Electric and Magnetic Properties of Atomic Layer Deposited ZrO 2 -HfO 2 Thin Films

Atomic layer deposition method was employed to deposit thin ﬁlms consisting of ZrO 2 and HfO 2 . Zirconia ﬁlms were doped with hafnia and vice versa, and also nanolaminates were formed. All depositions were carried out at 300 ◦ C. Most ﬁlms were crystalline in their as-deposited state. Zirconia exhibited the metastable cubic and tetragonal phases by a large majority, whereas hafnia was mostly in its stable monoclinic phase. Magnetic and electrical properties of the ﬁlms were assessed. Un-doped zirconia was ferromagnetic and this property diminished with increasing the amount of hafnia in a ﬁlm. All ﬁlms exhibited ferroelectric-like behavior and the polarization curves also changed with respect to the ﬁlm composition. terms of the

HfO 2 and ZrO 2 thin films have many applications, for example HfO 2 films are used as heat mirrors for energy-efficient windows, 1 both ZrO 2 and HfO 2 can serve as optical coatings, 2-4 waveguides, [5][6][7] gas sensors, [8][9][10] gate dielectrics in transistors. [11][12][13] Proposed applications also include novel memory materials, such as resistive switching memory elements [14][15][16][17] and ferroelectric memories. [18][19][20] ZrO 2 and HfO 2 have also been recognized as potential magnetic materials, mostly when doped with some other material, for example Mn-doped zirconia, 21 Fe, Co and Ni doped zirconia, 22 Sc, Ti, Ta, Fe and Co doped hafnia. 23 HfO 2 and ZrO 2 thin films have been prepared via reactive sputtering, 24,25 pulsed laser deposition, 2,26 electron beam evaporation, 27,28 chemical vapor deposition 29,30 and spray pyrolysis. 31,32 HfO 2 thin films have been atomic layer deposited (ALD) from hafnium tetrakis(dimethylamide) and water on Si(100) substrates in the temperature range of 205-400 • C. 33 Hafnium tetrakis(ethylmethylamide) and water were used as precursors for depositing the films on Si(100), 34 TiN 19 or Ni, Pt and Pd. 35 Atomic layer deposited hafnia has been made from HfCl 4 and H 2 O at 300 • C. 36,37 Film growth at 300 • C was analyzed for comparison, using different metal precursors HfCl 4 and HfI 4 , and the oxidizer was water. 38 HfCl 4 -H 2 O and HfI 4 -O 2 processes were compared in the temperature range of 300-600 • C. 39 HfO 2 and ZrO 2 have been fabricated, using Hf[N(CH 3 ) 2 ] 4 (TDMA-HF) and ZyALD ((tris(dimethylamino)cylcopentadienyl)-zirconium), as respective Hf and Zr precursors, and O 3 as oxidizer, at 260 • C and 285 • C. 20 ZrO 2 films have been atomic layer deposited using (CpMe) 2 ZrMe 2 and (CpMe) 2 Zr(OMe)Me as Zr precursors, and O 3 as the oxidant at 300 or 350 • C. 40 ZrO 2 films from tris(dimethylamino)cyclopentadienylzirconium CpZr(NMe 2 ) 3 and H 2 O, were investigated using real-time characterization of the growth process at substrate temperatures ranging from 120 to 350 • C. 41 ZrO 2 growth has been carried out in the case of ZrCl 4 and H 2 O as precursors 37 and this process has also been monitored in real time. 42 Zirconia has also been deposited from TEMAZ (tetrakis(ethylmethylamide) zirconium) and water. 19,35 Atomic layer deposited HfO 2 and ZrO 2 nanolaminates 20 and mixtures 19 were investigated and found to be ferroelectric. Microstructural evolution of ALD-grown HfO 2 -ZrO 2 nanolaminates was studied. 37 Atomic layer deposited Hf 0.5 Zr 0.5 O 2 films have also shown to be ferroelectric. 35 This study is devoted to ALD of mixed or laminated thin films consisting of ZrO 2 and HfO 2 . The purpose was to evaluate the strucz E-mail: kristjan.kalam@ut.ee ture and composition of the deposited films and how they influence the magnetic and electrical properties, more specifically, the materials ability to polarize in magnetic and electric fields. The hypothesis was that films consisting of ZrO 2 and HfO 2 should exhibit both ferroelectric and ferromagnetic properties and the goal was to find out the structure and composition, for which both of these properties appear in the same material.

