaries. In addition, in BE PFM on domain boundaries, a fitting amplitude spike locates near central frequency but does not correspond to actual contact resonance frequency (Figure 5b). As a result, appropriate edges are distinguishable on the BE PFM contact resonance frequency map (Figure 5i) but not on the BE AFAM contact resonance frequency map (Figure S5b in the Supporting Information), which is almost identical to that obtained for BE PFM in the rest. The identity of the contact resonance frequency maps confirms the absence of the electrostatic contributions in our experiment and the presence of the topographic crosstalk in BE PFM data. Normalization of complex BE PFM amplitudes on BE AFAM amplitudes allows to eliminate the topography contribution (Figure 5j) and to clearly visualize domains in the amplitude map. By overlapping phase and amplitude maps, one can associate vibrating areas with polarization orientation (Figure 5l). Piezoresponse hysteresis loops measured by band-excitation technique in an arbitrary point of the area confirms the ferroelectric nature of the TiN/HZO/TiN structures (Figure 5c). The domain structure dynamics of the trained TiN/HZO/ TiN ferroelectric capacitors were studied by piezoresponse mapping after the application of bias voltage pulse (Figure 6, Supporting Information, Section S4.2). One should take into account the drift of the sample during scanning, resulting in some distortion of the obtained maps. The switching by the voltage pulses of +3 V (−3 V) drives the polarization of the HZO layer to the up (down) direction (Figure 6a,g). A gradual change in the ratio of the areas occupied by the opposite domains is observed (Figure 6n). Equal areas are observed upon the approximated values of the voltage pulses ∼−1.47 and ∼+1.35 V, which is very close to the coercive voltages of the 8824

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces

specific polarization orientation and the polarization reversal in polar phase grains occur. The PFM results along with the TEM analysis for the pristine vs trained TiN/HZO/TiN ferroelectric capacitors indicate the electrically stimulated m → o phase transition during the wake-up process.

in the order parameter of the o-phase in- and out-of-plane of the sample. Recently, it has been shown from theoretical calculations36,37 that one of possible polarization switching pathways in FE-HfO2 within each operation cycle is the o → t → o phase transition. We do not see any regions without a relatively strong piezoresponse signal on the BE PFM/AFAM maps of the intermediate domain structures. However, the texture of the amplitude maps is smooth (not binary), for example, in the middle of the domain, the amplitude is higher than in the periphery. This can be explained both by the mechanical coupling with neighboring domains of opposite polarity and by the o → t → o phase transition that gradually moves from the center of the “domain” to its periphery. Considering the enhancement in the mechanical coupling between domains by the passive top TiN layer, we are inclined to consider that the switching process proceeds in a conventional way, i.e., like in other ferroelectrics, the polarization reversal in the o-phase grains is due to the nucleation and growth of seeds with the opposite polarization. The o → t → o phase transition in the HZO film during each switching cycle at domain walls is possible but cannot be detected if the intermediate t-phase is unstable in the PFM excitation field. The revealed domain structure dynamics shows that the local built-in electric fields in the HZO films are still present after wake up by 104 cycles. This fact is illustrated by the domains with preferred up and down polarizations, which are marked by dashed and dotted lines, respectively, in Figure 6. These domains asymmetrically contribute to the repolarization current peak (inset in Figure 3a) and increase the range of the coercive voltage. The picture revealed via microscopic domain structure dynamics study corroborates the FORC data describing the wake-up process.16,19,22 Indeed, it was found by the FORC technique that the cycling leads to the disappearance of built-in fields, presumably associated with two types of domains. The microscopic study of domain structure dynamics allowed to find domain residues responsible for remnant built-in fields in trained structures.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.7b17482. Additional information and figures related to the growth and characterization of the ferroelectric TiN/ Hf0.5Zr0.5O2/TiN devices, as well as the full description of the combined BE PFM/AFAM technique (PDF)



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Anastasia Chouprik: 0000-0003-3672-4791 Author Contributions

The manuscript was written through contributions of all of the authors. All of the authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was performed using equipment of MIPT Shared Facilities Center with financial support from the Russian Foundation for Advanced Research Projects and the Ministry of Education and Science of the Russian Federation (Grant No. RFMEFI59417X0014). The ALD growth was supported by Russian Science Foundation (Project No. 14-19-01645-P). Work at the UNL was supported by the National Science Foundation (NSF) through Materials Research Science and Engineering Center (MRSEC) under Grant DMR-1420645. A.G. also acknowledges the support by the Center for Nanoferroic Devices (CNFD), a Semiconductor Research Corporation Nanoelectronics Research Initiative (SRC-NRI) under Task ID 2398.002, sponsored by NIST and the Nanoelectronics Research Corporation (NERC).



