Ferroelectric properties of ion-irradiated bismuth ferrite layers grown via molecular-beam epitaxy

BiFeO₃ crystalizes in a rhombohedrally distorted perovskite structure (space group #161, R3c) and exhibits the combination of ferroelectricity and spin-canted weak ferromagnetism.^1–3 At room temperature, the polar and magnetic order parameters are coupled. As a result, when ferroelectric domains are poled, magnetic moments reorient deterministically.^2,4,5 Voltage-controlled magnetism is important for enabling low-power spintronic devices that operate efficiently and independently of current-based switching mechanisms, including Oersted induction and spin-transfer torque.⁶ Utilizing BiFeO₃ for practical voltage-controlled spintronics, however, requires overcoming reliability challenges currently limiting ferroelectric-based devices.⁷ Key among these is reducing through-film leakage.^8–10 

In this letter, we investigate the influence of growth conditions and postdeposition ion-irradiation treatments on the composition, structure, and ferroelectric properties of epitaxial Bi_xFe_2−xO₃ layers grown via adsorption-controlled ozone molecular-beam epitaxy. Structural characterization reveals that stoichiometric films, for which bismuth and iron concentrations are equal (i.e., x ≈ 1.00), exhibit the highest crystalline and ferroelectric domain perfection. The leakage characteristics of these high-quality layers are found to be similar to defective layers grown near the single-phase field boundaries (i.e., x ≠ 1.00). Through film leakage is dramatically reduced by irradiating both stoichiometric and nonstoichiometric samples with He⁺ ions.

Bi_xFe_2−xO₃ layers are grown to a thickness of ≃200 nm on SrRuO₃-coated (110)_o-oriented DyScO₃ substrates (o subscripts denote orthorhombic indices in the nonstandard Pbnm setting) via adsorption-controlled molecular-beam epitaxy in a Veeco GEN10 system (base pressure P_Base = 1 × 10⁻⁸ Torr = 1.3 × 10⁻⁶ Pa). SrRuO₃ is selected as an epitaxial bottom electrode¹¹ due to its relatively low electrical resistivity (170 μΩ-cm at room temperature)¹² and lattice match with the DyScO₃ substrate (lattice mismatch m = 0.6%). The SrRuO₃ electrodes are deposited as described in Refs. 12 and 13 to thicknesses of ≃20 nm. Bi_xFe_2−xO₃ films are grown subsequently without breaking vacuum at growth temperatures T_s between 550 °C and 650 °C, estimated using a thermocouple in indirect contact with the growth surface.

Iron (99.995% pure) and bismuth (99.9999% pure) are simultaneously¹⁴ supplied to the growth surface from effusion cells operating at temperatures near 1250 °C and 650 °C, respectively. Iron fluxes are fixed at J_Fe = 2.4 × 10¹³ cm⁻² s⁻¹, yielding bismuth ferrite deposition rates of 1.1 nm/min; bismuth fluxes are varied between J_Bi = 4.8 × 10¹³ cm⁻² s⁻¹and 38.4 × 10¹³ cm⁻² s⁻¹, corresponding to J_Bi/J_Fe flux ratios spanning 2 through 16. High bismuth fluxes (J_Bi/J_Fe ≫ 1) promote the incorporation of bismuth into the growing film, compensating for the desorption of volatile BiO_x species.¹⁵ 

Elemental oxygen is supplied via atmospheric mixtures consisting of 20% O₂ and 80% O₃. The deposition pressure is maintained at 1 × 10⁻⁵ Torr (1.3 × 10⁻³ Pa), corresponding to an equivalent^16,17 ozone flux of $JO3≈1016cm−2s−1$⁠. The use of strong oxidants^18,12,19 and high ozone fluxes (⁠$JO3/JFe≈500$⁠) suppress the formation of oxygen vacancies, which engender mobile electrons.²⁰ 

Initial film growth experiments focus on determining the effects of incident flux ratios 2 ≤ J_Bi/J_Fe ≤ 16 and deposition temperatures 550 °C ≤ T_s ≤ 650 °C on the composition and structure of the Bi_xFe_2−xO₃ layers grown on SrRuO₃/DyScO₃(110)_o.

