ITO thin films investigation. The key technology for the state-of-the-art ITO-based on-chip electro-optic modulator is ITO thin film with a strong compromise between carrier concentration and optical properties. Using typical value of ITO films carrier mobility (15 cm2/(V·s)) [18, 32], we calculated (ρ = 1/µcNce) its resistivity which has to be in the range from 2 to 8×103 Ω·cm. First, we compared resistivity, optical properties and surface morphology (Fig. 1) of four 20 nm-thick ITO films: 1) high temperature IBA e-beam evaporation (film1); 2) room temperature IBA e-beam evaporation with following 300C annealing (film2); 3) high temperature IBA e-beam evaporation with following 300C annealing (film3); 4) room temperature e-beam evaporation with following 300°C annealing (film4). We observe strong variation in films resistivity measured by the four-probe method with respect to evaporation technology and annealing parameters (Fig. 1a, detailed description see Supplementary). The resistivity of ITO film deposited at room temperature without annealing (not shown) is more than 101 Ω·cm, which indicates insufficient film crystallization and large amount of interstitial non-activated oxygen [29–31]. All the evaporated ITO films are ultra-flat with the RMS film surface roughness 0.7 ± 0.2 nm. Analyzing high-quality SEM images of the films one can observe strong texture growth for annealed films (Fig. 1d, films 2–4) and homogeneous microstructure for the film evaporated under elevated temperature without subsequent annealing.
Optical constants of the films 1–3 are measured in the wavelength range from 400 to 1600 nm (Fig. 1b, 1c); film4 was not studied as the career concentration of this film was too high. Evidently, film1 has much higher absorption in the whole visible range (Fig. 1c), which can be explained by low film crystallization during evaporation. Strong texture of film2 indicates an intense crystallization process that results in the lowest absorption over the entire wavelength range. ITO film3 also has a strong crystalline texture (Fig. 1d) and relatively low extinction coefficient in the visible region. The carrier mobility of film3 is higher than for films 1 and 2 (calculated based on the Drude–Lorentz model and spectroscopic ellipsometry experimental results ≈ 30 and 11–15 cm2/(V·s)), which can be explained by a less defective crystalline structure (see Discussion). The carrier concentrations for ITO films 1–3 is calculated as Nc = 1/ρeµ and equals to 3.4·1020 cm− 3, 3.7·1020 cm− 3 and 5.5·1020 cm− 3 correspondingly. According to previously reported data [18] films 1 and 2 are promising candidates for electro-optic modulators in both n-dominant (Mach-Zehnder) and k-dominant (electro-absorption) regions. The value for
film3 is the closest to epsilon-near-zero point (ENZ) at λ = 1550 nm (Figs. 1b, 1c, Drude-Lorentz model for ENZ). This type of ITO films can be used for electro-optic elements which require dielectric constant switching through ENZ point [33, 34].
However, low-loss n-dominant electro-optic modulator requires ITO films with higher resistance values (from 2·10− 3 to 8·10− 3 Ω·cm). We clearly observe that ITO films resistance is directly related to films crystalline structure and morphology. We investigate two methods to control ITO films resistance: 1) Ar/O2 gas mixture variation during IBAD evaporation with subsequent annealing; 2) annealing with different temperatures. Due to minimum extinction coefficient and acceptable resistivity, we choose a deposition recipe for film2 for further improvement. We studied two ITO films evaporated under room temperature (IBAD in Ar/O2 mixture with 2 sccm and 4 sccm Ar flow) without and with subsequent 20 minutes annealing in an argon atmosphere at 300°C (Fig. 1e). It can be seen that films annealing allows controllably reduce their resistance to the desired region. ITO films with relatively high resistance (more than 10− 2 Ω·cm) have an amorphous or nanocrystalline structure (Fig. 1e, no crystalline phases were found during SEM EBSD analysis). Films with relatively low
resistance (less than 10− 3 Ω·cm) have a crystalline structure (grains with hidden boundaries were detected by SEM EBSD diffraction analysis). Films with middle resistance (from 10− 3 to 102 Ω·cm) have a crystalline structure with the inclusions of amorphous phases. Finally, we investigated the dependence of ITO films resistance and crystalline structure (room temperature IBAD in Ar/O2 mixture (Ar flow = 4 sccm and O2 flow = 12 sccm)) on annealing temperature and atmosphere (Fig. 2a). One can observe that ITO film annealed at 300°C stays amorphous, while increasing the temperature by only 20°C immediately initializes the crystallization. We marked stepwise crystallization behavior which leads a well-defined crystalline ITO structure when annealed temperature becomes more than 500°C. Under higher temperatures (more than 1000°C) the films are damaged, melt and break into islands under surface forces. It should be noted that annealing in the atmosphere with a small fraction of oxygen increases resistance significantly (Fig. 2a, light blue point). In this case, the film
has a crystalline structure, in contrast to the partial crystallization (Fig. 1e), which is observed for the films with suitable resistivity annealed in a pure argon atmosphere. SEM EBSD analysis (Fig. 2b) demonstrates that increasing the annealing temperature in argon atmosphere (points 1 to 3) leads to increasing the pole density of reflecting planes orientations of ITO films (see inverse pole figures (IPF)). It indicates the appearance of preferred grains orientations in ITO films. Moreover, for the films annealed in oxygen-diluted atmosphere (point 4) the orientation of the grains is different which gives additional ways to control ITO films properties.