Experimental
The films studied in this work were grown in a low-pressure flowtype ALD reactor. 43 HfCl 4 , 99,9%, supplied by Alfa Aesar, was used as the hafnium precursor and was evaporated at 162 • C. Zirconium precursor, 99,9% ZrCl 4 , supplied by Aldrich, was evaporated at 160 • C. Both metal precursors were evaporated from a glass boat inside the reactor. Nitrogen, N 2 (99.999% purity, Eesti AGA AS), was applied as the carrier and purging gas. Water was used as an oxidizer and the ALD reactions were carried out at 300 • C.
Films were deposited by alternating metal and oxygen precursors in sequential ALD cycles with certain cycle ratios. For example, hafnia was doped with zirconia in the following manner: 10 ALD cycles of HfO 2 was deposited, onto which 1 ALD cycle of ZrO 2 was deposited and this sequence was repeated 9 times. Total amount of cycles was kept at about a 100 for each deposition. Cycle ratios of HfO 2 and ZrO 2 were varied as such: 10:1, 10:3, 10:5, 10:10, 5:10, 3:10, 1:10, 2:1. Also, two nanolaminates were prepared: 50 cycles of hafnia + 50 cycles of zirconia and vice versa. The cycle times for ZrO 2 were 4-4-2-8 s for the sequence ZrCl 4 pulse -N 2 purge -O 3 pulse -N 2 purge. Cycle times for an analogous HfO 2 growth cycle were 4-3-2-7 s.
Films were grown on Si(100) cleansed and etched 44 and highlydoped conductive Si substrates covered by 10 nm thick TiN film grown by chemical vapor deposition. The films, which were deposited on TiN substrates for electrical measurements, were also supplied with Ti/Al electrodes (area 0.204 mm 2 ) electron-beam evaporated on top of the films, with ca. 30 nm thick Ti layer in direct contact to the films. The structure to conduct electrical measurements was, from top to bottom, Al/Ti/Functional layer/TiN/Si/Al. Spectroscopic ellipsometer (SE), model GES5-E, was used for the measurements of the films thicknesses and refractive indexes. Ellipsometric data was modelled in the range of 1.3-4.5 eV using the Cauchy dispersion model. X-ray fluorescence (XRF) spectrometer Rigaku ZSX 400 and program ZSX Version 5.55 was used to measure the elemental composition of films. The crystal structure was evaluated by grazing incidence X-ray diffractometry (GIXRD), using a X-ray diffractometer SmartLab Rigaku with CuKα radiation, Table I. Thicknesses and cation ratios of samples deposited with variable HfO 2 :ZrO 2 ALD cycle ratio. Two HfO 2 +ZrO 2 double layers were also deposited, with 50 ALD cycles of one constituent oxide followed by 50 cycles for another oxide. which corresponds to an X-ray wavelength of 0.15406 nm. Surface morphology of films was evaluated by scanning electron microscopy (SEM) using a Dual Beam equipment FEI Helios NanoLab 600. Electrical polarization measurements were performed by means of an Agilent DXO-X 3104 digital oscilloscope with a built-in wave generator. The standard Sawyer-Tower experiment was carried out by applying a periodic triangular-shaped stimulus and recording the voltage loops data from the oscilloscope. Charge values were obtained from the sensed voltage across a stated capacitance.
Magnetic measurements were performed using Vibrating Sample Magnetometer (VSM) option of the Physical Property Measurement System 14T Quantum Design by scanning the magnetic field from −1 to 1 T parallel to the film surface at room temperature.