Figure 1(a) is a variability chart²¹ showing film bismuth fractions x determined from Bi_xFe_2−xO₃ layers using Rutherford backscattering spectrometry.^22–24 Measured x values span 0.90 (J_Bi/J_Fe = 4, T_s = 650 °C) through 1.05 (J_Bi/J_Fe = 16, T_s = 600 °C). Increasing T_s and reducing J_Bi/J_Fe result in lower film bismuth factions x; in particular, we find that a 50 °C increase in T_s has an effect on x equivalent to a two-fold reduction in J_Bi/J_Fe. The loss of bismuth at higher temperatures results from thermally activated BiO_x desorption.^25–27 

The phase composition and structural perfection of as-deposited films are investigated using x-ray diffraction (XRD) θ-2θ scans and reciprocal space maps (RSM). The findings are summarized as a function of incident metal flux ratios J_Bi/J_Fe and deposition temperatures T_s in Fig. 1(b). Three structurally distinct growth regions are observed, comprised of phase-pure Bi_xFe_2−xO₃ layers [white region, Fig. 1(b)] as well as two mixed-phase Bi_xFe_2−xO₃ regions, one with Fe₂O₃ inclusions and the other with Bi₂O_2.5 precipitates [blue and red regions, Fig. 1(b)]. The single-phase regime spans a wide temperature window in excess of 100 °C, which has been modeled kinetically and shown to narrow rapidly with decreasing oxidant pressures.¹⁵ Together with film compositional measurements [Fig. 1(a)], a single-phase-field width spanning x = 0.90–1.05 is established, in close agreement with prior reports for layers grown via pulsed-laser deposition.²⁸ Within the single-phase field, RSMs performed about symmetric film reflections reveal two categorically distinct peak shapes [Fig. 1(b) insets]: at lower deposition temperatures and bismuth fluxes, the fundamental film reflections display diffuse features; at higher T_s and J_Bi/J_Fe values, the peaks exhibit coherent profiles. In the remainder of this letter, we focus on the latter set of films, for which the structural quality is superior.

Figure 1(c) shows diffracted θ-2θ x-ray intensities collected near 002_p film and substrate peaks from Bi_xFe_2−xO₃ layers (p subscripts denotes pseudocubic indices) grown along the isoflux J_Bi/J_Fe = 16 and isotherm T_s = 650 °C lines. As the film bismuth fraction x is increased, film reflections shift to slightly lower 2θ values, resulting in out-of-plane lattice parameter values a_⊥ which grow linearly with composition x [Fig. 1(d)].²⁹ The variation in a_⊥ values suggests a changing concentration of point defects within the films.

The combination of reflection high-energy electron diffraction (RHEED) and atomic force microscopy (AFM) is employed to investigate the surface morphology of epitaxial Bi_xFe_2−xO₃ layers deposited on SrRuO₃-coated DyScO₃(110)_o substrates. Figures 2(a)–2(j) are in situ RHEED patterns collected along $100p$ and $110p$ azimuths as a function of film bismuth fractions x between 0.90 and 1.07; corresponding ex situ AFM topography maps are shown in Figs. 2(k)–2(o).

Stoichiometric Bi_xFe_2−xO₃ layers with x = 0.99 exhibit electron reflection patterns characterized by circular spots. The spots lie on half-circles [double arcs, Figs. 2(c) and 2(h)] on which the specular condition is satisfied. Both the position and shape of the spots indicate two-dimensional step-flow growth³⁰—a conclusion which is corroborated by AFM height maps [e.g., Fig. 2(m)], which consist of atomically smooth surfaces (root-mean-square surface fluctuations ρ_rms = 0.2 nm) with well-defined terraces and unit-cell-tall step edges.

Increasing x to 1.05 causes the reflected spots to elongate vertically into streaked ellipses [ovals, Fig. 2(d)]. Concurrently, sixth-order stripes [rectangles, Fig. 2(i)] as well as split Kikuchi lines [arrows, Fig. 2(d)] also appear. Together, these features suggest reconstructions of surface atoms and broadening of surface widths following the formation of mosaic blocks.^31,32 Surface topographies are determined via AFM [Fig. 2(n)] to consist of shallow fractal-like features,³³ yielding a surface roughness of ρ_rms = 0.9 nm.

For higher bismuth fractions of x = 1.07, arrays of bulk diffraction spots are observed [squares, Figs. 2(e) and 2(j)]. Bulk diffraction is a hallmark of three-dimensional growth³⁴ and occurs when glancing electrons penetrate through surface protrusions.³⁵ Such features are visible in AFM height images, including Fig. 2(o), and are attributed to Bi₂O_2.5 grains.²⁶ The protrusions also lead to rough surfaces for which ρ_rms = 11.0 nm, exceeding the roughness values obtained here for stoichiometric bismuth ferrite layers by over an order of magnitude.

Systematic changes in the morphologies of the layers are also observed for bismuth-deficient layers. RHEED and AFM images collected from Bi_xFe_2−xO₃ films with x = 0.95 exhibit half-order spots [rectangles, Fig. 2(g)] indicating doubling of surface unit cells and shallow islands characteristic of the layer-by-layer growth in the presence of limited adatom diffusivity across step-edge barriers^36,37 [Fig. 2(l)]. For films with x = 0.90, bulk diffraction spots are detected in RHEED patterns [squares, Figs. 2(a) and 2(f)] and pits are observed in AFM height maps [Fig. 2(k)]. The surface roughness for bismuth deficient layers spans ρ_rms = 2.7 nm (x = 0.95) through 7.4 nm (x = 0.90).