We assume that IBAD films have some structural crystalline features and represent chemically heterogeneous systems before annealing. Since the energy in IBAD processes is higher, crystal nuclei formation occurs during evaporation process (which is confirmed by EBSD analysis for non-annealed thin film – Fig. 3c). Based on the SEM image analysis (Fig. 3a, 3b) one can assume that during annealing the crystallization front propagation from nucleus along the growth channel occurs along preferred orientations similar to the processes occurring during the liquid- phase epitaxy [36]. In this case, the front movement can form opposite diffusion flow which is perpendicular to the growing surface (Fig. 3d). In other words, during ITO crystallization a depletion zone appears, where oxygen diffuses to build a crystal lattice. SEM images (Figs. 3a, 3b) show 2–6 channels radially emerging from each central point, which can be oxygen diffusion channels.
We observed that annealing of ITO films deposited without ion assistance causes greater resistance decrease (Fig. 1a). The part of the ITO structure crystallizes during evaporation (Fig. 3c). For ion-assisted processes, oxygen atoms and ions outside the crystal lattice are embedded into the volume of the film [29–31]. This affects further crystallization during annealing when oxygen annihilation with vacancies occurs [35]. Wherein the crystalline phase is more localized in the films evaporated with ion-assistance.
Thus, our deposition technique involves the usage of low energies and higher oxygen concentrations, which distinguishes it from other methods described in recent scientific papers [25, 26]. These conditions favor the formation of a non-crystalline film, as evidenced by the low carrier concentration (Nc < 4·1018 cm− 3). The films deposited at higher energy and lower oxygen concentration without further annealing show higher carrier concentrations (Nc > 1·1019 cm− 3) and a higher degree of crystallization. This is because, during the deposition, oxygen has time to integrate into the crystal lattice vacancies and initiate the process of material crystallization, whereas in our case it appears in the form of embedded oxygen outside the crystal lattice. Annealing at a controlled temperature gave certain energy for oxygen activation, which brought the film to certain values of the carrier concentration: Nc in the range from 1 to 10·1020 cm− 3. Moreover, as described above, crystallization occurs due to oxygen channels that are formed in the film.
Electro-optical modulation. We fabricate MOS capacitors (Si/SiO2/ITO/Me, Fig. 4a) based on ITO films 1, 2, 3 with aluminum (Al, work function is 4.2 eV) and silver (Ag, 4.8 eV) electrodes. MOS capacitors with Al electrodes demonstrate the breakdown voltage of less than 20 mV (Fig. 4b, red vertical curve). When replacing Al electrode to Ag one, a significantly higher breakdown voltage of 17 V can be observed with the leakage current of less than 5 nA (Fig. 4b, blue curve). Whereas MOS capacitor without electrodes (the probes contacted directly with the ITO film) showed breakdown voltage of 29 V (Fig. 4b, yellow curve). The high noise level in this case can be explained by poorer contact between the probe and ITO film surface. Based on these measurements of MOS-capacitors current-voltage characteristics we assumed that Al/ITO pair forms an energy barrier at their interface. When electrical potential is applied, an electron emission occurs due to smaller Al work function compared ITO (4.2 eV versus 4.7 eV, respectively – Fig. 4a – the targeted area). That means 0.5 eV energy barrier (\(\varDelta {\phi }_{ms}={{\Phi }}_{ITO}-{{\Phi }}_{Al}\)) is appeared at Al / ITO interface. The emission current density in this case is defined as:
\(j={j}_{ITO\to Al}-{j}_{Al\to ITO}=\frac{1}{4}q{n}_{S}{v}_{0}\left({e}^{\beta Vg}-1\right),\)
|
(1)
|
where q is an electron charge, ns = n0e–βΔφ – surface concentration in a semiconductor/metal interface, β = q/kT, υ0 – thermal velocity of electrons. In this case the free path of electrons in ITO equals to 17.8 nm:
\({l}_{e}=\frac{h}{\rho {q}^{2}N{\lambda }_{e}},\)
|
(2)
|
where h – Plank’s constant, 1/λe = (3π2N)1/3.