Results and Discussion
Film growth and composition.-Total amount of ALD cycles was kept near one hundred for each deposition to obtain comparable thicknesses of films, except for un-doped reference films. Thicknesses, however, varied quite substantially -from 10 to 22 nm (Table I). These results allowed one to believe that for various cycle ratios the growth rates are quite different. The lowest thicknesses were measured for the cycle ratio of 10:3, regardless of which oxide was deposited in majority (Table I). This is probably due to three ALD cycles not being enough to complete the growth of a continuous layer, but also inhibiting the growth of other oxide more than one ALD cycle does.
Hafnia doped with zirconia systematically exhibited larger thickness values for the same cycle ratios and cycle amounts than zirconia doped with hafnia. This was in spite of un-doped zirconia exhibiting higher growth rate than un-doped hafnia in the case of 200 ALD cycles. Examining the experimental Hf/(Hf+Zr) ratios, it appears that in the very beginning of growth, i.e. during the first 10 cycles, hafnia grows faster than zirconia.
Investigating the nanolaminates, it seems that, during the first 50 cycles, zirconia grows equally well on Si(100) substrate and hafnia, but hafnia grows faster on Si(100) than on zirconia.
Film structure.-All the samples were investigated in their asdeposited state. Hafnia was, by a vast majority, in its stable monoclinic form (PDF Card 01-075-4290). A low amount of cubic hafnia (PDF Card 00-053-0560) was also present (Fig. 1). Zirconia, on the other hand, was cubic or tetragonal -the width and intensities of the reflections in the patterns obtained and presented in this work did not allow one to clearly distinguish between the two lattice types (PDF Card 01-077-3168). A low amount of monoclinic zirconia (PDF Card 01-075-6446) was also distinguishable in the un-doped zirconia. Doping has evidently stabilized the metastable phase and monoclinic phase did not appear. Also, in un-doped zirconia, a peak attributable to the 103 reflection of orthorhombic ZrO 2 appeared at 56 • (Fig. 1). (PDF Card 00-034-1084). Films, which consisted of both metal oxides, exhibited the same GIXRD peaks as both un-doped zirconia and hafnia reference films, but the peaks in the mixed and laminated films were less intense. This can be due to the lower thicknesses of the doped films and nanolaminates, but also the degree of crystallinity is lower in such films, because mixing the oxides makes it more difficult for either single oxide to crystallize. Some patterns of mixed films had almost no distinguishable GIXRD peaks at all. Refractive index of un-doped zirconia and hafnia were found to be 1.97 and 1.90 (at 633 nm), respectively. These values are slightly lower than values commonly presented in literature. For example, HfO 2 , grown by ALD at 300 • C by Aarik et al., 45 has possessed refractive index of 2.00. Aarik et al. 46 have also measured ALD-grown ZrO 2 with refractive index of 2.20. All doped films had even lower refractive indexes, most likely due to the lower density of the films (Fig. 2). The two films with recognizably low refractive indexes (below 1.6) both had significantly lower densities than other films (about 35% porosity). The lowest refractive index was measured in the case of about 60% Hf in the total amount of cations. Scattered data in the graph, i.e. different refractive indexes for the films with quite similar composition, could be explained with fluctuations in crystallinitysamples with refractive indexes higher than the overall trend are prob-ably more strongly crystallized. Small fluctuations in composition, i.e. relative content of constituent metals, as well as deposition parameters, may induce fluctuations in phase composition, crystal growth and electron density. These fluctuations may become complicated to conveniently determine by X-ray diffraction studies, but they may induce notable changes in optical density of the solid layer.
It is noteworthy, that despite marked scatter in refractive index values, one may recognize an implication of dependence between refractive index and composition (Fig. 2). It is likely, that the films with dominating content of either HfO 2 or ZrO 2 were grown relatively dense and in these films the effective refractive index tended to be higher. On the other hand, the films in which the amounts of HfO 2 and ZrO 2 were comparable, i.e. the most strongly mixed films such as those with Hf/(Hf+Zr) = 0.5..0.6, were grown as the least dense ones with the most distorted lattices as also implied by the GIXRD results. Therefore, in the films with comparable Hf and Zr contents the refractive index had to become slightly lower, as confirmed by the measurements (Fig. 2).