Collectively, the direct- and Fourier-space analyses establish a rich morphological phase diagram in which topographical features vary systematically and depend sensitively on film composition. This makes RHEED a sensitive in situ monitor for characterizing the growth of Bi_xFe_2−xO₃ layers in real time.

Figures 2(p)–2(t) are lateral piezoforce microscopy (PFM) images showing ferroelectric domain morphologies of Bi_xFe_2−xO₃ layers grown as a function of the film bismuth fraction x. Stoichiometric layers with x = 0.99 exhibit bimodal contrast variations corresponding to two ferroelectric domain variants [Fig. 2(r)], in agreement with prior reports.¹³ The domains assemble into one-dimensional stripes with remarkable long-range order and an in-plane periodicity along [001]_o of $λ[001]p=$ 296 ± 20 nm. Increasing x to 1.05 and 1.07 progressively disrupts the uniformity of the stripe pattern. Concurrently, array periodicities decrease to $λ[001]o$ = 245 ± 32 (x = 1.05) and 235 ± 44 nm (x = 1.07). Increased disorder and reduced domain widths are also observed with decreasing x in bismuth deficient layers; in these cases, $λ[001]o$ = 260 ± 38 (x = 0.95) and 213 ± 42 nm (x = 0.90). Because of the increased pattern disorder observed near the single-phase field boundary, we hypothesize that pattern irregularities are due to lattice imperfections, including the cluster of point defects which accommodate film nonstoichiometry. The varying domain periods reflect an interplay between polar stiffness and depolarization effects.^38–41 Near x = 0.99, polar stiffness reduces domain wall densities by penalizing the regions containing rapidly varying polarization. Conversely, near the single-phase field boundaries, depolarization effects dominate, suppressing stray fields by bringing domain walls closer together.

The ferroelectric properties of the Bi_xFe_2−xO₃ films are quantified using metal-ferroelectric-metal capacitor structures. Platinum top electrodes, in the form of 40-μm-diameter circles, are defined lithographically. Polarization P vs electric field E measurements are performed by applying trains of bipolar triangular pulses to the devices at a frequency of 10 kHz using a Precision Multiferroic Tester (Radiant Technologies, Inc.).

In the as-deposited state, the films display pronounced leakage which impede ferroelectric poling. Leakage in bismuth ferrite arises from a combination of factors, including domain-wall conductivity⁴² as well as electron and hole donor defects such as oxygen vacancies²⁰ and Fe²⁺-based complexes.⁴³ Ion irradiation was recently demonstrated as a successful avenue for increasing the resistivity of leaky ferroelectrics.^44,45 We adopt a similar strategy and bombard our patterned structures with 3.0 MeV He²⁺ ions. At this energy, the ions penetrate to a mean depth of ∼15 μm, damaging the film lattice but preserving the film chemistry.

Figure 3(a) shows P(E) curves obtained as a function of irradiation doses D between 0.3 × 10¹⁵ and 1 × 10¹⁶/cm² from a stoichiometric Bi_xFe_2−xO₃ film grown with x = 0.99. In contrast to P(E) curves measured from as-deposited heterostructures [for reference, also shown in Fig. 3(a)], devices irradiated with D ≥ 0.3 × 10¹⁵/cm² exhibit clear signatures of ferroelectricity, manifested in the form of hysteresis loops. For 0.3 × 10¹⁵ ≤ D ≤ 1 × 10¹⁵/cm², small residual leakage causes the hysteresis loop to be open, but increasing D further causes the loops to close completely, reflecting progressively decreased leakage. The suppression of leakage in irradiated layers is attribute to the formation of carrier scattering and trapping defects.⁴⁵ 

Figure 3(a) also demonstrates that, as irradiation doses are increased, ferroelectric coercive fields grow from E_c = 0.15 MV/cm (D = 0.3 × 10¹⁵/cm²) to 0.90 MV/cm (1 × 10¹⁶/cm²), following the exponential relationship E_c = 0.14e^0.19D (here, the units for E_c and D are MV/cm and 10¹⁵/cm², respectively). Measured E_c values are comparable to those reported for epitaxial BiFeO₃ films deposited on SrRuO₃/DyScO₃(001)⁴⁵ and SrTiO₃:Nb(001)²⁸ but are larger than the 0.08 MV/cm value obtained for free-standing bismuth ferrite membranes, for which domain walls move unobstructed by epitaxial strain.⁴⁶ The larger coercive fields of irradiated devices are attributed to the formation of domain-wall-pinning defects.