CONCLUSIONS The static domain structure in thin (10 nm) ferroelectric polycrystalline HZO layer in functional ferroelectric capacitors was studied by the combined BE PFM/AFAM technique. By normalizing the BE PFM data to the mechanical contact response simultaneously measured by the BE AFAM, we have found that the effect of the polycrystalline surface morphology of the TiN/HZO/TiN ferroelectric capacitor on the domain mapping is weak. The total area of trained HZO film with Pr = 17 μC/cm2 takes part in polarization reversal. The domain structure evolution following the applied voltage during capacitor operation has been elucidated. It was found that the reversal of the polarization vector in the polar o-phase occurs by the nucleation and growth of the opposite polarization domains. Equal ratio of domains with opposite direction of the polarization vector is observed upon the averaged values of the switching voltage pulses ∼−1.47 and ∼+1.35 V, which is very close to the coercive voltages of the relaxed structure measured by PUND. The presence of the domains with strong built-in electric fields has been directly revealed by PFM. In the pristine HZO films was found the mixture of polar phase grains with random polarization orientations and regions with low piezoresponse. During very first voltage pulse, both the transition of whole passive area in the ferroelectric phase with



ABBREVIATIONS HZO, Hf0.5Zr0.5O2; ALD, atomic layer deposition; BE PFM, band-excitation piezoresponse force microscopy; BE AFAM, band-excitation atomic force acoustic microscopy; FE, ferroelectric; FeFET, ferroelectric field-effect transistors; CMOS, complementary metal-oxide semiconductor; HZO, Hf0.5Zr0.5O2; TEM, transmission electron microscopy; SAED, selected-area electron diffraction; PUND, positive up negative down



REFERENCES

(1) Böscke, T. S.; Muller, J.; Brauhaus, D.; Schroder, U.; Bottger, U. Ferroelectricity in hafnium oxide thin films. Appl. Phys. Lett. 2011, 99, No. 102903, DOI: 10.1063/1.3634052. (2) Mueller, S.; Mueller, J.; Singh, A.; Riedel, S.; Sundqvist, J.; Schroeder, U.; Mikolajick, T. Incipient Ferroelectricity in Al-Doped HfO2 Thin Films. Adv. Funct. Mater. 2012, 22, 2412−2417. (3) Olsen, T.; Schroeder, U.; Mueller, S.; Krause, A.; Martin, D.; Singh, A.; Mueller, J.; Geidel, M.; Mikolajick, T. Co-sputtering yttrium