SEM images of HfO 2 and films which consist mostly of HfO 2 indicate that regions with larger grains alternate with smaller grains (Fig. 3). Since the substrate was Si(100) and there is no reason to assume a heterogeneous distribution of nucleation centers, it can be supposed that the deposition parameters used in these experiments favored the growth of larger grains, but they could not form yet due to the limited time for the nucleation process. As the composition of the films changed toward higher amounts of Zr, the grain sizes became more even.
Electrical and magnetic properties.-Polarization charge in the films exhibited weak tendency to saturate upon increasing the electric field strength (Fig. 4). This effect can possibly be attributed to ferroelectric behavior, but it is evident that interfacial polarization markedly contributes to the charge polarized due to leakage current through the films. Correlation between the polarization values in such loops and values of leakage current has been shown previously. 44 Electric field values, at which similar polarization charges could be recorded, increased with higher hafnia content (Fig. 4). This is possibly due to hafnia being in the stable monoclinic form. Hafnia can be regarded  as more stable and less defective than zirconia, hence the preference to form a monoclinic phase. A less defective material also exhibits lower leakage, which probably accounts for the lowering of charge polarization values with the increasing content of hafnia.
The same possible reasons, why hafnia exhibits lower values for charge polarization, may explain, why it does not show ferromagnetic properties. The chemically more stable hafnia is less defective, has less oxygen vacancies and therefore does not magnetize as well as zirconia, which is ferromagnetic. Figure 5 shows hafnia having the lowest magnetization values and the magnetization increasing with the zirconia content. Un-doped zirconia film still did not exhibit as high magnetization as the nanolaminate consisting of 10 nm thick hafnia and 8 nm thick zirconia, which had a saturation magnetization of 5 · 10 −6 emu and a coercivity of 50 Oe. Figure 6 illustrates the magnetic and electric properties of the aforementioned nanolaminate and the inset on the left panel of Figure 6 is an enlargement of the scale near the origin of the graph to indicate coercivity values. One can see, that ferromagnetic properties can be observed in a material which also demonstrates electronic charge-voltage loops with some contribution from ferroelectric-like polarization.
In regard and to compare with literature data, Böscke et al. 18 have grown HfO 2 thin films doped with silicon (oxide) by ALD using tetrakis(ethylmethylamino)hafnium, Hf[N(C 2 H 5 )CH 3 ] 4 , and Figure 5. Magnetization versus applied magnetic field curves measured at room temperature for metal oxides described in Table I. tetrakis(dimethylamino)silane, Si[N(CH 3 ) 2 ] 4 , with ozone as precursors between TiN electrodes. The 8.5 nm thick films were crystallized after annealing at 1000 • C. The films containing less than 4 at.% silicon consisted of a mixture of tetragonal and orthorhombic phases of HfO 2 and demonstrated optimized ferroelectric performance with remnant polarization above 10 μC/cm 2 and coercive field of 1 MV/cm.
Müller et al. 19 deposited HfO 2 and ZrO 2 mixture films to the thickness of 9 nm, varying the pulsing ratio for Hf[N(C 2 H 5 )CH 3 ] 4 and Zr[N(C 2 H 5 )CH 3 ] 4 precursors. These films were crystallized upon TiN top electrode deposition at 500 • C. Müller et al. 19 studied the effects of Zr:Hf ratio as well as the measurement temperature on the film structure and ferroelectric behavior. They observed that decrement in the relative content of Zr was followed by gradual transition between tetragonal, orthorhombic and monoclinic phases. They also observed that the application of TiN electrode tended to promote the tetragonal to orthorhombic crystallization pathway instead of that of tetragonal to monoclinic. They have measured the polarization charge hysteresis widths of approximately 2 MV, and, at nearly equal amounts of Zr and Hf, reached remanent polarization value 17 μC/cm 2 , but noticed also that, due to the increasing leakage currents, the electrical measurements had to be limited to 400 K.