Hysteresis loops measured from Bi_xFe_2−xO₃-based devices⁴⁷ bombarded with 3 × 10¹⁵ ions/cm² are presented as a function of film bismuth fractions x in Fig. 3(c). Remanent polarizations are approximately constant at 62 ± 6 μC/cm², independent of the bismuth fraction. This is consistent with a spontaneous polarization of P_s = 107 ± 10 μC/cm² along $111p$⁠, the polarization direction in bismuth ferrite. As x is increased, however, coercive fields E_c vary substantially, decreasing from 0.57 (x = 0.90) to 0.31 MV/cm (x = 0.99) before rising to 0.35 MV/cm (x = 1.05). The reduced E_c values observed near stoichiometry are consistent with the lower defect concentration and higher structural perfection of these films (see Fig. 2).

When cyclically poled, bismuth ferrite layers exhibit fatigue-induced failure. The mechanisms responsible for fatigue are diverse: conducting filaments form causing electrical shorts,⁴⁶ charge injection at ferroelectric/electrode interfaces suppresses domain nucleation,^48,49 and pinned domains grow in size.⁵⁰ To characterize the endurance of our Bi_xFe_2−xO₃ layers and determine n, the number of polarization cycles tolerated before breakdown, we employ a 10 kHz rectangular waveform with variable bias amplitudes between ±0.35–0.65 MV/cm to ensure complete poling during testing. Figure 3(d) shows n as a function of the film bismuth fraction x. For bismuth-rich films with x = 1.05, repeated poling leads to breakdown above 3 × 10² cycles. As x is decreased, n grows exponentially to 6 × 10³ (x = 0.99) and 4 × 10⁴ (x = 0.90). The observed n values are typical of ferroelectric capacitor structures in which at least one of the electrodes is a metal⁵¹ and can be enhanced by exclusively employing epitaxial conducting oxide electrodes.⁵² The combination of oxide electrodes, ion irradiation, and bismuth deficient films thus provides an avenue for prolonging the reliability of Bi_xFe_2−xO₃ capacitors.

Commensurately strained Bi_xFe_2−xO₃ layers grown on SrRuO₃-coated DyScO(110)_o substrates using adsorption-controlled ozone molecular-beam epitaxy are employed to investigate the role of defects, introduced by varying synthesis conditions and by performing postgrowth ion bombardment, on the chemical composition, structural characteristics, domain morphology, and ferroelectric attributes of bismuth ferrite. Within the explored ranges of growth temperature 550 °C ≤ T_s ≤ 650 °C and incident bismuth-to-iron flux ratios 2 ≤ J_Bi/J_Fe ≤ 16, a single-phase field with bismuth fractions x spanning 0.90 through 1.07 is established. The varying film compositions are accompanied by topographical features that include pits (x = 0.90), mounds (x = 0.95), terraces (x ≈ 1.00), fractals (x = 1.05), and protrusions (x = 1.07); each feature produces unique diffraction signatures in RHEED suitable for monitoring film growth in real time.

Film polarization morphologies generally consist of two domain variants arranged in stripe patterns. Pattern perfection and geometry depend sensitively on point defect profiles, with the widest domain widths and most periodic structures occurring near stoichiometry (x ≈ 1.00). In the as-deposited state, all films display excessive leakage which impede ferroelectric poling when tested using fabricated 40-μm-diameter platinum-capped capacitor structures. By performing postgrowth ion irradiation treatments, leakage is suppressed, yielding closed polarization-vs-field hysteresis loops. Remanent polarizations are constant at ∼60 μC/cm² and independent of film composition; coercive fields are reduced near stoichiometry, where the bismuth and iron concentration are equal. Bismuth deficiency is demonstrated as an avenue for enhancing the endurance of Bi_xFe_2−xO₃-based ferroelectric devices.

This work was supported in part by the Semiconductor Research Corporation (SRC) as nCORE task 2758.003 and NSF under the E2CDA (Grant No. ECCS-1740136) programs. S.S. acknowledges support from the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Materials Sciences and Engineering Division under Contract No. DE-AC02-05-CH11231 (Materials Project program KC23MP) for the development of ferroelectric thin films and ion-beam-induced defect studies. L.W.M. acknowledges support from the National Science Foundation under Grant No. DMR-1708615. This work made use of the Cornell Center for Materials Research (CCMR) Shared Facilities, which are supported through the NSF MRSEC program (Grant No. DMR-1719875). Substrate preparation was performed in part at the Cornell NanoScale Facility, a member of the National Nanotechnology Coordinated Infrastructure (NNCI), which is supported by the NSF (Grant No. ECCS-1542081).