The calculated electric current in this barrier for chosen electrode area (approximately 2 mm2) is about 0.19 A (at applied voltage of 20 mV – Eq. 1). Additionally, free electrons can be accelerated by an external electric field and pass through the ITO layer (the electron mean free path is comparable to the film thickness), which can lead to oxide destruction and breakdown.
Next, we investigated ITO film optical characteristics switching under applied voltage. We decided to use MOS structure without electrodes in order to achieve better sensitivity and higher applied voltage (Fig. 5a). Ellipsometry parameters psi and delta (ψ, Δ) were carefully measured in a wide wavelength range from 400 to 1600 nm for multilayer stacks with the ITO films13. We observe a higher influence of the applied voltage on Psi and delta at longer wavelengths in the range from 1400 to 1600 nm (Fig. 5c, d). The dependencies of the applied voltage (data validation was performed only from 0 to + 3 V for the film 3) on ellipsometry characteristics (ψ, Δ) were measured at a standard telecom wavelength of 1550 nm (Fig. 5b).
The change in the film1 ellipsometry parameters is found to be lower compared with the film2, despite the fact these films have close values of the carrier concentration. Presumably, this can be explained by the presence of electron traps in the less crystallized structure of film1, which prevents the charge accumulation. The film3 possesses the strongest electro-optical effect because of the highest carrier concentration value. Based on carried measurements we studied ITO thin film optical parameters (see Supplementary). The complex refractive index of the accumulation layer was calculated using the
Drude-Lorentz model in the wavelength range from 550 to 1600 nm (Fig. 5e). We compared Δn and Δk for different ITO films at a wavelength of 1550 nm (Table 2, for details see Supplementary). The refractive index change for them is several times less than was mentioned in the recent works (Table 3). We assume, that this is the result of a big millimeter-scale electrode area which leads to weaker accumulation (depletion) in the ellipsometer spot area, because the charge coming from the probe distributes unevenly. Finally, we calculated (see Supplementary), that such device implementation can potentially provide an electro-optical modulation frequency of approximately 5 GHz (with silver electrodes placed on top of the ITO layer). In further work, we suppose device implementation with single-crystalline Ag electrode [37, 38], which could abate losses induced by SPP propagation at electrode interfaces.
Table 2
Comparison of refractive index and extinction coefficient change for the various ITO films at λ = 1550 nm.
parameter
|
film 1
|
film 2
|
film 3
|
Δn (+ 16 V)
|
0.053
|
0.048
|
0.101
|
Δn (-16 V)
|
0.030
|
0.053
|
0.098
|
Δk (+ 16 V)
|
0.030
|
0.028
|
0.119
|
Δk (-16 V)
|
0.016
|
0.056
|
0.121
|
Δn
|
0.083
|
0.101
|
0.199
|
Δk
|
0.046
|
0.084
|
0.240
|
n (at 0V)
|
1.342
|
1.308
|
0.675
|
k (at 0V)
|
0.226
|
0.141
|
0.329
|
Table 3
Comparison of devise parameters and ITO characteristics in recent works related to electro-optical modulation.
Device parameters
|
ITO characteristics
|
Device
|
IL, dB
|
Ldevice, um
|
Vswitch, V
|
tITO, nm
|
Nc, cm-3
|
nITO
(1550 nm)
|
kITO
(1550 nm)
|
ΔnITO
(1550 nm)
|
ΔkITO
(1550 nm)
|
Electroabsorption design
|
[3]
|
1
|
5
|
± 5
|
10
|
1.1·1019
|
1.964
(1310 nm)
|
0.002
(1310 nm)
|
0.922
(1310 nm)
|
0.271
(1310 nm)
|
[36]
|
-
|
1700
|
− 5
|
50
|
1.87·1020
|
1.75
|
0.195
|
0.08
|
0.075
|
Mach-Zehnder design
|
[13]
|
6
|
32
|
± 6
|
10
|
2.29·1020
|
1.45
|
0.18
|
~ 1
|
-
|
[37]
|
-
|
2
|
± 16
|
10
|
2.3·1020
|
1.62
|
0.15
|
0.15
|
0.1
|
[27]
|
6
|
1.5
|
± 13
|
10
|
3.13·1020
|
1.44
|
0.12
|
0.44
|
0.11
|
MOS-stack ellipsometry investigation
|
[23]
|
-
|
-
|
+ 5
|
40
|
3.63·1020
|
0.8
|
0.25
|
0.02
|
0.015
|
Our work
|
-
|
-
|
± 16
|
20
|
3.79·1020
|
1.31
|
0.14
|
0.1
|
0.084
|