Lin et al. 35 have also targeted Zr 0.5 Hf 0.5 O 2 films, containing equal amounts of Zr and Hf, grown by ALD from Hf[N(C 2 H 5 )CH 3 ] 4 and Zr[N(C 2 H 5 )CH 3 ] 4 precursors. Annealing at 500-550 • C was necessary to crystallize the structure. Electrical measurements revealed coercivity values even exceeding 1 MV and remanent polarization charge 19 μC/cm 2 . Annealing series was provided up to 750 C. Lin et al. 35 observed that, although the contribution from orthorhombic phase seemed to increase with temperature, the heat-treatment at 750 • C destroyed the ferroelectric-like polarization, likely due to the development of leakage pathways along grain boundaries.
A study by Weeks et al. 20 was devoted to ALD of HfO 2 -ZrO 2 films from tetrakis(dimethylamino)hafnium, Hf[N(CH 3 ) 2 ] 4 ; tris(dimethylamino)cyclopentadienylzirconium, CpZr(NMe 2 ) (Cp = C 5 H 5 , Me = CH 3 ); and ozone as precursors. Within the study, 8 nm thick mixtures as well as nanolaminates were fabricated, constituting of 1-4 nm thick ZrO 2 and HfO 2 layers of equal thicknesses. The as-deposited stacks had again to be annealed at 500 • C for 10 min to promote crystallization. The main phase stabilized was either tetragonal or orthorhombic decided on the basis of either tetragonal 011 or orthorhombic 111 reflection detectable at 30.5 deg in XRD patterns. The hysteresis width in charge polarization loops was ca. 2 MV and the remanent polarization, as reported, was in the range of 20-39 μC/cm 2 .
It is to be noted, that magnetic properties of ferroelectric hafnium zirconium oxides have not been measured in the works referred to above. In terms of structural and electrical characteristics, however, the results described in our study are to certain extent analogous to those obtained earlier. In the present study, the films were crystallized and measured in as-deposited state, since the layers were purposefully grown using chalcogenide precursors. As a common feature, the crystalline structure may not defined as unambiguously that belonging to any particular polymorph of either zirconia or hafnia, but a mixture of tetragonal and orthorhombic phases is quite likely. The width of charge polarization hysteresis tends to be quite similar to those recorded elsewhere, so the coercivity as defined, may reach 1 MV/cm. On the other hand, the polarization charge at zero external field is considerably higher than that reported earlier, up to tens of mC/cm 2 . One can thus consider marked role for leaking charge interfering the polarization mechanisms. Further modifications of the structure and improvements in the performance of functional layers after, e.g., optimization of annealing parameters and/or electrode materials would form possible matter for further studies.

Summary
ZrO 2 and HfO 2 films, their various mixtures and two-layer nanolaminates were deposited by ALD, using ZrCl 4 , HfCl 4 and H 2 O. Un-doped ZrO 2 was found to be mostly in the cubic or tetragonal form, whereas un-doped HfO 2 was mostly monoclinic. Doped films exhibited the cubic zirconia structure if Hf/(Hf+Zr) ≤ 0.46 and monoclinic hafnia if Hf/(Hf+Zr) ≥ 0.9. Films with compositions in between aforementioned values had low peaks implying both structures or were nearly amorphous. Un-doped zirconia exhibited ferromagnetic behavior and this property diminished with the increasing amount of hafnia in the films. The highest saturation magnetization value 5 · 10 −6 emu and a coercivity of 50 Oe was obtained from a nanolaminate grown with 50 ALD cycles of HfO 2 and on top of it 50 ALD cycles of ZrO 2 , which resulted a 10 nm HfO 2 + 8 nm ZrO 2 film and a cation ratio Hf/(Hf+Zr) = 0.55 and exhibited very low diffraction peaks attributable to the cubic phase. Some films exhibited both distinctive ferromagnetic characteristics and electrical polarization curves with ferroelectric component.