8825

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826

Research Article

ACS Applied Materials & Interfaces into hafnium oxide thin films to produce ferroelectric properties. Appl. Phys. Lett. 2012, 101, No. 082905. (4) Muller, J.; Boscke, T. S.; Brauhaus, D.; Schroeder, U.; Bottger, U.; Sundqvist, J.; Kucher, P.; Mikolajick, T.; Frey, L. Ferroelectric Zr0.5Hf0.5O2 thin films for nonvolatile memory applications. Appl. Phys. Lett. 2011, 99, No. 112901. (5) 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-Based Films. Adv. Mater. 2015, 27, 1811−1831. (6) Schroeder, U.; Yurchuk, E.; Müller, J.; Martin, D.; Schenk, T.; Adelmann, C.; Popovici, M.; Kalinin, S.; Mikolajick, T.; et al. Impact of different dopants on the switching properties of ferroelectric hafniumoxide. Jpn. J. Appl. Phys. 2014, 53, No. 08LE02. (7) Chernikova, A.; Kozodaev, M.; Markeev, A.; Matveev, Y.; Negrov, D.; Orlov, O. Confinement-free annealing induced ferroelectricity in Hf0.5Zr0.5O2 thin films. Microelectron. Eng. 2015, 147, 15−18. (8) Chernikova, A. G.; Kuzmichev, D. S.; Negrov, D. V.; Kozodaev, M. G.; Polyakov, S. N.; Markeev, A. M. Ferroelectric properties of full plasma-enhanced ALD TiN/La:HfO2/TiN stacks. Appl. Phys. Lett. 2016, 108, No. 242905. (9) Kozodaev, M. G.; Chernikova, A. G.; Korostylev, E. V.; Park, M. H.; Schroeder, U.; Hwang, C. S.; Markeev, A. M. Ferroelectric properties of lightly doped La:HfO2 thin films grown by plasmaassisted atomic layer deposition. Appl. Phys. Lett. 2017, 111, No. 132903. (10) Chouprik, A.; Chernikova, A.; Markeev, A.; Mikheev, V.; Negrov, D.; Spiridonov, M.; Zarubin, S.; Zenkevich, A. Electron transport across ultrathin ferroelectric Hf0.5Zr0.5O2 films on Si. Microelectron. Eng. 2017, 178, 250−253. (11) Hwang, C. S. Atomic Layer Deposition for Semiconductors; Springer: New York, 2014. (12) Clima, S.; Wouters, D. J.; Adelmann, C.; Schenk, T.; Schroeder, U.; Jurzak, M.; Portois, G. Identification of the ferroelectric switching process and dopant-dependent switching properties in orthorhombic HfO2: A first principles insight. Appl. Phys. Lett. 2014, 104, No. 092906. (13) Huan, T. D.; Sharma, V.; Rossetti, G. A., Jr.; Ramprasad, R. Pathways towards ferroelectricity in hafnia. Phys. Rev. B 2014, 90, No. 064111. (14) 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 oxide ferroelectric thin films. Appl. Phys. Lett. 2013, 103, No. 192904. (15) Sang, X.; Grimley, E. D.; Schenk, T.; Schroeder, U.; LeBeau, J. M. On the structural origins of ferroelectricity in HfO2 thin films. Appl. Phys. Lett. 2015, 106, No. 162905. (16) Pešić, 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-cycling behavior of the HfO2 based ferroelectric capacitors. Adv. Funct. Mater. 2016, 26, 4601−4612. (17) Zarubin, S.; Suvorova, E.; Spiridonov, M.; Negrov, D.; Chernikova, A.; Markeev, A.; Zenkevich, A. Fully ALD-grown TiN/ Hf0.5Zr0.5O2/TiN stacks: Ferroelectric and structural properties. Appl. Phys. Lett. 2016, 109, No. 192903. (18) Martin, D.; Müller, J.; Schenk, T.; Arruda, T. M.; Kumar, A.; Strelcov, E.; Yurchuk, E.; Müller, S.; Pohl, D.; Schröder, U.; Kalinin, S. V.; Mikolajick, T. Ferroelectricity in Si-doped HfO2 revealed: a binary lead-free ferroelectric. Adv. Mater. 2014, 26, 8198−8202. (19) Grimley, E. D.; Schenk, T.; Sang, X.; Pešić, M.; Schroeder, U.; Mikolajick, T.; LeBeau, J. M. Structural changes underlying fieldcycling phenomena in ferroelectric HfO2 thin films. Adv. Electron. Mater. 2016, 2, No. 1600173. (20) Fengler, F. P. G.; Pešić, M.; Starschich, S.; Schneller, T.; Künneth, C.; Böttger, U.; Mulaosmanovic, H.; Schenk, T.; Park, M. H.; Nigon, R.; Muralt, P.; Mikolajick, T.; Schroeder, U. Domain Pinning: Comparison of Hafnia and PZT Based Ferroelectrics. Adv. Electron. Mater. 2017, 3, No. 1600505.

(21) Schenk, T.; Schroeder, U.; Pešić, M.; Popovići, M.; Pershin, Y. V.; Mikolajick, T. Electric field cycling behavior of ferroelectric hafnium oxide. ACS Appl. Mater. Interfaces 2014, 6, 19744−19751. (22) Schenk, T.; Hoffmann, M.; Ocker, J.; Pešić, M.; Mikolajick, T.; Schroeder, U. Complex Internal Bias Fields in Ferroelectric Hafnium Oxide. ACS Appl. Mater. Interfaces 2015, 7, 20224−20233. (23) Lomenzo, P. D.; Takmeel, Q.; Zhou, C.; Fancher, C. M.; Lambers, E.; Rudawski, N. G.; Jones, J. L.; Moghaddam, S.; Nishida, T. TaN interface properties and electric field cycling effects on ferroelectric Si-doped HfO2 thin films. J. Appl. Phys. 2015, 117, No. 134105. (24) Stancu, A.; Ricinschi, D.; Mitoseriu, L.; Postolache, P.; Okuyama, M. First-order reversal curves diagrams for the characterization of ferroelectric switching. Appl. Phys. Lett. 2003, 83, No. 3767. (25) Richter, C.; Schenk, T.; Park, M. H.; Tscharntke, F. A.; Grimley, E. D.; LeBeau, J. M.; Zhou, C.; Fancher, C. M.; Jones, J. L.; Mikolajick, T.; Schroeder, U. Si Doped Hafnium Oxide  A “Fragile” Ferroelectric System. Adv. Electron. Mater. 2017, 3, No. 1700131. (26) Dehoff, C.; Rodriguez, B. J.; Kingon, A. I.; Nemanich, R. J.; Gruverman, A.; Cross, J. S. Atomic force microscopy-based experimental setup for studying domain switching dynamics in ferroelectric capacitors. Rev. Sci. Instrum. 2005, 76, No. 023708. (27) Stamm, A.; Kim, D. J.; Lu, H.; Bark, C. W.; Eom, C. B.; Gruverman, A. Polarization relaxation kinetics in ultrathin ferroelectric capacitors. Appl. Phys. Lett. 2013, 102, No. 092901. (28) Schroeder, U.; Pešić, M.; Schenk, T.; Mulaosmanovic, H.; Slesazeck, S.; Ocker, J.; Richter, C.; Yurchuk, E.; Khullar, K.; Müller, J.; Polakowski, P.; Grimley, E. D.; LeBeau, J. M.; Flachowsky, S.; Jansen, S.; Kolodinski, S.; van Bentum, R.; Kersch, A.; Künneth, C.; Mikolajick, T. In Impact of field cycling on HfO2 based non-volatile memory devices, Proceedings in Solid-State Device Research Conference (ESSDERC); IEEE, 2016; pp 364−368. (29) Vasudevan, R. K.; Balke, N.; Maksymovych, P.; Jesse, S.; Kalinin, S. V. Ferroelectric or non-ferroelectric: Why so many materials exhibit “ferroelectricity” on the nanoscale. Appl. Phys. Rev. 2017, 4, No. 021302. (30) Balke, N.; Maksymovych, P.; Jesse, S.; Herklotz, A.; Tselev, A.; Eom, C.-B.; Kravchenko, I. I.; Yu, P.; Kalinin, S. V. Differentiating Ferroelectric and Nonferroelectric Electromechanical Effects with Scanning Probe Microscopy. ACS Nano 2015, 9, 6484−6492. (31) Kim, S.; Seol, D.; Lu, X.; Alexe, M.; Kim, Y. Electrostatic-free piezoresponse force microscopy. Sci. Rep. 2017, 7, No. 41657. (32) Kalinin, S. V.; Rodriguez, B. J.; Kim, S.-H.; Hong, S.-K.; Gruverman, A.; Eliseev, E. A. Imaging mechanism of piezoresponse force microscopy in capacitor structures. Appl. Phys. Lett. 2008, 92, No. 152906. (33) Jesse, S.; Kalinin, S. Band excitation in scanning probe microscopy: sines of change. J. Phys. D: Appl. Phys. 2011, 44, No. 464006. (34) Jesse, S.; Kalinin, S. V.; Proksch, R.; Baddorf, A. P.; Rodriguez, B. J. The band excitation method in scanning probe microscopy for rapid mapping of energy dissipation on the nanoscale. Nanotechnology 2007, 18, No. 435503. (35) Hong, S.; Colla, E. L.; Kim, E.; Taylor, D. V.; Tagantsev, A. K.; Muralt, P.; No, K.; Setter, N. High resolution study of domain nucleation and growth during polarization switching in Pb (Zr, Ti)O3 ferroelectric thin film capacitors. J. Appl. Phys. 1999, 86, 607−613. (36) Barabash, S. V.; Pramanika, D.; Zhai, Y.; Magyari-Kopeb, B.; Nishi, Y. Ferroelectric Switching Pathways and Energetics in (Hf,Zr)O2. ECS Trans. 2017, 75, 107−121. (37) Materlik, R.; Kunneth, C.; Kersch, A. The origin of ferroelectricity in Hf1‑xZrxO2: A computational investigation and a surface energy model. J. Appl. Phys. 2015, 117, No. 134109.

8826

DOI: 10.1021/acsami.7b17482 ACS Appl. Mater. Interfaces 2018, 10, 8818−8826